The generic status of Oxyanthus gossweileri (Rubiaceae) from Angola

Oxyanthu5 gossweileri from Angola has been described from a single specimen collected in 1906. A search in herbaria holding Angolan collections has revealed an additional flowering specimen, allowing a sound morphological and pollen morphological investigation. It is concluded that 0. g055weileri must be kept in the Gardenieae Gardeniinae but deserves generic segregation. A new genus Ganguelia is described and the necessary new combin~ alion made. G. g055weileri is a pyrrhophytic geofrutex with densely hairy, suborbicular leaf blades with subpalmale venation. ChoroJogically, Ganguelia is endemic in the western part of the Zambezian Region; in the Rubiaceae, the mono specific Calanda (Knoxieae) has a very similar narrow distribution. Ganguelia seems to exhibit two poJlen morphological features which are not yet observed among the Gardenieae genera with tetrad poilen, namely, the ± acalymmate condition of the tetrads and the colpoidorate grains. However, these latter occur together with the pororate ones characteristic for Gardenieae tetrads. A part of the tetrads are decussate; this arrangment, not yet reported from the Rubiaceae, was previously overlooked in several genera and is hence not a peculiarity of the new genus.


Introduction
III a recent investigation of the genus Oxyanrlws (Sonke 1990), superv ised by o ne of us (E.R.), lhe species OxyantJws go.uweileri from Angola was excluded and its position left doubtfu l, main ly because the type examined was too poor for an in-dep th examination. A search in herbaria holding Angolan collections (COl. LISC, LISJC, LISU and ZT) has revealed one additional flower ing specimen in two sheets. Allhough fruits are sti ll not available, a reinvestigati on of this remarkable species is now possible and is presented here.

Material and Methods
Only one herbarium specimen was traced, in addition to the lype , as u result of the above-mentioned search (Teixeira et al. 8903 , see Taxonomy for whereabouts). This specimen was used fo r the dissection of one mature flowe r bud.
A small quantity of pollen, removed from a single flower, was availab le and for the sake of security. a preliminary preparation was made by embedding some grains in Hoyer's medium. The proper pollen sample was acetolysed according to Reitsma (1969) and observed by both light and scanning electron microscopy (S.E.M. Jeol6400, 15 kV). Broken grains were obtained by shaking the pollen suspension in the presence of glass beads (Huysrnans et aJ. 1994). Because orbicules have recently been found to be rather common in Rubiaceae (Huysmans er al. 1995). one open anther was additional ly critical~point dried to observe unacctolysed pollen and 10 check the possible presence of orbicules on the interior anther wall. Most of the pollen in the dried anther was removed using a fm e cactus needle and pollen measurements were made on S.E. micrographs of acetolysed grains. Palynological termi nology follows Punt ef of. ( 1994), whereas morphological terms are after Robbrecht (1988).

Observations
Habit and structure 0. gossweileri is a geofrutex having mass ive woody underground parts and tender supraterranean twigs (Figure 1 A). T hese arc hardly branched and very congested. reaching 1css than 10 <.:m in Icng.h.
Vegetative parIs. Leaves ( Figure IB) are dense ly hairy and have suborbicular blades borne on slender petioles 1.5-2 em long. Blades have a :!: emarginate apex (the midvein often lenninating in a small mucro) and a widely cuneate base. Blades arc coriaceous to almost succulen t. Their mesophyll has many idioblasts filled with tan nins (these are also abundant in ot her organs) and others filled with crystal sand. Leaf vena tion is as un usual for Rubiaceae as is the abovedescribed shape, namely that the lateral nerves tend to arise from the blade bases (subpa lm ate).
Sripules are inlerpetiolar, ovate with acute lip. They are free to the base (not sheathing at all).
Fertile region. Flowers are solitary and probably first terminal, later pushed in a pseudo-axillary position; they are subtended by ovate bracteoles with acute tip.
The flowers (Figure 1 D) are pentamerous and have left -contorted aestivation. The calyx (Figure 1 E, 1-1) is long-tubu lar and prOVided with narrowly triangular lobes. The corolla consists of a slender tube 4-5 cm long. crowned by narrow lobes of abou t 2 em in length.
There arc five anthers (Fig. IF) sessile in the uppermost part of the corolla tu be, their point of attachment situated at If) from their base. Their connective protrudes in a small apical triangular sterile appendage.
The un ited style in mature bud slightly surpasses the anthers, which are sessile just below the throat. It is almost thread-like, but its upper part is sligh tly swo llen; it temlinates in two short, papillar stigmatic arms (Figure 1 G). An annular disk surrounds the base of the style.
The ovary is two-locu lar wit h axile placentation; the two peItate, slightly elongated placentas are adnate to ± the entire septum . Each placenta bears about 15 ovules. Fruits are unknown.
In dumen tum and collelers . Almost all parts of the plant, vcge~  (alive as well as fertile, are densely covered with long soft hairs. Stipulc insides, however. arc glabrous, except for a basal row uf hairs intermingled with collelers. Bracteoles have silky hairs and no colleters insidc.
Insid~ the calyx, there are small groups of 3-5 colleters below cach sinus between the Jobes. Hairs are pluricellular and belong to the cylindrical type.
Pollen . The pollen grains arc united in penn anent tetrads. bcmg tetrahedral (± 70%; Figure 2A. B) or decussate (± 30%; Figure 2C); a single rh9mboidal uniplanar tetrad has also been observed . The size of the decussate tetrads is slightly larger (60-66 ~) than the more commo n tetrahedrally shaped tetrads (53-60 ~m). The individual grains measure ± 34 lIDl. The tetrads arc slightly acalymmate: fusion of the individual grains is moderate, and a suture line, caused by differences in tbe structure of the tec-tum , is visible between them ( Figure 2B). The tectum, hence, forms a continuous envelope around the unit, but is more loose in between the monads. Internally (proximally), the exine is extremely thin ( Figure 2H); o ur broken grains do not permit the observation of more details (transmission electron microscope studies would be needed for better understanding).
Individual grains vary from mostly 3-pororate ( Figure 2F) to occasionally more or less 3-colpoidorate ( Figure 2E ). In the latter case, colpi are very short with rounded ends. Apparently the transition from colpoidoratc to pororate is explainable by reduction of the ectocolpi until Ihey arc congruent with the endopori. The apertures meet (wo by two at six points on the tetrad (Fischer's law). The diameter of the ectoporus (in pororate grains) varies [rom 4.5 to 5.7 J...III1 . Colpoidorate grains show a s hort eClocolpus with irregular edges and a granular colpus mem - branc . The endoporus ( Figure 2H) has a fai nt costa which is only slightly protruding.
T he smooth tectum is perforate to rnicrofoveolatc, with granules on the nexine in the puncta ( Figure 2D) ; it is sUPBorted by ve ry shon columellae. Near the apertures and on the distal poles, the tectum is microfovcolatc, i.e. the columellae are almost completely or completely lacking (l'igure 20). The rounded perforations arc up to 1 IJII1 in diameter (;::: puncta). On the nexine , small granUles or frcc-standing bacula are often visible in the puncta.
Then! are no su prateclal processes. The exine is very thin « 1 jJffi); the thickness of the sexine is equal or slightly larger than that of the !lexine. The endcxine is homogeneous and humpy; no granu les occur.
No o rbicules have been observed on the inner anther wall .
Floral biology. Hardly any information is available, and only fully mature fl ower buds (just before ant hesis) were collected. The anthers in these buds were observed to have opened and released pari of their pollen. It adheres to the swollen upper part of the style lying between the anthers. Pro terandry and secondary pollen presentation consequently seem to be presen t. The very long and narrow floral tubes combined with the presence of a disk suggest sphingid pollination.
Ecology o. gossweileri is. according to its known geographic distributio n, restricted to White's (1983) mapping unit 47, which is a mosaic of Brachyslegia bakerana thicke t and edaphic grassland.
The species is probably pyrrhophytic or, at least, fire lolerant, as indicated by the woody underground parts and the signs of burning present in the two examined specimens. Pyrrhophytes and geo[rutices are common in the grasslands of the central pia· teau, where fires are frequent. and usually flower before the end of the dry season .
O. gossweileri occurs on sandy soils. fanning groups with 1-2 m diame ter, and prod uces flowers simultaneo usly wi th the pro· duction ofleaves on new shoots in September and October, i.e. at the end of the dry season.

Distribulion
The spec ies is apparently endemic to a narrow part of south-central Angola (districts Bie and Cuando-Cubango). The two examined specim ens were collected in the central African Plateau, at about 300 km from o ne another. Gossweiler 3275 was collected on the left margin of the Quiriri river, near Cassuango. i.e. in the eastern part of the region sometimes designated as the Ganguelas (Diniz 1973). The locali ty (known as Ca.o;soango) is shown on Gosswciler's Carta fitogcognHica de Angola (1939) near Longa (district Cuando-Cubango). However, recent maps do not mention this locality and no informa tion o n its loc ation co uld be ob tain ed from the local populations duri ng an expedition to this area in 1960 (EJ. Mendes, unpublished data in cluded in the card catalogu e of Angolan locali ties for internal use at L1SC). The more recent specimen Teixeira el al. 8903 was collected further north , betwee n Cuemba and Munhango (district Bii).

Discussion
The generic pasilian of 0. gossweileri S. M oore originally assigned the species to Oxyanthus. without proper comm ent on its generiC position ('a very remarkable plant on account of its lowly habit and small leaves '). His cho ice was surely based on floral ressemblances.
There is ample evidence to place thi s documented remarkable Angolan species in the same tribe as Oxyanthus, namely, the Gardenieae. The large flowers in pseudo-axillary position, contorted aestiva ti on, and multi-ov ulate placentas make a pOSit ion in S.-AfrTyctskr. PJalltk. 1996, 62(1) this tribe obvious, despite the absence of fruits, which have high val ue for tribal diagnosis in the Ixoroideae. Our finding Ihat the poilen grains are in pennanent tetrads corroborates this, and excludes the Gardcnieae -Diplospori nae. Tn Rubiaceae, tetrad pollen is very rare. It is restricted to one of the three genera of the Henriquezieae (a neotropical tribe needing no further consideration here), and to 13 of the more than 60 genera of the Gardenieae -Gardenii nae. In this subtribe. tetrads occur especially in the African representatives, and characterize the 11 following African genera : Preussiodora, Pleiocoryne, Oligocodon, Caiochone, Masslliaria, Macrosphyra, Afractogyne, Oxyanthus, Milriostigma, Euclinia and Gardenia .
Tetrads are not kn own from tropical Asi an Gardeni inae; in the New World. they are restricted to Casasia and Randia (Robbrecht & Puff 1986, Table 1).
With the exception of Gardenia, and including the genus Oxyanlhus to which the species was assigned, the 11 African Gardeniinae genera with tetrad pollen mentio ned above are typical rain -forest genera s howing a multi tude of morphologi cal differences with 0. gossweileri. Oxyanthus, the genus to which the species was originally assigned, should be ruled out, in particular because of the habit (in Oxyanrhus only trees and large shrubs, except for one species being a monocaul dwarf shrub), inflorescences (many-flowered in Oxyanthus), leaf-morphology (in Oxy-anOws large ± elliptic leaves having relatively short petioles and pcnninerved acuminate blades up to 70 em long), the anthers (in Oxyanthus generally a well-developed connective having apical .a.ru! basal sterile appendages) and the style (in Oxyanlhus the stigmatiC lobes above the sWQllen style apex are minute and nondiverging). In addition, Oxyanlhus is an essen tially Guinea-Congolian and Afromontane forest genus.
Only Gardenia contains a large number of species which are well adapted to Sudano-Zarnbezian conditions, among them even three geofrutescent species, G. subacaulis (widely distrib· uted in the eastern part of the Zarnbezian region), G. brachythamnus (Z ambezian region: Botswana and Zambia) and G. linneae (a species of the eastern part of the Sudanian region, known from the Central African Republic. Chad and Sudan). G. subacaulis and G. linneae are clearly aSSignable to Gardenia, notably subgenus Bergkias (Verdcourl 1979). The latter author even considered G. silbacaulis so closely related to G. temifo/ia that ' the specimens [of G. subacaulisJ are certainly not just variants of G. lernifolia due directly to burning.' I t is obvious that o. gossweileri is definitely not a member of Bergkias having characteristic, more stout flowers with short tubes, wide lobes and exserted anthers.
There are two reasons which make a further comparison of  Figure   191) shows several detail differences, such as the distribution of colleters over the whole surface of the base of the calyx tubes.
More importantly. the above-mentioned geofrutescent Gardenia brachythamnus also has narrow corollas. This former Randia has first been related to Xeromphis (;::: Catunaregam) by Keay (1958) and was later transferred to Gardenia (Launert 1960). This geofrutex is a rather tall, branched shrub with decumbent to ± erect stems fr om long rhizomes. Its miniature, coriaceous leaves are tufted on short shoots which also bear the tenninal, solitary flowers. G. brachythamnus needs detailed morphological and anatomical investigation, but seems c losely related to the African Gardenias with small, leathery. few -seeded fruits. such as G. jiorii.
Finally and most importantly. !!ll species of Gardenia have parietal placentation. It is no t very likely that it should have shifted 10 axile placentation in one single species, deviating in other important features. We consequently believe that O. gossweileri is neither an Ox),ontlllls nor a Gardenia and arc convinced thai it needs segregation in a new genus. Chorologicai facts somewhat support this conclusion, although little is known about the degree of generic endemism of this part of Angola. W hite (1983: 87) stated that there are only a few endemic genera in the Zambczian region as a whole; he enumerated on ly six genera, all monospeciflc. but restricted his list to woody ones. In the Rubiaccae, in total 13 genera (woody and herbaceous ones) were found to be endemic of the Zambezian region (Robbrechl, unpoblishcd). Only one of Ihcm, the equally monospccifi c Calallda (tribe Knoxieae, distribution map in Puff & Robbrecht 1989), is restricted to Angola; it has a generic distribution pattern fully comparable to that of the genus recognized here. It shou ld be stressed that the area where the two geofrutesce nt taxa Calanda and O.gossweileri occur is the most important diversity centre for this growth form in tropical Africa (White 1976). In Africa, the Rubiaceae seem to be tbe family containing the largest number of gcofruti ces; this life form has evolved ind ependently many times and in nine different tribes (Robbrecht 1988: 36).
A last question remains to be answered: is the separate generic status supported by pollen morphological characteristics? Persson (1993) gave a general palynological overview of the subtri bc Gardeniinae. His generic descriptions are concise. but many genera are ill ustrated by S.E. mi crographs. Pollen of the Gardeniin ae from Gabon was studied by Keddam -Malplanche ( 1985). The two genera discussed above, Oxyanthus and Gardenia, arc thus documen ted paJyn o Jogica lly. Keddarn -Malplanche (1985: 42) discussed the pollen morphology of nine representatives of Oxyanrhus from Gabon. O. gos.nveileri fits into the general description of the genus, save for the thickness of the exine, varying between 1.75 and 5 ).Ul1 in Oxyanrhus, while in this case it never exceeded 1 ).Ul1.
As regards the o ther African Gardeniinae genera with tetrads, deSCriptions andlor illu strations are available for all genera (Robbrecht 1978: Preussiodora;Persson 1993). They all have 3por(or)ate grai ns but most of these genera deviate from the species we studied in having a reticulate sex ine with sometimes a decreasing lumina size towards the poles (Alraclogyne, Calochone, Euclinia, Macrosphyra, and Preussiodora ). Oligocodon and Mirrioscigma pollen also differ conspiciously from O. gossweileri in sexi ne ornamentation: the fanner genus has perforate pollen beset with spines and verrucae, pollen of the latter is rugulate with branched co lumellae. Only the description of Pleiocoryne pollen (no illustrations) resembles the studied species in a few important characters: perforate sexine with no columellae, a very thin exine (OA )..lID), and 'a granulate membrane near the 21 ora.' This li anescent genu$ from West and Central Africa was documented by Kirkbride & Robbrecht (1984); it shows numerous morphological differences with the species discussed here , however, for example in the many-nowercd inflorescences.
Decussate tetrads have not explicitly been reported from the Rubiaceac . Keddam-MalpJanche (1985) did depict tetrahedral and decussate tctrad$ on her S.E. micrographs of Oxyanchus, and only decussate tctrad$ on her illustrations of Gardenia, hut she did not mention at all the occurrence of these two types o f tetrads. Persson (1993) illustrated only te trahedral tetrad s but made no menti on of the arrangmen t in IUs descriptions . We infer that decussate tetrads were previously overlooked in rubiaceous literature but are relatively rare ; tbey are at present reported from Oxyanthus, Gardenia and our new genus.
On the whole, the tetrad-possessing Gardeniinae are rather stenopalynous and not studied well enough to make a more sound comparison with o. gossweileri possible . II nevertheless seems that we have detected the following two characteristics not yet observed in other genera of the subtribe and consequently possibly supportin g the separate gener ic statu s: (l) the tetrads being slightly acalymmate, and (2 ) co lpoidorate grains occu rring together with pororate ones.