Taxon sampling included those taxa in the DNA based phylogenies of Doyle et al. (1997), Hu et al. (2002), Hu and Chang (2003), Jansen et al. (2008) and Li et al. (2014 – but not including several species they recognised in the Callerya cinerea complex). Three chloroplast regions were included in the study. Two protein coding genes: matK and rbcL, and the intergenic spacer ndhJ-trnF. One nuclear gene region was also included in the study ITS1, 5.8S and ITS2. Fresh DNA was extracted from the previously unsampled Sarcodum scandens (Tables 1, 2). For the ITS dataset, 12 additional sequences representing the Callerya group and 26 outgroup sequences were included from GenBank; the matK dataset comprised an additional five Callerya group sequences and 17 outgroup sequences from GenBank and the rbcL dataset, a further two ingroup and 12 outgroup sequences from GenBank (Table 2). Millettia japonica has also been confused with Wisteria floribunda in DNA sampling (see GenBank KT119544) and as a result of this, we have chosen to include three different verified samples in this study in order to ascertain its placement within the Callerya group.

Outgroup taxa (Table 2) for each analysis comprised several accessions that represented taxa from other Tribes within the IRLC e.g. Hedysareae, Galegeae and Trifolieae (Lewis et al. 2005) and several from outside the IRLC, e.g. Robineae, Loteae, Sesbaneae, Millettieae, Abreae, Phaseoleaeand Indigofereae. The genus Schfflerodendron Harms was selected as the outgroup upon which to root all trees owing to its position in the LPWG (2017) RAxML Maximum Likelihood analysis. Its position as sister to Callerya atropurpurea – and these two to Glycyrrhiza – in turn all sister to the Callerya group and the rest of the IRLC suggest that it is the most appropriate candidate for choice as outgroup. Additional outgroup sequences were generated of W095 (see codes to samples, Table 2) Schefflerodendron usambarense (Tribe Millettieae), W113 Lotus uliginosus (Tribe Loteae) and W115 Austrosteenisia glabristyla (Tribe Millettieae) with the addition of 14 other legume genera from GenBank representing additional Tribes all of which sit outside the IRLC (Table 2).

We generated 49 sequences of the nrDNA ITS spacer region, including one for the outgroup taxon Schefflerodendron usambarense. Sequence data was also generated for three plastid markers: 51 Callerya group sequences from the ndhJ-trnF cpDNA intergenic spacer, 53 sequences from the matK gene and 57 sequences from the rbcL gene. Sequences of three outgroup taxa (i.e. Austrosteenisia glabristyla, Lotus uliginosus and Schefflerodendron usambarense) were also obtained for these three plastid markers (see Table 2). Summary statistics of support levels at critical nodes of all trees generated in this study (Suppl. material 1 Figs S1–S6), derived from Maximum Likelihood (ML) analysis and Bayesian inference (BI) analysis, are shown in Table 3.

The DNA extraction protocol for all 54 samples (with numbers from W002 to W115) and the seven samples labelled W1, W2, W3, W5, W6, W8 and W10 (Table 2) used a modified CTAB protocol (Doyle and Doyle 1987). DNA extraction from herbarium specimens followed the protocol used by Särkinen et al. (2012) with some minor amendments.

For all accessions labelled W002 to W115 (Table 2), a circa 800 bp part of matK was amplified with the previously unpublished primers designed by Ki-Joong Kim: 1RKIM-f – ACCCAGTCCATCTGGAAATCTTGGTTC and 3FKIM-r – CGTACAGTACTTTTGTGTTTACGAG (Dunning and Savolainen 2010). PCR reactions were performed in 25μl volumes containing final concentrations of 1× Bioline Biomix Red, 0.35μM of each primer, 0.2mg/ml BSA (bovine serum albumin), 4% v/v DMSO (dimethyl sulfoxide) and 20ng DNA template. Cycling conditions were 94 °C for 120s, then 35 cycles of 94 °C for 30s, 48 °C for 30s, 72 °C for 60s, and finally 72 °C for 7 minutes.

For all accessions (Table 2) the gene rbcL was amplified with primers 1F (Fay et al. 1997) and 1460R (Fay et al. 1998). To amplify degraded and/or low quality DNA two overlapping shorter fragments were amplified with the original primers and internal primers 636F and 724R (Fay et al. 1997). PCR reactions for all primer combinations were performed in 25μl volumes containing final concentrations of 1× Bioline Biomix Red, 0.35μM of each primer, 0.2mg/ml BSA (bovine serum albumin), and 20ng DNA template. Cycling conditions for the reactions using primers 1F and 1460R were 94 °C for 120s, then 30 cycles of 94 °C for 60s, 51 °C for 30 s, 72 °C for 120s, and finally 72 °C for 7 minutes. For the shorter fragments this protocol was modified by decreasing the elongation time to 90s and increasing the number of cycles to 40 for weaker amplicons.

Again for all accessions, the intergenic spacer ndhJ-trnF was amplified with the primers ndhJ and TabE using the PCR protocol listed in Shaw et al. (2007). Low quality or degraded DNA necessitated the utilisation of primers that amplified two shorter, overlapping segments of this region. We designed two additional primers internal to the ndhJ-trnF spacer in order to overcome this problem: 456F – ATGGGCCGGATTCTATTTGT and 725R – TGATTAGTGGTCTAGATCATCAT. The PCR protocol for the shorter fragments was the same as above, apart from increasing the number of cycles to 40 for weaker amplicons.

For all accessions the nrDNA Internal Transcribed Spacers (ITS1 and ITS2) were amplified with primers ITS4 and ITS5 (White et al. 1990; Baldwin et al. 1995) or with 17SE and 26SE (Sun et al. 1994). The PCR reactions for ITS4 and ITS5 were performed in 25μl volumes containing final concentrations of 1× Bioline Biomix Red, 0.5μM of each primer, 0.2mg/ml BSA (bovine serum albumin), and 10 - 25 ng DNA template. Cycling conditions were 94 °C for 120s, then 30 cycles of 94 °C for 60s, 48 °C for 60 s, 72 °C for 90s, and finally 72 °C for 7 minutes. The PCR reactions for 17SE and 26SE were performed in 25μl volumes containing final concentrations of 1× Bioline Biomix Red, 0.35μM of each primer, 0.2mg/ml BSA (bovine serum albumin), and 20ng DNA template. Cycling conditions were 94 °C for 120s, then 40 cycles of 94 °C for 30s, 63 °C for 60 s, 72 °C for 60s, and finally 72 °C for 7 minutes.

Sequencing of 44 taxa for ITS and 54 taxa for ndhJ-trnF, matK and rbcL were performed at GATC Biotech (www.gatc-biotech.com; Konstanz, Germany).

For the seven accessions labelled W1 to W10 (see Table 2) DNA extractions used a similar protocol to that mentioned above. PCR amplifications were performed using the same primers as already mentioned for nrDNA ITS, cpDNA matK and cpDNA rbcL with the following different protocol: PCRs were performed in 25 μL volumes, containing 12.5 μL DreamTaq PCR Master Mix (2×) (4 mM MgCl2; Thermo Fisher Scientific, Waltham, MA, USA), 0.5 μL of each primer (100 ng μL−1), and 1 μL DNA template. TBT-PAR [trehalose, bovine serum albumin (BSA) and polysorbate-20 (Tween-20)] was added to reduce the inhibitory effects of polysaccharide and phenolic compounds (Samarakoon et al. 2013).

For the accessions W1 to W10 (Table 2) all amplifications were performed on a 9700 GeneAmp thermocycler (ABI, Warrington, UK). All PCR products were purified with either the QIAquick PCR kit (Qiagen, Hilden, Germany) or the Nucleospin Extract II kit (Machery-Nagel, Düren, Germany), following the manufacturers’ protocols. Cycle sequencing reactions were performed in 5 μL reactions using 0.5 μL BigDye Terminator cycle sequencing chemistry (v3.1; ABI) and the same primers as for PCR. Complementary strands were sequenced on an ABI3730 automated sequencer.

Sequences of each region were edited and compiled in Geneious (version 8.1.9; Kearse et al. 2012) and aligned with the MUSCLE algorithm (Edgar 2004) implemented in AliView (Larsson 2014). The ends of the alignments were trimmed to the point where all sequences were present and base calls were unambiguous.

Phylogenetic analyses were conducted on the plastid, ITS and combined plastid/ITS matrices using two approaches, Maximum likelihood (ML) and Bayesian inference (BI). For the ML approach, we used the software RAxML (v. 8.2.8; Stamatakis 2014) as implemented on the CIPRES portal (www.phylo.org) with 1,000 rapid bootstrap replicates followed by the search of the best ML tree; the GTRCAT model was used and all the other parameters were set as default settings. The Bayesian Markov Chain Monte Carlo (MCMC) approach was performed using the software MrBayes (version 3.2.6; Ronquist and Huelsenbeck 2003) as implemented on the CIPRES portal. The best-fit DNA substitution model was tested using JModelTest 2 (version 2.1.6; Darriba et al. 2012) as implemented on the CIPRES portal. The General Time Reversible (GTR) model with a proportion of invariable sites and a gamma shape to account for rate heterogeneity among sites (GTR+I+G) was selected for both partitions. The analyses were run twice each for 10 million generations and sampled every 1000^th generation. The MCMC sampling was verified using Tracer (Rambaut and Drummond 2009) and was considered adequate when the effective sampling size was higher than 200. A burn-in period of one million generations was applied to each run. The remaining trees from both runs were compiled using the ‘‘allcompat’’ option in MrBayes to produce a maximum credibility tree with Bayesian posterior probabilities (BPP) for each node. In both combined ML and BI analyses, the plastid and ITS partitions were allowed to have partition-specific model parameters. Schefflerodendron usambarense was designated as outgroup taxon in all analyses. Support values for nodes of critical taxa in the Discussion are shown in Table 3.

The morphological key to the species was based on examination of living material in cultivation in UK and USA and in the wild in China, Japan, Laos, Myanmar, Thailand and Vietnam. Herbarium specimens were examined including the collection of all relevant genera in the Callerya group at K and BM. Online collections were examined at the Chinese Virtual Herbarium, CVH (http://www.cvh.ac.cn/en); JSTOR Global Plants (https://plants.jstor.org/); Herbarium, Muséum National d’Histoire Naturelle, Paris, MNHN (https://science.mnhn.fr/institution/mnhn/collection/p/item/search/form?lang=en_US); Herbarium Royal Botanic Garden, Edinburgh, RBGE (http://data.rbge.org.uk/search/herbarium/) and Nederlandse Natuurhistorische Collecties, Naturalis (http://bioportal.naturalis.nl/). See Appendix 1 for a full list of all specimens used as the basis for the new generic descriptions. Herbarium acronyms follow Thiers (2019, http://sweetgum.nybg.org/science/ih/). A full list of all Herbaria cited is found in the acknowledgements. A list of the critical characters measured for this study is shown in Tables 4, 5.

Schot (1994) segregated her species of Callerya into two groups based on the presence or absence of stipels and, when present, whether they were persistent or caducous. We have found no evidence that stipel presence or absence has any taxonomic significance in Calleya s.l. Lôc (1996) and Wei and Pedley (2010) segregated species on the basis of the presence or absence of an indumentum on the dorsal surface of standard petals. We concur with Lôc (1996) and Wei and Pedley (2010) but in addition regard the inflorescence type and various floral, fruit and seed types to be equally significant in delimiting taxa (see the Key to the Genera and Table 4 for a list of significant characters). The key to fruiting specimens of Callerya s.l. (Lôc 1996: 56) emphasised stipellae characters as distinctive, an observation for which we find no support.

Using our revised generic concepts and species assigned to them (Table 1) and comparison of morphological characters (Table 4), the synapomorphies diagnosing the Callerya group are: the lianescent habit (except for the tree species Adinobotrys atropurpureus and A. vastus); flowers inserted singly on the axis in either axillary or terminal racemes and/or panicles, and bracts either fully or largely enclosing the flower buds at the inflorescence apex, which are usually longer and often wider than the buds. There are, however, some exceptions. Floral bracts are caducous at anthesis in most of the Callerya group except in Adinobotrys atropurpureus, Nanhaia speciosa, Serawaia strobilifera and Wisteriopsis where they are persistent.

Gibbosities, which are small protuberances that develop beneath the leaf pulvinus above the stipule where it is attached to the stem, are absent in most of the Callerya group but are present in both Wisteriopsis, Nanhaia and Serawaia (Table 4; Fig. 2R–S). Bracteoles may be found either at the base of the calyx or along the pedicel. They are present in most genera but absent in Endosamara, Sigmoidala, Kanburia and Afgekia. They are also absent in Wisteria frutescens (Table 4).

Genera in the Callerya group often differ from each other (Table 4, Fig. 2) according to the presence of callosities at the base of the standard petals. Callosities occur in five distinct types:

a) Boss callosities form two slightly raised domes or swellings on either side of the midline of the standard lamina, at the point of its upward flexion above the claw (Fig. 2A). The standard in the latter case often appears to be smooth but the bosses hold the two wing petals close to the standard prior to anthesis. Boss callosities are found in Adinobotrys, Endosamara, Sigmoidala, Sarcodum, Nanhaia, Serawaia and Wisteriopsis and occasionally in Callerya and Whitfordiodendron;

b) Arched callosities are paired half-moon or crescent shaped arches forming ridges of hardened tissue that curve up from the base towards the midline over the staminal sheath (Fig. 2E). These are found only in Austrocallerya;

c) Ridge callosities form a straight ridge or rim of hardened tissue on either side of the midline of the standard near the base (Fig. 2B) and are seen in most Callerya s.str. species, Whitfordiodendron, Kanburia, Padbruggea dasyphylla and in the North American Wisteria frutescens;

d) Papillate callosities are those where a pair of papillate projections protrude from the area of hardened tissue on the surface usually at the point of upward flexion of the standard lamina above the claw (Fig. 2H). These are present in Afgekia, Padbruggea filipes (see Dunn 1911: 195 as Adinobotrys filipes) and all the Asian species of Wisteria (see Valder 1995: 26 as “auricles”);

e) Corniculate callosities are present in the two species of Afgekia (Fig. 2I). Uniquely in the Callerya group, these two species have, in addition to a basal papillate pair, a second pair of corniculate or horned callosities which project out from the lamina above the basal pair.

There are notable differences in the fruits and seeds among the genera. In Endosamara and Sarcodum the exocarp separates from the endocarp and some degree of separation also occurs in Wisteriopsis. In Endosamara the pods are clearly septate with transverse walls between each seed, forming loments (see Endosamara racemosa in Geesink 1984: 63 Pl. 1, 5; Lôc and Vidal 2001: 17, Pl. 3). In Sarcodum the sausage- shaped or botuliform pods which initially have a fleshy exocarp, are also fully septate but do not form loments (see Sarcodum scandens in Geesink 1984: 63 Pl. 1, 7; Lôc and Vidal 2001: 9, Pl. 1). The North American Wisteria frutescens has nonseptate pods. In all other genera in the Callerya group the endocarp is subseptate with seeds making indentations in the surrounding pith while areas between the seeds appear as irregular, often indistinct transverse septa. In Afgekia the funicle as well as the hilum (see Afgekia sericea funicle in Geesink 1984: 64 Pl. 2, 10; Fig. 2J–K) are both significantly longer than those of other taxa in the group with the exception of Padbruggea (Fig. 2D) and Austrocallerya (Fig. 2G).

Callerya Endl., Gen. Pl. Suppl. 3: 104 (1843)

Endlicher (1843) first described the genus Callerya based on Marquartia tomentosa Vogel, which had been described that same year from southern China (Vogel 1843: 37). Endlicher used the new name Callerya to replace Vogel’s Marquartia, the replacement name being necessary because the generic name Marquartia Hassk. had already been provided a year earlier for a species of Pandanus by Justus Karl Hasskarl, assistant curator of the Buitenzorg Botanic Garden [Bogor] on Java (Hasskarl 1842: 14). Hasskarl’s Marquartia is moreover, now considered as synonymous with Pandanus L.f. (1782).

In addition, the description and illustration of the species on which Vogel based his new but superfluous generic name Marquartia [i.e. M. tomentosa] (Vogel 1843) had already been described by Bentham as Millettia nitida Benth. (Bentham 1842: 484). Vogel’s illustration clearly shows the leaves with persistent stipules, each with five leaflets, long persistent floral bracts and densely sericeous ovary, all characters diagnostic of that species. This meant that not only the generic name Marquartia but also Callerya tomentosa (Vogel) Endl. had to be replaced according to Art. 11.4 of the ICN (Turland et al. 2018). The replacement Callerya nitida (Benth.) Geesink was published formally by Geesink within his revision of Tribe Millettieae (Geesink 1984: 82). Callerya nitida is thus the type species of the genus Callerya. Endlicher’s description of Callerya mentioned the compressed, woody and leathery pod which he stated contained either a few seeds or a single seed which was ovoid-circular and flattened-compressed. Callerya nitida (Fig. 2L–N) does indeed possess flattened pods with up to ten compressed seeds (see Schot 1994: 28, Fig. 3). The species may also become a scandent shrub but is not arborescent as Endlicher implied.

Geesink (1984) recognised that species of Callerya had true paniculate inflorescences. He also sank three genera; Adinobotrys Dunn, Padbruggea Miq. and Whitfordiodendron Elmer into his redefined Callerya and transferred two Sections created by Dunn (1912a) from Millettia – Sect. Eurybotryae Dunn and Sect. Austromillettia Dunn – into Callerya and created a new monotypic genus Endosamara Geesink from Sect. Bracteatae Dunn based on Robinia racemosa Roxb. (1832).

Callerya was revised by Schot (1994) and was treated as belonging in Tribe Millettieae (see Table 1). Schot recognised 19 species from China, south-east Asia and Australasia and since then many more species have been described, bringing the total number to 33 (Sirichamorn et al. 2016). All species sensu Schot (1994) except C. atropurpurea and C. vasta, are vigorous climbing or scandent woody shrubs. Inflorescences are paniculate with either axillary racemes or secondary panicles and frequently possess rather thick to woody inflorescence axes with prominent bud scars. Floral bracts are generally short and can be narrow or broad according to species and are in most species caducous, rarely persistent. Flowers may be white, green, red, brownish-yellow, lilac, pink or deep purple. In the type species Callerya nitida, the wings are distinctly shorter than the keel. Fruits, which can be either flattened or inflated are velutinous and ribbed, occasionally wrinkled or smooth and the seed chambers are subseptate. Seeds 2–9, large, ovoid to ellipsoid (Tables 1, 4).

Schot (1994: 2) chronicled the transfer of species from other genera into Callerya and included eleven species from Millettia, the first two of which were Pterocarpus australis Endl. Prod. Fl. Norfolk (Endlicher 1833: 49) and Pongamia atropurpurea Wall. (Wallich 1830: 70).

Although the genus Pterocarpus is placed in Tribe Dalbergieae (Klitgaard and Lavin 2005), the species P. australis has been shown to belong in the Callerya group (Li et al. 2014). Pongamia Adans. is now treated as being synonymous with Millettia, with the type P. pinnata (L.) Pierre being transferred to Millettia pinnata (L.) Panigrahi by Panigrahi and Murti (1989). Pongamia atropurpurea which was transferred into Callerya by Schot (1994: 15) has also been found to belong in the Callerya group (Li et al. 2014). Schot (1994) further transferred the Australian species originally described as Wisteria megasperma F.Muell. (1858) into Callerya. This too has been confirmed to belong within the Callerya group (Li et al. 2014).

In their analyses using combined data from chloroplast trnK and matK sequences, Hu et al. (2000) showed that Callerya reticulata was sister to a clade supported by BS 100% with Wisteria frutescens and W. sinensis. In a later paper using sequence data from nuclear DNA ITS spacers and a larger sampling of Callerya and Wisteria as well as Afgekia filipes, Hu et al. (2002) found that Callerya was polyphyletic occurring in four different clades. Wisteria frutescens was strongly supported sister to W. brachybotrys, W. sinensis and W. floribunda with BS 100%. The Wisteria clade was sister to a clade with Callerya megasperma, C. australis and Afgekia filipes with strong BS (85%) support. In a later analysis Hu and Chang (2003) using the more conserved rbcL chloroplast gene, found that Callerya vasta was early branching to a clade of Wisteria sinensis sister to Afgekia sericea while Endosamara racemosa was sister to Millettia japonica.

Schot included nine synonyms within her concept of C. cinerea, a species that she recognised to have a wide distribution from Nepal in the west to the Chinese coast in the east (Schot 1994: 17). In our analyses we have utilised two specimens of Callerya cinerea, one from Thailand, the other from China, (Table 2) but these sheets lack diagnostically significant fruiting material and may not equate fully with the holotype material seen at Kew from Bangladesh, Sylhet, Chittagong, (Wallich 5888; K000881022). Within the C. cinerea complex, leaflet number, pod thickness and seed shape and number appear to be important characters. Wei and Pedley (2010) split C. cinerea into groups of species based on leaflet number: 3–5 in C. tsui, C. dorwardii and C. sphaerosperma and 5–7 in the other seven species. Wei and Pedley (2010) also segregated the species on the degree of pod inflation: flattened with lenticular seeds in C. congestiflora, C. dielsiana and C. longipedunculata; inflated with globose seeds in C. cinerea, C. dorwardii, C. gentiliana, C. oosperma and C. sericosema. We have only been able to sample material of two of the resurrected species recognised by Wei and Pedley (2010) that were included in C. cinerea by Schot (1994), namely C. dielsiana and C. oosperma. A further investigation of this group of Chinese species is needed to fully assess species delimitations. If all these species belong together with C. nitida and C. cochinchinensis as indicated both by Li et al. (2014) and from our preliminary results here, it seems that Callerya s.str. might comprise as many as twelve species.

Padbruggea Miq. Fl. Ned. Ind. 1(1): 150 (1855)

Miquel (1855: 150) described the genus Padbruggea including the statement:

“legumen oblongum, stipitatum crassum? exalatum” [legume oblong, possibly on a thickened stipe and not winged] and he particularly noted the presence of callosities on the standard petal “vexillum infra medium quidem texturae crassioris ac perinde subfoveolatum” [standard with a thickened area just below the middle with a somewhat pitted texture]

Reference to the unwinged nature of the fruit was in comparison to some species of Pterocarpus Jacq. (Tribe Dalbergieae) whose fruits have a distinct wing-like exocarp. Miquel also noted that he had not seen mature fruits.

Our examination of the species described as P. dasyphylla Miq. (1855), revealed that the pods were readily distinguished from others in the Callerya group by their inflated but broadly flattened-cuboid shape with distinct longitudinal ridges and furrows and by the 1 or 2 compressed obovoid seeds possessing long strap-shaped hila 18–36 × 4–7 mm (Table 5).

In his protologue of Adinobotrys filipes, Dunn (1911: 196) noted that in its appearance the species had more slender pedicels than those in what he considered to be the closely related A. erianthus (here in Whitfordiodendron) and that it approached Padbruggea in its auriculate standard (Dunn 1911). Dunn’s latter reference may also refer to the papillate callosities present on the standard of A. filipes but which are absent on the smooth standard of A. erianthus. In her monograph on Callerya, Schot (1994: 3) commented:

Craib (1928) argued that Padbruggea and Whitfordiodendron were congeneric. He based his arguments on the intermediate position of Adinobotrys filipes Dunn (now Afgekia filipes Geesink). This species resembles in habit mostly Padbruggea dasyphylla, but has the generic characters of Adinobotrys.

Craib recombined Adinobotrys filipes in Padbruggea, along with a good measure of uncertainty as to whether he believed the species really belonged in that genus or in Adinobotrys (Craib 1928: 397). He also postulated that Elmer’s Whitfordiodendron may belong in Padbruggea thereby highlighting the morphological difficulties with respect to these taxa faced by later workers such as Geesink (1984).

Schot in her synonymy of Callerya dasyphylla also included Milletia oocarpa Prain, distinguished from P. dasyphylla by its ovoid as opposed to compressed obovoid fruits and M. maingayi Baker which differs in its more numerous, smaller and more densely tomentose leaflets (Schot 1994: 20). We have recognised this as Padbruggea maingayi in this paper (see below).

The status of Afgekia filipes has long been debated as it has true panicles as opposed to racemes and shorter calyx teeth than those of the other two species of Afgekia (see Table 5). It also has a single pair of papillate callosities on the standard as opposed to two pairs found in both A. sericea and A. mahidoliae. Afgekia filipes has entirely glabrous anthers as opposed to anthers with a basal tuft of hairs and it has much larger fruits and seeds (Table 5). It was originally described as Adinobotrys filipes Dunn on the basis of its large single seeded pods (Dunn 1911: 195).

Geesink (1984: 76) transferred Adinobotrys filipes into Afgekia adding:

the general habit, the shape of the calyx, and the glabrous anthers are indeed similar to certain species of Padbruggea. It differs in the absence of bracteoles and the long pedicels. The pods were unknown until 1975, but then it appeared that the seeds showed an elongated fleshy funicle with a corresponding elongated hilum.

Geesink (1984: 76) concluded that the morphology of A. filipes indicated that it was a less derived species than A. sericea and A. mahidoliae and alluded to its affinities with Padbruggea within which it had been placed by Craib (1928) along with three other species in Adinobotrys.

In his transferral of the species into Afgekia, Geesink noted the apparent absence of bracteoles, the length of the pedicels and the elongated hilum on the seeds (Geesink 1984: 77). Both Lôc and Vidal (2001) and Wei and Pedley (2010) followed Geesink, maintaining the species in Afgekia. Sirichamorn (2006) examined 37 living specimens of A. sericea, 50 specimens of A. mahidoliae and 32 specimens of A. filipes from wild material in Thailand and found that A. filipes posseses bracteoles, that pedicel length among the species overlaps and that the overall size of the seeds are three times that of the other two species of Afgekia, i.e. c. 80 mm vs. 15–25 mm long (Tables 4, 5).

Prathepha and Baimai (2003) mentioned the existence of Afgekia filipes in their RAPD and nucleotide sequence analyses of A. mahidoliae and A. sericea, although they did not state their reason for excluding the species. Sirichamorn (2006), based on morphometric and molecular data, clearly showed that Adinobotrys filipes does not belong with Afgekia (Table 5).

We have examined material of both Afgekia filipes and Callerya dasyphylla and agree that there are indeed similarities between the two species. We have confirmed Sirichamorn’s (2006) discovery that Afgekia filipes does have short, linear bracteoles that are attached at the base of the calyx (Table 5). Both species possess inflated fruits with a velvety indumentum and oblique longitudinal ridges and furrows but those of C. dasyphylla are broader and flatter, with the dorsal midline flanked by two large folds or flanges that meet at the apex. Our results show that Afgekia filipes, originally described by Dunn (1911) in Adinobotrys, belongs in the genus Padbruggea and it is reinstated in that genus here (see Taxonomic treatment below) following Craib (1928). The diagnostic characters of Padbruggea filipes are shown in Table 5.

Whitfordiodendron Elmer. Leafl. Philipp. Bot. 2: 689, 743 (1910)

Elmer (1910) in his protologue of the illegitimate but valid name Whitfordia scandens stated that he had only seen young not mature fruits but that they were “thick, hard, canescently velvety and 1-seeded”. He also noted the puberulent dorsal surface of the standard petal of the deep red flowers (Elmer 1910: 691). The generic name Whitfordia was already utilised for the fungal genus Whitfordia Murrill (Murrill 1908: 407), a synonym of Amauroderma Murrill (Murrill 1905: 366) and although Whitfordia Elmer (with W. scandens) was described entirely in English, together with that of the transfer to Whitfordiodendron as an erratum in an appendix to the same volume, the name is nevertheless still valid. Under the International Rules of Nomenclature adopted in Vienna in 1905 and Brussels in 1910, a Latin diagnosis was a requirement for valid publication of a name of a new taxon on or after 1 January 1908 (Art. 36). As a result of discrepancies and disagreements between the American Code of Botanical Nomenclature [see Bull. Torrey Bot. Club 34(4): 167–178 (1907)] and the Vienna Rules and Brussels Rules, a rapprochement was made in the Cambridge Rules (1935) changing the implementation date to 1935. As a result, Elmer’s Whitfordiodendron based on W. scandens Elmer is validly published (J. McNeill pers. comm.).

Despite the nomenclatural wrangles on the validity and usage of Adinobotrys versus Whitfordiodendron (discussed by Merrill [1934: 159] who independently concluded that Art. 38 of the Cambridge Rules validated Elmer’s Whitfordiodendron), Merrill proceeded to describe the new species Whitfordiodendron sumatranum Merr., which he stated was close to W. myrianthus (i.e. to W. nieuwenhuisii (J.J.Sm.) Dunn). Merrill (1934: 160) also made what he believed to have been four new combinations in Whitfordiodendron but he was evidently unaware that W. atropurpureum, W. erianthum, W. myrianthum and W. nieuwenhuisii had already been combined in Whitfordiodendron by Dunn (1912b: 364).

Our morphological examination has revealed that two species previously included within Callerya share a suite of characters with Whitfordiodendron scandens. The most notable characters are: a) the flowers borne on extremely short pedicels 0.5–3 mm long vs. (2–)3–8 mm long in Callerya s.str.; b) the inflated, ovoid, rugose to ridged or ruminate pods with 1–3 seeds (if more than one-seeded then these often becoming fused together, Fig. 2P–Q) vs. pods flattened or if inflated then convex around seeds and contracted between them, the seeds being separate in the pod in Callerya s.str.; and c) most significantly, the sericeous keel petals which are particularly densely hairy along their lower margins (keel glabrous in Callerya s.str., see Table 4).

Based on nrDNA ITS sequence data, Li et al. (2014) showed that Callerya eriantha, C. scandens and C. nieuwenhuisii formed a clade with BS (100%) which is sister to C. eurybotrya and C. reticulata.

Adinobotrys Dunn, Bull. Misc. Inform. Kew 1911(4): 194 (1911)

The genus Adinobotrys was described by Dunn (1911: 194) in comparison to Millettia and Padbruggea with the statement:

“affinis Millettieae Wight et Arn. sed ovario stipitato, legumine monospermo indehiscente differt [related to Millettiae Wight & Arn. but differs by having a stipitate ovary and indehiscent one-seeded pod]”.

Dunn made a further distinction between Padbruggea (which he understood to comprise P. dasyphylla and P. maingayi) and Adinobotrys, stating that the inflorescence in Padbruggea was lax and that Padbruggea lacked any appendages on the wings and keel petals (Dunn 1911: 197). Dunn (1911) included five species in Adinobotrys without assigning any one of them as the type species; A. erianthus (Benth.) Dunn, A. filipes Dunn (see above), A. nieuwenhuisii (J.J.Sm.) Dunn, A. myrianthus Dunn and A. atropurpureus (Wall.) Dunn. Geesink (1984: 83) typified Adinobotrys on the species A. atropurpureus, the only species of the five which is a tree and not a liana. The following year Dunn (1912b) added A. scandens (Elmer) Dunn in the belief that Elmer (1910) had not validly published the name under Whitfordiodendron. Dunn, recognising the uncertainty of the validity of Whitfordiodendron, validated Elmer’s names firstly in Whitfordiodendron (including W. scandens) and then into his new genus Adinobotrys (Dunn 1912b: 364, 365).

Schot (1994) included A. atropurpureus, A. erianthus and A. nieuwenhuisii in Callerya and recognized A. myrianthus as conspecific with C. nieuwenhuisii. Schot did not include A. filipes in Callerya as she treated the species as belonging in Afgekia (Schot 1994: 3).

Adinobotrys has several morphological characters that separate it from the other allied genera within the Callerya group; trees vs. lianas, stipules 2–4 mm long, floral bracts short, c.1–3 mm long, standard petal glabrous (although this is not unique to Adinobotrys), pods inflated with glabrous, rugose surfaces and large ovoid seeds with short elliptic or circular hila (see Geesink 1984: 64 Pl. 2, 14).

Results from sequence data of nuclear ITS and chloroplast matK showed that Callerya atropurpurea was placed sister to the rest of the Callerya group (Li et al. 2014).