Studies in the Justicia and Siphonoglossa ( Acanthaceae ) species of southern Africa : final conclusions

On the basis of seed surface micromorphology, pollen and inflorescence type, as well as characters of the capsule, seven sections can be distinguished in the southern African species of Justicia. The relationship of these to the sections recognised by Graham is discussed. A key to the sections and a synopsis of their main distinguishing characters are presented. One new section, sect. M inima lmmelman, is described. A short note on the generic position of the Siphonoglossa species of the region, following on from a previous paper, is also given.


INTRODUCTION
In the preceding four papers on the southern African species of Justicia and Siphonoglossa, the various states of indumentum, pollen and seed surface were described.Their possible value both in identifying species and delimiting sections within Justicia, was discussed.One of the points made throughout was that none of these organs could be used in isolation in the delimiting of sections.
The species studied were chosen on the basis of their geographical distribution rather than as being (putatively) representative of the whole genus.As was stated by Graham (1988), small sample size created difficulties in the delimitation both of sections and of the genus Justicia.She studied 58 species in detail, but concluded that 'It is over-optimistic to hope that all the remaining known species of Justicia will fit neatly into the sectional classification proposed here -However, it is hoped that this study will provide the basis for future work.'The present study of the southern African representatives of Justicia and the closely related genus Siphonoglossa used Graham's work as a starting point for discussion, since it gives a worldwide perspective on the genus.Graham's sectional treatment will be discussed in the light of results obtained from the detailed study of the ± 30 southern African taxa traditionally included in Justicia and Siphono glossa.
Some confusion surrounds the designation of a type species in Justicia.A summary of the problems involved was given by Stearn (1971), and the problem is also dealt with by Graham (1988).She considered the type species to be J. hyssopifolia L., as did Hedrén (1989), and this decision is followed here.Seed testa surface, indumentum and pollen were examined under the Scanning Electron Microscope.Inflorescence structure was also examined.Various charac ters of leaves, bracts, calyx, corolla, capsules and habit were taken into consideration in the conclusions reached.Nomenclature of the sections was taken from Graham (1988), and her sectional synonomy and type species are not repeated here.

RESULTS AND DISCUSSION
In this series of studies, striking correlations between inflorescence form, seed testa morphology and pollen were found in the genus.Largely on this basis, the species studied could be grouped into seven sections, one of them new.
The primitive inflorescence is considered to probably be a thyrse which has been reduced in various ways.Where it has been reduced to sessile cymes/single flowers, these are in some cases re-aggregated to form a more or less dense, terminal synflorescence.
The pollen in the genus is three-or two-colporate or two-porate, with entire or areolate margocolpi.The sexine is reticulate to lophate in all species, with a pseudocolpus (colpoid streak) on either side of the colpus (lmmelman 1989a).The area between the colpus and the pseudocolpus (the margocolpus, see lmmelman 1989a) is either entire or divided into circular areolae.In the southern African taxa, as well as in seven tropical African species examined later, the inflorescence and pollen types are closely correlated and valuable in dividing the genus into sections.
Seed surface, described simply as 'rough' in earlier accounts of the genus, is one of the characters that is used to distinguish Justicia from the related genus Monechma Hochst.(Munday 1980).It is, however, very varied in Justicia, and under the SEM reveals a number of useful characters, a fact also noted by Graham (1988).In a number of cases seed surface also correlates with the sections, which lends weight to their naturalness.It is considered unlikely that a number of characters, taken from three different organs, viz.pollen, inflorescence and seeds, would have evolved in parallel.
A number of other characters were considered, and some were found to reinforce these divisions, especially capsule texture.Capsule shape is mainly useful at generic level.Munday (1980) considered a clavate capsule charac teristic of Monechma, as it was constant in the genus, being correlated with the two-seeded condition.Some Justicia species (also two-seeded) also have this type of capsule, but in most cases it can be used to distinguish the two genera.(It is possible that Monechma should be considered congeneric with Justicia, as no one character or combination of characters could be found to completely separate the two genera, but this will not be discussed here.) The capsule of Justicia is usually cylindrical-clavate with an apiculate apex and a short or long stipe.The stipe is laterally flattened.The fertile portion contains four seeds, sometimes reduced to two or one, each subtended by a hook-shaped retinaculum.The texture of the capsule varies from woody and thick-walled (sects.Raphidospora, Justicia) or papery and thin-walled (sects.Tyloglossa, Minima), most species of sect.Hamieria and sect.Ansellia.Intermediate between these two extremes are the capsules of, for example, sect.Betonica and some species of sect.Hamieria.
As stated in the introduction, the sections were proposed on the basis of a study of southern African material only.However, Justicia is probably a tropical genus, with southern Africa being at the southern end of its distribu tion range in Africa and it was therefore considered important to assess whether the apparent pattern seen in the southern African species also existed north of the Cunene and Limpopo Rivers.As a test, therefore, seven tropical species (listed in lmmelman 1989a) were selected from a number of the sections, their probable section being first assessed by examining the type of inflorescence present.The pollen of each was then viewed under the SEM and it was found that in every case it was of the pollen type expected for the section in which they had provision ally been placed (lmmelman 1989a).It was not, unfor tunately, possible to examine seeds of the species.The inflorescence and pollen are considered in most respects to be the most primitive in southern Africa.The seeds of the two southern African species, however, differ greatly and, in J. glabra at least, may be considered highly advanced, as they are covered with long barbed scales (lmmelman 1990a).
All except one species of Justicia seen have anthers with the lower locule spurred, J. campylostemon being an occasional exception.A few specimens were seen where the anther locules were not spurred, but only mucronate.This is the regular condition in Siphonoglossa.However, as this is variable within J. campylostemon, and its other characters are typical of Justicia, it was not considered a species of Siphonoglossa.J. bolusii is probably most similar to some East African species of Justicia, e.g.J. cordata (Nees) T. Anders., a common species in East Africa.J. bolusii and J. cordata are very similar in appearance, both have a pouched throat in the corolla, a clavate capsule and a similar inflorescence, in addition to the unusual smooth seeds.Graham (1988) states that the pollen of J. cordata is 2-colporate, but under the SEM, it was seen to be 3-colporate as in J. bolusii (Immelman 1989a).It is reasonable to surmise that the section to which J. bolusii and J. cordata belong was once more widely distributed in Africa, J. bolusii being a relict of this distribution.
The pouched corolla throat of J. bolusii and J. cordata is like that described for Monechma (Munday 1980) and so are the 1-or 2-seeded capsules and smooth-surfaced seeds.It was therefore considered whether J. bolusii should not be better placed in that genus than in Justicia, but the inflorescence and pollen are not like those of Monechma.Monechma usually has sessile cymes or single sessile or pedicellate flowers and has two-colporate areolate pollen, whereas J. bolusii has inflorescences as described above and 3-colporate pollen with entire margocolpi.The one or two smooth seeds and pouched corolla throat are considered to be a parallel adaptation in Monechma, rather than indicating relationship between it and J. bolusii.
Sect.Justicia sensu Graham (1988) also includes J. orchioides L.f. and J. platysepala (S.Moore) P.G.Mey.This has been accepted here with some hesitation, as it is felt that differences exist.However, the whole of sect.Justicia is in need of revision, which would have to includc more tropical African species.Graham states that J. orchioides and J. platysepala have 'type 4' pollen, i.e. 2-porate with two raised bands (entire margocolpi), which applies to these species but not to J. bolusii or J. cordata, which are both 3-colporate.These two groups do not resemble each other in inflorescence or seed testa either.It is possible to place them in separate sections.
Within the second group (J.orchioides, J. thymifolia, J. platysepala and J. guerkeana) there is a tendency to increased woodiness and to a reduction in the size of the bracts and the number of flowers in the cyme, the three trends progressing together.J. guerkeana and J. platysepala have numerous flowers in the cyme, broad bracts with green centres and white membranous margins, and rather soft shrubby stems, whereas J. thymifolia is woody, has small bracts and still has multiflowered cymes.J. orchioides and J. cuneata invariably have small bracts, solitary axillary flowers and are hard-stemmed (in J. orchioides sometimes spiny) shrublets.It is not suggested that any of these is the direct ancestor of any other species.
The common and widespread J. betonica (South Africa to India) and the narrowly endemic J. montis-salinarum (Soutpansberg of the Transvaal) have pollen similar to that of sect.Raphidospora, being 3-colporate with entire margocolpi, but the cymes have been reduced to 1 (rarely 3) axillary flower(s) each.These flowers have been secon darily aggregated to form a dense, terminal 'spike'.In J. betonica there are large, colourful, imbricate pherophylls and prophylls, wherease in J. montis-salinarum there apears to have been even further reduction, with both pherophylls and prophylls being small and subulate.In both species, the seed testa is drawn into sinuate, dentate structures, which in J. betonica have a micro-structure of colliculate individual cells and in J. montis-salinarum are finely striate.A sinuate or dentate testa was not seen in any other species in the genus.Although the two species placed in this section are very different superficially, their similar pollen and inflorescence structure, and their dis tinctive seed testa, indicate they should be placed together.

IV.
Sect.Tyloglossa (Hochst.)Lindau (sect.IV in Graham: 390).Species included: J. petiolaris (Nees) T. Anders, (with three subspecies), J. flava (Vahl) Vahl and J. kirkiana T. Anders.J. minima is a rare endemic from the Waterberg of the Transvaal.It is not here considered closely related to any of the southern African species, and its affinities probably lie north of the South African border.Its pollen is similar to, though much smaller than that of sect.Tyloglossa (above), but its inflorescence is a terminal, unbranched, lax 'spike' (actually a reduced series of cymes forming a terminal florescence) similar to that of sect.Ansellia.Both pherophylls and prophylls are greatly reduced, and the calyx is four-lobed.reticulate micro-pattern of the seed surface did not aid in placing the seed surface, as the reticulate pattern is scattered in the genus, e.g.J. campylostemon and J. capensis Thunb., and therefore was not considered significant.This species does not key out in Graham's key to sections, the nearest being sect.Rostellaria subsect.Anisostachya (if the bracts are ignored) or subsect.Ansellia (if the number of calyx lobes, the pollen and the details of the seed are ignored).After comparing the South African species with his descriptions of the South American taxa, I would agree that they do not belong together in a common genus, separate from Justicia.Rather the South African (and other African) taxa should be placed either in a section/sub section of Justicia (the latter being the course followed by Hedrén), or in a genus with a name other than Siphonoglossa, the type species of Siphonoglossa being a South American species.Aulojusticia, which already exists, would be a suitable name.Before either of these is done, however, it would be necessary to compare the South African species with the South American species and the tropical African species placed in Justicia This well-defined section has inflorescence and flowers resembling those of the sect.Minima, but the inflores cences are axillary not terminal, the calyx is five-not four-lobed and the pollen is quite different.The two upper (adaxial) lobes of the calyx are noticeably longer than the three lower (abaxial) lobes.The pollen, which is the smallest in the genus in southern Africa, is areolate and either 2-porate or shortly 2-colporate.In J. crassiradix there were short colpi present and it is considered probable that the porate condition seen in J. anagalloides and J. anselliana is derived from the colporate by gradual reduction of the colpus length.

VI
Justicia and Siphonoglossa were studied, and sections elucidated.The area termed 'southern Africa' is that defined by the Flora of southern Africa, i.e. the area south of the Cunene and Limpopo Rivers, including Botswana.Dried material was examined from the following herbaria: BOL, GRA, KMG, KNP, NBG, NH, NU, SAM, STE and WIND.Living material was collected on trips to the northern Transvaal, northern Natal and Zululand, the eastern Cape and the north-central part of Namibia.Some of this material was grown in the nursery of the National Botanical Institute, Pretoria.
From the three-colporate pollen with entire margo colpi and the laxly branching inflorescence of sect.Raphidospora, a number of lines of specialization can be traced.Inflorescenceof stout-peduncled axillary elongated thyrses or nearly sessile or flowers solitary, pedicelled or sessile.Pherophylls and prophylls reduced and subulate or large and lanceolate to broadly obovate.Flowers hooded or upper line straight, with or without pouched throat.Pollen two-or three-colporate, with entire margocolpi.Capsule relatively large, hard and wcxxly, with a long stipe.Seeds one (two), smooth or rough.Shrubby, in J. bolusii with suckers from stems below ground.J. bolusii is a rare eastern Cape endemic, and in flowers, seeds and pollen, differs from other species of the section in the southern African region.The pollen type is the same as sect.Raphidospora.J. bolusii and J. campylostemon were formerly placed together in sect.Gendarussa by Clarke (1901) (= sect.Raphidospora), but J. bolusii differs from both J. campylostemon and J. glabra in the single smooth seed, the pouched corolla throat and the un branched inflorescence.
terminal, a dense or lax aggregate of sessile axillary cymes.Pherophylls and prophylls reduced, oblanceolate to obovate.Pollen 3-colporate, areolate.Capsule of medium size, texture papery, stipe short.Seeds 4, reticulate, with crystals in testa.This is a well-defined section, one of the major unifying characters of the section being the seeds, which under the SEM are seen to have one to many cubic or rectangular crystals in each cell of the testa.This character is unique to the section.The pollen is 3-colporate and areolate.The inflorescence in the section is basically composed of sessile cymes, as in the preceding section, but these arc aggregated into a terminal compound 'spike' with numerous flowers at each node.The pherophylls and prophylls are somewhat reduced, usually spatulate and often with very long-stalked glandular hairs (lmmelman 1990b).Though the inflorescence type of J. petiolaris subsp.incerta and subsp.bowiei tends towards that of the sect.Hamieria, being more open than in subsp.petiolaris, the pollen and seeds are quite different from those of sect.Hamieria.V. Sect.Minima lmmelman, sect.nov.Type species: J. minima A. Meeuse.Inflorescentia terminalis elongata simplex spiciformis e cymis flore solitario redactis composita.Pherophylla prophyllaque redacta subulata.Calyx quadrilobatus.Pollinis grana tricolporata areolata.Semina quatuor, testa reticulata.Inflorescence a terminal, elongate, unbranched spike like aggregate of cymes reduced to a single flower each.Pherophylls and prophylls reduced and subulate.Calyx four-lobed.Pollen 3-colporate, areolate.Seeds 4, testa reticulate.
. Sect.Hamieria (Solms-Laub.)Benth.(sect.V in Graham: 591).Taxa included: J. capensis Thunb., J. protracta (Nees) T. Anders, (with two subspecies), J. heterocarpa subsp.dinteri (S.Moore) Hedrén, J. parxibracteata lmmelman and J. odora (Forssk.)Vahl.A very complete synonymy for the section is given inHedrén:  62 (1989).Inflorescence of scattered sessile axillary cymes, or these reduced to a single flower.Pherophylls and prophylls foliose, bracteoles reduced and subulate or absent.Pollen 2-colporate, areolate.Capsule of medium size, texture papery, stipe short or, in one species, hard and woody with long stipe (J.cuneata) with J. odora and J. capensis intermediate.Seeds 4, or only one in the dimorphic capsules of J. heterocarpa subsp.dinteri and (rarely) J. protracta subsp.protracta.Seed surface varied, usually papillate or reticulate.Both pollen and inflorescence in this section are specialized with respect to the condition seen in sect.Raphidospora, the pollen being 2-colporate and areolate and the inflorescence reduced to pedunculate or sessile axillary cymes with up to five flowers, scattered in the leaf axils.The seeds of four of the five species in the section have a micro-pattern of a single papilla in the centre of each cell of the testa, but in J. capensis the seed testa is reticulate.I would agree with Hedrén (1989) in considering J. capensis and J. odora as belonging in sect.Hamieria, while being in some respects very different from the majority of species in the section.A feature of interest in this section is the dimorphic capsules that are present in two members, regularly in J. heterocarpa subsp.dinteri and encountered once in J. protracta subsp.protracta.These are also occasionally seen in Siphonoglossa leptantha (Nees) lmmelman subsp.leptantha.Specimens of J. protracta and S. leptantha showing this condition are preserved at PRE and NU (Balkwill 217 (NU), Balkwill 649 (NU, PRE), lmmelman s.n.(630291 in PRE)).As well as the normal four-seeded capsule, a smaller, one-seeded, non-opening (indehiscent) capsule is produced, with four irregularly serrate wings.The seed does not differ from those in the normal capsules except in being larger (lmmelman 1990a).This was also seen by Hedrén (1989) in a number of tropical African taxa in sect.Hamieria.The possession in common of a specialised type of capsule is evidence of a close relation ship between Justicia sect.Hamieria and Siphonoglossa.The seed of Siphonoglossa, like that of most of the species of sect.Hamieria, is papillate, the pollen is simi lar, and its inflorescence is also of many-flowered scattered sessile axillary cymes.In Hedrén (1989), the J. striolata species group of sect.Hamieria corresponds to Siphono glossa as recognised by me (non Oerst, lmmelman 1989b), and he has placed Siphonoglossa sensu Moore non Oerst. in the synonymy of sect.Hamieria.He found it to separate rather widely from the rest of the section on MDS ordination, but nevertheless retained it in Justicia sect.Hamieria.Though I gave it as my opinion (lmmelman 1989b) that there are reasons for not combining the two genera, if Siphonoglossa sensu Moore is placed in Justicia, I feel it might best be as a separate section close to sect.Hamieria.Note on the genus Siphonoglossa: when I published my paper on Siphonoglossa and Aulojusticia (lmmelman 1989b), I was not aware of the existence of a series of papers by Henrickson & Hilsenbeck (1979) and Hilsenbeck (1990a, 1990b) on the South American representatives of Siphonoglossa.Hilsenbeck's opinion is that there are three, or possibly four diverse taxa included under the name Siphonoglossa, which are held together mainly by having in common a long narrow corolla tube (Hilsenbeck 1990b).