Phylogenetic studies in the Hoya group (Apocynaceae, Marsdenieae): the position of Anatropanthus and Oreosparte

Recent molecular phylogenetic studies have shown that Hoya is paraphyletic without Absolmsia, Cle­ mensiella, Madangia, and Micholitzia. These genera have been placed in synonymy with Hoya, but the monophyly of Hoya sensu lato relative to other genera of the broader Hoya group (Dischidia, Anatropanthus and Oreosparte, the latter two never included in a molecular phylogenetic analysis) remained unclear. Furthermore, no analysis has included both a significant sample of the Hoya group and outgroup genera of Marsdenieae to test the monophyly of the Hoya group and its position within the tribe. To address these gaps, we assembled two data sets: (1) the chloroplast trnT­trnL­trnF locus from 110 species and (2) three chloroplast loci (trnT­trnL­trnF, atpB­psbA spacer and matK) and two nuclear loci (nrDNA ITS and ETS) from 54 species. The Hoya group is monophyletic and nested in an Asian/Australian clade of Marsdenia s.l. The genus Hoya is paraphyletic unless Anatropanthus, Dischidia and Oreosparte are included. However, current evidence is not sufficient to synonymize Dischidia and Oreosparte with Hoya. Support for synonymy of Anatropanthus with Hoya is strong and the new name H. insularis is proposed. A clade of three new species with Oreosparte-like morphology is sister to the rest of the Hoya group and is described as the new genus Papuahoya.


Introduction
Hoya R. Br. (Fig. 1A, B) (Marsdenieae, Asclepiadoi deae) is the largest genus in Apocynaceae, comprising 350 -450 species of Asian and Australasian succulent epiphytic and terrestrial vines and shrubs (Rodda 2015) that are highly prized horticulturally for their showy "wax" flowers. The flowers of Hoya are characterized by the presence of a staminal corona with the proximal (apical) part of the lobe entire, distal (basal) part of the lobe with an "anther skirt" and revolute margins containing nectaries. The anther guide rails lack inner edges and the pollinia have a pellucid margin along the outer edge (Wanntorp & Kunze 2009;Endress & al. 2019).
The Hoya group -Hoya has been associated with a number of smaller genera including Absolmsia Kuntze (1 sp.) (Fig. 1E,F) in what has come to be termed the "Hoya group" (Omlor 1996(Omlor , 1998Wanntorp & Forster, 2007;Meve & al. 2009). A combination of characters separate these genera from the rest of Marsdenieae including succulent and epiphytic growth form (vs. woody and terrestrial), highly condensed, persistent inflorescences that re-flower multiple times before they senesce (vs. expanded, once-flowering deciduous inflorescences), valvate corolla lobes in bud (vs. contor ted), and narrow, spindle-shaped seeds (vs. flattened, ovate seeds) (Omlor 1998). However, many of these diagnostic characters are variable within the Hoya group (e.g. H. coronaria Blume is a terrestrial vine; H. lanceolata Wall. ex D. Don has ephemeral inflorescences).
The Hoya group has always received strong support as monophyletic in molecular phylogenetic analyses that have sampled at least two of the included genera along with other Marsdenieae genera (Potgieter & Albert 2001;Verhoeven & al. 2003;Meve & Liede 2004;Wanntorp & al. 2006aWanntorp & al. , 2006bSurveswaran & al. 2014). However, the studies with the best sampling of Hoya group taxa have had the sparsest sampling of other Marsdenieae genera ) and vice versa (Verhoeven & al. 2003;Meve & Liede 2004).
In the most comprehensive analysis of the Hoya group to date (Wanntorp & al. 2014), Hoya was not supported as monophyletic relative to Absolmsia, Clemen siella, Dischidia, Mandangia or Micholitzia. The larger infrageneric sampling of Hoya sensu stricto in Wanntorp & al. (2014: fig. 3, 4) does not allow evaluation of the monophyly of Hoya relative to Dischidia because the latter was designated as the outgroup. In a taxonomically reduced dataset (their fig. 5A), Dischidia is nested within Hoya but with minimal bootstrap support. All of these genera except Dischidia have been synonymized with Hoya (Wanntorp & Forster 2007;Wanntorp & Meve 2011). Floral characters including pollinia with pellucid margins, anther guide rails without inner edges, and presence of nectaries inside the revolute margins of the outer processes of the staminal corona lobes, i.e. the "anther skirt", support the inclusion of Absolmsia, Mandangia and Micholitzia in Hoya s.l. (Wanntorp & Forster 2007;Wanntorp & Kunze 2009). Clemensiella lacks these traits but shares similarities of terrestrial growth form and pollinium morphology with species of the atypical H. sect. Eriostemma Schltr. (Wanntorp & Meve 2011). With the inclusion of Clemensiella, Hoya s.l. becomes very heterogeneous, with a circumscription that includes most of the morphological variation within the larger Hoya group. This extension of Hoya mirrors a trend toward expansion of generic boundaries consequent to molecular phylogenetic analyses of other large Apocynaceae genera including Ceropegia L. (Bruyns & al. 2015), Cynanchum L. (Khanum & al. 2016) and Vincetoxicum Wolf (Liede-Schumann & al. 2012).
Anatropanthus Schltr. and Oreosparte Schltr. -Anatro panthus, Heynella and Oreosparte are the most poorly known genera of the Hoya group (Omlor 1998). No authentic material of these three genera was available for study until O. celebica Schltr. was neotypified (Rodda & Omlor 2013) and a detailed description and illustrations were provided. Oreosparte shares all the diagnostic characters of the Hoya group, but its corona and pollinaria are distinct from both those of Dischidia and Hoya. Oreosparte celebica has erect staminal corona lobes with bifid apices and without revolute margins; the pollinarium has very narrow caudicles and pollinia without pellucid margins. In contrast, Hoya has corona lobes with entire apices and generally revolute margins, while Dischidia typically has inverted anchor-shaped staminal corona lobes. The pollinaria of both Dischidia and Hoya generally have well-developed, broad caudicles and the pollinia of Hoya typically have pellucid margins. Other undescribed species with the "Oreosparte floral phenotype", i.e. presence of urceolate corollas and stipitate gynostegium with erect corona lobes, have been discovered in herbaria or collected in the field and are also sampled. Anatropanthus borneensis Schltr. was known only from its type specimen collected in 1901 in Sarawak, Malaysia, and destroyed during the fire of the Berlin Herbarium in 1943 (Hiepko 1978;Nicholas 1992). It was recently collected in Kalimantan, Indonesia. Its flowers have a peculiar, curved pedicel that is similar to that of H. retrorsa Gavrus & al. The corolla is unique in the Hoya group: tubular, apically inflexed and terminating in erect, lanceolate lobes (Fig. 2). The gynostegium has oblong, concave corona lobes, erect and attached at the back of the anthers. The pollinia are erect, with a pellucid margin all along the outer margin as generally seen in Hoya. Given the morphological heterogeneity of Hoya, a molecular phylogenetic analysis is necessary to ascertain whether Anatropanthus and Oreosparte should be maintained as separate genera from Hoya.
Hoya sections -In addition to the difficulty of drawing the generic boundaries of Hoya, no complete infrageneric system has been published to date. Infrageneric groups in Hoya are circumscribed based on the shape of the corolla (campanulate, urceolate, rotate, revolute), the corona (size and shape of the staminal corona lobes and their inner and outer processes) and the pollinaria (size and shape of corpusculum, pollinia and caudicles, presence of pellucid margin of the pollinia). The first infrageneric classification of Hoya s.l. was published by Miquel (1856) Burton (1985Burton ( , 1995Burton ( , 1996aBurton ( , 1996bBurton ( , 1996c and Kloppenburg (1993Kloppenburg ( , 1994, who used up to 21 sections. A critical revision of the infrageneric classification of Hoya has never been published. While sections such as H. sect. Eriostemma are supported as monophyletic in molecular analyses, others such as H. sect. Cyrtoceras and H. sect. Plocostemma are not ). In the most recent phylogeny, Wanntorp & al. (2014) divided Hoya s.l. into six unnamed clades, some of which are diagnosable by morphology (e.g. growth form, pollinium and corona structure, nectar colour) and/or biogeogra-phy, but only two of these can be readily aligned with previously published sections: H. sect. Acanthostemma (Blume) Kloppenburg (Omlor 1998;Forster 2000;Livshultz 2003aLivshultz , 2003b. Combinations in Dischidia exist for species of all of these genera. Typical of the taxonomic history of other genera within Asclepiadoideae, most of these segregates were diagnosed by the divergent structure of the stami- nal corona relative to the membranous, inverted-anchorshaped lobes characteristic of Dischidia s.s. However, later taxonomists, again consistent with trends across the subfamily, recognized a diversity of corona morphologies within Dischidia s.l. (Rintz 1980;Livshultz & al. 2005). In a molecular phylogenetic analysis of 46 ingroup species and eight outgroup Marsdenieae species (including Hoya, Marsdenia R. Br. and Telosma Coville) based on the second intron of the nuclear gene Leafy, there was strong support for monophyly of Dischidia s.l. including Collyris, Conchophyllum, Dischidiopsis, Leptostemma and Oistonema (Livshultz 2003b).
The most frequently used infrageneric classification of Dischidia divides it into three sections based on leaf morphology: D. sect. Dischidia with unmodified, laminar leaves; D. sect. Conchophyllum (Blume) K. Schum. with concavo-convex, shell-shaped, ant-house leaves; and D. sect. Ascidiophora K. Schum. with dimorphic leaves, producing both unmodified, laminar leaves and pouchshaped, ant-house leaves (Livshultz 2003b). Molecular phylogenetic evidence indicated that D. sect. Dischidia is paraphyletic to a clade that includes all sampled species of D. sect. Ascidiophora and D. sect. Conchophyl lum, while relationships between the latter two taxa were unresolved (Livshultz 2003b).
In this study, we test (1) the monophyly and phylogenetic position of the Hoya group in an analysis that includes both a representative sample of Hoya group taxa and other Marsdenieae genera, and (2) the current circumscription of Hoya, specifically asking if there is sufficient evidence for expanding the synonymy of Hoya to include Anatropanthus, Dischidia and Oreosparte. We sample A. borneensis, O. celebica Schltr. and other putative Oreosparte species for the first time. We include a substantially expanded sample of Di schidia, including its type species D. nummularia R. Br., and of Marsde nia species relative to previous analyses Wanntorp & al. 2014).

Material and methods
Sampling Matrix 1 (110 taxa) -To test the position of the Hoya group and Oreosparte within Marsdenieae, we modified the trnTL spacer, trnL intron, trnLF spacer dataset of Meve & Liede (2004), which includes the largest generic sample of Marsdenieae published to date (9 of 27 currently recognized genera; Endress & al. 2019;Espírito Santo & al. 2019). We excluded the species of Periplocoi deae, which are only distantly related to Asclepiadoideae (Straub & al. 2014), the single unidentified Marsdenia species, and the Hoya group species, and then added 54 species of Marsdenieae, primarily of the Hoya group, and a sample of Vincetoxicum flexuosum (R. Br.) Kuntze (Asclepiadeae). We included 12 Dischidia species, representing the morphological diversity of the genus including morphologies diagnostic of the synonymized genera  Wanntorp & al. (2014) where the complete six-locus dataset was available and where the identity of the species could be verified by exa mining the voucher specimen. Early-diverging lineages of Hoya were more densely sampled than highly nested ones. For other Marsdenieae, we added one species of Jasminanthes, namely J. maingayi (Hook. f.) Rodda [Marsdenia main gayi (Hook. f.) P. I. Forst.] and four additional species of Marsdenia, including M. ridleyi P. I. Forst, a species that displays "Oreosparte floral phenotype", to increase sampling of this morphologically heterogeneous genus.
Sampling Matrix 2 (54 taxa) -In a second analysis, we reduced outgroup sampling and increased sequence sampling to investigate inter-and intrageneric relationships with the Hoya group. We limited the taxon sample to the 54 Marsdenieae species used in Matrix 1 and used Jas minanthes maingayi, Marsdenia flavescens A. Cunn. and M. rostrata R. Br. to root the tree. We added the chloroplast trnHpsbA spacer and part of the matK gene as well as the nuclear 5′-ETS and ITS loci to all samples.
The new specimens for the present study were obtained during fieldwork in Papua New Guinea, from the extensive living research collections at Singapore Botanic Gardens (Singapore) and Nong Nooch Tropical Botanical Garden (Thailand) and from herbarium specimens at E and SNP (herbarium codes according to In dex herbariorum; http://sweetgum.nybg.org/science/ih/). Identifica tion of specimens was carried out by consulting the relevant taxonomic literature including all protologues and comparing our collections with reference herbarium materials at the herbaria A, BISH, BK, BKF, BM, BRUN, FI, G, HBG, IBSC, K, KEP, KUN, L, M, MO, P, SAN, SAR, SING, SNP, TO, UC, US, W and WRSL. Vouchers are listed in Appendix 1 (in Supplemental Content online).
The PCR products were purified using Wizard® PCR and gel clean-up system (Promega Corporation, Madison, Wisconsin, U.S.A.), according to the manufacturer's recommendations. AITBiotech Pte Ltd, Singapore, performed sequencing. Forward and reverse reads were assembled with Geneious Version 8.0 (Biomatters LLC) and the new sequences deposited in GenBank (Appendix 1 in Supplemental Content online).
Alignment and matrix construction -Sequences of each locus were aligned with the ClustalW (Larkin & al. 2007) plugin in Geneious prime 2019.0.4 (https://www .geneious.com/) using default parameters and adjusted by eye to correct obvious mis-alignments. Regions of ambiguous alignment were removed with GBLOCKS (Talavera & Castresana 2007) run on the GBLOCKS server version 0.91b (http://molevol.cmima.csic.es/castresana /Gblocks_server.html). For the trnTtrnL and trnLF matrices, GBLOCKS was accessed on 5 May 2019 and sites selected using the following criteria: minimum number of sequences for a conserved position: 56; minimum number of sequences for a flanking position: 56; maximum number of contiguous non-conserved positions: 8; minimum length of a block: 5; allowed gap positions: with half. For ITS, ETS, psbAtrnH and matK, GBLOCKS was accessed on 5 May 2019 using the following selection criteria: minimum number of sequences for a conserved position: 28; minimum number of sequences for a flanking position: 28; maximum number of contiguous non-conserved positions: 8; minimum length of a block: 5; allowed gap positions: with half. Indels were not coded as characters because they are not modelled by the GTR family of models.
Incongruence -Each of the six loci was analysed independently, then concatenated into a nuclear matrix and a chloroplast matrix, and finally into a combined nuclear plus chloroplast matrix. Incongruence between the nuclear and chloroplast matrices and parsimony and ML analyses was assessed by identifying contradictory clades with moderate to high bootstrap support (BS > 75).
Parsimony tree searches, consensus tree calculation and bootstrap -Analyses were conducted with PAUP 4.0a (Swofford 2002). To find most parsimonious trees, a heuristic search with TBR branch swapping of 1000 random starting trees was conducted, saving up to 10 equally parsimonious trees per iteration, followed by swapping to completion on all equally parsimonious trees, or until 10 000 trees were saved. The resulting trees were used to construct a strict consensus. The bootstrap analysis consisted of 1000 resampled replicates, with TBR swapping on one random starting tree, saving a maximum of 20 equally parsimonious trees per replicate and calculating the strict consensus tree from each replicate.
Maximum likelihood tree searches and bootstrap -Analyses were conducted with RAxML 8.2.11 (Stamatakis 2014) as implemented on Geneious prime 2019.0.4 (https://www.geneious.com/). The GTR plus GAMMA model of nucleotide substitution was applied in all steps of the analysis. The search for maximum likelihood trees combined the tree search and the rapid bootstrap analysis (-f a) using 10 000 rapid bootstrap iterations followed by a tree search through ML. Datasets were not partitioned.

Results
Sequencing -In total 227 new sequences were generated for this study (Appendix 1 in Supplemental Content online), including 15 from species previously sampled by Wanntorp & al. (2014), two from Hoya corymbosa Rodda & Simonsson, previously sampled in , two from H. papaschonii Rodda, previously sampled in Rodda & Ercole (2014), and 209 from 34 newly sampled species.
Matrices -Summary statistics are shown in Table 1. Taxon sampling was complete for each locus. As judged from the number of aligned positions removed by GBLOCKS, the alignment of the psbAtrnH locus had by far the most gaps and alignment ambiguity (only 314 of 820 aligned positions retained for analysis, Table 1). For Matrix 2 (54 taxa), the ITS locus contributed the largest number of PICs (160) and the trnLF locus the fewest (37) ( Table 1).
Incongruence -There were no moderately to strongly supported (BS > 75) incongruences between parsi mony and ML analyses of any data matrix (data not shown). The only moderately to strongly supported incongruence between chloroplast and nuclear loci concerns the position of Dischidia milnei Hemsl., which was supported as sister to D. major (Vahl) Merr. by the cp loci (ML BS 90 ) versus sister to the rest of the ant-house-leaved Dischidia species by the nuclear loci (ML BS 97). In the combined analysis, D. milnei is placed in the position favoured by the nuclear loci but with poor support (ML BS 57) (Fig.  4). Support for the sister-group relationship of D. milnei and D. major appears to come primarily from the trnT trnL locus, which has two unambiguous synapomorphies Willdenowia 50 -2020 that favour this relationship. There are no unambiguous synapomorphies for this relationship in parsimony analyses of any of the other three chloroplast loci (data not shown).
Topology -ML topologies are shown in Fig. 3 and 4 with BS support (ML/parsimony) indicated at each node. Nodes absent from the parsimony strict consensus tree are indicated with "−". Only the ML BS support will be mentioned in the descriptions below. Analysis 2: 54taxon matrix (Fig. 4) -The inclusion of four additional loci resulted in greater resolution and support for relationships within the Hoya group than in the 110-taxon matrix (compare Fig. 3 and 4), and the topology of the Hoya group will be discussed based on the combined chloroplast and nuclear analysis (Fig. 4). To facilitate comparison, for the Hoya clades, we used the clade names of Wanntorp & al. (2014: fig. 3, 4) and highlighted the species that they sampled in bold italics in our Fig. 4.
Three taxa, including Hoya urniflora and two putative Oreosparte species form a strongly supported, monophyletic (BS 100) clade (Oreosparte I), sister to the rest of the taxa in the Hoya group that form a moderately supported clade (BS 79). Within this clade, four strongly to moderately supported clades can be recognized. The first (Oreosparte II, BS 100) includes the type of Oreosparte as well as Marsdenia ridleyi and another putative Oreo sparte sp. 3.
Oreosparte II is sister, but without support, to Di schidia s.l., which is strongly supported as monophyletic (BS 100). The Dischidia clade includes species originally described in the segregate genera Dischidiopsis  The rest of Hoya forms a moderately supported clade (BS 80) where eight of the subclades identified by  and Wanntorp & al. (2014) can be delimited. However, clade J has only one species, H. cumingiana Decne, and H. imperialis Lindl. is not included in any clade. All but clade IV are moderately to strongly supported.

Discussion
This study is the best-sampled analysis of the morphological and taxonomic diversity of the Hoya group conducted to date, including for the first time the enigmatic Anatropanthus and Oreosparte. The 110-taxon analysis (Fig. 3) is completely congruent with the tribal-level topology published by Meve & Liede (2004) and clearly shows that the Hoya group clade (BS 99), including Ana tropanthus, Dischidia s.l., Hoya s.l. and Oreosparte, is nested within Marsdenieae in a clade with other Asian and Australasian species. The Hoya group is paraphyletic unless one Marsdenia species (M. ridleyi) is included. By increasing sampling of Marsdenia s.l. from six to 11 species, our result also highlights the polyphyly of the current concept of Marsdenia s.l. (Forster 1995 Hoya group phylogeny and taxonomy -Generic delimitation within the Hoya group (Fig. 4) remains problematic. Our analysis shows much the same topology, with the exception of "clade P" sensu Wanntorp & al. (2014) nesting within clade VI (Fig. 4), and the same ambiguities, evident in the studies previously published using the same loci (Wanntorp & al. 2006;Rodda & Ercole 2014;Wanntorp & al. 2014). Dischidia s.l. is strongly supported as monophyletic (BS 100), but Hoya s.l. is unsupported. To complicate matters further, Oreosparte celebica and the species with "Oreosparte floral phenotype" sampled do not form a monophyletic clade but are subdivided into two clades. Clade Oreosparte I is sister to the rest of the Hoya group and includes H. urniflora and two new species from Papua New Guinea. Clade Oreosparte II is sister to Dischidia (Fig. 4) and includes the type of the genus as well as M. ridleyi and a new species from Borneo. Our analysis provides strong evidence that the "Oreo sparte floral phenotype" has also evolved independently in H. hamiltoniorum within clade III of Hoya s.l. The floral morphology of the former Clemensiella species is also very similar (Meve & al. 2009). While this lack of resolution among the primary branches of the Hoya group clade has been interpreted as evidence of a rapid radiation (Wanntorp & al. 2014), it may also be a matter of insufficient character sampling. For example, the position of Eustegieae had been controversial based on molecular matrices of few loci such as this one (sister to Ceropegieae plus Marsdenieae, BS 75, Fig. 3) or sister to Asclepiadeae (BS 76) (Surveswaran & al. 2014), but was resolved with high support as sister to Asclepiadeae in a plastome analysis (Straub & al. 2013). Taxonomic undersampling may also contribute to the lack of support (Zwickl & Hillis 2002). While we have sampled the geographic and morphological diversity of Hoya s.s., we still may not have sampled all early-diverging lineages, and we have not sampled Heynella. Oreosparte I and Oreosparte II are separated geographically, the first from Papua New Guinea, the second from West Malesia. Additionally, species of Oreosparte II have bifid corona lobe apices, whereas species of Oreo sparte I have entire corona lobe apices. We therefore recognize Oreosparte I as the new genus, Papuahoya Rodda & Simonsson.
Because of the lack of support for relationships among Oreosparte I and the Dischidia and Hoya clades, we consider the evidence insufficient for placing Oreosparte and Dischidia in synonymy with Hoya s.l. (Fig. 4). Willdenowia 50 -2020 Anatropanthus borneensis is nested within Hoya clade V with high support (Fig. 4). Its tubular corolla is very unusual, but corollas in Hoya can be particularly diverse and new species with unusual corollas are still being discovered, e.g. H. versteegii Simonsson & Rodda from New Guinea is the first species in the genus with an infundibuliform corolla with a long, narrow tube. Other characters of Anatropanthus are already found among Hoya species. The long, linear leaves of A. borneensis are similar to those of H. acicularis T. Green & Kloppenb., also from Borneo; the recurved pedicels are similar to those of H. retrorsa; and the pollinia have an evident pellucid margin, as commonly observed in the majority of Hoya species. Anatropanthus borneensis is therefore transferred here to Hoya.
Dischidia phylogeny and taxonomy -Phylogenetic relationships within Dischidia are congruent with those found by Livshultz (2003b) in an analysis of the nuclear second Leafy intron. Aside from the relationships of the ant-house-leaved species, discussed above, the phylogeny supports the recognition of Dischidia s.l., including the synonymized genera Conchophyllum (D. astephana, morphology similar to D. milnei), Dischidiopsis (D. pa rasita), Leptostemma (D. hirsuta, D. latifolia) and Oisto nema (morphology similar to D. latifolia), erected on the basis of atypical corona morphologies. The division into three sections based on leaf morphology is also not supported because both D. sect. Conchophyllum and D. sect. Dischidia are paraphyletic (Fig. 4). The sister-group relationship of two laminar-leaved species, D. antennifera and D. nummularia, with the ant-house-leaved clade (BS 100) is supported by a potential vegetative synapomorphy: presence of prominent wax chimneys around the stomata, particularly evident on the abaxial leaf surfaces, and a diagnostic floral character: absence of papillate epidermal cells on the adaxial surface of the corolla lobes. The sister-group relationship of D. latifolia and D. parasita is congruent with a number of morphological characters. Both species are relatively robust vines with larger leaves (compared to most other Dischidia species) with both opposite and alternate phyllotaxis; other potential synapomorphies include fleshy corona lobes with abaxial sulci and pollinaria with very short caudicles. While the larger clade that includes these two species plus D. acutifolia and D. tomentella is weakly supported (BS 72), it is consistent with the presence of alternate phyllotaxy in seedlings of D. acutifolia. Dischidia acutifolia and D. tomentella have similar floral and inflorescence morphology. Dischidia tomentella is endemic to karst in N Malaysia and S Thailand (Rintz 1980), often growing epilithically on exposed rock surfaces rather than epiphytically (Livshultz, pers. obs.). It may have evolved from isolated populations of the widespread, lowland species D. acutifolia that adapted to the more challenging edaphic conditions on karst via evolution of smaller, more succulent leaves, greater pubescence and slower growth.
Hoya s.l. phylogeny and taxonomy -Eight of the nine clades recognized in Hoya (Fig. 4) were also identified by Wanntorp & al. (2014). Hoya corymbosa and H. ignorata, forming clade X, were not sampled by Wanntorp & al. (2014). Our sampling of Hoya species is insufficient to provide a strong basis for an updated subgeneric classification of the genus, but five already recognized sections can be identified.  (4) clade M, with two Asian representatives. A much more comprehensive sampling including the type species of all the sections and subsections described so far would be necessary to verify whether any of these four clades represent a published infrageneric entity.

Conclusions
Our analysis is the first to include a comprehensive sampling of Anatropanthus, Dischidia, Hoya and Oreosparte without a significant amount of missing data, as well as numerous outgroups, in a comprehensive phylogenetic analysis. Anatropanthus is strongly supported as nested in Hoya within clade V (Fig. 4) and is here transferred to Hoya as H. insularis.
The available data show once again that Hoya is paraphyletic unless Dischidia and Oreosparte are synonymized (Fig. 4). However, the relationships among Hoya and Oreo sparte clade II and Dischidia s.l. are not supported. Current evidence is not sufficient to synonymize Dischidia and Oreosparte with Hoya. A phylogenomic approach is needed to clarify relationships among these taxa.
Oreosparte is strongly supported as belonging to the Hoya group (Fig. 3), but its species are separated into two clades, one of which is described as a new genus, Papuahoya. The Hoya group is placed within a grade of Asian and Australasian Marsdenieae (Fig. 3). Our results underline the polyphyly of the current concept of Marsdenia (Fig. 3). Rodda & S. Rahayu, nom. nov. (Fig. 2 Remarks -The type specimen of Anatropanthus borne ensis was lost in the fire that destroyed the Berlin Herbarium in 1943 (Omlor 1998). No duplicates have been traced and it is likely that only a single specimen was made -Schlechter (1908) stated "I found very little material in bloom". The illustration in the protologue (Schlechter 1908: t. 2) is therefore designated as the lectotype. Description -Epiphytic climber (occasionally hemi-epiphytic in mossy forest), with white latex in all vegetative parts. Roots basal and adventitious. Stems pubescent. Stipular colleters present, 1 at each side of base of petiole. Lamina lanceolate to ovate, stiff and chartaceous, pubescent turning glabrescent on old leaves, basal colleters present; venation pinnate.

Papuahoya bykulleana
Distribution -Known only from the type locality in Morobe Province of Papua New Guinea.
Ecology -Recorded at 1500 -1700 m on two ridges in primary mossy forests, where it grows on mossy ground, at the base of tree trunks near the ground or as an epiphyte. Papuahoya bykulleana is often hemiepiphytic as it starts growing in mossy areas, at the base of a tree or on moss-covered shrubs and continues growing tightly attached onto the tree trunk, or climbing on small shrubs, upward toward better-lit areas. It is absent either further up or down the mountain, even on the same slope.
Etymology -Named after Gunilla Bykulle of Sweden, who contributes to N. Simonsson's work in Papua New Guinea.
Distribution -Known only from the type locality in Morobe Province of Papua New Guinea.