Bergbambos and Oldeania, new genera of African bamboos (Poaceae, Bambusoideae)

Abstract Two new monotypic genera, Bergbambos and Oldeania are described for African temperate bamboo species in the tribe Arundinarieae, after a comparison of their morphological characteristics with those of similar species from Asia. Morphological differences are supported by their isolated geographical distributions. Molecular evidence does not support the inclusion of these species in related Asian genera, recognising them instead as distinct lineages. New combinations Bergbambos tessellata and Oldeania alpina are made.


Introduction
While Asian temperate bamboos have received critical attention over recent decades (Stapleton 1994, Wong 1995, Li et al. 2006, the generic placement of the temperate bamboos of Africa has not been properly addressed. Th ere seem to be only two temperate bamboo species on the African mainland, currently enumerated most frequently as Th amnocalamus tessellatus (Nees) Soderstrom & R.P. Ellis and Yushania alpina (K. Schum.) W.C. Lin. Th ese species are in tribe Arundinarieae Nees ex Asch. & Graebn., a group also known as the northern temperate clade, identifi ed as a strongly supported monophyletic group from the fi rst molecular analyses of bamboos onwards (Watanabe et al. 1994, Zhang 1996. Tribe Arundinarieae contains woody bamboos with semelauctant synfl orescences (lacking a capability for indeterminate growth from buds subtended by the basal spikelet bracts), ebracteate or partially bracteate synfl orescence paraclades (reduced sheathing subtending infl orescence branches) and 3 stamens in each fl oret. Th ey constitute ca. 800 of the ca. 1400 woody bamboos, and are found in Asia, Africa, and the USA, having a montane or subtropical to temperate distribution.
Molecular studies reviewed by Bamboo Phylogeny Group (2012) suggest that semelauctant infl orescences with 3 stamens and reduced branch sheathing have evolved from tropical bamboos at least twice, once to give the northern temperate clade Arundinarieae of Asia and Africa, spreading to N America, and on separate occasions in Central & South America within the Bambuseae Kunth ex Dumort., principally to give Chusqueinae Bews, with these characters also evolving on a smaller scale within the Arthrostylidiinae Bews and Guaduinae Soderstr. & R. P. Ellis as well.
Most of the older 3-stamened species were placed at some time in Arundinaria Michx., which has 529 combinations, but that genus is now widely recognised as containing only 3 species, all from the Southeast USA (Stapleton et al. 2004, Zeng et al. 2010, Bamboo Phylogeny Group 2012. Treatments of the other species of tribe Arundinarieae vary, according to the breadth of generic concept used, and which characters are given greatest weight. For example, the group of Asian species morphologically closest to Arundinaria could be placed (Zhang et al. 2012) either in a polyphyletic broad interpretation of Arundinaria (e.g. Li et al. 2006), in a polyphyletic broad interpretation of Bashania Keng f. & T.P. Yi (e.g. Keng and Yi 1996), or in the monophyletic Sarocalamus Stapleton (Stapleton et al. 2004, Bamboo Phylogeny Group 2012. Th e morphologically more distinct species are currently placed in other genera, 27 of which were recognised by Bamboo Phylogeny Group (2012), out of a total of 42 genera that have been described within the tribe.
Th ere appears to have been a rapid and relatively recent diversifi cation within bamboos with 3 stamens, including tribe Arundinarieae (Stapleton et al. 2009, Zhang et al. 2011, Kelchner and BPG 2013, especially those found in montane and temperate areas such as the Andes, the Himalayas, and Northeast Asia. Th ere have also been several reports of hybridisation, reviewed by Triplett et al. (2010) and Zhang et al. (2012). Hybridisation may well have been common in the bamboos, as mechanisms to avoid it have not been documented. Recent rapid diversifi cation and hybridisation, combined with long generation times, appear to have limited the ability of DNA analyses to resolve phylogenetic patterns and defi ne well supported groups for taxonomy, especially at the generic level (Stapleton et al. 2009, despite reasonable or sometimes very substantial morphological variation. In the absence of reliable molecular analyses, for the purpose of descriptive treatments of bamboo species (Li et al. 2006, Wong 1995, Dransfi eld 2000, Widjaja 1997) a more traditional morpho-geographic approach has been maintained in the classifi cation of Asian bamboos. It has only been possible to use molecular data for the elimination of blatantly polyphyletic groups, rather than the determination of monophyletic ones. Attempts to group the genera substantially (e.g. Clayton andRenvoize 1986, Chao andRenvoize 1989) have resulted in polyphyletic and paraphyletic groups, or clades with weak support that are inconsistent in diff erent analyses.
Rapid recent diversifi cation seems to have spawned a host of small groups, often distinguished by relatively minor characters. Combining them together into a few large genera has not been possible without establishing excessively variable genera that are diffi cult to defi ne and demonstrably polyphyletic. On the other hand recognising only half of the genera described would still lead to a generic concept that is unusually narrow in the grasses. Th e latter procedure has been followed (Bamboo Phylogeny Group 2012), largely because it has been found unavoidable if a functional binomial classifi cation system is to be maintained. Th is is necessary in order to allow pragmatic fi eld identifi cation, and subsequently improve sustainable utilisation and conservation of these species, many of which have a limited range of distribution and are threatened by changes in land use and climate. A substantial proportion of the woody bamboos are yet to be described, and the lack of a functional and stable nomenclatural system for fi eld identifi cation has been a major factor preventing their recognition.
Only two species of temperate bamboo have been described from the African mainland. Th amnocalamus tessellatus (Nees) Soderstrom & R.P. Ellis is from mountains in southern Africa, while Yushania alpina (K. Schum.) W.C. Lin is from mountains in several countries across tropical Africa. Y. alpina was described initially in Arundinaria, and T. tessellatus was soon transferred into that genus from Nastus Juss. Th ey were more recently moved into the morphologically closer Asian genera, Th amnocalamus Munro and Yushania Keng f., the geographically closest representatives of which are found in the Western Himalayas, Map 1 and Map 2.
Th ree further, less well known species, Th amnocalamus ibityensis (A. Camus) Ohrnb., Yushania madagascariensis (A. Camus) Ohrnb. and Yushania humbertii (A. Camus) Ohrnb. (including Yushania ambositrensis (A. Camus) Ohrnb.) were described from Madagascar. Th amnocalamus ibityensis has been considered conspecifi c with T. tessellatus (Chao & Renvoize, 1989), but it would appear to have substantially diff erent branch sheathing. Th e two Yushania species would appear to share characteristics with Y. alpina, but their culms, branching and culm sheaths are not known. Y. ambositrensis resolved in a clade with Y. alpina (Triplett, 2008), but it is not clear how closely related they really are to Y. alpina, or to each other, and which species names should be recognised. Further fi eld work on temperate species of Madagascar is required, as existing collections are incomplete, although any such species may have already become extinct.

Comparison of morphological characters
Systematics within the grass family has traditionally given greater weight to fl oral than to vegetative characters. Th is has often led to polyphyletic genera in the bamboos, the superfi ciality of their similarities and their separate origins only being revealed by in-depth morphological investigations and/or molecular studies. In order to allow deeper, more objective morphological comparisons and to allow inclusion of consist-ent and accurate vegetative as well as fl oral characters in descriptions, the morphology of woody bamboos has been reviewed in depth (Stapleton 1997, available online). Recent bamboo treatments (Judziewicz et al. 1999, Li et al. 2006, Triplett et al. 2006, BPG 2012 have employed these revised concepts and terminologies, and they are followed here.
Th e characters and character states considered important at the generic level for distinguishing the two African species from similar Asian genera are given in Table 1.

Thamnocalamus tessellatus
Previous generic placements of Th amnocalamus tessellatus were based upon an incomplete knowledge of its morphology. Nastus tessellatus Nees was described before its fl owers were known, and transferred into Arundinaria (Munro 1868) simply as it bore "very great resemblance" to that genus. Later discovery of its fl owers has shown that it indeed has 3 stamens, rather than the 6 of Nastus, but it has pachymorph rhizomes (see Stapleton 1997: fi g.1) rather than the leptomorph rhizomes of Arundinaria.
It was transferred into Th amnocalamus largely on the basis of leaf anatomical characters by Soderstrom and Ellis (1982), who found that A. tessellata shared 10 characters out of 11 with Th amnocalamus spathifl orus (Trin) Munro, while it only shared 7 characters with Fargesia nitida (Mitford) Keng f. However, A. tessellata also shared only 5 characters with Th amnocalamus aristatus E.G. Camus, while the possibly conspecifi c T. spathifl orus and T. aristatus themselves only shared 6 out of 11 characters. When morphological characters other than those of leaf anatomy, along with more recent molecular results are taken into account, it would appear that the anatomical characters used by Soderstrom and Ellis (1982) are more informative at the level of species or below rather than at generic level.
Th e synfl orescence of T. tessellatus has been well illustrated in Hooker's Icones Plantarum (Prain 1913: Tab 2930 http://www.botanicus.org/page/1349516), and by Soderstrom and Ellis (1982). When examined closely, it can be seen that the synfl orescence of T. tessellatus has similarities to those of both Th amnocalamus and Fargesia Franchet, see Table 1, as they are compressed, and are associated with several supporting sheaths. However, while Th amnocalamus has loose racemose panicles, T. tessellatus, like Fargesia, consistently bears short racemes. Th ese are structurally very similar to those of Fargesia, but diff erences arise in the arrangement of the fl orets. In Fargesia the racemes are held tightly within imbricating sheaths, which can extend well beyond the spikelets. Development within the sheaths forces them to emerge to one side and appear unilateral, with the pedicels tightly pressed against the rhachis. Th ose of T. tessellatus are more cylindrical, the spikelets not so constricted by the sheaths, and the pedicels are free to develop in a normal distichous fashion, Figure 1.
In addition, in T. tessellatus the pedicels are scabrous, the glumes of each spikelet are basally tight and contain no vestigial bud remnants, and the racemes are usually largely ebracteate. Th e usually single fertile fl orets also distinguish T. tessellatus from  other Th amnocalamus and Fargesia species, but this character should be treated with caution as it can be a specifi c as well as a generic character.
Th amnocalamus tessellatus also has vegetative characteristics that distinguish it, notably from Asian members of Th amnocalamus, (see Table 1). A close inspection of the branching reveals not the pattern seen in species such as Th amnocalamus crassinodus (T.P. Yi) Demoly, but instead the substantial reduction in sheathing seen in Fargesia, Yushania, and Borinda Stapleton, Figure 2.
Th e branches of Th amnocalamus tessellatus are subequal, arranged side by side through strong compression of the basal internodes of the central branch, accompanied by loss of some of the sheaths at the nodes, Fig. 2A, cf Th amnocalamus crassinodus, Fig. 2B. Th is allows lateral branch prophylls to be seen side by side without any intervening sheaths. Th ese patterns were contrasted by Stapleton (1991;1994: fi g. 1;1997: fi g. 2), and also illustrated for T. tessellatus by Soderstrom and Ellis (1982: fi g.1, fi g. 4).
In addition to the synfl orescence and branching, T. tessellatus also diff ers in minor details that are harder to quantify, including the more varied orientation of the foliage leaves, and the delicate appearance of its oral setae and their more varied orientation, Figure 3.
Th us in terms of vegetative macro-morphological characteristics important at the generic level, T. tessellatus is closer to Fargesia than to Th amnocalamus, but can be distinguished from both. In general appearance it resembles a coastal species of Pleioblastus Nakai from Japan, with rather loose clumps, erect culms with short branches bearing coarse, irregularly arranged foliage with persistent sheaths. Th is contrasts with the delicate foliage leaves, all oriented towards the light on pendulous branches seen in Himalayan species of Th amnocalamus and in Fargesia. Th is is likely to be associated with the more open ecological habitat in which T. tessellatus is found, rather than the darker forest understorey habitats of Asian Th amnocalamus and Fargesia species.

Yushania alpina
Th e synfl orescence of Y. alpina is practically indistinguishable from those of several Asian and American bamboos, including species of Arundinaria, Sarocalamus, and Yushania-an open panicle with nearly complete reduction in sheathing at points of branching so that it is essentially ebracteate. However, the sheaths are often reduced to small tough bracts, as well as the more delicate sheath remnants or tufts of hairs seen in Yushania alpina is more distinct vegetatively. Reaching heights of up to 20m in its natural habitat, the tall, very erect culms are potentially much larger than those of any Asian species of Yushania, which only reach a maximum height of about 7m. Culm nodes and branching also diff er substantially from those of Asian species of Yushania, Figure 4.
Branches vary in size more than those of Asian Yushania species. Th e central branch is strongly dominant, and the fi rst two lateral branches are also strong. Th e orientation of the branches is less erect than those of most species of Yushania, becoming nearly horizontal. Above the branches the internode is distinctly sulcate, much more prominently than is seen in Asian Yushania species, as a result of the development of strong branches. Moreover there is often a dense ring of short, partially developed aerial roots at nodes in the lower part of the culm, often extending into the mid-culm region as well. Th is character is only known in species of Chimonocalamus Hsueh & T.P. Yi, and the leptomorph-rhizomed Chimonobambusa Makino among the Asian temperate bamboos. Th e roots are not as sharp and thorn-like as those seen in Chimonocalamus and Chimonobambusa, but they can be very distinct and prominent. Nodes have a distinct infranode between the culm sheath attachment and the supranodal ridge, which is well developed, Figure 5.
In its natural habitat, the open stands have a widely spaced appearance closer to that of a species of Phyllostachys Siebold & Zucc., rather than the denser thickets of   Asian Yushania species, because the rhizomes have consistently long necks, giving solitary culms rather than the denser clusters of pluricaespitose culms seen in Asian species of Yushania.
Branch structure and sheathing is diffi cult to distinguish from that of Yushania or Fargesia. Although the prophyll is usually 2-keeled, there is replication side by side of lateral branch initials without intervening sheaths. In this way it diff ers fundamentally from Chimonocalamus, which has only 3 branches and full sheathing.

Discussion
Th e morphological diff erences between Th amnocalamus tessellatus, Yushania alpina and other representatives of these and similar Asian genera suggest that although the two African bamboos share several characters and presumably common ancestors with Asian bamboos, they are not as closely related to their Asian relatives as previously thought.
Th e morphological distinctions are supported by geographical isolation, (see Maps 1 and 2). Long-distance dispersal of temperate bamboos is highly unlikely because of a lack of any specialized seed dispersal mechanism or dormancy, brief viability of seed, exacting habitat requirements, and extremely infrequent fl owering (Stapleton et al. 2004).
Together the morphological distinctions and geographical isolation justify the recognition of two new genera, following the existing relatively narrow generic concepts applied in the northern temperate clade, tribe Arundinarieae.
Th e new genera are keyed out below along with their 6 morphologically closest relatives in the tribe including the two Asian genera with distinct nodal thorns, as well as the North American type genus of the tribe, Arundinaria, and its Asian analogue, Sarocalamus.

. Bergbambos
Suffi cient data is now available to test whether this classifi cation would gain support from molecular phylogenetic evidence. Th ese two African species were not clearly resolved with Asian representatives of any genera in any molecular studies. For example, in a comparison of ITS sequences (Guo et al. 2002), Th amnocalamus tessellatus did not resolve with the type species of Th amnocalamus, T. spathifl orus, and its position varied between topologies. In the nuclear ribosomal ITS analysis of Hodkinson et al. (2010), Yushania alpina did not group with other Yushania species or closely with any other taxon. Weak associations between Yushania alpina and Chimonocalamus species were found by Guo and Li (2004) and Triplett (2008), which is interesting as they share possession of aerial roots developed into thorn-like structures, although they diff er in other ways. However, neither Yushania alpina nor Th amnocalamus tessellatus resolved with putative relatives in these or similar genera of temperate bamboos in the most comprehensive studies undertaken so far, using sequences from 8 regions of cpDNA in 146 species and 26 genera (Zeng et al. 2010), and 108 bamboos from 25 genera using plastid DNA and nuclear GBSSI gene sequences (Zhang et al. 2012).
Th e molecular data would suggest that their inclusion in Asian genera would render those genera polyphyletic. Because their monotypic status is considered likely they could not be supported as monophyletic groups themselves in a classifi cation based solely on molecular phylogeny. However, Zeng et al. (2010) and Zhang et al. (2012) considered them both to represent distinct lineages, and it is not possible to place them in well supported meaningful monophyletic groups except the tribe Arundinarieae. Th erefore while the molecular data would not allow the diagnosis of monophyletic genera for the African bamboos following a strict Hennigian cladistic analysis, neither their current placement in Th amnocalamus and Yushania, nor placement in any other existing genus receives any support either. Continuing to include these bamboos in Asian genera causes serious problems when describing or distinguishing between those genera.
Although woody bamboos are considered to have evolved originally in Gondwanaland rather than Eastern Asia , these African representatives are nested within the northern temperate clade, the tribe Arundinarieae, with a largely Asian distribution. Th is is estimated to have diverged from other woody bamboos around 23 mya ), 29 mya (Bouchenak-Khelladi et al. 2010 or 37.5 mya (Christin et al. 2008), but to have radiated only ca. 9 mya (Bouchenak-Khelladi et al. 2010) 10 mya , or 19 mya (Christin et al. 2008). Peng et al. (2013) after sequencing 95% of the Phyllostachys edulis genome found evidence of whole genome duplication 7-12 mya, supporting the more recent dates.
Collision of tectonic plates has been suggested as a likely cause of this rapid radiation (Stapleton et al. 2009. African and Indian plates met the Eurasian plate around that time, allowing a biotic interchange and subsequent radiation and diversifi cation of Gondwanan elements into a wealth of new habitats. However, the temperate ancestors of these two African bamboo genera seem to have diverged around the same time that temperate bamboos arrived in Eastern Asia. Inclusion of endemic temperate bamboos from S India, Sri Lanka and Madagascar in a molecular phylogeny is required before any conclusions can be drawn as to where bamboos from the northern temperate clade fi rst evolved, but there seems no evidence for an African origin, and it seems more likely that temperate bamboos radiated from India to Asia, Africa, and N America.

Nomenclature
Bergbambos Stapleton, gen. nov. urn:lsid:ipni.org:names:77131102-1 http://species-id.net/wiki/Bergbambos Remarks. Diff ering from Arundinaria and Sarocalamus and similar to Th amnocalamus and Fargesia in its short-necked pachymorph rather than leptomorph rhizomes, and its compressed synfl orescences. Diff ering from Borinda and Th amnocalamus in its racemose rather than paniculate synfl orescence branching. Diff ering from Fargesia in the distichous rather than unilateral arrangement of spikelets in the racemes, the spikelets usually having only one fertile fl oret, and the scabrous pedicels. Diff ering from Th amnocalamus in the branch complement with reduced sheathing, and from Fargesia in the more varied orientation of the leaf blades.
Name Bergbambos from the Afrikaans name (Bergbamboes) in South Africa. Th is genus would appear to be monotypic, confi ned to the mountains of South Africa, Lesotho and Swaziland. Oldeania Stapleton, gen. nov. urn:lsid:ipni.org:names:77131103-1 http://species-id.net/wiki/Oldeania Remarks. Diff ering from Arundinaria and Sarocalamus and similar to Yushania in its long-necked pachymorph rather than leptomorph rhizomes, though similar to all in its open panicles. Diff ering from Yushania in its sulcate culm internodes, fewer, more horizontal branches, culm nodes with well developed supra-nodal ridge and often thorn-like aerial roots. Similar to Chimonocalamus in its panicles and thorn-like roots at culm nodes, but diff ering in its multiple branches with reduced sheathing and sulcate culm internodes. Rhizome pachymorph, long-necked, giving open stands and solitary culms. Culms to 15(-20) m tall, diam. to 6(-10) cm, erect to nodding, terete with shallow sulcus above branches, smooth, nodes with prominent supranodal ridge, in lower to mid culm a nodal ring of dense, short, hard, and thorn-like aerial roots often well developed. Mid-culm branch complement initially with 3-5 main branches, spreading, sheathing reduced. Culm sheaths deciduous, tough. Leaf sheaths several to very many, blades thick. Synfl orescence semelauctant, paniculate, branch sheathing reduced to hard bracts, soft sheath remnants or hairs. Spikelets pedicellate with several fertile fl orets, pedicel scabrous. Empty glumes 2, bud remnants present or absent, fertile glumes 4-8. Lemma and palea similar in length. Stamens 3, fi laments free. Stigmas 2. Lodicules 3.
Name Oldeania from the Maasai common name (Oldeani) in Tanzania. Currently only the type species can be reliably placed in the genus, which thus has a distribution across tropical Africa from Cameroon in the west to E Africa, where it occurs from Ethiopia south to Tanzania. Th ere is a possibility that species from Madagascar will be placed in this genus when they are better known, but they may be conspecifi c or even introduced. It provides important montane wildlife habitats and food, notably for the critically endangered Mountain Gorilla, Gorilla beringei beringei.
Th e holotype, G.A. Fischer 672, was destroyed by fi re during the 1939-1945 World War. No trace of the type collection or any duplicate has been found in surviving components of the Berlin collections, nor in other herbaria, including the Hamburg collections taken to Russia and recently repatriated (Poppendieck pers. comm.). Th e likelihood of substantial infraspecifi c variation, the possibility of further species, and the lack of other collections from the type locality together make it inadvisable to select a neotype or epitype until new collections have been made.