The study of congruency between phylogenies of interacting species can provide a powerful approach for understanding the evolutionary history of symbiotic associations. Orchid mycorrhizal fungi can survive independently of orchids making cospeciation unlikely, leading us to predict that any congruence would arise from host-switches to closely related fungal species. The Australasian orchid subtribe Drakaeinae is an iconic group of sexually-deceptive orchids that consists of approximately 66 species. In this study, we investigated the evolutionary relationships between representatives of all six Drakaeinae orchid genera (39 species) and their mycorrhizal fungi. We used an exome capture dataset to generate the first well-resolved phylogeny of the Drakaeinae genera. A total of 10 closely related Tulasnella Operational Taxonomic Units (OTUs) and previously described species were associated with the Drakaeinae orchids. Three of them were shared among orchid genera, with each genus associating with 1–7 Tulasnella lineages. Cophylogenetic analyses show Drakaeinae orchids and their Tulasnella associates exhibit significant congruence (P < 0.001) in the topology of their phylogenetic trees. An event-based method also revealed significant congruence in Drakaeinae-Tulasnella relationships, with duplications (35), losses (25), and failure to diverge (9) the most frequent events, with minimal evidence for cospeciation (1) and host-switches (2). The high number of duplications suggests that the orchids speciate independently from the fungi, and the fungal species association of the ancestral orchid species is typically maintained in the daughter species. For the Drakaeinae-Tulasnella interaction, a pattern of phylogenetic niche conservatism rather than coevolution likely led to the observed phylogenetic congruency in orchid and fungal phylogenies. Given that many orchid genera are characterized by sharing of fungal species between closely related orchid species, we predict that these findings may apply to a wide range of orchid lineages.

Bayesian inference analyses were conducted using MrBayes 3.2.6 (Huelsenbeck & Ronquist, 2001; Ronquist & Huelsenbeck, 2003) to construct phylogenetic trees, with 25 % burn-in, 2,500,000 generations, a tree was sampled every 200 generations with four chains, using the GTR+G model of nucleotide substitution. Convergence of runs was confirmed when the average standard deviation of split frequencies was <0.01 with effective sample sizes >200. Tulasnella phylogenetic trees were visualized in Figtree 1.4.3 (Rambaut, 2016) and midpoint rooted.