Aspergillus fumigatus and aspergillosis: From basics to clinics

Abstract

The airborne fungus Aspergillus fumigatus poses a serious health threat to humans by causing numerous invasive infections and a notable mortality in humans, especially in immunocompromised patients. Mould-active azoles are the frontline therapeutics employed to treat aspergillosis. The global emergence of azole-resistant A. fumigatus isolates in clinic and environment, however, notoriously limits the therapeutic options of mould-active antifungals and potentially can be attributed to a mortality rate reaching up to 100 %. Although specific mutations in CYP51A are the main cause of azole resistance, there is a new wave of azole-resistant isolates with wild-type CYP51A genotype challenging the efficacy of the current diagnostic tools. Therefore, applications of whole-genome sequencing are increasingly gaining popularity to overcome such challenges. Prominent echinocandin tolerance, as well as liver and kidney toxicity posed by amphotericin B, necessitate a continuous quest for novel antifungal drugs to combat emerging azole-resistant A. fumigatus isolates. Animal models and the tools used for genetic engineering require further refinement to facilitate a better understanding about the resistance mechanisms, virulence, and immune reactions orchestrated against A. fumigatus. This review paper comprehensively discusses the current clinical challenges caused by A. fumigatus and provides insights on how to address them.

Introduction

Species within the genus Aspergillus have been long exploited as an invaluable biotechnological resource to produce pharmaceuticals, food and food ingredients, and enzymes (reviewed in Meyer et al. 2011). Among species within this genus, Aspergillus fumigatus is the most ubiquitous fungal species in the environment (reviewed in Kwon-Chung & Sugui 2013). Aspergillus fumigatus can withstand and survive in a wide range of pH and temperature and its hydrophobic cell wall allows this species to be efficiently dispersed by even slight air currents. Similarly, a number of features allow this species to be the most predominant mould species causing infections in humans (Kwon-Chung & Sugui 2013). Small conidia size allows penetration to the lower respiratory tract system and escaping clearance by mucociliary forces, presence of melanin in the cell wall enables withstanding reactive oxygen species and phagocytosis, and abundance of negatively charged sialic acid on the surface permits A. fumigatus to effectively bind to the basal lamina proteins once inside the host lung (reviewed in Kwon-Chung & Sugui 2013). Although conidia can be easily cleared by counteracting host mechanisms in the lung, A. fumigatus can cause a wide range of infections in both immunocompromised and immunocompetent individuals (Denning & Chakrabarti 2017), including an estimated annual number of 16 million pulmonary-infections (Denning et al. 2013, 2016) with fatal outcomes in many hundred thousand patients annually (Brown et al. 2012, Lowes et al. 2017).

Triazoles are the first-line antifungals used to treat patients suffering from aspergillosis and have brought down mortality rates to 30 % or lower in invasive aspergillosis, which is almost always fatal if untreated (Neofytos et al. 2009). Extensive use of azoles in various sectors ranging from agriculture and industry to clinics, however, promotes selective pressure allowing emergence of azole-resistant A. fumigatus (ARAF) isolates in numerous niches (Verweij et al. 2015). Subsequently, deposition of such ARAF spores in the lung of immunocompromised patients may cause azole-resistant invasive aspergillosis (IA), often in patients that have never been prescribed any azoles, resulting in treatment guidelines favouring initial treatment with liposomal amphotericin B in settings with high prevalence of azole resistance (i.e. > 10 %) (Thom & Church 1926). Therefore, the high mortality rate and wide range of infections together with the emergence of ARAF isolates severely complicates the management of patients suffering from aspergillosis. In this review, we discuss the current paradigm and challenges of aspergillosis, and subsequently provide suggestions to more effectively tackle these challenges utilising worldwide initiatives across multiple disciplines.

Taxonomy and prevalence of clinically important Aspergillus species

History

The genus Aspergillus has a long history which dates back to Micheli’s “Nova Plantarum Genera” of 1729. Micheli, being a priest, noted that the fungus he described resembled the shape of an aspergillum (sprinkler of holy water), hence the name Aspergillus. The genus gained more attention from 1850 onwards, because it was recognised as a causal agent of spoilage, human disease and producer of useful metabolites. Due to its economic significance, the taxonomy of the genus was studied various times in history. In 1926, Thom and Church brought all available material on Aspergillus together and published the first major monograph on the genus (Thom & Church 1926). This monograph was revised in 1945 (Thom & Raper 1945) and subsequently in 1965 (Raper & Fennell 1965). Their taxonomic schemes were based on macroscopic (e.g., conidial colour and growth rates) and microscopic characters (vesicle shape, presence/absence of metulae). These monographs were the standard until the introduction of the molecular techniques in the 1990’s. Due to DNA sequence analysis, and to a lesser extent extrolite analysis, morphologically well-defined species turn out to consist of multiple species. Nowadays, a polyphasic approach, integrating different kinds of data and information (phenotypic, genotypic and phylogenetic), is the standard for describing new species in Aspergillus. To date, an updated Aspergillus monograph is lacking; however, there are some more recent taxonomic overviews on various Aspergillus sections, e.g., sect. Aenei (Varga et al. 2010), Aspergillus (Chen et al. 2017), Cervini (Chen et al. 2016a, 2016c), Circumdati (Visagie et al. 2014), Clavati (Varga et al. 2007), Flavi (Frisvad et al. 2019), Flavipedes (Hubka et al. 2015), Fumigati (Samson et al. 2007a), Nidulantes (Chen et al. 2016a, 2016c), Nigri (Samson et al. 2007b, Varga et al. 2011), Polypaecilum (Tanney et al. 2017), Restricti (Sklenář et al. 2017), Terrei (Samson et al. 2011) and Usti (Houbraken et al. 2007).

Nomenclature and Aspergillus

The International Code of Nomenclature for Algae, Fungi and Plants (ICN) governs the naming of fungi (McNeill et al. 2012). For a long time, dual nomenclature was used and asexually reproducing fungi got separate names from their sexual states. When strictly following these old rules, the name of the sexual morph had priority over the asexual morph name. For example, Neosartorya fumigata and Petromyces flavus should be used instead of the more well-known names A. fumigatus and A. flavus. The separate naming of these morphs was debated for many years and the principle “One fungus, One name” was introduced on January 1, 2013. In practice this means after that date, a fungus can only have one name. Nowadays, the name Aspergillus is used in a broad sense. Species producing (different) sexual morphs and previously described in teleomorph genera (e.g., Emericella, Eurotium, Neosartorya, Petromyces) are treated as synonyms (Kocsubé et al. 2016). The single name nomenclature led to various name changes. Many of the clinically relevant species were already known under its current Aspergillus name and therefore these changes did not have a big impact in the field of medical mycology. For example, in medical mycology Aspergillus nidulans was already a well-known name, while Emericella nidulans was more commonly used in food and indoor mycology. In some cases, the species epithet already indicates the connection between the old and current name (e.g., A. chevalieri/Eurotium chevalieri, A. fischeri/Neosartorya fischeri, A. udagawae/Neosartorya udagawae), but in other cases this is less obvious (e.g., A. glaucus/Eurotium herbariorum, A. montevidensis/Eurotium amstelodami and A. thermomutatus/Neosartorya pseudofischeri). In order to help the users with these changes in nomenclature, a list of all accepted species was prepared for Aspergillus (and related genera) (Houbraken et al. 2020).

Classification of Aspergillus

The genus Aspergillus is classified in the family Aspergillaceae, order Eurotiales (Houbraken & Samson 2011). There is a long tradition of using an infrageneric classification in Aspergillus; these are names of taxa between the ranks of genus and species (e.g., subgenera, sections, series). Using morphological characters, Raper & Fennell divided Aspergillus in 18 groups; however, these groups do not have any standing nomenclature and should not be used anymore (Raper & Fennell 1965). To avoid confusion and to promote taxonomic stability, a formal infrageneric classification system was introduced by Gams et al. (1985) and they replaced the “group” structure by a subgeneric and sectional structure. Nowadays, the genus is subdivided in six subgenera, 27 sections and 87 series (Houbraken et al. 2020). These formal infrageneric ranks are not commonly used in medical mycology. Articles often refer to various other informal ranks, for example “species complexes” (e.g., “A. fumigatus species complex”) (Salsé et al. 2019, Dos Santos et al. 2020), “cryptic species”, “cryptic A. fumigatus” (Wiederhold et al. 2018, Rivero-Menendez et al. 2019a) “species clades” (e.g., A. fumigatus-clade) (Balajee et al. 2005, 2007a), “sensu lato” (e.g., A. fumigatus sensu lato) (Li et al. 2014, Hagiwara et al. 2019) and “sensu stricto” (e.g., A. fumigatus sensu stricto) (Li et al. 2014, Monteiro et al. 2019)). The main disadvantage of using these informal ranks is lack of consensus. For example, it is not clear whether “A. fumigatus sensu lato”, “A. fumigatus species complex”, “cryptic A. fumigatus” and the “A. fumigatus-clade” are actually representing the same (group of) species. It is therefore recommended to use, when possible, a formal classification system of subgenera, sections and series. In the case of A. fumigatus, it is recommended to refer to A. fumigatus (the species), series Fumigati (for A. fumigatus and the related species Aspergillus fischeri, A. fumigatiaffinis, A. fumigatus, A. fumisynnematus, A. laciniosus, A. lentulus, A. novofumigatus, A. oerlinghausensis, A. spinosus, A. takakii) or section Fumigati (59 species) (Houbraken et al. 2020).

Identification of Aspergillus species from pure culture

In the last decade, there is steep increase of the number of accepted Aspergillus species (Houbraken et al. 2020). The driving forces behind this steep increase are twofold: firstly, there is a large diversity and high interest in this genus and secondly, phenotypically well-known species are turn out to be species complexes that are genetically and evolutionary distinct (Chen et al. 2016a, 2016c, 2017, Houbraken et al. 2016a, Sklenář et al. 2017). Morphology was for a long time the mainstay in Aspergillus identification (Raper & Fennell 1965). As morphologically well-defined species turned out to be species complexes, accurate phenotype-based identification became more difficult and unreliable. There are often only small differences between species within a complex and sometimes they need to be grown on special agar media to observe those differences. Identification based on phenotypic characters is therefore challenging, even for experienced mycologists and (well-trained) staff of routine labs. These phenotypically similar species are also referred to as “cryptic species”; however, they can be identified using a molecular based approach. These phenotypically closely related (cryptic) species can have strikingly different patterns of antifungal susceptibility patterns against the most important antifungals, including triazoles and amphotericin B (AMB), and some of these species are intrinsically resistant or have acquired resistance against these antifungals (Alastruey-Izquierdo et al. 2013, 2014, Escribano et al. 2013, Negri et al. 2014, Iatta et al. 2016, Heo et al. 2017, Talbot & Barrs 2018, Zoran et al. 2018, Salah et al. 2019, Mendoza et al. 2020, Glampedakis et al. 2021). For correct identification of Aspergillus species, calmodulin gene sequencing is recommended (Samson et al. 2019), and partial β-tubulin gene sequencing can be used as an alternative. ITS sequencing lacks resolution and is therefore not suitable. The public databases are well-stocked with calmodulin gene sequences and 96.9 % of the 446 accepted Aspergillus taxa are represented with a calmodulin gene sequence in GenBank (Houbraken et al. 2020). Actually, the taxonomic position of the species lacking a calmodulin sequence needs to be determined (Houbraken et al. 2020) and are unlikely to be relevant in medical mycology. Besides an overview of the accepted species, also calmodulin and β-tubulin references sequences are given.

Since its introduction into clinical microbiology diagnostics, MALDI-TOF MS has become the standard workhorse system for the identification of bacteria and yeasts (Kostrzewa 2018). Because of their rigid cell wall and the phenotypic variability by sporulation, the identification of moulds has been shown to be more challenging. Different approaches for sample preparation have been developed to overcome these hurdles, e.g., tube extraction of proteins before spotting on the MALDI-TOF MS target or liquid culturing to avoid sporulation. A prerequisite of successful identification of moulds are extensive libraries containing high-quality reference spectra of well characterised strains. This has led to a number of dedicated, user-specific databases and database supplements besides the libraries supplied by manufacturers (Sanguinetti & Posteraro 2017, Patel 2019). A study of US academic centres using a database established at the NIH, a solid media extraction method and a challenge set of 80 clinical mould isolates demonstrated the requirement of instrument optimisation and high standardisation for mould identification across different laboratories (Lau et al. 2019). MALDI-TOF MS has been successfully applied to Aspergillus spp. identification in several studies. Thereby, it could be demonstrated that many but not all rare and cryptic species can be correctly identified if they are represented well in the according database (Vidal-Acuña et al. 2018, Imbert et al. 2019, Américo et al. 2020). Closely related Aspergillus species are sometimes difficult to differentiate by MALDI-TOF MS because of their similar spectral pattern, today, but this can be improved by further extension of databases and utilisation of alternative identification algorithms.

Taxonomic notes on A. fumigatus and other clinically relevant Aspergilli

Aspergillus section Fumigati

The species in this section produce uniseriate, columnar conidial heads in shades of green and flask shaped vesicles (Raper & Fennell 1965). Traditionally, the identification of these related species is performed using the colony patterns and the morphology of the conidiogenous structures, conidia, ascomata and ascsopores (Raper & Fennell 1965). However, clinical isolates can be markedly abnormal being more floccose with fewer conidia and more recent taxonomic studies showed that section Fumigati species can have a highly similar morphology. For accurate identification, a sequence-based approach is therefore recommended (Samson et al. 2007a). Aspergillus section Fumigati includes 63 species and the species of the section are thermotolerant. Aspergillus fischeri, A. fumigatus and A. oerlinghausenensis are able to grow at 50 °C (Houbraken et al. 2016b), while other species of the section have lower maximum growth temperatures (e.g., 45 °C: A. fumigatiaffinis, A. fumisynnematus, A. lentulus, A. novofumigatus; 42 °C: A. felis) (Balajee et al. 2007a). The sexual morph is of the neosartorya-type (Samson et al. 2007a). The most well-known member of this section is A. fumigatus, though other species in the section are also clinically relevant: A. felis, A. fischeri, A. fumigatiaffinis, A. fumisynnematus, A. hiratsukae, A. laciniosus, A. lentulus, A. novofumigatus, A. pseudoviridinutans, A. spinosus, A. thermomutatus, A. udagawae and A. viridinutans (Sugui et al. 2014, Frisvad & Larsen 2016). Taxonomic evaluation of A. pseudofelis and A. parafelis (also known as A. felis) showed that these species are included in the genetically diverse A. felis (Hubka et al. 2018) (Fig. 1).

Fig. 1.

Cladogram of the genus Aspergillus and the relationship between sections and subgenera. A selection of the species mentioned in the text are given in brackets in bold font after the section name. Adopted from Houbraken et al. (2020) with permission.

Aspergillus fumigatus is reported as the most prevalent species from this section in different countries (Binder & Lass-Flörl 2013). Other species of section Fumigati, the so-called cryptic A. fumigatus, have been increasingly identified in the clinical setting in the last two decades, which is because of the increasing use of the polyphasic approach for the discrimination and identification of Aspergillus species in clinical samples. Based on multilocus comparative sequence analysis, other species from this section are recovered from 3 to 6 % of patients with IA. Among cryptic species, A. felis, A. lentulus, A. thermomutatus and series Viridinutantes members (= A. viridinutans complex) are the most common isolates from clinical disease in human (Balajee et al. 2009, Alastruey-Izquierdo et al. 2013, 2014, Escribano et al. 2013, Negri et al. 2014, Sabino et al. 2014, Frisvad & Larsen 2016, Talbot et al. 2018, Paccoud et al. 2019, Yagi et al. 2019).

Aspergillus fumigatus biosynthesizes a variety of secondary metabolites such as fumagillin, fumitoxins, fumigaclavines A & C, fumitremorgins, gliotoxin, trypacidin, pseurotins, helvolic acid, pyripyropens, methyl-sulochrin, verruculogen, fumiquinazolines. Several of these metabolites may cause serious health hazard (Frisvad & Samson 1990, Fujimoto et al. 1993, Larsen et al. 2007, Frisvad & Larsen 2016), though none of them are actually regulated mycotoxins. Some of these metabolites are involved in impairing the host immune system (Steenwyk et al. 2020) e.g., gliotoxin has been shown to inhibit the host immune response (Sugui et al. 2007, Spikes et al. 2008).

Other Aspergillus sections with clinically relevant species

Besides the section Fumigati species, a wide variety of other Aspergilli are clinically relevant. Most of them belong to the species-rich sections Flavi, Nidulantes, Nigri, Terrei and Usti. The taxonomy of these sections is well-studied and correct identification using sequence data should therefore not be problematic (Houbraken et al. 2020).

The taxonomy of section Flavi was recently updated and contains 35 species (Visagie et al. 2014, Frisvad et al. 2019, Houbraken et al. 2020), of which nine species are known to cause infection in humans (Hedayati et al. 2007, Frisvad et al. 2019, Rudramurthy et al. 2019, Alshehri & Palanisamy 2020). Aspergillus flavus is the main and most commonly occurring species of the section. Most of the reports of A. oryzae in clinical settings are likely to be erroneous. Aspergillus oryzae is the domesticated form of A. flavus and they can be regarded as conspecific. It is impossible to reliably differentiate A. oryzae and A. flavus using morphology and calmodulin or β-tubulin gene sequencing. The differentiation between both species is mainly driven by an applied aspect: A. oryzae is extensively used in food fermentations (e.g., soy sauce, sake) and for the production of enzymes, and these industries do not want to use the name A. flavus, which has a strong association with aflatoxin production. As a consensus, strains that do not produce aflatoxin and have a food fermentation or biotechnological background can be identified as A. oryzae; wild-type strains are A. flavus.

The clinically most important species of section Terrei is A. terreus, which is the second or third most common cause of IA in immunocompromised patients (Lass-Flörl et al. 2005, Blum et al. 2008). Of the 17 accepted species in section Terrei (Houbraken et al. 2020), five are human pathogens (A. alabamensis, A. citrinoterreus, A. floccosus, A. hortae (= A. hortai) and A. neoafricanus) (Zoran et al. 2018, Lackner et al. 2019). According to a recent report, infections caused by the A. terreus species complex were identified in 21 countries and 38 centres, and account for 5.2 % of all mould infections (Risslegger et al. 2017). However, a high incidence of A. terreus infections was reported in Innsbruck (Austria) and Houston (USA) (Lass-Flörl et al. 2007).

Species belonging to section Nigri (“the black Aspergilli”) are phenotypically very similar. Eight (A. brasiliensis, A. carbonarius, A. japonicus, A. luchuensis (= A. acidus), A. niger (= A. foetidus), A. tubingensis, A. uvarum, A. welwitschiae (= A. awamori) of the 30 accepted species in section Nigri are reported to cause infections in humans. The identified pathogenic species from this section are generally reported as the third leading causative agents of IA (Samson et al. 2014, Huang et al. 2017). Within this section, A. niger sensu stricto is the most prevalent clinical isolate (68.4 % cases vs. A. tubingensis, 31.6 % cases) (Balajee et al. 2009). However, recent analyses based on β-tubulin and calmodulin gene sequencing revealed a shift toward other cryptic species, including A. tubingensis, and A. welwitschiae, in different countries (Iatta et al. 2016, Hedayati et al. 2019, Alshehri & Palanisamy 2020, Carrara et al. 2020, Takeda et al. 2020).

The majority of the 74 accepted species in section Nidulantes are isolated from the soil, plant material, or the indoor environment (Sklenář et al. 2020). Eleven species have been isolated from patients with Aspergillus infections, of which A. nidulans was reported as the main agent of IA in different countries (Gabrielli et al. 2014, Chrenkova et al. 2018, Seyedmousavi et al. 2018, Tavakoli et al. 2020). Aspergillus nidulans was also reported as the second most frequently encountered mould in patients with chronic granulomatous disease characterised by sudden invasive features (Blumental et al. 2011, King et al. 2016b, Khalid & Ali 2018). Section Versicolores is a synonym of section Nidulantes, and series Versicolores is nowadays used. The clinically relevant species A. sydowii and A. creber belong to series Versicolores (Borgohain et al. 2019, Alshehri & Palanisamy 2020); of all 25 members in the section Usti, A. calidoustus is most often reported as the causal agent of invasive infections. Prior to the description of A. calidoustus, clinical strains were attributed to A. ustus and A. calidoustus are easy to differentiate, since the latter grows rapidly at 37 °C, while the former does not (Balajee et al. 2007a, Varga et al. 2008). Other members of the section isolated from proven/probable IA cases include A. granulosus, A. pseudodeflectus and A. ustus; A. insuetus, A. keveii, A. puniceus, A. pseudodeflectus and A. ustus were reported from respiratory samples (Fakih et al. 1995, Glampedakis et al. 2021).

In addition to the species mentioned above, other taxa can also cause infections in humans. Others include A. chevalieri, A. costiformis, A. glaucus, A. montevidensis, A. proliferans and A. pseudoglaucus (sect. Aspergillus) (Aznar et al. 1989, Naidu & Singh 1994, Traboulsi et al. 2007, Hubka et al. 2012, Alshehri & Palanisamy 2020); A. insulicola, A. melleus, A. ochraceopetaliformis, A. ochraceus, A. persii, A. sclerotiorum, A. subramanianii and A. westerdijkiae (sect. Circumdati) (Novey & Wells 1978, García-Martos et al. 2001, Harima et al. 2004, Brasch et al. 2009, Zotti et al. 2010, 2015, Hubka et al. 2012, Babamahmoodi et al. 2015, Bongomin et al. 2018, Seyedmousavi et al. 2018, Amri et al. 2020); A. flavipes (sect. Flavipides) (Seyedmousavi et al. 2018); A. tanneri (sect. Tannerorum) (Seyedmousavi et al. 2018); A. candidus (sect. Candidi) (Bongomin et al. 2018); and A. penicillioides and A. conicus (sect. Restricti) (Sklenář et al. 2017).

Virulence, immunology and pathogenesis of Aspergillus

In its natural environment, Aspergillus behaves as a saprobe that survives under different stress conditions. Likewise, during human infection, it has evolved adaptive mechanisms that allow it to withstand the unfavourable conditions in the lungs and to counter environmental changes in temperature, pH, water and nutrient balance, oxidative stress, and host molecules with antifungal properties.

Among the many virulence traits exhibited by Aspergillus, its plasticity in nutrient acquisition and metabolism confers a major advantage for growth during fungal infection under conditions of limited nutrient availability (Brock 2009, Blatzer & Latgé 2017). In addition, in experimental models of A. fumigatus infection, sites of hypoxia are commonly observed in the lungs, highlighting a remarkable ability to survive and thrive in conditions of low oxygen (Grahl et al. 2011, Kowalski et al. 2019). Besides these and several other relevant traits elicited in the context of infection (reviewed in Latgé & Chamilos 2019), the cell wall is a unique virulence factor, since it protects A. fumigatus from external aggression, while at the same time, it plays an active role in infection by influencing and modulating the immune response of the host (Latgé et al. 2017, van de Veerdonk et al. 2017). Owing to its dynamic structural properties according to morphotype, growth stage, and environmental conditions, the fungal cell wall is the main source of fungal ligands that activate the immune system (Latgé 2010). The physical barrier of the respiratory tract affords the first line of defence against inhaled conidia of Aspergillus, after which the respiratory epithelium is the initial point of contact with inhaled conidia (Filler & Sheppard 2006). Indeed, an increasing body of evidence has revealed a critical role of the airway’s epithelium in fungal clearance (Amich et al. 2020, Seidel et al. 2020) and production of cytokines and antimicrobial peptides (Bellanger et al. 2009, Sharon et al. 2011, Richard et al. 2018). Under certain conditions conidia escape the respiratory epithelium and are then challenged by cells of the innate immune system, including resident alveolar macrophages and dendritic cells (van de Veerdonk et al. 2017). Germinating conidia that escape macrophages are eliminated by recruited neutrophils and monocytes. Neutrophil extracellular traps (NETs) contribute to the innate host defence in vivo and neutrophils exert a considerable variety of antifungal effector functions, which include recognition, phagocytosis, intracellular clearance mediated by both oxidative and non-oxidative mechanisms, secretion of antimicrobial molecules and the release of neutrophil extracellular traps (NETs) (Urban & Backman 2020). Failure to prevent conidial germination results in hyphal growth, tissue invasion, and marks the initiation of fungal disease. Innate immune cells express a vast repertoire of pattern recognition receptors (PRRs) that recognise pathogen-associated molecular patterns in the fungus and activate effector functions, including phagocytosis and the production of proinflammatory cytokines and chemokines that orchestrate innate and adaptive immune responses (Patin et al. 2019). IL-8, also known as neutrophil chemotactic factor, is produced by macrophages and epithelial cells as an important chemoattractant for neutrophils, also during early phases of IA, where conidia are killed by local alveolar macrophages, and has been extensively used as biomarker for invasive aspergillosis (Winn et al. 2003, Camargo & Husain 2014, Gonçalves et al. 2017, Heldt et al. 2017, 2018, Jenks et al. 2019d). Up-regulation of gene transcription by A. fumigatus proteases has been suggested as cause of increased release of IL-8 by A549 pulmonary epithelial cells and primary epithelial cells (Borger et al. 1999). Other studies have shown that in vitro opsonization of A. fumigatus conidia with H-ficolin, L-ficolin and M-ficolin, which play essential roles in pathogen recognition and complement activation through the lectin pathway, potentiate IL-8 secretion of A549 lung epithelial cells (Houser et al. 2013, Bidula et al. 2015, Ghufran et al. 2017).

The family of C-type lectin receptors (CLRs) is the best-studied with regard to antifungal immunity (Brown et al. 2018). For example, the importance of dectin-1 in the recognition of β-1,3-glucan and activation of downstream immune responses has been confirmed in patients with recurrent fungal infections carrying an early stop codon polymorphism (Ferwerda et al. 2009). This polymorphism results in a truncated form of dectin-1 lacking several amino acids, with a detrimental effect on recognition of β-1,3-glucan and cytokine production after fungal stimulation (Ferwerda et al. 2009, Cunha et al. 2010). As a result, this polymorphism was found to predispose hematopoietic stem-cell transplant (HSCT) recipients to the development of IA in different patient cohorts (Cunha et al. 2010, Chai et al. 2011, Fisher et al. 2017). More recently, another CLR named MelLec was identified as the receptor for fungal melanin (Stappers et al. 2018). Macrophages from carriers of a polymorphism in the cytoplasmic tail of MelLec displayed a generalised defect in the production of cytokines after fungal stimulation. Likely owing to this defect, HSCT recipients receiving grafts from affected donors displayed a markedly increased risk for invasive pulmonary aspergillosis (IPA) after transplantation (Bassetti et al. 2020, Donnelly et al. 2020).

The efficiency of fungal recognition also relies largely on the opsonization by different soluble pattern recognition molecules, including collectins, pentraxins, ficolins and components of the complement pathway (Bidula & Schelenz 2016). One molecule that has received a great deal of recent attention in the field of aspergillosis is the long pentraxin-3 (PTX3) (Foo et al. 2015). This molecule binds microbial moieties from a wide range of microorganisms, including A. fumigatus (Garlanda et al. 2002). Accordingly, genetic variation in PTX3 was identified as a major risk factor for IPA after HSCT (Cunha et al. 2014), an association that was validated in independent cohorts of recipients of HSCT (Fisher et al. 2017) and solid organ transplant (Cunha et al. 2015, Wojtowicz et al. 2015), as well as in patients with chronic obstructive pulmonary disease (Cunha & Carvalho 2018, He et al. 2018). Mechanistically, genetic variants in PTX3 compromised the normal expression of the protein in the lungs and, at a cellular level, the antifungal effector mechanisms of neutrophils (Cunha et al. 2014). The impact of PTX3 deficiency on neutrophil function was confirmed in a recent study describing a similar association in patients with acute myeloid leukemia undergoing chemotherapy courses without pre-existing neutropenia (White et al. 2018).

The interaction of Aspergillus with the immune system is being harnessed to propose novel and improved fungal diagnostics, but also the implementation of clinical models aimed at the effective prediction of infection in high-risk patients. A recent study evaluating the combination of multiple genetic and clinical factors into a predictive model has demonstrated that such information could be used to successfully guide pre-emptive therapy in haematological patients (White et al. 2018). Besides improved diagnostics, functional analyses of genetic variation influencing susceptibility to aspergillosis may assist in the design of innovative and personalised immunotherapeutic approaches. This is illustrated by the preclinical evidence showing that genetic PTX3 deficiency can be rescued by the administration of the recombinant protein, a finding that supports its personalised use in specific patients at high-risk of infection. In conclusion, the success of novel diagnostic and immunotherapeutic approaches for aspergillosis may benefit from personalisation based on the interindividual variability in antifungal immune function.

Clinical significance of Aspergillus

Infection in humans

Aspergillus is the most common cause of mould infections in humans and can cause a variety of serious diseases in both immunocompetent and immunocompromised patients (Lass-Flörl 2019). The most clinically relevant Aspergillus species is A. fumigatus, followed by A. flavus, A. terreus and A. niger. Non-invasive infections in immunocompetent patients (e.g. with cystic fibrosis or post-tuberculosis) are allergic sinusitis or allergic bronchopulmonary aspergillosis (ABPA), fungal balls in the sinus or lung, chronic pulmonary aspergillosis, otitis externa or onychomycosis (Denning et al. 2018). For invasive infections, the respiratory tract is the most common primary site of IA –due to inhalation of conidia– but any organ might be involved as a single organ infection or as part of dissemination. Sino-nasal and cerebral aspergillosis may occur particularly in immunocompromised patients (Reischies & Hoenigl 2014). Aspergillus endocarditis is rare and risk factors include prior valve replacement, indwelling central venous catheters or broad-spectrum antibiotic treatment (Aldosari et al. 2020). Hematogenous spread to the spleen leads to either infarction or abscesses (Smolovic et al. 2018). Aspergillosis of the kidney is rare and derives from hematogenous dissemination or ascending from pan-urothelial aspergillosis or secondary to obstructive uropathy. Gastrointestinal tract IA may occur when the mucosal barrier is impaired (Chasan et al. 2013). Other rare manifestations of IA are (vertebral) osteomyelitis, arthritis, or subacute thyroiditis. Endophthalmitis may be consequent to intraocular surgery or trauma of the eye or after hematogenous spread and is associated with poor ocular prognosis (Dave et al. 2020).

Cutaneous aspergillosis may be caused by inoculation into disrupted skin or secondary to hematogenous dissemination (Lass-Flörl 2019). Primary extrapulmonary invasive aspergillosis often requires surgery in addition to systemic antifungal therapy (Reischies & Hoenigl 2014, Dave et al. 2020). Depending on the type of immunosuppression of the host, invasive pulmonary aspergillosis may present primarily angio-invasive in those with neutropenia, or primarily airway-invasive in those with corticosteroid associated immunosuppression, resulting in distinct radiological and clinical manifestations (Bergeron et al. 2012, Jenks et al. 2019b). The most common site of hematogenic spread is the central nervous system leading to brain abscess, stroke or less frequent to meningitis, and associated with devastating mortality rates (Hoenigl & Krause 2013).

Emergent susceptible population to acquire pulmonary aspergillosis

Mould-active prophylaxis has shown some success in reducing IA in patients with traditional risk factors for IA, such as those with underlying hematologic malignancy and prolonged neutropenia (Cornillet et al. 2006, Duarte et al. 2014). However, the prevalence of IA continues to increase in non-neutropenic patients with severe underlying diseases, including those in intensive care units (Pappas et al. 2010, Eigl et al. 2015, Bassetti et al. 2018, Schauwvlieghe et al. 2018b), solid organ transplant recipients (Lewis & Kontoyiannis 2009), those receiving systemic glucocorticoids (Chamilos et al. 2018), those with underlying structural lung damage (Prattes et al. 2014), those with advanced liver cirrhosis and liver failure (Prattes et al. 2017), those receiving tyrosine kinase inhibitors such as ibrutinib (Lenczuk et al. 2018), those with solid cancers (Yan et al. 2009, Bassetti et al. 2018), and others (Guinea et al. 2010, Prattes et al. 2014, Bassetti et al. 2018, Ghez et al. 2018).

Aspergillus species, especially A. fumigatus, can cause co-infection with viruses, including cytomegalovirus, and –importantly– influenza virus leading to complication of management of patients inflicted (Schauwvlieghe et al. 2018b). In one multicentre study from the Netherlands and Belgium, invasive pulmonary aspergillosis was diagnosed in 83 (19 %) of 432 patients admitted with influenza to the ICU, a median of 3 days after admission to the ICU (Schauwvlieghe et al. 2018b), and independently associated with mortality. In another study from Canada, IA complicated only 7.2 % of influenza associated respiratory failure ICU admissions, although the rate varied between 0 and 23 % between influenza seasons (Schwartz et al. 2020).

In November 2019, a novel virus termed severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) was first identified as the causative agent of pneumonia from a cluster of individuals with pneumonia in Wuhan City, Hubei province, China (WHO 2020). Since the first reported case, SARS-CoV-2, which causes the disease now called coronavirus disease 2019 (COVID-19), has spread throughout China and to almost every country in the world as of May 2020. Although the vast majority of COVID-19 cases are mild to moderate, up to 20 % of patients with symptomatic COVID-19 may develop acute respiratory distress syndrome (ARDS) (Wang et al. 2020, Wu & McGoogan 2020, Xu et al. 2020). There is increasing evidence that COVID-19 patients are at risk of developing COVID-19 associated invasive pulmonary aspergillosis (CAPA) co-infection, with more than 100 cases reported to date (Blaize et al. 2020, Dupont et al. 2020, Gangneux et al. 2020, Hoenigl 2021, Koehler et al. 2020b, Lescure et al. 2020, Mitaka et al. 2020, van Arkel et al. 2020, Verweij et al. 2020, Bartoletti et al. 2021, Prattes et al. 2021, White et al. 2021). The pathophysiology of COVID-19 is not well understood, but leukopenia, lymphopenia and T-cell perturbations, including immune dysregulation impacting Th2 as well as Th1 responses in severe COVID19, are frequently observed among symptomatic patients (Chen et al. 2020a, 2020b, Huang et al. 2020, Zheng et al. 2020), and may predispose patients to fungal diseases. Furthermore, the utilisation of – often high dose – systemic and inhaled glucocorticoids, which may further predispose to opportunistic infections such as CAPA, has been described in close to 50 % of patients with COVID-19-associated ARDS (Wang et al. 2020). This immune dysregulation together with epithelial lung damage stemming from COVID-19 immunopathology may facilitate Aspergillus superinfection.

From Wuhan, epidemiological studies indicate that invasive fungal infections may occur in 4–5 % of COVID-19 episodes requiring ICU admission (Baxter et al. 2011, Yang et al. 2020). In a cohort from China, fungal infections were diagnosed in seven (3 %) out of 221 COVID-19 patients, all of whom were admitted to the ICU (Zhang et al. 2020). Given that in Wuhan, galactomannan (GM) testing is rarely available and fungal diagnostics are sparse (Chindamporn et al. 2018), this is likely an underestimate of the true burden of IA in patients with SARS-CoV-2 requiring ICU admission. A study from the United Kingdom reported a 14.1 % incidence of pulmonary aspergillosis among 135 ICU patients (White et al. 2021). Higher rates of CAPA were recently reported in a single centre study from the Netherlands, where a high incidence (19.4 %) of putative aspergillosis was observed in a cohort of 31 mechanically ventilated ICU patients with COVID-19 (van Arkel et al. 2020). In another study from Germany, five of 19 (26 %) consecutive critically ill COVID-19 patients with ARDS were diagnosed with putative CAPA, highlighting that rates may be comparable to those observed in association with severe influenza (Koehler et al. 2020b). In a study from Italy, probable CAPA was diagnosed in 30 (27.7 %) patients after a median of 4 (2–8) days from intensive care unit (ICU) admission (Bartoletti et al. 2021). In a study from France, putative CAPA was reported in 33 % of 27 critically ill COVID-19 patients in an enriched population undergoing bronchoscopy (Alanio et al. 2020). Finally, the highest rate of CAPA has been noted in Belgium reaching 35 % (7/20) of the COVID-19 patients presented with ARDS (Rutsaert et al. 2020). The vast majority of reported CAPA cases lacked EORTC/MSGERC host factors, highlighting the need for improved criteria for defining IPA in non-neutropenic patients, as reported elsewhere (Jenks et al. 2019b). Additional cases of fatal CAPA were reported from Argentina, Australia, Austria, Brazil, France, Ireland, Italy, Pakistan, Switzerland, United States and many other countries (Blaize et al. 2020, Mitaka et al. 2020, Prattes et al. 2021). Importantly, three cases of azole resistance have been reported, indicating that ARAF is emerging also among the ICU population at risk for IPA (Meijer et al. 2020, Ghelfenstein-Ferreira et al. 2021, Mohamed et al. 2021).

As clinical presentation and imaging findings of COVID-19 and IPA may overlap (fever, shortness of breath, cough, unspecific infiltrates and consolidations, halo sign), biomarker and culture based diagnostic work-up is essential. Serum GM may have imperfect sensitivity of 20 % and below (Alanio et al. 2020, Koehler et al. 2020b). While reasons for the lower sensitivity in CAPA versus influenza associated pulmonary aspergillosis are unknown, treatment with chloroquine, which exhibits in vitro activity against A. fumigatus (Henriet et al. 2013), may have explained the lower sensitivity in some of the earlier studies, given that exposure to antifungals is a well-known factor that decreases the sensitivity of GM-testing and may explain the lower sensitivity compared to influenza associated IPA (Eigl et al. 2015).

Future studies are needed to evaluate other blood tests for CAPA, including Aspergillus PCR (Egger et al. 2020), β-D-glucan (Heldt et al. 2018), and the two newly CE-marked point of care tests, the Aspergillus GM lateral flow assay (LFA) and the Aspergillus-specific lateral flow device test (Eigl et al. 2015, Jenks et al. 2019c, 2019e, Mercier et al. 2020, Wahidi et al. 2020). Further complicating diagnosis of CAPA is the extremely limited role of bronchoscopy in COVID-19 as this aerosol generating procedure increases the risk of exposure for patients and personnel (Jenks et al. 2020), although detailed instructions on how to safely perform bronchoscopy have been published (Koehler et al. 2020c). In some centres, however, collection of tracheal aspirates remains the preferred method for diagnosis. Although Aspergillus can be detected in sputum and tracheal aspirates in CAPA-patients, its presence might reflect oral pharyngeal colonisation as Aspergillus is considered a core component of the basal oral mycobiome (Krüger et al. 2019). Importantly, GM-testing, presented in detail below, which is an important tool for IPA diagnosis in BALF specimens and represents active growth (Eigl et al. 2017), is not validated for upper respiratory tract specimens.

These early findings suggest IA may be an important and underrecognised complication of SARS-CoV-2 infection, due to the absence of typical host factors. Since bacterial and fungal superinfections are difficult to distinguish from severe COVID-19 based on clinical or imaging findings, histopathology has a central role in determining prevalence and also outcomes of CAPA (reviewed in Arastehfar et al. 2020a). However, autopsies of COVID-19 patients are rarely performed due to the risk of aerosol transmission. Criteria for defining COVID-19 associated aspergillosis have been developed and will help classifying this important disease (Koehler et al. 2020a). The frequency of post-COVID-19 aspergillosis is likely to differ significantly between hospitals and geographic sites, and environmental factors may also play a large role in increasing exposure beyond what would normally be encountered within hospitals and ICUs. The rapid spread of COVID-19 to a non-immune population has been seen in temporary facilities/hospitals rapidly assembled that do not adhere to the rigorous ventilation requirements that are present within permanent hospitals. These temporary sites are essential to increase healthcare capacity; however, dust and construction-related increases in ambient air spore counts will very likely increase patient colonisation with Aspergillus and other fungal species predisposing to infection. Finally, drug-drug interactions may limit the use of voriconazole (Jenks et al. 2019a), the gold-standard treatment for IA in the ICU. Future studies need to evaluate effectiveness of isavuconazole (Jenks et al. 2018), and new antifungals currently under development like fosmanogepix or olorofim (Kupferschmidt 2019), which may have comparable efficacy without the same burden of drug-drug interactions, while other drugs such as rezafungin or ibrexafungerp may offer options for combination therapy or even prophylaxis.

Diagnosis of aspergillosis in clinic – Serology to PCR

For the diagnosis of IA, culture and microscopy are essential, but show limited sensitivity. Detection of the fungal cell wall component galactomannan (GM) has therefore become the imperfect gold standard (Hoenigl et al. 2012, 2013b, Duettmann et al. 2014, Eigl et al. 2017, Rawlings et al. 2019) as it is more sensitive than culture. Galactomannan is a polysaccharide that exists primarily in the cell walls of Aspergillus species (Zhou et al. 2017) and a commercially-available double sandwich enzyme immunoassay (EIA) utilises the monoclonal antibody EB-A2 (Platelia™, Bio-Rad, Marnes-la-Coquette, France) to detect the GM antigen. It is approved by the U.S. Food and Drug Administration (FDA) for testing of GM from serum and bronchoalveolar fluid (BALF). This antibody detects multiple epitopes on galactofuranose side chains that link to the mannan backbone (Kudoh et al. 2015), although galactofuranose is not specific to Aspergillus and is present in other fungi such as Fusarium species (Tortorano et al. 2012), Penicillium species (Huang et al. 2007), and Histoplasma species (Wheat et al. 2007). The current sensitivity and specificity of BALF GM is 82 % and 81 %, respectively (Rawlings et al. 2019). The optimal optical density index (ODI) threshold is debatable, although the FDA considers an ODI of ≥ 0.5 to be positive for GM in both serum and BALF.

Galactomannan from BALF has shown better diagnostic performance for IA than GM from blood, particularly in patients on mould-active antifungal prophylaxis (Heldt et al. 2018). In a systematic review investigating the accuracy of GM from BALF for the diagnosis of IA in immunocompromised patients, the sensitivity and specificity of GM was 0.88 and 0.81, respectively, at an ODI of 0.5; at an ODI of 1.0 the sensitivity was 0.78 and specificity 0.93 (de Heer et al. 2019). Particularly in non-neutropenic patients a higher cut-off of 1.0 ODI in BALF may be preferable (Prattes et al. 2014) as false-positive results may occur with the lower cut-off (Martinelli et al. 2019). Given the airway-invasive growth, BALF GM is always the preferable option for IA diagnosis in non-neutropenic patients, and GM from BALF has found to be superior to GM from serum (Bassetti et al. 2020).

Molecular tests such as PCR (Heldt et al. 2018, Prattes et al. 2018, Jenks et al. 2019f, Rawlings et al. 2019) have emerged as alternative options to diagnose IA and are widely used (Hoenigl et al. 2014b, Buchheidt et al. 2017), although there is a lack of standardisation (White et al. 2010) and a large variation in diagnostic performance across studies and settings (White et al. 2015b, Springer et al. 2016). PCR is most useful in high-risk groups such as neutropenic patients who are not receiving mould-active prophylaxis, where a negative result is reassuring in ruling out IA. In other settings including non-neutropenic patients and patients at low risk for IA, like GM the utility of PCR testing is limited, particularly from blood (Egger et al. 2020), and overall it suffers from poor precision, with a specificity of 76 % (Arvanitis et al. 2014). Performance of blood PCR is particularly poor in patients on mould-active prophylaxis (Egger et al. 2020).

Two point-of-care tests are now available for the diagnosis of IA. The Aspergillus-specific Lateral Flow Device (LFD) test (OLM Diagnostics, Newcastle Upon Tyne, United Kingdom) detects an extracellular mannoprotein antigen secreted exclusively during active growth of Aspergillus species via the JF5 monoclonal antibody (Hoenigl et al. 2014b, 2018, Prattes et al. 2015, Orasch et al. 2017). Another new test, the sōna Aspergillus GM-LFA (IMMY, Norman, OK, United States) detects GM but has a shorter turnaround time compared to the conventional GM ELISA test. In patients with hematologic malignancy, both the LFD and LFA have a comparable sensitivity and specificity to GM from BALF for diagnosing IA (Heldt et al. 2018, de Heer et al. 2019, Jenks et al. 2019c, 2021, Mercier et al. 2020), and sensitivity increased further when combined with inflammatory markers, triacetylfusarinine C, or PCR (Hoenigl et al. 2019, Jenks et al. 2019d, 2019e, Rawlings et al. 2019). In non-neutropenic patients, the LFA and LFD have demonstrated a sensitivity and specificity between 60–70 % when used alone and up to 80 % when used in combination (Jenks et al. 2019b), with tendencies towards better performance of the LFA (Jenks et al. 2020). Particularly the LFA, but also the LFD have also shown promise for diagnosing IA in serum samples from patients with haematological malignancies (Jenks & Hoenigl 2020). Unmet needs for the diagnosis of IA include a true point-of-care test that can be done at the bedside or in the clinic in the matter of minutes. In addition, improved diagnostic algorithms to diagnose IA in non-neutropenic patients are needed as well.

The detection of Aspergillus serum precipitin antibodies (i.e., subsets of IgG and IgM antibodies) is useful for the diagnosis of ABPA (Agarwal et al. 2013). The detection of Aspergillus IgG antibodies via commercial ELISAs or POC tests (Richardson & Page 2018) is the mainstay diagnostic test for aspergilloma (Hope et al. 2005) and chronic pulmonary aspergillosis (CPA) (Denning et al. 2016), when used in conjunction with pulmonary imaging, but Aspergillus IgG lacks the specificity in the diagnosis of IA (Richardson & Page 2017), and GM and also the LFD have limited sensitivity in those with CPA (Salzer et al. 2018). Aspergillus IgG are also present in Aspergillus bronchitis, Aspergillus nodule and chronic rhinosinusitis, and can be used for treatment stratification in CPA (Denning et al. 2018).

Clinical treatment of aspergillosis

Triazoles, particularly voriconazole, isavuconazole and posaconazole for invasive infections, and voriconazole or itraconazole for chronic infections, are the first line antifungal agents used to treat aspergillosis (Denning et al. 2016, Cornely et al. 2019a, Jenks et al. 2019a). Voriconazole and isavuconazole exert fungicidal activity by inhibiting ergosterol biosynthesis. Ergosterol is one of the main structural components of the fungal cell membrane, allowing membrane fluidity, proper positioning and function of membrane-integrated proteins, and the cell cycle. Azoles bind and inhibit at the active site of the cytochrome P450 enzyme lanosterol 14-α-demethylase, which contains a heme cofactor and is encoded by two homologous genes, CYP51A and the CYP51B (a third homolog CYP51C has been found in A. flavus) (Dudakova et al. 2017). This induces the accumulation of 14-methyl sterols, such as lanosterol, which alters functions of cell membrane leading to fungal growth arrest. Furthermore, inhibition of ergosterol biosynthesis results in both accumulation of toxic sterol intermediates and creation of cell wall carbohydrate patches that extend to the plasma membrane (Geiβel et al. 2018). Drug-drug interactions may limit the use of voriconazole and other triazoles not only in the ICU setting, but also in patients with haematological malignancies where some newer drugs, including ibrutinib, venetoclax, and midostaurin, may complicate the use of triazoles (Groll et al. 2017, Tapaninen et al. 2020).

The therapeutic approach of IA has been changed with the development of the second-generation mould-active triazoles, voriconazole, posaconazole and isavuconazole in addition to the first-generation azole itraconazole. The chemical structure of voriconazole closely resembles fluconazole and shares a similarity with isavuconazole, while posaconazole more closely resembles itraconazole. Voriconazole was derived from fluconazole by replacing one triazole moiety in fluconazole with a 4-fluoropyrimidine group and adding an α-methyl group (Bellmann & Smuszkiewicz 2017). Posaconazole derives from itraconazole through the replacement of the chlorine substituents with flourine in the phenyl ring, as well as hydroxylation of the triazolone side chain (https://pubchem.ncbi.nlm.nih.gov/). Unlike the other second-generation triazoles, isavuconazole is administered as a prodrug; the isavuconazoilum sulfate ester which is hydrolysed rapidly by serum esterases, is highly water soluble and does not require the addition of a beta-cyclodextrin to facilitate solubility (Jenks et al. 2018).

The availability of both intravenous and different oral formulations of triazoles increases the therapeutic options and improves their pharmacokinetics (Table 1) (Cornely et al. 2019b). The variable pharmacokinetics (80–100 %) of voriconazole (oral solution, tablet and intravenous [IV] formulation) and the oral solution of posaconazole due to erratic hepatic metabolism and absorption, respectively, have been improved with the new formulations of posaconazole i.v./tablet and isavuconazole i.v./capsules (∼50 % variation) (Jović et al. 2019). The bioavailability of oral formulations ranges from poor with the “old” posaconazole oral solution to high with isavuconazole, and intake with fatty food is most important for the posaconazole oral solution (Hoenigl et al. 2014a). All four azoles exhibit non-linear pharmacokinetics because of saturable absorption or metabolism except isavuconazole (Bellmann & Smuszkiewicz 2017). They are highly protein bound (> 98 %) except voriconazole (58 %). All azoles are characterised by a large volume of distributions (3–11 lt/kg) and they are extensively metabolised with minimal amounts of parent drug excreted renally or hepatobiliarily. Their half-lives and total drug exposure area under curve (AUC) varies from short with voriconazole (6–12 h and 13–16 mg∗h/l) to long with isavuconazole (110–130 h and 98–121 mg∗h/l, respectively) (Table 1). As will be discussed in detail later in this review, the emergence of ARAF isolates threatens the efficacy of azoles, and lipid formulations of amphotericin B, as well as echinocandins (which should be used in combination with another antifungal drug) are alternative treatment options for IA (Patterson et al. 2016). Surgery in addition to systemic antifungal therapy plays an important role in the treatment of primary extrapulmonary invasive aspergillosis (Aldosari et al. 2020).

Table 1.

Pharmacokinetic and pharmacodynamic properties of anti-Aspergillus azoles.

Parameters	Voriconazole	Posaconazole	Isavuconazole	Itraconazole
Chemical structure
Molecular weight	349.3	700.8	717.77	705.64
Water solubility (mg/mL)	0.0978	0.012	0.0162	0.00964
Log D (pH 7.4)	1.8	2.15	4.14	> 5
Formulations	Oral solution, tablet, iv	Oral solution, tablet, iv	Capsules, iv	Oral solution, tablet, capsules, iv
Standard dose	LD: 6 mg/kg q12h, MD: 4 mg/kg q12h or 200 mg PO	Oral = 200 mg q6h/q8h or 400 mg q12h Tablet / iv = 300 mg q24h	LD: 200 mg q6h for 2 d; MD: 200 mg q24h	Oral = LD: 800 mg q12h/g24h for 2 d, MD: 200 mg q12h IV = LD: 200 mg q12h for 2d, MD: 200 mg q24h
Dose adjustment	< 40 kg = 50 % MD dose, RI = avoid iv1, HI = 50 % MD, IR = 300mg q12h	RI = avoid iv1	No	RI = avoid iv1
Tmax	1.43–1.81	Oral = 3, tablet = 2.2, iv = 1.5	Oral = 2–3, iv = 0.75–1	Oral = 5, Iv = 1
Bioavailability	90 % (healthy) / 83 % (patients)	Oral = 8–47 %	98 %	Capsules = 22 %; oral solution = 55–92 %
Effect of food	22 % reduction	Oral = ↑ with high fat meal/low pH, tablet = ↓ mucositis, ↑ with high fat meal	No	Capsules = ↑ with high fat meal/low pH, oral solution = ↑ empty stomach
Protein binding	42–58 %	98–99 %	99 %	99.8 %
Vd (Lt/kg)	2–4.6	Oral = 3.7, tablet = 5, iv = 20	6.42	11
Tissue penetration	Brain, ELF	Alveolar cells, liver kidney, lung, myocardium	Muscle, fat, liver, brain	Skin, fat, liver, lung, kidney, spleen, bone, muscle
CL (mL/min/kg)	3–8.3	9.39–16.4 L/h	1.9–5	5.1
Hepatic/Urinary (metabolites and drug) excretion %	20 (M)+< 1(D) / 80 (M)+< 2(D)	77 (M)+< 5(D) / 14 (M)+< 5(D)	46 (M) / 46 (M)+0.38(D)	54 (M)+< 0.03(D) / 35 (M)+3–18(D)
Metabolizing enzymes	CYP2C19 > CYP2C9,CYP3A4	UGT1A4	CYP3A4, CYP3A5, UGT	CYP3A4, CYP2C9, CYP2C19
T1/2 (h)	6–12	27–35	110–130	24
Cmax (mg/L)	4.84 (300 mg), 5.27 (400 mg)	Oral = 0.851, tablet = 1.96, iv = 3.28	2.6	0.5–2.3
AUC0–24 (mg∗h/mL)	13.21–16.38	Oral = 17.24, tablet / iv = 34.3	98–121	29.2
Variation in AUC0–24	82 %	Oral = 82 %, tablet / iv = 35–40 %	10–43 %	30–60 %
Pharmacokinetics	Non-linear (saturable metabolism)	Oral = non-linear (saturable absorption), tablet / iv = linear	Linear	Non-linear (saturable absorption and metabolism)
tAUC/MIC (50 % survival in animals)	17.61–21.96	167–178	25	NA
TDM targets (mg/L)	Prophylaxis: Cmin > 0.5; therapy: Cmin > 1–2; toxicity: Cmin < 4–6	prophylaxis: Cmin > 0.5–0.7; therapy: Cmin > 1–1.25	NA	Prophylaxis: Cmin > 0.5; therapy: Cmin > 0.5–1; toxicity: Cavg < 17 (bioassay), < 3.5 (HPLC)

NA, not applicable.

¹Unless an assessment of the benefit/risk to the patient justifies the use of intravenous formulations.

Antifungal combination therapy with voriconazole or amphotericin B and an echinocandin is often employed as primary or salvage therapy for management particularly of refractory aspergillosis (Elefanti et al. 2013). Resistance to first line triazole antifungal agents among Aspergillus species has resulted in the increased use of second-line monotherapy with echinocandin drugs (caspofungin, micafungin or anidulafungin) (Aruanno et al. 2019). Echinocandin class drugs inhibit the cell wall biosynthetic enzyme β-(1,3)-d-glucan synthase (Perlin 2015), and were initially approved by the FDA for the treatment of invasive aspergillosis refractory to conventional therapy (Johnson & Perfect 2003). Given a strong 20-year history of safety and efficacy, it is being used increasingly in patients being treated for chronic pulmonary aspergillosis. A recent meta-analysis of 12 studies covering 380 patients who received IV antifungals, either amphotericin B (n = 143) or an echinocandin (n = 237) reported a response of 58 % for amphotericin B and 62 % for echinocandins (micafungin). Echinocandins, especially micafungin are well-tolerated and effective prophylactic antifungal agents used in patients with hematologic diseases at high risk for invasive mould infections (Ziakas et al. 2014, Park et al. 2019).

Worldwide emergence of azole-resistant A. fumigatus: environmental and clinical routes

Azole drugs play a major role in prevention and treatment of infections caused by Aspergillus species. Azole drugs belong to a wider group of compounds called demethylation inhibitors (DMIs) whose common target is the 14-α sterol demethylase. Demethylation inhibitors are widely used in the clinical setting, both as treatment and prophylaxis, but also in the agriculture setting due to their high-efficiency and broad-spectrum activity (Price et al. 2015, Hollomon 2017).

Several azole-based fungicides used in agriculture to protect cereals and wine from phytopathogenic moulds have a similar chemical structure to medical triazoles and the development of cross-resistance between them has been proved (Snelders et al. 2012, Zhang et al. 2017). A large number of demethylation inhibitor fungicides have been used intensively in agriculture and medicine (human and veterinary) since 1970 (Rochette et al. 2003). Though azole fungicides are not used to target A. fumigatus, it transpires that many of these fungicides are active against A. fumigatus, a condition that led to the emergence of resistance. More than 30 azole fungicides have been studied for their in vitro activity against A. fumigatus, including propiconazole, bromuconazole, epoxiconazole, difenconazole, and tebuconazole. In this context, one of the presumed routes for triazole-resistance emergence is through selection pressure from the DMIs used as fungicides in the environment (Verweij et al. 2007, 2009). The resulting azole resistant isolates are associated with a particular resistance mechanism constituted by a variable number of tandem repeat (TR) integrations in the CYP51A promoter followed or not by point mutations in the coding gene (Snelders et al. 2008). There is a lot of evidence that supports the idea of an environmental driven mechanism, such as the fact that these TR azole resistant strains (TR₃₄/L98H, TR₃₄/L98H/S297T/F495I, TR₄₆/Y121F/T289A, and TR₅₃) have been detected throughout the world (Garcia-Rubio et al. 2017) but showed shorter genetic distances among them than with other azole-susceptible and non TR-resistant isolates, which suggests that they could have developed from a reduced set of clonally related strains (Snelders et al. 2008, Camps et al. 2012b, Garcia-Rubio et al. 2018b). Besides, the fact that TR azole resistant strains have been isolated from azole-naïve patients with IA also endorses the existence of this environmental resistance route, indicating that infections would have been caused by the inhalation of already ARAF spores harbouring aforementioned mutations in CYP51A (Snelders et al. 2009, Verweij et al. 2009). However, other single mutations occurring in CYP51A arise during the course of azole therapy (Howard et al. 2006, 2009, Albarrag et al. 2011, Camps et al. 2012c, Wiederhold et al. 2016).

Nevertheless, there are also some findings that support that TR azole-resistance mechanisms do not seem to be restricted to the environment. A clinical case of fatal aspergillosis caused by an A. fumigatus strain that developed a TR₁₂₀ insertion in the CYP51A promoter during long-term azole treatment has been recently reported using both STRAf typing and whole-genome sequencing (Hare et al. 2019). This challenges the existence of a link between resistance mechanisms and specific routes of resistance selection and may fade the presumed boundaries between the environmental and clinical routes of resistance. In line with these facts, strains carrying G432S and G432A mutations have been isolated from azole-naïve patients (Howard et al. 2006, 2009, Alanio et al. 2011, Albarrag et al. 2011), while strains carrying TR₅₃ (and also TR₁₂₀) mutations have been isolated from patients exposed to azole antifungals in prior treatments (Hodiamont et al. 2009, Hare et al. 2019). Furthermore, studies about how these supposedly environmental resistance mechanisms originated are still scarce, although it has been hypothesised that environmental conditions or even a more complex reproductive method, such as sexual reproduction, could play a role. In addition, the dispersion of A. fumigatus spores from the human lung into the environment has been suggested lately as a possible transmission path in cystic fibrosis patients as an alternative transmission route from patient to environment (Engel et al. 2019). Although azole resistance may predominantly originate from environmental sources, further research is warranted in order to gain a deeper knowledge about how azole resistance emerges and is transmitted, which has implications for patient management.

The worldwide clinical burden of ARAF

The increasing burden of azole resistance on a global scale notoriously limits the therapeutic options to treat aspergillosis (Denning & Perlin 2011). Over the last two decades, a rapid local and global emergence of triazole resistance has been observed. The first ARAF isolates were obtained from two patients treated with itraconazole, one of whom died early 1990, in California, in a case that dates back to the late 1980s (Denning et al. 1997b). A Dutch study later reported three ARAF isolates recovered after long-term itraconazole treatment from a lung transplant recipient in 1997 (Verweij et al. 2002). Moreover, a study in France found four itraconazole-resistant isolates with high itraconazole minimum inhibitory concentrations (MICs) values, > 16 mg/L, in 1999 (Dannaoui et al. 1999b). Later, in 2007, a comprehensive study of nine cases of azole-resistant IA found that four out of nine patients had never previously been treated with azole antifungals (Hussain et al. 2007).

Studies have investigated the distribution of ARAF in relation to the TR₃₄/L98H mutation, which was first found in the Netherlands in 1998 (Snelders et al. 2008). Indeed, surveillance studies and case series recently reported the global presence of ARAF harbouring TR₃₄/L98H, with reports from Europe, the Middle East, including Australia, Tanzania, Kuwait and Iran, North and South Asia, including China and Japan, Australia and Tanzania; Africa, and North and South America, including Brazil and Columbia (Mellado et al. 2007, Verweij et al. 2007, Snelders et al. 2008, Howard et al. 2009, Lockhart et al. 2011, van der Linden et al. 2011, 2013, Chowdhary et al. 2012a, 2015, Rath et al. 2012, Seyedmousavi et al. 2013b, Astvad et al. 2014, Ahmad et al. 2015, Choukri et al. 2015, Fuhren et al. 2015, Liu et al. 2015, Steinmann et al. 2015, Vermeulen et al. 2015, Wu et al. 2015, 2020, Chen et al. 2016b, Jensen et al. 2016, Lazzarini et al. 2016, Mushi et al. 2016, Nabili et al. 2016, Perveen et al. 2016, Wiederhold et al. 2016, Koehler et al. 2017, Montesinos et al. 2017, Prigitano et al. 2017, Toyotome et al. 2017, Abdolrasouli et al. 2018a, 2018b, Berkow et al. 2018, Denardi et al. 2018, Pinto et al. 2018, Riat et al. 2018, Seufert et al. 2018, Talbot et al. 2018, Gonzalez-Lara et al. 2019, Rivero-Menendez et al. 2019b, Tsuchido et al. 2019, Li et al. 2020, Pontes et al. 2020) (Table 2). These studies describe the most recent discoveries of the TR₄₆/Y121F/T289A resistance mechanism involving environmental mutations, which the Netherlands first reported in 2009 (van der Linden et al. 2013); another report revealed that a US patient was recovering from TR₄₆/Y121F/T289A A. fumigatus infection in 2008 (Wiederhold et al. 2015).

Table 2.

Diversity and prevalence of CYP51A mutations causing azole resistance in clinical A. fumigatus isolates.

MIC values (μg/mL)					Azole exposure prior resistance		Azole therapeutic failure
Country (year; number of isolates)	Amino acid substitutions (total number) (references)	VRZ	PSZ	ITZ	Present	Naïve	VRZ	PSZ	ITZ
the Netherlands (1994–2013; 186), Germany (2002–2018; 111), Italy (1995–2006, 2013–2015; 28), China (2008–2016; 19), Belgium (2011–2016; 18), Taiwan (2011–2018; 16), Denmark (2010–2014; 12), India (2008–2014; 11), UK (1992–2017; 11), USA (2001–2017; 7), Iran (2003–2016; 5), Tanzania (2016; 5), France (2012; 3), Spain (2014–2018; 3), Brazil (2014–2018; 3), Pakistan (2016; 3), Japan (2016–2018; 3), Switzerland (2016; 2), Portugal (2010–2016; 2), Kuwait (2015; 2), Mexico (2014–2017; 2), Australia (2015–2017; 1)	TR34 / L98H (T297S / F495I) (453) (Mellado et al. 2007, Verweij et al. 2007, Snelders et al. 2008, Howard et al. 2009, Lockhart et al. 2011, van der Linden et al. 2011, 2013, Rath et al. 2012, Seyedmousavi et al. 2013b, Ahmad et al. 2015, Chowdhary et al. 2015, Liu et al. 2015, Vermeulen et al. 2015, Wu et al. 2015, 2020, Chen et al. 2016b, Jensen et al. 2016, Lazzarini et al. 2016, Mushi et al. 2016, Nabili et al. 2016, Perveen et al. 2016, Wiederhold et al. 2016, Koehler et al. 2017, Montesinos et al. 2017, Prigitano et al. 2017, Abdolrasouli et al. 2018b, Berkow et al. 2018, Denardi et al. 2018, Pinto et al. 2018, Riat et al. 2018, Seufert et al. 2018, Talbot et al. 2018, Gonzalez-Lara et al. 2019, Rivero-Menendez et al. 2019b, Tsuchido et al. 2019, Pontes et al. 2020)	0.5–>16	0.125–>8	>4–32	Yes FLZ VRZ ITZ PSZ	Yes	Yes	Yes	Yes
the Netherlands (2009–2013; 29), Belgium (2011–2016; 11), Denmark (2010–2014; 6), Germany (2015–2018; 3), USA (2001–2016; 3), China (2010–2016; 3), France (2014; 2), Spain (2014–2018; 2), UK (2016; 1), Portugal (2010–2016; 1), Argentina (2009; 1), Japan (2013; 1), Taiwan (2011–2018; 1)	TR46 / Y121F / T289A (64) (Verweij et al. 1998, Kuipers et al. 2011, Lockhart et al. 2011, Vermeulen et al. 2012, 2015, Montesinos et al. 2013, van der Linden et al. 2011, 2013, Astvad et al. 2014, Chen et al. 2015, 2016b, Peláez et al. 2015, Steinmann et al. 2015, Jensen et al. 2016, Vazquez & Manavathu 2016, Wiederhold et al. 2016, Lavergne et al. 2017, Moore et al. 2017, Isla et al. 2018, Pinto et al. 2018, Seufert et al. 2018, Rivero-Menendez et al. 2019b, Li et al. 2020, Wu et al. 2020)	>8–>32	0.125–2	1–>16	ITZ, VRZ PSZ FLZ	Yes	Yes	Yes	Yes
The Netherlands (2009; 1)	TR53 (1) (Hodiamont et al. 2009)	16	0.5	16	ITZ		Yes/No	No	Yes
Denmark (2016; 1)	TR120 (1) (Hare et al. 2019)	4	0.5	16–>16	Yes VRZ PSZ		Yes	Yes	ND
Germany (2018; 8), Japan (2012; 8), Italy (1995–2006; 7), UK (1992–2009; 6), USA (2001–2014; 5), China (2001; 3), Australia (2015–2017; 2), the Netherlands (2007–2009; 2), India (2011–2014; 2), Spain (2014–2018; 2), Denmark (2010–2014; 1)	G54R/E/W (46) (Chen et al. 2005, Bueid et al. 2010, van der Linden et al. 2011, Camps et al. 2012c, Tashiro et al. 2012, Chowdhary et al. 2015, Wiederhold et al. 2015, 2016, Jensen et al. 2016, Lazzarini et al. 2016, Talbot et al. 2018, Rivero-Menendez et al. 2019b)	0.125–4	1–>16	1–>16	Yes ITZ PSZ VRZ	ND	Yes	ND	Yes
UK (1992–2007; 16), USA (2001–2014; 2)	G138C/S (18) (Howard et al. 2006, 2009, Albarrag et al. 2011, Wiederhold et al. 2016)	16–8	1–4	>8–16	Yes ITZ VRZ	ND	Yes	ND	Yes
The Netherlands (2007–2009; 3), India (2012–2016; 3), Denmark (2010–2014; 2), UK (1992–2009; 2), USA (2015–2017; 1), Japan (2009–2011; 1)	P216L (12) (Howard et al. 2009, Bueid et al. 2010, Camps et al. 2012c, Hagiwara et al. 2014, Ahmad et al. 2015, Jensen et al. 2016, Berkow et al. 2018, Dabas et al. 2018)	0.25–2	0.06–2	2–>16	ND	ND	ND	ND	ND
The Netherlands (2007–2009; 8), USA (2001–2014; 1), Italy (2013–2015; 1)	F219I/S/L (10) (van der Linden et al. 2011, Wiederhold et al. 2016, Prigitano et al. 2017)	0.25–8	0.25–>16	>16	Yes ITZ	ND	Yes	Yes	Yes
UK (1992–2009; 10), USA (2001–2014; 4), the Netherlands (1994–2007; 1), Denmark (2010–2014; 1), Sweden (1997–1998; 1), Japan (2017–2018; 1), China (2001; 1)	M220V/K/I/R/W (19) (Chen et al. 2005, Snelders et al. 2008, Howard et al. 2009, Bueid et al. 2010, Jensen et al. 2016, Wiederhold et al. 2016, Dabas et al. 2018, Tsuchido et al. 2019)	0.5–4	0.5–>4	>4–>16	Yes ITZ				ITZ
Japan (2000; 1)	F332K (Asano et al. 2011)	0.25	0.5	32	ND	ND	ND	ND	ND
USA (2001–2014; 4), UK (1992–2009; 3), Belgium (2015–2016; 2), Spain (2011; 1), Japan (2017–2018; 1), France (2006; 1)	G448S (12) (Howard et al. 2009, Bellete et al. 2010, Bueid et al. 2010, Peláez et al. 2012, Wiederhold et al. 2016, Montesinos et al. 2017)	0.5–>16	0.25–4	0.5–>16	Yes ITZ VRZ	ND	Yes	ND	Yes
UK (1992–2007; 2), India (2012–2016; 2)	Y431C (4) (Howard et al. 2006, 2009, Albarrag et al. 2011, Dabas et al. 2018)	0.5–4	0.06–2	>8–16	ND	ND	ND	ND	ND
China (2011–2014; 1), France (2006–2009; 1)	G432A/S (2) (Alanio et al. 2011, Liu et al. 2015)	0.25–2	0.25–0.5	4–16	ND	ND	ND	ND	ND
UK (1992–2007; 2)	G434C (2) (Howard et al. 2006, 2009, Albarrag et al. 2011)	4	1	>8	No	Yes	ND	ND	ND
UK (1992–2017; 70),# China (2008–2009; 16), Spain (2014–2018; 17), Japan (2009–2018; 8), Denmark (2010–2014;7), USA (2015–2017; 7), Italy (1995–2006, 2013–2015; 6), Taiwan (2011–2018; 6), the Netherlands (1994–2013; 4), Belgium (2011–2016; 4), Portugal (2008–2016; 4), Sweden (1997–1998; 2), South Korea (2012–2013; 2), Iran (2003–2016; 2), Brazil (2008–2009; 1), Czech Republic (2008–2009; 1), Germany (2015; 1)	WT (158) (Chryssanthou 1997, Snelders et al. 2008, Bueid et al. 2010, Lockhart et al. 2011, Tashiro et al. 2012, Hagiwara et al. 2014, Fuhren et al. 2015, Steinmann et al. 2015, Vermeulen et al. 2015, Wiederhold et al. 2015, 2016, Jensen et al. 2016, Lazzarini et al. 2016, Nabili et al. 2016, Lavergne et al. 2017, Montesinos et al. 2017, Prigitano et al. 2017, Abdolrasouli et al. 2018b, Berkow et al. 2018, Isla et al. 2018, Pinto et al. 2018, Seufert et al. 2018, Tsuchido et al. 2019, Won et al. 2020)	0.25–16	0.06–16	0.5–>16	ITZ	ND	ND	ND	Yes

MIC, minimum inhibitory concentration; FLZ, fluconazole; ITZ, itraconazole; PSZ, posaconazole; VRZ, voriconazole.

In the following years, many more studies have been reported, from Argentina, China, Europe (Portugal, Spain, UK), Japan, Taiwan and US (Table 2) (Lockhart et al. 2011, van der Linden et al. 2013, Astvad et al. 2014, Steinmann et al. 2015, Vermeulen et al. 2015, Chen et al. 2016b, Jensen et al. 2016, Wiederhold et al. 2016, Lavergne et al. 2017, Montesinos et al. 2017, Isla et al. 2018, Pinto et al. 2018, Seufert et al. 2018, Li et al. 2020, Wu et al. 2020). Wild-type isolates have developed resistance to itraconazole in many countries and continents, such as the UK and the USA and in Europe and South Asia (Chryssanthou 1997, Snelders et al. 2008, Bueid et al. 2010, Lockhart et al. 2011, Tashiro et al. 2012, van der Linden et al. 2013, Hagiwara et al. 2014, Fuhren et al. 2015, Steinmann et al. 2015, Vermeulen et al. 2015, Jensen et al. 2016, Lazzarini et al. 2016, Nabili et al. 2016, Wiederhold et al. 2016, Prigitano et al. 2017, Abdolrasouli et al. 2018b, Berkow et al. 2018, Seufert et al. 2018). Other single nucleotide polymorphisms, in gene positions G54, M220, and G448 of the CYP51A gene, have been reported to be frequently observed in patients with chronic pulmonary aspergillosis, Invasive aspergillosis bronchitis, aspergilloma, and chronic cavitary pulmonary aspergillosis (CCPA) treated long term with azole antifungals, as well as several clinical treatment failures (Chryssanthou 1997, Chen et al. 2005, 2016b, Snelders et al. 2008, Howard et al. 2009, Bellete et al. 2010, van der Linden et al. 2011, Camps et al. 2012c, Tashiro et al. 2012, Chowdhary et al. 2015, Wiederhold et al. 2016, Montesinos et al. 2017, Dabas et al. 2018, Denardi et al. 2018, Riat et al. 2018, Seufert et al. 2018, Talbot et al. 2018, Tsuchido et al. 2019). Moreover, other single point mutations associated with resistance to azole antifungals have also been reported: G138C, F219I, P216L, G432S, and G432A (Howard et al. 2006, 2009, Bueid et al. 2010, Albarrag et al. 2011, van der Linden et al. 2011, Camps et al. 2012c, Hagiwara et al. 2014, Jensen et al. 2016, Wiederhold et al. 2016, Berkow et al. 2018, Dabas et al. 2018) (Table 2).

While matched control studies involving patients infected with azole-resistant and/or azole-susceptible isolates have not been conducted, patients with azole-resistant Aspergillus infections are at high risk for therapeutic failure. In the Netherlands surveillance study, a high mortality rate was reported among culture-positive ICU patients with ARAF; 14 patients, 10 of whom died, were identified with azole-resistant IA and several underlying conditions, such as autoimmune hepatitis, allogeneic stem cell transplant, hematologic malignancy non-small cell lung cancer, and chronic obstructive pulmonary diseases (COPD) (van Paassen et al. 2016). Two patients died in Belgium of IA that progressed to cerebrospinal aspergillosis (Vermeulen et al. 2012). Other cases support these findings, including a recent German study in which seven of eight azole-resistant IA patients experienced failed therapeutic treatment and died (Salsé et al. 2019). It has also been shown that CPA patients have failed azole treatment due to azole resistance (Howard et al. 2009, Steinmann et al. 2015), while several other case series reported mortality rates of 50–100 % in patients with triazole-resistant IA. Resistance rates as high as 29 % have been observed in specific patient populations, such as critically ill patients (van Paassen et al. 2016, Verweij et al. 2016a).

Worldwide burden of ARAF in the environment

In recent years, an increasing proportion of A. fumigatus isolates has been observed to be resistant in patients (Table 2) and environments (Table 3) due to the presence of mutations in the CYP51A gene (Snelders et al. 2008). Numerous fungicides were able to inhibit wild-type strains, and some azole fungicides were active against wild-type strains but ineffective against isolates with the TR₃₄/L98H mutations, which have high MICs. In the Netherlands, these fungicides were introduced between 1990 and 1996, just before the first clinical TR₃₄/L98H strain was found in 1998 (Snelders et al. 2008). Indeed, evidence of an ARAF environmental route of acquisition was first found in the Netherlands (Zhang et al. 2017). As a result, it has become evident that azole resistance has a potentially global distribution and is therefore a global problem (Mortensen et al. 2010, Verweij et al. 2016a).

Table 3.

Diversity and prevalence of CYP51A mutations causing azole resistance in environmental A. fumigatus isolates.

MIC1 values (mg/L)					Azole exposure prior resistance		Source of isolate
Country (year of isolation; number of isolation)	Amino acid substitutions (total number) (references)	VRZ	PSZ	ITZ	Present	Naive
The Netherlands (2002–2019; 203); Denmark (2009; 4); India (2011–2014; 56); Iran (2013; 5); China (200–2015; 4); Thailand (2014–2015; 8); Italy (2011–2018; 61); Romania (2013–2014; 16);UK (2009–2018; 43); France (2010–2017; 62); Taiwan (2014–2018; 33); Tanzania (2013–2014; 11); Australia (2013–2017; 1); Kuwait (2013–2015; 9); USA (2015; 38); Germany (2012–2013; 45); Colombia (2015; 2)	TR34/L98H (S297T/F495I/Q141H/S52T) (601) (Verweij et al. 2009, Mortensen et al. 2010, Chowdhary et al. 2012b, 2014a, 2014b, Badali et al. 2013, Ahmad et al. 2014, 2015, Bromley et al. 2014, Prigitano et al. 2014, Wang et al. 2014, Bader et al. 2015, Sharma et al. 2015, Chen et al. 2016b, 2019b, Nabili et al. 2016, van der Linden et al. 2016, Álvarez-Moreno et al. 2017, 2019, Hurst et al. 2017, Jeanvoine et al. 2017, Lavergne et al. 2017, Tangwattanachuleeporn et al. 2017, Rocchi et al. 2018, Talbot et al. 2018, Trovato et al. 2018, Tsitsopoulou et al. 2018, Sewell et al. 2019, Ahangarkani et al. 2020)	0.5–>32	0.25–3	>4–>16	Yes Bromuconazole, cyproconazole, difenoconazole, epoxiconazole, hexaconazole, metconazole, penconazole, propiconazole, tebuconazole, tricyclazole, triticonazole	No/Yes	Air, patient room, water filter, soil, corps, seeds, rose pot compost, compost. Hospital garden, cucurbit fields
Iran (2017–2018; 3); The Netherlands (2009–2011; 14); UK (2018; 6); Tanzania (2013–2014; 4); France (2014–2017; 31); Colombia (2015–2018; 38); India (2012–2013; 6); Germany (2012–2013; 6); Taiwan (2016–2018; 3); Greece (2016–2017; 1)	TR46 / Y121F / T289A (112) (van der Linden et al. 2013, 2016, Chowdhary et al. 2014a, 2014b, Bader et al. 2015, Sharma et al. 2015, Álvarez-Moreno et al. 2017, 2019, Lavergne et al. 2017, Rocchi et al. 2018, Chen et al. 2019b, Sewell et al. 2019, Siopi et al. 2020)	>2–≥16	0.25–2	>8–>16	Yes	No/Yes	Air, patient room, water filter, soil, corps, seeds, rose pot compost, compost, raisins, grapes. Hospital garden, cucurbit fields, farmland
India (2013–2014; 1); Romania (2013–2014; 7); Tanzania (2013–2014; 13); Australia (2013–2017; 2); Germany (2012–2013; 1); Thailand (2014–2015; 2)	G54R/E/W/A (26) (Bader et al. 2015, Sharma et al. 2015, Tangwattanachuleeporn et al. 2017, Talbot et al. 2018)	0.25–>32	0.25–2	16–>32	Yes Bromuconazole, cyproconazole, difenoconazole, epoxiconazole, hexaconazole, metconazole, penconazole, tebuconazole, tricyclazole, triticonazole	ND	Hospital soil, woody debris of trunk hollows, garden soil
Australia (2013–2017; 41); Germany (2012–2013; 1); Taiwan (2014–2016; 5); Colombia (2015–2018; 5)	WT(52) (Bader et al. 2015, Álvarez-Moreno et al. 2017, 2019, Talbot et al. 2018, Wang et al. 2018b)	1–>32	0.125–1	0.25–>32	Yes	No/Yes	Azole-naïve and azole exposed soil, air
Italy (2011-2012; 1) Australia (2013-2017; 1); Germany (2012–2013; 1)	F46Y; M172I/V; N248K/T; D255E (3) (Prigitano et al. 2014, Bader et al. 2015, Talbot et al. 2018)	1–>32	0.5–2	>16–>32	Yes	ND	Cucurbit fields
France (2014–2016; 20)	P216L (Lazzarini et al. 2016, Jeanvoine et al. 2017)	0.5	0.5	>16	Yes	ND	Sawmills and air
Germany (2012–2013; 1)	M220I (Bader et al. 2015)	1–>32	0.125–1	0.125–>32	ND	ND	Soils
Colombia (2015–2018; 4)	TR53 (van der Linden et al. 2016, Álvarez-Moreno et al. 2017, 2019)	4	ND	4–>8	Yes	No	Flower fields, soil and greenhouse soil

¹MIC, minimum inhibitory concentration; ITZ, itraconazole; PSZ, posaconazole; VRZ, voriconazole.

Such environmental route-related mutations are present in many geographically diverse countries and continents (Table 3). Many countries have reported environmental ARAF isolates harbouring TR₃₄/L98H, including Australia, China, Colombia, Denmark, France, Germany, India, Iran, Italy, Kuwait, the Netherlands, Romania, Taiwan, Tanzania, Thailand, UK and the US (Verweij et al. 2009, Mortensen et al. 2010, Chowdhary et al. 2012b, 2014a, Badali et al. 2013, van der Linden et al. 2013, Ahmad et al. 2014, 2015, Bromley et al. 2014, Prigitano et al. 2014, Wang et al. 2014, Bader et al. 2015, Sharma et al. 2015, Chen et al. 2016b, Nabili et al. 2016, van der Linden et al. 2016, Álvarez-Moreno et al. 2017, Hurst et al. 2017, Jeanvoine et al. 2017, Tangwattanachuleeporn et al. 2017, Rocchi et al. 2018, Talbot et al. 2018, Trovato et al. 2018, Tsitsopoulou et al. 2018, Chen et al. 2019b, Sewell et al. 2019, Ahangarkani et al. 2020) (Table 3). Additionally, several studies have reported ARAF isolates harbouring TR₄₆/Y121F/T289A in Iran, the Netherlands, the UK, Tanzania, France, Colombia, India, Germany, and Taiwan (Chowdhary et al. 2014a, 2014b, Bader et al. 2015, Sharma et al. 2015, Nabili et al. 2016, van der Linden et al. 2016, Álvarez-Moreno et al. 2017, Lavergne et al. 2017, Rocchi et al. 2018, Chen et al. 2019b, Sewell et al. 2019). Moreover, wild-type strains have been reported in several countries, including Australia, Taiwan, Germany, and Colombia, for which the MICs values of itraconazole, voriconazole, and posaconazole ranged from 0.125 to > 32 (Bader et al. 2015, Álvarez-Moreno et al. 2017, 2019, Wang et al. 2018b).

Increasingly, other point mutations in azole-resistant strains have been reported in studies in Europe, North America, and Asia (Prigitano et al. 2014, Bader et al. 2015, Sharma et al. 2015, van der Linden et al. 2016, Álvarez-Moreno et al. 2017, 2019, Jeanvoine et al. 2017, Tangwattanachuleeporn et al. 2017, Talbot et al. 2018). Therefore, it is crucial to change processing practices to reduce the use and spread of azole fungicides in the environment that result in cross-resistance with medical azoles. Nonetheless, while TR₃₄/L98H and TR₄₆/Y121F/T289A mutations are currently the most prevalent mutations associated with the environmental route, ARAF strains without the CYP51A gene mutations may also emerge from the environment (Table 3). Indeed, the evidence demonstrates that environmental azole-resistance is increasing due to azole fungicide drugs in the environment; further study is needed.

Antifungal tolerance in A. fumigatus

Antifungal tolerance is different from antifungal resistance and a relatively new concept for medical mycology. Among the human pathogenic fungi, antifungal tolerance has been mainly studied in Candida albicans and Candida glabrata. Tolerance is defined as the ability of a subpopulation of cells (> 1 %) in a drug-susceptible strain to persist or grow in presence of drug concentrations above the minimal inhibitory concentration (MIC), often resulting in “trailing growth” in MIC assays. This is in contrast to heteroresistance or persistence, where very rare cells (< 1 %) grow above the MIC (Delarze & Sanglard 2015, Berman & Krysan 2020). Candida albicans tolerance is most commonly seen with azole antifungals and is quantified by disk diffusion or broth microdilution assays; in C. glabrata tolerance occurs readily with echinocandins. Tolerance is affected by strain genetics, especially chromosomal rearrangements or mutations in genes that participate in core stress pathways; it is also influenced by differences in growth conditions such as pH, temperature and nutrient availability and in physiological differences between genetically identical cells (Berman & Krysan 2020). The clinical relevance of antifungal tolerance in Candida species is still unresolved, with some studies failing to find a connection (Rueda et al. 2017), whereas others show a positive correlation between tolerance and disease persistence (Astvad et al. 2018, Healey & Perlin 2018, Rosenberg et al. 2018).

In A. fumigatus, tolerance has been best characterised for the echinocandin antifungals: micafungin, anidulafungin and especially caspofungin, which inhibit fungal β-1,3-d-glucan synthase activity, thereby disrupting cell wall integrity (Patil & Majumdar 2017). Echinocandin tolerance is defined by partial growth inhibition or trailing, under increasing drug concentrations, followed by an unusual phenomenon called the “paradoxical effect”, during which hyphal growth intensifies despite increasing drug concentration (Wagener & Loiko 2017) until complete inhibition (MIC) at very high drug levels. Growth inhibition is characterised by the formation of stubby, highly budded compact colonies in which the growing hyphae undergo tip lysis followed by regenerative intrahyphal growth initiating from viable internal compartments (Moreno-Velásquez et al. 2017), with the minimal effective concentration (MEC) defined as the lowest drug concentration that induces compact colony formation. The mechanism(s) underlying A. fumigatus tolerance and paradoxical growth are not fully understood. A factor in overcoming echinocandin stress is the upregulation of chitin synthesis and a large increase in cell wall chitin levels (Walker et al. 2015). Three signalling pathways, the high osmolarity glycerol (HOG), HSP90/calcineurin and cell wall integrity pathway, upregulate chitin synthesis in response to echinocandin treatment (Wagener & Loiko 2017). Genetic or pharmacological targeting of these pathways blocked the upregulation of chitin synthesis and abolished A. fumigatus tolerance and paradoxical growth. Another important event that occurs after sustained echinocandin exposure is the relocalisation of glucan synthase from the vacuoles to the hyphal tips (Moreno-Velásquez et al. 2017). This is accompanied by increased cell wall β-1,3-d-glucan, and a reduction in chitin (Loiko & Wagener 2017). In animal models, increased drug levels elevate fungal burden but have no effect on overall survival (Wagener & Loiko 2017). Compact fragmented colonies are seen by histology, which may increase CFU counts while decreasing tissue penetration, organ injury and mortality (Petraitiene et al. 2002). Thus, the clinical relevance of echinocandin tolerance and the paradoxical effect is not yet clear.

In contrast to the literature on Candida spp., very few reports describe azole tolerance in aspergilli. Trailing reported for some clinical isolates of A. flavus and A. niger was dependent on inoculum size and growth medium, respectively (Mosquera et al. 2001, Wang et al. 2018a). Trailing also was described for A. fumigatus exposed to the allylamine antifungal terbinafine (Moore et al. 2001). No trailing growth has been reported with amphotericin B exposure. Recently, an important new mechanism for tolerance and clinical resistance to echinocandins emerged in A. fumigatus in which it was reported that caspofungin induces a change in the membrane sphingolipid content rendering glucan synthase insensitive to the drug (Satish et al. 2019).

In summary, for A. fumigatus, antifungal tolerance to echinocandins is associated with physiological changes in cell wall and membrane composition; tolerance to antifungal azoles remains to be investigated.

Azole resistance mechanisms

Azole resistance in A. fumigatus is mainly associated with the acquisition of mutations in and overexpression of CYP51A, and overexpression of efflux pumps, as discussed in detail below. Adaptation to a new environment before acquisition of stable mutations in the azole drug target, e.g., CYP51A, requires orchestration of a rapid, robust, and coordinated response that allows the cell to thrive despite the presence of drugs. It has been shown that mounting an appropriate physiological response and phenotypic plasticity to itraconazole may take 60 min (Hokken et al. 2019). Among the most highly expressed genes 30 and 60 min after itraconazole exposure are efflux pump genes mdr1 and mdr4; hmg1, encoding a 3-hydroxyl-3-methylglutaryl-CoA 1 reductase from the mevalonate pathway; and ERG6, the most upregulated gene at all time points studied, involved in ergosterol biosynthesis (Hokken et al. 2019) (Fig. 2). Further, increased expression of hmg1 may lead to an increased production of mevalonate, a precursor in sterol biosynthesis, which can positively regulate the overexpression of ERG6, resulting in increased eburicol levels (Yasmin et al. 2012, Hokken et al. 2019). The higher activity of ERG6 also results in the higher quantity of eburicol, the substrate for CYP51A, which also results in higher expression of CYP51A to keep up with the substrate elevation. The detailed mechanisms of azole resistance are discussed below (Fig. 2).

Fig. 2.

The ergosterol pathway involves multiple enzymes and is tightly regulated by the key rate-limiting enzyme Hmg1 producing mevalonate. Lanosterol produced by ERG6 is the substrate catalysed by CYP51A and CYP51B, while azoles competitively occupy the catalytic site and hence reduce the ergosterol synthesis. Adopted from Moreno-Velásquez et al. (2017), with permission.

Role of CYP51A and CYP51B in azole resistance

In Candida species, azole drugs target ERG11 catalyses demethylation of 14-α-lanosterol. In A. fumigatus, ERG6 catalyses this reaction and acts upstream of the triazole drug target cytochrome P450 51 (Fig. 2). Cytochrome P450 51 is encoded by two isoforms, CYP51A and CYP51B (Alcazar-Fuoli et al. 2008, Hokken et al. 2019), which share 59–63 % similarity (Warrilow et al. 2010, Hargrove et al. 2015). A gene essentiality study in A. fumigatus based on the gene expression levels revealed that CYP51A encodes the major enzyme required for mycelial growth; the biological function of CYP51B remains elusive (Hu et al. 2007). Further, deletion of CYP51A abrogates fluconazole resistance, but this effect is not observed when CYP51B is deleted (Mellado et al. 2005). In addition, CYP51A weakly binds to triazoles, which is in contrast with tighter binding observed for CYP51B (Warrilow et al. 2010). Collectively, these observations indicate that CYP51A is the major enzyme in the ergosterol biosynthesis pathway and azole resistance. The presence of both CYP51A and CYP51B is vital to cell survival, and deletion of either isoform is compensated by the presence of other, without major apparent abnormalities in cell morphology (Hu et al. 2007). Further, the intrinsic resistance of Aspergillus to fluconazole is thought to be mediated by I301T substitution in CYP51A and a higher expression of CYP51A than CYP51B upon fluconazole exposure (Blosser & Cramer 2012, Leonardelli et al. 2016).

Mutations in CYP51A and CYP51B

Mutations in the CYP51A gene identified in clinical and environmental ARAF isolates are listed in Table 2, Table 3. Most of these mutations within the CYP51A coding sequence are accompanied by tandem repeats (TRs) in the promoter region, such as TR₃₄/L98H or TR₄₆/Y121F/T289A (TR-mediated azole resistance mechanism is detailed in the following section) (Fig. 3). The effect of the mutations on MIC values and CYP51A structure requires high-resolution structure and/or simulation analysis. Although the crystal structure of CYP51B has been determined, that of CYP51A has not yet been defined (Hargrove et al. 2015). Therefore, the association of such mutations with azole resistance is mainly derived from heterologous expression experiments and simulation studies, in which CYP51A from human, Mycobacterium tuberculosis, and Saccharomyces cerevisiae, and CYP51B from A. fumigatus are used as models to evaluate the effect of specific amino acid substitutions on protein structure (Liu et al. 2016, Nash & Rhodes 2018). According to in silico modelling, the impact of an amino acid substitution on CYP51A structure depends on the position and the substituted residue (Liu et al. 2016). For instance, substitutions of G54, L98, M220, and Y431 decrease the binding affinity of CYP51A to azoles, while substitutions of G432 and also L98 reduce the stability of CYP51A, which can lead to conformational changes in the substrate and/or inhibitor binding pocket by causing dramatic changes in specific loop structures close to these sites (Liu et al. 2016, Nash & Rhodes 2018). Furthermore, L98H substitution reduces hydrogen bond formation between the residue at site 98 and polar side chains of adjacent residues, which could prevent docking of triazoles in the binding pocket (Nash & Rhodes 2018). Typically, L98 changes are accompanied by amino acid substitutions of S297 and F495, which are both adjacent to the binding pocket and, hence, may synergistically confer azole resistance (Liu et al. 2016). Conversely, G54, G138, and M220 are close to the opening channels 1 and 2, which are close to the ligand access channel. Based on the same analysis, substitutions of amino acids located on the periphery of the protein, such as E130D, L252L, S400I, F46Y, M172V, N248T, D255E, L358L, E427K, and C454C, do not cause pronounced conformational changes as they are far from the critical residues and, therefore, do not play a role in azole resistance (Liu et al. 2016).

Fig. 3.

Molecular mechanisms contributing to triazole resistance observed in A. fumigatus. The scheme represents an A. fumigatus cell - with particular focus on the nucleus (N), the endoplasmic reticulum (ER), and the plasma membrane (PM), which depicts the most relevant mechanisms of triazole resistance in this fungus. In A. fumigatus, a major role in ergosterol biosynthesis is played by the sterol demethylase CYP51A (a1). The CYP51A gene is regulated positively by the SrbA protein, which activates its expression by binding to two Sterol Regulatory Elements (SRE) in the promoter region. When ergosterol biosynthesis is repressed, the access of SrbA to SREs is prevented by both the CBC complex and the HapX transcription factor binding to regulatory elements located downstream of SREs, resulting in negative regulation of CYP51A expression. The sterol demethylase CYP51A, whose native substrate is eburicol, an intermediate of ergosterol biosynthesis, is the target of azole drugs (a2). As a result, changes in CYP51A sequence or expression are associated with increased MIC to triazoles. Amino acid substitutions in either the ligand binding site or the catalytic site (b1) modulates triazole binding affinity to CYP51A (b2). A different mutation that can be found in combination with SNPs in the CYP51A gene is the presence of tandem repeats (TRs) in the promoter region, resulting in an expansion of the SREs, unimpeded SrbA binding, and ultimately hyper-activation of CYP51A expression (c1). The same outcome was observed in the case of the P88L mutation in the HapE subunit of the CCAAT-binding complex (CBC) complex, which diminishes CBC binding affinity and its negative regulation of CYP51A expression, although this genotype had only been observed in the clinical isolates in which it was first described (d1). In both cases, the increased amount of the CYP51A enzyme prevents saturation by triazoles and sustains ergosterol biosynthesis (c2 and d2). As for other pathogenic fungi, overexpression of either ATP-binding cassettes (ABC) or Major Facilitator Superfamily (MFS) type drug efflux pumps had been observed among triazole-resistant clinical isolates, which prevents the accumulation of active concentration of drug in the cell. In particular, the transcription factor AtrR positively regulates the expression of the ABC transporter CDR1B (d1 and d2). Notably, AtrR is also involved in the positive regulation of CYP51A. A clinically relevant mutation of a different kind is the one affecting the Hmg-CoA reductase encoded by hmg1, which takes part in ergosterol biosynthesis by converting Hmg-CoA into Mevalonate. Hmg1 has a conserved Sterol Sensing Domain (SSD) involved in regulation of sterol biosynthesis. Mutations in the SSD result in a dysregulation of the sterol pathway that eventually translates to an increased cellular ergosterol production and triazole resistance (f1 and f2).

Resistance mechanisms that do not involve CYP51A mutations

Upregulation of CYP51A

Upregulation of CYP51A expression is an important mechanism of azole resistance. It is partly mediated by a steroid regulatory element-binding protein (SREBP), SrbA. Apart from CYP51A, SrbA also controls many other genes involved in sterol biosynthesis, adaptation to hypoxic conditions, virulence, normal cell polarity and hyphal morphogenesis, iron uptake, nitrate assimilation, and heme biosynthesis (Willger et al. 2008, Chung et al. 2014, Dhingra & Cramer 2017). Following complex processing in the endoplasmic reticulum and Golgi apparatus, the N-terminal DNA-binding domain of SrbA, a basic helix-loop-helix leucine zipper transcription factor, is liberated and translocated to the nucleus, where it binds to steroid regulatory elements (SRE) and activates the transcription of target genes (Fig. 3) (Willger et al. 2008, Chung et al. 2014, Dhingra & Cramer 2017). SRE elements typically contain two SrbA recognition sites, SRE1 and SRE2 (Willger et al. 2008, Gsaller et al. 2016). Hence, TRs identified in ARAF isolates (TR₃₄, TR₄₈, TR₅₃, and TR₁₂₀) act as additional SrbA-binding motifs, leading to the recruitment of additional SrbA molecules, and increased expression of srbA and CYP51A, and, to a lesser extent, CYP51B (Willger et al. 2008, Gsaller et al. 2016). Indeed, the activity of SrbA is similar to that of Upc2 in C. albicans, which upregulates the expression of ERG11 (Willger et al. 2008, Gsaller et al. 2016). It should be noted that CYP51A is not exclusively regulated by SrbA, since CYP51A expression is not completely inhibited in mutants lacking srbA (Blosser & Cramer 2012). Of note, a 1822-bp insertion (type II transposon Aft1) was identified upstream of the start codon of CYP51A in an ARAF isolate overexpressing CYP51A; however, its role in azole resistance remains to be determined (Albarrag et al. 2011). Finally, although upregulation of CYP51B in ARAF isolates appears to be rare, a baseline and/or induced overexpression of CYP51B has been observed in a limited number of clinical isolates lacking mutations in CYP51A (Buied et al. 2013).

CCAAT-binding complex-mediated azole resistance

Whole-genome sequencing of an ARAF strain with wild-type (WT) CYP51A isolated from a Dutch patient identified HapE, a new factor involved in the regulation of sterol synthesis (Camps et al. 2012a). HapE is a subunit of the CCAAT-binding complex (CBC); it harboured the amino acid substitution P88L in the clinical isolate (Camps et al. 2012a).

CBC is a trimeric transcription factor complex (HapB, HapC, and HapE), which together with the monomeric transcription factor HapX regulates sterol synthesis by binding at positions –293 to –289 and –275 to –269 upstream of the CYP51A start codon, respectively (Fig. 3) (Gsaller et al. 2016). Knock-out analysis of CBC subunit genes and hapX, and heterologous expression of HapE^P88L result in increased triazole MIC values and overexpression of CYP51A, HMG-CoA synthase (paralog of erg13A and erg13B), and HMG-CoA reductase (paralog of hmg1), indicating that these transcription factors act as repressors of genes involved in sterol biosynthesis (Gsaller et al. 2016). Further studies revealed that the N-terminal DNA-binding domains of CBC and HapX physically interact with one another (Hortschansky et al. 2015). Further, experiments under iron-limiting conditions demonstrated that the initial binding of CBC to a CBC motif allows the recruitment of HapX to HapX motif (5′-GAT-3′) located 11–12 bp downstream of the CBC motif (Hortschansky et al. 2015). Interestingly, although both SrbA and CBC competitively bind to the same SRE site, position –293, the binding affinity of SrbA is 8-fold higher than that of CBC (Gsaller et al. 2016). Further, SrbA shows a higher binding affinity for the original motif located in position –293 than to the additional motifs in isolates with TRs, located in position –327 (Gsaller et al. 2016). These studies implicated other determinants involved in azole resistance, which negatively regulate the expression of azole drug target and other genes involved in ergosterol biosynthesis.

The role of efflux pumps

Identification of a relatively large number of clinical ARAF isolates lacking CYP51A mutations and comparative genomics studies in yeasts prompted the investigation of alternative mechanisms of azole resistance. This led to the discovery of the role of efflux pumps in azole resistance. Efflux pumps are categorised into two main classes, the major-facilitator superfamily (MFS), encoded by 278 genes, and ATP-binding cassette (ABC) proteins that require ATP for activity, encoded by 49 genes (Loiko & Wagener 2017).

An initial comparative genomics analysis revealed that the two paralogs CDR1A and CDR1B (also known as abcC) are orthologous to C. albicans CDR1, and abcA the same as Afumdr1 (Fraczek et al. 2013). The efflux pump genes atrI and mdrA partially contribute to azole resistance in clinical isolates, while atrF appeared to be upregulated in environmental ARAF isolates (Meneau et al. 2016). Studies focused on generating AFAR strains in vitro (UV-irradiated and itraconazole-treated strains) and those exploring the response of A. fumigatus to itraconazole and voriconazole revealed a pronounced overexpression of several efflux genes, namely, Afumdr1, Afumdr3, Afumdr4, and atrF (Nascimento et al. 2003, da Silva Ferreira et al. 2004, 2006b, Hokken et al. 2019). Of note, the triazole MIC values of isolates lacking Afumdr1 and mfs56 showed a slight change, which raises the question whether these proteins are the main efflux pumps involved in triazole resistance (Fraczek et al. 2013). Several studies dissecting the molecular mechanisms underpinning triazole resistance in clinical ARAF isolates, however, showed that atrF, mfsC, and CDR1B (orthologous to CaCDR1), especially CDR1B and atrF, are greatly upregulated in these isolates (Slaven et al. 2002, Fraczek et al. 2013, Meneau et al. 2016, Paul et al. 2017, Sharma et al. 2019, Wu et al. 2020). This observation is further supported by the observation that heterologous expression of CDR1B in S. cerevisiae isolate lacking Pdr5, or deleting this gene from an ARAF isolate carrying TR₃₄/L98H, TR₃₄, or L98H, results in a profound decrease of voriconazole MIC values (Paul & Moye-Rowley 2013, Paul et al. 2017). Furthermore, these efflux pumps have different substrate specificity, i.e., a narrow or broad-spectrum and functionality, with CDR1B showing the broadest substrate specificity (Esquivel et al. 2020).

Attempts to identify the regulator of CDR1B expression resulted in the identification of AtrR, (ABC transporter–regulating transcription factor). AtrR is a Gal-4 type Zn₂-Cys₆ cluster-containing transcription factor, which shares homology with CgPdr1 (Fig. 3) (Hagiwara et al. 2017). Interestingly, AtrR is responsible for the upregulation of CYP51A and also CYP51B, as predicted by experiments that suggested that proteins other than SrbA also control CYP51A expression (Blosser & Cramer 2012, Paul et al. 2019). Intriguingly, the AtrR response element (ATRE) is located within the 34-bp repeat element and, hence, both AtrR and SrbA share overlapping binding sites (Fig. 3) (Paul et al. 2019). Therefore, it is plausible that both SrbA and AtrR cooperatively mediate the upregulation of CYP51A expression, as well as the expression of several genes that control the ergosterol biosynthesis pathway (Hagiwara et al. 2017, Paul et al. 2019). Importantly, deletion of atrR resulted in azole hypersensitivity of a strain with CYP51A with G54E substitution, which indicates that AtrR contributes to azole resistance even in isolates with mutated CYP51A (Hagiwara et al. 2017). Strains lacking both atrR and srbA are viable (Hagiwara et al. 2017), which may indicate the presence of other transcription factors that control the expression of other efflux pumps.

Although the regulator of atrF expression remains to be identified in A. fumigatus, a recent study using whole genome sequencing of a laboratory-derived voriconazole-resistant A. flavus isolate revealed that yap1 controls the expression of atrF (Ukai et al. 2018). Yap1 is a bZIP master-regulator transcription factor involved in the oxidative stress response (Ukai et al. 2018). Interestingly, the voriconazole-resistant A. flavus isolates harboured the mutation Yap1^L558T, which resulted in atrF overexpression, and reverting Yap1^L558T to WT form and deletion of atrF significantly decreased the voriconazole MIC values (8- to 16-fold) (Ukai et al. 2018). The gain-of-function mutation resulted in changes in the C-terminus of Yap1, leading to a constitutive localisation of this transcription factor in the nucleus, binding to a putative Yap1 response element (YRE) at positions –462 to –456 relative to the start codon of the target genes, and overexpression of the target genes (Ukai et al. 2018). Of note, YRE is also present upstream of CYP51A, but this gene was not greatly upregulated in the respective azole-resistant isolate, which suggests that Yap1 does not control the CYP51A. This finding encourages evaluation of the yap1 sequence in ARAF isolates overexpressing atrF.

Hmg1 and azole resistance

Losada et al. (2015) identified Hmg1 as another player in azole resistance, in an experiment involving successive in vitro exposure of A. fumigatus to various azole compounds. Hmg1 is bound to the endoplasmic reticulum membrane via an N-terminal anchor domain linked to the catalytic site via a linker (Fig. 3) (Sever et al. 2003, Friesen & Rodwell 2004). In the experiment of Losada et al. (2015), all voriconazole-resistant progenies had specific amino acid substitutions in the sterol-sensing domain of Hmg1. Sterol negatively regulates the activity of Hmg1, and in the presence of high sterol levels, the membrane-bound domain of Hmg1 is targeted to proteasome-mediated proteolysis (Sever et al. 2003). Therefore, it is plausible to associate mutations in the sterol-sensing domain with an increased enzyme stability, which would lead to sterol overproduction and, potentially, azole resistance (Sever et al. 2003, Friesen & Rodwell 2004, Losada et al. 2015, Jiang et al. 2018). Indeed, this was the case with ARAF clinical isolates from Japan and the US (Hagiwara et al. 2018, Rybak et al. 2019). These isolates, lacking CYP51A mutations but harbouring mutations in hmg1, produced more ergosterol and were more susceptible to polyenes, such as AMB, than isolates without mutations in hmg1. Other studies conducted in India and Taiwan identified similar mutations in hmg1 in clinical ARAF isolates (Sharma et al. 2019, Wu et al. 2020). Intriguingly, a study conducted in the US revealed that a high proportion of isolates with WT CYP51A (11/21; 52 %) harbour mutations in the hmg1 portion encoding the sterol-sensing domain (Siopi et al. 2017). Although initial ectopic expression experiments failed to associate the discovered hmg1 mutations with azole resistance, a recent study using CRISPR-Cas9 methodology revealed that F262_del, S305P, and I412S dramatically increase the triazole MIC values (Rybak et al. 2019). Similarly, other genes involved in ergosterol biosynthesis, including ERG6, are mutated and may potentially contribute to triazole resistance, although these mutations are not as prevalent as those of hmg1 (Hagiwara et al. 2018).

Master regulators of azole resistance

It is still unclear how genes involving ergosterol biosynthesis and SrbA, AtrR, CDR1B, and Hap complex genes are regulated on a larger scale, and which master regulators control their expression. In a recent study, potential master regulators that could simultaneously be involved in azole resistance and pathogenicity were systematically analysed (Furukawa et al. 2020). The authors showed that negative cofactor two (Nct2), consisting of the NctA and NctB subunits, regulates ergosterol biosynthesis and iron-responsive pathways by co-localising and interacting with the TATA-box located upstream of the target genes (an estimated nearly 30 % of coding genes in A. fumigatus). Interestingly, nctA and nctB mutants are not only pan-azole and AMB resistant, but they present no fitness cost as their pathogenicity is comparable with that of the WT (Furukawa et al. 2020). The controversial AMB resistance despite a modest increase of ergosterol content could be explained by an upregulation of oxidative stress-reducing enzymes and the notion that altered cell wall morphology may act as a barrier to AMB penetration in these mutants (Furukawa et al. 2020). These findings warrant future studies to assess the role of loss-of-function mutations in Nct2 complex genes in clinical ARAF isolates.

Additional mechanisms of azole resistance

In addition to the already mentioned major mechanisms of azole resistance, some other, relatively rare, mechanisms are also implicated in azole resistance. Damage resistance protein 1 (Dap1) is a cytochrome b₅-like heme-binding protein that regulates the function of CYP51A and ERG5. It is located at the endoplasmic reticulum membrane and is composed of three subunits, DapA, DapB, and DapC. DapA stabilises CYP51A and ERG5, allowing electron transfer, while DapB and DapC suppress electron transfer and prevent the activity of target proteins through depletion of heme (Song et al. 2016). Of note, although DAPA and DAPC co-localise at the endoplasmic reticulum membrane, and form complexes with CYP51A and ERG5, DAPB is located in the nucleus (Song et al. 2016, 2017). Gene deletion analysis revealed that ΔdapA was susceptible and ΔdapC was more resistant against itraconazole, while the itraconazole susceptibility of ΔdapB was indistinguishable from the parental strain (Song et al. 2016). The observation that even upon azole stress Dap1 family proteins remain at the endoplasmic reticulum membrane indicate that other transcription factors that translocate to the nucleus may regulate the expression of these genes (Song et al. 2017). In keeping with this anticipation, it was revealed that SrbA is required for the overexpression of dapA and dapC, and Dap1 family proteins per se do not sense ergosterol depletion, indicating that the expression of Dap1 protein genes is controlled by SRE (Song et al. 2017).

According to a recent study, mutation in a gene encoding farnesyl transferase (Afcox10^R243Q) and loss of algA, a component of the calcium signalling pathway, leads to itraconazole resistance (Wei et al. 2017). Interestingly, a collection of ARAF clinical isolates with WT CYP51A harbour several amino acid substitutions in Afcox10 but their effect on azole resistance was not evaluated (Sharma et al. 2019). A mismatch repair gene (MMR, also known as MSH2) plays an important role in facilitating the acquisition of drug resistance in C. glabrata (Healey et al. 2016). Unlike C. glabrata, however, the clinical and environmental Aspergillus isolates do not harbour many nonsense mutations in MSH2. Nonetheless, deleting msh2 profoundly impacts genetic stability, antifungal resistance, and virulence in A. fumigatus (Dos Reis et al. 2019). Finally, some studies have implicated OrmA, the rate-limiting enzyme of the sphingolipid biosynthesis pathway (Zhai et al. 2019), b5 CybE (Misslinger et al. 2017), mitochondrial dynamics (Neubauer et al. 2015), and oxidoreductase HorA (Kroll et al. 2016), in azole resistance.

It should be noted that triazole resistance is a multifactorial phenomenon, involving alteration of the drug target, and upregulation of the drug target and efflux pumps (Nascimento et al. 2003, Fraczek et al. 2013). In Candida species belonging to the CTG clade, such as C. albicans, gain-of-function mutations in UPC2, TAC1, and MRR1 result in upregulation of ERG11 and efflux pump genes (Gsaller et al. 2016). Since genes with similar function exist in A. fumigatus, e.g., srbA and atrR, it would be interesting to explore the presence of such gain-of-function mutations in ARAF isolates, and their association with CYP51A and CDR1B overexpression. Further, gain-of-function mutations in CgPDR1 lead to increased virulence and immune evasion and, hence, their implications for A. fumigatus represent an interesting area for future investigations (Vale-Silva et al. 2013). The most important factor that would facilitate gene expression analysis in A. fumigatus is gene characterisation, since most genes remain uncharacterised, which makes the interpretation of transcriptomic studies challenging (Hokken et al. 2019). Nonetheless, the current understanding of azole resistance in A. fumigatus offers a wide range of potential targets that can inspire the development of novel and potent antifungal drugs.

Azole resistance and in-host fitness cost of A. fumigatus

Aspergillus fumigatus exhibits tremendous phenotypic, physiologic, and genomic plasticity, which allows it to adapt to and survive azole exposure. However, bona fide azole resistance requires the acquisition of permanent mutations in azole resistance-conferring genes and/or rewiring of the transcriptomic landscape to enable fungal persistence and survival in the presence of antifungal drugs within the host. Occurrence of such changes, in turn, results in therapeutic failure and dramatically increases the mortality rates, to over 80 % and even up to 100 % in real-life clinical settings (van der Linden et al. 2011, van Paassen et al. 2016). However, since resistance-conferring mutations negatively affect the catalytic activity of key enzymes, such as CYP51A, the mutated isolates may show fitness defects in the absence of azoles compared with the WT population. Indeed, the growth rate and conidia production by isolates harbouring various mutations and with large chromosomal deletions are markedly lower than those of isogenic susceptible isolates (Hagiwara et al. 2014). Further, as shown in in vivo studies, ARAF isolates harbouring HapE with P88L are less virulent than azole-susceptible isogenic ancestors and WT isolates, and exhibit a 4-h growth delay relative to susceptible and WT isolates (Arendrup et al. 2010). In keeping with these observations, ARAF isolates are not detected following discontinuation of the azole therapy, while resistant isolates reappear following itraconazole treatment in the clinical setting (Chen et al. 2005). By contrast, ARAF isolates carrying mutations in CYP51A, especially those with TRs, may harbour additional mutations in the genome, acting as a compensatory mechanism, which allow them to thrive within the host and/or the environment (Verweij et al. 2016b). Indeed, no significant differences in the sterol (and ergosterol) content of several of azole susceptible and ARAF isolates were detected in one study (Alcazar-Fuoli et al. 2008). These observations highlight the notion that while mutations might affect the docking of an antifungal at the enzyme active site, they might not affect the binding of the sterol substrate, as also suggested in simulation studies (Nash & Rhodes 2018). Further, the presence of two copies of the CYP51 gene might enable rapid ergosterol biosynthesis and increased azole resistance (Hu et al. 2007). The presence of multiple paralogs impacts the fitness cost, which is unexpected, based on what is known about yeasts species, such as Candida. Consistent with this notion, in one study, mutant ARAF isolates were persistently isolated from clinical samples of a patient following discontinuation of azoles (Tashiro et al. 2012), and isolates carrying TR₄₆/Y121F/T289A exhibit the same growth rate and conidia production on PDA medium as WT isolates (Hagiwara et al. 2016).

There are multiple possible explanations for such contradictory observations. First, since the fitness cost is an outcome of accumulation of multiple mutations in the genome (Verweij et al. 2016b), assessment of the effect of various mutations requires a study of isogenic isolates, which only differ with respect to the presence of mutations of interest in a locus of interest. Accordingly, WT status should not be assigned solely on the sequencing of CYP51A or few genes. Indeed, to obtain reliable data, studies focusing on fitness-cost evaluation should utilise a well-defined WT isolate whose entire genome has been sequenced. Second, the fitness cost may vary depending on the gene and mutation studied, e.g., the virulence attenuation observed in HapE^P88L strains (Arendrup et al. 2010) vs. lack of pronounced virulence attenuation in isolates lacking nct2 (Furukawa et al. 2020). Third, pronounced growth defects in vitro do not always mirror isolate behaviour in vivo, since, as shown in some studies, the in vivo virulence of mutant strains with a growth defect in vitro does not significantly differ from that of WT isolates (Furukawa et al. 2020), which reflects the complex nature of growth in the host. Further complicating matters is simultaneous recovery of both, azole-susceptible and -resistant isolates from clinical samples, which undermines the notion of predominance of a single genotype in a specific ecological niche. Indeed, a remarkable 20 % rate of isolation of mixed susceptible-resistant colonies from clinical samples was reported in one study (Fuhren et al. 2015), which reinforced the analysis of multiple isolates per clinical sample to verify the concomitant presence of isolates with different susceptibility profiles (Camps et al. 2012c).

Resistance to other classes of antifungals

As explained above, the continuous increase in the prevalence of triazole-resistant A. fumigatus in the clinic promote physicians to use other antifungal drugs, most notably AMB and echinocandins. Although the degree of resistance may vary depending on the fungicidal and fungistatic nature of these antifungals, as is the case with azoles, it is rational to assume that the selective pressure exerted by these antifungals will allow the selection of drug-resistant A. fumigatus isolates. Overall, the AMB resistance is a rare phenomenon among patients with IPA and although not generalising the case, some studies have shown that AMB MIC did not differ among IPA patients with/ without AMB exposure (Moosa et al. 2002). The AMB resistance rarity also may explain the lack of knowledge on underlying resistance mechanisms involved in A. fumigatus, which appears to be associated with less drug uptake and a higher catalase activity in A. terreus (Blum et al. 2013). Although counterintuitive, the ergosterol level does not seem to differ among AMB-resistant and AMB-susceptible A. terreus isolates (Blum et al. 2013). To close this knowledge gap, developing in vitro AMB-resistant A. fumigatus isolates followed by unravelling the subcellular mechanism may provide some insights on this context.

Similar to AMB, echinocandin resistance is rarely reported among clinical isolates of A. fumigatus (Arastehfar et al. 2020b) and few studies conducted in this regard have shown that acquisition of mutations in a hotspot of the FKS gene, which encodes the catalytic subunit of β-1,3-D-glucan synthase (Jiménez-Ortigosa et al. 2017), along with the changes in the lipid profile surrounding β-1,3-D-glucan synthase (Satish et al. 2019) may serve as the possible cellular factors underlying echinocandin resistance.

Antifungal drug resistance detection in A. fumigatus: from phenotypic assays to MALDI-TOF MS

As discussed, mortality rates are notably high in patients infected with A. fumigatus azole resistant strains (van der Linden et al. 2011, Lestrade et al. 2019), evincing the importance of making an early resistance detection in order to start an appropriate therapy. Antifungal susceptibility testing has the ultimate goal of helping clinicians to anticipate the chance of treatment success or failure. This generally follows the “90/60” rule in which approximately 90 % of the infections caused by susceptible isolates and 60 % of those due to resistant isolates respond to therapy (Rex & Pfaller 2002). Antifungal resistance can be in vitro detected by performing broth microdilution assays such as those developed and standardised by the Clinical and Laboratory Standard Institute (CLSI) (CLSI 2008) and the European Committee on Antimicrobial Susceptibility Testing (EUCAST) (Arendrup et al. 2017a), considered as the reference for yeasts and moulds. These methods determined the MIC, defined as the lowest drug concentration required to inhibit fungal growth. However, some antifungals are only able to partially inhibit certain species, causing changes in their hyphal growth. Such is the case of echinocandins on moulds, for which the two methodologies propose the determination of the minimum effective concentration (MEC), the lowest concentration of antifungal resulting in hyphal morphological changes (Kurtz et al. 1994). MEC assessment has the difficulty of being subjective and requiring expertise in microscopic observation in order to determine what is considered as aberrant growth. Even though there are some methodological differences between CLSI and EUCAST procedures, their results have been proved to be comparable (Espinel-Ingroff et al. 2013) and allow the categorisation of the strains as susceptible or resistant by applying the established clinical breakpoints (CBPs). Although EUCAST has defined drug-related and species-related CBPs for A. flavus, A. fumigatus, A. nidulans, A. niger and A. terreus (EUCAST 2020), CLSI has recently adopted CBPs for voriconazole only for A. fumigatus species and uses epidemiological cut-offs (ECVs) to discriminate between wild-type susceptible strains and others with acquired resistance to other antifungal drugs (https://clsi.org/media/3682/m61ed2_sample.pdf, Espinel-Ingroff & Turnidge 2016).

In addition, several complementary methods have been commercialised and are easily carried out in the daily routine of a clinical microbiology laboratory, which include colorimetric endpoint methodologies such as Sensititre YeastOne, Micronaut-AM and the XTT assay, or agar-based methods using strips with a gradient of antifungal concentrations (Etest or MIC Test Strips) or four-well plates for azole (VIPCheck) and/or echinocandin susceptibility testing. The performance of these methods has been compared with the reference standard with generally positive results. Sensititre YeastOne yielded high essential agreement rates with CLSI for itraconazole, voriconazole, posaconazole and amphotericin B (Castro et al. 2004, Guinea et al. 2006, Mello et al. 2017), although its concordance for echinocandins was lower so its use is not recommended in this case (Siopi et al. 2017). Micronaut-AM showed good categorical agreements (≥96 %) with CLSI for anidulafungin, amphotericin B, voriconazole and itraconazole, being able to detect azole resistance in A. fumigatus (Nuh et al. 2020). As echinocandin susceptibility testing for Aspergillus spp. using reference methods is not easy, the XTT-based assay proved to be a feasible alternative and showed promising results when compared with EUCAST MEC values (Meletiadis et al. 2020). Etest and MIC Test strips have been confirmed to be reliable alternatives for antifungal susceptibility testing for Aspergillus species, as well as for detecting A. fumigatus azole resistance, after obtaining categorical and essential agreements of ≥90 % when correlating their results with those yielded by CLSI (Guinea et al. 2008, Martos et al. 2010, Lamoth & Alexander 2015) and EUCAST (Arendrup et al. 2017b, Idelevich et al. 2018). Azole and echinocandin resistance in A. fumigatus is successfully detected using four-well agar plates, showing comparable results to those obtained with the reference methodologies (Arendrup et al. 2017c, Buil et al. 2017b, Tsitsopoulou et al. 2018, Meletiadis et al. 2019). EUCAST has proposed several recommendations for their use as a screening procedure for the detection of azole resistance (Guinea et al. 2019).

Nevertheless, these methodologies have important limitations, especially for moulds: while EUCAST and CLSI are time-consuming and laborious, colorimetric and agar-based methods are easier to perform. Besides, all of them require a fungal pure culture. This can be a critical issue due to the usual low recovery rate of Aspergillus species in culture from clinical samples, which also leads to the underestimation of azole resistance rates (van der Linden et al. 2016). Although the potential of culturing high volume sputum samples yielded positive results for patients with chronic and pulmonary aspergillosis (Vergidis et al. 2020), the development of alternative non-culture-based techniques is essential.

In this context, molecular tools detecting resistance mutations directly from clinical samples are proving to be complementary to phenotypic assays by reducing turnaround times for the initiation of an effective therapy (Jenks et al. 2019f). Several in-house and commercial PCR-based methodologies to directly detect mutations in the A. fumigatus azole-resistance related gene CYP51A and its promoter, including the most frequent point mutations (G54, G138, M220, G448, L98H, Y121F, T289A) and tandem repeat insertions (TR₃₄ and TR₄₆), have been developed (Table 4). While some consist in conventional PCR assays (Spiess et al. 2012, 2014, Postina et al. 2018), the vast majority are real-time PCR based methods that avoid the delay associated with sequencing, something also successfully achieved with loop-mediated isothermal amplification (LAMP) assays (van der Linden et al. 2010, Denning et al. 2011, Zhao et al. 2013, 2016, Chong et al. 2015, 2016, White et al. 2015a, 2017, Dannaoui et al. 2017, Montesinos et al. 2017, Schauwvlieghe et al. 2017, de Groot et al. 2018, Denis et al. 2018, Guegan et al. 2018, Morio et al. 2018, Mikulska et al. 2019, Yu et al. 2019, Pelzer et al. 2020). Pyrosequencing also shows promising results, as it has the advantage of being adaptable to other genes of interest (Trama et al. 2005, van der Torre et al. 2020). However, one of their main limitations is that they need to be very sensitive and specific to detect the low concentration of Aspergillus DNA and prevent cross-reactivity with human DNA. Besides, CYP51A is a single-copy gene, which impairs its amplification, although this can be improved by the use of a nested PCR assay in order to achieve a higher sensitivity (Denning et al. 2011, Zhao et al. 2013, 2016). Although other in-house and commercial molecular methods for detecting CYP51A mutations in A. fumigatus have been successfully developed, they have not been evaluated in clinical samples yet (García-Effron et al. 2008, Klaassen et al. 2010, Bernal-Martínez et al. 2017, Wang et al. 2019, Fungiplex® 2020). The emergence of CYP51A-WT strains increasingly being identified in clinical settings in addition to a diverse range of mutations occurring in the CYP51A challenge the applicability of such tools. Despite the limited nature of resistance-associated mutations, there may be a role for whole-genome sequencing as an alternative strategy (see section 13).

Table 4.

Studies using in-house and commercial methodologies for the direct detection of azole resistance in A. fumigatus from clinical samples.

Method	Type of assay	CYP51A mutations detected	Clinical samples (n)	CYP51A amplification (false positives)	Additional results	References
Conventional PCR + sequencing	In-house	TR34, L98H, M220	BAL fluids (6); tissue (2)	100 %	75 % WT; 12.5 % TR34/L98H, 12.5 % L98H	Spiess et al. 2012
	In-house	TR34, L98H, M220, TR46	Blood (25); BAL fluids (120); CSF (19); tissue (17)	100 % (39.8 %)	58.6 % WT; 1.1 % TR34/L98H; 0.5 % L98H	Spiess et al. 2014
	In-house	TR34, L98H, M220, TR46, Y121F, T289A	BAL fluids (22)	74.2 % (65.1 %)	7.6 % not sequenced; 1.5 % TR34/L98H	Postina et al. 2018
			Tissue (15)	68.9 % (54.5 %)	7.8 % not sequenced; 1.1 % L98H; 3.3 % TR46/Y121F/T289A; 2.2 % TR34/L98H
			CSF (15)	39 % (28.7 %)	10.3 % not sequenced
Pyrosequencing	In-house	G54	Blood (56)	3.6 %	3.6 % WT	Trama et al. 2005
	In-house	All of those described	Respiratory samples			van der Torre et al. 2020
Real-time PCR	In-house	TR34, L98H	Sputum (1); tissue (4)	60 %	60 % TR34/L98H	van der Linden et al. 2010
	AsperGenius®	TR34, L98H, Y121F, T289A	BAL fluids (77)	18.2 %	15.6 % WT; 1.3 % TR34/L98H; 1.3 % TR46/Y121F/T289A	Chong et al. 2015
	AsperGenius®	TR34, L98H, Y121F, T289A	Serum (72)	16.7 %	6.9 % TR46/Y121F/T289A; 5.6 % L98H; 2.8 % TR34/L98H; 1.4 % Y121F	White et al. 2015a
	AsperGenius®	TR34, L98H, Y121F, T289A	BAL fluids (201)	33.8 %	28.3 % WT; 3.5 % TR34/L98H; 1.5 % TR34/L98H+WT; 0.5 % TR46/Y121F/T289A	Chong et al. 2016
	MycoGENIE®	TR34, L98H	Respiratory samples (88); serum (69)		0 % TR34/L98H	Dannaoui et al. 2017
	AsperGenius®	TR34, L98H, Y121F, T289A	BAL fluids (100)	20 %	17 % WT; 3 % unspecified mutations	Montesinos et al. 2017
	AsperGenius®	TR34, L98H, Y121F, T289A	BAL fluids (91)	49.5 %	34.1 % WT; 8.8 % TR34/L98H; 3.3 % TR34/L98H +WT; 3.3 % TR46/Y121F/T289A	Schauwvlieghe et al. 2017
	AsperGenius®	TR34, L98H, Y121F, T289A	Plasma (86)		100 % WT	White et al. 2017
	AsperGenius®	TR34, L98H, Y121F, T289A	Serum (9); tissue (8)	76.5 %	76.5 % WT	de Groot et al. 2018
	MycoGENIE®	TR34, L98H	BAL fluids (31)		0 % TR34/L98H	Denis et al. 2018
	AsperGenius®	TR34, L98H, Y121F, T289A	Sputum (119)	47.9 %	47.9 % WT	Guegan et al. 2018
	MycoGENIE®	TR34, L98H			0 % TR34/L98H	Guegan et al. 2018
	MycoGENIE®	TR34, L98H	Respiratory samples (147)		0 % TR34/L98H	Burckhardt & Zimmermann 2018
	AsperGenius®	TR34, L98H, Y121F, T289A	BAL fluids (22)	59.1 %	56.8 % WT; 2.3 % TR34/L98H	Postina et al. 2018
			Tissue (15)	46.7 %	43.4 % WT; 3.3 % TR46/Y121F/T289A
			CSF (15)	41.7 %	41.7 % WT
	MycoGENIE®	TR34, L98H	BAL fluids (123)		0.8 % TR34/L98H	Mikulska et al. 2019
	AsperGenius®	TR34, L98H, Y121F, T289A	BAL fluids (23)	65.2 %	52.2 % WT; 8.7 % TR34/L98H; 4.3 % TR46/Y121F/T289A	Pelzer et al. 2020
Nested PCR + real-time PCR	In-house	TR34, L98H, G54, G138, M220	Sputum (29)	100 %	48.3 % TR34/L98H; 31 % L98H; 6.9 % TR34; 6.9 % TR34/L98H+M220; 6.9 % M220	Denning et al. 2011
	In-house	TR34, L98H, G54, G138, M220, G448	BAL fluids (94)	64.9 %		Zhao et al. 2013
	In-house	TR34, L98H, G54, G138, M220, G448	Respiratory samples (97)	39.2 %		Zhao et al. 2016
LAMP	In-house	TR34	Clinical samples (11)		100 % TR34/L98H	Yu et al. 2019

LAMP, loop-mediated isothermal amplification; BAL, bronchoalveolar lavage; CSF, cerebrospinal fluid; WT, wild-type.

Taking all of these into account, new molecular assays should intend to cover a broader range of azole resistance-related mutations and mechanisms, as not all of the CYP51A reported alterations are detected by the available methodologies, while increasing their sensitivity in order to become a feasible option to detect azole resistance in more clinical settings. Besides, there are no molecular options for the direct detection of azole resistance in Aspergillus species other than A. fumigatus, which should be further studied and developed.

Since MALDI-TOF MS has been introduced as a routine identification tool, laboratories are interested to explore this rapid technology as an alternative methodology potentially accelerating antifungal susceptibility testing (Burckhardt & Zimmermann 2018). A simple approach using specific marker peaks in mass spectra of resistant microorganisms could be applied to a number of specific resistances in certain bacteria, but this approach was yet not successful for fungi. Several functional assays have been proposed for antibiotic susceptibility testing in bacteria, e.g., a test for the hydrolytic degradation of β-lactam antibiotics, the observation of incorporation of stable isotope-labelled metabolites, or the utilisation of MALDI-TOF MS as semi-quantitative read out for microbial growth in the presence of an antibiotic drug. Less work has been performed until today to apply such MALDI-TOF MS based assays to antifungal resistance. Most of the studies in this regard have focused on the detection of resistance in yeasts by MALDI-TOF MS. Detection of peak pattern changes after incubation of cells in presence of an antifungal has been demonstrated to determine antifungal resistance in Candida species (Delavy et al. 2019). In two studies, this method was also applied to test susceptibility in Aspergillus species. One study described the detection of caspofungin resistance in A. fumigatus and A. flavus strains (De Carolis et al. 2012). The same approach was applied to strains of Aspergillus species and voriconazole (Gitman et al. 2017). Although the results were in good agreement with reference methods, there was no obvious advantage over traditional methods, in particular because of the similar time to result. A promising method which has been shown to detect antifungal resistance in Candida by semi-quantitative detection of fungal growth in the presence of the drug after only several hours of incubation (Vatanshenassan et al. 2018, 2019) has not yet been applied to Aspergillus yet, according studies should be performed.

Application of typing techniques to identify infection and resistance routes

Strain genotyping is considered one of the most basic tools in the clinical setting since it fulfils many needs, among which the establishment of epidemiological relationships between isolates stands out. As in many other settings, typing methodologies have had an important impact in the aspergillosis field, since they have been used, among many others, for outbreak analysis (Menotti et al. 2005, Doll et al. 2017), environmental monitoring of the isolates that constitute a specific population (Deng et al. 2017, Fan et al. 2020), also for patient monitoring in order to study how clinical strains evolve under drug pressure within the antifungal therapy (Escribano et al. 2017) or to assess local and global Aspergillus spp. epidemiology (Garcia-Rubio et al. 2018a, 2018c, Choi et al. 2019). Thus, the molecular analysis of the genetic and epidemiological relationship between environmental and clinical strains could potentially assess strain origin and route of transmission. Besides all these applications at the subspecies level, molecular typing methods have also been used at the genus level for discriminating between species and also for the definition and recognition of new fungal species (Klaassen & Osherov 2007).

Different methodologies have been developed to genotype Aspergillus species strains (Table 5). However, due to its clinical significance, most of them have been implemented for A. fumigatus strains (Latgé & Chamilos 2019). Classically, genotyping techniques can be grouped in two different categories; methods either based on PCR amplification and sequencing, which are described in detail below, or based on non-coding repetitive sequences paired with restriction fragment length polymorphisms, such as random amplified polymorphic DNA (RAPD) (Loudon et al. 1993) amplified fragment length polymorphism analysis (AFLP) (Warris et al. 2003), and restriction fragment length polymorphism analysis (RFLP) (Neuveglise et al. 1996). The latter ones show a poor inter-laboratory reproducibility which is the reason why they have been replaced by other techniques. Thus, the selection of the most appropriate method in each context will highly depend on the technical resources of a particular setting (Klaassen & Osherov 2007).

Table 5.

Advantages and disadvantages of technologies used for typing clinical and environmental A. fumigatus isolates.

Assay	Methodology	D∗	R∗∗	References
RAPD	Random amplified polymorphic DNA	Low	Low	Bertout et al. 2001
SSDP	PCR typing method combining RAPD and specific DNA primers	Low	Low	Mondon et al. 1997
MLEE	Multilocus enzyme electrophoresis	Low	Low	Rodriguez et al. 1996, Bertout et al. 2000
AFLP	Amplified fragment length polymorphism analysis	Moderate	Low	Warris et al. 2003
RFLP	Restriction fragment length polymorphism analysis	Low	Low	Neuveglise et al. 1996
MLP	Microsatellite length polymorphism	High	Moderate	Bart-Delabesse et al. 1998, Bertout et al. 2001
RISC	Retrotransposon insertion-site PCR amplification	Moderate	Moderate	de Ruiter et al. 2007
MLST	Multilocus sequence typing	Moderate	High	Bain et al. 2007
STRAf	Short coding tandem repeats	High	Moderate	de Valk et al. 2005, Guinea et al. 2011, Escribano et al. 2015, Fan et al. 2020
CSP	Cell-surface protein sequencing	Moderate	High	Balajee et al. 2007b, Levdansky et al. 2007
TRESPERG	Sequencing of tandem repeats surface protein coding genes	High	High	Garcia-Rubio et al. 2018a, Fan et al. 2020
WGS	Whole Genome Sequencing	High	High	Hagiwara et al. 2014, Garcia-Rubio et al. 2018b, Puértolas-Balint et al. 2019

D∗, discriminatory power; R∗∗ reproducibility considering stability and availability.

Although not a bona fide gold standard technique, the microsatellite analysis assay called STRAf is the most popular and widely used technique used to type A. fumigatus, which stands for short tandem repeats of A. fumigatus (de Valk et al. 2005). This assay, developed more than a decade ago, is based on a panel of microsatellites divided into three multicolour multiplex PCRs. Each multiplex reaction amplifies three di-, tri-, or tetra-nucleotide repeats, respectively. One of the biggest advantages of this technique is the multicolour multiplex approach which allows large numbers of markers to be tested in a short period of time, which is why this assay is a very suitable tool for large-scale epidemiological studies. Moreover, it can even be used to genotype A. fumigatus isolates directly from clinical samples, such as formalin-fixed paraffin-embedded tissues or serum samples (de Groot et al. 2018). The Simpson’s diversity index of this assay (Hunter & Gaston 1988), which shows the discriminatory power of the methodology, is really high –0.9994 (de Valk et al. 2005), 0.988–0.995 (Escribano et al. 2015), 0.984 (Guinea et al. 2011) and 0.9993 (Garcia-Rubio et al. 2018a, 2018c). From a methodological point of view, the STRAf assay presents some major difficulties associated with sizing of the obtained PCR products; high-resolution equipment such as capillary-based or acrylamide-based electrophoresis platforms is required to translate the fragment electrophoretic mobility to their repeat number. However, this mobility is dependent on many critical factors such as the presence, or not, of denaturing compounds, the matrix, the run temperature, the sequence of the fragment, the fluorescent labels, the sizing marker, etc. (Klaassen & Osherov 2007). In order to get exchangeable typing results between laboratories, it is necessary to run allelic ladders for calibrating every platform (de Valk et al. 2009). Also, the low-level instability of two of the markers (de Groot & Meis 2019), together with the availability of the required laboratory technology, dedicated software and personnel specifically trained for its performance comprise some of the main disadvantages of this assay (Klaassen & Osherov 2007, Garcia-Rubio et al. 2016).

As a result, the development of novel and more accessible typing methods has been encouraged. One simple and rapid single-locus typing method was developed based on sequencing the coding tandem repeats present on the cell surface protein (CSP) gene (Balajee et al. 2007b, Levdansky et al. 2007). This method has a lower discriminatory power than microsatellite-based typing – 0.78 (Klaassen et al. 2009), 0.83 (Gao et al. 2013) – and that is why a new typing method, called TRESPERG assay, was described combining four coding tandem repeat markers (Garcia-Rubio et al. 2018a, 2018c). This assay has a sufficiently high discriminatory power to compete with STRAf, and its main advantage is that it does not require trained personnel, specific equipment, or software for analysis, as it only consists of conventional PCR amplification and Sanger sequencing (Garcia-Rubio et al. 2018a, 2018c). Moreover, TRESPERG assay clustered tandem-repeat (TR) azole resistant strains in a better manner than STRAf assay did compared to whole genome sequencing (WGS) studies. Many authors have described these TR azole resistant strains as genetically more closely related than other A. fumigatus isolates (Camps et al. 2012b, Abdolrasouli et al. 2015, Garcia-Rubio et al. 2018b, Wang et al. 2018b). This fact is supported by TRESPERG results in which every A. fumigatus TR isolate tested grouped in only one cluster, endorsing their genetic closeness (Garcia-Rubio et al. 2018a, 2018c).

Finally, whole genome sequencing (WGS) has recently emerged as an invaluable tool for the analysis of genetic differences between A. fumigatus strains and has turned into the typing technique with the highest resolution, becoming increasingly affordable and widely available (Hagiwara et al. 2014, Garcia-Rubio et al. 2018b, Puértolas-Balint et al. 2019). The details of this technology is presented below.

Whole-genome sequencing applications in the clinic

The phenotypic traits of an Aspergillus strain, including its virulence potential and its ability to survive drug exposure, are ultimately encoded in its genome. Reference genome sequences for A. fumigatus, and A. nidulans are available since 2005 (Galagan et al. 2005, Nierman et al. 2005), and have undoubtedly served to sustain major advances in the field. However, as many other fungi, Aspergillus have very plastic genomes and phenotypes, with clinically relevant traits varying widely across isolates. The lowering costs and continuing developments in high-throughput sequencing have recently allowed zooming into the specific genomes of particular strains, revealing how genomic and phenotypic plasticity are connected. In addition, our ability to access full genomic sequences in a timely and cost-effective manner are opening new avenues for clinical applications such as species identification and diagnosis of resistance potential (Consortium & Gabaldón 2019). Currently, three major sequencing platforms exist that differ in their functionalities and suitability for different purposes (Table 6). Their combined use has enabled deciphering genome variation in the Aspergillus genus with increasing level of resolution. For instance, reference genomes for virtually all major species in the genus have been produced in the last few years (Kjærbølling et al. 2018, 2020, Vesth et al. 2018). More recently, re-sequencing of evolved isolates has served to estimate mutation rates in the three major species (Álvarez-Escribano et al. 2019) – A. flavus (4.2 × 10^-11 mutations per site and mitotic division), A. fumigatus (1.1. × 10^-11) and A. nidulans (4.1 × 10^-11), which can be used, for instance, to more precisely date the divergence between two strains or the origin of clinical outbreaks. Other sequencing studies have shown that relationships between genome and phenotypic plasticity can be mapped by comparing genomes and phenotypes from different isolates (Bastos et al. 2020, Drott et al. 2020). The sequencing of serially collected isolates from the same patients with aspergillosis can reveal mutations acquired during the infection process (Hagiwara et al. 2014) and whole genome sequencing can identify mutations leading to azole resistance, particularly in resistant isolates not bearing mutations in CYP51A (Ukai et al. 2018, Sharma et al. 2019). Finally, another promising application of high-throughput sequencing is the direct detection of pathogenic species through targeted barcode sequencing or whole genome shotgun sequencing of complex samples, without the need of isolation, allowed by the high sensitivity of high-throughput sequencing approaches coupled to the increasing resolution power of comparisons with sequence databases (McTaggart et al. 2019).

Table 6.

Comparison of the main features of the three most widely used high-throughput sequencing platforms. Data has been compiled from vendors information and the literature mentioned in the text.

Platform	Advantages	Disadvantages	Comments
Illumina	–Ultra-high throughput –Low cost per sequencing coverage –Low error rate (<0.1 %) –Benchtop versions available (i.e. MiSeq, NextSeq).	–Short read length (< 300 bp)	–The high throughput and low error rate makes it ideal for re-sequencing projects when a reference is available. –Optimal for epidemiological studies where focus is on sequence variation of many samples. –Not suitable to analyze highly repetitive regions
Oxford Nanopore	–Ultra-long reads (median 20–50 kb, reaching up to 100 kb or more). –Simple and small instrumentation, portable.	–Medium throughput –High error rates in raw sequences (5–40 %)	–Read length enables easy assembly and taxonomic assignment of single reads. –Needs correction (updated versions are more accurate), often in combination with Illumina. –Cheap and portable equipment opens possibilities for clinical use.
PacBio	–Ultra-long reads (median (8–10 kb, up to 60 kb or more)	–Medium throughput –High error rates in raw sequences (10–15 %) –Large equipment	–Similar as above, but less amenable for in site operation, given the size and complexity of the equipment.

Although these developments are very promising, the implementation of whole genome sequencing in the clinic faces important challenges which limits its implementation (Greninger 2018, Consortium & Gabaldón 2019, Kidd et al. 2019). Three major clinical applications of genome sequencing are considered: (i) genome-wide profiling of resistance conferring mutations, (ii) genome-wide molecular epidemiology for the study of outbreaks and, (iii) detection and identification of pathogens from complex, patient samples. For the three applications, analytical methodologies are ready, as we have seen in the examples above. However, the translation of these methodologies from an academic study to a clinical context faces many issues, of which we will discuss the more relevant. The first issue is cost. Although the prices of high-throughput sequencing equipment continue to drop, they are still far away from those of other routine analyses in the clinical mycology lab, particularly when one considers the combined costs of reagents, equipment purchase and maintenance, computational infrastructure, and necessary personnel (Greninger 2018). Another important aspect is time, particularly for diagnostic purposes, less so for epidemiological studies. Most sequencing approaches require multiple steps before sequencing can be run (isolation, culture, DNA extraction, library preparation, etc.). These factors coupled with the required bioinformatics analyses that are often not fully automatised, require high-level expertise, which delay results over the limits that are reasonable in the context of clinical needs. This problem can be aggravated if the sequencing and bioinformatics resources are not on site, which still is currently the case for most hospitals. Finally, despite many developments in computational tools and databases, they are still not mature for routine infection control and outbreak investigations. For instance, many public genome repositories are not curated, leading to wrong annotations that can lead to errors in species identification (Stavrou et al. 2018). In our view, full integration of such methodologies into the clinics requires the following developments: (i) equipment should evolve into smaller, more robust and easier to handle devices minimising expert dedication and maintenance cost; (ii) these should be coupled to a computer storage and artificial intelligence-driven computation system that will automatically process data into clinically meaningful results, this system could be on-site or securely accessed remotely; (iii) curated databases and pipelines should be developed that are directed to specific needs in the clinics; and (iv) expert personnel with bioinformatics and microbial genomics expertise should be incorporated into the clinical system. Some of these developments, particularly (i) and (ii), are progressing significantly thanks to the push of personalised medicine applications based on the human genome. However, (iii) and (iv), only partially overlap with other medical applications of genomics, and require a specific microbial genomics focus, and expertise. We envision that joint international efforts with the participation of regulatory authorities, researchers, and clinicians will help to make progress through pilot proof of concept studies and the standardisation of methodologies.

Genetic toolbox used to identify antifungal resistance and virulence determinants

Inducing mating in A. fumigatus under laboratory conditions is a time-consuming process (O’Gorman et al. 2009), and studies of gene function have therefore heavily relied on site-directed mutagenesis (O’Gorman et al. 2009). However, the effect of the cell wall on the uptake of exogenous DNA and the low rate of homology driven repair (HDR) which ranges from 1–10 % in different strains makes gene manipulation of A. fumigatus an uphill road (Krappmann et al. 2006). Different techniques have been used in the attempt to find a balance between increasing transformation efficiency and limiting ectopic genome integration (Sánchez & Aguirre 1996, Sugui et al. 2005, Szewczyk et al. 2006); usually, polyethylene glycol-mediated transformation of protoplasts is preferred for site-directed mutagenesis (Brakhage & Langfelder 2002). Deleting essential players of the non-homology end-joining (NHEJ) pathway such as Ku70 (Krappmann et al. 2006) and Ku80 (da Silva Ferreira et al. 2006a) dramatically increases HDR frequency, so that 0.5–1 kb homology arms can efficiently mediate gene deletion (Krappmann et al. 2006). Although metabolic markers are available (Weidner et al. 1998, Xue et al. 2004) dominant selection markers are preferred, especially when using animal models, because position effects due to the marker integration site may ultimately affect fungal survival within the host (Liebmann et al. 2004, Greenstein et al. 2006). Commonly used markers (resistance towards hygromycin, phleomycin, or pyrithiamine) have been combined with site-specific recombinase systems to allow for marker recycling and targeting of multiple genes in the same background (Punt & van den Hondel 1992, Kubodera et al. 2002, Krappmann et al. 2005, Hartmann et al. 2010). Using a split-marker approach and a cloning-free fusion PCR strategy to assemble the transformation cassette further streamlined the process (Gravelat et al. 2012, Furukawa et al. 2020). Recently, the generation of a library of 484 transcription factor null mutants in a Δku80 A. fumigatus strain offered insights into the complex transcriptional regulation of azole response in this species (Szewczyk et al. 2006). Comprehensive libraries like this are a powerful resource to dissect pathways involved in antifungal resistance, virulence, and nutritional versatility, which all contribute to the multifaceted pathogenicity of A. fumigatus (Ries et al. 2018, Pérez-Cantero et al. 2020). Nonetheless, using NHEJ-deficient strains in animal models calls for caution due to possible genome instability/sensitivity to abiotic stress, and reconstitution of a functional ku80 gene is not feasible in large-scale efforts (Cairns et al. 2016). Moreover, dissecting the - still largely unclear - molecular basis of azole resistance in clinical isolates requires the ability to manipulate wild-type strains (van der Linden et al. 2015). Functional redundancy further muddies the water, and the sequential deletion of gene family members, albeit possible, is a painstaking process. These drawbacks can be overcome with the implementation of CRISPR-Cas9 editing technology (Fig. 4) (Morio et al. 2020). In 2016, microhomology-mediated end joining (MMEJ) was exploited to efficiently replace a Cas9-targeted gene with an hph cassette flanked by 28-bp homology arms in a NHEJ competent strain (Zhang et al. 2016). The same year, the expression of a ribozyme-flanked sgRNA from a RNA pol II promoter was combined with a split-marker approach to induce a single-nucleotide deletion in a Δku80 strain containing an integrated tetracycline-inducible CAS9 (Weber et al. 2017). Alternatively, Cas9-sgRNA ribonucleoproteins (RNP) can be used in conjunction with MMEJ to induce gene replacement with a hph marker in NHEJ competent strains (Al Abdallah et al. 2017). Multiplexing and protein-tagging have also been accomplished (Zhang et al. 2016). Recently, two gene-free intergenic safe haven regions were discovered in A. fumigatus, to which CAS9/sgRNA expressing constructs or selection markers may be directed, thus resolving potential position effects resulting from random ectopic integration (Pham et al. 2020). Notably, CRISPR-Cas9 finally allows high efficiency site-directed mutagenesis of wild-type strains. However, marker integration in the genome was originally required, either for replacing the target gene (Zhang et al. 2016, Al Abdallah et al. 2017) or for selecting the CRISPR elements (Weber et al. 2017). Last year, Ballard and colleagues (2019) tweaked the systems developed by Nødvig et al. (2015) and Weber et al. (2017) into a two-plasmid system for introducing SNPs into a clinical isolate without marker integration. Single-base CRISPR-Cas9 editing streamlined the association of amino acid substitutions with azole resistance in A. fumigatus, and opens up the way to the discovery of new clinically relevant and CYP51A-independent resistance mechanisms (Umeyama et al. 2018, Ballard et al. 2019).

Fig. 4.

CRISPR-Cas9 technology in A. fumigatus. The vignettes illustrate representative methods for CRISPR–Cas9 genetic manipulation of A. fumigatus. (a) MMEJ can be used to integrate an HPH cassette into a desired locus by using short homology arms. The strain is first transformed with a plasmid expressing Cas9 and containing a PYR4 marker and then with an in vitro transcribed sgRNA and the HPH cassette. (b) The need to set up a suitable system to express the CRISPR elements can be circumnavigated by using CRISPR RNPs, in which two different crRNAs and tracrRNA are assembled in vitro with Cas9 and then transformed into the cells to target the upstream and downstream regions of YFG. MMEJ results in the integration of a HPH cassette into the targeted gene. (c) Ballard and colleagues tweaked the systems developed by into a two-plasmid system for introducing SNPs into a clinical isolate without marker integration. The AMA1 sequence supports the replication of the plasmid harbouring CAS9 in A. fumigatus, which confers resistance to hygromycin. A different plasmid carries a cassette for the expression of a ribozyme-flanked sgRNA from a A. nidulans RNA pol II promoter; after expression in A. fumigatus, the self-splicing activity of the rybozymes releases the mature sgRNA. This plasmid contains a PTRA split marker, interrupted by the same protospacer sequence that is being targeted on the gene of interest. After transformation of both plasmids and a RT containing the SNP to introduce in YFG, Cas9 targets both the protospacer on the desired locus in the genome and the twin protospacer interrupting the split marker. HDR then simultaneously mediates the insertion of the SNP into YFG and the reconstitution of the PTRA marker, thus allowing for selection of the transformants on pyrithiamine without marker integration. (d) The RNP system can also be exploited to affect gene expression, and it was recently used to replace a native promoter with a constitutive hspA promoter by transforming the cells with the RNP particle and a repair template carrying HPH, hspA, and homology arms flanking the insertion site. Af, A. fumigatus; CRISPR, clustered regularly interspaced short palindromic repeats; crRNA, CRISPR-RNA; HDR, homology-directed repair; HDV, human hepatitis delta virus; HH, hammerhead; MMEJ, microhomology-mediated end joining; RNA–Cas9 protein complex; sgRNA, single- guide RNA; tracrRNA, trans-activating RNA; YFG, your favourite gene. HA, Homology Arms; RT, Repair Template. Panels a and b were adopted from Morio et al. (2020) with permission.

Recently, the versatility of the CRISPR-Cas9 technology was exploited to dissect the molecular players of triazole resistance in a collection of A. fumigatus clinical isolates (Rybak et al. 2019). Cas9-sgRNA complexes were used to induce site directed mutagenesis and promoter replacement and to show that – surprisingly – mutations in CYP51A and overexpression of either CYP51A/CYP51B or the efflux pump abcC could not recapitulate the triazole MIC observed. Instead, the combination of CRISPR-Cas9 technology with a split hygromycin B marker approach confirmed that clinically occurring mutations in the sterol-sensing domain of the 3-hydroxy-3-methyl-glutaryl-coenzyme A (HMG-CoA) – observed in more than half of the clinical isolates examined – were indeed responsible for the triazole resistance observed (Rybak et al. 2019).

Non-editing applications developed in other fungi may also be adapted in the future to A. fumigatus, including the use of a catalytically dead Cas9 fused to transcriptional regulators to either activate (CRISPRa) or repress (CRISPRi) gene expression, which could address some of the shortcomings of RNA interference in this species (Cairns et al. 2016, Schultz et al. 2018). Disease outcome in patients at high risk of IA also depends on a timely diagnosis (Latgé & Chamilos 2019). The CRISPR technology may revolutionise the field of diagnostics with regards to nucleic acid detection and antimicrobial resistance profiling directly from clinical samples (Myhrvold et al. 2018, Quan et al. 2019). Overall, we may be at a turning point in the study of virulence and antifungal resistance, as the full potential of CRISPR-Cas9 removes long-standing roadblocks and re-shapes the A. fumigatus genetic toolbox.

Application of murine and insect models to identify virulence determinants and antifungal resistance

In vivo infection models are essential for understanding pathogenesis, dissecting host-pathogen interactions, and identification of pathogen and host traits that contribute to susceptibility and disease development. For Aspergillus species, and A. fumigatus in particular, infection models for a variety of mammalian species, especially mice, rats, and rabbits have been used to model different types of disease caused by Aspergillus (Table 7). Technical aspects and the need for standardisation have been extensively reviewed recently by several authors (Paulussen et al. 2014, Desoubeaux & Cray 2017, 2018, Banfalvi 2018, Mirkov et al. 2019). Therefore, we will focus here on (i) differences between leukopenic and non-leukopenic mouse models of IA, as well as murine models of systemic aspergillosis, regarding the contribution of distinct fungal factors to virulence, (ii) the use of in vivo models, including insect models, to determine the impact of antifungal resistance, and (iii) the role of comorbidities.

Table 7.

Examples of mammalian models for different types of disease caused by Aspergillus.

Type of disease	Host species	Route of infection or exposure	Immunosuppression	References1
Invasive aspergillosis, pulmonary	Mouse	Intranasal	Cortisone acetate; cyclophosphamide; cyclophosphamide and cortisone acetate	Sarfati et al. 2002, Ibrahim-Granet et al. 2010, Wong et al. 2017, Morio et al. 2020
		Oropharyngeal	Triamcinolone acetate	Shepardson et al. 2014
		Intratracheal	Anti-Ly-6G / anti-Ly6C antibody	Shibata et al. 2014, Rolle et al. 2016
		Inhalation chamber	Cyclophosphamide and cortisone acetate	Sheppard et al. 2004, Steinbach et al. 2004
	Rabbit	Intratracheal	Hydrocortisone and cyclophosphamide; cyclosporin A and methylprednisolone	Chilvers et al. 1989, Berenguer et al. 1995
	Rat	Intratracheal	Cyclophosphamide and cortisone acetate	Schmitt et al. 1988, Chandenier et al. 2009
		Intrabronchial	Cyclophosphamide	Leenders et al. 1996
		Intrapulmonal	Cyclophosphamide	Habicht et al. 2002, Becker et al. 2006
Invasive aspergillosis, systemic	Guinea pig	Intravenous	Cyclophosphamide and triamcinolone acetate	Kirkpatrick et al. 2000
	Mouse	Intravenous	None; cyclophosphamide	Johnson et al. 2000, Sarfati et al. 2002
	Rabbit	Intravenous	Cyclophosphamide; cyclophosphamide and triamcinolone acetate	Patterson et al. 1988, 1991
Invasive aspergillosis, cerebral	Mouse	Intracranial, Intracerebral	Cyclophosphamide	Chiller et al. 2002, Clemons et al. 2012
	Rat	Intracisternal	None	Zimmerli et al. 2007
ABPA	Mouse	Intranasal, intratracheal or inhalation	None	Hogaboam et al. 2000, Kurup et al. 2001, Ramaprakash et al. 2009, Hoselton et al. 2010, Fei et al. 2011, Moretti et al. 2014
Aspergilloma	Rabbit	Intrapulmonal	None, but surgically induced artificial stenosis of the bronchus and the ligature of pulmonary artery	Sawasaki et al. 1967
Keratitis	Mouse	Intracorneal	None	Clark et al. 2016
	Rabbit	Intrastromal	Triamcinolone acetate (locally)	Komadina et al. 1985
Endocarditis	Guinea pig	Intravenous	None	Martin et al. 1997

¹Selected references with detailed description of materials and methods.

Respiratory infection of leukopenic mice or mice receiving high-dose corticosteroids represent the most commonly used models to study IA (Desoubeaux & Cray 2017). These two models represent the two main groups of human patients at risk for IA, oncological patients undergoing stem cell transplantation or aggressive chemotherapy and patients after solid organ transplantation, respectively. Both models differ fundamentally in the role of the host response in pathogenesis: Leukopenia prevents substantial influx of immune cells into the lung and results in fungal growth unrestricted by the host responses, but driven by the fungal ability to germinate and acquire nutrients from the host tissue for growth (Stephens-Romero et al. 2005, Ibrahim-Granet et al. 2010, Kalleda et al. 2016). In contrast, large numbers of immune cells are recruited to the lungs of corticosteroid-treated mice, and while partially restricting fungal growth, immune cell recruitment contributes to pathogenesis in this model (Ibrahim-Granet et al. 2010, Kalleda et al. 2016). Thus, fungal factors required for growth in the lung generally impact virulence in both models, whereas determinants involved in the interaction with immune cells might only affect virulence in corticosteroid models. An example for this dichotomy is the secondary metabolite gliotoxin; while reduced gliotoxin production attenuates virulence in corticosteroid-treated mice, no significant effect was observed in the leukopenic model (Kupfahl et al. 2006, Sugui et al. 2007, Spikes et al. 2008). Another example is the transcription factor DvrA that negatively regulates virulence in cell culture and the corticosteroid model but not in leukopenic mice (Ejzykowicz et al. 2010). IA furthermore differs from intravenous, systemic infection of mice that targets the liver, spleen and kidney (Jouvion et al. 2012, Paulussen et al. 2014). Organ-specific differences in nutrient supply are the likely reason why deletion of hcsA, involved in lysine biosynthesis, renders the mutant attenuated in IA but not systemic infection models (Schobel et al. 2010). These examples demonstrate the relevance of using different models for aspergillosis to fully dissect the contribution of fungal factors to host-pathogen-interaction.

While animal models have been and continue to be essential for studying pathogenesis of aspergillosis and identification of fungal virulence determinants, ethical concerns and practical considerations have led to the development and increasing use of invertebrate models, especially insects like adult Drosophila melanogaster and Galleria mellonella larvae (reviewed in Lionakis & Kontoyiannis (2012), Binder et al. (2016)). Important aspects of the innate immune response are conserved between insects and mammals (Medzhitov 2001, Müller et al. 2008, Sheehan et al. 2018), but insects lack lymphocytes and an adaptive immune system. While different levels of immunosuppression can be achieved by genetic manipulation in Drosophila (Neyen et al. 2014) or application of drugs (Chamilos et al. 2008), Galleria mutants are not yet available and non-treated insect hosts are commonly used. Furthermore, insect anatomy is dramatically different from mammals, and while both Drosophila and Galleria can be infected via different routes (Lionakis & Kontoyiannis 2012, Kavanagh & Sheehan 2018), it is not possible to mimic pulmonary infection. Thus, taking into account that the mode of immunosuppression and the route of infection determine the relevance of some fungal factors in mammalian aspergillosis models, results obtained in insect models cannot be directly translated to humans without confirmation in mammalian hosts. In fact, it yet remains to be determined which murine aspergillosis model the insect hosts resemble best, although infection of insects, which share several aspects with systemic infection of mice (Kavanagh & Sheehan 2018). Other limitations include the difficulties of infecting Drosophila with an accurate dose, and the necessity to maintain Drosophila at temperatures below 30 °C. Despite this, insect models have not only been used successfully to identify fungal virulence traits, but also in testing antifungal therapy.

Due to the lack of ethical restrictions, the short generation time, and limited space needed for insect models, they are ideally suited for comparison of a large number of fungal strains, semi-high throughput screening of novel antifungal compounds, and evaluation of combination therapies (reviewed in Kavanagh & Sheehan 2018 and Jemel et al. 2020). One limitation in many studies is however the lack of information on compound concentration at the site of infection. Although this has been addressed for Galleria in several studies (Forastiero et al. 2015, Maurer et al. 2015, 2019, Astvad et al. 2017, Kloezen et al. 2018), systematic analyses of pharmacokinetics in infected larvae (Jemel et al. 2020) and a comparison of larvae from different vendors are lacking. In addition, it remains unclear to which extent the metabolisation of antifungal compounds is comparable in insects and mammals. This is important for azoles, which target fungal cytochrome P450 oxidase and are also metabolised by the host (Sugar & Liu 2000, MacCallum & Odds 2002, Mavridou et al. 2010).

Animal models are essential for the development of therapies for aspergillosis, and especially murine models have been widely applied (reviewed in Paulussen et al. (2014), Lewis & Verweij (2017)). Both murine (reviewed in Lewis & Verweij (2017)) and insects (reviewed in Jemel et al. (2020)) models have more recently also been used to address consequences of antifungal resistance for treatment outcome (Table 8). Important developments are the use of coinfection with susceptible and resistant strains or species (Alcazar-Fuoli et al. 2015), the analysis of the consequences of resistance on fungal fitness in vivo (Lackner et al. 2018), and characterisation of strains evolved during chronic infection (Ballard et al. 2018).

Table 8.

Consequences of antifungal resistance of Aspergillus species in murine models and Galleria.

Species/mutation	Antifungal	Main finding	References
Murine models
A. fumigatus CYP51A (M220K, G54W)	Itraconazole, posaconazole	Systemic infection: resistance did not impact fitness and might increase virulence	Lackner et al. 2018
A. fumigatus CYP51A (M220K, G54W)	Itraconazole, posaconazole	IPA: resistance did not impact fitness	Valsecchi et al. 2015
A. fumigatus CYP51A (G448S)	Voriconazole	IPA: resistance was associated with reduced treatment efficacy	Krishnan-Natesan et al. 2012
A. fumigatus CYP51A (M220I, G54W, TR34/L98H)	Isavuconazole	Systemic infection: resistance was associated with reduced treatment efficacy in a dose-dependent manner	Seyedmousavi et al. 2015a
A. fumigatus CYP51A (M220K) and unknown	Caspofungin, posaconazole	Systemic infection: resistance was associated with reduced treatment efficacy in a dose-dependent manner	Arendrup et al. 2008
A. fumigatus	Posaconazole	IPA: resistance was associated with reduced prophylaxis efficacy	Seyedmousavi et al. 2015b
A. fumigatus	Itraconazole	Systemic infection: resistance was associated with reduced treatment efficacy	Denning et al. 1997a, 1997b, Dannaoui et al. 1999a, 2001
A. fumigatus CYP51A and Fks1	Posaconazole	IPA: resistance was associated with reduced treatment efficacy in a dose-dependent manner	Mavridou et al. 2010, Lepak et al. 2013
A. fumigatus	Anidulafungin, voriconazole	Systemic infection: resistance was associated with reduced treatment efficacy in a dose-dependent manner	Seyedmousavi et al. 2013a
A. fumigatus sequential isolates from CGD patient	anidulafungin, Posaconazole	Systemic infection: resistance was associated with reduced treatment efficacy	Arendrup et al. 2010
A. terreus	Voriconazole	Systemic infection: resistance was associated with reduced treatment efficacy	Salas et al. 2013
A. terreus	Amphotericin B, Itraconazole	Systemic infection: only itraconazole was effective	Dannaoui et al. 2000
A. flavus CYP51C	Voriconazole	Systemic infection: efficacy of voriconazole depended on drug exposure but correlated inversely with MIC	Rudramurthy et al. 2017
Insect models
A. fumigatus CYP51A (M220K, G54E, G54W, TR/L98H)	Posaconazole, voriconazole	In vitro resistance associated with reduced treatment efficacy	Forastiero et al. 2015
A. fumigatus CYP51A (M172V, N248T, D255E, E427K, F46Y)	Voriconazole	Resistance did not impact virulence but was associated with reduced treatment efficacy	Garcia-Rubio et al. 2018a
A. fumigatus CYP51A (N248K/V436A, Y433N substitution)	Itraconazole, posaconazole, 8voriconazole	Resistance did not impact virulence but was associated with reduced treatment efficacy	Chen et al. 2019a
A. fumigatus CYP51A (G54, M220, TR/L98)	N/A	Resistance did not impact virulence	Gomez-Lopez et al. 2014
A. fumigatus sequential isolates from CGD patient	N/A	No association between virulence and resistance profile	Ballard et al. 2018
A. lentulus vs A. fumigatus	Voriconazole	Aspergillus lentulus could not be eradicated by voriconazole treatment in single and mixed infections	Alcazar-Fuoli et al. 2015
A. terreus; resistance mechanism not defined	Amphotericin B	In vitro susceptibility to L-AMB correlated with in vivo outcome	Maurer et al. 2015

One aspect that has not been extensively addressed in animal models for aspergillosis is the presence of comorbidities. Due to differences in anatomy and physiology, insect models are innately limited regarding options to model underlying diseases that affect humans at risk for aspergillosis. However, although murine models for various oncological and metabolic disorders, and infections that might occur in association with aspergillosis exist, they have not been applied to aspergillosis research. One example is aspergillosis in COVID-19 patients, discussed above. Murine models of influenza infection have been described (Thangavel & Bouvier 2014), as well as Mycobacterium tuberculosis, but a combination with aspergillosis has not yet been published. In fact, healthy young mice that are rendered immunocompromised shortly before infection might not adequately represent elderly patients with complex comorbidities receiving a variety of medications that might possibly interfere with antifungal therapy. They furthermore do not reflect the group of “non-immunocompromised” patients that develop IA (Stevens & Melikian 2011). The challenge in addressing comorbidities in animal models is that it makes the models significantly more complex and likely more difficult to standardise. Addressing this requires a combination of expertise for aspergillosis models and animal models for the respective comorbidity. This can likely be only realised by collaboration of groups across disciplines and might require a shift from the predominant use of murine models to other mammalian species like rabbits, which more closely reflect human immune response. Although laboratory mice are convenient to use and their genetic tractability allowed identification of distinct genetic polymorphisms associated with an increased risk for aspergillosis (Garlanda et al. 2002, Gresnigt et al. 2018), their small size limits repeated sampling and the course of the standard models for IA in mice is acute. For chronic types of aspergillosis, several models for ABPA have been developed (summarised in Takazono & Sheppard (2017)). In contrast, only a single model for chronic airway colonisation of immunocompetent mice has been described (Nawada et al. 1996, Urb et al. 2015, Wang et al. 2017) that has not yet been widely applied to study the chronic forms of aspergillosis. This model could be useful to address the possible role of Aspergillus in chronic pulmonary diseases such as cystic fibrosis (King et al. 2016a) and chronic pulmonary obstructive disease (COPD) (Gago et al. 2019). As mouse models for COPD and cystic fibrosis exist (Vlahos & Bozinovski 2014, Semaniakou et al. 2018), these could be combined with models of aspergillosis to gain more insight into pathogenesis.

In summary, a variety of in vivo models for aspergillosis using different host species have been described. Insect models are valuable tools for questions that require a large number of strains or compounds to be screened and can be used to generate hypotheses later to be tested in mammalian models. Mice are the most widely used mammal to study aspergillosis, with practical advantages, but also some limitations. A variety of murine models have been established, that allow addressing different types of aspergillosis. For some aspects of aspergillosis, especially in the context of different comorbidities and also co-infections, it will however be necessary to refine existing models or to develop new high order models to adequately mimic human disease physiology and pathogenesis.

Therapeutic drug monitoring and its role in adjusting optimal azole drug dosage in the clinics

The pharmacokinetic/pharmacodynamic index that describes azoles activity against Aspergillus is the time/AUC (Lepak & Andes 2014) with an fAUC/MIC 0.33–25 corresponding to 50 % survival in mice or net stasis in fungal burden in lung (Lepak & Andes 2014). Preclinical and clinical studies have demonstrated exposure-effect and -toxicity relationships for voriconazole, posaconazole and itraconazole and targets for therapeutic drug monitoring (TDM) for treatment, prophylaxis and toxicity have been determined (Table 1), although most of them were based on low quality evidence (Ashbee et al. 2014). However, the benefit of TDM in treatment response as well as in adverse events of voriconazole have been demonstrated in a randomised controlled clinical trial (Park et al. 2012). Furthermore, TDM could prevent the development of resistance which usually occurs after 4 m (range 3 to 23 mo) of azole therapy with itraconazole, posaconazole (oral solution) and voriconazole (Arendrup et al. 2010, Camps et al. 2012c) as subtherapeutic levels may be associated with emergence of azole resistance (Howard et al. 2010, Moazam et al. 2020).

Given the large number of patients (40–60 %) having unpredictably subtherapeutic/undetectable levels of voriconazole, itraconazole and posaconazole (oral solution) (Park et al. 2012, Hoenigl et al. 2013a, 2014a, Prattes et al. 2016, Yi et al. 2017), TDM is recommended for most patients even for susceptible isolates (Arendrup et al. 2020). Patients for whom TDM is particularly recommended are those with erratic absorption (e.g., due to non-compliance, mucositis, diarrhoea), distribution and elimination (e.g., due to altered pathophysiology, genetic predisposition, insufficiencies, extracorporeal devices), potential drug interactions, poor response, difficult-to-treat infections (difficult sites of infection, resistant isolates) and those belonging to special patient population (e.g., neonates, obese, elderly) (Hoenigl et al. 2014a, Lenczuk et al. 2018). Although the tablet/iv formulation of posaconazole provides sufficient exposure in most patients, the steady state with the current dosing regimens is reached after 7 d of therapy (Dekkers et al. 2016), although posaconazole levels (solution) obtained on day 3 to 5 showed high correlation with day 7 levels (Prattes et al. 2016). Considering the extra delay for TDM and dose adjustment, it may take > 2 wk until the new steady states are reached and verify that target levels are attained particularly for drugs with long half-lives like isavuconazole (Cornely et al. 2019b). Such a delay may be detrimental for invasive infections particularly in neutropenic patients since mortality increases from 41 to 90 % if effective antifungal therapy is delayed by 10 d (von Eiff et al. 1995). Population pharmacokinetic models (Hennig et al. 2006, van Iersel et al. 2018, Shi et al. 2019) and algorithms for early TDM (after 48 h) can be used to identify patients with subtherapeutic levels and speed up the dose adjustment process (Dekkers et al. 2016, Prattes et al. 2016). TDM can be coupled with monitoring fungal biomarkers like galactomannan (von Eiff et al. 1995) and PCR (Moazam et al. 2020) in order to optimise efficacy of azole therapy.

Azoles’ efficacy is reduced by resistance which is often associated with high therapeutic failure rates (∼90 %) (Howard et al. 2009, van der Linden et al. 2011). Most resistance strains have mutations in the CYP51A gene either tandem repeats in the promoter region (TR₃₄, TR₄₆, and TR₅₃) and/or single point mutations (mainly in codons 54, 98, 138, 220 and 448) whereas several other non-CYP51A gene mutations have been described (CYP51B overexpression, overexpression/modification of efflux pumps, mutations in other genes) (Dudakova et al. 2017). TDM could be used to optimise azole exposure against azole-resistant isolates. An exposure-MIC relationship has been demonstrated in experimental pharmacodynamic models (Siopi et al. 2014) and in a retrospective study (Troke et al. 2011) with a recommended TDM target for voriconazole of Cmin/MIC 2 for CLSI and 1 for EUCAST. The probability of a standard voriconazole dosing regimen to attain the pharmacokinetic/pharmacodynamic (PK/PD) target for an isolate with EUCAST MIC 2 mg/L is ∼40 % requiring trough levels > 2 mg/L, which are feasible, whereas for isolates an MIC of 4 mg/L the probability drops to < 5 % requiring trough levels > 4 mg/L, which are usually associated with increased toxicity (Siopi et al. 2014). Similarly, the probability attaining the corresponding PK/PD targets of posaconazole and isavuconazole is very low for isolates with MICs > 0.5 mg/L (Seyedmousavi et al. 2014) > 2 mg/L (Espinel-Ingroff & Turnidge 2016), respectively. However, even if a clinically relevant PK/PD target has been determined, MIC-guided TDM approaches should consider the variation in MIC and in the PK/PD target. Isolates with voriconazole MIC 2 mg/L may harbour either the TR₃₄/L98H or the M220K/R/V CYP51A mutations with MICs 2–16 mg/L and 1–4 mg/L, respectively (Seyedmousavi et al. 2014, Arendrup et al. 2017c). Thus, an isolate with MIC of 2 mg/L may have a real MIC of 4 mg/L or higher depending on the underlying resistance mechanism and the corresponding MIC distribution of mutants leaving no space of optimisation via TDM. Given that resistant isolates have usually high modal MICs to itraconazole (> 16 mg/L), voriconazole (4 mg/L for TR₃₄/L98H and other mechanisms, > 16 mg/L for TR₄₆/Y121F/T289A) and isavuconazole (8 mg/L for TR₃₄/L98H, > 16 mg/L for TR₄₆/Y121F/T289A and other mechanisms though with wide variation) but lower to posaconazole (0.5 mg/L) (Meletiadis et al. 2012, Buil et al. 2018a, 2018b), posaconazole could be a good candidate for dose optimisation against azole-resistant aspergillosis. High dose posaconazole has been used to treat IA by azole-resistant isolates with posaconazole MIC 0.25–2 mg/L in patients (4/7 patients survived and 3/7 died from underlying diseases) (Mouton et al. 2018) and in dolphin after TDM (Bunskoek et al. 2017). A high dose regimen of isavuconazole (400 mg o.d.) has also been proposed for wildtype isolates with MIC 2 mg/L (Espinel-Ingroff & Turnidge 2016) previously considered resistant but now belonging to the new category of area of technical uncertainty as isolates with posaconazole MICs 0.25 mg/L previously considered intermediate susceptible (Arendrup et al. 2020). An MIC-guided TDM has been proposed using the epidemiological cut-off value for the wild-type isolates and 4 × MIC for the non-wild type isolates (Mouton et al. 2018) limiting thus the role of TDM for azole dose optimisation against even the low-level azole-resistant isolates.

In addition to reduced efficacy, subtherapeutic levels may also be associated with emergence of azole resistance particularly in the setting of chronic aspergillosis or HSCT prophylaxis where long azole therapy is required (Steinmann et al. 2015, Moazam et al. 2020). Azole induced resistance has been described in vitro for isolates with non-CYP51A mutants and it has been associated with the overexpression of CYP51B gene and CDR1B efflux pump after exposure to itraconazole (Buied et al. 2013, Fraczek et al. 2013). Thus, it was hypothesised that azole-induced resistance which increases slightly the azole’s MICs allows fungal survival until the development of stable resistance via mutation in target genes. Slight increases in MICs of posaconazole and itraconazole with ultimately development of resistance due to CYP51A overexpression has been reported in isogenic isolates recovered from patients after prolonged azole therapy (Arendrup et al. 2010). Those isogenic resistant isolates are unusually, harbour single point mutations (Camps et al. 2012c) or other non-CYP51A mutations (Howard et al. 2010). Subtherapeutic levels have been associated with the emergence of isogenic resistant isolates with non-CYP51A mutations and of non-isogenic resistant isolates harbouring the TR₃₄/L98H mutation (Howard et al. 2010). The PK/PD relationships and the clinical relevance of these phenomena should be further explored in order to estimate the potential role of TDM in preventing in situ development and de novo emergence of resistance.

Several methods have been used for TDM of azoles. Although chromatographic techniques (HPLC, LC-MS) are sensitive and specific, they are expensive, they have slow runtimes, they are not widely available and usually performed in central laboratories (Ashbee et al. 2014). Bioassays are cheap and simple to perform by each laboratory and can be adapted in different clinical settings including combination therapy (Siopi et al. 2016) but they lack specificity and measure total drug activity including any active metabolites. An immunoassay for quantitative measurement of voriconazole levels in serum samples has been developed and can be used in non-specialised centres (van der Elst et al. 2013). A dry spot blood technique has been developed for TDM of posaconazole, voriconazole and fluconazole (van der Elst et al. 2013). The DBS technique is simple, can be performed at home with finger prick blood, requires small sampling volumes, is less invasive and samples can be shipped at room temperature to the laboratory for analysis. A more innovative technique for TDM is the biosensor technology which provides real-time monitoring and dose adjustment of antimicrobials in a minimally invasive fashion (Rawson et al. 2018). An aptamer for sensing azole antifungal drugs has been developed moving forward this technology (Wiedman et al. 2017).

In conclusion, TDM of azoles is important in order to identify and optimise drug exposure in patients with subtherapeutic levels which are often observed in a large subset of patients treated particularly with voriconazole, itraconazole and the oral solution of posaconazole. Optimising azole exposure against low-level azole-resistant isolates in clinical settings is challenged by the precise and rapid determination of pathogen’s MIC and the underlying resistance mechanism, the turnaround time of drug levels and the dose optimisation techniques for attaining steady target levels. Azole-induced resistance is an area that requires further research to understand the PK/PD relationships and define target levels. The development of bedside point-of-care tests for measuring drug levels is crucial for efficient TDM strategies. Because of azoles’ complex pharmacology and non-linear pharmacokinetics, dose adjustment should be coupled with TDM early on therapy and frequent monitoring of drug levels depending on the disease and patient status.

Treating azole-resistant aspergillosis in the clinic

As discussed, the triazoles including voriconazole, posaconazole, and isavuconazole are recommended first-line antifungal therapy for IA, with itraconazole an option for the treatment of chronic pulmonary aspergillosis. Unfortunately, azole resistance has been an emerging problem over the last two decades, and treatment relies on either monotherapy with lipid formulations of amphotericin B or an echinocandin or combination therapy with an echinocandin or triazole plus lipid formulations of amphotericin B. Recent recommendations from the Netherlands (Schauwvlieghe et al. 2018a) outline the importance of primary treatment with lipid formulations of amphotericin B in settings where azole resistance is above 10 %, with a later switch to an azole if the isolate turns out to be susceptible to azoles. Several promising new antifungal agents are on the horizon that may serve an important role in the treatment of azole-resistant aspergillosis.

Amphotericin B appears inferior to the triazoles in the treatment of aspergillosis. In a large randomised trial comparing amphotericin B deoxycholate to voriconazole in patients with hematologic malignancy, the group receiving amphotericin B had lower rates of treatment response and survival at 12 wk compared to those who received voriconazole (Herbrecht et al. 2002). When treating azole-resistant aspergillosis with amphotericin B, lipid formulations are preferable to amphotericin B deoxycholate due to a decreased risk of toxicity and fewer infusion-related side effects (Hiemenz & Walsh 1996). In addition, a dose of 3 mg/kg of liposomal amphotericin B daily is preferable to 10 mg/kg daily given that no benefit has been seen and higher rates of nephrotoxicity associated with higher doses (Cornely et al. 2007). While the Infectious Diseases Society of America (IDSA) and European Conference on Infections in Leukemia (ECIL-6) do not specifically give recommendations on the treatment of azole-resistant aspergillosis, the European Society of Clinical Microbiology and Infectious Diseases (ESCMID) gives liposomal amphotericin B a strong recommendation for the treatment of voriconazole-resistant IA (MIC = 2 mg/mL), a moderate recommendation for the treatment of chronic pulmonary aspergillosis and a strong recommendation for the treatment of aspergillosis with an MIC to voriconazole > 2 mg/mL (Hiemenz & Walsh 1996, Ullmann et al. 2018).

The echinocandins are not preferable for first-line treatment of IA, although they can be used in salvage therapy or combination therapy as discussed in more detail below. While caspofungin is approved by the FDA for the treatment of IA, other echinocandins such as micafungin and anidulafungin are not FDA approved, although they have activity against Aspergillus species and appear equally efficacious to caspofungin. In one study of patients with IA who were intolerant to or refractory to other preferable agents, caspofungin showed utility as salvage therapy (Maertens et al. 2004). The ESCMID gives monotherapy with caspofungin or micafungin a marginal recommendation for the treatment of voriconazole-resistant aspergillosis (MIC > 2 mg/mL) (Ullmann et al. 2018) and a marginal recommendation (caspofungin) and a moderate recommendation (micafungin) for the treatment of progressive chronic pulmonary aspergillosis with triazole resistance (Hiemenz & Walsh 1996).

Combination therapy with liposomal amphotericin B and caspofungin has shown some benefit compared to monotherapy with liposomal amphotericin B for the treatment of IA. In a small prospective study of patients with hematologic malignancy, this antifungal combination was associated with favourable overall response and increased survival at 12 wk, although the survival difference was not statistically significant (Caillot et al. 2007). This combination has also shown promise in retrospective studies as salvage therapy (Aliff et al. 2003, Marr et al. 2004), although conclusions were limited due to the study design and small sample sizes.

Combination therapy with voriconazole plus an echinocandin has shown promise for the treatment of IA in some studies (Marr et al. 2004, 2015, Viscoli 2004, Singh et al. 2006), although again conclusions for this approach are limited based on study design and small sample sizes. In addition, none of these studies specifically looked at the treatment of azole-resistant aspergillosis. Still, the ESCMID gives combination therapy with voriconazole plus an echinocandin strong recommendation for voriconazole-resistant aspergillosis (MIC > 2 mg/mL) and voriconazole plus anidulafungin a moderate recommendation and posaconazole plus caspofungin a marginal recommendation for voriconazole-resistant aspergillosis with an MIC > 2 mg/mL (Ullmann et al. 2018).

New horizon

There are several new antifungal agents with promise for the treatment of azole-resistant aspergillosis. Fosmanogepix targets the highly conserved enzyme Gwt1 which catalyses an early step in glycosylphosphatidylinositol anchor biosynthesis, compromising cell wall integrity and fungal growth. This compound has been evaluated in vivo (Zhao et al. 2019) and in the murine model of IA (Gebremariam et al. 2019) and a phase II study is underway (ClinicalTrials.gov 2020b). Olorofim (previously F901318) is a novel member of the orotomide class that targets dihydroorotate dehydrogenase (DHODH), an important enzyme for pyrimidine biosynthesis. It has been evaluated in vitro (Buil et al. 2017a, du Pré et al. 2018), in the murine model of IA (Hope et al. 2017), and is currently being evaluated in a phase II clinical trial (ClinicalTrials.gov 2020a). Ibrexafungerp and rezafungin are other novel antifungals that show synergy when used in combination with azoles. Ibrexafungerp has shown to improve azole susceptibility for azole-resistant strains of A. calidoustus and A. terreus when used in combination. Rezafungin has shown some activity against azole resistant Aspergillus in a mouse model (Miesel et al. 2019). Further clinical data on these compounds is needed before determining their place in the treatment of azole-resistant aspergillosis.

Future perspectives

In the near future, there is a need for antifungal susceptibility testing of Aspergillus species to become the standard of care around the world. To accomplish this goal, more rapid and sensitive methods to detect ARAF need to be developed. Optimally, these methods would be applied as rapid tests and detect a broader spectrum of resistance markers directly in the clinical sample. Knowledge about the epidemiology of Aspergillus susceptibility patterns will represent a cornerstone for guiding the appropriate selection of antifungal prophylaxis and treatment. Given that more areas may be burdened with high rates of environmental triazole resistance, triazoles may not be universally recommended as primary antifungal treatment, but instead, treatment choice may depend on local epidemiology of ARAF. New antifungal agents, such as ibrexafungerp, that are currently in clinical stage evaluation with novel mechanisms of action may have central roles in treating these azole resistant infections, as well as species like A. terreus which are often less-susceptible to amphotericin B.

Funding disclosure

AA, RGR, and DSP were supported by NIH AI 109025. MH was supported by NIH UL1TR001442. AC was supported by the Fundação para a Ciência e a Tecnologia (FCT) (CEECIND/03628/2017 and PTDC/MED GEN/28778/2017). Additional support was provided by FCT (UIDB/50026/2020 and UIDP/50026/2020), the Northern Portugal Regional Operational Programme (NORTE 2020), under the Portugal 2020 Partnership Agreement, through the European Regional Development Fund (ERDF) (NORTE-01-0145-FEDER-000013 and NORTE-01-0145-FEDER-000023), the European Union's Horizon 2020 Research and Innovation programme under grant agreement no. 847507, and the “la Caixa” Foundation (ID 100010434) and FCT under the agreement LCF/PR/HP17/52190003. DJA was supported by CF Trust Strategic Research Centre TrIFIC (SRC015), Wellcome Trust Collaborative Award 219551/Z/19/Z and the NIHR Centre for Antimicrobial Optimisation.