Paraholcoglossum and Tsiorchis, Two New Orchid Genera Established by Molecular and Morphological Analyses of the Holcoglossum Alliance

Background Holcoglossum is a small orchid genus of 12 species ranging from SW China to Thailand and NE India. Although molecular and morphological analyses have been performed to establish the phylogenetic relationships within this genus, the interspecific relations and its relations with allied genera, such as Rhynchostylis, Aerides and Vanda, remain unclear. Methodology/Principal Findings In addition to morphological analysis, maximum parsimony, maximum likelihood, and Bayesian inference analyses were performed based on fragments of the nuclear ITS and chloroplast trnL-F and matK genes of 31 taxa (15 Holcoglossum, 14 Aeridinae, 2 outgroups) representing all major clades of the Holcoglossum alliance. The results suggest that Holcoglossum is triphyletic, comprising three clades: the Holcoglossum clade, its sister clade, and a distant clade more closely related to Rhynchostylis, Aerides, and Vanda than to the Holcoglossum clade. The Holcoglossum clade is further divided into three subclades; the genetic distances between these three subclades also support this delimitation. The molecular conclusion is consistent with their distinct morphological characters. Conclusions We propose that the latter two clades comprise two new genera, Paraholcoglossum and Tsiorchis, and Holcoglossum clade divides into three sections. In addition, a new section, Holcoglossum sect. Nujiangensia, and a new species, Holcoglossum linearifolium, are proposed. Some new combinations are made, and a new scheme is provided for the classification of all species of Holcoglossum, Paraholcoglossum, and Tsiorchis.


Introduction
Holcoglossum is a small genus of Orchidaceae (Aeridinae, Vandeae) comprising about 12 species ranging from China to Vietnam, Laos, Thailand, Myanmar, and NE India [1]. After it was established by Schlechter [2], Holcoglossum remained monotypic for over 50 years until Garay [3] delimited the genus and transferred Vanda kimballiana Rchb. f. and Vanda rupestris Hand.-Mazz. to it. In 1982, Tsi [4] added two species, Holcoglossum flavescens (Schltr.) Z. H. Tsi and H. junceum Tsi, though the latter was reduced to a synonym of Ascocentrum himalaicum (Deb, Sengupta & Malick) Christenson by Christenson in 1987 [5]. After an examination of the ''Vanda-Aerides alliance'', Christenson [5] assigned Vanda subulifolia Rchb. f. and Vanda amesiana Rchb. f. to Holcoglossum, and further divided Holcoglossum into two sections, H. sect. Kimballianum and H. sect. Holcoglossum, based on plant habit and flower number per inflorescence. Seidenfaden [6,7] disagreed with Christenson's treatment, however, and suggested that a new genus is necessary for Vanda amesiana and V. subulifolia. For the latter 2 species, Jin [8] proposed a new subgenus of Holcoglossum, H. subgen. Brachycentron. This subgenus is characterized by its lip shallowly saccate at the base, with a ridged callus on the front margin of the sac mouth, and an oblong or linear stipe with its base as broad as its apex. In the typical subgenus H. subgen. Holcoglossum, the lip has a conspicuous spur at the base and a crested or fleshy callus on the mid-lobe. The stipe is tapered, with one end broader than the other. Jin [8] further divided the latter subgenus into two sections: H. sect. Holcoglossum and H. sect. Sorotylos. Based on nuclear ITS and chloroplast trnL-F and matK analyses, Fan et. al. [9] suggested that Holcoglossum is a monophyletic genus consisting of three major subclades, albeit including two debatable species, H. amesianum and H. subulifolium. They even treated H. kimballianum and H. wangii as being closely related to the two debatable species, and grouped them together into the same clade (tropical clade). Morphologically, however, H. kimballianum and H. wangii are quite different from H. amesianum and H. subulifolium, and they were once regarded as belonging to two different sections [5,10] or even two different subgenera [8]. In fact, according to the molecular evidence provided by Fan et al. [9], the systematic positions of H. kimballianum and H. wangii are still not clear. This treatment is problematic, however, and clarification requires further molecular analysis or repeated tests for further generic delimitation and infrageneric classification.
In this study, phylogenetic relationships between Holcoglossum, allied genera, and infrageneric taxa were assessed using longer gene sequences for nuclear ITS, chloroplast trnL-F, and matK than employed previously [9], in combination with morphological analyses.

Results
The DNA sequences of 31 taxa were obtained and analyzed. The DNA sequences of 15 species (14 species of Holcoglossum and 1 species of Papilionanthe) were newly obtained except for the matK fragments from three species. Most of the sequences we obtained are longer than those used in previous research [9]. Detailed sequence information is listed in supplementary Table S2, and can also be accessed from GenBank. The aligned length, the indels' information, the numbers of variable sites and parsimony informative sites, tree statistics for the maximum parsimony (MP) analysis, and the best-fit model selected by Modeltest are given in Tables 1 and 2. Genetic distances between species and between sections of Holcoglossum are given in Table 3.

ITS analysis
This analysis strongly supports Holcoglossum auriculatum as a sister to Papilionanthe teres and P. biswasiana, the posterior probabilities (PP) is 100%. Holcoglossum can be divided into three major clades.

Morphological analyses
The morphological character matrix phylogenetic tree of Holcoglossum supports the division of Holcoglossum into three clades: the first clade consisting of H. kimballianum and H. wangii, the second consisting of H. amesianum, H. subulifolium and H. auriculatum, and the third including the remaining species of Holcoglossum.

Combined dataset analysis
In the present study, we also combined ITS, trnL-F, and matK into a single dataset. The strict consensus trees strongly support the division of Holcoglossum into three clades (PP100%). The first clade, consisting of H. kimballianum and H. wangii, is strongly supported as a sister to the genera Rhynchostylis and Vanda (PP100%). The second clade, consisting of H. amesianum, H. subulifolium, and H. auriculatum, is strongly supported as a sister to the first clade (PP100%). These two clades are not included in main clade of Holcoglossum. The last clade includes the remaining species of Holcoglossum (PP100%) and is further divided into three subclades or sections.  Fig. 1; Supplementary Figs. S10, S11).
In the present study, we also combined molecular data and morphological characters in a single dataset. The BI consensus tree of this dataset has similar topology to that constructed from the combined data set from ITS, trnL-F, and matK analysis, but most of the clades and subclades are better delimited and more strongly supported (Fig. 2).
Among the 14 analyzed species in six genera (Aeridinae), the species of five genera cluster together in the cladogram forming several related clades with the exception of one genus, Ascocentrum (A. ampullaceum), and two species, Aerides flabellata and Aerides odorata, which all cluster with Vanda forming the Vanda clade.

Genetic distance analyses
In general, the genetic distances (GD) between species within a genus are lower than those between intergeneric species. The mean genetic distance between all the species is 0.0311 when outgroups are included, and the value is 0.0245 when outgroups are excluded. Within Holcoglossum, most of the genetic distances between species are lower than 0.0200 (see Supplementary Table  S1). The average genetic distance of all the species of Holcoglossum is 0.0154. In the Holcoglossum clade, the GD values between species are given in Table 3 (the mean genetic distance is 0.0137). The genetic distances between H. linearifolium and the other species are generally higher than distances between other species pairs, which strongly supports the treatment of H. linearifolium as a new species distinct from all known species.

Data of ITS, trnL-F and matK analyses
In the present study, ITS, trnL-F, and matK were used to resolve the phylogenetic relationships within the genus Holcoglossum. We developed five sets of primers to sequence some difficult fragments, but there were still only partial fragments of the matK for three species (H. weixiense, H. rupestre, H. sinicum) (see Supplementary  Table S2), and the fragments of the trnL-F for H. kimballianum and H. subulifolium have some differences from those reported in a previous study [9]. In the present study, the trnL-F of H. subulifolium lost 291 bp compared to the sequence described previously [9], so we used the previous sequence. We also used the new sequence in our analysis, however, and results in the combined dataset analysis were the same (see Supplementary Figs. S12, S13, S14). The fragments of H. kimballianum in the previous study [9] had one more insert (19 bp) than in our study, so both of them were used for the present analysis. Therefore, these five fragments were accessed from GenBank (see Supplementary Table S2). In total, the fragments we obtained were almost the same as those of Fan et al. [9], but longer (see Supplementary Table S2).
In this study, we found that it was better to utilize ITS and trnL-F than matK to resolve interspecific relationships within the genus Holcoglossum, consistent with the previous study [9]. The ITS and trnL-F produced approximately equivalent tree topologies (see Fig. 1 and Supplementary Figs. S1, S2, S3, S4, S4, S6) that agreed with the main clades established by morphological characterization (see Supplementary Fig. S10). The consensus tree of matK showed inexact results (see Supplementary Figs. S7, S8, S9), which may be due to insufficient variations in the sequence of matK. Many studies had shown that at lower taxonomic levels (tribes and below), non-coding plastid markers (often the trnL intron and the trnL-F intergenic spacer) and ITS (ITS1-5.8S-ITS2) nuclear ribosomal DNA spacer [11] may yield better results than protein-coding genes such as rbcL [12,13], matK [14], psaB [15], or ycf1 [16]. Conversely, at the family level, plastid proteincoding genes have been the primary focus, including as rbcL [12,13], matK [14], psaB [15], and ycf1 [16]. Obviously this study focused on lower taxonomic levels (tribes and below), so it was reasonable that the consensus trees of matK showed inexact results. When a combined ITS, trnL-F, and matK dataset was used to analyze these relationships, however, the best tree topology was obtained, and was consistent with the delimitation based on their morphological characters. Using this combined dataset, most of the clades and subclades were better delimited and more strongly supported. From the present study, the matK fragments were still useful in helping to resolve the relationships below the generic rank when combined with ITS and trnL-F.   Table 3. Interspecific genetic distances of Holcoglossum clade based on combined datasets of ITS, trnL-F, and matK. In this study, phylogenetic analyses were performed under maximum likelihood (ML), maximum parsimony (MP), and Bayesian inference (BI) for each dataset. The BI, MP, and ML trees of each dataset had similar topological structures, meaning that the analysis had very good repeatability and that the experimental data were reliable. The differences between the BI, MP, and ML trees were the values of bootstrap percentages (BP) or posterior probabilities (PP) in each node; in general, the values in the BI tree were higher than that in MP and ML trees. In this paper, we will use the BI trees to demonstrate phylogenetic relationships between Holcoglossum and its allied genera. The MP and ML trees are presented in the Supplementary Information (see Supplementary Figs. S1, S2, S3, S4, S5, S6, S7, S8, S9, S10, S11, S12, S13, S14).

Phylogeny of Holcoglossum
In total, 15 Holcoglossum species, including a new one, were analyzed. Analysis of combined chloroplast and nuclear data strongly supported (PP100%) the tripartite division of Holcoglossum into the Holcoglossum clade, its sister clade, and a distant clade ( Fig. 1 and Supplementary Figs. S10, S11, S12, S13, S14). Similar results were obtained by combining molecular data and morpho- logical characters in a single dataset (Fig. 2). On these grounds, the trees of ITS (Supplementary Figs. S1, S2, S3), trnL-F (Supplementary Figs. S4, S5, S6) and the combinational dataset produced highly similar tree topologies. The main clades established were supported by these datasets. The phylogenetic relationships between Holcoglossum and allied genera were well resolved, and all data sets supported the division of Holcoglossum into three clades. Although the consensus trees of matK (Supplementary Figs. S7, S8, S9) showed inexact results, it supported the trees of trnL-F, showing that H. wangii and H. kimballianum form a new clade outside the Holcoglossum clade. The BI consensus trees of ITS, trnL-F, and combinational dataset were similar, but most of the clades and subclades were better delimited and more strongly supported by the combinational dataset, implying that combinational datasets may be better than single genes for phylogenetic analysis.
From the tree of the combinational dataset, the Holcoglossum clade was a sister to Papilionanthe, a genus of some ten species, while H. subulifolium, H. auriculatum, H. amesianum, H. kimballianum, and H. wangii were not included in the Holcoglossum clade. They formed two clades outside the main Holcoglossum clade. In fact, H. kimballianum and H. wangii were more closely related to Vanda than to the main Holcoglossum clade in the trnL-F and matK dataset. Moreover, the clade sequences are considerably different from each other, especially in the trnL-F sequence. In order to explain the differences between them, we listed some obvious differences in the trnL-F aligned file (see Supplementary Figs. S15, S16, S17, S18 for details). This aligned file revealed that each of the three clades had their own unique characters and could be easily distinguished. In particular, H. kimballianum and H. wangii are quite distinct from the other species. Moreover, the phylogenetic relationships established between species in the three clades by analysis of morphological characters were strikingly similar to the relationships defined by the combined molecular data (Fig. 2). Therefore, we conclude that some species must be classified into two new genera, Tsiorchis and Paraholcoglossum, distinct from Holcoglossum. This conclusion may also be drawn from Jin [8] and Fan et al. [9], but they treated them only at subgeneric rank. As a result, Tsiorchis and Paraholcoglossum are proposed to be outside the Holcoglossum clade.
Paraholcoglossum comprises three species, Paraholcoglossum subulifolium, P. auriculatum, and P. amesianum, while Tsiorchis consists of two species, Tsiorchis kimballiana and T. wangii. Holcoglossum includes the remaining ten species. The three genera show obvious morphological differences. In Holcoglossum, the white lip is conspicuously spurred at the base, with a crested or fleshy callus at the broad base of the mid-lobe, and the pollinia are porate and directly attached to a tapering stipe (see Supplementary Fig. S19). In Paraholcoglossum, the lip is shallowly saccate at the base, with a ridged callus at the entrance to the sac, and the mid-lobe is clawed at the base. The pollinia is porate in Paraholcoglossum, but the stipe is oblong, with both ends almost equal in width (see Supplementary Fig. S20). Tsiorchis is quite distinct from Holcoglossum and Paraholcoglossum, having a cleft, not a porate pollinia; each pollinium has a distinct caudicle attached to a common stipe; the mid-lobe of the lip is purple or has purple markings and the side-lobe is yellow with dark-purpled markings (see Supplementary  Fig. S21). Within the genus Holcoglossum, the infrageneric system proposed by Jin [8] and Fan et al. [9] accurately reflect many interspecies relationships, though some changes are made in our system (Fig. 1,  2). We also analyzed the genetic distances between the sections of Holcoglossum.  Supplementary Table S3). This result is consistent with topological analysis. The morphological characters and molecular evidence both support the division of Holcoglossum into three subclades (Figs. 1, 2). The first subclade, Holcoglossum sect. Holcoglossum, includes four species, H. tsii, H. quasipinifolium, H. omeiense and H. lingulatum, all of which are characterized by their mid-lobe of lip arising from the upper third of the spur, ligulate mid-lobe twice as long as the spur, mid-lobe with 2-7 lamellae adaxially (see Supplementary Fig. S19. a-c). The second subclade, Holcoglossum sect. Nujiangensia, is a new section with two species, H. nujiangense and H. linearifolium sp. nov., and is characterized by a mid-lobe of lip arising from the middle of the spur, obovate almost as long as the spur, and 3-5 crested lamellae extending down to the middle of the spur (see Supplementary Fig. S19 Supplementary Fig. S19. f, g).
We also assessed the genetic distances between the individual species of main Holcoglossum clade. The mean interspecies genetic distance is 0.0137 (Table 3) and pair-wise distances range from 0.0000 (H. omeiense and H. lingulatum) to 0.0313 (H. linearifolium and H. quasipinifolium), while the genetic distances between H. linearifolium and other species range from 0.0090 to 0.0313, with a mean (0.0215) significantly higher than the average interspecific genetic distance within the genus Holcoglossum. The genetic distance between H. linearifolium and H. nujiangense is 0.0090, which is also much higher than that between most other species pairs, such as H. weixiense and H. sinicum (0.0037), H. tsii and H. rupestre (0.0053), or H. rupestre and H. weixiense (0.0047) ( Table 3). Considering these molecular results as well as morphological characters, H. linearifolium is proposed as a new species that differs from its ally H. nujiangense by having a stem 4-5 cm long, filiform, soft leaves about 1 mm thick and often 30-40 cm long, side-lobes of lip narrowly ovate (not triangular), and mid-lobe obovate 8-10 mm in length.
Among the 14 analyzed species in six genera (Aeridinae), the species of five genera cluster together in the cladograms forming several related clades. However Aerides flabellata, Aerides odorata, and Ascocentrum ampullaceum cluster together with Vanda to form the Vanda clade, probably due to some species missing partial fragments of trnL-F or matK region. (see Supplementary Table  S2). Papilionanthe teres and P. biswasiana cluster together with the Holcoglossum sect. Nujiangensia. The results indicate that analysis of longer DNA sequences is helpful for resolving the relationships below generic rank.

Conclusion
In the present study, we investigated the phylogenetic relationships between Holcoglossum and allied genera, as well as the infrageneric classification of the genus Holcoglossum based on both morphological characters and molecular evidence. Holcoglos-sum is triphyletic and can be divided into two major sister clades and a distant clade. The three clades are regarded as three genera: Holcoglossum, Paraholcoglossum, and Tsiorchis. The latter two are published here as new genera. Paraholcoglossum includes three species and is related to Holcoglossum, but is quite distinct in many respects. Tsiorchis includes two species and is more closely related to Vanda than to Holcoglossum. The genus Holcoglossum itself comprises three subclades, all recognized here at sectional rank: H. sect. Holcoglossum with four species, H. sect. Sorotylos with four species and H. sect. Nujiangensia with two species (Fig. 1, 2).
The new genus is akin to Holcoglossum, from which it differs by its lip saccate at base, with a ridged callus on the front margin of the sac mouth, mid-lobe clawed at base, and stipe oblong, with its base as wide as apex.
? Description. Epiphytic plants. Stems often elongate. Leaves broadly subterete to terete, adaxially channeled. Inflorescence racemose or paniculate (with 1-2 branches), with several to many flowers; bracts much shorter than pedicellate ovary; flowers open widely; sepals and petals subsimilar; lateral sepals usually slightly larger; lip saccate at base, trilobed, with a ridged callus at the front margin of the sac mouth; mid-lobe clawed at base; column short, winged, with a very short but distinct foot; pollinia 2, waxy, globose, porate, attached by an oblong stipe to a broad viscidium.
The new genus is subsimilar to Holcoglossum, Rhynchostylis, and Vanda. It differs from Holcoglossum by its cleft (not porate) pollinia, each with a distinct caudicle, purple-mottled lip and cylindrical spur; from Rhynchostylis by its terete leaves, trilobed lip with conspicuous side-lobes at base, footless column and shorter and broader stipe; from Vanda by its terete leaves, distinct caudicle and cylindrical spur.

Materials
In total, 15 species of Holcoglossum were analyzed. Fourteen of them were sampled in this study, while one, H. tsii, was accessed from the GenBank. Fourteen additional species, representing six closely related genera, Aerides, Ascocentrum, Neofinetia, Papilionanthe, Rhynchostylis, and Vanda, were accessed from the GenBank and were treated as an ingroup in order to test the monophyly of Holcoglossum and to interpret its genetic relationships. Two African orchids of Vadeae, Jumellea sagittata, and Microterangis hariotiana, were chosen as outgroups [17,18]. For information regarding the assessment, see Supplementary Table S2 for details.

Amplification and sequencing
Total DNA was extracted from fresh material or silica-gel-dried plant tissue with a Multisource Genomic DNA Miniprep Kit (Axygen Biosciences) following the manufacturer's instructions. The amplification reaction included total DNA, primers, Ex-Taq buffer, and Ex-Taq DNA polymerase (Takara Bio). The polymerase chain reaction (PCR) profile consisted of an initial 5 min pre-melt stage at 95uC, then 30 cycles of 30 s at 95uC (denaturation), 30 s at 50-55uC (annealing temperature was determined by primer's need), and 1-3 min at 72uC (extension time was determined by length of the aim DNA region), followed by a final 10 min extension at 72uC.
The amplification of the ITS region was performed using the primer pairs ITS A and ITS B [19]. The trnL-F region was amplified with primers c and f [20] or two sets of primers developed specifically for this study (Table 4). For matK sequences, amplification was performed using the primer pair matK-19F and trnK-2R [19], and some fragments were amplified using three sets of primers developed for this study (Table 4). To check the quality of the amplified DNA, PCR products were run on 1.5% agarose gels. The gels with target products were excised, purified using DNA Gel Extraction Kits (Axygen Biosciences), and sequenced by BGI Americas Corporation.

Sequences editing and assembling
Both forward and reverse sequences and electropherograms were edited and assembled with DNASTAR (http://www.dnastar. com/). DNA sequences were aligned with MEGA5.05 [21] under the Muscle model and then adjusted manually with MEGA5.05 [21]. Aligned sequences are available from the corresponding authors upon request.

Morphological analyses
To explore the phylogenetic positions of the Holcoglossum alliance by morphological classification, we constructed a morphological character matrix consisting of 41 characters of 31 taxa (see Supplementary Morphological Character Codes S1and Table S4 for details).

Data analyses
The datasets included ITS, matK, trnL-F, and morphological characters, a combination of ITS, matK and trnL-F, and a combination of all four. Insertions, deletions, and some unavailable sequences were treated as missing. Phylogenetic analyses were performed under maximum likelihood (ML), maximum parsimony (MP), and Bayesian inference (BI). The best fit model for each dataset was selected by Modeltest 3.7 [22] under the Akaike Information Criterion (AIC)(see the Table 3 for the detail).
The MP analyses were performed using PAUP* version 4.0b10 [23]. All characters were equally weighted and unordered. Test settings included 1,000 replications of random addition sequence and heuristic search with tree bisection-reconnection (TBR) branch swapping. The length of tree, consistency indices (CI), and retention indices (RI) are given in Table 2. The ML analysis was computed by RAxML version 7.2.8 with 100 bootstrap replicates and settings as described in Stamatakis et al. [24]. BI analysis was performed using MrBayes 3.1.2 [25]. The best-fit model for each dataset was selected by Modeltest 3.7. In the combined dataset of ITS, matK and trnL-F, the model was also according to the best fit model for each individual dataset. The following settings were applied: sampling frequency = 100; temp = 0.1; burn-in = 10,000; and the number of Markov chain Monte Carlo (MCMC) generations = 4,000,000. The first 10,000 trees were discarded as burn-in to ensure that the chains reached stationarity. Majority-rule consensus tree was constructed on those trees sampled after generation 1,000,000. We also utilized MEGA5.05 [21] to estimate genetic distances between the species and sections of Holcoglossum treated under the Tajima-Nei model based on the combined molecular dataset.

Nomenclature
The electronic version of this document in itself does not represent a published work according to the International Code of Botanical Nomenclature [26], and hence the new names contained in the electronic version are not effectively published under that Code from the electronic edition alone. Therefore, a separate edition of this document was produced by a method that assures numerous identical printed copies, and those copies were simultaneously distributed (on the publication date noted on the first page of this article) for the purpose of providing a public and permanent scientific record, in accordance with Article 29 of the Code. Copies of the print-only edition of this article were distributed on the publication date to botanical or generally accessible libraries of the following institutions (BM, COL, GH, HUA, K, MEXU, MO, NY, QCA, QCNE, USM). The separate print-only edition is available on request from PLoS (Public Library of Science) by sending a request to PLoS ONE, Public Library of Science, 1160 Battery Street, Suite 100, San Francisco, CA 94111, USA along with a check for $10 (to cover printing and postage) payable to ''Public Library of Science''. In addition, new names contained in this work have been submitted to IPNI, from where they will be made available to the Global Names Index. The IPNI LSIDs (Life Science Identifiers) can be resolved and the associated information viewed through any standard web browser by appending the LSID contained in this publication to the prefix http://ipni.org/.
The online version of this work is archived and available from the following digital repositories: PubMedCentral (www.pubmedcentral.nih.gov/). Morphological character codes S1

(DOC)
Table S1 Pair-wise genetic distance of all species based on combined datasets of ITS, trnL-F, and matK. (DOC)