Abstract
Epiphytism has evolved repeatedly in plants and has resulted in a considerable number of species with original characteristics. Because water supply is generally erratic compared to that in soils, succulent forms in particular are widespread in epiphytic species. However, succulent organs also exist in terrestrial plants, and the question of the concomitant evolution of epiphytism and succulence has received little attention, not even in the epidendroid orchids, which account for 67.6 % of vascular epiphytes.
We built a new time-calibrated phylogenetic tree of Epidendroideae with 203 genera treated in genus Orchidacearum, from which we reconstructed the evolution of epiphytism as well as traits related to water scarcity (stem and leaf succulence and the number of velamen layers), while testing for the correlated evolution between the two. Furthermore, we estimated the ancestral geographical ranges to evaluate the palaeoclimatic context in which epiphytism evolved.
Epiphytism evolved at least three times: 39.0 million years ago (Mya) in the common ancestor of the Malaxideae and Cymbidieae that probably ranged from the Neotropics to Southeast Asia and Australia, 11.5 Mya in the Arethuseae in Southeast Asia and Australia, and 7.1 Mya in the neotropical Sobralieae, and it was notably lost in the Malaxidiinae, Collabieae, Calypsoeae, Bletiinae and Eulophiinae. Stem succulence is inferred to have evolved once, in a terrestrial ancestor at least 4.1 Mya before the emergence of epiphytic lineages. If lost, stem succulence was almost systematically replaced by leaf succulence in epiphytic lineages.
Epiphytism may have evolved in seasonally dry forests during the Eocene climatic cooling, among stem-succulent terrestrial orchids. Our results suggest that the emergence of stem succulence in early epidendroids was a key innovation in the evolution of epiphytism, facilitating the colonization of epiphytic environments that later led to the greatest diversification of epiphytic orchids.
INTRODUCTION
Epiphytism, or the ability of some plants to grow on the surface of other plants, is a major component of tropical and subtropical forests on Earth (Benzing, 1990; Zotz, 2016), representing about 9 % of global vascular plant diversity (Zotz, 2013, 2016), and up to 50 % of the plant species in some tropical rainforests (Silvera and Lasso, 2016). Among epiphytes, orchids are of particular interest, as they alone account for 67.6 % of all epiphytic species (21 169 out of 31 311; Zotz et al., 2021), and almost all of them belong to the Epidendroideae (Zotz, 2013), the largest subfamily of Orchidaceae. Givnish et al. (2015) demonstrated that several factors, including an epiphytic lifestyle and tropical distribution, were associated with higher net diversification rates in orchids. Although they did not specifically address the evolution of epiphytism in the Orchidaceae, their taxon sampling (162 species belonging to 18 tribes) allowed them to identify at least one transition to the epiphytic lifestyle at the orchid family level, in the subfamily Epidendroideae, but this analysis was carried out at the level of subtribes only. Two other ancestral estimations of lifestyle evolution were conducted on more extensively sampled phylogenies, including 335 orchid species (Chomicki et al., 2015) and 312 Epidendroideae species (Freudenstein and Chase, 2015), respectively. Freudenstein and Chase (2015) detected at least three independent origins of epiphytism in Epidendroideae, challenging the primary single origin of the epiphytic lifestyle found by Givnish et al. (2015) and Chomicki et al. (2015). While the latter group includes mainly epiphytic taxa in the tribes Cymbidieae and Vandeae, the former group comprises both terrestrial lineages that are rather basal in the tree and more recent epiphytic lineages, such as the Dendrobiinae which include the genus Bulbophyllum. In addition, confirming the early assumption of Dressler (1981), these studies found multiple re-terrestrialization events (Chomicki et al., 2015; Freudenstein and Chase, 2015; Givnish et al., 2015). Indeed, several lineages nested in the Epidendroideae diversify on the ground and show features similar to those of related epiphytic lineages, suggesting that reversions from the epiphytic lifestyle to the terrestrial lifestyle may have occurred in more recent times. Finally, recent phylogenetic works reappraising the phylogenetic relationships in Epidendroideae using genomic molecular data (Givnish et al., 2015; Li et al., 2019; Pérez-Escobar et al., 2021; Serna-Sánchez et al., 2021) now provide a robust phylogenetic framework to assess the evolution of this group.
Although diverse and heterogeneous in terms of environmental conditions, epiphytic habitats are characterized by an irregular water supply for plants (Benzing, 1987; Cribb, 1999; Zotz, 2016; Hietz et al., 2022; Taylor et al., 2022). Water shortage in epiphytic habitats could explain why water-catching and water-storing organs are commonly observed among epiphytic plants, such as the tank-forming leaves holding water in some bromeliads (Givnish et al., 2014; Zotz, 2016). Most epiphytic orchids and many Araceae have one or more layers of dead epidermal cells surrounding the root, called a velamen, which helps to capture water running off the surface of the tree (Cribb, 1999; Zotz and Winkler, 2013; Stern, 2014). Thickened stems and leaves are also common to conserve and store water and nutrients to compensate for their irregular supply (Cribb, 1999; Ng and Hew, 2000; Stern, 2014; Yang et al., 2016; Zotz, 2016; Niechayev et al., 2019; Fu et al., 2022; Hietz et al., 2022). In addition, Crassulacean Acid Metabolism (CAM) is a water-conserving trait widespread among epiphytic plants (Luttge, 2004; Zotz, 2016; Niechayev et al., 2019; Fu et al., 2022; Hietz et al., 2022; Orlov et al., 2022). With ~2100 species, the pantropical genus Bulbophyllum in Epidendroideae illustrates the ecological and evolutionary success of these characters (Gravendeel et al., 2004; Gamisch and Comes, 2019). Even though traits related to water scarcity (including succulence) were found to be more pronounced in epiphytes than in terrestrial species (Rada and Jaimez, 1992; Zhang et al., 2015; Hietz et al., 2022), justifying an ‘epiphytic syndrome’ according to Hietz et al. (2022), these features are not exclusive to epiphytes, as they are also found in terrestrial plants (Zhang et al., 2015; Zotz et al., 2017; Hietz et al., 2022). For example, the velamen is a typical feature of epiphytic orchids, but is also found in numerous terrestrial genera in many families, with a priori no epiphytic ancestors from which they could have retained this feature (Zotz et al., 2017). Therefore, as stated by Zotz (2016): ‘in the absence of phylogenetic analyses, it usually remains unclear whether a feature really represents an adaptation, or whether previously evolved traits simply proved to be advantageous in an epiphytic environment’. The second scenario would correspond to a key innovation as defined by Miller et al. (2023): ‘an organismal feature that enables a species to occupy a previously inaccessible ecological state’.
Dressler (1981) hypothesized that water-deprived traits could have evolved as adaptations to drought in seasonally dry tropical climates in terrestrial orchids, later facilitating the transitions to the epiphytic niches. Indeed, an ancestral state estimation carried out by Freudenstein and Chase (2015) suggested that the succulent stems, called pseudobulbs, common in modern epiphytic orchids could indeed have arisen before the evolution of the epiphytic lifestyle, but the method they used nevertheless returned an uncertain state. In the present study, we aimed to test Dressler’s hypothesis; that is, did traits related to water scarcity appear in epiphytic lineages as adaptations (i.e. the adaptation hypothesis) or did they emerge first in terrestrial ancestors as features that facilitated access to novel ecological states, here the epiphytic environment (i.e. the key innovation hypothesis)? To address these questions, we (1) analysed the evolutionary history of the epiphytic lifestyle and water-related traits in a single state-of-the-art analysis, using the most recent and robust phylogeny and methods available; (2) ascertained the likelihood of an emergence of succulence traits prior to the epiphytic lifestyle and estimated if transitions from terrestriality to the epiphytic lifestyle were more frequent among drought-adapted lineages; and (3) identified the palaeoclimatic context of the evolution of epiphytism, namely when and where water-related traits and the epiphytic lifestyle arose and were lost.
MATERIALS AND METHODS
Phylogenetic analysis of Epidendroideae genera and divergence times
To produce a well-sampled, robust, time-calibrated phylogenetic reconstruction of the Epidendroideae at the genus level, we used a molecular dataset of three plastid genes (matK, psaB, rbcL). Sequences were retrieved from GenBank for 203 placeholders representing each genus of Epidendroideae listed in genus Orchidacearum (Pridgeon et al., 2005, 2009, 2014). Whenever possible, all gene sequences came from the same voucher specimen (69 % of the genera), and if not from different specimens of the same species (14 %), or from different species (17 %). GenBank accession numbers are provided in Supplementary Data Table S1. Name acceptance by the WCSP (2020) of vouchers and genera was systematically checked. Although synonyms of Dendrobium, Cadetia and Epigeneium were kept for reasons of fossil calibration (see below). Sequences of each gene were aligned using MAFFT in Geneious Prime 2021.2, manually checked and then concatenated, resulting in a molecular dataset of 4649 characters. The best-fitting partition scheme was selected using ModelFinder in IQ-TREE 1.6.12 (Kalyaanamoorthy et al., 2017).
A phylogenetic analysis based on these three genes is unlikely to provide strong support for the deep relationships between the genera of Epidendroideae, but tribe relationships in Epidendroideae have been studied several times using 74–78 plastid genes (Givnish et al., 2015; Li et al., 2019; Pérez-Escobar et al., 2021; Serna-Sánchez et al., 2021). Here we used the multispecies coalescent tree inferred from 78 plastid genes by Pérez-Escobar et al. (2021) as a backbone tree for tribe relationships. We used the classification of genus Orchidacearum (Pridgeon et al., 2005, 2009, 2014) to circumscribe tribes and assign genera to them, while correcting by the phylogenetic relationships of Pérez-Escobar et al. (2021). We hereafter used the supra-generic classification of Chase et al. (2015) for convenience, with the exception of the genus Coelia that we have kept in the subtribe Coeliinae (instead of Calypsoinae).
Dated tree inference from the molecular dataset was performed in BEAST 2.6.6 (Bouckaert et al., 2019), with the backbone tree input as multiple monophyletic constraints. The stem age of Earina and the crown age of Dendrobium (including Cadetia and Epigeneium) were calibrated with the fossils of Earina fouldenensis and Dendrobium winikaphyllum [both dated to 23.2 million years ago (Mya) (Conran et al., 2009)], respectively. The crown age of Epidendroideae was calibrated using the fossil of Succinanthera baltica estimated at least at 45 Mya (Poinar and Rasmussen, 2017). Despite having only been tentatively assigned to Epidendroideae but not to any extant tribe, this fossil allows to put a minimal age to the Epidendroideae subfamily. Calibration points were set with a log-normal distribution of mean = 1 and standard deviation (s.d.) = 1.25 for the two points at 23.2 Mya, and s.d. = 2 for the crown age of Epidendroideae. The higher s.d. on the calibration of the root of the Epidendroideae allows the node to take much older values, encompassing the uncertainty of both the age estimation and the phylogenetic position of Succinanthera baltica. We set a relaxed uncorrelated log-normal molecular clock, and the birth–death process as tree prior. A diffuse gamma distribution (α = 0.001 and β = 1000) was applied to both clock mean and birth rate priors. Other priors were left as default. Site model averaging using bModelTest 1.2.1 was performed, as recommended by Bouckaert and Drummond (2017), with mutation rate estimated and fixed mean substitution rate. We performed two coupled Markov chain Monte Carlo (MCMC) runs of 200 million steps each with eight chains, resampling every 20 000 steps. Stationarity and convergence of runs were assessed using Tracer 1.7.1, and trees were combined using LogCombiner 2.6.6 after discarding the first 12 % as burn-in. A maximum clade credibility (MCC) tree with median node heights was calculated using TreeAnnotator 2.6.6.
Lifestyle and water-related trait data set
To estimate their evolutionary history, we scored the occurrence of the epiphytic lifestyle and water-related traits (see below) at the generic level, using Genera Orchidacearum (Pridgeon et al., 2005, 2009, 2014). While many traits are related to water limitations, such as specific leaf area or water-use efficiency (e.g. Zhang et al., 2015; Hietz et al., 2022), only morphological water-related traits were available. We thus retrieved the presence of stem succulence [with two modalities if present: heteroblastic (one internode swollen) or homoblastic (several internodes swollen) pseudobulbs/corms]; the presence of leaf succulence (including coriaceous leaves); and the minimum and maximum number of velamen layers [also from Stern (2014) and Porembski and Barthlott (1988)] (Fig. 1). With regard to lifestyle, 35 genera included lithophytic species, but this lifestyle is sometimes equivocal (Zotz, 2013, 2016), and as none of the genera sampled were entirely lithophytic, we did not consider it. The genera were therefore assigned as epiphytic and/or terrestrial. With regard to leaf thickness, 14 genera (6.9 %) were leafless, i.e. mycoheterotrophic (Fig. 2). For all categorical traits, states were mutually non-exclusive, i.e. genera could include both epiphytic species and terrestrial species (see frequencies of each state in Supplementary Data Table S2). The number of velamen layers was missing for 66 genera out of 203 in the phylogeny (32.5 %).
Illustration of morpho-anatomical water-related traits (arrows) observed among epiphytic and terrestrial epidendroid orchids. (A) Homoblastic pseudobulbs (composed of several internodes) of the terrestrial Oeceoclades pulchra (Thouars) P.J.Cribb & M.A.Clem. (Cymbidieae, Eulophiinae). (B) Heteroblastic pseudobulbs of the epiphytic Bulbophyllum nutans (Lindl.) Rchb.f. species complex (Malaxideae, Dendrobiinae). (C) Succulent leaves of the epiphytic Holcoglossum pumilum (Hayata) L.J.Chen, X.J.Xiao & G.Q.Zhang (Vandeae, Aeridinae). Photograph by Sunoochi from Sapporo, Hokkaido, Japan, CC BY 2.0, https://commons.wikimedia.org/w/index.php?curid=83337314. (D) Multi-layered velamen; cross-section of a root of O. pulchra.
Evolution of lifestyle and succulent stems and leaves in epidendroid orchids. (A, B) Ancestral state estimation of lifestyles in the Epidendroideae genus tree with corHMM. Pie charts at nodes represent ancestral lifestyles and their probabilities under correlated evolution with leaf succulence. For succulent stems and leaves, only significant state changes, i.e. state probability >0.5 while <0.5 in the ancestral node, are indicated by coloured triangles (see Supplementary Data Figs S4A and S5 for full ancestral state estimations of stem and leaf succulence respectively). The character states of the present genera are represented on the right side of the tree by coloured boxes. In addition, genera with mycoheterotrophic species are indicated by a mushroom symbol. Non-monophyletic subtribes are indicated by stars after the subtribe name. (C, D) Diagrams of transition rates (in events/Myr) between lifestyles and presence/absence of (C) succulent stems or (D) leaves estimated under correlated evolution in corHMM with a symmetric and an all-rate-different transition matrix, respectively.
Test of correlated evolution between epiphytism and succulence traits
Because the epiphytic habitat is characterized by irregular water supply, an epiphytic lifestyle and succulent organs could have evolved in a correlated way. To test for their potential correlated evolution throughout the phylogeny, we used the discrete dependent/independent approach implemented in BayesTraits v.4.0.0. For a given trait, we compared the fit of a model where the epiphytic lifestyle and this trait evolved in a correlated way (‘dependent model’) to the fit of a model where both traits evolved independently (‘independent model’) using Bayes factors (BFs). We ran a reverse-jump (with an exponential prior of mean 10) MCMC analysis for each model, with 5 500 000 iterations and a burn-in of 500 000, and used the stepping stone sampler (Xie et al., 2011) with 500 stones and 5000 iterations per stone. Stationarity and convergence of two runs per model were assessed in Tracer v.1.7.1. The log BF was computed from the resulting marginal likelihoods, as 2 × (log marginal likelihood of the dependent model − log marginal likelihood of the independent model). We considered a BF > 2 to be significant support for the dependent model.
The test was replicated with another method, i.e. using the package corHMM v.2.7.1 in R v.4.2.1 (R Core Team, 2022), which is a Hidden Markov model designed to allow for the correlated evolution of several characters when estimating transition rates and inferring ancestral states on a phylogeny (Boyko and Beaulieu, 2021). Two symmetric transition matrices were used to correlate or decorrelate transition rates between states (Supplementary Data Table S3). The maximum likelihood of each model was computed with the corHMM function using 100 random restarts, and the best-fitting model was determined by AICc (corrected Akaike information criterion) comparison. Compared with BayesTraits, corHMM allows some heterogeneity in the transition rates across the whole phylogeny by assuming that the rates can depend on some unobserved (‘hidden’) traits.
Estimation of transitions between ancestral lifestyles and succulence traits
Freudenstein and Chase (2015) found that succulent stems could potentially have appeared earlier than the epiphytic lifestyle. To ascertain the likelihood of an appearance of succulence traits prior to the epiphytic lifestyle in a temporal framework, transition rates between states were first estimated with corHMM. Polymorphic taxa were included in the analysis, as corHMM can handle multiple states, although this is interpreted as uncertainty by the method rather than polymorphism. We tested models with transition matrices assuming either equal, symmetric or all-different transition rates, and including or not one hidden rate category. Then, 1000 stochastic character maps of the best-fitting model were generated using the makeSimmap function and then summarized to estimate the posterior probability of each state at ancestral nodes. The number of transitions between correlated states was estimated from the posterior probability distribution of trait changes from the stochastic character mapping.
Estimation of ancestral number of velamen layers and test of differences between lifestyles
For the mean number of velamen layers in the root, we considered the trait to be continuous. We used the R function phytools::anc.ML (Revell, 2012) with a Brownian motion model of continuous character evolution to estimate the ancestral states. The caper::pgls function was used to test if the mean number of velamen layers in epiphytic taxa is significantly different from that of terrestrial taxa, while accounting for the phylogenetic relatedness between species. As several genera are both epiphytic and terrestrial, we repeated the phylogenetic generalized least squares (PGLS) ten times, each time randomly assigning each polymorphic genus to be only epiphytic or terrestrial.
Ancestral ranges of the Epidendroideae
To address both the time periods and the biogeographical and palaeoclimatic context of the emergence of the epiphytic lifestyle, we estimated the ancestral biogeographical ranges of Epidendroideae on their time-calibrated MCC tree, using BioGeoBEARS (Matzke, 2013). Bioregions were defined as in Givnish et al. (2016): North America, Neotropics, Africa, Eurasia, Africa, Southeast Asia, Australia and Pacific. The models DEC*, DEC*+J, DIVA*, DIVA*+J, BayArea* and BayArea*+J (excluding a null range) were tested, with a maximal range size of 7, and thus 127 possible states. Time-stratified dispersal multipliers were set following Givnish et al. (2016) (Supplementary Data Table S4). The best-fitting model was chosen by AICc comparison (Table S5) and likelihood-ratio tests (Table S6).
RESULTS
Phylogenetic relationships and divergence times between Epidendroideae genera
All epidendroid subtribes sensuChase et al. (2015) were found to be monophyletic in the obtained phylogenetic tree of 203 genera of Epidendroideae, except for the subtribes Cymbidinae and Eulophiinae which were split into two and three clades, respectively. Posterior probability (PP) values and constrained nodes are presented in Supplementary Data Fig. S1. According to our time calibration based on three fossils (including a minimal age for the Epidendroideae based on the fossil described in Poinar and Rasmussen, 2017), the crown age of Epidendroideae was estimated at 47.4 Mya [95 % highest posterior density (95 % HPD) interval = 63.2–45.0 Mya]. Tribe crown ages and corresponding 95 % HPD intervals are detailed in Table S7.
Test of correlated evolution between epiphytism and succulence traits
The evolution of leaf succulence was found to be correlated with the lifestyle using both BayesTraits and corHMM (Table 1). The evolution of stem succulence was found to be correlated with the lifestyle with BayesTraits only, while corHMM returned a non-significant difference in AICc between the correlated and the uncorrelated models (Table 1). As both traits were significantly correlated using at least one method, we jointly estimated their evolutionary histories with the lifestyle (using correlated matrices of transition between states) in corHMM.
Tests of correlated evolution between the epidendroid orchid lifestyle (epiphytic/terrestrial) and stem (absence/presence) or leaf succulence (absence/presence) using two methods (BayesTraits and corHMM). The log marginal likelihood (LML) of the dependent or independent models, and the log Bayes factor (BF), were calculated using BayesTraits. We considered a BF > 2 to be significant support for the dependent (i.e. correlated evolution) model. Symmetric matrices were used in corHMM to correlate or decorrelate the two traits (see Supplementary Data Table S3 for more details on these matrices). A delta AICc between 0 and 2 indicates that the AICc values are not significantly different. A delta AICc > 2 indicates that the correlated model fits the data significantly better.
| BayesTraits | corHMM | |||||
|---|---|---|---|---|---|---|
| Tested correlation | LML of independent model | LML of dependent model | log BF | AICc correlated matrix | AICc uncorrelated matrix | delta AICc |
| Lifestyle~Stem succulence | −147.4 | −144.0 | 6.7 | 276.9 | 278.8 | 1.9 |
| Lifestyle~Leaf succulence | −183.1 | −172.8 | 20.7 | 324.4 | 344.5 | 20.2 |
| BayesTraits | corHMM | |||||
|---|---|---|---|---|---|---|
| Tested correlation | LML of independent model | LML of dependent model | log BF | AICc correlated matrix | AICc uncorrelated matrix | delta AICc |
| Lifestyle~Stem succulence | −147.4 | −144.0 | 6.7 | 276.9 | 278.8 | 1.9 |
| Lifestyle~Leaf succulence | −183.1 | −172.8 | 20.7 | 324.4 | 344.5 | 20.2 |
Tests of correlated evolution between the epidendroid orchid lifestyle (epiphytic/terrestrial) and stem (absence/presence) or leaf succulence (absence/presence) using two methods (BayesTraits and corHMM). The log marginal likelihood (LML) of the dependent or independent models, and the log Bayes factor (BF), were calculated using BayesTraits. We considered a BF > 2 to be significant support for the dependent (i.e. correlated evolution) model. Symmetric matrices were used in corHMM to correlate or decorrelate the two traits (see Supplementary Data Table S3 for more details on these matrices). A delta AICc between 0 and 2 indicates that the AICc values are not significantly different. A delta AICc > 2 indicates that the correlated model fits the data significantly better.
| BayesTraits | corHMM | |||||
|---|---|---|---|---|---|---|
| Tested correlation | LML of independent model | LML of dependent model | log BF | AICc correlated matrix | AICc uncorrelated matrix | delta AICc |
| Lifestyle~Stem succulence | −147.4 | −144.0 | 6.7 | 276.9 | 278.8 | 1.9 |
| Lifestyle~Leaf succulence | −183.1 | −172.8 | 20.7 | 324.4 | 344.5 | 20.2 |
| BayesTraits | corHMM | |||||
|---|---|---|---|---|---|---|
| Tested correlation | LML of independent model | LML of dependent model | log BF | AICc correlated matrix | AICc uncorrelated matrix | delta AICc |
| Lifestyle~Stem succulence | −147.4 | −144.0 | 6.7 | 276.9 | 278.8 | 1.9 |
| Lifestyle~Leaf succulence | −183.1 | −172.8 | 20.7 | 324.4 | 344.5 | 20.2 |
Gains and losses of the epiphytic lifestyle
The evolutionary history of the epiphytic lifestyle was consistent across corHMM models, i.e. either with models correlated to leaf (Fig. 2) or stem (Supplementary Data Fig. S2) succulence. The correlated model of evolution between the epiphytic lifestyle and leaf succulence (Fig. 2) inferred that the most recent common ancestor (MRCA) of Epidendroideae was probably terrestrial (P = 0.99) and ranged from the Neotropics to Southeast Asia and Australia (P = 0.58). In agreement with Freudenstein and Chase (2015), we found that the epiphytic lifestyle probably appeared at least three times independently in ancestral nodes, namely in (1) the MRCA of Malaxideae and Cymbidieae 39.0 Mya [P = 0.98, node PP = 1 (backbone node), 95 % HPD = 51.7–32.8 Mya] in Southeast Asia and Australia (P = 0.94, Fig. 3); (2) in the MRCA of Dendrochilum and Panisea in Arethuseae 11.5 Mya (P = 0.98, node PP = 0.98, 95 % HPD = 18.3–6.3 Mya) in Southeast Asia and Australia (P = 0.94, Fig. 3); and (3) in the MRCA of Sobralieae 7.1 Mya [P = 0.71, node PP = 1 (backbone node), 95 % HPD = 17.1–1.7 Mya] in the Neotropics (P = 0.98, Fig. 3).
Estimation of ancestral geographical ranges in relation to the evolution of epiphytism. Ancestral ranges were estimated in the dated phylogenetic tree of the 203 genera of epidendroid orchids using the BayAreaLike*+J model in BioGeoBEARS. The different geographical areas occupied by the present-day Epidendroideae genera are represented by unique colours. The pie charts at nodes represent the ancestral geographical ranges in proportion to their likelihood. These are solid-coloured when the range comprises a single area (e.g. Neotropics), or striped when the range comprises several geographical areas (e.g. a range represented by orange and light blue bands includes both Southeast Asia and Australia). The geographical ranges of extant genera are represented on the right-hand side of the tree by coloured boxes. Significant lifestyle changes, i.e. state probability >0.5 while <0.5 in the ancestral node, are indicated by coloured triangles (see Fig. 2 for full ancestral state estimation of lifestyles).
Likewise, we inferred at least five transitions from the epiphytic lifestyle to terrestriality (secondary terrestrialization) leading to diversification in (1) the MRCA of Oberonioides and Malaxis in Malaxideae 17.0 Mya (P = 0.89, node PP = 1, 95 % HPD = 25.9–10.1 Mya); (2) in the Collabieae excluding Eriodes 22.8 Mya (P = 0.97, node PP = 1, 95 % HPD = 32.9–14.5 Mya]; (3) in the MRCA of Calypsoeae 30.3 Mya (P = 0.83, 95 % HPD = 40.7–22.8 Mya); (4) in the MRCA of Bletiinae 12.4 Mya (P = 0.95, node PP = 1, 95 % HPD = 20.7–6.1 Mya] and (5) in the MRCA of Eulophia and Oeceoclades 12.6 Mya (P = 0.97, node PP = 0.83, 95 % HPD = 18.6–7.5 Mya).
Using the posterior probability distribution of trait changes from stochastic mapping, transitions from the terrestrial to the epiphytic lifestyles were estimated to have occurred three to seven times (95 % HPD interval) in the phylogeny with a median of four changes. Reverse transitions, i.e. from the epiphytic to the terrestrial lifestyle, were estimated to have occurred 10–16 times (95 % HPD interval) in the phylogeny, with a median of 13.
Gains and losses of drought-related traits (stem succulence, leaf succulence and velamen)
Ancestral state estimation of stem succulence was consistent either when considering the presence–absence of succulence (Supplementary Data Fig. S3) or when dividing the succulent stems into heteroblastic (one internode swollen) or homoblastic (several internodes in succession swollen) (Fig. 2; Fig. S4). The absence of stem succulence was inferred as ancestral in Epidendroideae (P = 0.80). Confirming the general assumption (Dressler, 1981; Ng and Hew, 2000), stem succulence was inferred to be ancestrally homoblastic, and to have appeared only once, in the MRCA of Nervilieae and Cymbidieae 43.1 Mya [P = 0.82, node PP = 1 (backbone node), 95 % HPD = 57.3–36.6 Mya] (Fig. 2; Fig. S4A) in the Neotropics, Southeast Asia and Australia (P = 0.45, Fig. 3) or in Southeast Asia and Australia only (P = 0.45, Fig. 3), thus pre-dating the epiphytic lifestyle by at least 4.1 Mya.
Heteroblastic succulent stems evolved repeatedly from homoblastic pseudobulbs at least eight times throughout the phylogeny (distribution of changes from stochastic mapping: 95 % HPD [9, 16], median = 12). Stem succulence appeared to have been lost several times. Transition rates (Supplementary Data Fig. S4B) indicated that homoblastic pseudobulbs tended to be either lost [0.02 events per million years (Myr)] or to evolve into heteroblastic pseudobulbs (0.011 events/Myr). Heteroblastic stem succulence derived predominantly from homoblasty (0.011 events/Myr) rather than from absence of succulence (3.5e-5 events/Myr) and was lost as often as it was acquired (0.013 events/Myr), though mostly at tips (Fig. S4).
Succulent leaves (Fig. 2; Supplementary Data Fig. S5) were inferred to have appeared several times convergently (distribution of changes from stochastic mapping: 95 % HPD [33, 52], median = 43 in epiphytic taxa; 95 % HPD [77, 112], median = 93 in terrestrial taxa), mostly at tips, and particularly during the Oligocene in epidendroid lineages that were probably already epiphytic at that time (i.e. in Podochileae, Epidendreae and Vandeae, and at least twice in Cymbidieae). However, even though succulent leaves evolved predominantly in epiphytic groups, some terrestrial taxa also evolved this trait, for example in the genera Oberonioides, Acrolophia or Cyrtopodium. In Vandeae, the appearance of leaf succulence 26.6 Mya (P = 0.76, node PP = 1, 95 % HPD = 36.5–19.3 Mya) was inferred to coincide with the loss of stem succulence (P = 0.86, node PP = 1, 95 % HPD = 36.5–19.3 Mya).
The number of velamen layers (Supplementary Data Fig. S6) tended to be ancestrally low in Epidendroideae, and seems to have increased in the MRCA of Panisea and Bulleyia in Arethuseae, in Dendrobiinae, in Collabieae, in Laeliinae and in most of Cymbidieae, as well as in Govenia (tribe Calypsoeae). In Vandeae, most genera are lacking data regarding velamen, thus inheriting an ancestral state without change, and hence some genera may actually have a high number of velamen layers in this tribe. As expected, primarily terrestrial tribes tended to have only a few velamen layers, but some secondary terrestrial genera, i.e. Govenia, Cyrtopodium and the subtribe Eulophiinae, have among the highest numbers of layers. All PGLS models indicated that terrestrial taxa tended to have fewer velamen layers than epiphytic taxa, but a high number of velamen layers was not significantly associated with the epiphytic lifestyle (range of F-statistic [2.2–0.099] on 1 and 135 d.f., P [0.14–0.75]).
Transition rates between correlated lifestyle and succulence traits
Using the PP distribution of trait changes from the stochastic character mapping, we estimated that transitions from terrestrial to epiphytic lifestyles only occurred in lineages with succulent stems [0.0087 events/Myr, Fig. 2C(i)], while such transitions were almost impossible in non-succulent lineages [1e-9 events/Myr, Fig. 2C(ii)]. Conversely, we reported that transitions from terrestrial to epiphytic lifestyles occurred in lineages without thickened leaves [0.0041 and 0.011 events/Myr, Fig. 2D(iii, iv)], with almost no transitions between terrestrial and epiphytic species with succulent leaves [1e-9 events/Myr, Fig. 2D(v)].
DISCUSSION
Where and when did the epiphytic lifestyle and succulence traits evolve?
The first occurrence of the epiphytic lifestyle in Epidendroideae was inferred at the end of the Eocene in a range encompassing the Neotropics, Southeast Asia and Australia (Figs 2 and 3). Over the course of the Eocene, the global climate gradually became cooler and drier (Bohaty et al., 2009; Bush et al., 2011), leading to a major retraction of megathermal forests (Bush et al., 2011) and to the development of new open and drier habitats (Bobe, 2006; Woodcock and Meyer, 2020). Even though geographical ranges were different between Aizoaceae and Epidendroideae, this origin coincides with the origin of Aizoaceae, an entirely African succulent group (41.5 Mya, 95 % HPD = 38.7–56.4 Mya; Arakaki et al., 2011; Klak et al., 2017). In several groups, succulence evolved in arid desert systems and in semi-arid systems such as savannas (Ringelberg et al., 2020; Anest et al., 2021). Likewise, stem succulence in Epidendroideae could have been an adaptation to drier, seasonal forests or savannas. With at least ancestral stem succulence (Fig. 2), epiphytism could have arisen in seasonally dry forests or dry microhabitats, perhaps primarily on cliffs and rocky areas as hypothesized by Dressler (1981), among taxa already adapted to drought. This conclusion is also consistent with Frenzke et al. (2016), who hypothesized that the terrestrial ancestor of epiphytic Peperomia (Piperaceae) already showed water-related traits potentially facilitating an epiphytic lifestyle such as succulence or CAM metabolism.
In Arethuseae, epiphytism probably appeared during the Miocene in the Southeast Asian and Australian range, when the climate was cooling again after the Mid-Miocene climatic optimum and moist megathermal forests were restricted to the tropical zone, with the exception of the Australasian region (Bush et al., 2011). In Southeast Asia, everwet climates predominated, but some regions could have been wet but seasonal, with open vegetation (Bush et al., 2011). Thus, it is unclear if epiphytic Arethuseae probably appeared in seasonal or everwet forests, and in addition the ancestral stem succulence could have facilitated the transition to the epiphytic habitat in either environment. Moreover, Coelogyninae are at present predominantly inhabiting everwet forests, and occur less frequently in areas with seasonal climates (Pridgeon et al., 2005).
Our interpretation that the epiphytic lifestyle could have evolved in seasonally dry forests generally contrasts with the presumed evolution of other epiphytic lineages, such as Bromeliaceae (Givnish et al., 2014), ferns (Schuettpelz and Pryer, 2009; Chen et al., 2022), Lycopodiaceae (Wikström et al., 1999) or thalloid liverworts (Bechteler et al., 2021), which were rather found to have originated mostly in rainforests or moist montane forests. In some of these other taxa, epiphytism is indeed unrelated to drought-tolerant traits, or drought-tolerant traits evolved as adaptations to the epiphytic lifestyle. Even among the Epidendroideae, the last appearance of the epiphytic lifestyle probably occurred in Sobralieae in the absence of any type of succulence 7.1 Mya (95 % HPD = 17.1–1.7 Mya), at the end of the Miocene, in the Neotropics. In tank bromeliads (Bromeliaceae), epiphytism evolved around 5.9 Mya in the late Miocene, in the Atlantic forest region of Neotropics, during a global cooling of the climate, synchronously with the uplift of the Serra do Mar which would have favoured cooler, rainier, more humid conditions in the Atlantic forest region (Givnish et al., 2014). Moreover, humid montane habitats were found to have favoured epiphytism in Bromeliaceae (Givnish et al., 2014). By the Miocene, mountain uplift had created abundant topographic relief across the Neotropics (Potter and Szatmari, 2009; Givnish et al., 2014; Martins et al., 2018). The lack of succulent organs in Sobralieae suggests that this tribe could have evolved in humid, montane habitats.
Nevertheless, the hypothesis that epiphytism in all Epidendroideae evolved in everwet forests should not be rejected: in Epidendroideae the pre-adaptation to drought would also have been beneficial in everwet habitats. Indeed, CAM photosynthesis, a water-conserving metabolism, could have been selected in high-rainfall habitats as has been observed for Bulbophyllum (Gamisch et al., 2021), and many species of Epidendroideae with succulent organs now grow in the shade of everwet forests (Pridgeon et al., 2005, 2009, 2014). On the other hand, in the genus Crassula (Crassulaceae) succulence is also found in species of mesic or wet environments, and has been suggested to be evolutionarily conserved after having been selected in ancestral dry microhabitats (Fradera-Soler et al., 2021).
Succulence traits as key innovations for the evolution of epiphytic epidendroid orchids
Stem succulence probably evolved prior to the epiphytic lifestyle, and epiphytes evolved far more frequently among lineages with thickened stems than among lineages without this feature (Fig. 2). It is likely that stem succulence emerged as an adaptation to water shortage in seasonally dry climates, and that it proved to be advantageous in an epiphytic environment. Moreover, even though stem succulence has sometimes been lost in epiphytic lineages, it was almost systematically offset by the appearance of leaf succulence, as in Vandeae or Pleurothallidinae for example (Fig. 2). We thus propose that stem succulence has been a key innovation [as defined by Miller et al. (2023): ‘an organismal feature that enables a species to occupy a previously inaccessible ecological state’] for the evolution of epiphytism in Epidendroideae.
The definition of a key innovation as stated by Miller et al. (2023) does not include an enhanced diversification, because, as they explain, ‘the expectation that key innovations should result in increased species richness or adaptive radiation is conceptually problematic [ … ] because it conflates two distinct evolutionary phenomena: diversification in species richness, and shifts in ecology’. Nevertheless, it would be interesting to test the impact of succulence traits on the diversification of epiphytes (Miller and Stroud, 2021), for example using trait-dependent diversification analyses such as Hidden State Speciation and Extinction models (Beaulieu and O’Meara, 2016; Herrera-Alsina et al., 2019; Nakov et al., 2019). However, to be accurate these analyses require robust species-level phylogenies, with a sampling fraction ideally >50 %, which is not the case of the phylogenetic tree we were able to produce until now. Nevertheless, Givnish et al. (2015) detected a shift towards higher diversification rates in the MRCA of Arethuseae and Cymbidieae, corresponding approximately to the first appearance of epiphytism, and subsequently including the majority of epiphytic Epidendroideae. Stem succulence, by facilitating the transition to the epiphytic lifestyle, would have indirectly enabled diversification bursts within the Epidendroideae.
In this study we focused on a few traits, but there are other traits that could also have significantly contributed to the evolution of epiphytism and the diversification of orchids, notably the water-conserving CAM photosynthesis (Silvera et al., 2009; Givnish et al., 2015; Silvera and Lasso, 2016) which is often associated with succulent organs (Niechayev et al., 2019). Even though recent studies did not find a link between CAM photosynthesis and an increase in diversification rates, at least in Bulbophyllum (Gamisch et al., 2021; Hu et al., 2022), an estimation of the evolution of CAM photosynthesis has been conducted by Silvera et al. (2009) at the subtribe and genus levels, then by Givnish et al. (2015) at the subtribe level, and in spite of high uncertainties in the ancestral state estimations and/or in tree topology, their results nevertheless indicate that CAM photosynthesis could also have appeared prior to epiphytism. In the future, further investigation of the evolutionary history of CAM metabolism in Epidendroideae could thus be interesting.
Secondary terrestrial genera mostly retained the epiphytic ancestral drought-related traits
Epiphytism has probably been lost multiple times during the Oligocene and the Miocene, in different geographical ranges (Figs 2 and 3). As reported by Chen et al. (2022) in the fern family Polypodiaceae, re-terrestrialization in Calypsoeae probably occurred during the Oligocene in Southeast Asia and the Neotropics, when glaciation led to a decrease of the area of broad-leaved forests and available habitats, probably increasing competition in upper canopies (Chen et al., 2022). Opening seasonal forests and savannas may have promoted the re-terrestrialization of epiphytes, by allowing more light to reach the floor than in broad-leaved forests, thus releasing competition for light in the understorey (Dressler, 1981; Wikström et al., 1999), and by allowing plants to occupy a wider range of habitats (Chen et al., 2022). In Collabieae, Malaxideae, Bletiinae and Eulophiinae, epiphytism was probably lost during the Miocene. This geological epoch was characterized by fluctuations in the global climate, with rapidly changing environments and temperatures, and by widespread topographic changes which created new habitats (Potter and Szatmari, 2009; Bush et al., 2011), enhancing the diversification of many lineages (Potter and Szatmari, 2009). Hence, as hypothesized by Chen et al. (2022) for Polypodiaceae, it is likely that new habitats created by climate and topographical changes could have favoured the re-terrestrialization and subsequent diversification of the secondary terrestrial clades in Epidendroideae, but also of the epiphytic taxa which seem to have diversified mostly during the Miocene (Supplementary Data Figs S7 and S8). In Eulophiinae, whose diversity is centred on Madagascar (Bone et al., 2015), re-terrestrialization occurred between 12 and 13 Mya, i.e. simultaneously with the appearance of cactiform stem succulence in the genus Euphorbia sections Goniostema, Denisophorbia and Deuterocalli, which represent about 70 % of the diversity of Euphorbia in Madagascar (Aubriot, 2012). Cactiform stem succulence has been found to prevail in areas with seasonal drought but a reliable season of precipitation (Evans et al., 2014), which means that seasonal dry forests may have been present in Madagascar at this time and thus could have been favourable to the re-terrestrialization of Eulophiinae.
If we examine the evolutionary history of succulence traits in Epidendroideae, it appears that, on the whole, secondary terrestrial taxa retained their ancestral succulent stems (Fig. 2). In Calypsoeae and Bletiinae, this stem succulence takes the form of underground corms, while in Malaxideae, Collabieae and Eulophiinae it is pseudobulbs (Pridgeon et al., 2005, 2009). However, pseudobulbs are sometimes more or less buried, for example in Collabieae genus Ipsea (Pridgeon et al., 2005; Descourvières, 2011), or in Eulophia graminea (Pemberton et al., 2008). In fact, burying succulent structures could better protect them from above-ground temperature extremes and drought, especially in seasonal climates (Pemberton et al., 2008; Kumar et al., 2022). In addition, Kumar et al. (2022) showed that some species in Malaxideae are secondary epiphytes, and that they also retained the perennating organ structure and leaf texture of their parent lineage. This tends to support the hypothesis that seasonal forests could have been the cradle to at least part of both epiphytes and terrestrial Epidendroideae diversity.
The current ecology of secondary terrestrials is also not exactly the same as that of primary terrestrial taxa. Indeed, unlike the primary terrestrial taxa, many secondary terrestrials are also found as lithophytes, or even as occasional epiphytes (Pridgeon et al., 2005, 2009, 2014). Even though there are differences between the lithophytic and epiphytic habitats, they still share similar rooting conditions (Zotz, 2016), and similar mycorrhizal fungal communities are shared by orchids between the two habitats (Xing et al., 2019; Qin et al., 2020). Moreover, lithophytic Malaxideae were not found on the rock itself but in the humus-rich and mossy substrate (Hermans et al., 2020), resembling some epiphytic conditions. Even among genera that were described as fully terrestrial in Genera Orchidacearum (Pridgeon et al., 2005, 2009, 2014), some may not always really root in the soil.
While epiphytes tend to have slightly more velamen layers than terrestrials, the re-terrestrialization did not particularly lead to a reduction of the number of velamen layers, and this number may have even increased in some secondary terrestrial clades (Supplementary Data Fig. S9). The velamen allows rapid uptake and long retention (increasing with velamen size) of water and nutrients, which is beneficial in environments with low and intermittent water and nutrient supply (Zotz and Winkler, 2013). Other important functions are mechanical protection and, in exposed roots, reduction of heat load (Zotz and Winkler, 2013; Zotz et al., 2017) and UV protection (Chomicki et al., 2015). Hence, velamentous roots may have been beneficial in secondary terrestrial taxa with a lithophytic-like ecology, whose roots can be exposed to a scarce supply of water and UV radiation as for epiphytic taxa. Furthermore, even rooted in soil, numerous taxa in Orchidaceae and other families are velamentous, as the velamen minimizes water loss in dry soils without damaging the root, unlike in other, non-velamentous, taxa (Zotz and Winkler, 2013; Zotz et al., 2017). Indeed, Zotz et al. (2017) found that terrestrial species were most prominent in seasonally dry habitats, which is consistent with our hypothesis of re-terrestrialization events in seasonally dry environments. In addition, Zotz et al. (2017) suggested that velamen pre-dated epiphytism, and thus could have been another key innovation (in addition to stem succulence as shown in this study) for the evolution of epiphytism, a view also supported by our results. This is also consistent with the hypothesis that epiphytism appeared in mostly seasonally dry climates.
Finally, considering that secondary terrestrials mostly retained their ancestral velamen and succulence traits, why are some of these taxa no longer found as epiphytes? The minute orchid seeds should not have constrained dispersal between ground and aerial environments, although different communities of suitable mycorrhizal fungi should (Martos et al., 2012). Indeed, given that orchids germinate with mycorrhizal fungi, we hypothesize that these taxa could no longer colonize the epiphytic habitat because of an evolutionary change in their association with mycorrhizal fungal symbionts, which are different between the ground and the trees (Martos et al., 2012). Mycorrhizal shifts are also suggested by the fact that all terrestrial lineages derived from epiphytic ancestors have evolved strategies of mycoheterotrophy by parasitizing soil-dwelling, ectomycorrhizal or saprotrophic fungi. Mycoheterotrophy is also often seen as an adaptation to shaded forest understorey (Martos et al., 2009). Likewise, even though light conditions of epiphytes span the entire gradient from deep shade to full radiation, many epiphytes may not tolerate shade, and low light in the understorey could thus promote epiphytism (Gravendeel et al., 2004; Zotz, 2013). Therefore, because most Epidendroideae taxa evolved epiphytism, and primary and secondary terrestrial taxa have often evolved mycoheterotrophy, access to light may have been a major constraint in the evolution of Epidendroideae, leading to the evolution of either epiphytism or mycoheterotrophy.
Assumptions and intrinsic limitations of methods
We identified two main methodological limitations intrinsic to the methods we used. First, even though our phylogeny is robust and congruent with previous studies based on a higher number of plastid markers, the divergence times were based on only three fossils (including a fossil with an uncertain phylogenetic position), while the Epidendroideae are very much diversified with no less than 23 246 species listed (WCSP, 2021). In addition, divergence time confidence intervals are large. Because we allowed the model to infer much older values than our fossil calibrations, such large intervals are partly artefacts of the probability distribution we used for calibration points combined with the relaxed molecular clock. Nevertheless, new orchid fossils would be beneficial to ascertain the divergence times of the subfamily.
Second, models of ancestral state estimation rely on several assumptions. For instance discrete state models assume that transition rates between states are constant throughout the phylogeny, even though hidden states in corHMM alleviate this assumption to some extent. Moreover, all transitions are likely to be slightly underestimated due to the processing of polymorphic traits as uncertainties and not as polymorphisms by corHMM; in other words, we only estimated transitions between sampled genera, but were not able to measure within-genus transitions.
CONCLUSION
Our analyses show that drought-related traits probably did not emerge as adaptations to the epiphytic lifestyle. Rather, epiphytes would have appeared among already drought-adapted terrestrial lineages in seasonally dry forests, possibly driven by light availability. In particular, ancestral stem succulence would have favoured the multiple colonization of aerial environments. This character would have evolved once in the form of a homoblastic pseudobulb, later evolving convergently towards heteroblastic pseudobulbs. When lost, stem succulence would have almost always been offset by leaf succulence, indicating that succulent organs could have actually been a key innovation for the emergence of the epiphytic lifestyle. Other physiological and morpho-anatomical traits, such as CAM photosynthesis, should be further investigated in combination with diversification analyses to further understand the macroevolutionary history of epiphytism in Epidendroideae.
SUPPLEMENTARY DATA
Supplementary data are available at Annals of Botany online and consist of the following.
Table S1: GenBank accession numbers of matK, psaB and rbcL sequences of 203 Epidendroideae genera. The three sequences for each genus originate either from the same voucher (‘voucher’, 69 % of genera), from different vouchers but same species (‘sp’, 14 %), or from different species (‘x’, 17 %). Sequences for the genus Angraecum were produced by Givnish et al. (2015) and are provided in their supplementary data, but were not available on GenBank. Table S2: Survey of morphological characteristics and lifestyle of the 203 genera of Epidendroideae sampled in the phylogeny, including two synonymized genera (Epigeneium and Cadetia, synonym: Dendrobium). Data compiled from Genera Orchidacearum (Pridgeon et al., 2005, 2009, 2014). Stem succulence includes corms and pseudobulbs. Leaf succulence includes leaves described as coriaceous or fleshy. Table S3: Symmetric (SYM) transition rate matrices between combined states used in corHMM for testing the correlation between lifestyle and binary morphological traits (stem and leaf succulence). Numbers indicate the transition rate category, with 0 meaning that transitions are not allowed between corresponding combined states. The decorrelated matrix allows movement from one state to the other inside a trait, whatever the state of the other trait. For example, rate category 1 allows transitions to occur at the same rate between the epiphytic and the terrestrial lifestyles, independently of the presence or absence of stem/leaf succulence. Table S4: Time-stratified dispersal multipliers between bioregions applied to BioGeoBEARS models, following Givnish et al. (2016). N: North America, T: Neotropics, E: Eurasia, F: Africa, S: Southeast Asia, A: Australia, P: Pacific. Table S5: AICc comparison of BioGeoBEARS models. d, e, j: values for dispersal, extinction and founder-event speciation parameters, respectively. delta AICc: difference of AICc between the model and the model with minimal AICc. AICc weight: proportion of the total predictive power of all models explained by the model. Table S6: Likelihood ratio-tests for nested BioGeoBEARS models. LnL alt, LnL null: log-likelihood of the data given the alternative or null model, respectively. DF alt, DF null: degrees of freedom of alternative and null models, respectively. p-value: p-value of the one-tailed chi-squared test. Table S7: Estimated crown age and 95 % highest posterior density (95 % HPD) interval of epidendroid orchid tribes (sensuChase et al., 2015) from our study versus Givnish et al. (2015) and Serna-Sánchez et al. (2021). Values from Serna-Sánchez et al. (2021) were retrieved from their relaxed molecular clock, birth–death divergence time analysis. Empty cells are tribes for which only one genus was sampled and therefore crown ages are not available. As Nervilieae are represented by only one genus in the three studies, their crown age is not available and therefore they are not included in the table. Our crown age estimates are within the 95 % HPD intervals found by Givnish et al. (2015) and Serna-Sánchez et al. (2021) for all orchid tribes, except for Neottieae, for which our age estimate is significantly younger than the 95 % HPD interval in Serna-Sánchez et al. (2021). Figure S1: Dated phylogenetic tree reconstructed from 203 genera in 14 tribes of epidendroid orchids. Branches with posterior probability <0.95 are coloured coral-red. Nodes that were constrained in the analysis are indicated by stars (in red when the quartet support for the main topology was <50 % in the coalescent-species tree of Pérez-Escobar et al., 2021). Figure S2: Ancestral state estimation of lifestyles in the Epidendroideae genus tree with corHMM. Pie charts at nodes represent ancestral lifestyles and their probabilities under correlated evolution with stem succulence. The character states of the present genera are represented on the right side of the tree by coloured boxes. In addition, genera with mycoheterotrophic species are indicated by a mushroom symbol. Non-monophyletic subtribes are indicated by stars after the subtribe name. Figure S3: Ancestral state estimation of the presence/absence of succulent stems in the Epidendroideae genus tree with corHMM. Pie charts at nodes represent ancestral presence and absence of succulent stems and their probabilities under correlated evolution with lifestyle. The character states of the present genera are represented on the right side of the tree by coloured boxes. In addition, genera with mycoheterotrophic species are indicated by a mushroom symbol. Non-monophyletic subtribes are indicated by stars after the subtribe name. Figure S4: Evolution of succulent stems in epidendroid orchids. (A) Ancestral state estimation of the forms of stem succulence in the Epidendroideae genus tree with corHMM. Pie charts at nodes represent ancestral stem succulence forms and their probabilities under correlated evolution with lifestyle. The character states of the present genera are represented on the right side of the tree by coloured boxes. In addition, genera with mycoheterotrophic species are indicated by a mushroom symbol. Non-monophyletic subtribes are indicated by stars after the subtribe name. (B) Diagram of transition rates (in events/Myr) between stem succulence forms estimated under correlated evolution in corHMM with an all-rate-different transition matrix. Figure S5: Ancestral state estimation of the presence/absence of succulent leaves in the Epidendroideae genus tree with corHMM. Pie charts at nodes represent ancestral presence and absence of succulent leaves and their probabilities under correlated evolution with lifestyle. The character states of the present genera are represented on the right side of the tree by coloured boxes. In addition, genera with mycoheterotrophic species are indicated by a mushroom symbol. Non-monophyletic subtribes are indicated by stars after the subtribe name. Figure S6: Ancestral state evolution of the mean number of velamen layers in the Epidendroideae genus tree with anc.ML. Missing data for the number of velamen layers are indicated by stars at tips before the genus names. Genera are missing data on the number of velamen layers inheriting the ancestral state without change, and hence some genera missing data may actually have a lower or a higher number of velamen layers. In addition, genera with mycoheterotrophic species are indicated by a mushroom symbol. Non-monophyletic subtribes are indicated by stars after the subtribe name. Significant lifestyle changes, i.e. state probability >0.5 while <0.5 in the ancestral node, are indicated by coloured triangles (see Fig. 2 for full ancestral state estimation of lifestyles). Figure S7: Lineage-through-time plot of terrestrial genera of Epidendroideae sampled in the phylogeny (75 genera). Figure S8: Lineage-through-time plot of epiphytic genera of Epidendroideae sampled in the phylogeny (153 genera).
ACKNOWLEDGEMENTS
We thank B. Bytebier and T. Stévart for constructive feedback on these results. We also thank the editors and the two anonymous reviewers for their constructive feedback.
