Weda, a new genus with two new species of Euphorbiaceae‐Crotonoideae from Halmahera (North Maluku, Indonesia) and phylogenetic relationships of the Australasian tribe Ricinocarpeae

During the environmental impact study for a proposed nickel mine near Weda Bay on Halmahera in North Moluccas (Maluku Utara Province), Indonesia, two unknown Euphorbiaceae were discovered. Morphological comparisons and molecular phylogenetic analyses using four markers (plastid trnL‐F and rbcL, and nuclear ribosomal internal transcribed spacer and external transcribed spacer) indicated that they should be recognized as constituting a new, distinct genus of two species, which are described and illustrated here as Weda fragarioides and Weda lutea. The new taxa are members of the Australasian tribe Ricinocarpeae in subfamily Crotonoideae, and they are most closely related to Alphandia. In contrast with the otherwise mostly sclerophyllous Ricinocarpeae, Weda possesses stellate to dendritic hairs, large, long‐petiolate, glandular leaves, and inflorescences with a pair of large, leafy, subopposite bracts. The two narrowly distributed species are distinguished from each other by vegetative and floral features, molecular data, and elevational preferences. Leaf elemental analysis of Weda indicated manganese, but not nickel, accumulation. Newly resolved generic relationships and potential morphological synapomorphies within Crotonoideae are discussed, and the circumscription of Ricinocarpeae is expanded from 7 to 11 genera.


Introduction
The exploration of remote and ecologically or edaphically unusual environments often yields significant taxonomic novelties, and this pattern is evident in Euphorbiaceae with several genera recently described (i.e., Gradyana, Athiê-Souza et al., 2015;Karima, Cheek et al., 2016;Tsaiodendron, Zhou et al., 2017;Incadendron, Wurdack & Farfan-Rios, 2017). The Malay Archipelago (Malesia) with its many islands and intense tectonic activity (including earthquakes and volcanism) is ecologically diverse and contains a high number of endemic species (van Welzen et al., 2005). Many of these islands are still undercollected and their floras poorly known, including the Moluccas in Indonesia (Fig. 1). Of the 804 gymnosperm and angiosperm species recorded for the Moluccas (from a total of 6616 indigenous Malesian taxa published in Flora Malesiana series 1; van Steenis and others, 1950 and later; see van Welzen et al., 2005), 71 are This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. endemic. A review of the checklists of the Malesian Euphorbiaceae (Airy Shaw, 1982;van Welzen, 2017) shows that 8 of the 81 recorded Moluccan species (in 26 genera) are endemic. The Moluccas are often treated as a monolithic biogeographic unit (i.e., same climate), but recently it was shown that the very different tectonic origins of North and South Moluccas warrant the recognition of two biogeographic provinces (Rutgrink et al., 2018).
Halmahera is the largest island in North Moluccas (Maluku Utara; Fig. 1), and it harbors economically important ores rich in nickel and cobalt. The mining company PT Weda Bay Nickel (PT = Perseroan Terbatas, and it refers in Indonesian to a limited liability company) plans to mine in Halmahera, one of the largest undeveloped nickel deposits in the world, with potential for extraction of 9 million tons nickel. Ultramafic soils, associated with such deposits, are known to harbor higher levels of plant endemism and diverse metal hyperaccumulators (e.g., Cuba, New Caledonia; Reeves et al., 1996;Jaffré et al., 2013). As a part of the procedure to procure mining permissions, an environmental impact study was needed, of which one aspect was a botanical inventory that was conducted under the guidance of the Missouri Botanical Garden. The mine site and neighboring areas in Halmahera were inventoried and the collected specimens were distributed to the herbaria of BO (Herbarium Bogoriense), L (Naturalis Biodiversity Center), and MO (Missouri Botanical Garden), whereas staff of BO and L provided most of the identifications, with further expert identifications obtained as needed. During identification at L, from the set of three separate collections from the vicinity of Weda Bay in the southeastern coast of Halmahera ( Fig. 1), unknown Euphorbiaceae appeared with the following characteristics: large, adaxial, basal glands on the leaf blades and fruits with a columella, which had apical remnants of a single ovule/seed per locule. The "Macaranga" specimen, referred to by Lopez et al. (2019bLopez et al. ( , 2019c, and other three specimens originally identified as "Pangium edule" (Achariaceae) proved to be a second new species, overlapping in several distinctive characters. Morphological and molecular analyses have shown that this material from Weda Bay constitutes a new genus comprising two new species within Euphorbiaceae, subfamily Crotonoideae Pax, and they are described below. We provide a generic-level molecular phylogenetic perspective of the core Crotonoideae, which is the most comprehensive perspective to date, illuminating the interesting biogeography, as well as certain problems in existing tribal classification, especially related to tribe Ricinocarpeae Müll.Arg. We also investigated whether the new genus might be a nickel hyperaccumulator, given its presence in other Euphorbiaceae (e.g., Leucocroton Griseb.; Reeves et al., 1996) and in other Malpighiales worldwide, including species of Phyllanthaceae (Reeves et al., 1996;van der Ent et al., 2016van der Ent et al., , 2018Bouman et al., 2018), Dichapetalaceae, Salicaceae, and Violaceae (van der Ent et al., 2015). For Halmahera, it was recently shown that one of these hyperaccumulators, a Rinorea sp. (Violaceae), was even capable of influencing the bacterial community around its roots (Lopez et al., 2019b).

Micromorphology and elemental analysis for heavy metals
Scanning electron microscopy (SEM) of untreated herbarium fragments and unacetolyzed pollen was conducted with a Zeiss EVO MA15 (Carl Zeiss SMT, Inc., Peabody, Massachusetts, USA) SEM at 12 kV after sputter coating with 3 nm of C and 8 nm of Au/Pd using a Leica EM ACE600 (Leica Microsystems GmbH, Wetzlar, Germany). An Orbis MC Micro-XRF Analyzer (EDAX Inc., Mahwah, New Jersey, USA), 20 kV, 600 μAmp, was used to scan a dried leaf fragment of 1.28 by 1 cm (Gushilman et al. 777, L) for a non-destructive elemental analysis.

Molecular phylogenetic sampling
To infer broad relationships of the Weda Bay material, we generated orthologous sequences of plastid rbcL and trnL-F (trnL intron and the 3′ intergenic spacer), and performed a preliminary analysis in the context of the full 179-tip Euphorbiaceae backbone phylogeny of Wurdack et al. (2005), which indicated nested placement in Crotonoideae. Our final taxon sampling of 102 tips (97 taxa) focused on Crotonoideae and included 47 of ca. 52 genera potentially belonging to clade C2, using sequence data from Wurdack et al. (2005), GenBank, and 149 newly generated sequences. A small outgroup sampling (seven taxa) represented the three other subfamily lineages without greatly increasing trnL-F alignment complexity. To improve resolution, especially for the Ricinocarpeae s.l. subclade with which the Weda Bay material grouped, we added nuclear ribosomal internal transcribed spacer (ITS) region and external transcribed spacer (ETS) region data. External transcribed spacer aims to resolve species-level relationships using recently designed primers that amplify multiple groups of Euphorbiaceae (Cardinal-McTeague et al., 2019;Wurdack, unpublished data). To reduce alignment problems with increasing sequence divergence, the ITS taxon sampling was limited to clades C1 and C2 (77 tips, 65 with data) and the ETS to Ricinocarpeae s.l. (28 tips, all with data plus two other clade C2 taxa). In addition, a fragment of plastid matK was sequenced for the Weda Bay material to produce DNA barcode reference data. Voucher information and GenBank numbers are provided in the Appendix I.

Phylogenetic analyses
For initial analysis, trnL-F sequences of the Weda Bay material were inserted into the 179-tip multiple sequence alignment (MSA) of Wurdack et al. (2005) using Mesquite ver.3.4 (Maddison & Maddison, 2018), and global Bayesian inference (BI) was conducted with MrBayes ver.3.2.6 (Ronquist et al., 2012) on the CIPRES gateway (https://www.phylo.org/) using standard options (most complicated model, nst = 6, rates = invgamma) and sampling every 1000 generations over 20 000 000 generations. Convergence was assessed with Tracer ver.1.6.0 (Rambaut et al., 2013) for effective sample sizes (ESS) >200, a 20% burn-in implemented, and the maximum clade credibility (MCC) tree (LogCombiner and TreeAnnotator in Beast package; Drummond et al., 2012) viewed with FigTree ver. 1.4.2 (Rambaut, 2014). For the final 102-tip matrix, the sampling of Wurdack et al. (2005) was reduced to 66 tips, the trnL-F alignment then collapsed empty columns, and the additional data were manually inserted under similarity criteria using Se-Al ver.2.0a11 (Rambaut, 1996(Rambaut, -2002. As noted in the study of Wurdack et al. (2005), the trnL-F alignment is complex with a mix of discrete indels and hypervariable regions of ambiguous alignment. For ITS and ETS, the MSAs were generated under the Q-INS-i refinement method of MAFFT ver.7.452 (Katoh & Standley, 2013) and then adjusted by eye using Se-Al, similarity criteria, and limiting matrix fragmentation. The ITS and ETS MSAs are relatively compact with mostly 1-3 bp indels. For ITS, the MSA is relatively homogeneous within clades C1 and C2, but it was challenging to align clades with each other; for ETS, the MSA was relatively homogeneous as the taxon sampling included only two tips (Aleurites, Tapoides) outside of Ricinocarpeae s.l. To examine the impact of ambiguously aligned regions, sensitivity analyses included three masking sets: (i) all data (aligned length 4354 bp and 38.3% total missing data), (ii) a strict set excluding 814 bp, which removed positions with >50% missing data (aligned length 3540 bp and 26.0% total missing data), and (iii) relaxed set excluding 567 bp (aligned length 3787 bp and total 30.6% missing data), which removed trnL-F hotspots and overlapping indels (a subset of the 814 bp strict set). All three masking set analyses yielded similar topologies and support values (differing <5%), and the relaxed set (567 bp excluded) was selected for final analyses. To examine the impact of model selection and ML program on results, the 567-matrix was run on PhyML ver.3.0 with Smart Model Selection ver.1.8.1 (GTR + I + G selected; Guindon et al., 2010;Lefort et al., 2017) and IQ-TREE (567-matrix partitioned by marker; Trifinopoulos et al., 2016). For final trees, the Bayesian MCMC analyses were implemented in MrBayes with two concurrent runs, each with four chains and sampling every 1000 generations over 50 000 000 generations, a 0.2 temperature coefficient, and a conservative 20% burn-in implemented. The final maximum likelihood (ML) analysis was performed with RAxML ver.8.2.12 (Stamatakis, 2014), as implemented on CIPRES XSEDE under GTR + I + G and clade support estimated by 1000 rapid bootstrap (ML-BS) replicates. Each marker was also analyzed separately with ML to determine any strongly supported topological differences that could indicate incongruence. A reduced 30-tip data set (all 28 Ricinocarpeae and rooted with Aleurites +Tapoides) with no exclusion sets (aligned length 3817 bp and 22.0% total missing data) was analyzed with RAxML as before to test the impact of reduced missing data (especially for ETS), fewer ambiguous regions, and greater computational thoroughness on clade resolution. The final 102-tip MSA with masking details is available from the Dryad Digital Repository (https://doi.org/10. 5061/dryad.nvx0k6dnw or https://datadryad.org).

Results
The elemental analysis to assess whether the first Weda Bay species might be a nickel hyperaccumulator showed no increased values (Fig. 2). The nickel present at low concentrations appeared to be evenly distributed, though less so above the major veins, which is consistent with the spread of this metal, as part of water-soluble salts, through the leaf due to evaporation (Antony van der Ent, pers. comm.). However, the leaf had manganese accumulation, especially along the midrib and major veins (Fig. 2). These findings are corroborated by Lopez et al. (2019bLopez et al. ( , 2019c for the second species (referred to as "Macaranga sp." in their papers), which accumulates some nickel but mainly manganese.
Search results of the matK and trnL-F sequences of the Weda Bay material against GenBank and/or BOLD using BLAST showed best matches with members of tribe Ricinocarpeae. In the 179-tip Euphorbiaceae-wide, 2-marker MCC tree (not shown, but compatible with fig. 4 in Wurdack et al., 2005), the Weda Bay material was recovered as a strongly supported (posterior probability, PP 1.0) sister group to Ricinocarpos. In our 102-tip, 4-marker analyses, the major lineages of Euphorbiaceae that represent potential subfamily clades were poorly resolved along the spine of the tree and Crotonoideae was not supported as monophyletic (Fig. 3A). The core Crotonoideae (excluding the articulated crotonoids, Gelonieae, and Adenoclineae s.l.) are strongly supported (PP 1.0, bootstrap percentage, BP 97) and comprise two strongly supported (PP 1.0, BP 100) sister clades C1 and C2 (sensu Wurdack et al., 2005); however, C2 has especially poor backbone resolution with none of its four constituent tribes resolved as monophyletic. The Weda Bay material was strongly supported (PP 1.0, BP 100) as a member of Ricinocarpeae s.l. and as a weakly supported sister group to Alphandia Baill. (PP 0.75, BP 60). Although resolution among the genera within Ricinocarpeae s.l. was poor, only Ricinocarpos and Baloghia were not recovered as monophyletic. The Bayesian and ML analyses were largely congruent with a handful of unsupported (BS <50) topological differences. The Ricinocarpeae s.l. resolution was similar between the 102-and 30-taxon RAxML analyses, with slightly better support for the Weda Bay material + Alphandia (BP 75) and a shift in the poorly supported node relating to the two Baloghia subclades (see Fig. 3B). There were no topological differences, with BP >70, among analyses of each individual marker or in partial combination as plastid versus ribosomal (see Figs. S1, S2). However, the ribosomal markers did have weakly supported differences in the deepest nodes of Ricinocarpeae s.l., and in particular Alphandia and the Weda Bay material had a nested rather than sister relationship (Fig.  S2). Our alternative analyses (i.e., differing exclusion sets, IQ-TREE, PhyML, partitioning schemes) showed a little impact of analysis details on topology or support values. The multiple accessions of each Weda Bay taxon have identical sequences for each marker within a species. Variation between species includes 3 differences (1 poly-A indel and 2 substitutions, all in the trnL intron) for trnL-F, no differences for rbcL, 13 differences (11 substitutions and 2 polymorphic sites) for ITS, and 9 differences (8 substitutions and 1 tandem duplication) for ETS.
Ricinocarpeae s.l., the focus of this study due to its relevance for placing the Weda Bay material, is the largest supported C2 subclade in our analyses and for which we propose a new tribal circumscription (Fig. 3A) that is expanded from the seven genera of Webster (2014). In our sensu lato circumscription, the tribe contains a distinct Australasian group of nearly 100 species in 11 genera (all 11 sampled), which include the Weda Bay genus (described below) and members of Ricinocarpeae s.s. (Bertya Planch., Beyeria Miq., Borneodendron Airy Shaw, Cocconerion Baill., Myricanthe Airy Shaw, Ricinocarpos Desf., Shonia R.J.F.Hend. & Halford), and Codiaeae subtribe Baloghinae G.L.Webster (Alphandia, Baloghia, Fontainea, but not Hylandia). Dimorphocalyx Thwaites, which was closely associated with Fontainea by Radcliffe-Smith (2001), is excluded from Ricinocarpeae s.l. We have refrained from revising the subtribal taxonomy due to low support at the deepest nodes. However, the Weda Bay taxa can be accommodated in subtribe Bertyinae Müll.Arg. Members of the tribe, and they present the following characteristics: even though morphologically diverse, frequently (but see exceptions below) share monoecy (rarely dioecious), reddish latex, stipules absent (rarely present), indumentum stellate (rarely simple or dendritic), inflorescences terminal (sometimes axillary or pseudoterminal), pistillate flowers with well-developed petals (rarely absent), stamens numerous (mostly 20-100), filaments often partly connate, anthers extrorse, and seeds carunculate (or fruit drupaceous). These features in combination distinguish Ricinocarpeae s.l. from other clade C2 taxa, and stipules appear especially diagnostic. Genera of clade C2 possess stipules except for Codiaeum Rumph. ex A.Juss., Garcia Vahl. and nearly all Ricinocarpeae s.l. (except Borneodendron), which are exstipulate. Oligoceras, known only from the type collection of flowering branches, is described as exstipulate (Radcliffe-Smith, 2001;Webster, 2014) but needs further study to rule out small, caducous stipules. The apical bud in Borneodendron is covered by a caducous, circum-axillary sheath, associated with the youngest verticillate whorl of three leaves, enclosing suppressed axillary buds (each apparently covered in tiny bud scales) and the younger leaf primordia (SAN 109845, L, US). This sheath has been interpreted as united interpetiolar stipules with similarities to Baloghia inophylla (G.Forst.) P.S.Green (Airy Shaw, 1963;van Welzen, 2012). Baloghia is usually described as exstipulate (Radcliffe-Smith, 2001;Webster, 2014), but some species bear structures that have been described as stipules (Baillon, 1858;Pax & Hoffmann, 1911) or caducous bud scales ("perulae pseudostipuliform, soon deciduous," fide Radcliffe-Smith, 2001). Baloghia spp. have a range of shoot apex morphologies, from clearly naked with no appendages in the New Caledonian taxa (e.g., B. alternifolia Baill., Baumann 15305, US; B. drimiflora (Baill.) Schltr., Bernardi 12534, US; B. pulchella Schltr ex Pax, Franc 2486, US) to variously sheathed. In B. inophylla (Johnson & Constable 52338, US), there are two decussate pairs of valvate axillary appendages enclosing a single pair of opposite leaves and leaving annular scars. The appendage pairs are dimorphic and perhaps not homologous, with the proximal pair (bud scales) scale-like and containing axillary buds, and the much larger distal pair (stipules), with hirsute margins and lacking axillary buds. In Australian Baloghia marmorata C.T.White, a cluster of 6-8+ imbricate scales with axillary buds, encloses a seasonal flush of otherwise naked leafy nodes (4-6 alternate to subopposite leaves), leaving annular scars (White 3588, US). In Australian Baloghia parviflora C.T.White, which otherwise vegetatively resembles B. marmorata, there are no imbricate scales (McPherson 6670, MO). The nature of stipules has been debated by morphologists; stipule axils are not expected to contain axillary buds (see Rutishauser & Sattler, 1986). However, other apical interpetiolar appendages containing axillary buds, similar to Borneodendron, are described as stipules (e.g., Rhizophoraceae; Gill & Tomlinson, 1969). Cocconerion spp., the strongly supported sister group to Borneodendron, possess verticillate whorls, densely packed with 6-10 leaves per node, which before leaf expansion form a tight stockade of abaxial midribs (vernation is involute, with blade rolls inside the bud), protecting the shoot apex without additional appendages (i.e., exstipulate).
Details of the intergeneric relationships recovered within Ricinocarpeae s.l. differ from the more sparsely sampled study of Tokuoka (2007). Webster (2014) excluded Alphandia from his Ricinocarpeae and classified it as an aberrant member of tribe Codiaeae. The pollen evidence used to justify this exclusion is less compelling than morphology, which unites them, and our results also support the inclusion of Alphandia within Ricinocarpeae, as circumscribed by Radcliffe-Smith (2001). Baloghia (3 of 15 species sampled) is clearly not monophyletic as Australian endemic B. marmorata groups with Fontainea (PP 1.0, BP 100). Baloghia has typical explosively dehiscent capsular fruit versus Fontainea with drupes. The fruiting voucher for B. marmorata, from which the leaf sample was directly obtained, appears to be correctly identified and has the characteristic dehiscent fruits of that species. The sclerophyllous Australian genera (i.e., Bertya, Beyeria, Ricinocarpos, Shonia) of Ricinocarpeae s.l. present many overlapping morphological characters (see table 1 in Halford & Henderson, 2005). Our results, though limited in power due to sparse species sampling and relatively poor resolution, suggest that the generic circumscriptions need further study. Shonia (three of four species sampled here) was described to accommodate taxa formerly in Beyeria that were considered morphological intermediates between Beyeria and Ricinocarpos (Halford & Henderson, 2005). They form a strongly supported clade (PP 1.0, BP 100), but the continued recognition of Shonia as a distinct genus will depend on relationships among the related sclerophyllous genera. Borneodendron, Cocconerion, and Myricanthe clearly fall within Ricinocarpeae s.l. and have no relationships with Picrodendraceae (formerly Euphorbiaceae, subfamily Oldfieldioideae), as had been suggested by Airy Shaw (1971Shaw ( , 1980 and Radcliffe-Smith (2001), based on some morphological similarities and biogeography. This hypothesis had discounted the major differences in ovule number and palynology between Picrodendraceae (two ovules per locule; pollen echinate) and Euphorbiaceae (one ovule per locule; pollen not echinate except in Cheilosoideae). The close relationship between Borneodendron and Cocconerion is well supported by morphology including the unusual shared features of whorled leaves and hairy anthers (Airy Shaw, 1971).
The Ricinocarpeae s.l. genera are discussed below, highlighting major morphological differences from and similarities to the new genus established for the Weda Bay material.
Most Ricinocarpeae s.l. and the Weda Bay material agree in the absence of stipules (see above for comments on Borneodendron) and the presence of stellate hairs (but also dendritic in the new genus). Unlike the Weda Bay material, the Australian and New Caledonian taxa are mostly reduced variously and are sclerophyllous, with leaf blades lacking the basal glands, being sessile to shortly petiolate, and being very narrow and small. The other taxa sampled across clade C2 all differ from the Weda Bay material in their very different inflorescences, and most of them possess stipules and/or a floral disc. Neotropical Garcia has no stipules and an illdefined disc, but it only has simple hairs and completely different inflorescences with terminal clusters of large flowers.
Within the mainly Australian/New Caledonian Ricinocarpeae s.l., Borneodendron (N Borneo) and the Weda Bay genus are the biogeographic outliers. The presence of the latter in North Moluccas is likely the result of plate tectonic movements. Halmahera consisted of an eastern and a western half for a long time, and it was a part of the Outer Melanesian Arc, which was close to northeastern New Guinea around 24 Ma (Coleman, 1997;Hall, 2002). Dispersal from Australia likely involved New Guinea. During the last 24 million years, parts of Halmahera moved along the north coast of New Guinea to their present position, west of New Guinea, where they united, and similar to the rest of North Moluccas, they differed in floristic composition from South Moluccas, which moved to the north from Australia (Rutgrink et al., 2018). The Weda Bay plants, endemic to Halmahera, likely originated prior to the union with South Moluccas, but a divergence dating analysis is needed to test this hypothesis. Due to the distinctiveness of the Weda Bay material with morphological and molecular divergence from potential close relatives (Fig. 3), we describe the specimens from Halmahera as a new genus with two new species and classify the genus in the subfamily Crotonoideae, tribe Ricinocarpeae, as circumscribed herein.
Some morphological features of the new genus, especially in the flowers and fruits, are poorly understood due to the few, sparsely reproductive specimens available (one flower per inflorescence branch). We therefore refrained as much as possible from dissecting flowers. Uncertainty remains in the interpretation of the receptacle of the staminate flowers. In the first species, it appears as a highly domed receptacle, and in the second species, it seems more as a union of the basal part of the filaments into an androphore. In the sexes of both species, no typical floral disc or disc glands were discovered. However, a narrow tissue band with a hirsute margin is present in the staminate flowers of the second species and may be a reduced disc (Fig. 4A). Finally, the very young seeds, as seen in the material of the second species, had an apical appendage with caruncle-like tissue, which was not seen in the far more mature seeds of the first species. This difference could be another factor to separate the two species or the appendage may disappear during seed development. Caruncles are usually present in Ricinocarpeae, but rare elsewhere in clade C2.
The trichome diversity in the new genus spans three basic types (simple, stellate, and dendritic), with clear differences between the species, notably in the presence/absence of long, simple trichomes (Figs. 4G-4J). The dendritic trichomes, each containing a variously elongate central axis with numerous arms (radii), are especially well developed in the second species. The stellate trichomes can have long, thin or shorter, fatter arms, or occasionally have a long central arm ( Fig. 4H; porrect sensu Webster et al., 1996). The stellate and dendritic trichomes have a large multicellular attachment to the leaf surface, but they are easily detached and they mostly weather off from older leaves (Fig. 4F). The pollen of both species is virtually identical and spheroidal, 40-48 μm dia. (polar:equatorial ratio 1.03; n = 10, via SEM), inaperturate, and with subunits at the surface densely spaced, free, and rounded at their tips (Figs. 4D, 4E). The pollen has a typical Croton structure, closely resembling that of Alphandia (see Nowicke, 1994; fig. 49). Based on our present floristic inventories, the two new species have a very limited distribution, and they only occur in ultramafic areas, which have a patchy distribution (Cock & Lynch, 1999;Lopez et al., 2019a). This means that their continued existence is extremely vulnerable to habitat destruction, especially when nickel mining commences. Protection of the Weda Bay area is of eminent importance, given the growing list of endemic taxa (e.g., Nepenthes, Cheek, 2015;Pandanus, Callmander et al., 2015).
Leaves alternate with long petioles, base of blades with two large, adaxial, basal glands or these at end of narrow lobes, venation raised, very distinct. Inflorescences axillary, cymose, functionally unisexual, with long peduncle and apically two unequal, subopposite, leaf-like, (sub)sessile, late-caducous bracts. Flowers with calyx and corolla, apparently lacking a disc. Staminate flowers with highly domed, hirsute receptacle with many short stamens or an androphore with a short free-filament part per stamen. Pistillate flowers with three-locular ovary, stigmas split, smooth adaxially. Fruits capsular, smooth, dehiscing loculicidally and septicidally. Seeds naked, marbled.
Trees, monoecious; latex unknown. Indumentum of pale brown, short, stellate to somewhat lepidote to dendritic hairs, parts also with long, whitish, simple hairs (especially in W. lutea). Stipules absent. Leaves alternate to sub-opposite, simple; petioles long, cylindrical, basally slightly pulvinate, apically hardly thickened; blade at insertion (not peltate) or at base of peltation (blade peltate) with two adaxial large elongate glands, margin entire with an occasional extending gland, surfaces smooth, penninerved or basally palmately nerved, venation raised on both sides, especially beneath, secondary veins looped and closed near the margin, higher order veins laxly reticulate. Inflorescences axillary, erect, cymose, bisexual, but usually unisexial functionally, basically dichotomous and first pistillate with a single, central pistillate flower, after fruit dehiscence, when the more horizontal, scorpioid branches develop, with per branch only staminate flowers with one or two flowering at a time; peduncle long, somewhat flattened, with two leaf-like, slightly subopposite bracts of unequal size at the distal end, lower one smaller, both late caducous during staminate phase; floral bracts either vestigial, hairy enations or acicular and apically glandular. Flowers five-merous, actinomorphic, pedicellate; sepals four or five, basally united, lobes imbricate, outer three smaller than inner one or two, apices rounded; petals ovate, five or six, contorted, fleshy, glabrous, apex rounded. Staminate flowers, highly dome-shaped receptacle or filaments partly united in an androphore, densely hairy with simple hairs; disc absent or not obvious to a vague ring around dome/androphore in W. lutea; stamens more than 30, filaments free, thread-like, with few hairs, anthers twothecate, dorsi-basifixed, connective very short and indistinct, thecae almost completely separate, opening extrorsely with longitudinal slits, glabrous; pistillode absent; pollen spheroidal, inaperturate, crotonoid. Pistillate flowers, partly seen in bud and in young fruit; sepals five; petals five; disc not seen; ovary three-locular, densely hairy, with a single ovule per locule, style short, hairy, stigmas three, broad, mostly bifid, glabrous, smooth, not papillate above. Fruits capsular, ellipsoid, slightly lobed, lobes higher than wide, smooth, dehiscing completely loculicidally and septicidally into six mericarps; wall thin, with a thin exocarp and a woody mesoand endocarp when dry; columella three-quetrous, persistent. Seeds three per fruit, smooth, marbled, naked, with an apical appendage with carunculoid tissue in the young material of W. lutea but without appendage (disappeared) in the mature seeds of W. fragarioides; hilum small; embryo unknown.
Distribution: The two species are each discovered from three collections made in central Halmahera (Indonesia, N Moluccas; Fig. 1).  Etymology: The genus name refers to Weda Bay where the specimens were collected.
Note: The leaf blade glands are typical, with the long peduncled inflorescences having two leaf-like large bracts, the highly domed receptacle or androphore of the staminate flowers, and the stellate to dendritic indumentum.  (Fig. 1).
Habitat and Ecology: Common on Blikep in primary to secondary forest along the road at 566-777 m altitude. Flowering in September-October and fruiting in June.
Etymology: Due to the arched receptacle in the staminate flowers, the epithet refers to the strawberry (Fragaria L.), where the receptacle becomes a highly domed fruit.
Conservation: The known distribution of the new taxon is limited, as exploration of Halmahera and nearby islands is highly incomplete (see Callmander et al., 2015).
It appears to be a local endemic and might at least be vulnerable, especially when the planned nickel mine begins operation.
Note: The specimen from Blikep Nu is slightly different, as the leaf blades are more elliptic, whereas the specimens of Bukit Limber are more ovate. In addition, the peduncle of the pistillate inflorescence of Blikep Nu has (next to stellate hairs) long simple hairs, not seen in the specimens of Bukit Limber, whereas many stellate hairs on the fruits in Bukit Limber have a central, much longer and stiffer arm, not seen in the fruits of Blikep Nu. Habit: small trees (reported by Bangun et al. 118 as a 3 m long liana), 10 m high, dbh to at least 10 cm; flowering branches 3-5 mm diameter, branchlets rather densely covered with stellate to dendritic hairs, rather persistent, and long (to 2.2 mm) simple hairs, latter caducous. Outer bark brown to brownish gray, rugose; inner bark pale green to red; sap clear or red and oxidizing black. Leaves dark brown from dense indument when young; petioles 8-12.8 cm long, 1-1.8 mm thick in middle, with brown stellate to dendritic hairs and yellow simple hairs; blade ovate-oblong, 8-25.5 by 4.8-16 cm, 1.2-1.7 times longer than wide (mean = 1.58, SD = 0.145, n = 17), base peltate for less than 1 cm or cordate, emarginate, with one or usually two elongated slender glands, often to one side, ca. 1.3 by 0.5 mm, often with a flat gland next to the base; lamina margin somewhat recurved, often with several, somewhat shorter glands (Fig. 4J), apex acuminate, acumen up to 2 cm long and minutely gland tipped, upper surface with stellate and simple hairs when young, stellate hairs persistent on midrib and basal part of major veins, lower surface slightly hairy with stellate to dendritic hairs and simple patent hairs, venation basally palmate, secondary veins 8-10 pairs to apex. Inflorescences, peduncle 8.4-14.4 cm long, with stellate hairs and often long simple hairs and few dendritic hairs; bracts ovate, 4.2-8 by 3.2-5.8 cm, subsessile, petiole up to 5 mm long, basally emarginate, margin entire, apex acuminate, venation well developed, like leaf blades, but fewer secondary veins, indumentum like leaves; branches at least 18 cm long (broken), along with branches acicular bracts, one or two per node, up to 2.5 mm long by 0.2-0.4 mm diam., ending in a globose gland, apically bent downward. Flowers, sepals basally more united than in other species, petals yellow. Staminate flowers ca. 12 mm diameter, receptacle highly domed or androphore; bud green; pedicel up to 11 mm long, with stellate hairs and some long simple hairs; calyx pale green, ca. 3.8 mm long, lobes ovate to oblong, ca. 2.3 by 1.2-1.4 mm, fleshy, outside with stellate hairs, especially in middle, inside glabrous; petals ca. 7 by 3.5 mm; narrow hirsute band (disc) at base of androecium (Fig. 4A); stamens yellow, filaments seemingly united in an androphore, free part of filaments thread-like, ca. 1 by 0.1 mm, glabrous, anthers elliptic, ca. 0.8 by 0.5 mm, glabrous. Pistillate flowers ca. 21 mm diameter; pedicel ca. 7.5 mm long, hairy; calyx with five lobes, lobes ovate, ca. 4.2 by 2 mm, with stellate hairs outside; petals oblong-obovate, ca. 11 by 5 mm; disc not observed; ovary densely stellate-hairy, ca. 3 mm high, tapering into a hairy style; stigmas ca. 4 mm long, seemingly only apically split. Fruits dehiscing, ellipsoid, ca. 15 by 14 mm, green, stellate and dendritic hairs brown, hairs with arms all of same length (no longer central arm); wall ca. 1 mm thick; sepals persistent; columella ca. 1.2 cm high, narrow, apically not much widened. Seeds very immature, apically with an extension of a white two-lobed caruncle-like structure. Distribution: Indonesia, N Moluccas (Maluku Utara), Halmahera, northeast of Weda, vicinity of Weda Bay at Sake South (Fig. 1).

Weda lutea
Habitat and Ecology: (Open) secondary forest at 78-103 m altitude. Flowering in October and November, fruiting in October.
Etymology: The epithet refers to the yellow color of the petals.
Conservation: The distribution of the Weda lutea is not adequately known, as species inventories of other areas in Halmahera and nearby islands are incomplete (see Callmander et al., 2015). It appears to be a local endemic and might at least be vulnerable; it is especially at risk because it seemingly occurs only at low altitude and will be especially endangered when the planned nickel mine begins operation.
Note: The stamens of W. lutea were difficult to observe due to the very few open staminate flowers. It is unclear as to whether the stamens are inserted on a highly domed receptacle (as is more the case in W. fragarioides) or form an androphore around whose base is a vague circular, very regular annular hirsute disc. The seeds have a poorly understood caruncule-like appendage, not observed in W. fragarioides. at Weda Bay for their efforts and company in the field. Herbarium Bogoriense (BO), especially Joeni Setijo Rahaju and Harry Wiriadinata, and the Missouri Botanical Garden (MO), especially Pete Lowry and Mary Merello, are thanked for their support for the organization of the inventory and handling of specimens; BO and Naturalis Biodiversity Center (L) are thanked for help with identifications, and the laboratory of Naturalis Biodiversity Center and the Smithsonian's Laboratories of Analytical Biology of the National Museum of Natural History for their respective molecular facilities. Tissues or DNAs were provided by BO, BRI, K, L, MO, NY, US, and WAG. The first author acknowledges the help of the Treub Maatschappij, the Society for the Advancement of Research in the Tropics, for supporting the Ornstein university chair in Tropical Plant Biogeography. The help of Max van Balgooy in the initial identifications is highly appreciated, and his remark "no name" was the accelerator for this research. Ismail Rachman (BO) is thanked for his help at Weda Bay and for his general assistance in identification of all collections. The help of Rob Langelaan and Dick van der Marel (both Naturalis Biodiversity Center) is highly appreciated with the elemental analysis. Hans-Joachim (Hajo) Esser (M) advised on the Latin names.

Author contributions
Peter van Welzen (L) identified and described the taxon and conducted phylogenetic analyses; Susana Arias Guerrero (L) and Deby Arifiani (BO) managed the identification of the specimens in Leiden and Bogor, respectively; Roderick Bouman (L) helped with molecular data; Iris Tabak (L), under the guidance of Marcel Eurlings (L), did the Leiden-based lab work; Peter Phillipson (MO) was one of the organizers of the inventory program and collected specimens with Tjut Bangun and Iska Gushilman (BO); Phillipson also provided photos of living plants; Esmée Winkel (L) drew the illustrations; and Kenneth Wurdack (US) generated the Smithsonian data, developed the final molecular data sets and alignments, and did the final molecular analyses. All the abovementioned researchers helped to write the manuscript.