A systematic study of the genus Pseudopeiitameris (Arundinoideae: Poaceae)

The genus Pseudopentameris Conert is examined morphologically and anatomically. A phenetic study of the morphologi­ ca lly variable species P. macrantha indicates that two taxa should be recognised. One of these. P. caespitosa N.P. Barker, is described as new. In addition, the study supports the inclusion of Pentameris obtusifolia in Pseudopentameris. The genus Pseudopentameris is re-delimited to accommodate the new taxa, and a key to species is provided. Details of the cytology, phylogeny and conservation status of taxa in the genus are also discussed.


INTRODUCTION
The genus Pseudopentameris Conert is endemic to the lowland fynbos of the southwestern Cape Province.South Africa.This genus was created by Conert (1971), who thought that two species of the genus Danthonia DC., D. macrantha Schrad.and D. hrachyphylla Stapf, merited recognition as a separate genus.This new genus was char acterised by unusually large spikelets which, while similar to both those of Pentameris P. Beauv.and Danthonia, dif fered in having many-nerved glumes.The nature of the f ruit of these two species also set them apart.
Studies of the leaf anatomy (discussed below) provided further evidence of their singularity (De Wet 1956. 1960: Ellis 1985a).Ellis (1985b) observed that the leaf anatomy of the taxon he called Pentameris sp.nov., bore a strong resemblance to the anatomy of the two known species of Pseudopentameris.and postulated a hybrid origin for this species.Additional studies on the fruit m orphology (Barker 1986(Barker . 1989(Barker . 1993) and leaf anatomy (presented below) also suggested that this unusual species has close affinities with the other species of Pseudopentameris.Be cause of the nomenclatural confusion surrounding this taxon (see I. Pseudopentameris obtusifolia below), past studies have referred to it by a number of names.Barker (1989) and Ellis (1985b) referred to it as Pentameris sp.nov.whereas Linder & Ellis (1990) refer to it as Pen tameris squarrvsa Stapf.Some herbaria (in particular PRE and STE) misapplied this name to the taxon now known as Pentameris oervphila N.P. Barker (1993), whereas BOL has used the synonym Pentameris squarrosa Stapf.This study encapsulates all these observations, and documents the anatomical and morphological variation within the genus, and presents the relevant nomenclatural modifications that are required as a result of these obser vations.

M ORPHOLOGY
Pseudopentameris is unusual in having two species (P.macrantha and P. obtusifolia) with aerial stems that pos sess the anatomy of a rhizome (H.P Linder unpublished).Until the anatomy and homology of these structures is better understood, the term 'stem' is used, as it serves to differentiate this structure from below-ground rhizomes and the annually produced flowering culm.
These stems are often branched, and the annual flow ering culms are produced from their apical buds.Sub sequent shoot growth takes place from what Linder & Ellis (1990) term caulescent innovation buds on the stems, which results in branched stems.This branching can lead to a shrub-like growth form in older plants.This type of growth form has also been described in some species of Pentameris (Barker 1993).In Pseudopentameris obtusifo lia these branched aerial stems are scandent.spreading out from the base of the plant, over and between the other vegetation.Linder & Ellis (1990) illustrate this species (named in their work as P. squarrosa) and consider iLs unusual growth form to be an adaptation enabling the plant to survive long interfire periods that are charac teristic of the fynbos biome.The two species which do not have these stems (P.hrachyphylla and P. caespitosa sp.nov.described below) produce the annual flowering culm from an underground rhizome that can become fi brous or woody in older plants.
In P. obtusifolia the stems are covered by the persistent remains of the leaf sheaths, the dead leaf blades being The inflorescence is a contracted panicle, which be comes somewhat lax during anthesis and comprises 8-80 spikelets.Spikelet size is variable between the species.The glumes are 3-nerved in P. obtusifolia (the lateral veins are visible only at the base of the glumes) and 5-9-nerved in the other species.The lemmas are variously pubescent, with a geniculate central awn.The lemma lobes are acu minate and variously adnate to the lateral bristle.The palea is bicarinate, bifid at the apex and longer than the lemma body.The lodicules are many-nerved (G.A. Verboom unpublished data), cuneate.glabrous or ciliolate.The latter condition is developed to the extreme in P. ob tusifolia, which has densely ciliolate lodicules, the cilia being as long as the lodicule bodies.
Each floret has three large anthers (up to 9 mm).The ovaries of all taxa possess what Barker (1990Barker ( . 1994) ) termed pseudostigmata.These structures have the appear ance of stigmatic hairs, and arise from the region between the points of origin of the style branches, although the style branches themselves appear to be absent (or alter natively have stigmata along their entire length, from the apex of the ovary upwards).As the pseudostigmata are easily deciduous, they are seldom observed on mature fruit.Their presence may.however, explain the references by De Wet (1954b) and Tomlinson (1985) to hairs which are decurrent on the inner side of the caryopsis and join over the top of the ovary.A similar description of these structures is given by Chippindall (1955).These features, although not labelled as pseudostigmata, appear in the fig ures published by Barker (1986Barker ( . 1990).The fruit is a cary opsis (Clayton & Renvoize 1986), up to 6 mm long, with a long, canaliculate hilum extending almost the full length of the fruit.The embryo is a quarter to a fifth the length of the fruit (Conert 1971;Barker 1986Barker , 1989Barker , 1990)).The surface of the caryopsis is reticulate (Barker 1986(Barker , 1989)).Barker (1990) noted substantial variation in the size of the floral parts in specimens identified as P. macrantha.This prompted an additional study in which measurements of six floral characters from 59 specimens were taken, including those exam ined in the earlier study.The phenetic package NTSYS-pc version 1.4 (Numerical Tax onomy and SYStematics package; Rohlf 1988) was used to perform a Principal Components Analysis (PCA).
The first three principal components (axes) account for 92.9% of the variation in the data.When the specimens are plotted out on the first two axes (accounting for 88.5% of the variation), two clusters are apparent (Figure 1).Each cluster contains specimens from a wide geographic area, and the geographic areas of the two clusters overlap extensively.Furthermore, specimens from each cluster have been observed to co-occur.Despite this sympatric distribution, a one-sided T-test showed that the differences in the means of each of the six characters from the two clusters were statistically significant at a 99.9% confi dence interval.Table 1 presents the sample means and standard deviations of these characters in each cluster.A character not included in the PCA, but overlaid onto the PCA plot, is that of growth form, as related to the presence of the branching stems.This feature was ob served in the field for some of the specimens included in the PCA, and the position in the PCA plot of these speci mens is shown in Figure 1 by means of open symbols.As can be seen from Figure 1, each cluster has a specific growth form.The cluster with the smaller floral parts in cludes specimens with branching stems, whereas the specimens with larger floral parts do not possess stems.These two clusters are considered here to be two species: P. macrantha and P. caespitosa N.P. Barker sp.nov.De scriptions and details on nomenclature of these two spe cies appear below.
of large, inflated abaxial epidermal cells, dumbbell-shaped silica bodies and the absence of abaxial micro-and macro hairs.This knowledge was supplemented in this study by a Scanning Electron Microscope (SEM) survey of the leaf epidermides of all the species in the genus.
The SEM study of both adaxial and abaxial leaf anat omy was carried out on leaf blade material of nine her barium specimens of P. macrantha, nine specimens of P caespitosa, five specimens of P brachyphylla and 11 specimens of P. obtusifolia.Leaf material was mounted on SEM stubs and then coated in gold-palladium.The specimens were examined using an ISI-SX-25 Scanning Electron Microscope.

LEAF ANATOMY
Pseudopentameris has been well examined anatomi cally by De Wet (1956, 1960), Renvoize (1981), Ellis (1985a), Tomlinson (1985) and Barker (1990).The genus as a whole possesses a sclerophyllous type of arundinoid leaf anatomy (Ellis & Linder 1992).Ellis (1985a) found minor anatomical differences between P. macrantha and P. brachyphylla.The anatomy of P. obtusifolia is only slightly different from the form er two species (Ellis 1985b).All the taxa share characters such as the presence The abaxial epidermis of all specimens examined lacked both microhairs and macrohairs and, as noted by Ellis (1985a), the abaxial epidermides of all the species are almost indistinguishable.However, occasional needle like prickles were observed.Figure 2A illustrates the abaxial epidermis of P. obtusifolia.
The adaxial epidermides of P. macrantha, P. caespitosa and P. brachyphylla were almost identical (Figure 2B, C), but the epidermis of P. obtusifolia differed, in that it had a denser distribution of prickles (Figure 2D).In all the species, the prickles are needle-like and angled almost par allel to the epidermis.These prickles appear to be modi fications of the epidermal cell surface (shown magnified in Figure 2C).These prickles are unlike those found in Pentameris (Barker 1993), Pentaschistis (Ellis & Linder 1992) and at least one species of Merxmuellera (Barker & Ellis 1991).The prickles of these latter taxa are slightly inflated basally, variable in length but longer than in Pseudopentameris, and protrude away from the epidermis.
Microhairs with minute, deflated apical cells were ob served at the bottom of the adaxial furrows (Figure 2E).These bicellular microhairs are similar, except in size, to those reported in Pentameris (Barker 1993).Microhair size ranged from 55-67 jim, with a mean of 64.8 pm samples), and (unlike in Pentameris) no species level size differences were apparent.No microhairs were observed on the abaxial leaf surface.
The presence, distribution and nature of microhairs in this genus has been a contentious issue . De Wet (1956, 1960) implies that, as Pseudopentameris has a panicoid epidermis, bicellular hairs (microhairs) would be present, presumably on the abaxial surface.In contrast, Renvoize (1981) considered microhairs to be absent, but later he described 'long slender papillae, which in a few instances appear to bear the remains of a thin-walled apical cell' which were associated with the adaxial surface (Renvoize 1986).This description matches the structure shown in Figure 2E, and it appears that these papillae are in fact microhairs.

CYTOLOGY
The cytology of the genus Danthonia was examined by De Wet (1953, 1954a, b), who was of the opinion that the base chromosome number of this group is x = 6.The subsequent division of the genus Danthonia resulted in limited cytological knowledge for each of the segregate genera.Additional research by Spies & Du Plessis (1988) and Du Plessis & Spies (1988) has increased cytological knowledge about the Danthonia segregates but no chro mosome counts of the species of Pseudopentameris have been published.From analyses of the relationship between stomatal size and chromosome counts, De Wet (1954a) predicted a count of 2n = 24.This figure is credible, as many of the African arundinoid genera display intraspe cific polyploidy (see the study on Chaetobromus by Spies et al. 1990).

CONSERVATION STATUS
Two species in the genus are mentioned in local con servation texts.Pseudopentameris obtusifolia (under the name of Pentameris obtusifolia) is listed by Hall et al. (1980) as a threatened species, the conservation status of which is given as 'uncertain'.However, owing to the no menclatural difficulties experienced with this taxon (dis cussed below and in Barker 1993) and the lack of voucher citations, the exact identity of this taxon is not certain.Pseudopentameris hrachyphylla is considered by Bond & Goldblatt (1984) to be rare, as it is restricted to the Cale don area.Both these species are restricted to relatively small areas at a low to mid altitude.Therefore, although not presently under threat, their narrow habitat range places them in a category that may best be described as 'potentially threatened'.

PHYLOGENETIC RELATIONSHIPS
The genus Pseudopentameris now contains four spe cies.These species are united by an unusual fruit mor phology and the presence of pseudostigmata (described above).This fruit morphology is rare in the southern Af rican Arundineae.The only other arundinoid taxon with a similar fruit type is Chaetobromus Nees, the fruit of which are somewhat smaller, but similar in all other re spects except the possession of pseudostigmata (Barker 1994).The close relationship between these two genera is further supported by DNA-based molecular studies (Barker et al. in prep.).As Pseudopentameris has both anatomical and morphological autapomorphies (the pres ence of inflated abaxial epidermal cells and the presence of the pseudostigmata on the apex of the ovary) the genus is distinct from Chaetobromus, and the two genera ought not to be combined.Autapomorphies for Chaetobromus include spikelets which disarticulate below the glumes and dimorphic florets.In Pseudopentameris, and many other arundinoid genera, the spikelets disarticulate above the glumes.Chaetobromus is the only arundinoid genus which has dimorphic florets: the basal ones are glabrous, the up per ones variously hirsute.
Affinities between Pseudopentameris and other 2-flowered arundinoids, Pentameris and Pentaschistis, have also been postulated (e.g.Ellis 1985b: Ellis & Linder 1992).Pen tameris is presently defined by fruit characters (an achene with an apical crown of hairs; Barker 1993), whereas Pen taschistis.as presently defined, has no obvious autapomorphy, and may include both Pentameris and Prionanthium Desv.(H.P. Linder pers.comm.).While the presence of 2flowered spikelets may indicate some degree of relationship of Pseudopentameris to these genera, other features such as the sclerophyllous type of arundinoid leaf anatomy suggest that these relationships extend beyond these taxa to genera such as Merxmuellera.The assessment of phylogenetic re lationships of Pseudopentameris and other arundinoid taxa therefore needs to be done in the broadest possible context, and should include all taxa and not be restricted to the south ern African genera.Such a study is presently under way (H.P. Linder et al. in   Pentameris squamosa Stapf: 515 (1899); Linder & Ellis: 91-103 (1990).Type: Caledon Div., Nieuw Kloof in Houw Hoek Mountains, Burchell 8076 (K, holo.; PRE.iso.!).
Plants perennial; scandent.sometimes branched.Stems branching, 0.4-1.5(-3.0)m long, woody and persistent, flowering culms produced annually from apical growth points.Leaves: sheaths appressed to culm, persistent; sheath mouth slightly bearded; leaf blades linear, 25-120 mm long, somewhat rolled and rigid, inflorescence pani culate.lanceolate and contracted.Spikelets 8-45, laterally compressed, 2-flowered; glumes two, more or less equal, 18-25 mm long, chartaceous.minutely scabrid, 3-nerved basally, central nerve extending to glume apex; lemma body uniformly pubescent, 4.5-6.0mm long, 9-nerved, nerves anastomosing into awn base and a 8-12 mm long lateral bristle; lemma lobes 1.5-3.0mm long, acute to acuminate, partly to fully adnate to bristles; central awn geniculate, scabrid, contorted basally, 7.5-11.0mm long from base to knee.12-16 mm long from knee to tip; palea longer than lemma body, bifid at apex.Lodicules two, densely ciliolate, cilia as long as lodicule body.Stamens three, 4.5-7.5 mm long.0 \a ry glabrous, with apical pseudostigmata, styles short to absent.Fruit 4 mm long.Flowering time November to January.The known distri bution of this species is indicated in Figure 3A.Schweickerdt (1938) points out that Steudel (1829) and Hochstetter (1846) probably described new taxa from dupli cate material collected by Baron von Ludwig.Hochstetter called his species Danthonia obtusifolia Hochst., whereas Steudel named the material Avena rigida Steud.Schweick erdt showed that the name Avena rigida had been used earlier by Marschall von Bieberstein (1808) and therefore Steudefs later homonym is rejected.The Ludwig specimen at OXF bears the name Avena rigida.and is considered to be the holotype or perhaps an isotype of this name.The specimen seen by Hochstetter was not traced, so the same specimen (Ludwig specimen at OXF) is considered to be probably the isotype of the name Danthonia obtusifolia.Hochstetter's (1846) D. obtusifolia has priority over Stapf's (1897) D. squarrosa.Schweickerdt notes that had Stapf seen the type material on which Steudel had based his description, he would have been able to identify his species as Danthonia obtusifolia Hochst.
The type of S tap f s Pentameris squarrosa is a Burchell specimen, Burchell 8076.There is a fragment of this specimen in PRE, which originates from Kew;.The enve lope is annotated by Schweickerdt.who wrote "a fair match of the type of Avena rigida Steud. in the Herb.Fielding.Oxford'.This fragment, consisting of two leaves and two spikelets.comprises sufficient material to allow comparison on the basis of a number of characters.It is a good match of the type specimen of Avena rigida.It can therefore be confirmed that Pentameris squarrosa is a synonym of Pseudopentameris obtusifolia.This nomenclatural confusion is reflected in past stud ies and curatorial practices (cited above).The use of these confusing names without voucher citations in red data books (e.g.Hall et al. 1980) results in further confusion and renders the information in such texts meaningless.In order to assist in the identification and curation of speci mens of this taxon.an extended list of voucher specimens is provided.The herbarium code JF is used for the her barium at Jonkerhoek Forestry Station.
The absence of a type specimen, and thus accurate flo ral measurement data, caused some problems in the allo cation of the epithet macrantha to the relevant cluster in the PCA study.Despite the lack of the type specimen, Schrader's (1824) description includes the spikelet length, given as 1.5 inches (37.5 mm).As the glumes extend well beyond any floret parts except the awns, this spikelet measurement is interpreted to equate to that of the glumes.As this figure (37.5 mm) is close to the mean glume si/e (37.14 mm) for one of the two clusters distinguished by the PCA.this cluster is allocated the epithet macrantha (Figure 1).
In contrast to the paucity of information provided by Schrader.Trinius (1827) provides an elegant description of this species, referring to an elongate woody rhizome and a branched culm.This description is accompanied by a detailed illustration showing a well-developed rhizome.These characters clearly match the morphology of the specimens in the rhizomatous cluster of the PCA.which Schrader's description implies represents P. macrantha.Unfortunately, although citing Schrader's description, Trinius does not provide a collector and number for the specimen on which his description or illustration is based, and mentions no type specimen.Therefore the link be tween his clear description and the Hesse specimen on which Schrader's description w as based is tenuous at best.The Trinius herbarium is presently housed at St Peters burg (Leningrad).The grass collection, including Trinius's types and the material on which his illustrations are based, have recently been databased (R. Soreng.Cornell Univer sity pers.comm).Computer records indicate that these include several specimens of Danthonia macrantha.al though the type status of these is not certain.All attempts to communicate with this herbarium have failed, and thus the unsatisfactory solution of neotvpification has been im plemented.The second cluster in the PCA is considered to be a new species and is described below.
Plants perennial with woody or fibrous basal under ground parts, caespitose.Stems absent.Culms annual.0.7-1.0m long, sometimes decumbent.Iwaves: sheaths basal, persistent, often purple, apprcssed to culm: sheath mouth somewhat auriculate; leaf blades 45-500 mm long, open and Hat or rolled.Inflorescence paniculate, lanceo late and contracted, upper culm and panicle branches pur ple.Spikelets 10-20.laterally compressed, 2-llowered.highly reduced third floret rarely present: glumes two.more or less equal, 33-60 mm long, 5-9-nerved.char taceous.minutely scabrid or glaucous, dark purple when young: lemma body cartilaginous, apically pubescent, basally glabrous, 8-10 mm long, 9 -1 1 -nerved, nerves anastomosing into awn base and 15-30 mm long lateral bristle: lemma lobes 2-4 nun long, acute to acuminate, partly to fully adnate to bristles: central awn geniculate, scabrid, contorted basally.10-18 mm long from base to knee.17-27 mm long from knee to tip: palea longer than lemma body, bifid at apex.Lodicules two, glabrous or sparsely ciliolate.Stamens three, up to 9 mm long.Ovary glabrous, with apical pseudostigmata, styles short to ab sent.f ruit 5-6 mm long.Flowering time October-November.The known distribution of this species is shown in Figure 3D.This species is characterised by both its larger floral structures and its caespitose.unbranched growth form, hence the specific epithet.The glumes, culms and panicle branches are often dark purple when young, turning a dark brown biscuit colour with age. in comparison to the pale yellow of the glumes of P. macrantha.Although occurring w ith P. macrantha in some localities (e.g.Cape Point Na ture Reserve, where they have been observed growing within a metre of each other).P. caespitosa flowers sub stantially later than P. macrantha.thus precluding any pos sible hybridisation.
In addition to the morphological differences described above, certain differences in the leaf anatomy have also been observed.In a phenetic study on a limited number of anatomical samples.Barker (1990) noted that speci mens of what was then named P. macrantha were sepa rable into two groups on the basis of three anatomical characters.One of these groups corresponds to the taxon here named P. macrantha and has a leat which is V-shaped in cross section, with narrow furrows between the ribs and prickles distributed over the exposed adaxial rib sur face.The other group corresponds to P. caespitosa (de scribed above) in which the leaf is U-shaped, the furrows are open and the prickles are restricted to the edges ot the adaxial ribs.
FIG URE 2.-L eaf blade SEM o f P seudopentam eris.A, D, P. o b tu sifo lia , E llis 2342. A. abaxial e p id erm is show ing large inflated ep id erm al cells and sm all prickles; D, deeply ribbed adaxial e p id erm is heavily adorned w ith prickles.B, P. b ra ch yp h ylla , B o u ch er 3 5 7 a , adaxial e p id e rm is show ing open furrow s and ribs w ith prickles.C , E, P. m a cra n th a , E llis 2515: C , adaxial prickles, note also d u m b b ell-sh ap ed silica bodies; E, m icrohair from adaxial ep id erm is, note m inute apical cell.Scale bars: A, 225 Jim; B, 240 ^im; C, 115 ^m ; D, 325 E, 17 Jim.