Taxonomic studies in the Acanthaceae: A reappraisal of the genera Duvernoia and Adhatoda in southern Africa

A reappraisal of floral and vegetative characters of southern African species of Adhatoda Miller and Duvernoia E. Mey. ex Nees is made with reference to selected tropical species of Adhatoda and to species of the closely related genus Justicia L. The generic status of Duvernoia is confirmed and the generic characters expanded. The southern African species of Adhatoda are clearly distinct from Duvernoia. Their relationship to the tropical members cannot be assessed until a full revision of the genus is undertaken. S. Afr. J. Bot. 1985, 51: 475-484


Introduction
Duvernoia E . Mey. ex Nees, a genus of two species confined to south-east Africa, is closely related to, and often regarded as inseparable from, the widespread genus Adhatoda Miller. A brief review of the history of these two generic names and examination of the generic descriptions revealed that distinctions between them have often been confused. Their relationship with Justicia L. is also subject to debate. Therefore, an attempt was made to reappraise the characters that have been used to distinguish these genera in the past, using southern African species of Adhatoda and Duvernoia together with a limited number of specimens of tropical species of Adhatoda and a sample of southern African species of Justicia.

History of the genera
Although the name was first used in pre-Linnaean literature {Tournefort 1700), the genus Adhatoda was first described by Miller {1754) and recognized as sfistinct from Justicia L. Two species of the so-called malabar nut, both tropical Asian woody species, were described, but the formal publication of specific binomials was delayed until Nees' account of the genus included the combination A. vasica, which we now propose as the lectotype of the genus. The original generic description included reference to the hooded flower and bilocular fruit with heart-shaped seeds. Duvernoia was described by Nees almost 100 years later {1847). Here Nees described D. adhatodoides, the type species, very fully and included anther details. In the same publication, Nees considered the genus Adhatoda but his concept of this was evidently much broader than that of Miller {1754) for he included within it the Linnaean genus Justicia and species now regarded as members of Monechma Hochst. and Siphonoglossa Oersted. Adhatoda sensu Nees was distinguished from Duvernoia by its tailed anthers and five-partite calyx, whereas Duvernoia anthers lack tails and the calyx was described as four-partite. Bentham & Hooker fiL (1876) ignored differences in anther appendages and amalgamated the two genera under Adhatoda. Lindau {1895), basing his generic concepts largely on pollen characters, reassorted the species into an expanded concept of Duvernoia, which included some species of Adhatoda sensu Nees, the balance being assigned to Justicia L. Clarke (1901) recognized two taxa, regarding them as subgenera of Adhatoda but remarked that its unusual calyx might be sufficient to support the recognition of the genus Duvernoia.
Brief references to these generic problems were made by Meeuse (1956) and Obermeyer {1962) but no detailed studies have been conducted recently. A re-examination of the generic 476 limits in this complex was therefore carried out to see if correlated features from diverse sources could replace the dependence on single-character classifications of the past. Nature of the pollen and of the anther thecae was not excluded, but was augmented by other data.
The study was based on southern African species of Duvernoia and Adhatoda and comparisons were drawn between these and some tropical species of Adhatoda, including representatives of the type species, A. vasica. In the absence of clear evidence of the type of Justicia , we used material of South African species currently regarded as true members of the genus (Clarke 1901) as a basis for this aspect of our comparisons. A list of the species selected is given in the Appendix.

Materials and Methods
Material for this study included herbarium specimens (see Appendix) and living material in cultivation locally or collected on field trips and preserved in 700Jo ethanol. This latter was particularly useful for studies of the inflorescences. Pollen was prepared for examination using a scanning electron microscope by shaking the pollen from dried anthers directly onto brass stubs covered with double-sided tape. Specimense were coated with gold-palladium and viewed in a Jeol 200X scanning electron microscope.

Characters appraised Inflorescence
The organization of the inflorescence in the Acanthaceae has often been neglected or described superficially. This is •, -A fr. Tydskr. Plantk ., 1985 , 51(6) particularly true of those members in which the inflorescences are congested and branching is partly concealed by bracts and bracteoles. Detailed description is particularly difficult when only herbarium ma~erial is available. Nevertheless, where inflorescence structure has been analysed, it has proved to be a useful taxonomic tool. Baden (1981) relied on this character to recognize sections within Anisotes Nees, Balkwill & Getliffe Norris (1985) found it valuable as a specific character in Hypoestes Soland. ex R. Br. and Immelman (1983) utilized it as one feature characterizing subdivisions within Justicia.
A synopsis of the inflorescence types was prepared after the dissection of each species and is presented in Figure 1.
A dichasium is assumed to be the ancestral form of these inflorescences and increase in the number of flowering nodes (Figure 1 A-C), is followed by the suppression of some of the higher order flowers. This is seen in the inflorescence of A. robusta C.B. Clarke (Figure lD

478
is formed in the axils of the bracteoles. In these highly derived spike-like thyrsiform inflorescences, the opposition of flowers and their multibracteate condition distinguish them from true spikes.
The species of Duvernoia are well characterized, and separated from others examined, by the abortion of the first order apical bud, and represent an offshoot from one of the main lines of inflorescence evolution. Their particular form of spike-like inflorescences is unique and apparently not matched in Adhatoda. In addition, many species of Adhatoda with strobilate spike-like inflorescences have very large, broadly ovate bracts (A. engleriana, A. schimperiana Nees, A. vasica) quite unlike those of Duvernoia.
Two highly modified derivatives of early stages in this proposed sequence are the inflorescences of J. bolusii C.B. Clarke ( Figure lBi) and J. campylostemon T. And.
The second trend is believed to involve the reduction of the dichasia to congested monochasia by the suppression of axillary buds, and the condensation of lateral axes. These reduced axillary inflorescences are aggregated into terminal synflorescences (AI -5). This line is associated with herbaceous perennials and thus is found in Adhatoda andromeda (Lindau) C.B. Clarke, A. densijlora (Hochst.) Manning as well as many species of Justicia.
Woody tropical species of Adhatoda have inflorescences of the sequence A-Fi, the exception being Adhatoda maculata C.B. Clarke ( Figure IAi). A. maculata also differed florally from other species examined and its position in the genus may prove to be incorrect on further study.

Floral features
These attractive flowers have a common basic floral organization which is illustrated in Figure 2.
The bracteate flower is markedly zygomorphic and bilabiate with a distinct corolla tube convoluted below (Figure 2Da-e) and expanding abruptly into a bifid upper rugulate lip and a tripartite lower lip with a distinct palate which may be prominently swollen. The palate is often faintly coloured and the veins deeply coloured, probably functioning as nectar guides.
The flowers are markedly protandrous, the style remaining enclosed within the rugula until after anthesis when it extends (Duvernoia) or arches downwards (Justicia) to expose the stigma. The flowers of Adhatoda and Justicia are also hercogamous in that the filaments reflex after anthesis carrying the anthers away from the palate or, rarely, tucking them into the corolla throat (J. protracta). In Duvernoia the anther filaments are restricted by their position in lateral rugulae but elongate conspicuously so that the anthers descend onto the lower lip. This alteration of the position of the anthers correlated with protandry must favour xenogamy.

(i) Calyx
Despite Nees's statement (1847) that the calyx of Duvernoia was four-partite, we found five-partite calyces in this and all the genera studied. There is, however, a distinction between the deeply divided calyces of Adhatoda and the campanulate calyces of Duvernoia which have lobes approximately as long as the tube.
(ii) Corolla Aspects of corolla symmetry and aestivation feature prominently in the subdivision of the family (Lindau 1895) and often serve to distinguish genera. Within the Justicieae the corolla is very uniform. In the course of this study, S.-Afr. Tydskr. Plantk., 1985 , 51(6) however, attention was drawn to the form and venation of the upper lip ( Figure 2C), the degree of subdivision and shape of the lobes of the lower lip and the nature of the palate. These subtle differences provided interesting taxonomic data.
The southern African and all but one tropical species of Adhatoda examined have markedly hooded, arched upper corolla lips without lateral channels, and with the lateral traces joining the marginal traces in the proximal half of the lip ( Figure 2C). The lower lip is deeply or shallowly divided with a prominent rugose, channelled palate restricted to the hypochile ( Figures 2B and 3B & C). Adhatoda macu/ata differed in that its palate was narrow and extended into the epichile and the lateral and marginal traces do not fuse ( Figure  3A). The upper lip of Duvernoia is shallowly hooded and possesses lateral channels in which the filaments are held (shaded in Figure 3E). The lateral traces of this lip fuse with the marginal traces very near to the apex ( Figure 3E). The lower lip is divided at least half way and the lateral lobes are falcate. A smooth palate is raised into two tumuli ( Figure 3E). This form of palate was also seen in J. campylostemon ( Figure  3F) but other features of the corolla did not match those of Duvernoia. No Justicia species examined had lateral channels, nor were the lateral and marginal traces free for more than half their length. Although frequently as deeply divided as in Duvernoia, the lower lip differed in that the lobes were obtuse or truncate and not falcate ( Figure 3D). Most southern African Justicia species have a rugose palate with conspicuous coloured veins ( Figure 3D).

(iii) Anther
Characteristics of the androecium feature prominently in the taxonomy of the Acanthaceae. Stamen number is used at the subfamily level (Bremekamp 196S); the number of anther thecae distinguishes genera such as Hypoestes (monothecate) from the allied genera Peristrophe and Dicliptera (bithecate); the relative heights of the anther thecae distinguish between Adhatoda and Duvernoia (Dyer 197S) and the presence of anther appendages characterizes Thunbergia and Hygrophila and is a feature of the Justicieae.
Once again a careful comparison of this feature revealed that it is not a constant characteristic of Justicia although when used with caution it may help to distinguish this genus from Duvernoia and Adhatoda. Variants on the anther form in Justicia are illustrated in Figure 4 (A-D).
The length of the tail on the lower theca in Justicia is extremely variable. Those of J. campylostemon are so short that the anther is scarcely tailed, but this species is anomalous in other respects too. In all of the Justicia species examined, however, the anther thecae are disposed slightly one above the other. Adhatoda anther thecae are also obliquely inserted on the connective and at least the lower theca is briefly tailed ( Figure 4E-G). The southern African species ( Figure 4E) are very similar, in this respect, to the tropical species including the type A . vasica. In Duvernoia both anther thecae are slightly elongated but scarcely tailed ( Figure 4H&I). The  character 'anther tailed' must therefore be used with caution.
(iv) Pollen Radlkofer (1883) was the first to distinguish different pollen types in this eurypalynous family and Lindau (189S), with almost fanatical zeal, applied differences in pollen in his classification of infra-familial groups. Pollen studies are therefore a prerequisite of any taxonomic study in the family and are undoubtedly of immense value, although perhaps not as useful as generic markers at the tribal level. Lindau (/.c.), probably as a result of the limited number of species examined, described Justicia pollen as tuberculate, but recent studies (lmmelman 1983) demonstrated that this is not invariably so. Indeed, most of the recent studied on pollen of acanthaceous genera have resulted in revisions of the original concepts of taxa within the family (De 1968;Baden 1981;Immelman 1983).
(v) Ovary, fruit and seed Hitherto unrecorded characteristics of the ovary, fruit and nectariferous disc lend weight to the opinion that Duvernoia is a separate natural unit. The nectariferous disc in the two species of this genus is conspicuously cupulate and encloses almost half the ovary. The rounded, finely downy ovaries ( Figure 6A&B) contrast markedly with the elongate, sparsely pubescent ovaries of the other species examined (Figure 6).
The nectariferous disc in species of Adhatoda is not as conspicuous ( Figure 6C-F). Justicia species have distinctive auriculate discs (Figure 60-J).
Within the Justicineae, stipitate fruits are the norm although in Justicia there is a series from long-stipitate through intermediates to subsessile capsules (Figure 7 A -D). The latter are not known in Duvernoia or Adhatoda and appear to be an apomorphic feature.
Crystals in the cells of the testa noted in Justicia by Immelman (1983) apparently do not occur in either Duvernoia or Adhatoda.
In Duvernoia the capsules have a stipe which is very long when compared with that in Adhatoda and a characteristic facies in dorsal view ( Figure 7F) which is not found in other long-stipitate ovaries. These differences are of degree only but support the close relationship between D. aconitijlora and D. adhatodoides and their separation from other genera examined.

Leaf epidermal features
The taxonomic and phylogenetic value of epidermal trichomes in the Acanthaceae have been demonstrated by many (Metcalfe & Chalk 1950;Singh & Jain 1975;Ahmad 1978 and references therein). In this study, we recognized three types of glandular and three types of eglandular trichomes on the leaves ( Figure SA-F). Most hair types were found in more than one of the genera under discussion, but type B trichomes S. Afr. J. Bot., 1985, 51(6)  characterized the genus Duvernoia. In addition, the leaf epidermal cells in Duvernoia were unusually small and had straight walls ( Figure 9A) while those of Adhatoda and Justicia were larger and had ripple walls ( Figure 9B).

Conclusions
The genus Duvernoia is reliably distinguished from related genera by its campanulate calyx which is unique in the Justiceae. We have found that this character correlates with several other features and suggest that Duvernoia can be defined as follows: (i) Small trees or strongly branching shrubs, (ii) inflorescences subverticillate, on nude peduncles, determinate by the abortion of the first order apical bud, (iii) bracts never subrotund nor imbricate and never partly concealing the flowers, (iv) calyx 5-fid, tube campanulate, (v) upper corolla lip with two lateral channels containing the filaments; the lateral veins anastomosing with the marginal traces in the distal third of the lip, (vi) lower corolla lip deeply lobed, lateral lobes falcate, (vii) palate of two smooth tumuli, (viii) anthers with thecae at slightly different heights and slightly divergent; thecae acute or equally acuminate at the base, (ix) pollen prolate bilateral, 2-colporate; margocolpi entire, (x) disc cup-shaped, not auriculate,   (xi) capsules long-stipitate, slender and truncate in dorsal view, pubescent, (xii) leaf epidermal cells with straight walls, and (xiii) leaf epidermal trichomes of type B and glandular trichomes with 8 or more head cells. The limits of Adhatoda are rather wide and a comprehensive survey over its entire range may well support segregates of at least sectional rank, if not separate genera. At present the genus seems to be characterized as follows: (i) Shrubs or perennial herbs with a woody rootstock, (ii) inflorescences axillary, open or congested, usually cymose, sometimes sessile and congested so as to become pseudo-terminal synflorescences, (iii) bracts in southern African species, linear-lanceolate and not concealing the flowers but in tropical species, large subrotund and partly concealing the flowers, S.-Afr. Tydskr. Plantk., 1985, 51 (6) (iv) calyx 5-partite almost to the base, (v) upper corolla lip deeply naviculate, rarely shallow, lateral veins anastomosing with the marginal traces within the proximal half of the lip, (vi) lower corolla lip fused almost to the tip or deeply 3-partite, lateral lobes arched, usually oblong and obtuse, rarely tapering, (vii) palate median, channelled, rugose, usually restricted to the hypochile, (viii) anthers with thecae at slightly different heights, slightly diverging, rounded or with the lower thecae tailed at the base, (ix) pollen prolate, 2-or 3-colporate, margocolpi entire or areolate, (x) ovary ovoid, variously pubescent; disc cupulate, not auriculate, (xi) capsule stalked, pubescent, (xii) leaf epidermal cells with ripple walls, and (xiii) leaf epidermal trichomes of type A and glandular trichomes with up to 8 head cells. Adhatoda vasica and its two close associates A. schimperiana and A. engleriana form a discrete group characterized by their woody habit, pedunculate, strobilate inflorescences ( Figure  IE&F), subrotund imbricate bracts and 2-aperturate areolate pollen. Further investigation may reveal that the genus Adhatoda should be restricted to species that share these distinctive features. The southern African species are perennial herbs, have synflorescences of variously reduced axillary cymes ( Figure IAI-5), linear lanceolate bracts, and 3-aperturate pollen with entire margocolpi. Until a revision of the genus throughout its range is complete, it would be premature to recommend that these species be removed from Adhatoda.
The origins of the members of this tribe are probably in tropical arborescent forms and Duvernoia and Adhatoda from tropical Africa probably represent early derivatives of this woody core. The southern African species of Adhatoda may represent the most highly derived members of their genus in that the arborescent habit has been replaced by a suffrutescent habit in which robust woody perennial rootstocks give rise to annual flushes of short flowering stems. Our observations over the past two growing seasons support the theory that this derived habit has evolved in response to fire, for unless the surrounding grass cover is removed, the annual flush of flowering shoots is not produced. Justicia species have perhaps diverged furthest from the ancestral woody stock as they invaded drier habitats and acquired an herbaceous habit, derived pollen, derived anthers and highly condensed inflorescences. An anomaly in this genus is the shrubby woodland species, J. campylostemon.
To sink Adhatoda and Duvernoia into Justicia would conceal what appears to be considerable evolutionary divergence.  In his synonymy of the species A. vasica, Nees (1832) included, amongst others, J. adhatoda L. and A. zeylanensium Herm. Both of these were based on a specimen in the Hermann Herbarium and the same material was the source of Miller's description of the genus (1754). It thus seems certain that the Hermann gatlrering numbered 16, is the type of the genus and that the earliest correct combination is Adhatoda vasica Nees. The Hermann Herbarium is now housed in the British Museum of Natural History.