Contribution to the taxonomy of Garcinia (Clusiaceae) in Africa, including two new species from Gabon and a key to the Lower Guinean species

Abstract Garcinia has some 260 species and is often regarded as a genus with a difficult taxonomy. No recent treatment is available for the botanically rich Lower Guinea phytogeographical region. This study aims at partly filling this gap. First, several taxonomic problems are solved. Garcinia chromocarpa is reduced to a variety of Garcinia quadrifaria. Garcinia gnetoides and Garcinia granulata are both synonyms of Garcinia quadrifaria. Garcinia zenkeri is a synonym of Garcinia densivenia and lectotypes are being designated for both names. Garcinia brevipedicellata is a synonym of Garcinia afzelii, as is Garcinia antidysenterica for which a lectotype is designated. Second, two new species endemic to Gabon are described: Garcinia gabonensis Sosef & Dauby and Garcinia obliqua Sosef & Dauby. Finally, an identification key to all species present in the Lower Guinea region is provided. A few remaining West African species names could not be placed with certainty, because the type material was lost or not traced yet. One is a Rutaceae while the remaining three are provisionally to be regarded as synonyms of Garcinia smeathmannii.


Introduction
genus received comparatively little attention from taxonomists because of its supposed complexity. Th is has resulted in an unsatisfying situation where taxonomic treatments have often only been focussing on a particular region (in Floras for example) and as a result contain various contradictions and errors. Th is obviously further feeding the idea that Garcinia is a complex and 'diffi cult' genus. Plant collectors, but also foresters, ecologists, etc., are often satisfi ed when they have concluded that their specimen belongs to the genus Garcinia and when they try to identify their material down to an individual species, they are bound to make many errors due to the unresolved taxonomic backbone.
In tropical Africa, the Clusiaceae (often in their old concept of Guttiferae) have been treated for West Africa (Hutchinson et al. 1954), the Congo Basin region (Bamps 1970a), East Africa (Bamps et al. 1978), and the Zambesian region (Robson 1960). However, no treatment exists yet for the notoriously richest Lower Guinean region (Sosef 1996, Küper et al. 2004). Many West African species extend their range to the east into Lower Guinea while many East and Central African ones do so to the west. Combined with a comparatively high local endemism in Lower Guinea, possibly caused by the presence of Pleistocene rain forest refugia (Sosef 1994, 1996, Linder 2001), this area counts as a diversity hotspot for many angiosperm families (Beentje et al. 1994, Barthlott et al. 1996, Küper et al. 2004, Linder et al. 2005. Besides that, it is also the region where taxonomic views and treatments of 'West' and 'East' meet and not seldom appear to disagree. During the preparation of a treatment of the Clusiaceae for the revitalized Flore du Gabon (Sosef and Florence 2007), several of these taxonomic 'problems' within Garcinia were tackled, and it was decided to slightly broaden the scope of that study. Th e resulting present article clarifi es the confusing taxonomy amongst several African members of the sections Xanthochymus (Roxb.) Pierre and Tagmanthera Pierre, describes two interesting new species from Gabon, provides an identifi cation key to the Garcinia's within the high diversity region of Lower Guinea, and hopes to draw attention to the need of more elaborate taxonomic studies in Garcinia.

Materials and methods
Th e authors studied the available Garcinia material in BM, BR, BRLU, K, L, LBV, MO, P and WAG (herbarium abbreviations follow Index Herbariorum, http://sweetgum.nybg.org/ih). Additional material was consulted through the JSTOR Plant Science website (http://plants.jstor.org). In 2008 and 2009, during tree plot inventories and general collecting undertaken in Gabon, the second author collected numerous Garcinia specimens and made many useful fi eld observations.
Th e conservation status of the two new species was assessed using the IUCN (2001) category criteria. Extent of occurrence and area of occupancy were estimated using Arcview 3.3 and Conservation assessment tools (Moat 2007).

The Garcinia quadrifaria complex
On the correct authors of the name Garcinia quadrifaria Th e species now known as Garcinia quadrifaria was fi rst described by Oliver (1868: 168) as Xanthochymus ? quadrifarius. It was transferred to Garcinia by Pierre (1883: 4) who seems to credit Baillon (1877: 404) for the name and also cites Oliver (1868: 168). Many sources (including IPNI, (http://www.ipni.org), Tropicos (http://www. tropicos.org), etc.) refer to G. quadrifaria with Baill. or Baill. ex Pierre as the authors of this combination. Actually, Baillon (1877) only states that Xanthochymus should be united with Garcinia, and refers to Oliver (1868), but does not make any new combinations. Th erefore, the correct author combination for G. quadrifaria is (Oliv.) Pierre.

On the distinction between G. quadrifaria and related species
Th e species G. quadrifaria belongs to the section Xanthochymus, characterized mainly by having staminal bundles with fi laments only partly fused and globose anthers (Jones 1980). G. quadrifaria shows several quite striking characters. Most obvious are the quadrangular and narrowly though distinctly winged twigs, the usually terminal infl orescence composed of a simple and short rachis set by 4 rows of overlapping scale-like bracts (somewhat reminiscent of an infl orescence of a Gnetum species), the 5-merous fl owers, a fruit with a verrucose exocarp and white latex. It shares these features with the following, until recently still recognized, species: G. chromocarpa Engl., G. gnetoides Hutch. & Dalziel, G. granulata Hutch. & Dalziel and G. le-testui Pellegr. In the past, two more species names, G. parva Spirlet and G. echirensis Pellegr., have been treated as synonyms of G. chromocarpa (Bamps, 1970a(Bamps, , 1970b. After careful examination of the characters of G. quadrifaria and G. chromocarpa, we have come to the conclusion that the only distinction between the two is the minute puberulence on the bracts, pedicels and fruits of G. chromocarpa, where those of G. quadrifaria are glabrous. Although at fi rst glance there might be a geographical distinction, G. chromocarpa in the Congo Basin, west to Gabon and Cameroun (Bamps 1970c), and G. quadrifaria in W.-Africa east to Cameroon and Gabon, this could not be upheld because of the presence of true G. quadrifaria in eastern DR Congo (North Kivu: Léonard 2337; Kivu: Gutzwiller 1868) and true G. chromocarpa in Ivory Coast (Breteler 6126,J.J. de Wilde & Leeuwenberg 3442). Also, when studying the literature (Engler 1908, Pellegrin 1959, Bamps 1970a, 1970b one might conclude that G. chromocarpa has 3-locular ovaries and 3-lobed stigmas, while G. quadrifaria would have 2-locular ones bearing 2-lobed stigmas. Although indeed all true (glabrous) G. quadrifaria specimens studied had 2-lobed stigmas, various true (puberulent) G. chromocarpa ones with 2-lobed stigmas (D.W.Th omas 4874, Walters & Niangadouma 1264, White 767 and others) were observed. Finally, from the literature it seems that fl owers of G. chromocarpa may have shorter pedicels than those of G. quadrifaria. Again, after proper examination of all the material available this turns out to be incorrect.
We therefore conclude that since the diff erences between the two taxa are minimal, they cannot be upheld as diff erent species. We do, however, want to distinguish them, and the level of variety seems most appropriate since there is no geographical separation. Because the name G. quadrifaria has priority over G. chromocarpa, this leads to the following new combination: Th e type material at K, Chevalier 15157, consists of a plant carrying terminal infl orescences with a dense mass of many racemes composed of a short rachis with closely set bracts which are glabrous. In G. quadrifaria each raceme normally appears solitary. An old note attached to the type already states it might well be a galled infl orescence, because a larva was observed inside. Hutchinson and Dalziel are well aware of its potentially diseased nature which shows from their remark in the Flora of West tropical Africa (Hutchinson and Dalziel 1927). However, in their more elaborate 1928 publication they do not mention the galled infl orescence. Th ey cite two other specimens apart from the type (Chevalier 15620 and Vigne 222) and these have normal, solitary racemes. Besides that, they cite Xanthochymus quadrifarius A.Chev. non Oliv. as belonging here. So, they were aware of the fact that this plant had some relation to that species, transferred to Garcinia by Pierre in 1883 (see above). Th eir plants had no fruits. In the same 1928 publication they describe another new species, G. granulata, citing a single specimen, Unwin & Smythe 58, that bears only fruits which are verrucose and glabrous.
After careful examination of all material at hand, we cannot conclude otherwise than that both G. gnetoides and G. granulata represent the same species known to us as G. quadrifaria and thus are synonyms of the latter. Th e fact that the fi rst two are in fact synonymous was already concluded by Hawthorne and Jongkind (2006). Some sources cite the publication of the names G. gnetoides and G. granulata in Kew Bulletin (Hutchinson and Dalziel 1928) as the place of valid publication, and not that of one year earlier in the Flora of West tropical Africa (Hutchinson and Dalziel 1927). In the latter publication the authors indeed seem to indicate the names will be formally published later on by adding "ined.", meaning ineditus (unpublished), behind "Kew Bull.". However, this does not render their 1927 publication invalid. Article 34.1b of the Code (McNeill et al. 2006) does not apply, because they do accept the taxa.
So, in conclusion, the new situation is as follows: Finally, G. le-testui, a rare species endemic to southern Cameroon and Gabon, seems suffi ciently distinct from G. quadrifaria being larger in most parts (notably wider wings on the twigs, larger leaves, longer pedicels, etc.). Most diff erences being related to size, it would not be surprising if G. le-testui turns out to be a polyploid of G. quadrifaria.
On the status of G. densivenia and G. zenkeri Two more species of the section Xanthochymus were described by Engler (1908): G. densivenia and G. zenkeri. Th e fi rst was based on two collections from Cameroon: Zenker 2397 (with fl owers) and Zenker 2547 (with fruits). Th e angular twigs and coriaceous leaves are reminiscent of G. quadrifaria. However, the fruit wall is smooth (not verrucose) and the infl orescence is both terminal and axillary and composed of bundles of very short racemes (up to 10 mm). Occasionally the infl orescence is composed of a branched raceme. In our view, this renders the taxon suffi ciently distinct to recognize it at species level. To date, no lectotype has been chosen from among the two syntypes, and because the fruit character is the most striking distinction, the best choice would be Zenker 2547. With the duplicate at B lost, we propose to select the duplicate at G, which seems to have the fi nest fruits, as the lectotype. G. zenkeri was also based on two Zenker collections from Cameroon: Zenker 1120 (with fl owers) and Zenker 3247 (with fruits). According to the African Plants Database (http://www.ville-ge.ch/musinfo/bd/cjb/africa), G. zenkeri would be a synonym of G. quadrifaria. Th is view is probably based upon Pellegrin (1959), who only cites the fi rst syntype. Studying material of both syntype collections (especially the fi rst with numerous duplicates in various herbaria), we indeed confi rm the presence of angular twigs and coriaceous leaves which point to G. quadrifaria. However, we also observe the presence of infl orescences composed of bundles of short racemes, often axillarily positioned, just as in G. densivenia. Th e only duplicate of Zenker 3247 (the syntype with fruits) available to us is sterile, and so we were unable to verify whether the fruit wall is smooth as in G. densivenia. On the other hand, Engler l.c. does not mention a verrucose structure of the fruit wall, a feature we believe he would certainly not have missed. We therefore render it most likely that the fruit he observed had a smooth wall. In all, we conclude that G. zenkeri and G. densivenia are synonyms. Both being published in the same publication, we may choose one of the names as being the valid one, and we have picked G. densivenia because the type material is better, showing all diagnostic features. G. zenkeri also needs a lectotypifi cation, for which we have taken Zenker 1120. Although all duplicates we have seen lack fl owers, it is by far the most widely distributed collection. Since, again, the material at B was lost and that at G seems the best among the remaining duplicates available, we have chosen that to be the lectotype. Th e above leads to the following situation: Th e material now identifi ed as G. densivenia, shows a remarkable variation in the distinctiveness of the tertiary venation. In the type as well as the paratype collection this venation is indeed, as the name indicates, strikingly dense and prominent. However, in most of the remaining material we observed a large continuous variation towards leaves where the tertiary venation was even hardly visible. Because otherwise, the material is quite uniform, we assume this character to be highly variable within the species, and possibly also depending on the way in which the material was treated and dried after collecting.

Garcinia densivenia
A second remarkable variation was observed in the shape of the fruits. Th ese can be subglobose to distinctly 5-lobed and 'pumpkin-like'. Th e lobed feature was even already mentioned by Engler in his protologue: "Baccae ….. leviter 5-lobae….." and "schwach 5 lappigen Früchte…..". However, again, we found no other characters to correlate with this feature. Moreover, the label of Bos 3639 specifi cally mentions that the shape of the fruits he collected are "globose to shallowly 5-lobed and resembling a pumpkin". We thus assume that these observations also illustrate within-species variation, and might be related to the number of seeds that develop within a single fruit.
The Garcinia afzelii complex Th e section Tragmanthera is characterized by 4-merous fl owers and staminal bundles that are completely fused carrying a row of ellipsoid anthers at their tip (Jones 1980). Four of its species, G. afzelii Engl., G. brevipedicellata (Bak.f.) Hutch. & Dalziel, G. lujae De Wild. and G. mannii Oliv., are closely related because they share a unique feature: the presence of anthers with locellate (septate) thecae. After studying the material, we noticed it is fairly easy to split it into two groups based on a leaf venation character: 1) G. afzelii and G. brevipedicellata having leaves with the lateral veins (3-)4-11 mm apart and gradually but distinctly curving up towards the margin and eventually running almost parallel to it, and 2) G. lujae and G. mannii with dense lateral veins, only 1-2(-3) mm apart, that run almost straight to the margin, or curving up just before it, where they join up in a blunt angle with an intramarginal vein running just inside of (at 0.5-1 mm) the actual margin. Bamps (1970b) already noticed that the latter two species are very closely related. Th e only remaining diff erences we could fi nd is the colour of the petals (red to deep orange or sometimes yellow in G. mannii and always yellow in G. lujae) and the length of the staminal bundles (as long as the pistillode in G. mannii and clearly overtopping the pistillode in G. lujae). Although these distinctions are weak, there is no geographical overlap between the two taxa, G. lujae in the Democratic Republic of the Congo and G mannii from southern Nigeria to Gabon, which made us decide to uphold them at species level, but further investigations are needed.
However, we could not fi nd any clear diff erence between the remaining two species, G. afzelii and G. brevipedicellata. Hutchinson et al. (1954) give the density of the lateral veins (more dense in G. afzelii) as the only diff erence, but we have clearly observed this feature to be variable, even within a single specimen. Also, the name of G. brevipedicellata suggests it has short pedicels (although this taxon was originally described as a variety of G. mannii and so this would be a diff erence with that taxon), but again, those in G. afzelii are quite variable and we observed a continuous variation. Finally, there might be a diff erence in fl ower colour, since the petals of G. afzellii are reported to be yellow to pale green, while label data indicates the fl owers of G. brevipedicellata are "yellow with a red centre". Th e latter observation is indeed correct, because the petals are yellow and the pistillode or ovary has an orange to red colour. Both species have staminal bundles that are longer than the ovary. So, none of the presumed diagnostic characters can be confi rmed and we conclude that both names refer to the same species. Since G. afzelii is the older one, that is the accepted name for the taxon.
Finally, the status of the name G. antidysenterica A.Chev. is unclear. At present, it is regarded as a synonym of G. afzelii (Hutchinson et al. 1954, Pellegrin 1959, African Plants Database at http://www.ville-ge.ch/musinfo/bd/cjb/africa). Th e protologue text provides information on three collecting localities referring to Chevalier's own specimens: "entre Nze et Danané", "entre le Morénou et l'Indénie" and "bassin du moyen Comoé". No holotype is designated, but the protologue is accompanied by a photo of one of the sheets: Chevalier 21200, collected at "Danané", and so this seems to be the most logical choice for the lectotype. However, this collection (duplicates at P and K) turns out to be G. epunctata Stapf, as was already discovered by Bamps (1969: 364). While the protologue clearly states that the anthers have locellate thecae, those of G. epunctata and hence of Chevalier 21200 are not. Th erefore, choosing the latter collection as the holotype should be avoided, following Article 9.17 of the Code (McNeils et al. 2006; lectotype in serious confl ict with the protologue while other elements are available). Chevalier (1920) provides a list of specimens which according to him belong to G. antidysenterica. Th e only other collection that carries one of the remaining two localities, "vallée du Moyen-Comoé", is Chevalier 22571. Th erefore, this seems the most obvious choice for the lectotype. It has at least three duplicates (BR, K (2x) and P) and the specimen at P should be regarded as the most original material. At least the BR and K duplicate carry an original label with the name Garcinia antidysenterica A.Chev. (Th e P material was seen at an early stage of the project, but due to the closure of the Paris herbarium the exact label data could not be checked.) Th e above now leads to the following situation: Garcinia afzelii Engl., Bot. Jahrb. Syst. 40: 570 (1908 Within the section Tragmanthera some taxonomic questions remain to be solved, for example the distinction between the non-locellate species such as G. epunctata Stapf and G. preussii Engl., and the status of several other names now regarded as their synonyms. For now, we maintain the present status quo, just signalizing the need for a more in-depth study.

Two new endemic Garcinia species from Gabon
During the preparation of the Clusiaceae treatment for Flore du Gabon, material belonging to two new species turned up. Both are endemic to this country that has a plant endemism rate of ca. 11% (Sosef et al. 2006). Gabon is notoriously rich in species (see above), especially its lowland rain forest is reputedly the most species-rich in tropical Africa (Breteler 1990, Linder et al. 2005, Reitsma 1988) and novelties are still regularly discovered (Bissiengou and Sosef 2008, Ntore et al. 2010, Walters et al. 2011).
Distribution. Endemic to central Gabon, known only from the Ogooué-Ivindo and Ngounié provinces (see Figure 3a).
Habitat and ecology. Primary terra fi rme rainforest; at ca. 450-800 m altitude. Flowering in April, fruits observed in February.
Conservation status. G.obliqua is currently known from nine collections and six locations. Estimates of its extent of occurrence and area of occupancy are respectively ca. 9488 km 2 and 80 km 2 . One collection (Dauby et al. 1650) corresponded to a dead individual along a forestry road and all locations are currently found within logging concessions. Hence, we assume the extent of occurrence, area of occupancy, quality of habitat and number of sub-populations will decrease in the near future. We therefore assign a preliminary status of Vulnerable (Vu B1ab(i,ii,iii,iv,v)+B2ab(i,ii,iii,iv,v)).
Notes. Th e shape of the staminal bundles and the anthers, as well as the distinct foveola point to a relationship with G. smeathmannii and G. ovalifolia, both belonging to the section Rheediopsis Pierre (Jones, 1980). On the other hand, preliminary molecular data obtained by the second author suggest that G. obliqua is not related to these species. Th erefore, the results of an upcoming molecular study are awaited before a fi rm statement about the sectional position can be made.
Distribution. Endemic to southern and central Gabon, in the provinces of Moyen-Ogooué, Ngounié and Ogooué-Maritime (see Figure 3b).
Habitat and ecology. Primary or late secondary terra fi rme rain forest, along rivers or on ridges; at ca. 150-850 m altitude. Flowering in September to November, fruiting in September, November, December and February.
Conservation status. Currently, G. gabonensis is known from eleven collections and nine locations. Estimates of the extent of occurrence and the area of occupancy are respectively ca. 16 000 km 2 and 109 km 2 . Since nine of the eleven collections are within logging concessions or along main roads, we consider that continuing decline  in the extent of occurrence, area of occupancy, quality of habitat and number of subpopulations has occurred or will occur in the near future. We therefore assign a preliminary status of Vulnerable (VU B2ab(i,ii,iii,iv)).
Notes. For now, it remains unclear as to which section this species belongs. Most striking feature are the free stamens. According to the elaborate work of Jones (1980), this is characteristic for only two sections: sect. Teracentrum Pierre and sect. Rheedia sensu Jones. Species belonging to the fi rst, however, have their stamens inserted across a central mass and lack a pistillode. Th ose belonging to the second are known to occur, until now, only in Central and South America and on Madagascar…. Morphologically though, G. gabonensis seems very similar to other species in sect. Rheedia, and this might be the fi rst continental African representative of that section. A molecular study to further investigate this will be performed soon.
Additional specimens examined (all from Gabon). SE of Sindara, km 12 from Camp Chantier Waka to Ngounié River,

Key to the Lower Guinean species of Garcinia
For most of the continental tropical African regions, an identifi cation key to the species of Garcinia exists: Hutchinson et al. (1954) for West Africa, Bamps (1970a) for Central Africa, Bamps et al. (1978) for East Africa, and Robson (1961) for the Zambesian region. Th e lack of such a key for the notoriously species-rich Lower Guinean region has hindered proper identifi cation of specimens; a shortcoming we hope to overcome by providing the key below. Th e protologue states there are 20-30 stamens in 4 bundles (so some 5-8(-9) per bundle), fused until halfway, and leaves similar to G. ovalifolia, but with less pronounced veins and a cuneate base, on 1-1.5 cm long petioles. Flowers are positioned in glomerules on the twigs, below the leaves. Th is description fi ts that of G. smeathmannii and we momentarily place this name under that species.
Th e protologue states there are 30-40 stamens in 4 bundles, fused almost to the top, a distinct foveola, leaves with 12-15 lateral veins and fl owers in many-fl owered bundles on the nodes. Th e sketch at BM shows large leaves (12-19 × 3.5-8 cm) with an acute apex and fl owers on pedicels of 1.5-2 cm. All this fi ts G. smeathmannii best, and for the moment we regard it as a synonym of that species. Th e protologue states the material concerns a fairly large tree (18-22 m), with twigs soon rounded, rugose petioles of 1.5-2 cm long with a distinct foveola, leaf blade coriaceous and shiny, male fl owers 4-7 in a fascicle, pedicel 2.5 cm, white petals, stamens 20, in 4 bundles, fused to halfway, alternating with verrucose disc lobes. Th e sketch at BM shows large elliptic-obovate leaves with slightly acuminate apex and a fasciculate infl orescence with three fl owers on 22-28 mm long pedicels.

Garcinia laurifolia
Again, this description, as well as the sketch, fi t G. smeathmannii and for now we regard it as a synonym of that species.