Notice: Wiley Online Library will be unavailable on Saturday 27th February from 09:00-14:00 GMT / 04:00-09:00 EST / 17:00-22:00 SGT for essential maintenance. Apologies for the inconvenience.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Understanding evolutionary and ecological processes requires accurate delimitation of species. Species are most commonly defined under the general lineage concept (GLC), where they are considered to be segments of diverging population-level lineages (de Queiroz, 2007). Within the umbrella of the GLC, various species concepts have been developed, including the morphological species concept (MSC), the biological species concept (BSC) and the phylogenetic species concept (PSC). The BSC is characterized by species representing populations that potentially can interbreed. The PSC includes the genealogical concordance phylogenetic species recognition concept, which uses phylogenetic concordance of multiple unlinked genes to identify evolutionary independence of lineages (Taylor et al., 2000). PSC is applied frequently in taxonomic groups where it is difficult to quantify morphological variation or perform mating studies. One such group is Tulasnella, a group of fungi where morphological identification of species is problematic.
Tulasnella includes putatively saprotrophic species on decayed wood (Roberts, 1999; Cruz et al., 2011). Some species are also encountered as ectomycorrhizas (Tedersoo et al., 2010) or orchid mycorrhizal symbionts (Dearnaley et al., 2012). However, many questions remain about species delimitation in this group. We therefore need a robust multilocus method for species delimitation to establish a framework for studying the evolution, ecology and physiology of orchid–fungus relationships.
It is now recognized that the most effective approach to species delineation is the integration of multiple datasets and analytical methods (Sites & Marshall, 2004; Leaché & Fujita, 2010; Yang & Rannala, 2010; Barrett & Freudenstein, 2011). Here we use evidence from six nuclear loci, two mitochondrial loci, orchid-host association and geographical location of samples in a multifaceted approach to delineate species of Tulasnella associated with the Australian orchid genera Chiloglottis, Drakaea, Paracaleana and Arthrochilus. Specifically, we employ gene tree construction methods to resolve fungal species boundaries; use coalescent species tree construction methods (using the programs *BEAST and BPP) to test for host and geographic association; and apply population genetic assignment methods to test for admixture between populations. Finally, germination data are used to explore the correlation between physiological traits and phylogenetic boundaries among Tulasnella. In light of the outcomes we evaluate the implications of multigene approaches for fungal species delimitation in this Tulasnella group.
Materials and Methods
Sampling of Tulasnella isolates
In total, 28 Tulasnella isolates from eight species of Chiloglottis, 17 isolates from seven Drakaea species, nine isolates from five species of Paracaleana and nine isolates from Arthrochilus oreophilus D.L. Jones were analysed (Table S1, Fig. S1). Samples from Chiloglottis, P. minor (R.Br) Blaxell and A. oreophilus were all collected from eastern Australia, while all other samples were collected from south-western Australia.
Six sequence loci for Tulasnella, which includes a mitochondrial (C14436; ATP) and five nuclear loci (C3304; ATP helicase, C10499; 26 proteasome regulatory complex, C12424; isocitrate dehydrogenase, C4722; CAS1, C4102; glutamate synthase), were amplified and sequenced as described previously (Ruibal et al., 2013). These loci were developed using sequences from a 3 kb pair-end sequence library (Roche) on a GS FLX 454 platform using GS XL70 sequencing chemistry (Ruibal et al., 2013). Additionally, the internal transcribed spacer (ITS) region, using primers ITS1 and ITSTul4, and the mtLSU were sequenced and edited following Roche et al. (2010).
A multiple sequence alignment was constructed using the alignment tool in Geneious pro v5.6.3 (Drummond et al., 2011) before performing manual checks and minor adjustments. Intronic regions that were difficult to align for loci C14436 (base pair positions 183–250), C4722 (13–86), C1242 (221–281) and C4102 (164–216) were deleted. Phylogenies of individual and concatenated loci were estimated with a maximum likelihood (ML) analysis using RAxML 7.0.3 (Stamatakis et al., 2008) and using Bayesian inference with MrBayes 3.1.2 (Ronquist & Huelsenbeck, 2003). Support for nodes was assessed for ML trees using 1000 pseudoreplicates of nonparametric bootstrapping in RAxML and with Bayesian Posterior Probabilities (BPP) in MrBayes. A GTR+G substitution model was used for all analyses as all other models are nested inside this model. Trees were visualized using FigTree v1.3.1 (http://tree.bio.ed.ac.uk/software/figtree/) and mid-point rooted. Topological similarity between trees of each locus was tested with a congruency test (de Vienne et al., 2007).
Sequence diversities and genetic divergence were calculated in Mega 5.05 (Tamura et al., 2011) (Table S2). For genetic divergence estimation we employed the p-distance as well as the Kimura 2-parameter (K2P) distances (Kimura, 1980) (pairwise deletion), which, despite some limitations (see Srivathsan & Meier, 2012), is recommended by the consortium for the Barcoding of Life (http://www.barcoding.si.edu/protocols.html) in order to standardize comparisons among studies.
Population structure analysis
While assignment tests are better known for their ability to cluster individuals into K populations (Manel et al., 2005), they are also useful for inferring species boundaries (Noble et al., 2010; Barrett & Freudenstein, 2011). Simulation and empirical studies indicate that Structure is generally robust even when the assumption of Hardy–Weinberg-Equilibrium (Pritchard et al., 2000) is not met (Martien et al., 2007; Behere et al., 2013). We tested for the presence of population and species level structure within fungi across the eight-locus sequence dataset using Bayesian genetic assignment as incorporated in Structure v2.3.3 (Pritchard et al., 2000; Falush et al., 2003) (Notes S1). Separate analyses were run for three sets of fungal isolates, representing the main clades revealed in the eight locus RAxML phylogeny (Fig. 1): (1) Chiloglottis; (2) Drakaea + Paracaleana; and (3) Arthrochilus.
Bayesian coalescent-based species delimitation: host and/or geographic association
Reciprocal monophyly or diagnostic states (e.g. fixed differences between putative species) are often used as criteria for species delimitation using molecular data (Sites & Marshall, 2004). By contrast, coalescent-based species delimitation methods do not require reciprocal monophyly (Fujita et al., 2012). To apply this approach, we first generated coalescent species trees (guide trees) using *BEAST 1.7.4 (Heled & Drummond, 2010). The species trees were inferred alternatively by incorporating the host species from which the Tulasnella was isolated or using the geographic origin of isolates. We then used the guide trees in a Bayesian species delimitation approach using the program BPP 2.0 (Yang & Rannala, 2010) (see Notes S1 for details).
The ability of a random subset of fungal isolates to germinate orchid seed was tested with germination trials of Drakaea, Chiloglottis and Paracaleana following Roche et al. (2010) (Notes S1).
Results and Discussion
Phylogenetic analyses of the eight individual loci showed similar topologies (Figs 2, S2–S8), with all loci passing the congruence test. Gene trees for individual loci (Figs 2, S2–S8) and for the combined eight locus matrix (Fig. 1) always gave strong bootstrap and Bayesian probability support for at least four Tulasnella clades; a clade each for Tulasnella from Chiloglottis and Drakaea + Paracaleana, plus three clades from Arthrochilus. There was also strong support for Tulasnella from P. minor as sister to Tulasnella from Drakaea, however, in loci C12424, C14436 and C3304, all grouped into one clade. Tulasnella isolated from Paracaleana in western Australia all fell within the Drakaea clade except the isolate from P. lyonsii. This isolate was sister to Tulasnella from Drakaea and all other Paracaleana in the combined sequence matrix (Fig. 1) and in five out of eight individual loci (Figs 2, S2–S8); however, strong support was not always achieved for this relationship. Comparing the same isolates, both the ITS and eight-locus concatenated data set resulted in strong support for three Tulasnella clades from Arthrochilus (Figs 1, 2).
Most mycorrhizal studies that attempt to delineate species via ITS sequence data rely on the rule of thumb that clades showing > 3% sequence divergence belong to different species (Nilsson et al., 2008; Peay et al., 2008; Hughes et al., 2009). However, levels of ITS variation can vary within fungal species, meaning the use of a threshold may not always accurately reflect species boundaries (Nilsson et al., 2008). Further, genetic thresholds do not explicitly take into account how the timing of speciation influences patterns of genetic differentiation (e.g. Matz & Nielsen, 2005).
Across the eight loci analysed (5015 bp), mean within host group K2P sequence divergences (converted to percentage) were 0.69% (Tulasnella from Drakaea + Paracaleana), 1.74% (Tulasnella from Chiloglottis) and 6.69% (Tulasnella from Arthrochilus). The mean within host group sequence divergence for the ITS ranged from 0.43% to 7.52% (Table 1). K2P sequence divergence (for ITS) between isolates from P. minor and Drakaea was 0.60% and was 1.3% between isolates from P. minor and P. lyonsii. Paracaleana lyonsii showed 1.4% sequence divergence to the Tulasnella clade consisting of isolates from Drakaea and Paracaleana (excluding P. minor). The individual divergences among Tulasnella isolates from Paracaleana and Drakaea, fall well within the range of ITS variation deemed to be typical of ‘within species variation’ (Hebert et al., 2003; Jacquemyn et al., 2010), where clades with < 3% sequence divergence among individuals are referred to as Operational Taxonomic Units (OTUs) and are considered likely to represent species. The percentage p-distances were comparable or even < K2P distances (Table 1), further supporting low genetic divergence within host groups. Although we do not suggest a universal sequence divergence threshold for Tulasnella, the phylogenetic analyses and a sequence divergence of < 1.8% within host groups suggest the presence of one Tulasnella species (OTU) each from Chiloglottis and Drakaea + Paracaleana, and three Tulasnella species (OTUs) from Arthrochilus (Fig. 1).
Table 1. Within host group and between host group K2P- and p-genetic (italicised) distances for Tulasnella species as calculated from the eight loci concatenated data set and (ITS)
Numbers in bold and in brackets represent within host group genetic distances whereas the remaining numbers represent between host group genetic distances.
K2P –within clade genetic distance of the clades identified in Fig. 1, that is, within the clade consisting of CLM084 and CLM085 = 0.20%; within the clade consisting of CLM007, CLM022, CLM027, CLM028, CLM091 and CLM092 = 0.85 ± 0.46%. K2P-genetic distance between these clades = 9.84 ± 0.16%. K2P-genetic distance between these clades and CLM031 = 17.51 ± 0.40%.
Assignment tests for the two mitochondrial and six nuclear loci indicated that the number of genetic clusters (K) among fungi from Chiloglottis was most likely K = 1 as indicated by the logeP(D) values (Fig. S10). For fungi from Drakaea + Paracaleana logeP(D) values indicated K = 1 or 2, with isolates from P. minor, P. lyonsi and P. terminalis exclusively making up one genetic cluster when K = 2 (inferred ancestry 0.67 (P. terminalis) to 0.99) (Fig. S11). logeP(D) and ΔK values were broadly congruent in the number of K-clusters inferred (Notes S1, Figs S10–S11). For Arthrochilus fungal populations, a single genetic cluster was inferred as the most likely scenario by the logeP(D) values although this was not supported by ΔK values (Notes S1, Fig. S12). Thus, Structure analyses indicated that one Tulasnella species (OTU) each is associated with Chiloglottis and Drakaea + Paracaleana, although the number of species associated with Arthrochilus could not be accurately determined.
Species tree inference and posterior probabilities: host and/or geographic association
The species trees and posterior probabilities inferred with *BEAST and BPP are shown in Fig. S9. Strong posterior probability support is inferred when values are ≥ 0.99. The *BEAST analysis with the 21 host-associated populations (Fig. S9a) provides strong posterior probability support (≥ 0.99) for four populations, one each representing Tulasnella from Chiloglottis, Paracaleana minor, Drakaea + Paracaleana from western Australia, and Arthrochilus. The one isolate from P. lyonsii formed an unsupported sister to the Tulasnella populations from Drakaea and Paracaleana from western Australia. With Bayesian species delimitation using BPP, a 10 species 95% credible model (Table S3) was identified with all prior combinations (Fig. S9a, Table S3). The remaining putative species were not strongly supported and were influenced by the choice of priors. Changing the guide tree topology did not alter overall posterior probability support for the hypothesized species obtained by BPP.
The *BEAST analysis using the geography-associated populations as priors provide strong (posterior probability = 1) support for four Tulasnella geographical populations including two from eastern Australia (those representing populations from Chiloglottis and a population from P. minor) and two from south-western Australia. The P. minor population (eastern Australia) formed a weakly supported sister to the Tulasnella populations from western Australia. Species delimitation analyses with BPP found further support (posterior probabilities ≥ 0.99) for geographic groups (Fig. S9b).
Bayesian species delimitation using the Chiloglottis guide tree supports all nodes, with posterior probabilities of one, except the Tulasnella from C. aff. jeanesii + C. valida node (Fig. S9c). The host-associated guide tree (Fig. S9a) also did not support the Tulasnella nodes from C. aff. jeanesii + C. valida. Thus, for Tulasnella from Chiloglottis, phylogenetic, coalescent (*BEAST) and Structure analyses provide strong support that all tulasnellas analysed from Chiloglottis belong to a single species. This contrasts with BPP, which recognized up to four lineages depending on which guide tree (Fig. S9a–c) was used. In light of the broad congruence across the phylogenetic, *BEAST and Structure analyses, we urge caution in relying on BPP alone for species delimitation.
The three step process for Bayesian species delimitation that we employed: (1) population structure analysis (Structure); (2) species tree reconstruction (*BEAST); and (3) posterior probabilities of species cluster inference (*BEAST and BPP); collectively indicated that one Tulasnella species (OTU) each is associated with Chiloglottis and Drakaea + Paracaleana (Fig. 1).
Although there was some variation in germination success among seed lots, all tested Tulasnella isolates from Chiloglottis germinated Chiloglottis, but not Drakaea or Paracaleana nigrita seed. Interestingly, all fungi from Drakaea and Paracaleana, including divergent fungi from both sides of the country, supported germination of both Drakaea and Paracaleana seed in vitro (Table S4). The results are definite, despite the well-known variability of seed germination trials between replicate isolates and seed batches (Phillips et al., 2011). We interpret the germination results as further evidence that Tulasnella from Drakaea and Paracaleana belong to the same species and as consistent with the phylogenetic evidence that all Tulasnella isolates from Chiloglottis belong to a single species
Conclusions and directions
This study has combined six newly developed DNA sequence loci in combination with the ITS and mtLSU loci to delimit Tulasnella species associated with several genera of Australian orchids. Given the pitfalls of single-criterion species delimitation, we applied a multi-method approach over multiple loci. Further, we used specified sequence divergence estimation methods to allow comparison with future studies. Under the general lineage species concept based on phylogenetics, coalescent species trees and germination data, the multilocus analyses provide further support to our previous ITS phylogenetic studies (Roche et al., 2010; Phillips et al., 2011). All Tulasnella isolates analysed from Chiloglottis belong to one species, whereas those from Drakaea and Paracaleana belong to a sister species (Fig. 1).
What are the implications of our findings, given that an eight-locus analysis is broadly congruent with the ITS locus in this group of fungi? Although PCR amplification of the ITS is sometimes difficult, given its widespread use across fungi spanning many phyla (Schoch et al., 2012), we suggest that there may be a case in Tulasnella, at least, where the ITS provides good phylogenetic support for species delimitation. However, more generally the use of multiple sequence loci remains optimal for species delimitation (e.g. Knowles & Carstens, 2007). We predict that as we move towards routine multilocus analyses, which are becoming increasingly more cost efficient with high throughput sequencing methods, that these new studies will largely consolidate, rather than re-write our current understanding of phylogenetic relationships in fungi.
C.C.L. and R.P. were funded by the Australian Research Council (LP098338), and R.D.P. by the Australian Orchid Foundation and the Holsworth Wildlife Research Endowment. The Department of Environment and Conservation issued permits to collect Declared Rare species of Drakaea. The authors thank Don Gomez for collection of Arthrochilus oreophilus; the Editor Marc-André Selosse, L. Tedersoo and two anonymous reviewers for helpful comments that improved the manuscript.