Availability of orchid mycorrhizal fungi on roadside trees in a tropical urban landscape

Urban expansion threatens biodiversity worldwide, therefore urban spaces need to be amenable to biodiversity conservation. On trees in urban environments, natural colonisation and successful translocation of epiphytic orchids are necessary to enhance urban biodiversity, and depend on the availability of compatible orchid mycorrhizal fungi (OMF). However, the extent of OMF presence and distribution, as well as niche requirements for the OMF, remain poorly studied. To identify and quantify OMF on urban trees as well as assess their suitability for native epiphytic orchids, we conducted high-throughput sequencing on tree bark and orchid root samples. OMF were detected at 60% of the study sites on 16% of 270 bark samples (from stem, fork, and branch microsites within each tree). OMF presence and richness on bark samples were related to multiple biophysical factors; in general, humus presence and precipitation levels were positively predictive of OMF presence and richness. We found Ceratobasidiaceae- and Serendipitaceae-associated OMF both on bark and within roots. Orchid species also showed differing mycorrhizal specificity. Sites associated with fungal genera Ceratobasidium, Rhizoctonia, and Serendipita were considered suitable habitats for seven orchid species. The results suggest that urban trees support OMF and are therefore suitable for native orchid species; however, OMF availability are largely constrained by biophysical factors. To maximise the likelihood of translocation success and consequent natural establishment, we propose that (micro)sites are screened for compatible OMF prior to any intervention.

factors that could influence OMF presence and richness on urban sites. As previous studies have reported orchid colonisations in urban environments, we expected OMF to be present on urban trees 18,19 . We also expected variations in OMF presence to be related to specific biophysical factors. Knowledge of OMF availability and associated niche requirements can help with the selection of suitable (micro)sites for future orchid conservation efforts.
The composition of OMF varied between sites. The more common Ceratobasidiaceae was present at 18 sites, of which five had Serendipitaceae present as well (Fig. 1). Both "grassland" and "urban" sites had the highest frequency of OMF occurrence while the former had the highest OMF richness (Fig. 2). "Secondary forest" sites had the lowest frequency of OMF occurrence and richness. The NMDS showed differentiation in fungal community composition between the three habitat-types, particularly "secondary forest" and "grassland" sites ( Fig. 3), which was statistically supported by ANOSIM test (Bray-Curtis; R = 0.16, P = 0.04).
The overall OMF richness across microsites was 12 OMF-OTUs (Richness Chao2 = 17.0 ± 8.1); the closeness of the OMF richness estimate (Chao2 diversity estimator) suggested that our sampling effort (n = 126) was adequate. The fork microsites had the highest frequency of OMF occurrence whereas the branch microsites had the highest OMF richness (Fig. 2). The stem microsites had the lowest frequency of OMF occurrence and richness.
At the microsite scale, habitat-type corresponded to OMF presence ( Orchid-site suitability was assessed at genus-level (see Table 2 and Fig. 1

Discussion
fungal communities associating with tree microsites and orchid roots. Knowing the presence as well as identity of OMF on microsites and within orchid roots can facilitate the process of selecting both suitable microsite and compatible orchid species for translocation attempts. By employing Illumina ® ITS-targeted sequencing approach on bark samples, we found several Ceratobasidiaceae-and Serendipitaceae-related taxa that are commonly associated with both terrestrial and epiphytic orchids 13,37 . Similarly, we detected both fungal taxa in the roots of native epiphytic orchids, as well as Tulasnellaceae-associated OMF. However, the latter was not detected on bark samples; this shortfall was likely due to the downstream effects of inhibitory compounds such as polysaccharides and phenolics, which are typically found in recalcitrant materials such as tree bark 56 . These inhibitors are known to negatively affect primer efficacy by disrupting the annealing process of the primer to the DNA template during amplification and/or sequencing process, i.e. competitive binding 57 . In this case, it is likely that the ITS1/ITS4-TUL primer combination is highly susceptible to this inhibitory effect. Nevertheless, these findings suggest that OMF that are associated with the study species may be present on urban roadside trees as well. Non-rhizoctonia OMF were also discovered on microsites; although largely unknown, this fungal group may potentially form mycorrhizal associations with native epiphytic orchid species. For instance, the basidiomycetous fungal genera-Mycena (Agaricales; sequence identity = 99%) and Marasmius (Agaricales; sequence identity = 90%)-were previously reported to form mycorrhizal relationships with epiphytic orchids, including species from the genus Dendrobium 25,58-60 . A specific ascomycete genus, Fusarium (sequence identity = 99%), was also detected on bark samples. Although ascomycetes are rarely involved in orchid mycorrhizas 13 , Salifah et al. 61 demonstrated that Fusarium sp. can establish functional mycorrhizae with Grammatophyllum speciosum Blume seeds. In this study, we chose a conservative approach of considering only rhizoctonia fungi (i.e. Ceratobasidiaceae, Serendipitaceae, and Tulasnellaceae) for analyses and excluded other possible OMF. Nevertheless, we hope to see more studies implicate OMF beyond the three rhizoctonia taxa, especially since increasingly more OMF, both basidiomycetous and ascomycetous fungi, are being discovered 37,62 . We also encourage future studies to include other primer sets that may improve the detection as well as coverage of OMF on bark samples. Some orchid species may exhibit differing mycorrhizal specificity. We found five orchid species associated with only one OMF-OTU, suggesting high mycorrhizal specificity 25,63 . Conversely, three orchid species associated with multiple OMF-OTUs, of which two formed associations with more than one rhizoctonia taxa, suggesting low mycorrhizal specificity. We also discovered a pair of orchid species from the genus Cymbidium associating with the same Serendipita OMF-OTU. Possible similarities in orchid mycorrhizal fungal preference among closely related epiphytic species were also reported in Martos et al. 25   www.nature.com/scientificreports www.nature.com/scientificreports/ influence a species' suitability for ex situ conservation. It is likely that species with low mycorrhizal specificity are preferable due to their less restrictive mycorrhizal preference 25 . Our evaluation of orchid-mycorrhizal specificity was based on roots sampled from translocated orchids only and thus may be limited. As the study species were part of Singapore's "orchid conservation programme", root sampling was constrained as well. Therefore, to improve the assessment of orchid-mycorrhizal specificity as well as the detection and coverage of OMF, we encourage prospective studies to involve more root samples collected from multiple sites, ideally (if possible), both natural and non-natural areas.
We also discovered non-mycorrhizal taxa that may benefit orchid development. Members of saprotrophic Hypocreales, Helotiales, and Pleosporales were identified on bark and root samples. These non-mycorrhizal taxa, which are frequently found within orchid roots, have been reported to indirectly amplify nutrient access to orchids via decomposition of local substrates [64][65][66] . However, the exact functions and effects of non-OMF on orchid physiology remain largely unknown and warrant further research attention 13,67 .  www.nature.com/scientificreports www.nature.com/scientificreports/ Mycorrhizal distribution and niche requirements. OMF were widespread, occurring at majority of the study sites. We found that surrounding land use influenced OMF presence, with grassland habitat-types having the highest frequency of OMF occurrence and richness. This occupancy pattern may imply that dispersal and subsequent establishment of orchid mycorrhizal fungal spores/inocula are largely unhindered at sites that are surrounded by large, open areas-in this case, managed turf-unlike forests, in which wind velocities are typically reduced 68,69 .
In general, OMF presence and richness were correlated with presence of humus, a substrate derived from decomposed organic matter such as dead plant parts, stemflow leachates, and plant exudates 70,71 . Humus commonly accumulates at tree forks, forming vital sources of nutrients and water for mycelial survival and growth 23,72 . This may be the primary reason why the fork microsite, which is structurally ideal for aggregation of organic debris, had the highest frequency of OMF occurrence. On the other hand, we observed OMF richness to be highest at the branch microsite. Given that tree branches tend to be exposed and structurally angled (cf. fork and vertical stem), it is possible that this part of the tree has the highest capacity for interception of diverse fungal spores/inocula and rain 73 . The open structure and orientation of tree branches however may not be ideal for humus accumulation (cf. fork); this may perhaps be the reason why the frequency of OMF occurrence on branch microsites was slightly lower than fork microsites.
Our results denote the central role of precipitation levels in determining OMF presence and richness at both microsite and site scale. Fungal establishment and mycelial development are particularly dependent on moisture conditions 72 ; Querejeta et al. 74 noted that fungal mortality increases with lower moisture, particularly during dry periods. Presence of mosses can enhance the water-holding capacity of microsites 75,76 . The association between richness of Serendipitaceae-associated OMF and both rainfall and moss presence suggests that members of this fungal taxon may be more sensitive to moisture limitation than that of Ceratobasidiaceae-associated OMF. The additional dependence on moisture conditions may explain their lower frequency of occurrence and richness, as well as limited distribution, as compared to Ceratobasidiaceae-associated OMF.
Overall, our findings indicate that OMF presence and richness vary in accordance to precipitation levels and humus presence. Both biophysical factors are also known to influence orchid germination and development 25 Table 3. Results of generalised linear models (GLMs) and generalised linear mixed models (GLMMs) illustrating the biophysical variables that were influential predictors of orchid mycorrhizal fungus (OMF) presence and richness at microsite-level and frequency of OMF occurrence at site-level. No predictor was influential in determining the frequency of occurrence of Ceratobasidiaceae at site-level. OMF presence models were fitted with a binomial or quasi-binomial error structure whereas OMF richness and frequency of occurrence models were fitted with a Poisson error structure. † GLM or ‡ GLMMs; error structures-b: binomial, qb: quasi-binomial, p: Poisson.
Scientific RepoRtS | (2019) 9:19528 | https://doi.org/10.1038/s41598-019-56049-y www.nature.com/scientificreports www.nature.com/scientificreports/ suggesting possible overlap between niche requirements of OMF and orchid species. Such OMF-orchid correspondence can benefit the microsite selection process, therefore improving the likelihood of a successful translocation. potential applications in orchid conservation. Our findings demonstrate the widespread presence and distribution of OMF-associated (micro)sites as well as their potential as orchid habitats. Based on the orchid-site suitability assessment, the similarities in orchid mycorrhizal fungal associations between bark and orchid root samples may suggest high amenability of native orchids towards translocation efforts, especially for species with low mycorrhizal specificity 20,78 . Such species are able to associate with different fungal lineages, and possibly multiple OMF partners, which can exploit various nutritional resources, and as a consequence, enhance the nutrient uptake of host orchid 66,79,80 . Orchid species that showed high mycorrhizal specificity may be advantageous for orchid conservation as well. Unique orchid-fungal affiliations may induce spatial differentiation among establishment sites (at tree or site scale), leading to possible co-occupancy of multiple species 81,82 .
The geographic distribution of OMF at landscape scale highlights the prevalence of sites suited for orchid species that associated with Ceratobasidiaceae-related OMF; thus, it might be sensible to focus on these species in future translocation attempts. Although more limited, Serendipitaceae-associated sites appear to have greater potential for orchid conservation as this fungal taxon was linked with five study species. To maximise this potential, we propose fungal inoculations on microsites that meet the niche requirements of Serendipitaceae-associated OMF, i.e. presence of humus, moss, and high precipitation levels (see McCormick et al. 47 ). In fact, this may be a good pre-translocation practice for any applicable species. Since high-throughput sequencing-based quantification of species abundance is prone to overestimation 83 , we did not quantify the orchid mycorrhizal fungal density/potential (abundance and vitality) of each site. Nevertheless, fungal density/potential can influence a microsite's capacity to support orchid establishment and development and therefore should be considered in future research 24,47 .
This study demonstrates that translocated orchids can form relationships with compatible OMF on host trees in non-forest habitats, in this case, urban parks (root sampling sites). De hert et al. 84 , Keel et al. 85 , and Waud et al. 78 observed mycorrhizal associations in germinated seeds that were sowed in novel, unoccupied habitats, i.e. sites without adults. Similarly, Downing et al. 20 detected OMF in orchids that were translocated beyond their natural ranges. These evidence suggest that communities of OMF in non-natural habitats, including urban landscapes, can support orchid establishment, survival, as well as development and thus should be considered in future ex situ conservation efforts. Ideally, these fungal communities should have the capacity to support both germinating seeds and mature orchids 86 .
This is the first study to explore OMF presence and distribution, as well as assess the biophysical factors that influence both aspects, on epiphytic microsites directly via bark samples, rather than indirectly via orchid roots. By employing metabarcoding and high-throughput sequencing, we were able to screen a large number of environmental samples and characterise fungal communities at a much higher resolution than before, thus making it an efficient and valuable management tool for large-scale detection of OMF, i.e. potential orchid colonisation sites. This tool may be particularly useful in dynamic or disturbed habitats whereby OMF are stochastically available [87][88][89] . Absence of obligate OMF must be dealt with if orchid conservation programmes are to be successful with lasting conservation benefits 90 . Applicable countermeasures include fungal isolation, preservation, and propagation as well as on-site inoculation of relevant fungal cultures 20,47,91,92 . Cevallos et al. 93 noted the influence of keystone species: a core OMF species in which mycorrhizal communities are built around. Given their fundamental role in mycorrhizal assembly, future conservation strategies should perhaps prioritise keystone OMF and regularly monitor (e.g., via molecular detection) for their continued presence 13 .
Ultimately, knowledge on specific OMF-primarily presence, geographical distribution, and niche requirements-is imperative for future plans of establishing orchids, especially endangered species, in either natural or urban landscapes 13,24,94 . Knowledge on another biotic dependency of orchids, specific pollinator(s), is equally crucial for orchid propagation and therefore continuity 15 . Hence we encourage future studies to focus on orchid-pollinator interaction as well, especially in urban habitats. With appropriate planning, urban environments can serve as viable habitats for epiphytic orchid species, and accordingly, help mitigate their extinction risks.

Methods
Species selection. We focussed on 11 native species (planted as part of the "orchid conservation programme"; see Yam et al. 23 for descriptions); except for Bulbophyllum vaginatum (Lindl.) Rchb. F. (nationally Endangered), C. bicolor (nationally Critically Endangered), and C. finlaysonianum (nationally Critically Endangered), all other species-Bulbophyllum blumei (Lindl.) J. J. Sm., B. medusae, Coelogyne mayeriana Rchb. f., C. rochussenii, D. aloifolium, D. leonis, G. speciosum, and P. cornu-cervi-are Presumed Nationally Extinct 95 (see Supplementary  Table S1). The average number of years since planting of all orchid species is 4.2 years. These species were selected primarily for three reasons: (1) differing conservation statuses, (2) easy access by bucket crane/ladder for root sampling, and (3) consistency in host tree, as all individuals were planted on Albizia saman trees. Albizia samannative to South America-is the most widely planted urban tree in Singapore 96 . It has distinctive characteristics such as wide, umbrella-shaped crown, leaf-free inner branches, and flaky bark that make this species favourable host trees for epiphyte colonisation 19 .

Sample collection.
Through collaboration with Singapore's National Parks Board, we generated a comprehensive list of roads planted with Albizia saman trees. From this list, 30 roads/sites were randomly selected, ten in each habitat stratum ("habitat-type") defined by the dominant surrounding land use type 17 (Fig. 1) www.nature.com/scientificreports www.nature.com/scientificreports/ (surrounded by man-made structures), grassland (surrounded by large areas of manicured turf grass), or secondary forest (surrounded by tree-dominated urban parks, lowland tropical forest, or young secondary forest patches occurring on previously degraded or cleared land) 97 . The dominant land use was defined as >50% of the area within an octagon of ~ 200 m diameter, drawn in Google Earth around the selected site (see Supplementary  Fig. S1). At each site, three individual trees were randomly selected and the location of each tree was recorded using a Garmin ® GPSMAP 60CSx GPS receiver (Garmin International, Inc., Olathe, KS, USA).
To quantify the presence of OMF on roadside trees, we sampled bark (~25 cm 2 per sample) from 90 urban trees in September 2016. For each tree, sample collection points within each of the three microsites-stem, fork, and branch-were randomly selected (total microsites = 270). We then employed a grid system and random number generator to select a specific quadrat for bark collection. Each individual sample was placed in a storage bag and stored at -80 °C prior to DNA extraction.
We sampled roots of translocated orchids to identify species-specific OMF (independent of bark samples). Approximately 5-6 orchid roots were collected from six different individuals per species. To avoid cross-contamination, root samples were collected using a sterilised scalpel and rubber gloves, placed in separate storage bags, and stored at -80 °C until DNA extraction. In total, 65 samples were collected from four different urban parks (see Supplementary Table S1); one B. medusae individual was not collected during the sampling period due to tree pruning.
Several biophysical variables were recorded at each tree: microsite location (stem/fork/branch), substrate (presence/absence of humus and moss), tree DBH (diameter at breast height; measured with a diameter tape, at 1.3 m above the base), mean ambient temperature, mean rainfall, and mean wind speed. Climate data were 5-year daily averages, calculated from collated weather records provided by Meteorological Service Singapore 98 . We also measured the nearest distance between individual trees located on the same road and distance of trees from nearest forest vegetation via Google Earth. Bark chemistry (e.g., pH, hydrocarbon content) was not quantified and analysed in this study.

Molecular analyses.
To extract fungal DNA from tree bark, the material of each sample was removed by scraping the surface with sterile scalpel. Approximately 50 mg of this material was ground using liquid nitrogen, sterile mortar, and pestle, followed by bead-based homogenisation in Bead Ruptor 24 (Omni International, Kennesaw, GA, USA). To extract fungal DNA from orchid roots, the root material was cut into 5 cm pieces, washed with Milli-Q ® water (Merck & Co., Inc., Kenilworth, NJ, United States) for 2 min, surface-sterilised in 1% sodium hypochlorite for 1 min, and rinsed with Milli-Q ® for 2 min. Six to eight root sections per plant were pooled (~50 mg) and homogenised using Bead Ruptor 24.
Total DNA was extracted from both homogenised bark and root samples using a modified CTAB protocol 99 . The fungal nuclear ribosomal internal transcribed spacer (ITS) sequences were amplified using primer combinations ITS86F/ITS4 100 and ITS1/ITS4-TUL 101 . Each primer was tagged with a 9 bp long sequence for specimen-to-sequence association and a dual indexing strategy was used. Two replicates of polymerase chain reaction (PCR) amplifications were conducted for each DNA extract. PCRs were conducted in 25 µl reaction volumes containing 10 µl of 100-times diluted DNA extract and 15 µl of GoTaq ® Colorless Master Mix (Promega Corporation, Madison, WI, USA). PCR temperature profile for ITS86F/ITS4 primer-pair was as follows: initial denaturation step at 94 °C for 3 min, followed by 30 cycles of denaturation at 94 °C for 45 s each, annealing at 59 °C for 45 s, and extension at 72 °C for 2 min. For ITS1/ITS4-TUL primer-pair, PCR conditions were as follows: initial denaturation step at 96 °C for 2 min, followed by 35 cycles of denaturation at 94 °C for 30 s each, annealing at 60 °C for 40 s, and extension at 72 °C for 1 min. The final cycle of both amplifications was followed by a 7-min extension at 72 °C. PCR products were verified on 1% agarose gel, quantified using Qubit ® 2.0 Fluorometer (Thermo Fisher Scientific, Waltham, MS, USA), pooled in equimolar quantities, and cleaned using MinElute PCR Purification Kit (Qiagen, Hilden, Germany). The pooled products were subjected to paired-end (2 × 300 bp) sequencing on an Illumina ® MiSeq sequencer (Illumina, Inc., San Diego, California, USA). Libraries were prepared using NEB Ultra II DNA Library Preparation and amplification-free protocol (New England BioLabs, Massachusetts, USA).
The sequence reads were merged using PEAR version 0.9.11 102 and processed using OBITools version 1.2.11 103 . Samples and replicates were demultiplexed using the ngsfilter command; only amplified regions remained for subsequent analyses. We then employed the obiuniq command to identify and cluster strictly identical sequences. Sequences with <1% sample-specific percentage occurrence were removed (potential amplification/sequencing errors). Sequences shorter than 100 bp were also omitted via the obigrep command. Given that there were two replicates of PCRs per sample, only sequences that passed the various filtering criteria in both replicates were retained for further analyses. The sequences were then aligned and clustered into operational taxonomic units (OTUs) at 97% sequence similarity criterion using VSEARCH version 2.6.0 104 . To identify the different OTUs, BLAST searches were performed on representative sequences to determine the closest match represented in GenBank (nt database). No ties-i.e. same OTU sequence giving hit to multiple species at same best identitywere encountered. A taxonomic assignment was given for a match of ≥85% identity (as done in Xing et al. 105,106 ). Based on current orchid mycorrhizal fungal knowledge, only OTUs related to rhizoctonia taxa (i.e. Ceratobasidiaceae, Sebacinales/Serendipitaceae, Tulasnellaceae; see Dearnaley et al. 13 and Jacquemyn et al. 27 ) were considered potentially mycorrhizal (i.e. OMF-OTU). Other OTUs were considered as non-rhizoctonia OTUs, which are endophytic fungi that associate with orchid roots that may or may not provide benefit to the host 107 . Note that we also tested sequence identification using ecotag in OBITools and EMBL database as well as BLAST to UNITE database, but overall, GenBank yielded better matches.
Additionally, orchid-site suitability was assessed based on similarity of fungal taxa at genus-level, i.e. generic correspondence between bark-and root-associated OMF-OTUs. Thus, if a specific OMF genus was detected on a bark sample-a representative of site-as well as in an orchid root sample, we considered the site to be a potential (2019) 9:19528 | https://doi.org/10.1038/s41598-019-56049-y www.nature.com/scientificreports www.nature.com/scientificreports/ habitat for the orchid species. We chose to compare fungal genera rather than species because (1) based on a recent large-scale study by Martos et al. 25 , tropical epiphytic orchids tend to have low mycorrhizal specificity, mainly associating with typical genera/species found in the three rhizoctonia families and (2) the fungal databases consist of numerous poor-quality fungal sequences-particularly species-level sequences-with low-resolution taxonomic annotations and subpar technical quality 108,109 . Hence to achieve a reasonable yet conservative compromise, a genus-based comparative approach was adopted.

Statistical analyses.
A non-metric multidimensional scaling (NMDS) analysis based on the Bray-Curtis dissimilarity matrix was used to visualize variation in orchid mycorrhizal fungi between habitat-types 110 ; we then performed a variance analysis of these distances using analysis of similarities (ANOSIM) 111 . We also used the nonparametric diversity estimator Chao2 112 to estimate the OMF richness of the microsite-level data set.
We applied either binomially/Poisson-distributed generalised linear models (GLMs) or generalised linear mixed models (GLMMs) 113 to explore the correspondence of biophysical factors with (1) the presence of OMF (any OMF-OTU) at the microsite level, (2) OMF richness (i.e. number of OMF-OTUs) at the microsite-level, and (3) frequency of OMF occurrence (i.e. total number of microsites with OMF-OTUs per site) at the site-level. Multicollinearity was assessed and highly collinear variables were removed from the models. All models were conducted at two levels: (1) overall (i.e. all orchid mycorrhizal fungal families) and (2) family-specific. When overdispersion was evident, the model was fitted with quasi-binomial or quasi-Poisson error structure. The explanatory variables for the models were all recorded biophysical variables. "Tree" was a random factor for all GLMMs. Akaike's information criterion (AIC) 114 was employed for step-wise simplification and model evaluation. Additionally, GLM deviance was estimated as goodness-of-fit. Spatial autocorrelation was examined using correlograms and variograms 115 .
NMDS, ANOSIM, and diversity estimator test were conducted on PAST Version 3.17 116 . All other analyses were conducted using R version 3.4.2 117 , including the packages "spatial" 118 and "car" 119 .