Phylogenetic Analysis of a ‘Jewel Orchid’ Genus Goodyera (Orchidaceae) Based on DNA Sequence Data from Nuclear and Plastid Regions

A molecular phylogeny of Asiatic species of Goodyera (Orchidaceae, Cranichideae, Goodyerinae) based on the nuclear ribosomal internal transcribed spacer (ITS) region and two chloroplast loci (matK and trnL-F) was presented. Thirty-five species represented by 132 samples of Goodyera were analyzed, along with other 27 genera/48 species, using Pterostylis longifolia and Chloraea gaudichaudii as outgroups. Bayesian inference, maximum parsimony and maximum likelihood methods were used to reveal the intrageneric relationships of Goodyera and its intergeneric relationships to related genera. The results indicate that: 1) Goodyera is not monophyletic; 2) Goodyera could be divided into four sections, viz., Goodyera, Otosepalum, Reticulum and a new section; 3) sect. Reticulum can be further divided into two subsections, viz., Reticulum and Foliosum, whereas sect. Goodyera can in turn be divided into subsections Goodyera and a new subsection.


Introduction
The genus Goodyera R. Br. (subtribe Goodyerinae; tribe Cranichideae; subfamily Orchidoideae) comprises ca. 40 species [1][2][3][4] and is widely distributed, including Asia, northeast Australia, Europe, South Africa, Madagascar, North America (including Mexico) and the southwestern Pacific islands [4]. Thirty-three species of Goodyera are recognized in China (with 12 endemics), showing widespread distribution mainly in south and southeastern China [4][5][6][7][8][9]. Goodyera is a genus of mainly terrestrial (rarely epiphytic) orchid which grows in shade on mossy rocks or along moist tracks of perennial mountain stream banks; it is characterized by the creeping rhizome, upper surface of leaves often with white or golden markings and veins, saccate labellum (glabrous or not internally), two sectile pollinia attached to a viscidium and a single stigmatic lobe [3]. Due to the remarkable markings on the leaves some taxa of this genus are known in horticulture as 'jewel orchids'. The markings, as well as the overall colouration of the leaves of this genus, vary considerably depending on habitat conditions, and hence can be cause of confusion in field identification.
Schlechter [10] divided Goodyera into two sections: Otosepalum (lateral sepals reflexed) and Eu-Goodyera (lateral sepals normally spreading), his treatment was followed by Seidenfaden [11][12] and Pearce & Cribb [13]. However, some other scholars did not adopt Schlechter's system because of the difficulty in defining the exact orientation of the lateral sepals. Lang [14] and Chen et al. [4] treated the markings on the leaves and whether the leaves are rosulate or not as important characters in distinguishing groups of species. Tian [15] did taxonomic studies of Goodyera in China based on morphological data where she also disagreed with Schlechter's stand on section division, and preferred the lip sac hairy or not as more prominent feature than the orientiation of lateral sepals. However, most classifications of Goodyera have been based on morphological attributes.
The advent of molecular techniques has dramatically advanced our understanding of the phylogenetic relationships in family Orchidaceae. The internal transcribed spacer (ITS) region of nrDNA possesses moderate interspecific variation and has been the primary source of characters for phylogenetic analysis at lower taxonomic levels [16]. In the previous systematic studies based on molecular data, Shin et al. [17] conducted a phylogenetic analysis of five Goodyera species from Korea based on the ITS region which indicated that Goodyera was monophyletic, and the ITS sequences of G. schlechtendaliana Rchb. f. and G. repens (L.) R. Br. were identical to one another. The achievements were limited because of limited sampling and single DNA marker. Chung [18] investigated the systematics of Goodyera mainly based on taxa reported from Taiwan and concluded that the genus was monophyletic according to morphological and cytological data as well as ITS sequences. He divided Goodyera into three sections (including the two sections proposed by Schlechter [10] and a new section, Reticulum S. W. Chung & C. H. Ou), which were further subdivided into seven subsections (Table 1). But in his research, only one single ITS marker was used and all those new sections and subsections ( Table 1) that he proposed without Latin diagnoses, which turned them invalid according to the Melbourne Code [19].
Juswara [20] utilized ITS and two chloroplast markers trnL-F and rpl16 sequences to study the phylogyne of Goodyerinae. In her study, Goodyera was polyphyletic, 11 Goodyera species were split into two subclades each cluster with other genera in Goodyerinae. She adopted Schlechter's classification and no further subsectional treatment was given.
As previous molecular systematics of Goodyera were largely based on samples from Tropical area [18,20] or utilized a single DNA marker (ITS) [17,18], the systematics of Goodyera is still unclear. Likewise, in other studies [21][22][23][24][25][26][27][28][29][30] only a few species of Goodyera have been included, which has not unable to resolve the phylogeny of Goodyera as a whole. The 33 species of Goodyera present in China represent ca. 83% of the total of 40 spp. in the genus worldwide, so the phylogenetic study on Chinese members of the genus is of high value in building up a global phylogenetic framework of Goodyera. Based on previous research [15,17,18,20], we conduct a comprehensive phylogenetic study of Goodyera based on DNA sequence data of ITS and two plastid regions (trnL-F, matK) in this study, with the aim of assessing the monophyly of Goodyera and shed light on its infrageneric relationships. Schlechter [10] Chung [18] Sect. Otosepalum Schltr. Sect. Otosepalum Schltr.

Ethics statement
All the samples were collected and processed in their respective countries, for all samples from China, DNA extraction and sequencing was done from live or fresh silica gel dried specimens, for all samples from outside China, FTA cards were used to collect the plant extract for DNA extraction and sequencing. So none was taken out of its respective country, none of the species studied belongs to rare, endangered or threatened species according to IUCN, all orchids studied during the current study are not included under CITES Appendix II, none of the samples was collected within protected areas, hence no permission was needed.

Taxon sampling
In total, we analysed 132 samples representing 35 species of Goodyera and 27 additional genera/48 species of related genera (55 accessions), including 64 sequences from GenBank. Pterostylis longifolia (Pterostylidinae) and Chloraea gaudichaudii (Chloraeinae) were chosen as outgroups on the basis of previous phylogenetic studies [23,28]. Five species of Goodyera [Goodyera brachystegia, G. fusca, G. makuensis Ormerod, G. malipoensis Q. X. Guan & S. P. Chen and G. wuana] from China could not be sampled in spite of repeated attempts. Voucher specimens were deposited at the Herbarium of East China Normal University (HSNU) and the Herbarium of Taiwan Forestry Research Institute (TAIF). Detailed voucher information is provided in S1 Table. DNA extraction, amplification and sequencing Genomic DNA was extracted from 10 mg of fresh or silica-dried tissue using a modified CTAB method [31]. Amplification was carried out on in a TAKARA TP600 thermocycler (TAKARA BIO INC, Japan) using 50 μl reactions containing 25 μl 2× Taq PCR Master Mix (BIOMIGA, China), 17.5 μl ddH 2 O, 2.5 μl of each primer (10 μM) and 2.5 μl of target DNA template (0.5 ng/μl). The ITS and trnL-F regions were amplified with two primers, but matK was amplified using 5 primers. The primers and amplification protocols for each DNA region are listed in S2 Table. PCR products were purified using a PCR purification kit (BIOMIGA, China). For each region, both strands were sequenced with the same primers as for the amplification, except for trnL-F for which two internal primers were used. Sequencing for this work was outsourced to Invitrogen Biotechnology Corporation (Shanghai, China), Majorbio Bio-Pharm Technology Corporation Limited (Shanghai, China) and the Beijing Genomics Institute (BGI, China). All sequences have been submitted to GenBank and accession numbers are listed in S1 Schlechter [10] Chung [18] Goodyera bifida (Bl.) Bl.

Goodyera velutina Maxim.
Goodyera robusta Hook. f. Phylogenetic analyses Sequences were firstly assembled and edited with Seqman (DNA STAR package, Madison, WI, USA) [32], aligned with Mega 5 [33] and then adjusted manually. Three datasets, namely ITS, the combined chloroplast dataset (matK and trnL-F) and the combined nuclear and chloroplast DNA sequences (ITS, matK and trnL-F) were analysed using Bayesian inference (BI), maximum parsimony (MP) and maximum likelihood (ML); all characters were treated as unordered and equally weighted. Indels were treated as missing data. The maximum parsimony (MP) analyses were performed with PAUP Ã version 4.0b10 [34]. A heuristic search with 1000 random addition sequence replicates, tree bisectionreconnection (TBR) branch swapping and the MulTrees (saving multiple trees in memory) option were performed. Bootstrap values were generated with 1000 bootstrap replicates with TBR branch swapping, with each replicate performing 100 random-addition sequence replicates and a limit of 1000 trees. Homoplasy levels were assessed by means of the consistency index (CI) and the retention index (RI). For the ML analyses, MrModelTest 2.3 [35] was used to select the most suitable model under the Akaike information criterion (AIC) [36]. The GTR+I+G model was selected as the best-fit model by MrMTgui 1.0 [37] for all datasets. Then the models were added in a command block after the data in the NEXUS file. A total of 1000 bootstrap replicates were performed using Garli v0.951-GUI [38]. Other parameters were set as default for the Garli searches. PAUP Ã version 4.0b10 was used for exporting tree files. Both for the MP and ML analyses, bootstrap values over 89% were considered as high support, between 71% and 89% moderate support and below 71% as weak support. The Bayesian inference analyses were conducted with MrBayes 3.1.2 [39]. GTR+I+G was selected as the best-fit model as for the ML analyses for all datasets. The analyses consisted of 3,000,000 generations of four simultaneous Monte Carlo Markov chains. We increased the number of generations until the average deviation of split frequencies fell below 0.01 [40]. Trees were sampled every 1000 generations; the samples prior to stationary were discarded as burn-in using Tracer v. 1.5 and the remaining trees were used to build a majority-rule consensus tree on which the posterior probabilities (PP) were shown. We defined PP values above 0.90 as high support, between 0.80 and 0.89 as moderate support, and below 0.79 as weak support.

Homogeneity test
Homogeneity between the ITS data and the combined chloroplast datasets (trnL-F and matK) was tested following Farris et al. [41] using the incongruence length difference (ILD) test, as implemented in PAUP Ã version 4.0b10 [34].

Characteristics of sequence data and inferred phylogenetic trees
One hundred and forty-three ITS, 90 trnL-F and 82 matK sequences were newly generated in this study. All these sequences have been submitted to GenBank. Sixty-four sequences were added to our analysis from GenBank. In total, we analysed 379 sequences of 187 accessions (132 of Goodyera and 55 of other genera).
Details of three datasets are shown in Table 2. For all datasets, the three phylogenetic methods yielded similar phylogenetic patterns, but the MP trees were the most resolved. Posterior Probabilities from the BI analysis and bootstrap values from both the MP and ML analysis are shown on the MP trees (Fig 1, S1 and S2 Figs).

Nuclear DNA dataset analysis
The ITS dataset included 183 ingroup taxa, two outgroups and 818 characters, of which 162 (19.80%) were variable and 347 (42.42%) were parsimony-informative. The analysis found 141 equally shortest trees with a length of 1662 steps, CI = 0.5078 and RI = 0.4922. The strict consensus of the 141 trees is shown in S1 Fig

Combined nuclear and chloroplast DNA dataset analysis
The result of the ILD test for the nrDNA and combined cpDNA showed incongruence between the two datasets (P = 0.01). However, the support of branches increased in the combined tree and the incongruence might disappear with more data [41]. So we included the combined dataset in our analysis. The combined dataset of the three markers (ITS, trnL-F & matK) had 4033  1. It showed a similar topology to the plastid DNA tree with the support values of some main clades increased, and it was more resolved than the ITS analysis. The combined tree of three markers indicated that Goodyera is closely allied to Erythrodes, Platythelys, Lepidogyne and Hylophila and these genera were more close to sect. Otosepalum. Goodyera procera formed an independent clade (O) (PP = 1, MP = 100%, ML = 100%). The strongly supported clade Q (PP = 1, MP = 99%, ML = 100%) and clade R (PP = 1, MP = 92%, ML = 96%) each consisted of two subclades (subclades Q 1 , Q 2 and subclades R 1 , R 2 , respectively). Goodyera vittata and G. biflora nested in subclade Q 1 with species of sect Goodyera, subsect. Reticulum. Three species viz., G. foliosa, G. velutina and G. henryi Rolfe formed a well supported subclade Q 2 (PP = 1, MP = 99%, ML = 100%). The complexes of G. schlechtendaliana (G. schlechtendaliana, G. daibuzanensis., G. bilamellata G. kwangtungensis and G. robusta) and G. repens (G. nankoensis, G. prainii Hook. f., G. marginata Lindl., G. pendula Maxim., G. yunnanensis Schltr., G. wolongensis and G. bomiensis) each formed a strongly supported subclade (R 1 and R 2 , respectively).

Circumscription of Goodyera
Goodyerinae had been recognized as a well-defined group by several authors [21,28] using molecular data from both chloroplast and nuclear markers. However, within the subtribe especially in Goodyera, the evolutionary relationships between taxa remain unresolved. Pridgeon et al. [3] pointed out that circumscription of Goodyera was problematic and that a better understanding of the infrageneric phylogeny was needed.
In this study, Goodyera turns out to be polyphyletic, in contrast with previous studies [17,18], which included only a small number of samples from other genera of Goodyerinae. Our analysis has revealed that there are at least six genera viz., Erythrodes, Kreodanthus, Microchilus, Platythelys, Lepidogyne and Hylophila that have a close relationship with Goodyera. In all our trees, these genera are embedded within the Goodyera clade. Erythrodes, Lepidogyne and Hylophila do share some morphological characters (such as lack of marked leaf surface) with Goodyera subsect. Otosepalum. Erythrodes, Lepidogyne, Hylophila and subsect. Otosepalum are all distributed in Malaysia, Indonesia, Philippines and some other tropical Asia areas. However, Erythrodes differs in the flowers, which always have a spurred lip lacking the numerous setose appendages often present in different species of Goodyera. Hylophila is distinguishable from Goodyera by its flowers with a large scrotiform lip, whereas Lepidogyne differs from Goodyera by its short, thick stems, and a projecting plate under the stigma of flowers. Platythelys differs in having smaller, fleshy flowers with a broad, flat, elliptic to suborbicular rostellum and the species of this genus are restricted to the tropics and subtropics of the New World. Further research is needed to clarify the relationships of Goodyera with other genera in Goodyerinae.
Our results are identical with previous researches, such as Juswara [20]. In her phylogenetic tree inferred from ITS, trnL-F and rpl16, Goodyera turned out to be polyphyletic. But it included only small number of species and the placement of G. vittata is totally different from us, which could be a case of misidentifications in her research. and because of sample limitation. So, further studies with more samples especially other genus of Goodyerinae are still needed.

Infrageneric relationships of Goodyera
According to morphological characters, a small number of Goodyera species were assigned to two newly established sections [10]. Chung [18] divided the Goodyera species in Taiwan into three sections and seven subsections based on ITS and chromosome numbers ( Table 1). The relationships revealed by three DNA loci in this study (Fig 1, S1 and S2 Figs) do not agree with the infrageneric classification of previous studies [10,18].
The species Goodyera procera formed a clade (Clade O in Fig 1) with high support. G. procera has narrowly ovate-elliptic leaves, spike inflorescence with very dense flowers that are not usually secund and mainly grows beside streams in forests instead of in humus like most other Goodyera species. In all our molecular analyses, G. procera did not form a clade with other species of sect. Otosepalum and sometimes has different chromosome numbers (2n = 22 or 42) with respect to other species of sect. Otosepalum (2n = 22 or 44). It should be treated as a separate section by taking all the molecular and morphological differences into consideration (H.Z. Tian & C. Hu, unpublished manuscript).
Five species are included in clade N (Goodyera thailandica Seidenf., G. fumata, G. rubicunda (Blume) Lindl., G. seikoomontana Yamam. and G. viridiflora). As we get from the phylogenetic trees, these species of Goodyera have a very close relationship with some other genera (Erythrodes, Lepidogyne, Hylophila and Platythelys). These species vary considerably in their flowers and vegetative parts. Goodyera seikoomontana and G. viridiflora share many similarities (big flowers, reflexed lateral sepals and long pollinia) and differ a lot from the other three species. The flowers of G. fumata, G. rubicunda and G. thailandica do share a similar size and shape of leaves and inflorescence with Erythrodes but differ in their absence of a spur. In this study, we assign these five species to sect. Otosepalum of Goodyera because of their saccate labellum. Further studies such as morphology and molecular phylogeny including more species within Goodyera and related genera need to be conducted to clarify their relationships.
In Fig 1, most species of clade Q have silver or gold veins and the lateral sepals are not opened. Clade Q is related to sect. Reticulum S.W. Chung & C.H. Ou. It is split into two well supported clades (Q 1 and Q 2 ). Most species of clade Q 1 have reticulate venation on leaves, lateral sepals are not opened and lip sacs do not extend up to the lateral sepals. Morphologically, these species show marked differentiating characters and can be easily distinguished from each other as well as other species of Goodyera. Chung [18] established two subsections (subsect. Biflora and subsect. Reticulum) based on ITS for these species with a single species G. biflora forming subsect. Biflora. This is not supported by any of our ITS, plasted or combined analyses (S1 Fig). Goodyera biflora was included in subsect. Biflora by having long tubular flowers and reticulate venation on the leaves [18]. In fact, between the long tubular flowers of G. biflora and the short tubular flowers of G. hispida, there are some transitional species such as G. vittata and G. hemsleyana. So this well supported clade Q 1 can be merged into one subsection (subsect. Reticulum).
Clade Q 2 is related to subsect. Foliosum. Some problems still remained in this clade. Goodyera foliosa formed two strongly supported clades with G. velutina and G. henryi respectively. Though G. velutina can be easily distinguished from G. foliosa by its white or pink mid-vein on leaf, the phylogenetic trees show complex relationships among them, which requires further studies.
All species in clade R can be placed in sect. Goodyera by morphological data and chromosome number (2n = 30 or 60). Only in this clade are there some species epiphytic, such as G. bilamellata, G. pendula, G. pranii, and sometimes G. schlechtendaliana.
Clade R split into two clades, R 1 and R 2 . Clade R 1 contained five species known as a species complex that share many similarities: white flowers like a flying dove and white marking on the upper surface of leaves (G. billamelata and G. robusta are exceptions). Due to the lack of abrupt interspecific variations, it is difficult to distinguish the species of this complex. In the previous studies [4,5], some species were merged. But because of lacking molecular data, those treatments are debatable. In this study, the phylogenetic trees show that this complex is not well resolved, and more samples and studies are needed. But it can be confirmed is that these species form a well-supported clade and should be separated from subsect. Goodyera by establishing a new subsection (H. Z. Tian & C. Hu, unpublished manuscript).
All species in clade R 2 share two features: almost glabrous inside the labellum and distribution at higher elevations. It is reasonable to treat R 2 as a unique subsection merged with species of subsect. Recurvum. The name of this subsection should be Goodyera because the type species of the genus Goodyera, G. repens, forms part of this clade.

Biogeography of Goodyera
The pollinarium fossil of Meliorchis caribea from Dominican Republic indicated that a minimum age of 15-20 Myr can be assigned to the subtribe Goodyerinae [42].
The genus Goodyera is worldwide and species are mostly distributed in temperate and tropical regions of Asia. Tian [15] divided the distribution patterns of Goodyera species from China into six types (North Temperate, Tropical Asia, Tropical Asia to Tropical Australasia, Tropical Asia to East Asia, East Asia and Endemic to China). As we can see from Fig 2, the new section of G. procera shows Tropical Asia distribution (Fig 2B). Sect. Otosepalum is mainly distributed in Tropical Asia and some species such as G. rubicunda, G. fumata southwards to Tropical Australia and Pacific Islands (Fig 2C). Meanwhile, due to the tropical habitat of these species, they are always tall and robust. Sect. Reticulum ranges from Tropical Asia to East Aisa (Fig 2E). The distribution of sect. Goodyera reflects almost all part of the genus ( Fig 2D) and this section comprises many alpine species as well as epiphytic species. Goodyera repens is the most widely distributed species in Goodyera. Tsiftsis and Papaioannou [43] pointed out that the southern distribution limits of G. repens is indirectly affected by soil variables through the establishment of mycorrhizal symbiosis and its distribution is found to be negatively correlated with the nutrient content of the soil. It is worthy to be mentioned that the microhabitat is really important for species differentiation in Goodyera. Most Chinese endemic species in subsect. Goodyera are affinis with G. repens and mainly distributed in Hengduan Mountain with relatively high altitude.
To know the geographic tracks of Goodyera species, more samples of Goodyerinae especially species from other parts of the world are needed. More comprehensive biogeographic analysis such as divergence time and ancestral area reconstructions can reveal details about this genus.

Conclusions
According to this study, Goodyera is polyphyletic and can be divided into four sections: sect. Otosepalum characterized by leaves without markings and lateral sepals reflexed; the single species G. procera forms a new section and characterized by green leaves and nearly spicate inflorescence; sect. Reticulum with subsect. Reticulum (having golden venation on the leaves) and subsect. Foliosum containing G. foliosa, G. henryi (three pale veins on the leaves) and G. velutina (one white, golden or pink mid-vein); sect. Goodyera with subsect. Goodyera (smaller flowers with almost glabrous lip sac) and a new subsection (white dove-like flowers with papillose lip sac). Further studies are needed especially in some species complexes.