Niche conservatism and evolution of climatic tolerance in the Neotropical orchid genera Sobralia and Brasolia (Orchidaceae)

Sobralia and Brasolia form a large complex of Neotropical Orchidaceae. Although the molecular and morphological studies allowed to increase the rate of work on the modern classification of the taxa, they still require the attention as remaining without complete revision. The niche similarity analysis between representatives of Sobralia and recently segregated from this taxon—genus Brasolia is presented. The ecological tolerance evolution within the group was investigated with molecular clock analysis and phylogeny as the background. The phylogenetic analysis has confirmed the previous results and placed Brasolia representatives in a single clade with Elleanthus and Sobralia core as a separated group. The molecular clock analysis suggests that Sobralia and Brasolia are relatively young groups that evolved between 8.5 and 8 million years ago. Distribution of suitable niches of studied species is generally congruent with the known geographical ranges of particular taxa. The calculated niche overlap did not indicate any correlation between niche overlap and species phylogenetic relationships and remains low for both intra- and intergeneric relationships. The reconstruction of climatic tolerance evolution indicated that the studied species of Brasolia and Sobralia are characterized by generally similar ecological tolerance for most of the analyzed variables.

The importance of ecology in speciation has been discussed since Darwin's 1 theory on the evolution of species. The divergent selection of ecologically affected characters can lead to the reproductive isolation of populations inhabiting separated niches. Grafen 2 suggested that niche similarity of closely related species is a result of common ancestry, as well as their shared environmental preferences. The recent advancements in the studies on ecological niches prompted scientists to intensify research on the relationship between fundamental niches and the evolution of various organisms [3][4][5] . Substantial dispute has emerged in the literature about phylogenetic niche conservatism (PNC), which was defined as the tendency of lineages to retain their ancestral niche-related traits through speciation events [6][7][8] . Although numerous scientists consider PNC pattern to be common in nature 9 , several case studies indicated that ecological and phylogenetic similarities often are not correlated 10,11 .
Without a doubt, the currently observed diversity of numerous orchids is the result of rapid radiation, which supposedly was the result of their adaptation to various pollinators [12][13][14] . However, the importance of environmental factors in the diversification of Orchidaceae has been poorly studied 15 .
In this paper, the niche similarity between representatives of Sobralia Ruiz & Pav. and recently segregated from this taxon, genus Brasolia (Rchb.f.) Baranow, Dudek & Szlach. was evaluated. Moreover, the evolution of ecological tolerances within these orchids was reconstructed. To analyze the impact of preferred ecological niches on the speciation and evolution of Sobralia and Brasolia, we conducted both phylogenetic study and ecological niche modeling.
The studied group consists of about 200 species with characteristic reed-like stems reaching from over a dozen centimeters to more than 10 m high, often forming dense clumps. The position and structure of the inflorescence www.nature.com/scientificreports/ are diversified. Along with the other morphological characters and the phylogenetic studies results, it was the argument for proposition of division Sobralia into two genera with Brasolia as a group with branching inflorescences and Sobralia with unbranched, often condensed inflorescences 16,17 . The flowers of Brasolia are often somewhat fleshy, persistent, often with contrasting colored fimbriate keels on the labellum, opening simultaneously or successively but always more than one at a time on the inflorescence and with developing floral buds visible. Flowers of Sobralia are more delicate, in most cases ephemeral and short-lasting, usually with two keels or ridges in the throat, sometimes with some other projections on the disc. Only species of the section Racemosae Brieger have elongated inflorescence with visible internodes. In this section, flowers at various stages of development can be observed at a time. The remaining species have condensed, cone-like inflorescences with imbricating floral bracts and usually one, sometimes two flowers at a time. The floral buds appear after the previous flowers fall off. The group with abbreviated inflorescences is the largest one and it is divided into few sections-the nominal one is represented by often tall, reed-like plants with shortened inflorescences covered by tightly imbricating floral bracts forming a cone-like structure. Their flowers are comparatively large as for the studied group, delicate, thin-textured, usually short-lasting, often one day only. The section Globosae Brieger is defined as a group of species with narrow, acuminate leaves, narrow base of the column and the cluster of floral bracts increasing in length as flowers are produced reaching 3-4 cm in length. The section Intermediae Brieger groups a complex of species with small flowers and inflorescences. Apart from the sections, there are few species or groups of species without the sectional membership. Sobralia macrophylla Rchb.f. and its allies are characterized by the clustered floral bracts hidden between the leaf-like bracts forming a funnel. Sobralia undatocarinata C. Schweinf. group is characterized by loose bract clusters and 2-5 relatively durable flowers produced simultaneously. The description of variety within Sobralia, requires also a mention of S. amabilis (Rchb. F.) L. O. Williams and S. callosa, which differ from the members of the nominal section by inconspicuous lip throat and the column structure 18 . The studied plants grow in various habitats as epiphytic, terrestrial or lithophytic herbs at various elevations in Central and South America 19 . Their usually relatively large flowers are pollinated by bees or hummingbirds 20 .
The previous studies concerning the phylogeny of the group didn't discuss the relations within the group and ecological aspects at the same time. The studied genera are object of taxonomical studies, however they still need more attention. The first steps to resolve the polyphyly of Sobralia have been made, and as a result, Brasolia was separated from Sobralia 17 . However, as it is known, with more data more accuracy can be achieved, we decided to expand the study with the ecological aspects analyses. The hypothesis was that Sobralia as a genus with much wider occurrence than Brasolia will also have wider potential ranges. Brasolia species, which mostly occur in Andean localities, are not so flexible as Sobralia-a genus with its representatives at much wider range of elevation and hence, in much more types of habitats. The same assumption was made for the ecological tolerance evolution analyses.

Results
Phylogenetic relationships. On the phylogenetic tree including the molecular clock analysis results The molecular clock analysis suggests ( Fig. 1), that Sobralia and Brasolia are relatively young groups that evolved between 8.5 and 8 Mya (Million years ago) together with Elleanthus forming a sister group for the clades composed by the two genera. The most intensive diversification of Sobralia and Brasolia has started about 3Mya. Their ancestor has diversified into two groups between 6.5 and 6 Mya.
Ecological niche modeling, limiting factors and niche overlap. All models received high AUC scores of 0.955-1.000 (Table S2). Mapped distribution of suitable niches of studied species created based on current climatic conditions is generally congruent with the known geographical ranges of particular taxa (Annex 1-6). Singular localities of few species (S. crocea, S. decora, S. fragrans, S. liliastrum, S. lindleyana, S. powellii, S. rosea, S. sessilis, S. valida) are outside of the potential range as calculated in MaxEnt.
In some cases, the suitable niche distribution overlaps with the known geographical range of these orchids or the calculated potential range is only slightly broader than known distribution of the plants (B. dichotoma, B. speciosa, S. crocea, S. leucoxantha, S. fragrans, S. lindleyana, S. luteola, S. valida, S. wisoniana and E. lancifolius). In other species, the calculated distribution of suitable niches is broader than recognized geographical range.
The distribution of most species is limited mainly by the annual mean temperature (bio1, 16 species), and the annual precipitation (bio12, 10 species). The range of four species is restricted by the precipitation in the driest month (bio14) and the equal number of taxa depends on the precipitation in the warmest quarter (bio18) and the precipitation of the coldest quarter (bio19). The occurrence of Sobralia pulcherrima and Brasolia ciliata is limited by the mean diurnal range (bio2). The temperature seasonality (bio4) is crucial for occurrence of Brasolia speciosa, Sobralia gloriosa, and Sobralia lancea. The presence of Sobralia macrantha depends on the precipitation seasonality (bio15). www.nature.com/scientificreports/   46 . Boxplots show the pairwise niche overlap between species belonging to the following groups: within each genus (Brasolia, Elleanthus, Sobralia) and between genera (Brasolia and Elleanthus, Brasolia and Sobralia, Sobralia and Elleanthus). The niche overlap value for each pair of species within a given group is represented by a dot. www.nature.com/scientificreports/ The calculated niche overlap (Fig. 2, Tables S2 and S3) did not indicate any correlation between niche overlap and species phylogenetic relationships and remains low for both intra-and intergeneric relationships. This is visible especially within Brasolia, whereas in the more species-rich genus Sobralia the general overlap is low but there are both species with very low overlap and some that reach values around 0.8 for both D and I indices. Concerning intergeneric relationships, the lowest overlap may be observed between Brasolia and Elleanthus. Many species belonging to Brasolia and Sobralia also exhibit a low overlap but this is not true for all pairs of species, as some receive overlap around 0.5 for both D and I indices. Niche overlap between Sobralia and Elleanthus is generally shifted towards more overlapping niches than between other pairs but still the values are low and their mean oscillates around 0.2 for D statistic and around 0.3 for I statistic.
Within the studied Brasolia species, the highest overlap is observed between the pairs B. cattleya-B. ciliata Ancestral climatic suitability and evolution of climatic tolerance. The reconstruction of climatic tolerance evolution indicated that the studied species of Brasolia and Sobralia are characterized by generally similar ecological tolerance for most of the analyzed variables and that their ecological divergence began c. 42.87 Kya ago. However, we did not find any correlation between the phylogeny and the ecological tolerance-in most cases, the closely related taxa are not characterized by more similar niches than more distinct species. However, both included Elleanthus species seem to have similar tolerance which may be different than in most of Brasolia and Sobralia species (bio4). It may suggest that the extended sampling could bring the new conclusions concerning the ecological tolerance-phylogeny relation. Noteworthy, most species share at least a part of climatic niches and we did not identify any species which adapted to completely separated values of analyzed variables (Figs. 3, 4).
The first conclusion is that some If considering the specific ecological factors, it was noticed that some of them are more limited for the studied species than the others. For temperature seasonality (bio4), almost all the studied species indicate the narrow ecological tolerance. On the other hand, considering annual precipitation (bio12), broad spectrum of tolerance for almost all taxa is observed.
It is worthy to emphasize the topology of the diversification of the studied group in order to the time scale and the phylogenetic relations.

Discussion
Our study indicated strong niche conservatism in Sobralia and its relatives. The representatives of various phylogenetic groups are limited by similar climatic factors and apparently their evolution was not directly related to the adaptation to various climatic variables. According to our analyses, the most closely related species do not share the most similar niches, whereas the general ecological tolerance in Sobralia and Brasolia is relatively narrow considering the tested environmental variables.
Out of the possible radiation triggers pointed by Dodson 21 , the most probable reason of diversification of Sobralia and its relatives was the adaptation to various pollinators. Flowers of Sobralia and Brasolia were reported to be pollinated by a variety of large solitary bees, especially by euglossine bees 22 . However, in some species (e.g. S. amabilis and S. callosa) pollen is transferred by hummingbirds 18 . The previously published data suggested that when separated by putative pollinator, different groups of Sobralieae showed affinities to the elevational preference of pollinators 23 .
The phylogeny based on the set of taxa used in the analyses (Fig. 1) confirmed the previous phylogenetic study devoted to the discussed taxa relationships 17,24 .
It is worthy to mention, that the molecular clock created for Sobralieae, indicated that the highest rate of diversification occurred in this group recently, between 2 Mya (Million years ago) and the present time. If compared with the phylogeny of euglossine bees 25 , the speciation within Sobralia was proceeded by the intensive radiation of the Euglossini. The diversification of insects began in the end of the Miocene and reached the highest rate in the mid-Pliocene, between ca. 5.5 and 2 Mya 25 . In our opinion these two events could be correlated and the diversification of Sobralieae was probably triggered by the increase of pollinators' diversity. However, to confirm such hypothesis a broader set of observations on Brasolia and Sobralia pollinators would be required.
In the case of the studied orchid group, epiphytism is not the widespread growth system and the majority of its species grows on the ground, hereby the adaptation to various microhabitats in trees 21 can be rejected as a radiation trigger. Moreover, Neubig 23 did not find any one-to-one mimicry related to the food deception within Sobralia, so this character should not be considered as related with rapid diversification of the studied orchids.
The associations of Sobralia and its relatives with mycorrhizal fungi have not been explored so far. However, considering the variety of environments where the studied orchids occur, including terrestrial and epiphytic www.nature.com/scientificreports/ habitat, as well as different types of soils, we suppose that the specialization of mycorrhizal partner did not prompt speciation events. Sobralia and its relatives can occur in both lowlands and mountain habitats. However, representatives of this group are most diverse in the montane regions of Neotropics. We believe that the orogeny of cordilleras of Central America and the Andes prompted radiation of pollinators and this event was directly correlated with evolution of pollinator specificity and diversification of Sobralia. For example, the intensive uplift of the Eastern Cordillera of the Colombian Andes occurred relatively recently 26 . By 4 Mya, this mountain range reached no more than 40% of its current elevation 26 . The surface uplift of 1400 and 1700 m in the northern Altiplano occurred between 18.7 and 5 Mya 27 . Moreover, several contractional deformations in Central America were dated as quite young 28 . For example, the age of Fila Costeña and Cordillera de Talamanca was estimated to 15-7 Mya and the age of Cordillera Central to 11-7.3 Mya 27 . Undoubtedly, considering the current distribution of some species, e.g. S. rosea, the Andes could serve as a North-South corridor for migration of some orchids.
According to our analyses, the studied group is relatively young and its highest rate of speciation began ca. 3 Mya. However, the intensive diversification of ecological niche tolerance in Sobralia and Brasolia started around 50 Kya. This recent evolution of preferred climatic niches was also indicated in some previous study on Orchidaceae 29 . Apparently, the divergence of ecological niche tolerance after the great species diversification is a repeatable phenomenon in tropical and subtropical orchids, whose evolution was driven mainly by adaptation to various pollinators. Since climate adaptation has significant consequences in the evolution and ecology of all living organisms, it would be interesting to evaluate the PNC in Sobralia and Brasolia pollen vectors and compare it with the diversification of climatic niches of orchids. Arnal et al. 30 proved that in host-specialized phytophagous insects, host relationships not always explain niche preferences of insects. So far, no study on the overlap of niches of tropical orchids and their pollinators was conducted and the correlation of the evolution of www.nature.com/scientificreports/ their ecological tolerances remains unexplored. Unfortunately, this assessment is currently not possible due to the scarce information about species pollinating studied plants. While all models of suitable niche distribution created for studied orchids received high scores of reliability tests, the potential ranges of some species are clearly broader than the registered geographical distribution of these taxa. Predictions of suitable climatic niches outside of environmental ranges are not uncommon in species which occurrence is related with numerous ecological constrains 31 and which dispersal is limited by various geographical and biotic barriers 32,33 . Detailed information about availability of specific mycorrhizal fungi and pollinators would have to be used to receive more precise models of potential distribution of the species studied. Unfortunately, the ecology of Sobralia and its relatives is so far poorly recognized and the information about its pollen vectors and symbiotic organisms is extremely limited. Here we evaluated exclusively the evolution of climatic niche of Sobralia and its relatives, not their realized niches which are also defined by the biotic interactions 34,35 .  Herzog 311 (B n.v.).

Taxonomic note
Notes. Although its morphology allows to classify it within Brasolia without doubt, the transfer was missed in the paper elevating Brasolia as a separate genus 17 . Thus, we propose the transfer here to keep all the species of the group within one genus.  (Table S1). To evaluate the ancestral climatic tolerance of the studied group, which is based on outcomes of phylogenetic research also information about distribution of two related species, Elleanthus ampliflorus Schltr. and E. lancifolius C. Presl, were gathered. The occurrence data were obtained from the information recorded with the specimens deposited in the herbaria AMES ( The georeferencing process followed Hijmans et al. 36 . The geographic coordinates provided on the herbarium sheet labels were verified. If there were no geographic coordinates on the herbarium label, we used the description of the locality where the plant was collected and assigned coordinates as precisely as possible to this location. The Google Earth (ver. 6.1.0.5001, Google Inc.) application was used to validate all the information gathered. A total of 786 localities that could be precisely located were used in the ecological niche modeling (ENM) (5-46 localities per species; supplementary material Table S1).
Ecological niche modelling and niche similarity. The ecological niche modelling was done using the maximum entropy method in MaxEnt version 3.4.3 37-39 based on presence-only observations of studied species. The duplicate presence records (records within the same grid cell) were removed using MaxEnt.
For the modelling, bioclimatic variables in 2.5 arc-minutes of interpolated climate surface were used. This approach was justified considering the precision of georeferenced records derived from information provided on herbarium material labels. Because some previous studies 40 indicated that usage of a restricted area in ENM analysis is more reliable than calculating habitat suitability on the global scale, the area of the analysis was restricted to 60°S-35.75°N-126.916°W-31.8749°W.
In this study, as in many other ecological studies, the most widespread source of climatic data was used. WorldClim 41 is commonly applied to produce species distribution models (> 15,000 citations). Of 19 bioclimatic variables ("bioclims", Table 1) available in WorldClim (version 2.10 42 some were removed because they were significantly correlated with one another (above 0.9) as evaluated by Pearsons' correlation coefficient computed using SDMtoolbox 2.3 for ArcGIS 43 . While the strategy of excluding highly intercorrelated variables has little influence on models derived from MaxEnt according to Feng et al. 44 , because of the reduction of multi-collinearity the following variables were excluded from further analyses: bio3, bio5, bio6, bio7, bio8, bio9, bio10, bio11, bio13, bio17 and bio18.
In all analyses, the maximum number of iterations was set to 10,000 and convergence threshold to 0.00001. The neutral (= 1) regularization multipler value and auto features were used. All samples were added to the background. The "random seed" option which provided a random test partition and background subset for www.nature.com/scientificreports/ each run was applied. The run was performed as a bootstrap, which is highly recommended for small-sample analyses, with 1000 replicates, and the output was set to logistic. The evaluation of the created models was made using the most common metric-the area under the curve (AUC) [45][46][47] . All operations on GIS data were carried out on ArcGis 10.6 (Esri, Redlands, CA, USA) and R 48 . The quantification of niche overlap between each pair of generated ENMs was done using the ENMTools package for R 48,49 utilizing two statistics-Schoener's D 50 and the I statistic 51 . Calculations were done according to the method proposed by Broennimann et al. 52 , which compares the similarity between two species in an environmental space that is represented by the first and the second axis of the principal component analysis (PCA) computed on bioclimatic variables.
To reconstruct ancestral ecological tolerances and predicted niche occupancy profiles (PNOs), the Phyloclim package was used 53 . This approach implements the methodology originally developed by Evans et al. 54 . PNO profile takes into account species probability of occurrence derived from ecological niche modeling and compiles a response to a particular environmental variable for each species. Ancestral ecological tolerances were computed from the phylogenetic tree and PNO using nonparametric approach and ancestral character estimation 54,55 . Phylogenetic analysis. The DNA sequences of ITS region used for molecular analyses were taken from Baranow et al. 17 and Neubig et al. 24 . They were aligned by Seaview 56 using MUSCLE algorithm 57 . The complete matrix consisted of 27 taxa representing 21 species of Sobralia, 4 species of Brasolia and 2 species of Elleanthus C. Presl. Data matrix is available from the corresponding author upon request. Gaps were treated as missing data. The molecular substitution model was calculated using MrModeltest 2.2 58 . In both cases, hLRTs and AIC as the best fitting were selected General Time Reversible with gamma distribution and invariable sites (GTR + G + I). For reconstruction of phylogeny we were using MrBayes 3.1.2 59 . The phylogenetic trees were constructed by applying the Bayesian inference using Markov chain Monte Carlo simulations. Four chains were run for 10,000,000 generations of two independent and simultaneous runs. The trees were checked for stability, until the average standard deviation of split ranges was smaller than 0.01. The first 25% trees were discarded as the burn-in. The remaining trees were used to construct a maximum clade credibility tree by using TreeAnnotator v1. 8.1 60 .
To estimate divergence times for representatives of Sobralia and Brasolia, molecular clock analysis was performed using BEAST 1.8.3 60 . First, based on the nuclear ITS sequences available in GenBank and Baranow et al. 17 , a matrix was prepared for all representatives of Sobralieae Pfitzer with Neottieae Lindl. as an outgroup. The calibration point was chosen according to Givnish et al. 61 as the time of the last common ancestor for all Epidendroideae (48.05 Mya years ago). Divergence times for representatives of Sobralieae were estimated with a log normal relaxed molecular clock using the Yule model of speciation. Two independent runs, each with 20,000,000 generations of Markov chain Monte Carlo (MCMC) were sampled every 1000 generations. The MCMC output files were probed by Tracer v1.6. The consensus tree files from two runs were obtained with LogCombiner v1.8.1 with a burn-in of 25% of the trees, then the final maximum clade credibility tree was summarized by TreeAnnotator v1.8.1 60 . The obtained results (supplementary material Fig. S1) served us as source of the calibration points to determine the exact divergence times of representatives of Sobralia and Brasolia. The time of the last common ancestor of Sobralieae tribe was determined to 12.96 Mya years ago (10.60 as 95% HPD lower and 16.50 as 95% HPD upper). The two additional points indicated the time of diversification of Sobralia (8.70 Mya) and others member of the tribe (10.98 Mya). We performed the analysis with a log normal relaxed molecular clock using the Yule model of speciation. Two independent runs, each with 10,000,000 generations of Markov chain Monte Carlo (MCMC) were sampled every 1000 generations. The MCMC output files were probed by Tracer v1.6. The consensus tree files from two runs were obtained with LogCombiner v1.8.1 with a burn-in of 25% of the trees, then the final maximum clade credibility tree was summarized by TreeAnnotator v1.8.1 60 .

Data availability
The herbarium specimens list used as the samples for the localities dataset is attached as a supplementary material. All the DNA sequences used for phylogenetic study were taken from previously published results and are available as the cited authors indicated in their works. The alignments used for the presented analysis are available from the corresponding author upon request.