Cheniella gen. nov. (Leguminosae: Cercidoideae) from southern China, Indochina and Malesia

For much of the last thirty years, the caesalpinioid genus Bauhinia has been recognised by numerous authors as a broadly circumscribed, ecologically, morphologically and palynologically diverse pantropical taxon, comprising several subgenera. One of these, Bauhinia subg. Phanera has recently been reinstated at generic rank based on a synthesis of morphological and molecular data. Nevertheless, there remains considerable diversity within Phanera. Following a review of palynological and molecular studies of Phanera in conjunction with a careful re-examination of the morphological heterogeneity within the genus, we have found strong evidence that the species of Phanera subsect. Corymbosae are a natural group that warrant generic status. We describe here the genus Cheniella R.Clark & Mackinder gen. nov. to accommodate them. It comprises 10 species and 3 subspecies, one newly described here. Generic characters include leaves that are simple and emarginate or bilobed; fl owers with elongate hypanthia which are as long as or much longer than the sepals; pods that are glabrous, compressed, oblong, indehiscent or tardily dehiscent; and with numerous seeds, the seeds bearing an unusually long funicle extending most of the way around their circumference. A further distinctive fl oral character was found to be a fl eshy disc on which the staminodes are mounted. An analysis carried out for this study reveals Cheniella to be characterised by a pollen type that is unique to the genus and previously unknown in the Leguminosae. Species diversity is richest in southern China, the full distribution extending westward to India and southand eastward through Indochina into Malesia.


Phanera circumscription and relationships
Until recently, Phanera was considered to comprise 120-130 lianescent species from southeast Asia and South America (Lewis & Forest 2005). However, following the segregation of the neotropical species as the genus Schnella (Trethowan et al. 2015), Phanera was eff ectively recircumscribed as a group of ca 90 species, restricted to southeast Asia. This genus, which is most diverse in Malesia, remains the largest of the segregate genera reinstated from Bauhinia s. lat. (Mackinder & Clark 2014). Nevertheless, Phanera is still a somewhat heterogeneous taxon. In particular, as several authors have already noted, the generic boundary of Phanera with regard to Lasiobema warrants more detailed research (Hao et al. 2003;Lewis & Forest 2005;Mackinder & Clark 2014;Trethowan et al. 2015). Sinou et al. (2009) indicated, albeit with weak support, that based on a phylogenetic analysis of nucleotide sequence data from the cpDNA trnL-F region, Asian Phanera are closely related to Lasiobema; although sampling in that study was very limited. Both the close relationship between Phanera and Lasiobema, as well as the wider infrageneric phylogenetic framework of Cercidoideae (the then Cercideae) resolved by Sinou et al. (2009), were congruent with the fi ndings of Hao et al. (2003) in a phylogenetic study of the Cercidoideae (as the Cercideae) based on nucleotide data from the Internal Transcribed Spacer (ITS) region. The phylogenetic structure of the main lineages as determined by those two studies is compared (Fig. 1). In Hao et al. (2003), the Schnella lineage was represented by an accession of Bauhinia glabra Jacq. now recognised as Schnella glabra (Jacq.) Dugand. In Sinou et al. (2009), several accessions of Schnella were sampled and placed together on a lineage labelled as American Phanera in Fig. 1 of that publication.
The study of Hao et al. (2003), designed to investigate the close relationship between SE Asian Phanera and Lasiobema, recovered a strongly supported clade containing 31 accessions, representing 27 species, of which 23 were species of Phanera and four were species of Lasiobema. Neither genus was resolved as individually monophyletic. A plausible evolutionary interpretation of these fi ndings is that neither Phanera nor Lasiobema are natural groups as currently delimited and that, given the morphological diversity that still exists within both genera, a further division of both genera is likely to be necessary if we are to achieve a classifi cation that refl ects monophyly. Indeed, Wunderlin et al. (1987) proposed A. Topology of the parsimony strict consensus tree based on the nrDNA Internal Transcribed Spacer region (Hao et al. 2003). B. Bayesian majority rule tree topology derived from an analysis of data from the cpDNA trnL-F region (Sinou et al. 2009). Note that the reinstatement of the genus Schnella (Wunderlin 2010) post-dates both phylogenetic studies. that the Asian species of Phanera (treated by him in that publication under Bauhinia subg. Phanera) should be further divided among seven sections, and that the species of Lasiobema be placed in three further sections. Larsen (1975) demonstrated two diff erent pollen types in Lasiobema which supported the hypothesis of de Wit (1956) that Lasiobema may be better recognized as two genera based on fl oral morphology.
Nonetheless, within Phanera s. lat., the species of Phanera sect. Corymbosae de Wit do form a morphologically homogeneous group that is readily identifi able as distinct from all other Phanera diversity. Five species of Phanera sect. Corymbosae were sampled by Hao et al. (2003) and were resolved as a strongly supported (100% bootstrap) monophyletic group and placed as sister to a clade containing all other accessions of Phanera and Lasiobema species. Here we present the case for the recognition of Phanera sect. Corymbosae as a genus in its own right.

Material and methods
The synoptical descriptions, the key to species, and distribution data were created using a combination of the existing literature, herbarium specimens at BM, E and K, and those available online via JSTOR (http:// plants.jstor.org) and the Chinese Virtual Herbarium (http://www.cvh.ac.cn/). Preliminary conservation assessments were made using the categories and criteria of IUCN (2001). Locality data were derived from specimens from the herbaria listed above as well as specimens detailed in the Flore du Camboge, Laos, et Vietnam (Larsen et al. 1980). It was not possible to use locality data from the Flora of China for the assessments, as only occurrence by province is recorded; thus the Extent of Occurrence (EOO) reported in this account may in some cases be smaller than the true value, which in turn may have the eff ect of slightly infl ating the category of threat.
The systematic arrangement of the species was created by inferring relationships based on morphological similarity, and on geographical distribution.
For pollen studies, mature buds were obtained from herbarium specimens held at the Royal Botanic Gardens, Kew (K) (Appendix 1). These were dissected in a 1% solution of nonionic wetting agent using a Leica Wild M8 microscope. Pollen was washed and then acetolysed (Erdtman 1960). The pollen grains were prepared for examination by scanning electron microscope (SEM) by pipetting acetolysed pollen exines in 95% ethanol onto specimen stubs. These were left to air dry before being sputter-coated with platinum using a Quorum 150TES sputter coater. They were examined using a Hitachi S-4700 II fi eld emission SEM. The pollen terminology follows the glossary published by Punt et al. (2007).
Of the samples used for the fl oral anatomy study with SEM, specimen Owens s.n. was stored in Kew Mix (5% formaldehyde, 5% glycerol, 37% water, 53% industrial methylated spirit), and fl owers of Nichols 1292 were taken from dried herbarium specimens and rehydrated in boiling water for 10 minutes. Specimens Clark 418, 426 and 458 were collected into and preserved in 70% ethanol. For scanning electron microscopy (SEM), fl owers and fl oral buds were dissected in 70% ethanol, dehydrated to 100% ethanol, and critical point dried using an Autosamdri-815B critical-point dryer. Dried samples were mounted onto aluminum stubs and coated with platinum in a Quorum Q150T sputter coater. SEM images were taken with a Hitachi S-4700-II cold fi eld emission SEM.
Secondary veins as defi ned in this manuscript are the leaf veins arising from the leaf base, and approximately parallel to the primary central vein. Measurements in square brackets are those derived from literature, not observed by the present authors.

Proposal to recognise Phanera sect. Corymbosae as a distinct genus
In his detailed account of the (then) 43 Malesian species of Phanera, de Wit (1956) fi rst delimited Phanera sect. Corymbosae to accommodate Phanera corymbosa Benth. (type) and Phanera glauca Benth. These two species have narrow, elongate hypanthia, the length of which greatly exceeds the length of the sepals, in contrast to all other species from the region, which have a hypanthium that is in most cases much shorter than, or occasionally equal to the sepals. De Wit (1956) also observed that the compressed fruits of Phanera corymbosa and Phanera glauca contained numerous seeds, many more than other species from the region. Both Phanera corymbosa and Phanera glauca have glabrous ovaries, a character that is unusual among species of Phanera. In 1988, T.C. Chen described three new legume species from China (in Bauhinia s. lat.) as Bauhinia ovatifolia T.C.Chen, Bauhinia damiaoshanensis T.C.Chen and Bauhinia quinnanensis T.C.Chen which also possess the combination of elongate hypanthia and a glabrous or subglabrous ovary (Chen 1988).
In a cladistic analysis of Bauhinia, Zhang & Chen (1992) expanded the Corymbosae group, which they defi ned as a series within Bauhinia, to include the three recently described species of Chen (1988), and additionally Bauhinia clemensiorum Merr., Bauhinia didyma H.Y.Chen, Bauhinia lakhonensis Gagnep. and Bauhinia touranensis Gagnep. We have found that all species included in Bauhinia ser. Corymbosae as defi ned by Zhang & Chen (1992) exhibit the elongate hypanthium, glabrous or subglabrous ovary, and the long, glabrous, indehiscent (or tardily dehiscent), many-seeded fruits that characteristic of the section.
In his description, de Wit (1956) described the seeds of the two original species as having a funicle that extends as a rim along more than half the circumference of the seed, a funicle type unknown in other species of Phanera. We have also seen seeds of Bauhinia clemensiorum, Bauhinia quinnanensis and Bauhinia tenuifl ora Watt ex C.B.Clarke and confi rm that they too have this unusual type of funicle. Flowers of the Corymbosae contain seven staminodes, of which fi ve are mounted on a fl eshy disc, a structure that appears to be unique within the context of Phanera s. lat. Zhang (1994) concluded Bauhinia sect. Corymbosae to be a natural group based on his study of leaf venation in Cercidoideae (the then Cercideae), describing the Corymbosae leaf as having "primary veins not marginal, the marginal ultimate venation looped, well developed areoles, and simple veinlets".
In this study, we raise the status of Bauhinia glauca subsp. tenuifl ora (Watt ex C.B.Clarke) K.Larsen & S.S.Larsen to the rank of species, bringing to 10 the total number of species which we consider to belong to Phanera sect. Corymbosae of de Wit (1956). Together, the 10 species of sect. Corymbosae form a morphologically homogeneous group that can be clearly distinguished from the rest of Phanera.
Pollen morphology in legumes has frequently been demonstrated to be taxonomically informative at the genus level, and overviews of the great diversity of pollen morphology found in subfam. Caesalpinioideae (as then circumscribed) have previously been documented by Graham & Barker (1981), and Banks (Banks & Klitgaard 2000;Banks et al. 2003;Banks et al. 2006). The overview of pollen in the subfam. Caesalpinioideae (former circumscription) shows that the greatest diversity in pollen structure occurs in the earlier branching taxa, while the taxa of later diverging lineages appear to have more constrained pollen morphology (Banks et al. 2003). The pollen of Bauhinia s. lat. is known to exhibit a great range of morphological features, in terms of ornamentation, apertures, and units of dispersal (dos Santos et al. 2012).
The monophyletic Cercidoideae is one of the earliest branching lineages in Leguminosae, sister to the rest of the family (LPWG 2013(LPWG , 2017. The pollen of the Cercidoideae has long been known to be diverse and varied, but problematic species and generic delimitations have previously hindered our understanding of the similarities, diff erences, evolution and taxonomic signifi cance of the various pollen structures found within the tribe, until the studies of Banks et al. (2013Banks et al. ( , 2014. Previous published studies of the pollen of Bauhinia s. lat. have mostly been restricted to regional accounts (Larsen 1975;Larsen & Larsen 1982, 1993Gamerro & Fortunato 2001) although Ferguson & Pearce (1986) carried out a wider survey of neotropical species. Ferguson & Pearce (1986) described two patterns of morphological variation in pollen and found that apertures and exine stratifi cation correlated with taxonomy, whereas ornamentation, pollen size and exine thickness showed a great deal of continuous variation and also examples of homoplasy. In many cases homoplasious morphological variation is correlated with fl oral morphology and pollination syndromes (e.g., supratectal gemmae and verrucae appeared to be correlated with bat pollination) (Ferguson & Pearce 1986) and has evolved iteratively in the tribes Cercideae and Detarieae (Banks & Rudall 2016).
Recent studies in the phylogenetics of Bauhinia s. lat. allowed us to re-evaluate pollen structures in this group (Banks et al. 2013(Banks et al. , 2014 and assess the morphological variation using a molecular based phylogenetic framework. These studies identifi ed pollen characters of potential taxonomic importance and assessed the evolution of pollen structure by examining the variation of pollen types in Bauhinia s. lat. within the phylogenetic context provided by the study of Sinou et al. (2009). The pollen of Cheniella R.Clark & Mackinder gen. nov. is revealed to have the structure fi rst described by Larsen (1975) as the "Glauca type". Cheniella gen. nov. pollen is isopolar, tricolporate, with apertures usually syncolporate ( Fig. 2A-E), and is oblate spheroidal to oblate in shape ( Fig. 2A, C-E). Ornamentation is coarsely rugulate to verrucate (Fig. 2F), with larger verrucae along the aperture margins ( Fig. 2B) and sometimes anastomosing at aperture margins, forming a defi ned margin to the apertures. Aperture membranes are verrucate, with coarser verrucae over the pores forming operculae ( Fig. 2A-E).
The palynological investigation carried out as part of this study indicates that the species of the sect. Corymbosae (four species sampled, of a total of nine at the time of study) share a pollen morphology that is unique within Bauhinia, and within the Cercidoideae. No other pollen in the Leguminosae is known to have the combination of syncolporate apertures and verrucate ornamentation (Fig. 2B). However, similar types of verrucate ornamentation have previously been seen in taxa pollinated by butterfl ies, moths, birds or bats (Banks & Rudall 2016). Hao et al. (2003) constructed a well resolved phylogeny of the subtribe Bauhiniinae (sensu Wunderlin 1979) in which 7 accessions were sampled that represented 5 species from sect. Corymbosae, including the type. The 7 accessions formed a strongly supported (bootstrap = 100%) monophyletic group placed as sister (bootstrap = 89%) to all other accessions sampled from Phanera and Lasiobema (Fig. 1).
The centre of diversity of sect. Corymbosae is southern China and northern Indo-China, the distribution being principally more northern and more seasonal than that of the majority of Phanera and Lasiobema, which extend into regions that are more tropical, lacking a dry season (Fig. 3).

Etymology
Named for Prof. Chen Dezhao (Chen Te-chao), who described three of the (then) fi ve species of Cheniella gen. nov. known from China. The fossil species Bauhinia cheniae Qi Wang, Z.Q.Song, Y.F.Chen, S.Shen & Z.Y.Li (Wang et al. 2014) from the late Eocene-Oligocene Ningming Formation in Ningming County (Guangxi Zhuang Autonomous Region, South China) was likewise named for Prof. Chen Dezhao.

Preliminary conservation assessment
Extent of Occurrence 322,120.231 km 2 = Least Concern (LC). Incorporating all subspecies, this species is distributed over a fairly wide area from southern China into northern Myanmar and India, and seems to be able to occupy a variety of habitats. It is therefore here considered to be Least Concern, although it should be noted that it is quite poorly represented in herbaria (ca 18 specimens known to the present authors), which could indicate that it is relatively uncommon within its range.
Key to the subspecies of C. quinnanensis comb. nov.

Taxonomic and geographical notes
Phanera glauca subsp. tenuifl ora var. murlenensis is here synonymised with C. quinnanensis comb. nov. on the grounds that the description does not diff er signifi cantly from that of C. quinnanensis comb. nov., and that the specimen's locality is in Mizoram state in northern India, which is adjacent to Assam state in which C. quinnanensis comb. nov. subsp. quinnanensis is known to occur.

Distribution
Endemic to India (Arunachal Pradesh).

Taxonomic notes
This taxon was described as Phanera glauca subsp. tenuifl ora var. gandhiana; however, we consider the variety to be more closely aligned with C. quinnanensis comb. nov. than with C. glauca comb. nov. The leaves are variable, some are similar in size and shape to C. quinnanensis comb. nov., and the ovary bears hairs at the base as in C. quinnanensis comb. nov., whereas the ovary of C. glauca comb. nov. is completely glabrous. Moreover, the dimensions of the hypanthium, pedicel, and stamens are more congruent with those of C. quinnanensis comb. nov. than with the taxon recognised by those authors as Phanera glauca subsp. tenuifl ora. The type collection of C. quinnanensis subsp. gandhiana comb. et stat. nov. is from Arunachal Pradesh state in northern India, which is adjacent to Assam state, in which C. quinnanensis comb. nov. subsp. quinnanensis is known to occur.
A noteable and unusual feature of this taxon is the leaf form, which is dimorphic. Leaves are ovate and emarginate, with the base rounded to subcordate, or ovate with the apex attenuate to several millimetres, and the base sub-hastate. Both leaf forms can occur on the same branch of an individual. The latter leaf form is very unusual (possibly unique) within Bauhinia s. lat., raising the possibility that this plant is an anomalous form of C. quinnanensis comb. nov.; however, in the absence of opportunity to study the material directly, or further evidence to support this hypothesis, we here maintain the distinct status of the taxon, reclassifying it as a subspecies of C. quinnanensis comb. nov.

Ecology
Open habitats on hill slopes; margin of disturbed forest on limestone; elevation 700-830 m a.s.l.

Field notes
Lianescent shrub to 4 m; calyx green, striped pink and white; petals white; staminodes bright red at base to white at tip; fi laments white; anthers red; sweetly fragrant.

Preliminary conservation assessment
Few herbarium specimens exist of this species, and it is not possible here to generate an EOO value due to limited access to locality data. The species is reported from Guangxi and Yunnan provinces (Dezhao et al. 2010), and from Vietnam; however, these two provinces are adjacent, and the Vietnamese specimen is from very near the border with Guangxi, so its total range may be small. The paucity of herbarium material available suggests that the species is uncommon. However, it must also be noted that the species occurs on limestone above 700 m a.s.l., which may constitute terrain that is diffi cult to access for collecting, and that this may account for the low number of specimens present in herbaria. We here consider the species to be Data Defi cient (DD).

Taxonomic notes
The authors have been unable to access the type material of the species for this study, nor any positively verifi ed material. A total of seven specimens identifi ed as Bauhinia ovatifolia are available via the Chinese Virtual Herbarium, but their identity is in question because they are, with one exception, fruiting specimens, and have measurements that do not agree with those given in the protologue of the species. The data from these specimens has therefore not been included in this account. During this study, a fl owering specimen (Harder et al. 4249) was examined by the authors and, based upon the description in the protologue, we believe that it represents C. ovatifolia comb. nov. Measurements of the Harder collection are included in the description given here, and the specimen has been used to create part of the illustration for the genus (Fig. 4).

Distribution
Endemic to Vietnam.

Ecology
Evergreen closed forest on yellow ferralitic soil, forest margin; elevation ca 100 m a.s.l.

Field notes
Flowers pink-red.

Ecology
Primary lowland forest, primary broad-leaved evergreen closed wet forest on limestone, disturbed agricultural fi elds, gorge, over rocks; elevation to 1,100 m a.s.l.

Preliminary conservation assessment
Extent of Occurrence 262,775.118 km 2 = Least Concern (LC). This species is fairly well represented in herbaria. It is distributed throughout northern Vietnam, extending into central Laos. Its primary habitat is forest, but it seems also able to inhabit open or disturbed areas. It is here considered to be of Least Concern (LC) due to its fairly broad range and apparent suitability to diff erent habitat types.

Taxonomic notes
The lectotype was selected for being the most complete specimen of the syntypes housed at Paris, where the author of the basionym worked, and where the majority of his collections are housed.

Preliminary conservation assessment
Extent of Occurrence 120,019.364 km 2 = Least Concern (LC). Cheniella corymbosa comb. nov. is known from southern China and northern Vietnam. Based on the EOO generated here, and the fact that it is moderately well represented in herbaria (more than 20 specimens have been seen for this study), the species is here considered to be of Least Concern (LC). Little is known about its ecology, but it is known to occur in forest at low altitudes, and may therefore be vulnerable to habitat destruction. Additionally, many of the specimens seen for this study are old (pre-1950), and the populations from which they were taken may no longer exist. A more detailed study of the conservation status of this species should be undertaken when further data becomes available.

Taxonomic Notes
The lectotype was selected from the three sheets of Wallich 5788 for being the most complete specimen, bearing leaves and both fl owers and fruit, whilst K000760722 also bears these but with fewer fl owers, and K000760726 bears only leaves and fruit.

Field notes
Staminodes and fi laments pink, anthers yellow-brown, stigma pink.

Preliminary conservation assessment
Extent of Occurrence 146,856.735 km 2 = Least Concern (LC). The species is found primarily in an area of southern-central Laos, and northern Thailand, with a single record in western Vietnam (near the boder with Laos), and a single record from Yunnan province, China. Without this recently discovered Chinese specimen, the EOO of the species would be considerably smaller, at 44,776.691 km 2 , in the category of Near Threatened (NT). However, this specimen shows that the species is more widespread than previously realised, and it is therefore considered here to be Least Concern (LC). It should be noted, however, that the habitat of C. lakhonensis is lowland forest, which may render it vulnerable to loss of populations due to the prevalence of deforestation in lowland areas. Larsen et al. (1980) stated the type to be Thorel s.n. from Thailand (P). It is not possible here to lectotypify the name as the present authors have not seen both syntypes.

Preliminary conservation assessment
This species occurs in Guangdong, Guangxi, and Hainan Provinces, and is thus relatively broadly distributed. It is not possible here to generate an EOO value for the species due to limited access to locality data. However, the fact that few herbarium specimens exist (10 known to the present authors) suggests that it is uncommon. If further information were available, it might be expected that a category of threat would be applicable, although the fact that the species occurs in three provinces implies that the category of threat would not be high. Given the lack of locality information available for C. didyma comb. nov., it is here determined to be Data Defi cient (DD).

Taxonomic notes
Lectotypifi cation has not been carried out as the authors have not seen all of the syntypes, and of those seen, none are fl ower-bearing.
The number of staminodes reported here derives from the protologue of Bauhinia didyma. Due to lack of access to material, the present authors have not been able to verify this. However, given that the number of staminodes in all other species of Cheniella gen. nov. is seven, it is likely that the number was misreported in the protologue, and that this species also possesses seven. It is presumed that some of the staminodes are mounted onto a raised fl eshy disc as with the other species in the genus, altough it has not been possible to observe this.

Ecology
Open forests; elevation to 1,850 m a.s.l.

Preliminary conservation assessment
This species is endemic to Guangxi Province. It is not possible to generate an EOO value for the species due to limited access to locality data. However, the facts that it is known to occur in only one province, and that few herbarium specimens exist (ca 10 known to the present authors) suggests that it is uncommon. If further information were available it is likely that a category of threat would be applicable to the species. Until such time, it is here listed as Data Defi cient (DD).

Ecology
Tropical evergreen forest, roadsides and open areas, thicket, disturbed hillside, mixed forest; elevation to 1,400 m a.s.l.

Preliminary conservation assessment
Extent of Occurrence 3,592,509.952 km 2 = Least Concern (LC). This is a widespread and apparently common species, found from central China to east India and the Gulf of Thailand, and is therefore considered here to be Least Concern (LC).

Taxonomic notes
Specimens C.B. Clarke 42304A, 42304B, 42304C and 42304D can be considered as remaining syntypes following the lectotypifi cation by Bandyopadhyay & Sharma (1992) of C.B. Clarke 42304E, because C.B. Clarke 42304 was cited in the protologue without the associated letters A, B, etc., and thus could refer to any of these specimens. Larsen et al. (1980) cited the type of Bauhinia glauca subsp. tenuifl ora to be Watt 6915 (Birmanie [Burma], E00064065!, K000760782!), but this is erroneous as this specimen is not one of the syntypes listed in the protologue.
The decision to recognise this taxon at the species level, rather than as a subspecies of Bauhinia glauca as it was designated by Larsen et al. (1980), was based on morphological diff erences, and on geographical separation (see Discussion). The morphological diff erences, highlighted in Table 1, consist of the indumentum or lack thereof on the hypanthium, the much more deeply bilobed leaf of C. glauca comb. nov., and diff erences between the size ranges of the fl oral parts and fruit characters. The diff erences between the two taxa are at least as great as those separating other species in the genus, and were therefore deemed to be suffi cient to merit the recognition of this taxon as a distinct species. Moreover, Larsen et al. (1980) did not explain their reasons for choosing to down-rank Bauhinia tenuifl ora, as it was originally described, to a subspecies of Bauhinia glauca.

Ecology
Open roadsides, by the sea, rain forest, in ravines and on slopes, evergreen forest, scrub; elevation to 1,700 m a.s.l.

Preliminary conservation assessment
Extent of Occurrence 920,396.728 km 2 = Least Concern (LC). This species is distributed over a fairly broad range, in several countries, and is represented by numerous herbarium specimens. It is therefore considered to be Least Concern (LC).

Taxonomic notes
The lectotype was selected for being the more complete specimen of the two, with more numerous and better preserved leaves.
Bauhinia tenuifl ora, formerly considered to be a subspecies of Bauhinia glauca, is here recognised as a distinct species, the reasons for which are outlined in the text for that taxon.

Specialised morphology of Cheniella gen. nov. (1): Pollination syndrome
As discussed in the description of the pollen of the Corymbosae earlier in this paper, the pollen type of Cheniella gen. nov. is indicative of pollination by butterfl ies, moths, birds or bats. Here we explore further evidence to explain the pollination syndrome of the genus.
The primary pollinators in the legume family are bees, the majority of these being the long-tongued Apidae and Megachilidae (Lewis et al. 2000). Butterfl y, moth, bird and bat pollination are also known to occur within the family, including subfam. Cercidoideae, and within Bauhinia s. lat. (Hokche & Ramírez 1990;Lewis et al. 2000).
With regard to pollination syndromes, caesalpinioid legume fl owers are in most cases less specialised than mimosoid or papilionoid fl owers. Within the Caesalpinioideae (excluding the mimosoid clade), the fl owers are usually open, with exposed pollen and nectar available to both specialised and nonspecialised pollen vectors (Arroyo 1981). However, a notable specialisation of the fl owers of Cheniella gen. nov. is the hypanthium, which is as long as, or longer than, the calyx lobes, and is in absolute terms longer than the hypanthium of most other species of Phanera s. lat. Selective eff ects of pollinators are known to generate diff erences in fl ower morphology (Lewis 1998;Pérez-Barrales et al. 2007), and in this case the elongated hypanthium of Cheniella gen. nov. is suggestive of a specialised relationship with particular species (or groups) of pollinator.
It is documented that nectar is produced and accumulated in the hypanthium of fl owers in the arborescent Bauhinia species (Hokche & Ramírez 1990), and the elongated hollow hypanthium of Cheniella gen. nov. is presumed to serve similarly as a nectar receptacle. Flowers that are specialised to attract bees or other insects with long tongues often have petals fused into a tube, and are typically "high reward" fl owers, off ering more nectar than open access fl owers (O'Toole & Raw 1991). In the case of Cheniella gen. nov., restricting access to nectar, a function served in other fl owers by a tubular corolla, could be served by the enlongated hypanthium. The long hypanthium of Cheniella gen. nov. is therefore most probably functionally related to particular pollinators, which are most likely to be long-tongued bees, moths or butterfl ies.
Evidence to support the hypothesis of pollination of the Corymbosae by bees and butterfl ies was provided by Lau et al. (2009), who studied visitation rate to the fl owers of Bauhinia corymbosa and Bauhinia glauca by invertebrates. They noted that the bright colour of the stamens and carpel in Bauhinia corymbosa, diurnal anthesis, low viscosity of the nectar, and the dense multi-fl owered infl orescences are all typical of butterfl y-pollinated fl owers, and that the fl ower characteristics of Bauhinia glauca are consistent with bee and butterfl y pollination. The sucrose dominated nectar of both Bauhinia corymbosa and Bauhinia glauca is consistent with pollination by long-tongued bees and butterfl ies. Lau et al. (2009) observed that C. corymbosa comb. nov. was visited primarily by one species of bee (Hymenoptera), and by three species of butterfl y (Lepidoptera), whilst Bauhinia glauca was visited by two species of bee and four species of butterfl y. Their fi ndings in terms of frequency of pollinator visits were that the fl owers of Bauhinia corymbosa and Bauhinia glauca received signifi cantly fewer visitations from potential pollinators than did the fl owers of Bauhinia championii. A possible reason for this is that the fl owers of Bauhinia corymbosa and Bauhinia glauca produce less nectar than those of Bauhinia championii, or that the nectar they produce is less accessible to a range of pollinators. This is compatible with the hypothesis that a longer hypanthium represents a specialised relationship with particular long-tongued pollinators.

Specialised morphology of Cheniella gen. nov. (2): Staminodes
A further noteable fl oral morphological feature of Cheniella gen. nov. is the unusual arrangement of the staminodes. There are seven of these per fl ower, two of which are inserted between the stamens, with the remaining fi ve in a separate group. The distinctive element of this arrangement within Cheniella gen. nov. is that the the group of fi ve staminodes is inserted on a fl eshy connective structure, which is raised above the hypanthium (Figs 5 & 6). This structure is not typical of the fl owers of Phanera s. lat., in which the staminodes usually arise individually, directly from the hypanthium, as in the example of Phanera yunnanensis (Franch.) Wunderlin (Fig. 6).
This staminodal ring may constitute a unique synapomorphy for Cheniella gen. nov. However, possibly homologous, or at least morphologically similar structures are reported in the literature for Phanera gracillima de Wit, and have been both reported in the literature ( de Wit 1956;Larsen et al. 1980;Hou et al. 1996) and observed on herbarium specimens by the present authors in Phanera japonica (Maxim.) H.Ohashi, Phanera wallichii (J.F.Macbr.) Thoth. and Phanera williamsii (F.Muell.) de Wit. We conclude that the fl eshy staminodal base is indeed a synapomorphy for Cheniella gen. nov., but whether it is unique within Phanera s. lat. has yet to be confi rmed.
The fl eshy staminodal disc in Cheniella gen. nov. is presumed to serve some function. In a study of staminodal function and evolution, Walker- Larsen & Harder (2000) noted that staminodes which do not serve some use within the fl ower probably interfere with the interactions between fl owers and pollen vectors, and should therefore quickly be lost to a taxon through the processes of evolution. The reduced number of stamens and staminodes in many species of Bauhinia s. lat. indicates that a process of this nature has occurred within this group. The fact that the staminodal disc appears to be a stable characteristic within Cheniella gen. nov. therefore implies a functional role, the precise nature of which is unknown, but is most likely related to pollination. Two common staminode roles highlighted by Walker- Larsen & Harder (2000) are pollinator attraction through the provision of attractants and rewards, and/or as a visual attractants.
Some species of Phanera s. lat., and particularly those of the potential segregate genus Lasiobema, possess fl oral nectariferous discs, which appear superfi cially similar to the staminodal disc of Cheniella gen. nov. The staminodes could be serving as a nectar secretor, as has been recorded in certain plant species such as within the Lauraceae Juss., Ochnaceae DC., Corynocarpaceae Engl., and Loasaceae Juss. (Walker-Larsen & Harder 2000). However, SEM images of the surface of the staminodal disc in Cheniella gen. nov. show that the structure lacks apertures through which nectar might be secreted (Fig. 7), whilst distinct apertures are visible on the nectariferous discs of Phanera s. lat. (Fig. 8). The staminodal disc of Cheniella gen. nov. appears therefore not to be homologous with the nectariferous discs found within Phanera s. lat., and its potential as a nectar-producing organ is discounted.
At maturity, the staminodal disc in fl owers of C. quinnanensis comb. nov., C. tenuifl ora comb. nov. and C. touranensis comb. nov. is visually distinct, being bright yellow in contrast to the staminodes which are white, and to the petals, which are pale yellow or greenish to white (Fig. 9). When the fl ower is old, the colour of the disc changes to dark pink-red, or dark brown (Fig. 9). In C. corymbosa comb. nov., the staminodes are pinkish red, contrasting with the white petals (Lau et al. 2009). The distinctive colour of the staminodes in all species of Cheniella gen. nov. observed by the authors indicates a function related to pollinator attraction. This hypothesis is supported by studies on Bauhinia yunnanensis (Phanera yunnanensis) by Wang et al. (unpubl. res.), in which staminodes were experimentally removed from fl owers of this species, which suggest that presence or absence of staminodes infl uences the composition of visiting pollinators and pollen removal and deposition.

Biogeography of C. glauca comb. nov. and C. tenuifl ora comb. nov.
The decision to re-establish C. glauca subsp. tenuifl ora at the species level is supported by morphological and geographical evidence. The data obtained from the specimens studied by the authors, and those seen online via the Chinese Virtual Herbarium, shows the morphological diff erences between C. glauca subsp. glauca and C. glauca subsp. tenuifl ora to be suffi ciently great to merit recognition of two distinct species. The characters that most clearly separate the two species are highlighted in bold in Table 1.
Geographically, the distributions of C. glauca comb. nov. and C. tenuifl ora comb. nov. are almost discrete, with very little overlap between the ranges (Fig. 10). Cheniella glauca comb. nov. has a more southerly distribution, in the Thai-Malay Peninsula, extending south as far as Sumatra and Java, whereas C. tenuifl ora comb. nov. occurs throughout Central and Northern Thailand, Cambodia, Laos, Vietnam, Myanmar, North-eastern India (the Seven Sister States), and Southern China. The transition between the two species occurs around the Isthmus of Kra, with a single record of C. glauca comb. nov. to the north of this region, in Myanmar. This distribution coincides with that of other plant species, many of which are confi ned to areas north or south of the Isthmus (Parnell 2013). The primary phytogeographical transition zone of the Thai-Malay Peninsula occurs, however, around the Kangar-Pattani Line at 6˚-7˚ north, ca 450 km south of the Isthmus of Kra, where the Indochinese seasonal dry forests give way to Malesian aseasonal evergreen tropical forest (Woodruff 2010). A secondary vegetation transition zone is found ca 500 km north of the Kangar-Pattani Line, at Wallace's line between Mergui and Samut Songkram, where evergreen rainforest gives way to mixed moist deciduous forest (Hughes et al. 2003;Woodruff 2010;Parnell 2013). The limits of the distributions of C. glauca comb. nov. and C. tenuifl ora comb. nov. intersect around this transition zone. However, it must also be noted that C. glauca comb. nov. is apparently not absolutely restricted to areas of rainforest and wetter, aseasonal climate: a single specimen is known from Myanmar, and one from Koi Samui, which are areas of higher seasonality (Richardson et al. 2012), and the Myanmar specimen is north of Wallace's line vegetation transition zone. The western side of the Peninsula is generally wetter than the east (Parnell 2013), which may partially explain the continuation of the range of C. glauca comb. nov. further north on the western side. In summary, the distributions of C. glauca comb. nov. and C. tenuifl ora comb. nov. indicate that they have distinctly divergent habitat and climate preferences, with C. glauca comb. nov. preferring a wetter, less seasonal climate, and C. tenuifl ora comb. nov. inhabiting drier areas of greater seasonality and lower rainfall. This distinction supports their recognition as separate species.
The limited molecular based phylogenetic analysis that has been carried out including C. glauca comb. nov. also provides some evidence to suggest that C. glauca comb. nov. and C. tenuifl ora comb. nov. should not be considered as conspecifi c: the study of Hao et al. (2003) included one accession each of C. glauca comb. nov. (as Bauhinia glauca 1) and C. tenuifl ora comb. nov. (as Bauhinia glauca 2) which were not resolved as sister lineages. Instead C. glauca comb. nov. was placed as sister to C. didyma comb. nov. albeit with low support (bootstrap = 63%).