Dendrochilum hampelii (Coelogyninae, Epidendroideae, Orchidaceae) traded as ‘Big Pink’ is a new species, not a hybrid: evidence from nrITS, matK and ycf1 sequence data

Abstract In 2013, an unidentified species of Dendrochilum appeared in cultivation under the commercial trade name ‘Big Pink’. Using sequences of the nuclear ribosomal ITS1-5.8S-ITS2 region and of the plastid matK and ycf1 genes, we examined the phylogenetic relationships between ‘Big Pink’ and six other species of the phenetically defined Dendrochilum subgen. Platyclinis sect. Eurybrachium. Separate and combined analyses (using Bayesian, Maximum Likelihood and Parsimony inference) showed consistent placement of the unidentified species within a statistically well supported clade. Furthermore, the multi-copy nrITS marker showed clear distinct peaks. Thus, we found no evidence that ‘Big Pink’ could be a hybrid. Against this background, and further supported by species-specific mutations in (at least) nrITS and ycf1, we formally describe ‘Big Pink’ as a new species under the name Dendrochilum hampelii. Morphologically, it is most similar to Dendrochilum propinquum, but it differs in a number of characters. Of the two cultivated individuals available for our study, one was of unrecorded provenance. The other allegedly originated from the Philippines. Observations of the species occurring in the wild in the Philippines in the northern provinces of Bukidnon and Misamis Oriental on the island of Mindanao confirmed this.


Introduction
Th e largely Malesian genus Dendrochilum Blume (Blume 1825(Blume -1827published 1825) accommodates ca. 275 species-most of them described based on fi eld inventories in the periods ca. 1900-1940and ca. 1985-2000(Pedersen 2007a). However, since the turn of the millenium, most new species of Dendrochilum have been described based on cultivated material of unrecorded provenance (cf. Pedersen 2011). Following the publication of protologues based on cultivated material, four of these species have been located in the wild (Cootes 2011). Th is is the case for D. coccineum H.A. Pedersen & Gravend. (Pedersen et al. 2004), D. croceum H.A.Pedersen (Pedersen 2005), D. quinquecallosum H.A.Pedersen (Pedersen 2007b) and D. undulatum H.A.Pedersen (Pedersen 2007b). Th e discovery of D. coccineum in the Philippines confi rmed a phylogeny-based hypothesis put forward by Pedersen et al. (2004). Similar phylogenetic inference of D. warrenii H.A. Pedersen & Gravend. (Pedersen et al. 2004) as probably originating from the Philippines and/or Sulawesi (Pedersen et al. 2004) is still awaiting confi rmation.
Th e formal description of new species of unknown natural distribution has undoubtedly served to stimulate the (partly successful) search for these species in the wild, thus demonstrating the relevancy of this practice-not least in a conservation context. Nevertheless, describing new species based on material in commercial trade also involves a few problems. Th us, Vermeulen et al. (2014) raised some issues of moral concern, and Pedersen (2011) emphasized that care should be taken not to accidentally describe artifi cial hybrids as new species. Until recently, this risk was negligible in connection with Dendrochilum since not a single Dendrochilum hybrid had been registered (cf. Pedersen 2011). However, now there are two Dendrochilum artifi cial hybrids registered in Th e International Orchid Register (http://www.apps.rhs.org.uk/horticulturaldatabase/ orchidregister/orchidregister.asp; accessed on 13 February 2015), originating from Bornean parental species. Th is demonstrates that human assisted hybridization between diff erent species of Dendrochilum is indeed possible-for which reason we must consider the possibility that seemingly undescribed species suddenly appearing in cultivation could in reality be artifi cial hybrids.
Since chloroplast DNA is usually maternally inherited and nuclear DNA is biparentally inherited in orchids, incongruences between nuclear and plastid gene trees might indicate past events of hybridization. For example, by comparing phylogenies based on cpDNA and nrDNA, Barkman and Simpson (2002) inferred a natural hybrid origin of D. acuiferum Carr whereas Gravendeel et al. (2004) detected natural hybrids within the more distantly related genus Pleione D.Don. Further evidence for hybridization could potentially come from signifi cant within-individual variation in multi-copy DNA markers.
Th is paper reports our study of an unidentifi ed Dendrochilum (trade name: 'Big Pink') that appeared in cultivation in 2013. A live plant presented to the Hortus botanicus in Leiden carried a tag indicating a Philippinese provenance. However, as the plant came from a commercial nursery that trades much material of unknown geographic origin, we felt this provenance was in need of verifi cation.
Synanthous infl orescences in combination with an entire rostellum, presence of stelidia and an apical wing on the column place the study plant in the phenetically defi ned subgenus Platyclinis Engl. as circumscribed by Pedersen et al. (1997). Within this subgenus, a fi rmly attached entire labellum and a stout and straight column with stelidia but without a foot place the plant in the phenetically defi ned section Eurybrachium Carr ex J.J.Wood, H.A. Pedersen & J.B.Comber (Pedersen et al. 1997). However, the morphology of 'Big Pink' does not match any previously described species in section Eurybrachium-implying that it should be formally described as a new species, provided it is not either an artifi cial or natural hybrid.
Altogether, we decided to examine 'Big Pink' in a molecular phylogenetic framework-and to describe it as a new species, if the results of the phylogenetic study could reject the possibility of 'Big Pink' being a hybrid.

Plant sampling and DNA extraction
Th e possible hybrid status of 'Big Pink' was tested using a molecular phylogenetic approach based on three markers, namely the biparentally inherited multi-copy nuclear ribosomal internal transcribed spacer (nrITS), and the maternally inherited plastid matK and ycf1 genes. Th e ingroup consisted of 'Big Pink' and six other species belonging to Dendrochilum subgen. Platyclinis sect. Eurybrachium (cf. Pedersen et al. 1997), see Table 1.
A live plant of 'Big Pink' was available from the Hortus botanicus in Leiden, whereas the remaining Dendrochilum plants sampled for this study were reared in the Botanical Garden, Natural History Museum of Denmark. For information on vouchers, see Table 1. Th unia alba Rchb.f. was chosen as outgroup, based on the placement of the genus Th unia as sister to Dendrochilum using sequences from nrITS, matK, trnL-F, and rbcL (Goldman et al. 2001, Gravendeel et al. 2001, van den Berg et al. 2005. Total genomic DNA was obtained from 50 mg of silica dried or fresh leaf tissue. In the case of 'Big Pink', the tissue was mechanically reduced to dry powder using liquid nitrogen; for all other taxa, it was ground in Lysing Matrix A tubes (MP Biomedicals) and extracted using the DNeasy Plant Mini Kit (Qiagen) following the manufacturer's protocol.

Amplification and Sanger sequencing
Th e ITS1-5.8S-ITS2 region of the nuclear ribosomal internal transcribed spacer (nrITS) was amplifi ed using primers 17SE (ACGAATTCATGGTCCGGTGAAGT-GTTC) and 26SE (TAGAATTCCCCGGTTCGCTCGCCGTTAC), as described by Sun et al. (1994). Subsequently, a M13 universal sequencing primer was added to the 5' end of the forward (TGTAAAACGACGGCCAGT) and reverse (CAGGAAACA-GCTATGAC) primers to improve Sanger sequencing effi ciency. Each PCR reaction consisted of 25 μl, containing the template DNA, CoralLoad PCR buff er (Qiagen), dNTPs, Taq DNA Polymerase (Qiagen), and both primers. Th e PCR reactions were carried out using a MyCycler Th ermal Cycler (Bio-Rad) or a C1000 Touch Th ermal Cycler (Bio-Rad). Th e thermal cycling protocol began with 5 min initial denaturation at 95 °C followed by 35 amplifi cation cycles, each with 30 sec denaturation at 95 °C, 30 sec annealing at 50 °C, and 1 min extension at 72 °C, which were concluded by a 7 min fi nal extension at 72 °C. Th e primers for the amplifi cation of the chloroplast matK region were also used by Gravendeel et al. (2001), and specifi ed as follows: 19F (CGTTCTGACCAT-ATTGCACTATG) and 881R (TMTTCATCAGAATAAGAGT), 731F (TCTG-GAGTCTTTCTTGAGCGA) and 2R (AACTAGTCGGATGGAGTAG). Th e PCR mix for the amplifi cation of matK followed that of nrITS, but with additional BSA. Th e thermal cycling protocol for matK PCR began with 5 min initial denaturation at 94 °C followed by 28 amplifi cation cycles, each with 30 sec denaturation at 94 °C, 30 sec annealing at 49 °C, 1 min extension at 72 °C, and ended with a 7 min fi nal extension at 72 °C.
Th e 3' end portion of the chloroplast ycf1 region was amplifi ed using primers newly designed in this study. Th e design was based on the ycf1 sequences data set of Neubig et al. (2008) available on NCBI GenBank (http://www.ncbi.nlm.nih.gov/ GenBank), specifi cally from the species of subfamily Epidendroideae. Th e sequences were aligned using Geneious 5.6.7 (Kearse et al. 2012), and conserved regions were identifi ed to be used as annealing sites. Th e ycf1 sequences produced in this study had a complete marker size of approximately 1.5 kb. Th e ycf1 region was amplifi ed using a Hot start PCR protocol with primers d147F (TGCAGCRAATTYATTTAT-GAGTC) and intR2 (GATTTGATTGGGATGATCCAAGG), d557F (TCAAGA-GATCAAACCATKCAATCA) and 1560R (CTCTACGACGTCTGGGAGATAG). Each PCR reaction consisted of 25 μl, containing DNA template, Phire Hot Start II DNA Polymerase and Phire buff er (Th ermoScientifi c), dNTPs, BSA, and both primers. Th e cycling condition started with 30 sec initial denaturation at 98 °C and followed by 30 cycles of 5 sec denaturation at 98 °C, 5 sec annealing at 65.5 °C, and 10 sec extension at 72 °C, concluded with a 1 min fi nal extension at 72 °C.
Sanger sequencing of the amplifi cation products was performed at Baseclear (http://www.baseclear.com/), using an ABI 3730xl sequencer (Applied Biosystems). All new sequences are deposited in NCBI GenBank (Table 1). All sequences of T. alba were obtained from NCBI GenBank.

Phylogenetic analyses
Raw Sanger sequencing results in the form of AB1 fi les were edited and contigged using Sequencher 5.3 sequence analysis software (http://www.genecodes.com). Th e ends of all data sets were trimmed to avoid character misinterpretation. Ambiguous bases were replaced with "N" in the data matrix. Th e sequences were aligned using Geneious multiple sequence alignment in Geneious 5.6.7 (Kearse et al. 2012) with subsequent manual adjustments. Missing data were replaced with "?".
Phylogenetic analyses were carried out by means of Maximum Parsimony and Maximum Likelihood using PAUP* and Bayesian methods using the software Bayesian Evolutionary Analysis and Sampling of Trees (BEAST ver. 1.8.0; Drummond et al. 2012). Both PAUP* and the BEAST program were used to analyze the nrITS, the combined plastid matK + ycf1, and the combined nrITS + matK + ycf1 data matrices. For the Bayesian analysis, the substitution and clock model was set as unlinked, and the chosen nucleotide substitution model was General Time Reversible (GTR), plus Gamma with 10 categories. Th e best fi t substitution model for each partition was determined using the Find Model web tool (http://www.hiv.lanl.gov/content/sequence/fi ndmodel/fi ndmodel.html). A lognormal relaxed clock model was used for each partition, and the chosen tree prior was the Yule speciation process (Gernhard 2008). Tree samplings were generated through Markov Chain Monte Carlo (MCMC), with the number of generations set to 20,000,000 and a tree sampling for every 1000 generations. Th ree consecutive replicates were done to assess the consistency of the method. Th e three consistent replicates were then combined into a single matrix using LogCombiner c1.8.0 (http://beast.bio. ed.ac.uk/logcombiner) and used to search for the best probable tree using the program TreeAnnotator ver. 1.8.0 (http://beast.bio.ed.ac.uk/TreeAnnotator) with a 20% burnin value to avoid the reduction of posterior probability (PP) values, and visualized using FigTree ver. 1.4.2 (http://tree.bio.ed.ac.uk/software/fi gtree/).

Results
Th e nrITS sequence alignment contained 854 positions with a mean ungapped length of 834 bp. Included in the alignment were the nrITS1 (236 positions), 5.8S RNA (166 positions), and mrITS2 (253 positions) regions. In total, the number of variable sites for the included positions was 164 (19.2%), of which 39 were potentially phylogenetically informative. Mean pairwise distances within the ingroup varied from 0.2-6.9%. Th ere were six synapomorphic indels, with a size ranging from 1-5 bp. All the sequences included in the nrITS matrix were complete except for T. alba, which lacked 110 characters.
Th e matK matrix was characterized by a fairly high number of missing data, mostly due to amplifi cation failures. Samples lacking approximately half (800 bp Th e ycf1 alignment consisted of 1,297 positions with mean ungapped length of 1065 bp. Th e number of variable sites was similar to that of nrITS, 244 (18.8%). Out of these positions, 30 were phylogenetically informative. Th ere was one synapomorphic indel with a size of 9 bp. Mean pairwise distances within the ingroup varied from 10-33.5%. Amplifi cation failures for D. septemnervium resulted in almost half of the desired ycf1 marker missing for this species.
Th e phylogenetic trees based on the combined matK + ycf1 matrix obtained by all three methods yielded overall stronger branch support relative to that of nrITS (Fig.  1A). Th e internal nodes gained high (PP = 0.97-1.0) support values, and the preceding replicates showed consistent topologies as well as well-supported clades. Clades presented by the matK + ycf1 tree were highly congruent with the nrITS tree, and with higher support value.
No hard incongruence was present between the nrITS and the plastid phylogenetic trees obtained by all three methods. Th e combined nrITS + matK + ycf1 matrix yielded a single tree with highly consistent topology and strong support values; the same clades as those in the separate nrITS and matK + ycf1 analyses were identifi ed (Fig. 1B). In all our trees, 'Big Pink' was positioned as a sister to D. tortile and D. septemnervium, which were very closely related, in terms of relative branch length (Fig. 1).  Analysis of the nrITS sequence alignment revealed a species-specifi c mutation of 'Big Pink' at position 567 (Fig. 2). Electropherograms of the ITS sequences of 'Big Pink' and its three closest relatives among the study species showed clear, distinct signals for positions included in the alignment, including the 'Big Pink' species-specifi c mutation site (Fig. 2).

Discussion
Nuclear sequence variation found was largely in agreement with previous studies in Coelogyninae (Gravendeel et al. 2001). In general, sequence divergence of nrITS between diff erent species of Dendrochilum is known to vary between 0.2-20% (Pedersen et al. 2004;Sulistyo, unpublished data). Additionally, the nrITS sequence obtained from D. hampelii is not likely to be a paralogous locus since branch lengths are similar to the sequences of other bona fi de species analyzed. Th e position of 'Big Pink' was in agreement with preliminary analyses based on larger sampling (Sulistyo, unpublished data), as well as with the analyses based on diff erent sequences. We found no evidence suggesting that the morphologically distinctive 'Big Pink' could be an artifi cial hybrid rather than a new wild species. Firstly, the separate phylogenetic analyses of nuclear and plastid markers proved highly congruent. Secondly, the electropherogram of the nrITS sequence of 'Big Pink' showed distinct single peaks with no indication of any heterozygosity of the specimen examined.
Genetically, 'Big Pink' possessed a number of automorphic mutations compared to the other species included in our small phylogenetic study. Th ese unique mutations were found in nrITS and ycf1, whereas one in matK is in need of verifi cation due to the alignment being characterized by a fairly high number of missing data. Morphologically, 'Big Pink' is most similar to D. propinquum Ames. Unfortunately, no material was available of D. propinquum for DNA sequencing for this study. However, 'Big Pink' has much larger fl owers (approximately twice the size of those of D. propinquum), its fl owers have petals that are 1.4-1.5 times as broad as the sepals (0.8-1.1 times in D. propinquum), its labellum is broadly cordate (broadly elliptic to ovate in D. propinquum), and the stelidia of its column are acute (obtuse in D. propinquum). Against this background, we describe 'Big Pink' as a new species below.
Contrary to the studies of Wanntorp et al. (2002) and Pedersen et al. (2004), the taxon sampling underlying the phylogenetic analyses in this study was too small to allow for any inference of a probable geographic origin of 'Big Pink'. However, in a preliminary study based on systematically and geographically much broader sampling of Dendrochilum s.l., individual phylogenetic analyses using sequences from nrITS, plastid matK, and ycf1 have all indicated 'Big Pink' to be nested within a clade nearly exclusively consisting of Philippine endemics (Sulistyo, unpublished data). In subsequent fi eld surveys by the second author, a total of 12 plants were observed in February 2014 in the northern provinces of Bukidnon and Misamis Oriental on the island of Mindanao at high elevation, a few of which were fl owering. In March 2015, only 4 individuals were observed in situ in the wild by the second and third author, and none were fl owering. Th ese observations confi rm that D. hampelii is indeed a Philippinese, almost certainly endemic species. It is conjectured to have entered cultivation in Europe through the many domestic markets in Southeast Asia that sell orchid species. For these markets, wild plants are harvested but traded as 'cultivated' to circumvent CITES legislation. All Coelogyninae are listed on Appendix 2. Despite this legal protection, illegal trade is continuing at international orchid shows and by web based orders from buyers of specifi c species or nursery owners hoping to incorporate desirable wild traits into new hybrids (Hinsley et al. 2015).

Taxonomic treatment
Based on these results, it is determined that "Big Pink" is a new species in need of recognition. Formally naming the species is relevant for horticulture and ex situ conservation, because the name provides an unambiguous way to refer to the species.Th e morphology of 'Big Pink' was described using terminology of the vocabulary and list of individual absolute terms in Stearn (1983), if relevant standardized according to the Orchidaceae glossary in Pedersen et al. (2011). Diagnosis. Th is new species is similar to D. propinquum Ames, but is distinguished by its larger fl owers with petals proportionally broader (1.4-1.5×) than the sepals, a broadly cordate labellum (6.8-8.0 × 7.2-7.6 m) and acute stelidia.
Etymology. Th e specifi c epiphet honours Georg Hampel, who was one of the fi rst to provide us with study material of the newly described species.
Distribution and ecology. Th e species occurs in the wild in the Philippines in the northern provinces of Bukidnon and Misamis Oriental on the island of Mindanao (Fig. 4C). It grows as an epiphyte at elevations approximately 1,200 m above sea level among mosses on the trunks and branches of trees. Fresh fl owers of plants observed in the wild were pale yellow whereas fresh fl owers of the cultivated plants studied were pinkish salmon-coloured. We do not consider this reason to describe them as a diff erent variety or forma as color dimorphism is known to occur in other Coelogyninae as well (Gravendeel 2000).
Reproductive biology. Th e live plant in Leiden fl owered in mid-December. Attempts to pollinate fl owers of D. hampelii were made using pollinia from the same fl ower and pollinia from a diff erent fl ower in the same infl orescence. None of these eff orts led to fruit formation. Th is indicates that D. hampelii is probably self-incompatible, as previously demonstrated for D. longibracteatum Pfi tzer (Pedersen 1995), although it should be noted that experimental pollination was severely challenged by the small size of the stigmatic cavity.
Conservation status. Although the species occurs in cultivation we as yet know very little about the distribution and abundance of D. hampelii in the wild. As such, we recommend the species to be considered for the Data Defi cient category of the IUCN Red List of Th reatened Species (IUCN 2012).