Recircumscription of Bredia and resurrection of Tashiroea (Sonerileae, Melastomataceae) with description of a new species T.villosa

Abstract Bredia (Melastomataceae) is an Asian genus that extends from central and southern mainland China to Taiwan and the Ryukyu islands. Molecular phylogenetic analyses reveal that the type of Bredia is nested in a clade of 20 species, while Tashiroea, a genus previously synonymized in Bredia, falls in another distantly related clade of 10 species. Our morphological survey shows that the two clades can be distinguished by several diagnostic features including leaf indumentum, texture, leaf surface sculpture under SEM, presence/absence of yellowish uniseriate trichomes, and capsule morphology. Based on molecular and morphological evidence, Bredia is recircumscribed and Tashiroea is resurrected. Description and a list of species are provided for the two genera with the description of a new species, T.villosa.


Introduction
Bredia Blume (Sonerileae, Melastomataceae) was originally described based on B. hirsuta Blume (Blume 1849), a species in Taiwan and the Ryukyu islands. Later, Ito and Matsumura (1899) established another genus, Tashiroea Matsumura, to accommodate two new species, T. yaeyamensis Matsum. and T. okinawensis Matsum., also discovered in the Ryukyus. Diels (1924Diels ( , 1932 recognized both genera. He added one Chinese species to Tashiroea and expanded Bredia by describing new species and transferring species from Blastus Lour., Otanthera Blume and Fordiophyton Stapf into Bredia. Li (1944) followed Diels's concept of Bredia but considered Tashiroea to be included within the limits of Bredia. He therefore synonymized Tashiroea, accommodating its three species within Bredia section Tashiroea (Li 1944). With additional descriptions of new species and transfers of other species (Hooker 1871;Cogniaux 1891;Diels 1924Diels , 1932Ohwi 1936;Li 1944Li , 1945Lauener 1972;Chen 1984a;Yeh et al. 2008;Zhao et al. 2017;Zhou et al. 2018), Bredia as currently circumscribed comprises 18-22 species according to different species delimitations (e.g. Chen 1984b;Chen and Renner 2007), and occurs from northern Vietnam and southern mainland China to Taiwan and the Ryukyu islands.
The generic circumscription of Bredia has long been problematic. Bredia is morphologically closely related to Phyllagathis Blume (Blume 1849;Ito and Matsumura 1899;Li 1944;Chen 1984a;Hansen 1992;Chen and Renner 2007), another Asian genus with overlapping geographical range. Delimitation of the two genera was traditionally based on staminal morphology (equal or only slightly unequal, isomorphic vs. dimorphic, subequal, or distinctly unequal stamens), which is rather obscure when species with subequal stamens are concerned. The vague generic boundary is reflected in the taxonomic history of some species that have been moved back and forth between the two genera by different authors (Diels 1932;Merrill and Chun 1940;Li 1944;Chen 1979Chen , 1984bHansen 1992;Chen and Renner 2007) and also in two phylogenetic studies based on very limited taxon sampling (Zeng et al. 2016;Zhou et al. 2018).
A recent molecular phylogenetic study, with extensive sampling of Bredia (19 species) and Phyllagathis (35 species), has shed new light on their generic delimitation (Zhou et al. 2019). Staminal characters traditionally used to separate Bredia and Phyllagathis were shown to be highly homoplasious (Zhou et al. 2019). Both genera as currently defined are not monophyletic. To facilitate discussion, we reconstructed a phylogenetic tree using the combined dataset (nrITS and chloroplast trnV-trnM) published in Zhou et al. (2019) plus eight newly sequenced species (see methods). As shown in Fig. 1. Bredia hirsuta, the type of Bredia is clustered within a clade of 20 species, viz. 13 of Bredia and seven of Phyllagathis, (hereafter referred to as the Bredia clade), whereas the two species originally published in Tashiroea, B. yaeyamensis (Matsum.) H.L. Li (=T. yaeyamensis) and B. okinawensis (Matsum.) H.L. Li (= T. okinawensis), fall in another clade comprising the new species, seven species of Bredia and two of Phyllagathis (hereafter referred to as the Tashiroea clade). The Bredia clade is close to Blastus, Fordiophyton and Plagiopetalum Rehder, while the Tashiroea clade is most closely related to Scorpiothyrsus H.L. Li and Driessenia axantha Korth. (Fig. 1). These findings are strongly corroborated by our chloroplast phylogenomic analyses (unpublished), indicating that the Tashiroea clade is an independent lineage distantly related to the Bredia clade. The same conclusion is also reached in a recently published study, although based on limited sampling of species (Kokubugata et al. 2019). Continued use of the non-monophyletic Bredia may hinder further study of this group and cause problems in describing new species, which has led us to update its circumscription.
In this paper, we revisited the morphological characters of the Bredia and Tashiroea clades in search of possible diagnostic characters. The results are presented below. Based on molecular and morphological evidence, Bredia is recircumscribed and Tashiroea is resurrected. Descriptions and a list of species are provided for the two genera with the description of a new species T. villosa. Thirteen new combinations are made.

Morphological comparison
All species of the Bredia clade (20 species) and the Tashiroea (Diels) H.L. Li (= T. sinensis Diels), B. yaeyamensis (= T. yaeyamensis), P. nudipes C. Chen, P. oligotricha Merr. and the new species T. villosa from the latter clade. Their habit, indumentum, shape and texture of the leaves, leaf surface sculpture under scanning electronic microscope (SEM), inflorescence type, stamen morphology, capsule morphology, habitat preference and geographical distribution were recorded. Data were obtained via field, herbarium and literature surveys as well as by observing living plants in the facilities of Sun Yat-sen University. Specimens of the two clades (GXMG,  GXMI, HNNU, IBG, IBK, IBSC, JJF, KUN, NAS, PE, SYS) or their high-resolution  photos (A, BM, CSFI, E, HAST, K, KYO, MO, NTUF, NY, PH, TAI, TAIF, TI, UC, WU) were examined. Habit, shape and texture of the leaves, and inflorescence type were obtained via visual observation. Stamen morphology and capsule morphology were determined by stereomicroscopic (Leica S8APO) examination, leaf epidermal features by desktop SEM (Phenom Pro), and indumentum by both equipment. During observation using Phenom Pro, fresh or dried tissues were directly mounted on stubs and examined without further processing. We failed to obtain the materials of two species recorded in Bredia, namely B. laisherana C.L. Yeh & C.R. Yeh and B. violacea H.L. Li, which were therefore not included in our molecular phylogenetic study. For the two species, we examined their protologue, images of herbarium specimens and color photos from the author of B. laisherana. Species circumscriptions basically follow Chen (1984b). Description of the capsule morphology mainly follows Hansen (1992), while the description of one type of trichomes (sessile glands with thin-walled heads) was taken from Wurdack (1986).

Taxonomy
Diels published several names in Melastomataceae in 1924 and 1932. Although he designated type specimens, he did not specify the place of deposition for them in the protologue. Even if we assume the holotypes existed in the Berlin herbarium (where Diels worked), they were probably destroyed during the Second World War. Therefore, we follow McNeill (2014) and designate lectotypes for those taxa according to the stipulations of Art. 9.12, 9.22 and 9.23 of the Code (Turland et al. 2018). Diels (1932) also lectotypified two names, Barthea cavaleriei H. Lév. and Fordiophyton tuberculatum Guillaumin, without citing the herbaria where they were housed. For Barthea cavaleriei, only one specimen of the gathering designated by Diels is known and is in the Royal Botanic Garden Edinburgh herbarium (E). We accept it as the lectotype of this name and cite the page number of Diels's lectotypification. For Fordiophyton tuberculatum, multiple specimens of the gathering designated by Diels (1932) are mixtures of more than one taxon. Based on Art. 9.17 (Turland et al. 2018), a lectotype corresponding most closely with the original description is designated for this name. Tashiroea, T. yaeyamensis and T. okinawensis were described simultaneously without designation of a type (Ito and Matsumura 1899). We also lectotypify these names here according to Art. 8.1, 8.2, 10.1, and 10.2 (Turland et al. 2018). In general, we select the herbarium sheet with the best-preserved leaves and flowers for lectotypification. Detailed reasoning where the choice may not be obvious is provided at the end of the taxon being discussed.

Results
The combined dataset contained 1768 characters. The phylogenetic tree resulted from ML analysis is shown in Fig. 1 with PP and ML bootstrap support values (BS) indicated at nodes. Phylogenetic relationships within Sonerileae/Dissochaeteae are near- Characters of the Tashiroea clade (10 species) and the Bredia clade (20 species) are summarized in Table 1. Species of the two clades are shrubs or shrublets, except that the latter clade includes some herbs. The mature leaves of species in Tashiroea clade are usually glabrous, stiffly papery to leathery and exhibit furrowed surface sculpture under SEM (Figs 2, 4 A-B, D), except that those of the new species T. villosa are densely pubescent and villous and without furrowed surface sculpture (Fig. 4C). The mature leaves of species in the Bredia clade are usually puberulous and papery and without furrowed surface sculpture (Figs 3, 4 E-L). Species in both clades have sessile glands and      shape (lanceolate to suborbicular), leaf base (cuneate to cordate), inflorescence type (cymose to cymose paniculate), stamen morphology (dimorphic to isomorphic), and staminal appendages (gibbose, tuberculate, or spurred) (Figs 2-3, 6-7). An ovary crown is present in most species of the Tashiroea clade at anthesis (Fig. 8A-D), but is usually evanescent, resulting in uncrowned capsules ( Fig. 8I-K). Exceptions are two species traditionally placed in Phyllagathis. Their crowns are persistent and enlarged, forming crowned capsules with an obpyramidal depression at the top (Fig. 8L). In species of the Bredia clade, an ovary crown is present at anthesis (Fig. 8E-H), and persistent and enlarged in old capsule, enclosing an inverted frustum-shaped depression ( Fig. 8M-P). Field and literature surveys revealed that species of both Tashiroea clade and Bredia clade prefer shaded or moist habitats in forests or along forest margin, often along stream banks, from 50 m to 2500 m alt. The Tashiroea clade occurs in southeastern mainland China and the Ryukyu islands, while the Bredia clade, aside from the above areas, also extends to the central and southwestern Chinese provinces of Guizhou, Hubei, Chongqing, Sichuan and Yunnan.
Of the two species without molecular data, B. laisherana is endemic to Taiwan and B. violacea occurs in northern Vietnam. The former species has glabrous stems and leaves, and uncrowned ovaries and capsules (Yeh et al. 2008), occurring in dense forests, while the latter has puberulous stems, leaves and inflorescences but no habitat record is available.

Morphological comparison
Species of the Tashiroea clade and Bredia clade are similar in habit, habitat preference, most indumentum types, and in having lanceolate, ovate to suborbicular leaves, a cuneate to cordate leaf base, cymose to cymose panicles, dimorphic or isomorphic stamens, and gibbose, tuberculate, or spurred staminal appendages (Figs 2-3, 5-7). Additionally, their range of distribution partly overlaps in southeastern mainland China, and the Ryukyu islands. Therefore, it is quite understandable that Li (1944) considered Tashiroea to be congeneric with Bredia, where he placed it in synonymy.
However, our survey of morphological characters has shown that the two clades can be distinguished by several diagnostic features, including leaf indumentum, texture, leaf surface sculpture under SEM, presence/absence of yellowish uniseriate trichomes, and capsule morphology. Species of the Tashiroea clade differ from those of the Bredia clade by the glabrous, stiffly papery to leathery mature leaves with furrowed surface sculpture under SEM (vs. puberulous and papery, without furrowed sculpture) (Figs 2-4), the presence of yellowish uniseriate trichomes on buds (vs. absence) (Fig. 5), and usually uncrowned capsules with a rounded or 4-humped apex, or rarely crowned with an obpyramidal apical depression (vs. capsules crowned, with an inverted frustum-shaped depression) (Fig. 8). Tashiroea villosa, the only exception in the Tashiroea clade with densely hairy leaves and without furrowed surface sculpture, nevertheless has yellowish uniseriate trichomes on buds and uncrowned capsules with a 4-humped apex characteristic of this clade. At least some of the above differences, such as indumentum and capsule morphology, had long been noticed by previous authors. Li (1944) established section Tashiroea in Bredia, whereas Hansen (1992) divided Bredia into four groups of species, placing Tashiroea in a separate group.

Species without molecular data
Two species previously placed in Bredia were not sampled in the molecular phylogenetic studies: B. laisherana from Taiwan and B. violacea from north Vietnam. Bredia laisherana was initially identified as B. quadrangularis by Yeh and Yeh (2006) and was later published as a new species (Yeh et al. 2008). It has glabrous stems and leaves, and uncrowned ovaries and capsules (Yeh et al. 2008), which are characteristic of the Tashiroea clade. Bredia vio-lacea has puberulous stems, leaves and inflorescences (Li 1945). Li (1945)

Conclusion
Morphological evidence and molecular phylogenetic data confirmed that the Tashiroea clade and the Bredia clade represent two distantly related lineages morphologically well differentiated from each other and from their possible relatives. We therefore resurrect the generic name Tashiroea Matsum for the former clade and redefine Bredia Blume to include the latter. For the species lack of molecular data, we place B. laisherana and B. violacea in Tashiroea and Bredia respectively based on morphology. Species circumscriptions basically follow Chen (1984b). A revision at the species level will be dealt with in another study.
Ito and Matsumura (1899) established Tashiroea based on T. yaeyamensis and T. okinawensis without designating a type. The original materials of both species conform to the protologue of Tashiroea. We designate T. yaeyamensis as the type of Tashiroea because this species bears much larger leaves and more flowers than T. okinawensis, which may facilitate future molecular and/or morphological analysis of the type materials.
Etymology. The specific epithet refers to the dense pubescent and villous indumentum.
Distribution and ecology. Tashiroea villosa is currently known from Pingnan, Jianou and Jianyang, northern Fujian, China (Fig. 9). It often grows in grasses and bushes along streamside at elevations of 900-1400 m.
Notes. Tashiroea villosa is discovered by Mr. Xiang-xiu Su. He is an amateur collector in Fujian who had made an important contribution to the description of this new species. We therefore include him as the author of this name. Tashiroea villosa is the sole species currently known in Tashiroea with densely puberulous and villous leaves (vs. glabrous) and smooth leaf surface sculpture (vs. furrowed). It is morphologically and phylogenetically closest to T. amoena. The two species are similar in height, leaf size and shape, inflorescence and stamen morphology. Tashiroea villosa is distinct from T. amoena in the dense indumentum covering the whole plant (vs. petioles and inflorescences pubescent or sometimes glabrescent) and much larger bracts (9-19 × 5-8 mm vs. 1-2 × 1 mm) in the florescence (Fig. 12). Geographically, T. amoena is widely distributed in southeastern China (Anhui, Fujian, Zhejiang, Jiangxi, Guangxi), whereas T. villosa occurs in northern Fujian where both species occur. Nevertheless, they have not been found to co-occur within the same habitat. Description. Shrubs, shrublets or herbs, erect, ascending or creeping. Stems terete or more or less 4-sided, sparsely to densely puberulous, rarely glabrescent. Leaves petiolate; leaf blade ovate, cordate, oblong, elliptic, ovate-orbicular, rarely lanceolate, papery, rarely submembranous, sparsely to densely puberulous or strigose, secondary veins 2-5 on each side of midvein, margin serrulate or entire. Inflorescences terminal, umbellate, cymes or cymose panicles. Flowers 4-merous. Hypanthium funnel shaped to campanulate. Calyx lobes conspicuous, linear-lanceolate to triangular. Petals pink or purplish red, ovate to oblong, more or less oblique, apex acute or acuminate. Stamens 8, unequal or subequal; filaments filiform; anthers dimorphic or isomorphic, subulate to oblong-linear, gibbose, tuberculate or spurred at base, rarely unappendaged abaxially. Ovary half inferior, crowned, ovoid, 4-celled. Style filiform; stigma apiculate. Capsule turbinate to cup-shaped, more or less 4-sided, crown persistent and enlarged, enclosing an inverted frustum-shaped depression at capsule apex. Seed numerous, minute, cuneate, densely granulate. ( Distribution. Twenty-one species: 15 in central and southern mainland China (Fujian, Guangdong, Guangxi, Guizhou, Hubei, Hunan, Jiangxi, Sichuan, Chongqing, Yunnan, Zhejiang), one in north Vietnam, five in Taiwan and one extending to the Ryukyu islands (Fig. 13).   Barthea esquirolii H. Lév., Repert. Spec. Nov. Regni Veg. 11(301-303): 494. 1913 (Basionym) In the protologue, Blume (1849) cited no specimen but the name "Rhexia fasikan" (should be Rhexia hasikan, see below) attached to a specimen from Von Siebold's herbarium. Ohba and Akiyama (pers. comm.) discovered one sheet with the name "Rhexia hasikan" in the herbarium in Leiden (L0170980). The sheet contains two col-lections, a smaller one covered by paper on the upper part and a larger one with four branches on the middle and lower part. According to Ohba (pers. comm.), the epithet "hasikan" comes from the Japanese name of B. hirsuta, viz. "Hashikan-boku". This sheet is regarded as the type material of B. hirsuta. We here designate the larger collection on the sheet as the lectotype of B. hirsuta.  Diels (1932) designated Delavay 5053 as the type without citing a specific herbarium. We located four duplicates of Delavay 5053 (E00285959, P02274729, P02274730, P02274731). Close examination of the specimens revealed that this gathering is a mixture of B. tuberculata and B. yunnanensis. Individuals on E00285959, P02274731 and the one on the right side of P02274729 conform to the protologue of B. tuberculata in the adaxially white punctate leaves and unequal and dimorphic stamens. The remaining four individuals (three on P02274730 and one on the left side of P02274729) conform to the holotype of B. yunnanensis in the absence of white spots on their leaves and the subequal and isomorphic stamens. According to Art. 9.11, 9.12, 9.14, and 9.17 (Turland et al. 2018), the lectotypification by Diels should be accepted and further narrowed to a single specimen that corresponds most closely with the original description. We therefore designate E00285959 as the lectotype as it conforms to the description and contains many well-preserved leaves and flowers.