Phylogenetics of Tribe Collabieae (Orchidaceae, Epidendroideae) Based on Four Chloroplast Genes with Morphological Appraisal

Collabieae (Orchidaceae) is a long neglected tribe with confusing tribal and generic delimitation and little-understood phylogenetic relationships. Using plastid matK, psaB, rbcL, and trnH-psbA DNA sequences and morphological evidence, the phylogenetic relationships within the tribe Collabieae were assessed as a basis for revising their tribal and generic delimitation. Collabieae (including the previously misplaced mycoheterotrophic Risleya) is supported as monophyletic and nested within a superclade that also includes Epidendreae, Podochileae, Cymbidieae and Vandeae. Risleya is nested in Collabiinae and sister to Chrysoglossum, a relationship which, despite their great vegetative differences, is supported by floral characters. Ania is a distinct genus supported by both morphological and molecular evidence, while redefined Tainia includes Nephelaphyllum and Mischobulbum. Calanthe is paraphyletic and consists four clades; the genera Gastrorchis, Phaius and Cephalantheropsis should be subsumed within Calanthe. Calanthe sect. Ghiesbreghtia is nested within sect. Calanthe, to which the disputed Calanthe delavayi belongs as well. Our results indicate that, in Collabieae, habit evolved from being epiphytic to terrestrial.


Background
The family of Orchidaceae is one of largest among seed plants, and several classification systems have been proposed to reflect their phylogeny and evolution (such as [1][2][3][4]). Recent molecular phylogenetic studies have shed new light on the relationships of Orchidaceae from genera to subfamilies (such as [5][6][7][8][9][10][11][12][13][14][15][16]). In Collabieae, as in many orchids, relationships have been traditionally defined based on subjective assessments of morphological characters, and phylogenetic relationships remain to be assessed using molecular data.
Recent results of cladistic analyses of combined DNA sequences have provided some new insights in the systematics of tribe Collabieae. Based on matK and rbcL, Goldman et al. [25] proposed that Mischobulbum, Nephelaphyllum, and Tainia should be transferred from Bletiinae (in the tribe Arethuseae) to the non-Arethuseae subtribe Collabiinae. Likewise, based on their analyses of ITS, matK, and trnL-F, van den Berg et al. [26] suggested that Collabiinae and Phajinae should be transferred to tribe Collabieae. Chase et al. [6] and Pridgeon et al. [19] tentatively redefined Collabieae to include 18-19 genera, pending future studies.
As currently defined, Collabieae include Acanthephippium, Ancistrochilus, Ania, Calanthe, Cephalantheropsis, Collabium, Diglyphosa, Eriodes, Gastrorchis, Hancockia, Ipsea, Mischobulbum, Nephelaphyllum, Pachystoma, Phaius, Pilophyllum, Plocoglottis, Spathoglottis, and Tainia [19] and shows a variety of vegetative and floral variation, such as plants with corms or pseudobulbs of one to several internodes or without storage organs; leaves petiolate or not and conduplicate or convolute; inflorescences lateral or terminal; pollinia varying from 2 through 4 to 8, and being soft or hard; viscidium present or absent. This diversity has led to difficulties in the circumscription of the tribe, as well as its subtribes and genera, and made problematic to infer the systematic position of the tribe among other Orchidaceae [19,27]. Moreover, previous molecular systematic studies sampled Collabieae only superficially, with just 7 species in 6 genera in van den Berg et al. [26] and 10 species in 8 genera in Goldman et al. [25]. Due to the sparse sampling and/or weak support for Collabieae in previous molecular systematic studies, subtribal and generic delimitation, as well as the phylogenetic position of Collabieae within Orchidaceae, remain unresolved.
Here we use DNA sequences of chloroplast genes rbcL and psaB, pseudogene matK, and the trnH-psbA region, and a broad sample of taxa across the Collabieae and Orchidaceae in order to: i) clarify the circumscription and systematic position of tribe Collabieae in Orchidaceae; ii) infer phylogenetic relationships within Collabieae; iii) elucidate the delimitation of several debatable genera.

Sequences characteristics
In this study, 45 DNA sequences of rbcL, 45 of matK, 33 of psaB and 35 of trnH-psbA were newly obtained. For each of the regions studied, aligned sequence lengths and other parsimony-related information are given in Table 1. The subfamily-wide matrix comprised 4674 aligned nucleotides of three chloroplast markers combined: rbcL (1362 bp), psaB (1666 bp), and matK (1646 bp).
In the reduced matrix, the aligned rbcL was 1343 bp in length, psaB was 1666 bp in length; both were without indels. The matK pseudogene was 1846 bp in length with 10 indels from 1 bp to 33 bp. The aligned trnH-psbA region was 1919 bp in length among Collabieae. We tried to amplify trnH-psbA from Risleya atropurpurea, but failed. It is possible that this region is lacking in R. atropurpurea. The combined dataset of four chloroplast markers was 6775 bp, and more than 4.9% of the characters were parsimony-informative (Table 1).

Subfamily-wide analysis of Epidendroideae
For the first analysis using the subfamily-wide matrix, under the Bayesian criterion, selection of a partition scheme based on BF favored the P5 partition scheme (see Table 2). The trees generated by BI were congruent with those of MP analysis except for poorly supported nodes along the backbone of the tree (Fig. 1, the MP strict consensus tree is not shown). Risleya atropurpurea is identified as a member of tribe Collabieae, not a member of tribe Malaxideae as previously thought [19].

Analysis of Collabieae
For the second analysis with the reduced matrix, P6 was selected as the best-fit partition scheme under the Bayesian criterion (see Table 2). The BI analysis yielded trees with topologies that were consistent with those retrieved by the MP analysis except collapsed nodes (Fig. 2). The parsimony analysis generated 7055 MPTs of 1,175 steps, with a CI of 0.702 and a RI of 0.739. The currently defined Collabieae is subdivided into three clades, of which clade III includes only the monotypic genus Eriodes and is sister to two other clades (BS = 77, PP = 1.00).
Clade II includes most sampled genera of Phajinae, and is subdivided into five subclades ( Figure 2): subclade V includes Calanthe sect. Calanthe, and C. sect. Ghiesbreghtia with strong support (BP = 96, PP = 1.00); subclade VI consists of C. sect. Styloglossum and the genus Cephalantheropsis with strong support (BS = 100, PP = 1.00); subclade VII includes Phaius without support; subclade VIII includes two species of Calanthe subgenus Preptanthe with strong support (BP = 99, PP = 1.00), and is sister to subclades V, VI and VII (BP = 74, PP = 1.00). Cladograms from the morphological data based on BI and MP analyses were poorly resolved ( Figure S1). The MP analysis of the morphological data found 51284 equally-shorter trees with a length of 107 steps, a CI of 0.355 and a RI of 0.725. The Table 2. Summary of Bayesian analyses and 2ln Bayes factor comparisons of partitioning strategies.

Delimitation of Collabieae
Our results indicate that, as currently delimited, Collabieae, consisting of Collabiinae, Phajinae, and Eriodes, is paraphyletic, with Risleya having been misplaced in Malaxideae [19]. A redefined Collabieae, including Risleya, is strongly supported as monophyletic ( Figure 1) and nested within a superclade consisting of Cymbidieae, Epidendreae, Podochileae and Vandeae in Epidendroideae (PP = 0.82). These results differ greatly from most previous taxonomic treatments about Collabieae (such as [3,4]) but agree well with the suggestions of van den Berg et al. [26] and the tribal circumscription by Pridgeon et al. [19].
The monotypic genus Risleya likely has been misunderstood due to its mycoheterotrophic habit and greatly reduced vegetative and floral parts. Although it was placed in Malaxideae by most authors due to the overall floral similarity with some members of Malaxideae [19,22,32,33], it is out of place in Malaxideae by having hairy roots, a rhizome, a cylindrical column, four waxy pollinia in two separate spheroid anther cells (i.e., lacking an anther cap proper), the pollinia attached to a common, large viscidium, a concave stigmatic area under the rostellum, and the elongate rostellum. Chen et al. [33] established a new subtribe to accommodate it. Our morphological examination indicated that some morphological characters, such as hairy roots, waxy pollinia with viscidium, and elongate rostellum, support Risleya as a member of Collabieae.
Pridgeon et al. [19] defined Collabieae based on several morphological characters. Our results of morphological characters reconstruction and morphological examination indicate that, as redefined here, Collabieae (including Risleya) is characterized by Calanthe-type velamen (hairy velamen), plants with rhizome, labellum usually saccate or spurred at base (except in Eriodes and Risleya), waxy pollinia, and elongate, flap-like rostellum.

Phylogenetic relationships within Collabieae
It is unexpected that Eriodes is sister to the remaining genera in Collabieae (BP = 77, PP = 1.00). Eriodes has been neglected by most authors since it was established (such as [23,34]), but can be easily distinguished from the remaining Collabieae by the combination of an epiphytic habit, distinct conical to globose pseudobulbs with 1-3 non-sheathing leaves, a lip without a spur, a column with a distinct foot, and eight pollinia. Most genera in Collabieae are terrestrial. Its distinctive morphology supports Eriodes as a sister of the remaining genera in Collabieae.
Acanthephippium is usually considered to be a member of Phajinae. Our results indicate that Acanthephippium is sister to other subclades of Collabiinae. This result remains to be tested further. Tainia is subdivided into two groups: one group, traditionally known as Ania (syn. Ascotainia), is sister to a clade formed by Hancockia (subclade II) and subclade I, whereas the other group is mixed with Nephelaphyllum and Mischobulbum in subclade I. These results agree well with their vegetative characters, such as shape and growing pattern of pseudobulbs, and vernation of the leaves. The group formed by Nephelaphyllum, Mischobulbum and Tainia spp. is strongly supported and is well characterized by morphological characters, such as ascending pseudobulbs articulated at their junction with either the petiole or scape, inflorescences arising on specialized leafless shoots, one convolute, petiolate, not sheathing leaf per pseudobulb, lip more or less concave at base, 8 pollinia, and no viscidium. Ania is characterized by conical to ovoid pseudobulbs often growing above ground, plicate leaf with long petiole and sheathing at base, spurred lip, and 8 pollinia without a viscidium. The monotypic Hancockia is characterized by a 1-flowered inflorescence arising from the top of the pseudobulb, pedicel 2.5 cm long, and pollinia with stipe and viscidium. Based on these findings, we support the separation of Tainia in two genera: Tainia (including Nephelaphyllum and Mischobulbum) and Ania.
Phajinae is subdivided into 4 subclades. Gastrorchis is sister to subclade V, which is supported by morphological characters, such as the few-leaved, pseudobulbous stem, pleated leaves, and eight waxy pollinia. Calanthe is subdivided into three groups, one consisting of sect. Calanthe together with sect. Ghiesbreghtia, the second consisting of sect. Styloglossum, and the third of subgenus Preptanthe. Section Ghiesbreghtia nests within section Calanthe in subclade V (Figure 2, 3), which is supported by gross morphological characters, such as their short stem-like pseudobulbs, persistent bracts, inflorescence more or less pubescent and flowers, and lip usually spurred. Calanthe delavayi has been considered as an intermediate between Phaius and Calanthe on column structure and sometimes transferred to Phaius [35], but our results indicate that it belongs to Calanthe sect. Calanthe (Figure 2, 3). Cephalantheropsis is nested within Calanthe sect. Styloglossum in subclade VI, which is supported by several morphological characters, such as an elongate rhizome, leaf sheaths forming a distinct pseudostem, and inflorescence arising from the base of the indistinct pseudobulb (Figure 2, S2).
Based on the morphological and molecular evidence, our results indicate that Calanthe is paraphyletic with respect to Cephalantheropsis, Phaius and Gastrorchis. Morphologically, Calanthe differs from Phaius in having the column more or less united to its apex with the lip, while Phaius is characterized by having the apex of the column free from the lip. However, our results indicated that adnation of the lip to the column evolved several times independently. Some species, including Calanthe delavayi, have an intermediate column type between these two states. There are two alternative approaches for the circumscription of Calanthe and its infrageneric groups. The first option is to consider each subclade in clade II as distinct genera, and narrow Calanthe to include only Calanthe sect. Calanthe and sect. Ghiesbreghtia in subclade V. In this approach, at least four genera, Calanthe, Cephalantheropsis, Gastrorchis, together with a new genus to include the former subgenus Preptanthe, should be recognized. The second option is to define Calanthe in a broad sense: for Calanthe s.l. to remain as monophyletic, then the genera Cephalantheropsis, Gastrorchis, and Phaius should have to be included in it.
Based on morphological and molecular systematics evidence, and in the interest of nomenclatural stability, we prefer to circumscribe Calanthe in the broad sense, including Calanthe s.s., Cephalantheropsis, Gastrorchis, and Phaius.

Evolution from an epiphytic to a terrestrial habit
A shift from the epiphytic to the terrestrial habit has evolved many times in Orchidaceae [4]. Collabieae is nested within a superclade consisting of more than 15 000 species (see [6]), of which 95% are epiphytic. However, Collabieae are predominantly terrestrial orchids, and even an alpine mycoheterotrophic genus has evolved, while the epiphytic Eriodes is sister to remaining terrestrial genera. Our results of morphological characters reconstruction indicated that terrestrial Collabieae have evolved from epiphytic ancestors (Figure 2, S2). This evolution in habit has been accompanied by a variety of pseudobulb shapes, ranging from heteroblastic, petiole-like in Tainia and Hancockia to homoblastic, fleshy and swollen in certain species of Phaius (e.g. P. takeoi), while pseudobulbs are even absent in some taxa. This great diversity in vegetative morphology and complicated evolutionary patterns of morphological characters have contributed to the confusion surrounding the systematics of the tribe Collabieae, which only now is becoming better understood.

Ethics statement
The species collected here are not included in the checklist of Chinese Protected Species. The fieldwork was conducted under the permission of the authority of each natural reserve, specifically Gaoligongshan National Nature Reserve (Yunnan, China), Xishuanbanna National Natural Reserve (Yunnan, China), Huanglianshan National Nature Reserve (Yunnan, China) and Wuzhishan National Nature Reserve (Hainan, China). No specific permits were required for the described field studies.

Taxon and gene sampling
For the subfamily-wide analysis, a total of 96 genera (Table S1 in File S1) were sampled, representing all tribes of subfamily Epidendroideae. In total, 128 accessions of Epidendroideae taxa were analyzed, including two accessions each of Phaius flavus, Risleya atropurpurea, and Tainia hookeriana. Outgroups include 4 species from subfamily Orchidoideae and 1 species from Cypripedioideae. We sequenced 4674 bp of chloroplast DNA, including the rbcL and psaB genes, and the matK pseudogene. All terminal taxa represent single species and include at least two of the three DNA markers. Voucher information and GenBank accession numbers are listed in Table S1 (in File S1).
A second series of analyses focused on the tribe Collabieae (see Table S2 in File S1). We sampled 14 out of 18 genera of Collabieae as circumscribed in Pridgeon et al. [19]. The reduced matrix included 49 ingroup species and 3 outgroups. We analyzed 6775 bp of chloroplast DNA, including the rbcL and psaB genes, the matK pseudogene, and the trnH-psbA region. Voucher information and GenBank accession numbers are listed in Table  S2 (in File S1).
The primers used in both series of analyses are listed in Table  S3 (in File S1).
For BI analyses, we partitioned our data a priori on the basis of gene identity and, for coding regions, codon position (Table S4 in File S1). Based on Bayes factors, the partitioning strategy (rbcL, matK, psaB, and trnH-psbA) was identified as optimal for our data and was applied in all subsequent Bayesian analyses. Initial analyses providing data for comparison of the different partition strategies were run for 3 000 000 generations, and analyses applying the final best-fit model were run for 5 000 000 generations. Runs were started from a random tree sampled every 1000 generations of the MCMC chain, with default priors and the option prset/ratepr set as variable. Each parameter estimation obtained from the results of two runs was checked in Tracer v1.5 (http://tree.bio.ed.ac.uk/software/tracer) to ascertain whether they had obtained proper effective sample size and to verify that stationary state had been reached. Trees from the first 10% of generations were discarded as burn-in. The remaining trees were combined to build a 50% majority-rule consensus tree. Bayesian inference was run on CIPRES [40]. The data matrix and phylogenetic trees have been submitted to TreeBase (http://purl. org/phylo/treebase/phylows/study/TB2:S14958).

Morphological data analysis
A total of 41 characters were included in the analysis (Table S5 in File S1). Characters were coded for 40 species representing most genera in Collabieae. Nine characters were constant and 27 were parsimony informative. A parsimony and Bayesian analyses with all characters equally weighted was conducted in PAUP* v4.0b10 [38] and MrBayes v3.2.0 [39]. The evolution of morphological characters was reconstructed using a maximum parsimony approach implemented in Mesquite v2.74 (http://mesquiteproject. org/mesquite/mesquite.html). File S1 Tables. Table S1. Taxa, voucher and GenBank accession numbers of Epidendroideae used in this study; Table  S2. Taxa, voucher and GenBank accession numbers of Collabieae used in this study; Table S3. Primers used for amplification and sequencing in this study; Table S4. Partition analysis of Bayesian inference; Table S5. Morphological data matrix for the phylogenetic analysis. (DOCX)