Radcliffea , a new genus of Euphorbiaceae sensu stricto from Madagascar

Introduction In a loan received from Paris of various Madagascan Euphorbiaceae was a specimen that had originally been provisionally assigned to Sterculiaceae, but had later been considered to be a probable new species of Givotia Griff. (Euphorbiaceae). The specimen was in staminate flower but, later, two collections in both staminate and pistillate flower of the same plant were sent to Kew from Wageningen (WAG). These were also provisionally identified as Givotia. However, the non-coherent petals and especially the staminal column bearing a pistillode at the apex showed that it was not a Givotia. Marginal foliar glands, present in all species of Givotia, were absent, and instead there were abaxial laminar glands irregularly distributed along the basalmost nerves (up to 5 mm from the margin of the leaf base). Apart from being positionally different, the glands in Givotia are also partially embedded in the leaf margin whereas in the new material they were slightly stalked so as to be easily broken off. The five stamens were half as long as the petals, whereas in Givotia they are about as long as the petals. The other two genera of tribe Ricinodendreae, Ricinodendron Mull. Arg. and Schinziophytum Hutch. ex Radcl.-Sm., are restricted to continental Africa. They have palmatisect or palmate leaves but are otherwise similar to Givotia. Molecular phylogenetic analyses (K. J. Wurdack, unpublished manuscript) place this new material in Crotonoideae sensu stricto in the highly supported (bootstrap percentage 93) subclade C2 (Wurdack et al. 2005) containing predominantly Old World genera, but a supported sister relationship could not be established even with four genes (plastid rbcL and trnL-F, nuclear ITS and PHYC; K. J. Wurdack, unpublished manuscript). Placement of the new material in the inaperturate crotonoids is also confirmed by pollen morphology (Fig. 1), showing a typical crotonoid exine pattern with apically rounded subunits (see Nowicke 1994). Sampling in the family-wide molecular phylogenetic analyses (Wurdack et al. 2005) of the eight tribes of Crotonoideae with inaperturate pollen (Radcliffe-Smith 2001) includes 45 of the 60 genera, all tribes and all subtribes except monogeneric Benoistiinae. Five further genera (Anomalocalyx Ducke, Borneodendron Airy Shaw, Cocconerion Baill., Dimorphocalyx Thwaites, and Pantadenia Gagnep.) are sampled in the 4-gene analyses (K. J. Wurdack, unpublished manuscript). Furthermore, Acidocroton Griseb. was shown by Berry et al. (2005) to belong in the predominantly New World clade that includes Croton L. Of the nine genera that remain unsampled, Madagascan Parapantadenia Capuron was considered congeneric with sampled Pantadenia by Webster (1994). It agrees with the new material in provenance and the unilocular, asymmetrical ovary, but differs in its simple indumentum, uniformly distributed abaxial leaf glands, leaves basally triplinerved to penninerved and the short, unbranched pistillate inflorescence. There are two unsampled genera of Ricinocarpeae: Alphandia Baill. has 20 35 stamens, no pistillode and a 3-locular ovary, and Myricanthe Airy Shaw has opposite leaves, lacks petals, a disc and a pistillode, and it has 60 80 stamens and a 3-locular ovary. Benoistia H. Perr. & Leandri, Reutealis Airy Shaw, and Tapoides Airy Shaw can be excluded because as members of tribe Aleuritideae their staminate sepals