The mycobiome of Australian tree hollows in relation to the Cryptococcus gattii and C. neoformans species complexes

Abstract Cryptococcosis is a fungal infection caused by members of the Cryptococcus gattii and C. neoformans species complexes. The C. gattii species complex has a strong environmental association with eucalypt hollows (particularly Eucalyptus camaldulensis), which may present a source of infection. It remains unclear whether a specific mycobiome is required to support its environmental survival and growth. Conventional detection of environmental Cryptococcus spp. involves culture on differential media, such as Guizotia abyssinica seed agar. Next‐generation sequencing (NGS)‐based culture‐independent identification aids in contextualising these species in the environmental mycobiome. Samples from 23 Australian tree hollows were subjected to both culture‐ and amplicon‐based metagenomic analysis to characterize the mycobiome and assess relationships between Cryptococcus spp. and other fungal taxa. The most abundant genera detected were Coniochaeta, Aspergillus, and Penicillium, all being commonly isolated from decaying wood. There was no correlation between the presence of Cryptococcus spp. in a tree hollow and the presence of any other fungal genus. Some differences in the abundance of numerous taxa were noted in a differential heat tree comparing samples with or without Cryptococcus‐NGS reads. The study expanded the known environmental niche of the C. gattii and C. neoformans species complexes in Australia with detections from a further five tree species. Discrepancies between the detection of Cryptococcus spp. using culture or NGS suggest that neither is superior per se and that, rather, these methodologies are complementary. The inherent biases of amplicon‐based metagenomics require cautious interpretation of data through consideration of its biological relevance.


| INTRODUC TI ON
Cryptococcosis is a potentially lethal mycosis affecting both humans and animals, caused by basidiomycetous yeasts in the Cryptococcus gattii and C. neoformans species complexes.
Cryptococcosis is acquired from the environment by the inhalation of basidiospores or desiccated yeast cells (Kwon-Chung et al., 2014). The C. gattii and C. neoformans species complexes inhabit various ecological niches and are associated with decaying organic material, such as wood, soil and pigeon excreta (Cogliati et al., 2016;Kidd et al., 2007;Nielsen, De Obaldia, & Heitman, 2007;Springer et al., 2014). The ecological role of Cryptococcus spp. is not well understood but their presence in decaying organic matter suggests that they might contribute to the process of decomposition (Voříšková & Baldrian, 2013).
TA B L E 1 Tree species from which the Cryptococcus gattii species complex has been isolated in Australia, including previously published data The C. gattii species complex has traditionally been regarded as tropical or subtropical, whereas the C. neoformans species complex is globally distributed. There is increasing evidence, however, that the C. gattii species complex is also prevalent in temperate climates (Chowdhary et al., 2012;Colom et al., 2012;Kidd et al., 2004). Ellis & Pfeiffer described a specific ecological association between the C. gattii species complex and eucalypt trees in Australia, most notably the river red gum (Eucalyptus camaldulensis) (Ellis & Pfeiffer, 1990), in a wide variety of temperate and subtropical locations.
Since then, the C. gattii species complex has been found globally in decaying wood, particularly inside trunk hollows and in various living tree species, suggesting that trees could be its primary natural habitat (Cogliati et al., 2016;Kidd et al., 2007;Lazera et al., 2000;Randhawa et al., 2008).
In Australia, eucalypts appear to be a key environmental niche for the C. gattii species complex (particularly C. gattii VGI), and the range of tree species from which it has been isolated continues to expand (Table 1). Koalas (Phascolarctos cinereus) exhibit comparatively high rates of both clinical and subclinical cryptococcosis, likely preceded by nasal colonization by Cryptococcus spp., with this association presumed to be related to their close association with eucalypts (Krockenberger, Canfield, & Malik, 2003). In recent years, several cases of cryptococcosis have been observed in freeranging koalas (Phascolarctos cinereus) inhabiting the Port Stephens (Schmertmann et al., 2018) and Liverpool Plains (Schmertmann et al., 2019) regions of New South Wales, Australia. As the Vancouver Island C. gattii VGII outbreak highlighted the potential for animals to represent key sentinels for human disease (Lester et al., 2004), we were prompted to conduct environmental investigations in both areas.
There has been ongoing speculation regarding whether the presence of specific fungal communities or critical species support the TA B L E 2 Detection of Cryptococcus spp. from 23 tree hollow samples in New South Wales, Australia by using conventional culturing and next-generation sequencing growth of Cryptococcus spp. in the environment. A study conducted in Africa attempted to characterize this notion but did not find any associations between the presence of any specific fungal taxa and Cryptococcus spp. (Vanhove et al., 2017). The fungal community residing in eucalypt tree hollows in association with the C. gattii species complex in Australia remains unexplored.
The detection of Cryptococcus spp. from the environment is important in the context of cryptococcosis in order to pinpoint potential infection sources, with culture being the primary means of achieving this. Staib's niger (Guizotia abyssinica) seed extract agar containing antibiotics is a differential and selective media that underpins the conventional detection of Cryptococcus spp. from environmental samples using mycological culture (Paliwal & Randhawa, 1978;Shields & Ajello, 1966). The method is based on Cryptococcus spp. colonies exhibiting the brown-color-effect, due to the production of melanin by cryptococcal laccase (a phenoloxidase enzyme; Staib, 1962). This method is dependent on the visual recognition of suspected cryptococcal colony-forming units (CFUs), which can be challenging when multiple fast-growing filamentous fungi are presented concurrently.
New culture-independent methods, such as next-generation sequencing (NGS), have the potential to detect any organism, including Cryptococcus spp., in the environment without relying on growth in vitro, therefore aiding in the identification of sources of infections.
In addition, they can also define the microbial communities present in environmental samples (Hamad et al., 2017;Taberlet, Coissac, Pompanon, Brochmann, & Willerslev, 2012;Tong et al., 2017). There has been a fundamental shift away from conventional DNA sequencing introduced by Sanger, Nicklen, and Coulson (1977), considered as first-generation sequencing technology, to newer methods, such as NGS. High-throughput sequencing technologies have revolutionized biological research and allowed for the in-depth characterization of microbial diversity, without the need for morpho-taxonomy (Creer et al., 2016).
Fungi might represent the largest genetic diversity among the eukaryotes, with an estimated 5.1 million species, including taxa ranging from unicellular yeasts and microscopic molds to large mushrooms (Blackwell, 2011 (Nguyen, Viscogliosi, & Delhaes, 2015;Tedersoo et al., 2014;Tonge, Pashley, & Gant, 2014). Currently, it is the standard tool and the most efficient method for culture-independent assessment of microbiomes, even if its broad application is still hampered by relatively high cost and the need for special bioinformatic analyses (Tang, Iliev, Brown, Underhill, & Funari, 2015). The approach combines the methodologies of DNA barcoding (Hebert, Cywinska, Ball, & deWaard, 2003) with high-throughput sequencing technology. It is based on the concept that each operational taxonomic unit (OTU) can be unequivocally identified using DNA barcodes. The general strategy involves the following: (a) extraction of DNA from an environmental sample or organism, (b) amplification of the species-specific DNA barcodes, (c) sequencing of the DNA amplicons, (d) analyses of the generated sequences using appropriate pipelines, and (e) taxonomic assignment of the detected sequences. PCRbased metabarcoding has become a rapid and accurate method to species level identification from complex environmental and clinical samples, without the requirement for culture, thereby providing unprecedented insights into the underlying biological diversity (Bik et al., 2012).
In this study, we used amplicon-based NGS as a tool to characterize the fungal community (mycobiome) of tree hollows, allowing assessment of the coexistence of Cryptococcus spp. with other fungal genera in eucalypt and other Australian native tree hollows in two areas of New South Wales (NSW), Australia. The two areas have recently been associated with cryptococcosis in koalas. We also compared conventional culture-based methods with NGS, to determine which was more sensitive at detecting pathogenic Cryptococcus spp.
in environmental samples.

| Sample collection
Debris and related material from hollows were collected from 23 trees selected randomly at multiple locations within the Port Stephens (9) and Liverpool Plains (14) regions of New South Wales, Australia (Table 2). Tree species from which samples were taken (1). Samples were collected as part of a disease investigation into cases of koala cryptococcosis. Generous amounts of material were collected from the interior of each tree hollow and placed into clean plastic bags, which were sealed and labeled then maintained at room temperature.

| Culture
Samples were inoculated onto niger seed extract agar as soon as possible by introducing a sterile swab, premoistened with sterile saline, into the sample and then gently rolling the swab across the agar plate. Plates were incubated at 27°C for a minimum of 7 days and monitored daily. Cryptococcus spp. CFUs were identified by the brown-color-effect and their yeast-like growth on niger seed agar.
Once suspected cryptococcal CFUs were observed, the agar plate was removed from the incubator and one or more CFUs were subcultured onto Sabouraud's dextrose agar for isolation of a pure culture, which was followed by DNA extraction (see below). Samples E2657, E2666, and E2704 had five, four and three isolates collected, respectively, from each primary isolation plate. In all other positive samples, only one isolate was collected. DNA extraction from cryptococcal isolates was performed using an adaptation of an established fungal DNA extraction method (Ferrer et al., 2001). Restriction fragment length polymorphism (RFLP) analysis of a URA5 PCR product was used as described previously (Meyer et al., 2003) to determine the cryptococcal species and molecular type of each isolate.

| DNA Extraction for NGS
DNA was extracted directly from tree hollow material by initially grinding approximately 20 g of each sample with liquid nitrogen using a mortar and pestle. This homogenized the sample and aided in breaking down both the cryptococcal capsule and fungal cell walls. DNA was then extracted from the ground portion of the sample using the DNeasy PowerSoil kit (Qiagen GmbH) following manufacturer's instructions.

| Sequencing
Sequencing of PCR amplicons was conducted with MiSeq ® System of Illumina (Illumina) by the Australian Genome Research Facility. The Illumina bcl2fastq 2.18.0.12 pipeline was used to generate the sequence data.
Pair-ends reads 2 × 300 bp were generated up to 0.15 GB per sample.

| Bioinformatics pipeline and analysis
Reads were processed according to the protocol described in the USEARCH documentation (https ://drive5.com/usear ch/manua l/ uparse_pipel ine.html) using USEARCH package (Edgar, 2010) version 10.0. The sets of OTU were generated, "zero-radius OTUs" (ZOTUs) (i.e., error-corrected (denoised) sequences), using the UNOISE algorithm including chimera filtering (Edgar, 2016b) to identify all correct biological sequences in the reads. The ZOTU table was normalized to the same number of reads per sample (5,000) prior to downstream analysis, to be able to compare data from different measurements.
All singletons were kept for downstream analysis. The taxonomy was predicted for ZOTU sequences using the SINTAX classifier (Edgar, 2016a) against the most recent combined (available on 04.03.2017) UNITE full dataset (Kõljalg et al., 2013) and ISHAM-ITS (Irinyi et al., 2015) database containing all relevant ITS sequences of Cryptococcus spp. The SINTAX algorithm predicts taxonomy by using k-mer similarity to identify the top hit in a reference database, supported by bootstrap values for all ranks in the prediction (Edgar, 2016a correction (Benjamini & Hochberg, 1995) was used to adjust p values for multiple comparisons to limit the probability of even one false discovery.

| Detection of Cryptococcus spp. by culture
Cryptococcal CFUs were observed after culture on niger seed agar in 14/23 (61%) tree hollow samples studied (

| Detection of Cryptococcus spp. using NGS technology
The bioinformatics tools were able to identify C. gattii or C. neoformans species complex specific reads in 16/23 (70%) samples. C. gattii species complex reads were detected in 16 samples, of which 12 were also positive for C. neoformans species complex reads. No Cryptococcus spp. reads were detected from seven samples (E2657, E2668, E2677, E2697, E2698, E2699, and E2711) (Figure 1a).    Cryptococcus (2.5%) were the most dominant. In Mortierellomycota, the most common genera were Mortierella (55.3%) and Gamsiella (2.6%). The overall ten most abundant genera in each sample are shown in Figure 6. The differential heat tree analysis (Figure 9) highlighted numerous taxa that differed in the abundance between trees with no Cryptococcus F I G U R E 7 Heat tree of the abundance of taxa at different ranks of the 23 tree hollows in New South Wales, Australia. The size and color of nodes and edges are correlated with the abundance of taxa. The central node is the total of all the other nodes in the tree for each phylum spp. reads compared with those with one or more reads. In trees with  It is also possible that cryptococcal DNA was detected using NGS but insufficient viable organisms were present in these samples for culture-based detection to be successful. This observation may have clinical relevance and should be considered carefully if NGS results are used as part of a disease investigation, as viable live yeast cells or basidiospores are required to initiate an infection.

| D ISCUSS I ON
The potential for environmental DNA to complicate NGS results is further explained later in the discussion. The culture-positive but NGS-negative results for Cryptococcus spp. are likely related to the multitude of biases encountered in amplicon-based metagenomics, as discussed later. It is also of note that the number of Cryptococcus spp. reads based on NGS did not appear to be consistent with the culture-based grading of low, moderate or heavy. These findings again suggest that abundance-based results are generally considered unreliable in amplicon-based metagenomics studies (Nguyen, Smith, Peay, & Kennedy, 2015;Tessler et al., 2017).
Many samples were found to contain C. neoformans VNI/VNII sequences using NGS, as this method is unable to distinguish between VNI and VNII since their ITS1 regions are identical (Figure 1b; F I G U R E 9 Taxa abundance tree with differential heat mapping in the presence of Cryptococcus gattii/C. neoformans species complexes in the mycobiome of 23 Australian tree hollows. The color of each taxon represents the log 2 ratio of median proportions of reads observed in C. gattii/C. neoformans species complex negative and positive samples. Taxa colored brown are more abundant in negative and those colored blue are more abundant in positive samples. Only significant differences are colored, as determined using a Wilcoxon rank-sum test followed by a Benjamini-Hochberg (FDR) correction for multiple comparisons Katsu et al., 2004). These findings were not supported by the culture results, as none of the isolates collected were identified as being members of the C. neoformans species complex using URA5-RFLP analysis. This could be due to the potential high error rate of NGS and a very high similarity between the ITS1 regions of the C. gattii and C. neoformans species complexes. Differentiating between them relies on only three polymorphic sites in the ITS1 region (Figure 1b; Katsu et al., 2004), which may also explain that C. neoformans VNI/ VNII reads were only obtained in samples that also had C. gattii VGI reads. However, the number of C. neoformans VNI/VNII reads suggested that they are unlikely to be solely attributable to sequencing error. It is also possible that members of the C. neoformans species complex were present in the samples but were in a quiescent state and therefore did not grow on the culture media (Hommel et al., 2019). The Cryptococcus reads after denoising were extracted, visually checked and their taxonomic assignments were confirmed using BLAST and pairwise alignment against the reference ITS sequences.
The choice of another target, more discriminatory between the C. gattii and C. neoformans species complexes, such as the entire ITS region or the URA5 gene (Meyer et al., 2003) (Damm, Fourie, & Crous, 2010;de Hoog, Guarro, Gené, & Figueras, 2000;Khan et al., 2013;Taniguchi et al., 2009). There is some speculation that yeasts may be overrepresented in NGS analyses due to their higher nucleus to cytoplasm ratio when compared to filamentous fungal species with longer cells (Lindahl et al., 2013).
Aspergillus and Penicillium spp. are well-known fungal genera associated with wood degradation (soft rot fungi) in nature, since they tolerate wide ranges of temperature, humidity, and pH, and attack a variety of wood substrates. Soft rot fungi are more common in hardwood, such as Eucalyptus spp., than in softwood which might be due to differences in the quality of the lignin (Hamed, 2013). Other genera, such as Scytalidium, Blastobotrys, Jaapia, or Mortierella, are also commonly found in decaying organic matter and produce enzymes that enhance the degradation of proteins in the wood of dead trees (Middelhoven & Kurtzman, 2007;Takahashi & Oda, 2008;Telleria, Dueñas, Melo, Salcedo, & Martín, 2015;Wagner et al., 2013).
Statistical comparisons between tree hollow mycobiomes were not considered relevant in this study, because the primary focus of the sampling was to investigate the connection between cryptococcosis cases of koalas and the environmental source of these infections. A more in-depth statistical analysis, requiring the systematic collection of further samples and additional information about trees and tree hollow characteristics will be subject of future studies. Further work of interest to the authors will be directed toward characterizing the mycobiome of the koala nasal cavity to determine how closely this reflects that of nearby tree hollows.
Although we identified 2,638 OTUs among the samples, only 7.5% were classified to species and 25.6% to genus level. Moreover, 34.2% of the OTUs were singletons which are largely due to the choice of the algorithm in the downstream analysis. Below order level, most OTUs remained unclassified without any taxonomic predictions. Our findings regarding the assignment success rate of OTUs agree with previous metagenomics studies carried out in a different environment (Schmidt et al., 2013;Soliman, Yang, Yamazaki, & Jenke-Kodama, 2017;Sun et al., 2015;Yuan et al., 2017).
The lack of a relationship between the presence of Cryptococcus and any other fungal genera based on the correlation analysis is consistent with a prior study that also found no significant relationships between Cryptococcus and any other fungal taxa in environmental samples (Vanhove et al., 2017). However, the differential heat tree analysis suggested some differences in the relative abundance of numerous taxa in samples with versus without Cryptococcus reads.  (Nguyen, Smith, et al., 2015;Tessler et al., 2017). Therefore, it is difficult to draw definitive conclusions from these findings, and further work, such as more systematic sampling and numerous technical replicates would be required to determine how reliable these potential associations are. We also did not attempt to find correlations between Cryptococcus and bacteria or nonfungal eukaryotes in Australian tree hollows, which most likely will also influence the composition of the mycobiome but are beyond the scope of the current work.
However, denoising algorithms present a challenge and another potential bias of NGS data interpretation, as defining an abundance threshold that differentiates correct sequences from random errors is difficult (Schirmer et al., 2015). The high number of singletons observed in the dataset might have been attributable to the settings in UNOISE in the analysis. It is also well recognized that inferences from metagenomics studies are greatly influenced and varied by the fields, laboratory and analytic techniques utilized (Majaneva, Hyytiäinen, Varvio, Nagai, & Blomster, 2015). Besides technical biases in sample preparation, DNA extraction and sequencing, there is also another level of complexity and biases in downstream analyses and databases.
Another major limitation of NGS metabarcoding is the lack of reference databases, which are necessary to determine the phylogenetic affiliation of sequence reads. Taxonomical assignments are only as good as the reference databases (Santamaria et al., 2012).
This study used the UNITE full dataset (Kõljalg et al., 2013) which also includes the ISHAM-ITS (Irinyi et al., 2015) dataset containing all relevant ITS sequences of Cryptococcus spp. The SINTAX taxonomy classifier (Edgar, 2016a) was chosen to predict the taxonomy of the identified sequences, as it achieved comparable or better accuracy than the popular RDP Naive Bayesian Classifier (NBC) (Wang, Garrity, Tiedje, & Cole, 2007).
Another bias of amplicon-based metagenomics is the uncertainty as to whether the detected DNA belongs to a live or a dead microbe. DNA is ubiquitous and stable in the environment and can account for roughly 10% of extractable phosphorus in soil (Turner & Newman, 2005). Extracellular DNA fragments can persist over time in many environments, allowing for their detection with highthroughput sequencing technology (Nielsen, Johnsen, Bensasson, & Daffonchio, 2007). Eucalypt hollows could protect environmental DNA from some forms of degradation, including extreme heat and UV-radiation.
Another interesting finding of this study is its presentation of the Based on our findings of discrepancies between culture and NGS-based identification of Cryptococcus spp. from environmental samples, neither approach can be considered definitive. Spiking samples with Cryptococcus spp. could prove a useful future experiment to further assess this. NGS potentially could not differentiate between the C. gattii and C. neoformans species complexes, no significant correlation between the presence of Cryptococcus spp. and other fungal genera or taxa could be identified, and abundance-based analyses were inconclusive. As expected, the mycobiome of Australian tree hollows reflected the microenvironment of decaying wood. Given the discrepancies between culture and NGS results and the multitude of potential biases in amplicon-based metagenomics, meaningful inferences are difficult to establish, and results must be interpreted with caution. Further improvements in NGS, such as whole-genome shotgun and long-read sequencing, together with appropriate data analysis pipelines and the extension of reference databases should significantly contribute to better characterization and understanding of such complex microbial community structures. Further work in this area should include assessing possible correlations between Cryptococcus spp. and bacteria, free living ameba, and nematodes.

ACK N OWLED G M ENTS
We

DATA AVA I L A B I L I T Y S TAT E M E N T
All raw reads from this study were submitted to NCBI's Sequence Read Archive under the BioProject accession PRJNA497337.