External fruit morphology of southern African Arundineae ( Arundinoideae : Poaceae )

Fruits of a number of taxa of all indigenous southern African arundinoid genera were examined by means of SEM. Size, shape, compression, surface sculpturing, embryo anti hilum features were recorded and fruits of all genera are illustrated. Results are compared to existing descriptions. The genera are placed in five informal groups according to similarities noted in the fruits.


INTRODUCTION
In his book o f grasses o f the British Isles, which is now in its third edition, C.E. Hubbard (1984) included a section on the 'seeds' o f the grasses, which gave illustra tions of the grains o f various genera, as well as a key to the genera based mainly on characteristics o f their fruits (Hubbard 1984(Hubbard . 1st edn 1954)).Despite this emphasis, fruit morphology is still a neglected aspect of grass systematics.In this study, observations on the fruits o f the southern African Arundineae are presented.

The trihe Arundineae
The tribe Arundineae belongs to the subfamily Arun dinoideae which Kellogg & Campbell (1987) consider to be polyphyletic-an assemblage o f basal groups and ev olutionary dead ends.
The classification by Clayton & R envoize (1986), which is followed in this work, provides a broad definition for the tribe Arundineae which encom passes most o f the genera in the subfamily.The tribe is a fragmented, hetero geneous group of numerous isolated or weakly linked gen era.Other workers divide the subfamily into numerous smaller tribes on the basis o f phenetic similarity (Watson 1990) or breeding systems (Conert 1987).
The Arundineae are cosm opolitan.O f the approxi mately 40 genera in the tribe, 16 occur in southern Africa.Two of these, Arundo L. and Ph rag mites Adans., are pan demic, and species of a third, Cortaderia Stapf, were in troduced into South Africa to control soil erosion on mine tailings dumps (Robinson 1984).

Terminology
There is as yet no general agreement on the terminol ogy to be used for describing the fruits of the Poaceae.In the Poaceae the ovary is uniloculate.After pollination and fertilisation (ignoring examples of apomixis) the ovule develops into a seed, and the ovary wall becomes a fruit coat or pericarp.In the majority of grasses, the seed and pericarp are fused, forming a grain (Clifford & Watson 1977).T hese authors equate this term to the term 'caryopsis' but point out that the grain may take the form of one o f a number of different structural types, such as 'achenes', 'nuts' or 'berries'.In the fruit of some grasses, the seed is free from the pericarp.Clayton & Renvoize (1986) refer to such a fruit as an achene.However, Sendulsky et al. (1987) consider it a modified caryopsis, formed by the collapse of either the endocarp or the endoand mesocarps at a late stage of development, and rec ommend that use of the term 'achene' in grass biology be discontinued.
In the light of the uncertainty in terminology, the grass grain will be referred to as a fruit in this study.When the necessary distinction between a grain with a free pericarp and a grain with an adnate pericarp has to be made, the term 'achene' will be used for the former, and 'caryopsis' for the latter grain type.

The fru it o f the Arundineae
Clayton & Renvoize (1986) describe the fruit of this tribe as a caryopsis, sometimes with a free or separable pericarp, rarely an achene.They name three genera within the tribe which possess the achene fruit type: Pyrrhanthera Zotov, Dregeochloa Conert and Pentameris P. Beauv.The latter two arc endemic to southern Africa and were included in this study.
SEM studies on the fruits of arundinoid genera by Barker (1986Barker ( , 1989Barker ( , 1990Barker ( , 1993, and , and  ered fruits of most species o f Pentameris P. Beauv.and Pseudopentameris Conert.Davidse (1988) exam ined the fruits of the genus Prionanthium Desv.These studies are discussed below.In this paper the external morphology of the fruits of the southern African arundinoids is explored.

Fruit collection
Mature fruits of several taxa in each genus were col lected from herbarium specimens housed in the National Herbarium in Pretoria (PRE).Taxa representing all en demic southern African genera were sampled, as well as taxa of the pandemic genus Phragmites Adans.Fruits o f Arundo L. were unfortunately not available for study.The specimens and species from which fruit were obtained are listed below.
As the fruits were dry, no additional desiccation pro cedures were followed.Specimens were mounted on the SEM stubs by means o f two-sided tape.They were then coated in gold-palladium and examined using an ISI-SX-25 Scanning Electron M icroscope.Photographs were taken using 6() x 70 mm format black and white Ilford FP4 100 ASA film.

OBSERVATIONS AND DISCUSSION
The salient features of the fruits examined are listed in Table 1, and these features are illustrated in Figures 1-5.Although only a few of the species in the larger gen era were examined, the illustrations show that the fruits o f different species within a genus are relatively uniform.
The differences at generic level are more distinct, al though some genera display strong similarities.These sim ilarities in fruit morphology may be used to form groups of genera on a purely phenetic basis, where char acter and structural homologies are assumed.Four such groups were assembled on the basis of shared fruit char acters.A fifth group comprises three remaining genera, the fruits of which resemble no other genera in the tribe.As the tribe is considered to be polyphyletic, these groups may represent various monophyletic lineages within the Arundineae.These informal groups, and the genera that are placed in each, are discussed below, and the salient features of these groups and genera are presented in Table 9   Genera in this group share fruits which are broadly obovate, strongly dorsiventrally com pressed, generally glabrous or almost so, have em bryos Vy to \the fruit length, and punctiform hi la.

Karroochloa Conert
Fruit of all four species o f this genus were exam ined and illustrated in this study (Figure 1A-D).The differ ences between the fruit o f the species exam ined is slight, but a larger sample size may reveal that these differences intergrade.
Karroochloa was separated from D anthonia DC. on the basis of, among other characters, differences in the fruit (Conert & Tiirpe 1969).The latter genus has fruit which are described as having a hilum half to three quar ters the length o f the fruit and an em bryo no longer than a third the length of the fruit.Further differences in the fruit morphology between Karroochloa and Danthonia in clude the size of the fruit, surface features, and nature o f the endosperm (Conert & Tiirpe 1969).Clayton & R envoize (1986) include Karroochloa, along with Merxmuellera Conert and several other non-African genera, in Rytidosperma Steud.However, the de scription of the fruit of Rytidosperma does not match that observed in Karroochloa in this study.Thus the placement of Karroochloa in this genus ought to be questioned, es pecially in the light of the strong apparent similarities be tween Karroochloa and Schism us, which is not included in the broad Rytidosperma complex.
Schismus comprises five species, four o f which occur in southern Africa.Fruit o f three of the five species were examined (Figure 1E-H).They are all very similar and closely resemble those o f Karroochloa.Results presented in Table 1 agree essentially with descriptions provided by Clayton & Renvoize (1986) and Conert & Tiirpe (1974).

Dregeochloa Conert
Dregeochloa was created by Conert (1966) to accom modate two Danthonia species with an unusual fruit mor phology: an obovate caryopsis with a punctiform hilum, and a membranaceous, removable pericarp.As mentioned in the introduction, Clayton & Renvoize (1986) consider the fruit of this genus to be an achene.The fruits of D. pumila were examined (Figure 2A.B).The only other species in the genus, D. calviniensis Conert. is extremely rare.
The pericarp appears to be flaking and easily rem ov able from the fruit (Figure 2A).It is possible that the fruit surface was damaged during manipulations, but the fra gility of the pericarp and the apparent ease of removal suggests that it is free from the layers beneath.This and other features recorded in Table 1 agree with the descrip tions by both Conert (1966) and Clayton & Renvoize (1986).Conert (1966), however, describes the pericarp as membranaceous w hereas C layton & R envoize (1986) refer to it as thick.
Dregeochloa is placed in this group on the basis of the shape and compression of the fruit and the large embryo.Conert (1971) observed that the fruit of this taxon is unlike that of any other arundinoid.The somewhat rugose surface and the fact that the fruit appears to be an achene, as noted by Clayton & Renvoize (1986) and corroborated here, suggests that the relationship o f this taxon to the Schismus-Karroochloa alliance is somewhat tentative.

G roup 2: Chaetobrom us and Pseudopentameris
The genera in group 2 have narrowly elliptic fruits with reticulate sculpturing, narrow, canaliculate hila extending almost the entire length of the fruit and embryos which are a fifth to a quarter the length of the fruit.These two genera have not been previously considered as closely re lated.

Chaetobrom us Nees
Chaetobromus comprises either two (Barker in Gibbs Russell et al. 1990), three (Clayton & Renvoize 1986) or four species (Chippindall 1955).Spies et al. (1990) have suggested that the genus may be a polyploid series, divis ible into two phenotypic groups: C. dregeanus and C. involucratus.Fruits of both these groups were examined and found to be identical (Figure 2C), and corroborate the descriptions of Clayton & Renvoize (1986).
Psei (dopentameris Cone rt Barker (1986Barker ( , 1989Barker ( , 1990) ) found the fruits of all spe cies to be almost identical, differing only in size.Figure 2D shows the embryo and surface features of the fruit of this genus.Fruits of this genus are unique in having stigmata-like apical appendages, termed pseudostigmata by Barker (1990, and in prep.).These structures, visible on the apices of the ovary, developing and mature fruit, are readily deciduous, being best observed in mature ovaries and developing fruit.These structures have not been ob served in Chaetobromus, but no immature fruits of this genus have been examined.The descriptions of the fruit of Pseudopentameris by Clayton &Renvoize (1986) andGibbs Russell et al. (1990) agree with details given in Table 1.The decision to transfer Pentameris obtusifolia (H ochst.)Schw eick. to Pseudopentameris (B ark er in prep.) was made partly on the basis of strong similarities of the fruit characters.

G roup 3: Tribolium and Urochlaena
The fruits of genera in this group are ovate, strongly dorsiventrally compressed, with a flaking pericarp.The embryo is a quarter to half the length of the fruit.
Tribolium Desv.This genus comprises cither 11 (Gibbs Russell et al. 1990) or nine species (Spies et al. 1992).Clayton & Renvoize (1986) consider this genus to be an outlier in the Arundineae, with at least superficial similarities to genera in the Eragrostidcae.
The fruits of this genus are described by Clayton & Renvoize (1986) as having a reluctantly separable peri carp.This description agrees with that given by Gibbs Russell et al. (1990), who describe the fruits us being small, with a short hilum and embryo, and a fairly loosely adherent pericarp.These descriptions match observations made in this study, where the fruits of three species in this genus were studied (Figure 2FL-G).As shown in Fig ure 2G, the rugose outer layer o f the fruit is separable, matching the decription given by Clayton & Renvoize (1986).The exposed surface below the pericarp is also rugose.

Umchlaena Nees
The fruit of the monotypic genus V w ch la en a is de scribed by Clayton & Renvoize (1986) as a caryopsis with a free pericarp, an observation repeated by Gibbs Russell et al. (1990).The embryo is described by these latter au thors as being large, while the hilum is short but large.The fruits of Urochlaena (Figure 2H) are similar to those of Tribolium.
Tribolium and Urochlaena thus form a natural group on the basis of all the fruit characters examined.In par ticular, the nature of the flaking and separable pericarp indicate a close relationship betw een the two genera.These observations augment the data from other studies suggesting strong similarities between Urochlaena and Tribolium (Clayton & Renvoize 1986;Ellis 1988;Spies et al. 1992).

Group 4: Centropodia, M erxm uellera, S typpeioch loa, Prionanthiurn and Pentaschistis
The fruits of the genera in this group share features such as a reticulate surface, a hilum approxim ately half the length of the fruit, an em bryo about a quarter to a third the length of the fruit and only slight com pression in either plane.However, the genera in this group are probably, at best, only distantly related, given the differ ences in floral and vegetative morphology.

Centropodia Rchb.
Two of the four species in this genus were examined.Conert (1962) describes the fruits o f Centropodia as being naked caryopses, with linear hi la and em bryos one third the length of the fruits.This description approxim ates the observations made in the present study (Figure 3A, B), although Conert's (1962) description of the hilum as being linear differs from the present study where it is shown to be shallowly punctiform (Figure 3B).
In view of the differences in photosynthetic pathways and thus leaf anatomy (Ellis 1984), the relationship of Centropodia to the other genera in this group should be considered as tentative and distant.

Merxmuellera Conert
At present, 20 species are recognised in this genus, two of which arc known only from Madagascar.I he fruits of three southern African species of the genus were stud ied.Each of these species cam e from a different part of the distribution range of the genus and from a different habitat.The fruits of all three species studied were re markably similar (Figure 3C-E).
The fruits of Merxmuellera have been described as brown, free, almost terete caryopses, two to three milli metres in length (Conert 1970).This description, though sparse, agrees with observations made in this study.Clay ton & Renvoize (1986) include Merxmuellera in Rytidospenna , and their description o f the fruits o f this latter genus is considered too broad to be applicable to M erx muellera alone.

Styppeiochloa De W inter
Styppeiochloa gynoglossa, the only southern African species in the genus, was investigated.De Winter (1966) described the fruits of Styppeiochloa as elliptic caryopses with a linear hilum three quarters the length of the grain and an embryo one quarter the length of the grain, a de scription matching observations made here.Clayton & Renvoize (1986) do not provide a description of the fruit, fhe fruit of S. gynoglossa is illustrated in Figure 3F-H.

Pentaschistis (Nees) Spach
Pentaschistis is the largest genus in the tribe, compris ing 72 species (Linder & Ellis 1990).The fruits of only six species, including those of P. pusilla (= Poagrostis pus ilia (Nees) Stapf) were studied.Despite the size of the genus, little variation was found in the external morphol ogy of the fruits investigated (Figure 4A -H ).
The fruits were found to conform with descriptions given by Clayton & Renvoize (1986) and Gibbs Russell et al. (1990).However, the pericarp description given by the latter workers (free to loosely adherent to fused) was found to be too broad; the pericarp appeared to be fused in all the species studied.

Prionanthiurn Desv.
Prionanthiurn comprises three species.It was revised by Davidse (1988) who described and illustrated the fruit o f two species.The genus shares many fruit characters with Pentaschistis, an observation supporting the hypoth esis that these two genera are closely related, as proposed by Linder et al. (1990) on the basis o f the presence of unusual glands in both these genera.In addition.Davidse (1988) considers the base chromosome number of x = 7, which is common to both genera, to also be an indication o f their close relationship.This is in contrast to Clayton & Renvoize (1986), who place Prionanthiurn close to Tribolium and Urochlaena.
( iroup 5: special cases Three genera are placed in this group, not because of any similarities in the nature of the fruit, but simply be cause their fruits resemble none of the fruits of the other genera discussed above.These 'misfits' are Phragmites, Pentameris and Elytrophorus.

Phragm ites Adans.
Phragmites is a pandemic genus with two species oc curring in southern Africa, P. australis and P. mauritianus.The fruits of these species examined were almost identi cal.As no detailed description of the fruit from past work ers was found, the observations presented here cannot be   compared to past descriptions.The fruits of this genus are illustrated in Figure 5A-C.
The fruits of Phragmites are unusual within the con text of the southern African Arundineae.Their small size undoubtedly assists in their dispersal by wind, the fruit being dispersed in situ in the floret.Further examination of the fruits of other reed-like grasses such as Arundo, Hakonechloa Honda and Molinia Schrank may assist in clarifying the relationships between Phragmites and other related reed genera with wind dispersed fruit.
The fruits of all except one of the nine species of Pen tameris have been investigated and illustrated (Barker 1986(Barker , 1989(Barker , 1990(Barker , 1993)).The fruit of P oreophila N.P. Barker is illustrated in Figure 5D.Pentameris possesses fruits w hich differ m arkedly from other southern African arundinoid genera, in both external and internal structure.The fruits of this genus are achenes with apical appendages in the form of hairs or hair-like structures.
These two characters clearly distinguish this genus from the closely related and probable sister genus Pen taschistis.Thus in a phylogenetic system based on overall morphology, Pentameris would belong in group 4. How ever, on the basis o f phenetic groupings employed here, it is treated as an exceptional taxon.
Elytrophorus differs from the o th er g enera o f the Arundineae in having dimorphic spikelets and a m em bra nous ligule (as opposed to the more usual arundinoid ligule comprising a row o f hairs).
Both species in this genus were examined, and the fruits of these two taxa were identical in all but size, those of E. globularis being far larger than those o f E. spicatus (Figure 5E-G ).Schweickerdt (1942) reported that the em bryo is almost as long as the grain, whereas Clayton & Renvoize (1986) consider the embryo to be only half the length of the fruit, a size more in keeping with that ob served here.Schweickerdt (I.e.) alludes to the caryopsis as 'showing the remains o f the pericarp', implying that the pericarp is separable.However, Clayton & Renvoize (1986) do not consider the fruit o f this genus to be an achene.Instead, the pericarp is described, som ewhat con fusingly, as being free.
The shape o f the fruit of these species is like that o f no other southern African arundinoid.and it is not possible to place it in any o f the other groups described above, which only serves to enhance the confusion o f this taxon's uncertain affinities: Clayton & Renvoize (1986) consider Elytrophorus to have uncertain affinities, and Chippindall (1955) placed it in the Eragrosteae.

CONCLUSION
This study has revealed that the diversity in the floral and vegetative m orphologies o f the genera of southern African Arundineae is parallelled by an equally diverse set of fruit morphologies.However, this diversity is gen erally minimal at infrageneric levels.Despite the diversity in fruit morphology, certain genera are united in almost all fruit characters exam ined (Tribolium and Urochlaena; Karroochloa and Schismus).W here this occurs, it is in terpreted as indicating a strong degree o f relatedness.Un fortunately, only this very lim ited phylogenetic inter pretation is possible using this data.This is because struc tural homologies are unproven, only southern African taxa are being examined, and insufficient characters are ob tained from a purely SEM -based study.The extension o f this study to include non-southern African taxa may, how ever, provide additional data for comparison.
This study is strictly descriptive.In many instances, the generic relationships proposed on the basis o f fruit morphology are speculative.Nonetheless, this study has provided valuable data for incorporation in a phyloge netic analysis which is presently being undertaken in the form of a combined molecular (a survey of chloroplast DNA sequence variation) and morphological survey of the entire tribe.This study has indicated the need for a thor ough examination of the nature of the fruit of grasses at the generic level.

7
of shape of fruit in outline; D-V.dorsi ventral; I^at.. lateral; +emhryo and hilum lengths are fractions of total fruit length.

Group 1 :
K arroochloa, Schism us and D regeochloa

FIGURE 3 .
FIGURE 3. Fruit of genera in Group 4 (Centropodia, Merxmuellera and Styppeiochloa spp).A, B, C. glauca: embryo side, showing large embryo reticulate surface features and remains of style; B, hilar side, showing punctiform hilum.C, M dura, hilar region, close-up of scalariformreticulate surface features.D, E, M. stricta: D, embryo side, showing embryo a third the length of fruit; E, broad, fairly short hilum and remains of styles.F-H, S. gynoglossa: F, reticulate surface features; G, embryo one third length of fruit; H, lateral view Scale bars A 370 Hm; B, 425 nm; C, D, 110 nm; E, 695 ^m; F, 65 ^m; G, H, 465 [im.