From Polyalthia to Polyalthiopsis (Annonaceae): transfer of species enlarges a previously monotypic genus

Abstract The genus Polyalthiopsis Chaowasku (Annonaceae) was a poorly known monotypic genus from Vietnam that was recently segregated from the highly polyphyletic genus Polyalthia s.l. The sister clade relationship between Polyalthiopsis and Miliusa was not well established in previous study. The phylogenetic position of two Polyalthia spp. from China, P. chinensis S.K.Wu ex P.T.Li and P. verrucipes C.Y.Wu ex P.T.Li, remains unresolved and is shown here to be phylogenetically affiliated with Polyalthiopsis. Phylogenetic analyses of six chloroplast regions (matK, ndhF, psbA-trnH, rbcL, trnL-F and ycf1; ca.7.3 kb, 60 accessions) unambiguously placed Polyalthia chinensis and P. verrucipes in the same clade with Polyalthiopsis floribunda (PP = 1, MPBS = 97%); the entire clade is sister to Miliusa with weak to strong support (PP = 1, MPBS = 54%). Polyalthia chinensis and P. verrucipes share several diagnostic characters with Polyalthiopsis floribunda, including the raised midrib on the upper surface of the leaf in vivo, conspicuous foliar glands when dried, petiole with transverse striations when dried and axillary inflorescences. The two species differ from Polyalthiopsis floribunda in having fewer flowers per inflorescence, longer linear petals and two ovules per carpel. On the basis of the combined molecular phylogenetic and morphological data, we propose two new combinations, Polyalthiopsis chinensis (S.K.Wu ex P.T.Li) B.Xue & Y.H.Tan and Polyalthiopsis verrucipes (C.Y.Wu ex P.T.Li) B.Xue & Y.H.Tan. The protologue of Polyalthia verrucipes did not include a description of the flowers, which we provide here. An updated description for the genus Polyalthiopsis and a key to species in the genus Polyalthiopsis is also provided.

H.Tan. The protologue of Polyalthia verrucipes did not include a description of the flowers, which we provide here. An updated description for the genus Polyalthiopsis and a key to species in the genus Polyalthiopsis is also provided.
Amongst the newly segregated genera, Polyalthiopsis Chaowasku is a poorly known monotypic genus from Vietnam . The type species Polyalthiopsis floribunda is known from only two field collections (Poilane 10052, P, A, BO, CMUB, HN, K, L, P; and Chaowasku 128, CMUB). The species was first collected in 1924 and described under the name Polyalthia floribunda Jovet-Ast (Jovet-Ast 1940). I.M. Turner (2016) subsequently transferred the name to Huberantha. Ninety years after the first collection, Chaowasku collected this species again in 2014 and was able to sequence it for a phylogenetic study. It was shown not to be congeneric with Huberantha and was instead retrieved (without statistical support) as sister to Miliusa, leading Chaowasku et al. (2018) to erect a monotypic genus, Polyalthiopsis Chaowasku, to accommodate it. With only one Polyalthiopsis species and limited DNA regions used in the phylogenetic study, the sister relationship between Polyalthiopsis and Miliusa was not well established. It is also difficult to identify important diagnostic characters for Polyalthiopsis with only one flowering collection and a single monocarp available for taxonomic comparison.
Several species names remain unresolved in Polyalthia and await assignment to specific genera (Xue 2013;Xue et al. 2012), including the Chinese endemics Polyalthia chinensis S.K.Wu ex P.T.Li and P. verrucipes C.Y.Wu ex P.T.Li. As with P. floribunda, these two species are represented by very few collections and lack adequate flowering and fruiting descriptions.
The name Polyalthia chinensis was published in 1976, based on a flowering collection (Qinghai-Xizang Exped. 74-4451, KUN & PE) from Mêdog, Xizang, China, in 1974(Li 1976Li and Gilbert 2011). A second sterile specimen was subsequently collected in 1983 (B. S. Li & S. Z. Cheng 2668, PE). Until now, the species was only represented by these two collections.
The relationship between these two species has been controversial. Both species are represented by very few collections, with P. chinensis lacking fruiting specimens and P. verrucipes lacking flowering specimens, rendering morphological comparison problematic. Based on the foliar glands and leaf venation, Hou and Li (2007) regarded the name P. chinensis as synonymous with P. verrucipes, although this treatment was rejected by Li and Gilbert (2011) in the Flora of China without explanation. Li and Gilbert (2011) included identical floral descriptions in their treatment of P. verrucipes and P. chinensis, but with no clear indication of the source of this information, casting some doubt over the floral description of P. verrucipes.
With limited morphological characters, especially the lack of flowers in P. verrucipes and the limited material available, the relationship between P. chinensis and P. verrucipes and their taxonomic placement has never been resolved. We therefore, carried out several field explorations to search for these two species. This resulted in new collections of Polyalthia verrucipes, including flowers, enabling clarification of the relationship between P. chinensis and P. verrucipes, as well as their phylogenetic position. As a consequence, we were able to enlarge the poorly known genus Polyalthiopsis, supplementing available descriptions and providing better support for its sister relationship.
Six chloroplast DNA regions (matK, ndhF, rbcL, psbA-trnH and trnL-F and ycf1) were sequenced for the above-mentioned four collections of Polyalthia chinensis and P. verrucipes. The samples, localities and GenBank accession numbers are listed in Appendix I.

DNA extraction, amplification and sequencing
Genomic DNA was extracted from herbarium materials using a modified cetyl trimethyl ammonium bromide (CTAB) method (Doyle and Doyle 1987). A single amplification protocol was used for amplification of the chloroplast regions: template denaturation at 94 °C for 5 min, followed by 35 cycles of denaturation at 95 °C for 30 sec; primer annealing at 50 °C for 1 min; and primer extension at 72 °C for 1 min, followed by a final extension step at 72 °C for 10 min. The primers used to amplify the psbA-trnH intergenic spacer were psbAF (Sang et al. 1997) and trnH2 (Tate and Simpson 2003); other primers are the same as those used by Thomas et al. (2012). PCR products were visualised using agarose gel electrophoresis. Successful amplifications were purified and sequenced on an Applied Biosystems 3730xl DNA Analyzer at Sangon Biotech (Shanghai) Co. Ltd., Guangzhou, China.

Alignment and phylogenetic analyses
Sequences were assembled and edited using Geneious ver. 5.4.3 (Drummond et al. 2010) and pre-aligned with the MAFFT (Katoh et al. 2002) plugin in Geneious using the automatic algorithm selection and default settings and, subsequently, manually checked and optimised. An inversion of 15 positions in psbA-trnH was identified and reverse complemented in the alignment, following a strategy previously applied by Pirie et al. (2006), to retain substitution information in the fragments.
Maximum parsimony (MP) analyses of the seven combined regions were conducted using PAUP ver. 4.0b10 (Swofford 2003). All characters were weighted equally and gaps treated as missing data. The most parsimonious trees were obtained with heuristic searches of 1,000 replicates of random stepwise sequence addition, tree bisection-reconnection (TBR) branch swapping with no limit to the number of trees saved. Bootstrap support (BS) was calculated following Müller (2005), with 10,000 simple stepwise addition replicates with TBR branch swapping and no more than 10 trees saved per replicate.
Bayesian analysis was performed using NSF Extreme Science & Engineering Discovery Environment (XSEDE) application of MrBayes ver. 3.2.2 (Huelsenbeck and Ronquist 2001;Ronquist and Huelsenbeck 2003) provided by the CIPRES Science Gateway (Miller et al. 2010). PartitionFinder2 was used to test the dataset for partitions (model of evolution: mrbayes; model of selection: AICc; scheme search: greedy) (Guindon et al. 2010;Lanfear et al. 2012Lanfear et al. , 2016. The best partition scheme suggested six partitions, based on DNA region identity with GTR+G chosen for matK, psbA-trnH, trnL-F and ycf1 regions and GTR+I+G selected for the ndhF and rbcL regions. Two independent Metropolis-coupled Markov Chain Monte Carlo (MCMC) analyses were run. Each search used three incrementally heated and one cold Markov chain and was run for 10 million generations and sampled every 1,000 th generation. The temperature parameter was set to 0.08. The mean branch length prior was set from the default mean (0.1) to 0.01 (brlenspr = unconstrained: exponential (100.0)) to reduce the likelihood of stochastic entrapment in local tree length optima (Brown et al. 2010). Convergence was assessed using the standard deviation of split frequencies, with values < 0.01 interpreted as indicating good convergence. Tracer ver. 1.6 (Rambaut et al. 2014) was used to determine whether the parameter samples were drawn from a stationary, unimodal distribution and whether adequate effective sample sizes (ESS) for each parameter (ESS > 200) were reached. The first 25% of samples (2,500 trees) were discarded as burn-in and the post-burn-in samples summarised as a 50% majority-rule consensus tree.

Morphological studies
Comparative morphological data were obtained from specimens deposited in KUN, IBSC and PE herbaria and from published literature. Field surveys were carried out in Menghai County, Yunnan Province, with voucher specimens deposited in HITBC and IBSC.

results
The concatenated alignment of the 60-terminal dataset consisted of 7,334 characters. The MP heuristic search retrieved four equally most parsimonious trees of 3,519 steps (consistency index, CI = 0.664; retention index, RI = 0.709).
The MP and Bayesian analyses resulted in similar topologies. The 50% majorityrule consensus tree resulting from the Bayesian analyses under the six partitioned model is shown as Fig. 1. The results are consistent with previous phylogenetic analyses of the family, with the backbone of the tribe Miliuseae unresolved as in previous studies.

Discussion
Polyalthia chinensis was regarded as a synonym of P. verrucipes by Hou and Li (2007). The phylogenetic positions of these two species are quite distinct, with the following relationship: (Polyalthia verrucipes, (Polyalthia chinensis, Polyalthiopsis floribunda)) ( Fig. 1). Our field collection of the flowers of Polyalthia verrucipes provides further evidence for the distinction between these two species.
Although these two species resemble each other vegetatively ( Fig. 2A, C), they differ in the number of flowers per inflorescence, the length and thickness of the pedicel and the colour of the petals. The inflorescences of Polyalthia chinensis have 1-2 flowers ( Fig. 2A, C), whereas those of Polyalthia verrucipes comprise a solitary flower (Figs 2D, 3F, G). The pedicel of P. chinensis is slender and ca. 7 mm long, whereas that of P. verrucipes is stout and shorter than 2 mm. The petals of P. chinensis are green (Li 1976), whereas those of P. verrucipes are white (Fig. 3F-H). The leaf also differs slightly, with the leaf lamina of P. chinensis (2.5-3.8 cm) narrower than that of P. verrucipes (2.5-5 cm) and slightly thinner. The morphological data are therefore congruent with the phylogenetic topology and our phylogenetic and morphological analyses support the hypothesis that both species are not conspecific, as suggested by Li and Gilbert (2011).
The monotypic genus Polyalthiopsis Chaowasku was published in 2018, based on Polyalthia floribunda collected in Vietnam . It was reconstructed as the sister group of Miliusa, but without statistical support. Polyalthiopsis, Huberantha and Miliusa have previously been retrieved as an unsupported to weakly supported clade in Chaowasku et al. (2018). Although Chaowasku et al. (2018) mentioned that a more comprehensive phylogenetic study, using the whole plastome sequence data, demonstrates the same topology with strong support, the result has yet to be published. Polyalthiopsis is also retrieved as sister to Miliusa in this study, with weak support in the maximum parsimony analysis (MP BS = 54%), but strong support in the Bayesian analysis (PP = 1). This sister relationship was also well supported in Xue et al. (2020) (ML BS = 86%, suppl. material 1: fig. S1). The relationship between Huberantha and the Polyalthiopsis-Miliusa collective clade is, however, not retrieved in this study. The long-recognised sister relationship between Miliusa and Huberantha in previous studies (Mols et al. 2008;Saunders et al. 2011;Xue et al. 2011Xue et al. , 2012Chaowasku et al. 2012Chaowasku et al. , 2014Chatrou et al. 2012;Guo et al. 2017) can be redefined here following the inclusion of Polyalthiopsis.
Polyalthiopsis Chaowasku is easily distinguished from most of the other genera in the tribe Miliuseae by its raised midrib on the adaxial leaf surface. When dry, such an adaxial leaf midrib appears slightly sunken. The raised midrib on the adaxial leaf surface is rarely observed in the Annonaceae, but is known from Artabotrys (Sinclair 1955 (Chaowasku et al. 2013;van Heusden 1995). Another distinct feature of Polyalthiopsis floribunda is the dried petiole with multiple transverse striations ). Not many Annonaceae species have this pronounced drying artifact. One more distinct feature is the obvious foliar glands on the leaf surface when dried (obvious in fig. 2C in Chaowasku et al. 2018). Foliar glands are also observed in Wuodendron B.Xue, Y.H.Tan & Chaowasku in Miliuseae ).
Based on one species with only two collections, the genus is not well described and compared and, hence, it is difficult to identify important diagnostic characters. Polyalthia chinensis and P. verrucipes are retrieved in the same clade as Polyalthiopsis floribunda in the molecular phylogeny (Fig. 1). Sterile material of these three species is very similar. The leaves are elliptic with a cuneate base and acute to acuminate apex, with brochidodromous venation and reticulate tertiary veins. The leaf midrib in all three species is furthermore raised adaxially in vivo ( Fig. 3C; raised midrib still visible in the specimen of P. chinensis), with multiple transverse striations on the dried petiole (Figs 3E, 4C) and obvious foliar glands on dried leaf surface (Figs 2E-H, 3D, 4B). Although all three species have axillary inflorescences, the number of flowers per inflorescence differs: P. chinensis has one to two flower(s) per inflorescence ( Fig. 2A, B), Polyalthia verrucipes has only one flower per inflorescence (Figs 2D, 3F-H), while Polyalthiopsis floribunda has 1-5 flower(s) per inflorescence. The shape of the petal also differs: the petals of Polyalthia chinensis and P. verrucipes are linear (Figs 2B, D, 3F-H, 4D, E), while those of Polyalthiopsis floribunda are elliptic-ovate . The carpel characters of Polyalthia chinensis and P. verrucipes also differ greatly from those of Polyalthiopsis floribunda: the former two species have two ovules per ovary and hence two seeds in each monocarp (Fig. 3K, M), whereas Polyalthiopsis floribunda has only one ovule per ovary (Jovet-Ast 1940;Chaowasku et al. 2018).
In conclusion, Polyalthia chinensis, P. verrucipes and Polyalthiopsis floribunda share axillary inflorescences, a raised midrib on the adaxial leaf surface (Fig. 3C), petioles with transverse striations when dry (Fig. 3E) and foliar glands on dried leaf surface (Figs 2E-H, 3D, 4B). These characters render the three species distinctive from other species in the tribe and are thus diagnostic for the enlarged genus Polyalthiopsis.
The present phylogenetic study shows that Polyalthia chinensis is strongly supported as sister to Polyalthiopsis floribunda (PP = 1, MP BS = 86%). The collective clade is strongly supported as sister to Polyalthia verrucipes (PP = 1, MP BS = 97%). The whole clade (comprising the three species) is weakly to strongly supported (PP = 1, MP BS = 54%) as sister to Miliusa. The morphological and molecular phylogenetic data therefore support the transfer of Polyalthia verrucipes and P. chinensis to Polyalthiopsis and the new nomenclatural combinations are proposed here.
As Polyalthia verrucipes was published based on fruiting material only (Li 1976), with the newly collected flowers, an updated description is presented. It is noteworthy that the floral description of P. verrucipes, published by Li and Gilbert (2011), does not correspond with the material we collected in the field, but is instead similar to that of P. chinensis.
As more species were included in the genus Polyalthiopsis, an updated description and a key to the three species is also provided. Description. Medium-sized to large trees. Young twigs glabrous. Leaves petiolate, blade elliptic, with glandular dots observable when dry, base cuneate, apex acute to bluntly (caudate-)acuminate; petiole with transverse striations when dry; upper surface of midrib raised in living plants, becoming slightly sunken when dry, lower surface of midrib raised; secondary veins rather faint in living plants, becoming slightly raised on both sides when dry, leaf venation brochidodromous; tertiary veins reticulate. Flower(s) in 1-to 5-flowered inflorescences, bisexual, pedicellate; inflorescences axillary, peduncle inconspicuous, bracts present. Sepals broadly ovate-triangular. Petals membranous-papyraceous to leathery. Outer petals elliptic-ovate or linear-lanceolate. Inner petals (narrowly) elliptic-ovate or linear-lanceolate. Stamens numerous per flow- er, connective truncate, covering thecae. Carpels numerous per flower; ovaries with 1 or more line(s) of hairs; stigma terete; ovule(s) 1 or 2 per ovary, sub-basal or lateral. Monocarps oblong to rhomboidal or cylindrical, stipitate, glabrous. Seed(s) 1or 2 per monocarp, cylindrical, surface smooth, raphe broadly sunken and partially slightly raised in middle, endosperm ruminations lamelliform.
Distribution. Three species, known from Xizang, Yunnan Provinces of China and Thừa Thiên-Hu, Ninh Thuận Provinces of Vietnam (Fig. 5  Description. Trees to 15 m tall (Fig. 3A). Branches greyish-black, glabrous. Petiole 3-7 mm long, 1-2 mm in diameter, glabrous, with transverse striations when dry (Figs 3E, 4C); leaf laminas oblong to oblong-lanceolate, 10-17 × 2.5-5 cm, base broadly cuneate or obtuse, apex acuminate (Figs 3B, 4A), both surfaces glabrous, thinly leathery, densely verrucate with foliar glands when dry (Figs 2G, H, 3D, 4B); upper surface of midrib raised when fresh (Fig. 3C), becoming flat or slightly sunken when dry (Fig. 2G), lower surface of midrib raised; secondary veins 15-18 on each side of midrib, delicate and prominent on both surfaces; tertiary veins reticulate. Inflorescences axillary, with solitary flower (Figs 2D, 3F, H, 4A, D, E). Pedicel 1-2 mm long, hispid, with one ovate bracteole at top, 2-3 ovate bracteoles at base (Fig. 3F, L). Sepals ovate, 2 × 2 mm, slightly reflexed, ciliate (Fig. 4F) (2004). Prior to this study, P. verrucipes was only represented in herbaria by two collections from Yunnan, China (two localities, both of which have subsequently been severely deforested). Our field survey in 2016 identified one population with dozens of individuals of dbh ca. 10 cm and dozens of young treelets in Manxi village, Menghai County. We made a second visit to the location in 2019 and found only a few individuals with dbh larger than 10 cm and few treelets. Herbicide had been used in that location. The bark at the bottoms of the tree trunks was damaged. The local farmers appear to clear the forests in this way for tea plantation and it is anticipated that the trees with damaged bark could not survive. We hope additional undocumented subpopulations will be found and protected, although further field investigation is needed to better understand the current status of populations. At present, we recommend that this species be regarded as critically endangered (CR) based on current IUCN Red List Categories and Criteria (IUCN 2012).