Diversity of clinical isolates of Aspergillus terreus in antifungal susceptibilities, genotypes and virulence in Galleria mellonella model: Comparison between respiratory and ear isolates

We analyzed the antifungal susceptibility profiles, genotypes, and virulence of clinical Aspergillus terreus isolates from six university hospitals in South Korea. Thirty one isolates of A. terreus, comprising 15 respiratory and 16 ear isolates were assessed. Microsatellite genotyping was performed, and genetic similarity was assessed by calculating the Jaccard index. Virulence was evaluated by Galleria mellonella survival assay. All 31 isolates were susceptible to itraconazole, posaconazole, and voriconazole, while 23 (74.2%) and 6 (19.4%) showed amphotericin B (AMB) minimum inhibitory concentrations (MICs) of ≤ 1 mg/L and > 4 mg/L, respectively. Notably, respiratory isolates showed significantly higher geometric mean MICs than ear isolates to AMB (2.41 vs. 0.48 mg/L), itraconazole (0.40 vs. 0.19 mg/L), posaconazole (0.16 vs. 0.08 mg/L), and voriconazole (0.76 vs. 0.31 mg/L) (all, P <0.05). Microsatellite genotyping separated the 31 isolates into 27 types, but the dendrogram demonstrated a closer genotypic relatedness among isolates from the same body site (ear or respiratory tract); in particular, the majority of ear isolates clustered together. Individual isolates varied markedly in their ability to kill infected G. mellonella after 72 h, but virulence did not show significant differences according to source (ear or respiratory tract), genotype, or antifungal susceptibility. The current study shows the marked diversity of clinical isolates of A. terreus in terms of antifungal susceptibilities, genotypes and virulence in the G. mellonella model, and ear isolates from Korean hospitals may have lower AMB or triazole MICs than respiratory isolates.


Introduction
The importance of invasive fungal infections caused by Aspergillus species has increased among immunocompromised patients due to their considerably high rates of morbidity and mortality [1,2]. Although Aspergillus fumigatus accounts for the majority of invasive aspergillosis, Aspergillus terreus has been recognized as an emerging pathogen of pneumonia, infective endocarditis, and disseminated infections with a high mortality rate [3][4][5][6]. Invasive pulmonary aspergillosis caused by A. terreus has been highlighted because it could rapidly progress in immunocompromised patients despite amphotericin B (AMB) therapy [3][4][5][6]. Invasive otomycosis by A. terreus can also develop in immunosuppressed patients, with lethal consequences if not treated properly [7,8].
In vivo and in vitro data indicate that almost all A. terreus isolates are innately resistant to AMB [9][10][11][12][13]. However, unusually AMB-susceptible A. terreus isolates have become a focus of attention, and the relationship between the AMB susceptibility and virulence of A. terreus isolates was investigated using mouse or Galleria mellonella models [14][15][16]. Although these studies included fewer than five isolates from respiratory specimens, they showed that isolates with an AMB minimum inhibitory concentration (MIC) of 0.5 mg/L were more virulent than resistant isolates with an AMB MIC of > 4 mg/L [14][15][16].
A recent Korean surveillance study reported a high frequency of isolation of AMB susceptible isolates of A. terreus from ear cultures [17]. A different pattern of AMB susceptibility of A. terreus isolates could be related to a great genomic diversity in A. terreus [18]. Genetic diversity in A. terreus species has been demonstrated by genotyping methods, such as random amplified polymorphic DNA, repetitive-sequence-based polymerase chain reaction (PCR), inter-simple sequence repeat PCR, and multilocus sequencing typing [19]. However, these techniques have poor inter-laboratory reproducibility and do not allow for the exchange of results between laboratories [19]. Instead, microsatellite-based typing, also referred to as short tandem repeat (STR)-based typing, has been introduced for genotyping of Aspergillus species and could yield a precise numerical result. There are only three reports of genetic analysis of A. terreus based on STR typing [20][21][22]. Although A. terreus is a frequently encountered species based on Korean surveillance data [17], the molecular epidemiology of clinical A. terreus strains has not been investigated in Korea.
The aim of this study is to investigate the antifungal susceptibility profiles, genotypic relatedness, and virulence for clinical isolates of A. terreus obtained from Korean multicenter study. Microsatellite typing was performed using seven STR markers and virulence was evaluated by G. mellonella survival assay; moreover, the associations of the former two parameters with antifungal susceptibility and genotype were investigated. In addition, we compared the genotypic relatedness, antifungal susceptibility profiles and virulence of respiratory isolates of A. terreus recovered from Korean hospitals with those of ear isolates.

Strains and antifungal susceptibility testing
A total of 31 clinical isolates of A. terreus from respiratory specimens (n = 15) and ear specimens (n = 16) were obtained from 6 South Korean university hospitals from January 2012 to August 2013. Species identification was performed by sequencing of the internal transcribed spacer (ITS) and β-tubulin regions [23,24]. The MICs of AMB, itraconazole, posaconazole, voriconazole for each isolate were determined by the CLSI M38-A2 broth microdilution method [25]. The MIC results of isolates were analyzed using the following recently described epidemiological cutoff values (ECVs): AMB, 4 mg/L, itraconazole, 2 mg/L; posaconazole, 1 mg/L; and voriconazole, 2 mg/L [26]. Isolates for which the MICs were higher than the ECVs were defined as non-wild type (non-WT).

Genotyping using microsatellite markers
In total of seven STR markers were used for microsatellite typing, with a modification of the method established by Rougeron et al. [20]. PCR was carried out using reaction conditions described previously with slight modification. Briefly, PCR involved an initial denaturation step of 95˚C for 10 min followed by 30 cycles of 30 s denaturation, 30 s of annealing at 60˚C and 1 min of extension at 72˚C, with final elongation at 72˚C for 10 min. The PCR products were diluted 10-fold with formamide and 1 μL of this diluted product was combined with 15 μL of formamide, with 0.3 μL of ROX 500 marker (Applied Biosystems Inc., Foster City, CA, USA). The amplicons were separated by size and detected on an ABI3130xL Genetic Analyzer platform equipped with a 16-capillary array (Applied Biosystems) as per the manufacturer's recommendations. Repeat numbers in each marker were assigned by using A. terreus NIH 2624 as a reference. Genetic similarity was calculated according to the Jaccard-similarity coefficient Jxy = a/(a+b+c), where Jxy is the measurement of the genetic similarity between isolates x and y, a is the number of the same genotype in both isolates, b is the number of amplified products at each STR marker observed in x but not in y, and c is the number of amplified products at each STR marker observed in y but not in x [27]. The matrix of similarity was analyzed by the unweighted pairgroup method using the arithmetic average (UPGMA) and a dendrogram was constructed using PAUP software (ver. 4.0b10; David Swofford, Smithsonian Institution, Washington DC, USA).

In vivo virulence study using Galleria mellonella
Virulence was evaluated in the insect model G. mellonella as described previously [14]. Briefly, groups of 20 larvae (~150 mg; S-worm, Cheonan, South Korea) were stored in wood shavings in the dark at 18˚C prior to use. The following three control groups were included: larvae injected with 10 μL phosphate buffered saline (n = 20), larvae that received needle injury only (n = 20), and untouched larvae (n = 20). A Hamilton syringe (25 gauge, 50 μL) was used to inoculate larvae with A. terreus and for introduction of treatments or control solutions into the larvae. To determine the virulence of clinical A. terreus isolates, larvae were infected with 5 × 10 6 conidia per larvae and survival was monitored up to 72 h post-infection at 37˚C. Experiments were performed in duplicate and re-tested when the difference between the two experiments was > 15%. Data from all experiments were combined to calculate the mean values of percent survival values.

Statistical analysis
Statistical anlyses were performed using GraphPad Prism 5 software (GraphPad Software Inc., La Jolla CA, USA). The Kruskal-Wallis test or Mann-Whitney test were used to determine differences in the survival at 72 h between groups. Survival rates were evaluated using Kaplan Meier survival curves and analyzed with the log rank (Mantel-Cox) method. Chi squared or Fisher's exact tests were used to determine differences in the number or percentage of isolates between groups. Differences were considered significant at P < 0.05. granted given the retrospective nature of the project. Patient information was anonymized and deidentified prior to analysis, and no information that could lead to patient identification was used.

Genotyping using microsatellite markers
A total of 31 A. terreus isolates were subjected to microsatellite typing using nine STR markers for genotype determination. However, two STR markers (3A and 3C) were excluded because loci 3A showed limited allelic variation (n = 3), and 27 of 31 (87.1%) isolates failed to amplify for loci 3C. Therefore, seven of nine STR markers were ultimately used for genotype determination in this study. Overall, the 31 isolates comprised 27 distinct genotypes: 24 genotypes (88.8%) were unique to a single isolate, and 3 genotypes (genotypes 1, 7, and 15) were shared by 7 isolates (11.2%) ( Table 2). Two strains of genotype 1, both of which were AMB non-WT (MIC > 4 mg/L), were isolated from respiratory specimens of two patients who were Table 1 CLSI ECVs were used to classify the strains as wild type (WT) or non-wild type (non-WT) in terms of susceptibility [26]. c Geometric mean MICs for the four antifungal agents differed significantly between the respiratory and ear isolates (all, P <0.005). hospitalized in the same hospital (hospital A) at a similar time. Of the three respiratory strains of genotype 7, two were isolated from the same hospital (hospital C), but 6 months apart. Two strains of genotype 15 were isolated from ear specimens from two different hospitals (hospital C and F). Overall, the 15 respiratory isolates yielded 12 genotypes and the 16 ear isolates yielded 15 genotypes, respectively. The genetic similarity of the 31 isolates was determined by calculating the Jaccard index. The values ranged from 0.0 to 1.0, with a mean ± SD of 0.18 ± 0.20. The average genetic similarity value was 0.33 ± 0.19 for the 16 ear isolates, and 0.22 ± 0.25 for the 15 respiratory isolates, which were significantly higher than that of the 31 isolates (0.18 ± 0.20) (P < 0.05) ( Table 3). The average genetic similarity value of the ear isolates did not differ significantly from that of the respiratory isolates. The A. terreus isolates exhibited high genetic diversity at 7 loci by microsatellite analysis (Fig 1). However, the dendrogram for the 31 isolates demonstrated a closer genotypic relatedness among isolates from the same body site (ear or respiratory tract); in particular, the majority of the ear isolates clustered together.

Virulence study using Galleria mellonella
The mean survival rates of G. mellonella infected with each of the 31 A. terreus isolates was 32.6%, with a range of 5% to 72.5% (

Discussion
This study characterized the antifungal susceptibilities, genetic relatedness, and virulence in a G. mellonella model of clinical A. terreus isolates obtained during laboratory-based surveillance of six Korean hospitals. Although A. terreus is known to be intrinsically resistant to AMB [9][10][11][12][13], our findings suggest that AMB susceptible clinical isolates of A. terreus are not uncommon, especially in ear specimens. Notably, we report for the first time that isolates from the same body site (ear or respiratory tract) have greater genetic similarity, and respiratory isolates are significantly less susceptible to AMB and triazoles than ear isolates. However, there was no difference in virulence between respiratory and ear isolates in a G. mellonella model.

Respiratory isolates (n = 15) Ear isolates (n = 16) Total isolates (n = 31)
Average between isolates x and y, a is the number of the same genotype in both isolates, b is the number of amplified products of each STR marker observed in x but not in y, and c is the number of amplified products of each STR marker observed in y but not in x. b P < 0.05, respiratory vs. ear isolates. Many investigations show that A. terreus has intrinsic resistance to AMB, with elevated MICs [9][10][11][12][13], but others have reported that A. terreus isolates have a wide spectrum of AMB MIC ranging from 0.125 mg/L to 16 mg/L [18,28,29]. Our results also showed the broad distribution of AMB MICs, ranging from 0.125 to 8 mg/L. According to the most recent report, 12-13% of A. terreus isolates worldwide have low AMB MICs ( 1 mg/L) [10,11,21]. In the present study, 74.2% (23/31) of the isolates had an AMB MIC of 1 mg/L, suggesting that AMB susceptibility is not uncommon among clinical A. terreus isolates in Korea. The high The dendrogram is based on a categorical analysis of seven microsatellite markers in combination with unweighted pairgroup method using the arithmetic average (UPGMA) clustering. The number on the tree indicates the branch length, showing the difference along a branch. All 31 isolates (R1 to R15 and E1 to E16) comprised 27 distinct genotypes (GT 1 to GT 27) by 7 microsatellite markers. However, when a cluster is defined as the isolation of two or more strains with a branch length distance of < 0.63, ear isolates comprise clusters II and V, and the respiratory isolates comprise clusters I, III, and IV, suggesting a closer genetic relatedness among isolates from the same body site (ear or respiratory tract). Five isolates (R7, R15, E1, E4, and R12) were unique to a single isolate, which did not cluster with other isolate as a branch length distance of < 0.63. See Table 2  proportion of AMB susceptible A. terreus isolates in the present study may be due to the high proportion of ear isolates in this Korean collection. Ear isolates showed significantly lower geometric mean AMB MICs (0.48 mg/L) than respiratory isolates (2.41 mg/L). The respiratory isolates exhibited AMB MICs of 0.5 to 8 mg/L, in line with other reports [30,31].
Of two previous studies using microsatellite typing, one reported a high genetic diversity in a large collection of geographically diverse A. terreus isolates from clinical and environmental sources in India [21], while the other suggested the existence of geographically predominant genotypes of A. terreus species complex in Austria [22]. In the present study, using the microsatellite typing method, the 31 isolates comprised 27 distinct genotypes, indicating that, with the exception of two Genotype 1 isolates, the A. terreus isolates in our collection are highly diverse and unlikely to originate from a common source. Why A. terreus isolates from ear cultures are more susceptible to AMB and triazoles than those from respiratory cultures is unclear, but the explanation may involve the genetic relatedness of the isolates. Notably, the average genetic similarity value of ear isolates was significantly higher than that of the total 31 isolates. The average genetic similarity value of respiratory isolates was also significantly higher than that of the total 31 isolates. Taken together, these findings suggest that Korean A. terreus isolates exhibit considerable genetic diversity without a predominant genotype, but ear isolates from South Korean hospitals comprise closely related genotypes with lower AMB or triazole MICs than respiratory isolates.
After Rougeron et al. [20] introduced microsatellite-based typing using nine markers for A. terreus, a multicenter study used this method to conduct genotyping on clinical and environmental isolates from India, North America and Europe [21]. In that study using nine markers, however, 17.2% (21/122) of isolates revealed no amplification at >2 loci (of all nine loci) after repeated attempts; therefore, 21 isolates were excluded from their study [21]. When we performed typing using all nine microsatellite markers, no isolates had amplification at >2 loci. Instead, all but four isolates showed no amplification at one locus (3C) after repeated attempts, and all but three showed the same genotype (93 in size) at locus 3A. These results may, in part, be influenced by the composition of isolates within a collection with geographic differences. Therefore, in the present study, of the nine microsatellite markers used, two (3A and 3C) were excluded. Overall, a combination of seven markers revealed that the 31 isolates comprised 27 distinct genotypes, and the dendrogram for the 31 isolates demonstrated a closer genotypic relatedness among isolates from the same body site (ear or respiratory tract); in particular, the majority of ear isolates clustered together, suggesting successful microsatellite testing using seven markers for all isolates of A. terreus from South Korean hospitals.
The virulence of AMB susceptible strains of A. terreus has been a focus of interest [14][15][16].
In an in vivo murine model of disseminated aspergillosis, AMB-susceptible strains (MIC, 0.5 mg/L) were more virulent than AMB-resistant strains (MIC, 4 mg/L) [14]. Similarly, in the G. mellonella model, two AMB-resistant strains showed lower virulence than three AMB-susceptible strains [16]. These results are in agreement with previous findings in a murine model [15], suggesting that G. mellonella facilitates in vivo screening of A. terreus. In the present study, the virulence of 31 clinical isolates of A. terreus in the G. mellonella model varied markedly, which confirms a previous report that AMB susceptibility in A. terreus is not necessarily associated with loss of in vivo virulence [14]. However, there was no statistically significant difference in the average 72-h survival rate of G. mellonella infected with AMBsusceptible and -resistant isolates (MIC 0.5 mg/L and > 4 mg/L, respectively) (Fig 2). Rather, a trend toward a lower survival rate of infected G. mellonella according to the increment of AMB MIC (38.8% for a MIC of 0.125-0.5 mg/L; 30.3% for a MIC of 1-4 mg/L; and 25.1% for a MIC of 8 mg/L, respectively) was observed. The reasons for these discrepant results are unclear, but the results might have been influenced by the composition of the isolates. One possible explanation is that the previous three studies tested a few selected isolates, but we assessed 31 clinical A. terreus isolates with a wide range of AMB MICs. Another possible explanation is the AMB-susceptible strains (AMB MIC, 0.5 mg/L) tested in the previous three studies were from respiratory cultures, while all AMB-susceptible strains in the present study were from ear cultures because all respiratory isolates had an AMB MIC ! 1 mg/L.
Although A. terreus is found frequently in the environment [32], infections caused by A. terreus are less common than those caused by A. fumigatus [5]. However, A. terreus infections are associated with a lower rate of a response to AMB therapy and are more frequently fatal than A. fumigatus infections [10]. Here we report for the first time that the antifungal susceptibilities, genotypes, and virulence of clinical A. terreus isolates vary markedly, and that respiratory isolates comprise diverse genotypes with higher AMB or triazole MICs than ear isolates. Because the main portal of entry and site of infection for Aspergillus in human hosts is the respiratory tract, respiratory isolates of A. terreus may be more likely to cause severe infections than ear isolates. Although we did not find a difference in virulence between respiratory and ear isolates, a significantly higher proportion of respiratory isolates than ear isolates was able to kill infected G. mellonella. These characteristics of A. terreus isolates were also supported by clinical information (S1 Table). All patients with ear isolates presented with otorrhea; but had a relatively good prognosis. Of 15 patients with respiratory isolates, 10 had evidence of lung lesions or bronchopulmonary infections, such as pneumonic infiltration, bronchiolitis, or tuberculous empyema, and 6 died mainly due to therapeutic failure from treatment with azoles or caspofungin. There would be several factors that contribute to aspergillosis, including both fungus and host related factors such as strain virulence and host pulmonary structure/immune status, respectively [33]. Therefore, given the marked diversity of A. terreus strains and their opportunistic behavior, we postulate that strain-specific factors contribute to colonization of different body sites, or to aspergillosis, depending on the patient population.