Blackwell Science, LtdOxford, UKBOJBotanical Journal of the Linnean Society0024-4074The Linnean Society of London, 2004? 2004 1463 257283 Original Article PHYLOGENY OF VANGUERIEAE H. LANTZ and B. BREMER Botanical Journal of the Linnean Society, 2004, 146, 257–283. With 5 figures Phylogeny inferred from morphology and DNA data: characterizing well-supported groups in Vanguerieae (Rubiaceae) HENRIK LANTZ1* and BIRGITTA BREMER1,2 1 Department of Systematic Botany, Uppsala University, Evolutionary Biology Centre, Norbyvägen 18D, SE-75236, Uppsala, Sweden 2 The Bergius Foundation at the Royal Swedish Academy of Sciences, PO Box 50017, SE-10405, Stockholm, Sweden Received October 2003; accepted for publication June 2004 The Vanguerieae is a species-rich tribe of the Rubiaceae with problematic generic circumscriptions. Here we address this issue by presenting a phylogeny in which strongly supported groups are identified. We use DNA sequence data from the nuclear ribosomal internal transcribed spacer (ITS) region, trnT-F sequences from the chloroplast genome, and 30 morphological characters. Over 70 taxa are sampled, representing 23 out of the 27 currently recognized genera. We also present a detailed discussion of the morphological variation present in the tribe, in which we focus on features that are taxonomically informative. We show that there are several strongly supported groups in the tribe, but that these rarely coincide with traditional genera. Canthium, Rytigynia, Tapiphyllum and possibly Pygmaeothamnus are polyphyletic, and Fadogia is paraphyletic. Keetia, Lagynias, and Multidentia are monophyletic with strong support, while the monophyly of Psydrax is weakly supported. Several morphological characters are mapped onto the phylogeny to visualize how these can be used to delimit monophyletic groups. Canthium subgenus Afrocanthium is given generic rank as Afrocanthium and 17 combinations for species in this genus are made. New combinations are also made for Canthium sensu stricto. © 2004 The Linnean Society of London, Botanical Journal of the Linnean Society, 2004, 146, 257–283. ADDITIONAL KEYWORDS: Afrocanthium – Canthium – classification – morphology. INTRODUCTION Rubiaceae is a species-rich family containing many taxonomically difficult groups at various taxonomic levels. Examples include Psychotria, one of the largest angiosperm genera (e.g. Nepokroeff, Bremer & Sytsma, 1999), the Hedyotis–Houstonia–Oldenlandia complex (e.g. Terrell, 2001), and the tribe Vanguerieae is clearly also one such group. Not identified or incorrectly determined Vanguerieae specimens are common in many herbaria. The tribe is recognized by a combination of characters, i.e. axillary inflorescences, valvate corolla aestivation, ovaries with solitary pendulous ovules, and most importantly a special kind of pollen presenter, a ‘stigmatic head complex’ (Igersheim, 1993), used in secondary pollen presentation. *Corresponding author. E-mail: henrik.lantz@ebc.uu.se This structure is a unique synapomorphy for the tribe and consists of a combination of the stigmatic lobes and surrounding tissue and usually looks like a globose to cylindrical knob on the apex of the style. While many other Rubiaceae species have secondary pollen presentation (Puff et al., 1996), none have the very specialized structure of Vanguerieae. This synapomorphy and the above-mentioned combination of characters makes the tribe easy to recognize as a group, but unclear generic delimitations and a profusion of species make it difficult to identify a specimen to genus and species levels. The tribe Vanguerieae contains 27 genera and about 600 currently recognized species, but this number is likely to change in the future. Many species remain to be described, especially from Madagascar and southeast Asia, and several already described are poorly known. So far, species from eastern and southern © 2004 The Linnean Society of London, Botanical Journal of the Linnean Society, 2004, 146, 257–283 257 258 H. LANTZ and B. BREMER Africa have been investigated most carefully, but the distribution of the tribe extends over all areas in Africa south of the Sahara into Madagascar, south and south-east Asia, the Pacific and Australia. Like most tropical Rubiaceae, Vanguerieae species are usually trees or shrubs. The most common exception from this type is the geofrutescent habit that Fadogia, Fadogiella and some other taxa share, but there are also lianescent shrubs, e.g. Keetia and Psydrax subgenus Phallaria. Flowers are usually small and inconspicuous with green to white corollas most common. Notable exceptions are found in the genera Ancylanthos and Temnocalyx, and in a few members of Fadogia that have £4-cm-long corollas in bright yellow and red colours. Few genera are delimited in such a way that they are easily recognized and some genera grade into each other, e.g. Pachystigma and Tapiphyllum (Bridson, 1996). Several of the characters used to delimit genera are unfortunately often continuous in nature, e.g. amount of indumentum and length of calyx lobes, or characters are shared by many genera, e.g. plurilocular ovaries and inflorescences fasciculate. This complicates the identification of Vanguerieae specimens. Without a better knowledge of the morphological variation, the circumscription of many species and genera will continue to be uncertain. Vanguerieae, with the exception of the large and widely spread genus Canthium and the West African genus Cuviera, was revised by Robyns (1928). In several cases the generic definitions published there still remain, although they are not unquestioned. While Ancylanthos (Bridson, 1996) and Temnocalyx (Verdcourt, 1981) have been reduced to the type species, only small changes have been implemented in other genera such as Pachystigma and Tapiphyllum, although they have been suggested to be weakly delimited (Verdcourt, 1981). Cuviera was later revised (Hallé, 1959), as were the African members of Canthium (Bridson, 1985, 1986, 1987a, b, 1992), of which the majority were transferred to Keetia, Multidentia, Psydrax and Pyrostria. Four subgenera of Canthium are now recognized (e.g. Bridson, 1987b, 1992), but the majority of the Asiatic Canthium species have not been formally referred to any of these subgenera or any other genus (but see comments in Bridson, 1987b, 1992; Wong, 1989). In the present study, we perform a combined phylogenetic analysis of one morphological and two molecular datasets in order to generate a well-resolved phylogeny in which strongly supported groups can be identified. Such groups can be used as the basis for an improved classification of the tribe. Of the two DNA regions that are sequenced, one is nuclear and one is from the chloroplast. The nuclear genome is represented by Internal Transcribed Spacer (ITS) sequences (cf. White et al., 1990) from the ribosomal DNA region, the chloroplast by the trnT-F region (cf. Taberlet et al., 1991). We also provide an overview of the morphological variation present in Vanguerieae in a phylogenetic context, to clarify which morphological characters can be used to delimit monophyletic groups. Of the 27 genera accepted today (Robyns, 1928; Bridson, 1998; Schatz, 2001), we include representatives from 23 in our analysis and also sample a majority of the often distinct infrageneric groups of the larger genera. Four monotypic genera are not represented (owing to a lack of material), but a discussion of their phylogenetic affinities based on morphological features is included in the discussion. MATERIAL AND METHODS TAXON SAMPLING A total of 69 ingroup taxa is included representing 23 genera (Table 1). The focus is on East African species, but taxa from other parts of Africa, south and southeast Asia, and the Pacific are also included. In phylogenetic studies of the subfamily Ixoroideae (Andreasen & Bremer, 2000), Alberta, Ixora and Mussaenda have been shown to be close to Vanguerieae and were therefore chosen as outgroup. We include two specimens of Ancylanthos rubiginosus due to the presence of an ITS paralogue for this species (see Results). In Lantz, Andreasen & Bremer (2002) Peponidium sp. was annotated as Pyrostria sp. and Pygmaeothamnus cf. chamaedendrum was incorrectly determined to Pygmaeothamnus zeyheri (Sond.) Robyns var. zeyheri Verdc. A few specimens determined only to cf. or sp. levels need further explanation. The specimen annotated as Leroya cf. richardiae has the typical fruits of Leroya and is very similar in vegetative characters to L. richardiae. It is undoubtedly close taxonomically, but its thicker indumentum compared to the type specimens of L. richardiae precludes its certain identification to species. Peponidium sp. is a specimen of an undescribed species from central Madagascar. It is dioecious, glabrous with small leaves (up to 30 ¥ 14 mm), has male fasciculate inflorescences with ‘false involucres’ (see Discussion on inflorescences below), and corolla throats congested with crisped hairs; thus, it is well in agreement with the diagnostic characters of Peponidium. Pygmaeothamnus cf. chamaedendrum is a geofrutex collected in South Africa with paired setulose leaves and large bilocular fruits. These characters support a clear affinity with Pygmaeothamnus chamaedendrum (Kuntze) Robyns var. setulosus Robyns (see Retief, 2003), but the lack of flowers inhibits us from determining the specimen to species. Vangueriopsis cf. longiflora is a fruiting specimen that exhibits some of the typical features of © 2004 The Linnean Society of London, Botanical Journal of the Linnean Society, 2004, 146, 257–283 PHYLOGENY OF VANGUERIEAE 259 Table 1. Voucher information and EMBL accession numbers. Herbarium abbreviations according to Holmgren, Holmgren & Barnett (1990). We were not able to sequence Fadogia cienkowskii for ITS and Ancylanthos rubiginosus (specimen B) for trnT-F. The trnT-F sequence for Pseudopeponidium ampijoroense was submitted in two parts owing to missing sequence and therefore has two accession numbers. New sequences are underlined Species Voucher Alberta magna E. Mey. Middleton s. n. 94101, Tonkin 200 (UPS) Bremer 3104 (UPS) Gillis 10838 (FTG) Zimba et al. 776 (UPS) Bingham 8652 (K) Mwiga 241 (UPS) Bremer et al. 4345 (UPS) Andreasen 36 (UPS) Bremer et al. 4259 (UPS) Kuchar 17410 (UPS) Bremer & Bremer 3686 (UPS) Luke & Luke 8373 (UPS) Luke & Luke 9045 (UPS) Bremer & Bremer 3704 (UPS) Borhidi et al. 85449 (UPS) Lantz 129 (UPS) Friis et al. 3671 (UPS) Luke & Luke 9032 (UPS) Luke et al. 9121 (UPS) McPherson 16297 (MO) Lorence 8672 (UPS) Chapman & Chapman 9109 (UPS) Lantz 101 (UPS) Taylor et al. 8318 (UPS) Bremer & Bremer 3799 (UPS) Gereau et al. 6010 (UPS) Lawton 1318 (S) Adam 20599 (UPS) Bremer 3083 (UPS) Luke 8341 (UPS) Luke 8347 (UPS) Bremer 3069 (UPS) Bremer & Bremer 3811 (UPS) Bremer & Bremer 3792 (UPS) Davis and Rakotonasolo APD 2501 (K) Bremer 3074 (UPS) Ixora coccinea L. Mussaenda erytrophylla Schumach. & Thonn. Ancylanthos rubiginosus Desf. (specimen A) Ancylanthos rubiginosus Desf. (specimen B) Canthium burttii Bullock Canthium ciliatum (Klotzsch) Kuntze Canthium coromandelicum (Burm.f.) Alston Canthium gilfillanii (N.E. Br.) O.B. Miller Canthium glaucum Hiern ssp. glaucum Canthium inerme (L.f.) Kuntze Canthium keniense Bullock Canthium lactescens Hiern Canthium mundianum Cham. & Schltdl. Canthium oligocarpum Hiern ssp. captum (Bullock) Bridson Canthium parasiebenlistii Bridson Canthium pseudosetiflorum Bridson Canthium pseudoverticillatum S. Moore Canthium siebenlistii (K. Krause) Bullock Cuviera angolensis Welv. ex K.Schum. Cyclophyllum barbatum (Forst.f.) A.C.Smith & S.P.Darwin Fadogia ancylantha Hiern Fadogia cienkowskii Schweinf. Fadogia elskensii De Wild. Fadogia tetraquetra K.Krause Fadogia verdcourtii Tennant var. verdcourtii Fadogiella stigmatoloba (K.Schum.) Robyns Hutchinsonia barbata Robyns Keetia gueinzii (Sond.) Bridson Keetia lukei Bridson Keetia venosa (Oliv.) Bridson Keetia zanzibarica (Klotzsch) Bridson ssp. zanzibarica Lagynias dryadum (S.Moore) Robyns Lagynias lasiantha (Sond.) Bullock Leroya cf. richardiae Cavaco Meyna tetraphylla (Hiern) Robyns ssp. comorensis (Robyns) Verdc. Multidentia concrescens (Bullock) Bridson & Verdc. Multidentia fanshawei (Tennant) Bridson Neoleroya verdcourtii Cavaco Pachystigma pygmaeum Robyns Peponidium horridum (Baill.) Arènes Peponidium sp. Plectroniella armata (K.Schum.) Robyns Pseudopeponidium ampijoroense Arènes Pseudopeponidium asosa Arènes Psydrax kraussioides (Hiern) Bridson Bidgood et al. 845 (K) Lovett et al. 3311 (K) 23339 SF (TEF) Pawek 12335 (BR) Labat et al. 2236 (K) Pettersson & Nilsson 752 (UPS) Bremer & Bremer 3790 (UPS) SF 33726 (TEF) Anthony 1390 (TEF) Lantz 114 (UPS) © 2004 The Linnean Society of London, Botanical Journal of the Linnean Society, 2004, 146, 257–283 EMBL accession number ITS/trnT-F AJ224842/AJ620118 AJ224826/AJ620117 AJ224823/AJ620116 AJ617747/AJ620119 AJ617748/– AJ617749/AJ620120 AJ617750/AJ620121 AJ315081/AJ620122 AJ617751/AJ620123 AJ617752/AJ620124 AJ315120/AJ620125 AJ617753/AJ620126 AJ617754/AJ620127 AJ315107/AJ620128 AJ617755/AJ620129 AJ617756/AJ620130 AJ617757/AJ620131 AJ617758/AJ620132 AJ617759/AJ620133 AJ315088/AJ620134 AJ617760/AJ620135 AJ315103/AJ620136 –/AJ620137 AJ315118/AJ719191 AJ315099/AJ620139 AJ315116/AJ620140 AJ315100/AJ620141 AJ315102/AJ620142 AJ315117/AJ620143 AJ617761/AJ620144 AJ617762/AJ620145 AJ315105/AJ620138 AJ315090/AJ620146 AJ315089/AJ620147 AJ617763/AJ620148 AJ315083/AJ620149 AJ315086/AJ620150 AJ315087/AJ620151 AJ617764/AJ719192 AJ315091/AJ620152 AJ617765/AJ620153 AJ315112/AJ620154 AJ315082/AJ620155 AJ617766/AJ719193 AJ719194 AJ617767/AJ620156 AJ617768/AJ620157 260 H. LANTZ and B. BREMER Table 1. Continued Species Voucher EMBL accession number ITS/trnT-F Psydrax livida (Hiern) Bridson Psydrax locuples (K. Schum.) Bridson Psydrax nitidum (Craib) Wong Psydrax obovata (Klotzsch ex Eckl. & Zeyh.) Bridson ssp. obovata Psydrax parviflora (Afz.) Bridson ssp. parviflora Psydrax schimperiana (A.Rich.) Bridson ssp. schimperiana Psydrax sp. A of F.T.E.A. Pygmaeothamnus cf. chamaedendrum (Kuntze) Robyns Pygmaeothamnus zeyheri (Sond.) Robyns var. zeyheri Pyrostria bibracteata (Baker) Cavaco Pyrostria hystrix (Bremek.) Bridson Pyrostria phyllantoidea (Baillon) Bridson Robynsia glabrata Hutchinson Rytigynia bagshawei (S.Moore) Robyns var. bagshawei Verdc. Rytigynia bugoyensis (K.Krause) Verdc. Rytigynia fuscosetulosa Verdc. Rytigynia mrimaensis Verdc. Rytigynia senegalensis Blume Scyphochlamys revoluta Balf.f. Tapiphyllum cinerascens (Welv. ex Hiern) Robyns var. cinerascens Tapiphyllum velutinum (Hiern) Robyns Vangueria apiculata K.Schum. Vangueria infausta Burch. Vangueria madagascariensis Gmelin Vangueria parvifolia Sond. Vangueriella spinosa (Schumach & Thonn.) Verdc. Vangueriopsis cf. longiflora Verdc. Lantz 109 (UPS) Bremer et al. 4290 (UPS) Ryding 599 (UPS) Bremer & Bremer 3762 (UPS) AJ617769/AJ620158 AJ617770/AJ620159 AJ315108/AJ620160 AJ315109/AJ620161 Gilbert & Phillips 9069 (UPS) Simon et al. 462 (UPS) Luke & Luke 9031 (UPS) Bremer & Bremer 3800 (UPS) Bremer et al. 4356 (UPS) Bremer 3036 (UPS) Bremer & Bremer 3791 (UPS) Taylor et al. 8486 (UPS) Hall & Amponsah 46545 (K) Borhidi et al. 84439 (UPS) Rwburindore 3536 (UPS) Frimodt Moller et al. NG117 (K) Luke & Luke 9020 (UPS) van den Berghen 8746 (BR) Chase 12564 (K) Milne-Redhead 3292 (BR) AJ315110/AJ620162 AJ617771/AJ620163 AJ617772/AJ620164 AJ315119/AJ620165 AJ617773/AJ620166 AJ315113/AJ620167 AJ315114/AJ620168 AJ315115/AJ620169 AJ617774/AJ620170 AJ315101/AJ620171 AJ315084/AJ620172 AJ315097/AJ620173 AJ617775/AJ620174 AJ315104/AJ620175 AJ617776/AJ620176 AJ315096/AJ620177 Emanuelsson 672 (S) Kårehed & Odhult 161 (UPS) Bremer et al. 4254 (UPS) Bremer 3077 (UPS) Bremer & Bremer 3771 (UPS) Merello et al. 1494 (K) Luke 8316 (UPS) AJ315098/AJ620178 AJ315095/AJ620179 AJ617777/AJ620180 AJ224839/AJ620184 AJ315092/AJ620181 AJ315085/AJ620182 AJ617778/AJ620183 V. longiflora, such as extremely long stipule apices and calyx lobes, and a ribbed fruit. However, it was collected in Kenya (K7, Shimba hills), far from the other known collections in Tanzania (T6, T7), and the specimen had no flowers. Although certainly close to V. longiflora, we cannot be certain that it is not an undescribed species. MOLECULAR METHODS Thirty-five specimens were amplified from herbarium material, the rest from silica-dried or fresh material. Extraction protocols and PCR cocktails are as outlined in Lantz et al. (2002), unless otherwise noted. In order to increase the likelihood that all possible ITS paralogues of different stabilities were amplified, 8–10% DMSO (Buckler, Ippolito & Holtsford, 1997) was included in all steps of the ITS amplification and sequencing. For the ITS region, primers ITS4 (White et al., 1990), ITS-I (Urbatsch, Baldwin & Donoghue, 2000), P17, and 26S-82R (Popp & Oxelman, 2001) were used in different combinations. The PCR reactions were run for 1 min at 97 ∞C followed by 35–50 cycles of 97 ∞C for 10 s; 48 ∞C-55 ∞C for 30 s; 72 ∞C for 20 s + 4 s/cycle; finishing with 72 ∞C for 7 min. The trnT-F region was amplified by various combinations of primers A, B, C, D, E, F (Taberlet et al., 1991), I (Bremer et al., 2002), but also primers A1, G, H, and rps4, published here. A1 (5¢-ACAAATGCGAT GCTCTAACC-3¢) is based on primer A, but moved four bases in the 3¢-direction. This modification reduces the number of possible base-pairings both within and between primer-molecules and simplified the amplification of the region between the trnT(UGU) and trnL(UAA)5¢-exon genes, which is usually hard to amplify. To construct the primer, sequences were produced by amplifying with primer rps4 (5¢-AAACGAG GTCCTCGRTAACG-3¢), located about 200 bases in the 5¢ direction from primer A, in combination with primer B or I. Primers G (forward; 5¢-GGAAAGCTGA CAGGGAGATA-3¢) and H (reverse; 5¢-TATCTCCCT GTCAGCTTTCC-3¢) are located in the intergenic © 2004 The Linnean Society of London, Botanical Journal of the Linnean Society, 2004, 146, 257–283 PHYLOGENY OF VANGUERIEAE spacer between trnT(UGU) and trnL(UAA)5¢-exon, approximately 500 bases downstream of primer A. The PCR cycle was constructed as follows: 95 ∞C for 2 min followed by 35–50 cycles of 95 ∞C for 35 s; 55 ∞C for 1 min; 72 ∞C for 2 min, finishing with 72 ∞C for 10 min. The sequencing reactions were either performed as in Lantz et al. (2002) or with the DYEnamic ET terminator Cycle Sequencing Kit (Amersham Biosciences) following the protocol of the manufacturer and run on a MegaBACE 1000 DNA Analysis System. Primers ITS2, ITS3, ITS4, ITS5 (White et al., 1990), P16 (Popp & Oxelman, 2001) and P25 (Lidén, Fukuhara & Axberg, 1995) were used for the ITS sequencing reactions. The sequencing reactions for the trnT-F region used the same primers as in the PCR amplification. For four specimens (Ancylanthos rubiginosus A and B, Canthium glaucum and Robynsia glabrata), it was impossible to produce readable ITS sequences using direct sequencing, which necessitated gel-cleaning of the PCR fragment and cloning. The full amount of PCR product was run on an agarose gel with TAE buffer. The gel was allowed to soak for 10 min in an ethidium-bromide solution and bands were visualized and cut out under a low power UV-light. DNA was extracted from the gel slices using the Ultrafree-DA kit (Millipore Corporation). We used the Topo Ta Cloning kit (Invitrogen life technologies) according to the manufacturer’s instructions for the cloning, and screened between five and ten colonies for each taxon by PCR followed by sequencing. MORPHOLOGICAL METHODS The morphological dataset is based on an investigation of herbarium material. With few exceptions the characters have been examined by the first author on actual material; only when material has been missing from the specimens (e.g. fruits or flowers) has the information been taken from literature. A list of specimens examined is available from the first author. We could not find flowering material for Canthium burttii, C. parasiebenlistii, Cuviera angolensis, Fadogia verdcourtii, Keetia lukei, Leroya cf. richardiae, Peponidium horridum, Pygmaeothamnus cf. chamaedendrum, Pyrostria hystrix, P. phyllantoidea, Rytigynia bugoyensis, R. fuscosetulosa, R. mrimaensis, Scyphochlamys revoluta, and Vangueriopsis cf. longiflora. Fruits were missing from our material of Peponidium sp. and Pseudopeponidium ampijoroense. Flowers, leaves and/ or fruits were boiled and examined under a mounted lens and measurements were taken. Hairs from inside the corolla and from the leaf surface were mounted in Hoyer’s solution (Anderson, 1954) and examined under a phase contrast microscope. The boiled material was preserved in Copenhagen Mixture (70 parts 95% ethanol, 29 parts distilled water and 1 part glyc- 261 erol). Thirty characters were coded and included in the morphological matrix (Appendix 2). When a character can be absent or present, and if present occurs in different states, we chose to divide the character into two (with separate coding; Lee & Bryant, 1999). All characters were traced on one of the most parsimonious trees (tree no. 1) using MacClade v.4.04 with ACCTRAN optimization. MacClade v.4.04 (Maddison & Maddison, 2001) was also used to calculate consistency index values for the morphological characters on the same tree used in the tracing. We used the Flora of Tropical East Africa (Verdcourt & Bridson, 1991) and Flora Zambesiaca (Bridson, 1998) as main taxonomic references because the results of most of the recently performed morphological/taxonomical studies of the tribe are included in these publications and they constitute the best general works on Vanguerieae in Africa. PHYLOGENETIC ANALYSES Sequences were aligned by eye. The phylogenetic analysis was performed using PAUP* v.4.0b10 (Swofford, 2002). All three datasets were included in a combined analysis but were also analysed separately. Phylogenetically informative insertion/deletion events (indels) were coded for the molecular datasets and given the same weight independent of size. We preferred not to introduce any hypothesis of evolution prior to the analyses, and thus all characters are treated as unordered. A heuristic parsimony search was performed with 100 random addition replicates, TBR and MULTREES on. Branch support was calculated using the jack-knife and bootstrap procedures implemented in PAUP*. The jack-knife analysis used a deletion frequency of 37.5%, 10 000 replicates, TBR branch swapping, 5 random addition sequence replicates, and MULTREES off. Bootstrap used 10 000 replicates, TBR branch swapping, 5 random addition sequence replicates, and MULTREES off. Bootstrap (bp) and jack-knife ( jk) values are listed in the text as (bp; jk). RESULTS SEQUENCE AND ALIGNMENT DATA Thirty-two new ITS sequences and 73 new trnT-F sequences were produced for the present study. ITS sequences in the ingroup varied in length from 611 bases for Psydrax obovata to 671 bases for Fadogiella stigmatoloba. When the PCR product of Fadogia cienkowskii was visualized on a gel, two bands were seen; one strongly amplified of about 400 bases and one faint band of a size comparable to other ITS sequences. We were unable to sequence any of the fragments. A paralogue was found for Ancylanthos rubiginosus characterized by a deletion of 13 bases in © 2004 The Linnean Society of London, Botanical Journal of the Linnean Society, 2004, 146, 257–283 262 H. LANTZ and B. BREMER ITS1. Two out of seven clones shared the deletion, but the deletion is not shared by any other species in the tribe and should not affect the phylogenetic position of the species. After alignment and removal of unalignable regions, the ITS matrix included 668 characters; 188 of these were informative, including 12 indel-characters. TrnT-F sequences in the ingroup varied in length from 1559 bases for Keetia zanzibarica to 1785 bases for Canthium gilfillanii. Alignment resulted in a matrix of 2118 positions for the trnT-F sequences of which 155 were informative, including 23 indelcharacters. All investigated species of Psydrax (apart from P. kraussioides which we were unable to sequence for this part) share a sequence of 40 bases not present in the other genera. The sequences of Fadogia elskensii, Neoleroya verdcourtii and Pseudopeponidium ampijoroense are partial as a consequence of difficulties in sequencing. MORPHOLOGICAL DATA All 30 morphological characters included in the analysis are informative. Characters 11 and 27 are nonreversed synapomorphies for the ingroup. The homoplasy of the characters as measured by the consistency index (CI) (Kluge & Farris, 1969; Goloboff, 1991) exhibit a wide range of values (from tree no. 1). The highest values, suggesting low levels of homoplasy, for the characters informative in the ingroup are found for supra-axillary spines (character no. 2; CI = 1.00), sex distribution (no. 4; CI = 0.67), bracts enclosing the inflorescence when young (no. 13; CI = 1.00), and filament reflexed (no. 23; CI = 1.00). Consistency index values are lowest for indumentum (no. 6; CI = 0.06), calyx lobe length (no. 16; CI = 0.08), corolla lobe apex tailed (no. 19; CI = 0.07), dark connective on the anthers (no. 24; CI = 0.08), and fruit size (no. 30; CI = 0.08), indicating high levels of homoplasy for these characters. PHYLOGENETIC ANALYSES The combined matrix of ITS, trnT-F and morphological data included 2816 characters. An heuristic search of the combined matrix (uninformative characters removed) resulted in 6207 most parsimonious trees with a length of 1332 (CI = 0.4062, RI = 0.7776). The strict consensus tree (Fig. 1) is well-resolved and contains several strongly supported clades. Canthium, Rytigynia, Tapiphyllum and possibly Pygmaeothamnus are polyphyletic, Fadogia is paraphyletic and the monophyly of Vangueria is uncertain. For several strongly supported groups we use informal names (see Fig. 1), some of which have already been suggested in Lantz et al. (2002). Some nodes in the tree (Fig. 1) have been marked by A, B, C, etc., which are referred to in the discussion. A tree in which the new classification here suggested has been implemented is presented in Figure 5. The two specimens of Ancylanthos rubiginosus group together with strong support (95; 94). We also ran a separate analysis of an ITS dataset where all clones of A. rubiginosus were included, and the clones form a monophyletic group (results not shown). It thus seems that the variation found for the different paralogues of A. rubiginosus does not pass the species boundary. The two molecular datasets were also analysed separately (results not shown). One strongly supported incongruence is seen in the Fadogia–Rytigynia group. Rytigynia bagshawei is sister to the rest of the group in the trnT-F tree but groups with Hutchinsonia barbata in the ITS tree. The result of the trnT-F analysis is favoured by the combined analysis. There is also some uncertainty concerning the position of Ancylanthos rubiginosus, and the taxa in the large-flowered group not included in the Vangueria group or in the Fadogia–Rytigynia group. Taxa with incongruent positions are restricted to the large-flowered group, and the incongruencies will be addressed in more detail in a later study on the large-flowered group (work in progress). The incongruencies do not affect the statements of homology made in the morphological discussion or the phylogeny for those species outside the large-flowered group. The unexpected position of Pygmaeothamnus cf. chamaedendrum in the spiny group caused us to re-extract and re-sequence the specimen to check for possible contamination. The new sequences were identical to the first ones and we can thus dismiss contamination as a cause behind the phylogenetic position of Pygmaeothamnus cf. chamaedendrum. A position in the spiny group is also supported by both the ITS and trnT-F datasets; there is no strongly supported topological incongruence indicative of hybridization. DISCUSSION As one of the main aims of this study is to give an overview of the morphological variation in the tribe, we will open this section with a rather comprehensive account on the various useful characters for delimitation of groups and genera. Characters much dependent on quality and age of the material, or characters that vary continuously and lack discrete states have been excluded from the phylogenetic analysis. However, because some of these characters are certainly useful at other levels and as we also wish to present a more inclusive discussion on the variation present in the tribe, the account encompasses more characters © 2004 The Linnean Society of London, Botanical Journal of the Linnean Society, 2004, 146, 257–283 263 PHYLOGENY OF VANGUERIEAE 95 94 87 95 100 100 64 70 73 79 80 85 78 83 90 94 76 81 73 82 C E 91 95 99 100 100 100 A B 88 94 81 89 100 100 92 97 96 98 D 86 92 58 64 72 78 63 77 100 100 92 96 96 98 69 80 100 100 100 100 89 94 100 100 100 100 <50 <50 99 100 94 97 51 64 <50 <50 <50 <50 90 95 97 98 96 98 69 73 75 82 55 71 77 70 77 71 78 53 67 58 63 53 51 Outgroup Vangueria group Multidentia Large-flowered group 58 68 Mussaenda erythrophylla Ixora coccinea Alberta magna Ancylanthos rubiginosus a Ancylanthos rubiginosus b Lagynias dryadum Lagynias lasiantha Pachystigma pygmaeum Tapiphyllum cinerascens Vangueria apiculata Vangueria infausta Vangueria madagascariensis Vangueria parvifolia Rytigynia fuscosetulosa Tapiphyllum velutinum Vangueriopsis cf. longiflora Pygmaeothamnus zeyheri Robynsia glabrata Canthium oligocarpum Cuviera angolensis Multidentia concrescens Multidentia fanshawei Fadogia ancylantha Fadogia verdcourtii Fadogia cienkowskii Fadogia tetraquetra Fadogia elskensii Fadogiella stigmatoloba Rytigynia senegalensis Hutchinsonia barbata Rytigynia bagshawei Canthium ciliatum Plectroniella armata Canthium coromandelicum Canthium glaucum Canthium inerme Meyna tetraphylla Pygmaeothamnus cf. chamaedendrum Rytigynia bugoyensis Vangueriella spinosa Rytigynia mrimaensis Canthium burttii Canthium keniense Canthium pseudoverticillatum Canthium lactescens Canthium parasiebenlistii Canthium siebenlistii Canthium gilfillanii Canthium mundianum Keetia gueinzii Keetia lukei Keetia venosa Keetia zanzibarica Canthium pseudosetiflorum Leroya cf. richardiae Neoleroya verdcourtii Pseudopeponidium ampijoroense Pyrostria hystrix Pyrostria phyllantoidea Pseudopeponidium asosa Pyrostria bibracteata Scyphochlamys revoluta Peponidium horridum Peponidium sp. Cyclophyllum barbatum Psydrax kraussioides Psydrax livida Psydrax locuples Psydrax parviflora Psydrax nitidum Psydrax obovata Psydrax schimperiana Psydrax sp. A of F.T.E.A. FadogiaRytigynia group Spiny group Canthium subgenus Afrocanthium Keetia Dioecious group Psydrax Figure 1. The strict consensus tree from the combined analysis (ITS + trnT-F + morphology). Bootstrap support values are above the branches, jackknife below. A, B, C, etc. refer to clades mentioned in the text. Type species of currently recognized genera in bold. © 2004 The Linnean Society of London, Botanical Journal of the Linnean Society, 2004, 146, 257–283 264 H. LANTZ and B. BREMER than used in the analysis. We include comments on how the morphological characters were coded for cases where we feel it is necessary to do so. CHARACTERS Habit All Vanguerieae species are woody. The geofrutices (Robbrecht, 1988) are sometimes mentioned as semiherbacous, because the aboveground parts of some species are only slightly woody or completely herbaceous, but the underground parts are always woody. Fadogia is a large genus characterized by this habit, but there are geofrutices in several other genera as well. Apart from Fadogia, geofrutices included in this study are Ancylanthos rubiginosus, Fadogiella stigmatoloba, Multidentia concrescens, Pachystigma pygmaeum, Pygmaeothamnus cf. chamaedendrum, P. zeyheri and Tapiphyllum cinerascens. All of these species belong in a strongly supported (100; 100) clade (A; Fig. 1), but the growth form has evolved several times in this clade (Fig. 2). An anatomical investigation of the wood of Vanguerieae (Lens et al., 2000a) revealed that the underground parts of Fadogiella stigmatoloba exhibit features characteristic of roots, while these structures in other geofrutices (including the closely related genus Fadogia) more resemble the aboveground parts. There is also some variation in the woodiness of the aboveground parts, and it is possible that the geofrutex character could be further divided, but further anatomical data are needed for such a division to be performed. The species of Keetia and Psydrax subgenus Phallaria, here represented by P. kraussioides, are without exceptions climbers or scandent shrubs. However, there are climbers in other genera as well (e.g. Canthium, Lagynias, Meyna, Multidentia and Vangueriella, and also a few species in Psydrax subgenus Psydrax), although only as rare exceptions to the otherwise common habit in those groups. The majority of the species in the tribe are not geofrutices or climbers but are usually described as either shrubs, treelets or trees, and the distinction between these growth forms is difficult to make. Treelet seems to be especially difficult to define; the term is usually used for large shrubs or small trees and is thus based on size and not on homology. Not only are the definitions of these habits vague and used differently by different authors and collectors (Coates Palgrave, 2002), but many species also have a very variable habit, further increasing the difficulty of defining states for the character. We code for three states: geofrutices, climbers, and shrubs and trees. Efforts to further divide the character into more states did not change the support of any groupings (results not shown), but it should be noted that one state (shrubs or trees) is perhaps more inclusive than the other ones. Spines and shoots There are at least two types of spines present in Vanguerieae. Synapomorphic for the spiny group (except Pygmaeothamnus cf. chamaedendrum; Fig. 2) are paired supra-axillary spines, i.e. they are positioned at a short distance from the nodes. The spines are usually present on all parts of the plant, but for some species (e.g. Canthium inerme) they are more common on sucker shoots or on seedlings. Another type can be seen for Canthium oligocarpum where the spines are present at the nodes, usually in whorls of three, and only occur on coppice shoots and young branches. Of the other spiny species included in this study (Fig. 2), Cuviera angolensis, Hutchinsonia barbata, Lagynias lasiantha and Robynsia glabrata only carry spines occasionally or only on older parts rarely collected and we have not examined any spiny material from these species. Given this lack of information, we code only for supra-axillary spines and not for any other type. Spines are also known from Rytigynia subgenus Rytigynia, a few South African Canthium species not placed to subgenus (Tilney, 1986; Bridson, 1992), and several south-east Asian Canthium species (Wong, 1989). Some species, especially in the spiny group, have brachyblasts (lateral shoots with contracted internodes). Similar but not identical shoots occur in several other genera, e.g. the lateral shoots of Pyrostria hystrix. For Canthium pseudoverticillatum the lateral shoots are extremely reduced which makes the leaves appear as in whorls of four. This is different from the other type of whorled leaves common in the tribe (e.g. Fadogia), in which the leaves are present on the main shoot at the same node with stipules present between all leaves and usually are in whorls of three. There is a correlation between a geofrutescent habit and whorled leaves, but the correlation is not absolute, and the characters are coded as two separate characters. Trichomes Hairs (trichomes) vary in abundance and on where they are positioned. Hairs on leaves, stems, fruits and outer parts of the corollas are all of the same type and exhibit only a slight variation in one individual. This external indumentum is not homologous to the types found inside the corolla (see Discussion on the corolla) or inside the stipules. This seems to be the case for most Rubiaceae species (Verdcourt, 1958; Robbrecht, 1988). There is also a strong correlation between the presence of hairs for the different structures, e.g. species with hairs on the fruits always have hairs on the leaves, species with glabrous leaves never have hairs on the outside of the corollas, etc. Creating separate © 2004 The Linnean Society of London, Botanical Journal of the Linnean Society, 2004, 146, 257–283 PHYLOGENY OF VANGUERIEAE 265 Mussaenda erythrophylla Ixora coccinea Anyclanthos rubiginosus a Ancylanthos rubiginosus b Lagynias dryadum Lagynias lasiantha Pachystigma pygmaeum Tapiphyllum cinerascens Vangueria apiculata Vangueria infausta Vangueria madagascariensis Vangueria parvifolia Rytigynia fuscosetulosa Tapiphyllum velutinum Vangueriopsis cf. longiflora Pygmaeothamnus zeyheri Robynsia glabrata Cuviera angolensis Canthium oligocarpum Multidentia concrescens Multidentia fanshawei Fadogia ancylantha Fadogia verdcourtii Fadogia cienkowskii Fadogia tetraquetra Fadogia elskensii Fadogiella stigmatoloba Rytigynia senegalensis Hutchinsonia barbata Rytigynia bagshawei Canthium ciliatum Plectroniella armata Canthium coromandelicum Canthium inerme Canthium glaucum Meyna tetraphylla Rytigynia mrimaensis Rytigynia bugoyensis Vangueriella spinosa Pygmaeothamnus cf. chamaedendrum Canthium burttii Canthium lactescens Canthium keniense Canthium pseudoverticillatum Canthium parasiebenslistii Canthium siebenlistii Canthium gilfillanii Canthium mundianum Keetia gueinzii Keetia lukei Keetia venosa Keetia zanzibarica Canthium pseudosetiflorum Leroya cf. richardiae Neoleroya verdcourtii Pseudopeponidium ampijoroense Pyrostria hystrix Pyrostria phyllantoidea Pseudopeponidum asosa Pyrostria bibracteata Scyphochlamys revoluta Peponidium horridum Peponidium sp. Cyclophyllum barbatum Psydrax kraussioides Psydrax livida Psydrax locuples Psydrax parviflora Psydrax nitidum Psydrax obovata Psydrax schimperiana Psydrax sp. A of F.T.E.A. Alberta magna Habit Supra-axillary spines Geofrutex Absent Shrub or tree Present Climber Polymorphic Polymorphic Equivocal Figure 2. Habit and spines (characters 1 and 2; Appendix 1) mapped onto one of the most parsimonious trees using MacClade v.4.01 with ACCTRAN optimization. The absence of a square at the end of the branches for some species indicates that the character state is unknown for that species. © 2004 The Linnean Society of London, Botanical Journal of the Linnean Society, 2004, 146, 257–283 266 H. LANTZ and B. BREMER characters for the presence of hairs on all mentioned structures would result in an overweighting of the presence of hairs and we have coded only for the presence of hairs on the leaf blades. When comparing species, hairs can have a varying number of cells, different colour and variable length. Unicellular hairs are only found in Psydrax (Tilney, van Wyk & Kok, 1988; Tilney, Kok & van Wyk, 1990). This distinction between unicellular and multicellular hairs is the only clear distinction we can make when looking at the variation present in the whole tribe. In more restricted groups, characters such as colour and exact number of cells are likely to be more useful. Leaves Size, shape and texture of leaves vary within the tribe, but are of use primarily at the species level. Leaf texture is of some use to delimit Psydrax, which, with very few exceptions, has coriaceous leaves. Apparent or obscure tertiary nerves have been used to diagnose some genera (Multidentia and Pyrostria, respectively), but the character is not otherwise useful at the generic level. Domatia are commonly present and vary in structure and amount of hairs present. This information is potentially useful taxonomically but has not been investigated in detail here (or elsewhere). Fadogia seems to lack domatia altogether, although domatia can be difficult to see in some species owing to a thick indumentum, and domatia are also very rare in the Vangueria group. Of the Vangueria group species studied here, only V. apiculata has domatia, and this only rarely. Anatomical leaf characters potentially useful at the genus level are differentiation and arrangement of parenchymatous cells and arrangement of collenchyma in the midrib area (Tilney et al., 1990). Amount of indumentum varies from glabrous to densely pubescent. We have coded one absent/present character and one character for amount of indumentum, if present. The criteria used for the states are indumentum present but not obscuring the surface or thick indumentum present, obscuring the surface at least below. This distinction has been used in the Flora of Tropical East Africa (Verdcourt & Bridson, 1991) and Flora Zambesiaca (Bridson, 1998) to key out some genera (e.g. Tapiphyllum). Stipules in Vanguerieae are always connate, but sometimes only slightly so. They usually consist of a basal broadly ovate to triangular part and a narrow apical part, but this differentiation is not always present. The narrow apices, when present, tend to become less pronounced with increasing age of the stipule, sometimes disappearing completely. Some species have stipule apices widest above the midpoint, and this has been coded as an absence/presence character. Presence of hairs within the stipules is phylogenetically informative. These hairs are completely lacking in Psydrax, rare in Keetia and also lacking in Canthium mundianum. It is one of the few morphological characters that indicate a relationship between Canthium subgenus Afrocanthium and Keetia. Sex distribution Hermaphrodite flowers are the most common type in the tribe, but there is also a large group of dioecious species (the dioecious group; Fig. 1). Pyrostria (incl. Pseudopeponidium, Peponidium, Leroya and Neoleroya; see Schatz, 2001), Scyphochlamys, and Canthium subgenus Bullockia, here represented by C. pseudosetiflorum, all belong to this group. Male flowers have a pollen presenter and sometimes rudimentary ovules and female flowers have sterile anthers. However, it is possible that on rare occasions a male plant could produce a few fruits or a female plant form some pollen (Bridson, 1987b), although no detailed study on the sex expression of these species has been made. Canthium mundianum is one of the rare recorded examples of species not belonging to this group that is not an obligate hermaphrodite. It is gynodioecious (Balkwill, Sebola & Robinson, 1996); single individuals either have hermaphrodite or female flowers. A study of Psydrax odorata (Skottsberg, 1945) noted an occurrence of functionally female flowers for some specimens of this species also. It is possible that further studies on the sexual expression of Vanguerieae would show that dioecious or gynodioecious flowers are more common than presently known. The number of flowers in the inflorescences is correlated with the sex of the flowers. Individuals with female flowers in the dioecious group tend to have one or two flowers in each inflorescence compared to individuals with male flowers that have up to 20 flowers. For Pyrostria p.p. (Pseudopeponidium) the number of corolla lobes also varies – female flowers usually having eight lobes and male have four to five. This sex-correlated variation complicates the coding for these species, and there are several possible solutions. Polymorphism has been defined as ‘variation within species that is (at least partly) independent of ontogeny and sex’ (Wiens, 1999). Based on this definition, the sex-correlated variation we see is different from the polymorphic variation present for other characters. This distinction would not be seen if we scored all states that are present for the sexcorrelated characters. For species that exhibit a sexcorrelated variation for a character, that character is coded as unknown (?). A difference in number of locules between male and female flowers is also known for many members of the dioecious group (Bridson, 1987b). The number of locules reported for male individuals (two to three) is based on the rudimentary ovaries that sometimes are visible. We suggest that the lesser © 2004 The Linnean Society of London, Botanical Journal of the Linnean Society, 2004, 146, 257–283 PHYLOGENY OF VANGUERIEAE amount of ovaries in male flowers is due to their being nonfunctional, and we therefore do not consider this variation to be sex-correlated variation in the same sense as number of flowers per inflorescence discussed earlier. The number of locules coded and included in the morphological matrix is based on functional structures, i.e. ovaries and fruits of female or hermaphrodite flowers. Inflorescences Inflorescences at anthesis can either be borne at nodes where the leaves have fallen or be subtended by leaves. The character closely approaches deciduous (leaves not present) or not deciduous (leaves present). Defining homologous structures in inflorescences is generally a difficult matter (Weberling, 1965), and this is definitely the case for the Vanguerieae. There is a lot of variation in the inflorescences, but apart from number of flowers per inflorescence and the presence of a discrete type of paired bracts on the peduncle (see below), we refrain from coding any more characters. Length of peduncle is a potentially useful character; some species have inflorescences that look fasciculate owing to a much shortened peduncle. Some have an elongate peduncle, but there are many intermediates. Length and presence of secondary inflorescence branches vary; several species have very reduced secondary branches resulting in an umbel-like inflorescence. Bracts and bracteoles occur in varying size and shape. Comparisons of these characters from the literature are further complicated by the imprecise usage of the terms. Here we use bracts for structures present on the peduncle and bracteoles for structures occurring further up in the inflorescence. Several species have paired bracts on the peduncle (e.g. Keetia gueinzii). Others have a structure more reminiscent of a connate sheath, usually splitting up with increasing age (e.g. Vangueria madagascariensis). Bracteoles are commonly unpaired, but can be paired or lacking. The paired, large bracts of Pyrostria and related genera are distinct. They completely surround the young inflorescences and later split into two lobes, and are in structure similar to the stipules. A further development has occurred in the genus Scyphochlamys, where the flowers are collected in an involucre (Verdcourt, 1983). In Pyrostria p.p. (Peponidium) a ‘false involucre’ (Arènes, 1960) is sometimes present at the base of the peduncle. We view these structures as homologous to the bracts of Pyrostria. Number of flowers has, with some hesitation, been included in the analysis. Defining discrete states for the character has been exceedingly difficult and some species fall in between the categories suggested and are polymorphic in the current coding. The character is informative for some genera (e.g. Fadogia and Keetia), and is often used in the description of genera. 267 Calyx Two features of the calyx have been coded; length of the lobes and length of the calyx limb tube. Calyx lobes are useful for identifying species, but seem to be of limited use for delimiting genera. Most species have small, shortly triangular or indistinct lobes, but in some genera the lobes are elongate (defined here as length > 1 mm and ratio of length : width > 1). Elongate lobes can either be linear, lanceolate, obovate, ovate, rounded or triangular. The distinction between the different shapes is hard to make and we have only coded for elongate or not elongate lobes. The calyx limb tube (the part of the calyx free from the ovary, but not lobed) can either be short, with the disk of the gynoecium clearly visible when a flower with the corolla removed is viewed from the side, or well developed and hiding the disk. Corolla Corolla aestivation is valvate. Most species are fivelobed, but four lobes are common in Keetia, Psydrax and Pyrostria, and also occur in Plectroniella, and all subgenera of Canthium apart from subgenus Lycioserissa. Some species are obligately four-lobed but many have a varying amount of either four or five lobes. Six lobes (or even more) are sometimes present in Fadogia, Multidentia and a few other genera. We include a character for the presence of appendages on the corolla lobe apices but do not make any distinction between appendages of different lengths. An apiculum is here interpreted as a short appendage and species with apiculate corolla lobe apices are scored as ‘appendages present’. The distinction between the two states is sometimes very hard to make and the character is certainly open to some interpretation. Corolla tube length varies from around 1 mm to 30 mm. There are two discrete groups; tubes usually <8 mm or tubes usually >9 mm. A further division can be made in the group with long tubes between robust, wider tubes and weaker, more slender ones. Species with long tubes are rare in the tribe and were formerly thought to be closely related (e.g. Robyns, 1928), especially those that also are more robust. These species were usually placed in Fadogia, Ancylanthos or Temnocalyx. Today, the importance of the character has been re-evaluated and the two latter genera have been reduced to the type species and the remaining species moved to several other genera (Verdcourt, 1981; Bridson, 1996). In the current circumscriptions, Fadogia houses most of the species with long and robust corolla tubes, but they can also be found in Ancylanthos, Fadogiella, Lagynias, Multidentia, Tapiphyllum and Temnocalyx. Hutchinsonia and some members of Cyclophyllum (Smith & Darwin, 1988; Reynolds & Henderson, 2001) have long but slender corollas, clearly different from the robust © 2004 The Linnean Society of London, Botanical Journal of the Linnean Society, 2004, 146, 257–283 268 H. LANTZ and B. BREMER type. There is a tendency for the whole clade (B; Fig. 1) consisting of the Fadogia–Rytigynia group and its sister group to have larger corollas compared with the rest of the tribe, both in length and width of the tube and length of the lobes. We therefore annotate this clade as the large-flowered group (Fig. 1), with the caveat that this is not based on a discrete state but instead on a trend, and is not without exceptions. Cyclophyllum is the only genus with long corolla tubes not in the large-flowered group. Most Vanguerieae species have hairs in the corolla tube. Diffusely spreading or crisped hairs present at the throat are common, and a ring of retrorse hairs at varying positions in the corolla tube is also often present. There are, however, intermediates and these can be difficult to assign to one of the mentioned groups. An investigation of the hairs with the help of a phase contrast microscope reveals that apart from the variation noticeable by eye, there are differences in shape and surface only visible under the microscope. We suggest, albeit with some hesitation, that there are indeed two kinds of hairs: verrucose and smooth. The verrucose type has a surface covered with elongate warts and is usually moniliform to a varying degree, at least towards the apex, but can also be straight in outline, tapering to the usually obtuse apex. The hairs are usually diffusely spreading but can be rather retrorse (and thus easily mistaken for the other kind of hairs, see below) and often form a tangled mass close to the opening of the corolla tube. The smooth kind has a smooth or slightly striate surface without warts, is straight or only slightly undulating in outline and has an acute apex. This type can be further divided into (1) retrorse hairs, straight, placed in a well-defined ring usually separate from other kinds of hairs, hairs parallel to each other, and (2) hairs usually diffusely spreading (often bent and with a wrinkled surface on herbarium specimens), only slightly retrorse or not at all, sometimes attached at the same point as verrucose hairs or alone and easily mistaken for the verrucose hairs. Verdcourt (1958) mentions the presence of flat, ribbon-like hairs inside the corolla of most Rubiaceae. The question of whether the two hair types discussed here, i.e. verrucose and smooth, also are flat is difficult to answer owing to the very transparent nature of the hairs, but the smooth kind does appear to be flat. As for the verrucose hairs, we can clearly see that the hairs have a volume, at least when they are moniliform. The two hair types co-occur in several species, but the two states of the smooth hairs never do, supporting our coding of the characters (Patterson, 1988). We have constructed three characters for these hairs: absence or presence of verrucose and smooth hairs, respectively, and type of smooth hairs if present. This investigation can be thought of as a first effort to more clearly understand the very complex and variable nature of the internal corolla hairs in Rubiaceae. Presence of smooth retrorse hairs is clearly informative (Fig. 3), but the verrucose hair character is more homoplastic (Fig. 3). Whether this homoplasy is the result of a mistaken homology assessment on our part or whether the character for some reason is subject to common reversals and parallel evolution is not known. We welcome further investigation into this character. Several species also have hairs on the outside of the corolla. In some cases the hairs are restricted to the tube or to the lobes, but they can also cover the whole outer part of the corolla. The type of hair present is the same type as other external indumenta. Androecium In Vanguerieae, the pollen is released from the anthers in bud directly onto the pollen presenter, which when the flower opens presents the pollen to the pollinators. The stamens are attached to the corolla close to the opening of the tube, alternate to the corolla lobes, and remain in the mature flowers. The anthers may be included in the corolla tube (or just exserted), exserted on long straight filaments, or exserted on reflexed filaments. This is a function of the length and orientation of the filament. Anthers exserted on reflexed filaments is the common situation in Psydrax, and a good synapomorphy for the genus. The orientation of the anthers also varies; some species have flexed anthers with the dorsiventral plane perpendicular to the axis of the flower. This is more a function of where the filament attaches on the anther rather than the length of the filament. All anthers are dorsifixed, but the point of attachment ranges from close to the base of the anther to the middle of the anther, and it is especially when the filament is attached close to the middle that the anther tends to be perpendicular to the axis of the flower. Filament attachment on the anther has not been included in the analysis because the variation is gradual. The point of attachment cannot easily be extrapolated from the orientation of the anthers, which varies depending on the developmental stage of the flowers and the quality of the material. Sterile apical appendages on the anthers are known from several genera in the tribe. The structure is correlated with type of pollen presenter, of which there are two types, type 1 and 2 (Igersheim, 1993; see Gynoecium section in the Discussion). However, we are unable to distinguish between the appendages present in some type 1 species (e.g. Canthium inerme) and those of the type 2 species (e.g. Vangueria parvifolia) and can find no discrete states for the character. There is too much variation in the apical appendages for the type of pollen presenter to be deduced from them. Darkened connectives on the dorsal face of the anthers are a feature of many species, but have been © 2004 The Linnean Society of London, Botanical Journal of the Linnean Society, 2004, 146, 257–283 PHYLOGENY OF VANGUERIEAE 269 Mussaenda erythrophylla Ixora coccinea Anyclanthos rubiginosus A Ancylanthos rubiginosus B Lagynias dryadum Lagynias lasiantha Pachystigma pygmaeum Tapiphyllum cinerascens Vangueria apiculata Vangueria infausta Vangueria madagascariensis Vangueria parvifolia Rytigynia fuscosetulosa Tapiphyllum velutinum Vangueriopsis cf. longiflora Pygmaeothamnus zeyheri Robynsia glabrata Cuviera angolensis Canthium oligocarpum Multidentia concrescens Multidentia fanshawei Fadogia ancylantha Fadogia verdcourtii Fadogia cienkowskii Fadogia tetraquetra Fadogia elskensii Fadogiella stigmatoloba Rytigynia senegalensis Hutchinsonia barbata Rytigynia bagshawei Canthium ciliatum Plectroniella armata Canthium coromandelicum Canthium inerme Canthium glaucum Meyna tetraphylla Rytigynia mrimaensis Rytigynia bugoyensis Vangueriella spinosa Pygmaeothamnus cf. chamaedendrum Canthium burttii Canthium lactescens Canthium keniense Canthium pseudoverticillatum Canthium parasiebenslistii Canthium siebenlistii Canthium gilfillanii Canthium mundianum Keetia gueinzii Keetia lukei Keetia venosa Keetia zanzibarica Canthium pseudosetiflorum Leroya cf. richardiae Neoleroya verdcourtii Pseudopeponidium ampijoroense Pyrostria hystrix Pyrostria phyllantoidea Pseudopeponidum asosa Pyrostria bibracteata Scyphochlamys revoluta Peponidium horridum Peponidium sp. Cyclophyllum barbatum Psydrax kraussioides Psydrax livida Psydrax locuples Psydrax parviflora Psydrax nitidum Psydrax obovata Psydrax schimperiana Psydrax sp. A of F.T.E.A. Alberta magna Smooth hairs Verrucose hairs Retrorse Absent Bent and wrinkled Present Figure 3. Smooth and verrucose hairs from the corolla inside (characters 21 and 22; Appendix 1) mapped onto one of the most parsimonious trees using MacClade v.4.01 with ACCTRAN optimization. An absence of a square at the end of the branches indicates that the character is either not present or not known for the species. © 2004 The Linnean Society of London, Botanical Journal of the Linnean Society, 2004, 146, 257–283 270 H. LANTZ and B. BREMER used to define the subgenera of Canthium (Bridson, 1992). Some species have no darkened connective or sometimes just a slight coloration present at the point of filament attachment. Others have the whole dorsal side dark or only the margin pale. The character is only just visible in recently collected or living material, which tends to be very pale; the colour darkens as the specimen dries. Gynoecium Style at least twice as long as corolla tube is a feature of all Psydrax and Keetia species, and is also known from a few other species that are not closely related (see Appendix 2). Structure of the pollen presenter, especially the amount by which the style is recessed into the presenter, has been used successfully for delimiting some of the genera earlier included in Canthium, e.g. Pyrostria (Bridson, 1987b), and Keetia (Bridson, 1986). In Pyrostria, the style is not recessed at all; in Canthium (as circumscribed by Bridson, 1992) it is slightly recessed; and in Keetia it is very deeply recessed. In some cases the difference between recessed and not recessed is very slight, and the character can in difficult cases be subject to opinion. The character is also somewhat connected to the shape of the pollen presenter, i.e. for the style to be deeply recessed the presenter has to be elongate. We code for recessed and not recessed, but do not include the shape of the presenter because it is correlated with the amount by which the style is recessed into the presenter. The variation in shape is gradual throughout the tribe and is also dependent on how deeply lobed the presenter is. The presenters are usually globose to cylindrical, but tend to look coroniform or mitriform when they are deeply divided by the stigmatic lobes. For all examined species, the number of stigmatic lobes is the same as the number of locules in the ovary (before any ovules are aborted). Based on developmental studies of flower buds in different stages of development, Igersheim (1993) suggested that there are two types of pollen presenters in the tribe and that these types are correlated with the presence of sterile apical appendages on the anthers. In type 1, the stigmatic surfaces are tightly pressed together in bud and not exposed, and the anthers lack or have very short appendages. This type is the most common in Vanguerieae. In type 2, the stigmatic surfaces are exposed from early development but protected by the sterile appendages of the anthers. According to Igersheim (1989, 1993), type 2 is present in Vangueria, Fadogia, Pachystigma, Tapiphyllum, Lagynias, Ancylanthos and part of Rytigynia, a group of genera that with a few additions agree with the large-flowered group found in this study (also discussed by Bridson, 1996). The type of presenter thus seems to be of taxonomic importance, but a lack of developmental data stops us from coding the character. The number of locules in the ovaries and fruits varies from two to 20, with more than five locules only regularly occurring in female individuals of Pyrostria and Scyphochlamys (Bridson, 1987b; Schatz, 2001). It is a useful character for defining larger groups of genera, e.g. the strongly supported suprageneric Vangueria group that with few exceptions has three to five locules. We divide the character into three states: 2 locules, 3–5 locules, and usually > 5 locules. As seen from Figure 4, plurilocular ovaries are likely to have evolved several times and bilocular ovaries is the plesiomorphic condition. Fruits and seeds All species have fleshy drupes with one to ten pyrenes. Fruits with one pyrene are, however, the result of faulty development because the number of locules is always two or more. Pyrene characters have been used successfully in several genera earlier included in Canthium (Bridson, 1985, 1987a, b). The number of pyrenes is a major factor in determining the shape of the fruit; fruits with two pyrenes tend to be bilobed, fruits with more pyrenes are more globose. The presence of a lid-like area surrounding the apical crest of the pyrenes is a useful character to differentiate Keetia from Psydrax (Bridson, 1986). This character has not been reported for the sister group of Keetia, Canthium subgenus Afrocanthium (Fig. 1). Other variation of the pyrenes includes degree of woodiness of the pyrene walls; Multidentia is partly defined by having thickly woody pyrenes (Bridson, 1987a). Size of the fruits varies; the bilocular fruits tend to be smaller than the plurilocular ones, with the notable exception of Multidentia, a bilocular genus with very large fruits (Bridson, 1987a). There are also some interesting characters to be gained from the embryos (Capuron, 1969; Bridson, 1985; Verdcourt, 1987; Bridson, 1996). Investigations of the embryos have revealed two states for the plane of the cotyledons. For Keetia and Psydrax, the plane is parallel to the ventral face of the pyrenes, while for most other Vanguerieae it is perpendicular. A difference in the ratio of the length of cotyledons to radicle has also been used, but there is some doubt about how constant both of the mentioned embryonic characters are within species (Verdcourt, 1987). Pollen Several papers have been published on the pollen morphology of Vanguerieae (Tilney, 1986; Havard & Verdcourt, 1987; Verdcourt, 1987; Igersheim, 1989; Tilney & van Wyk, 1997; Lens et al., 2000b; we follow the terminology used in the last reference). As it is clear from these studies that the pollen of © 2004 The Linnean Society of London, Botanical Journal of the Linnean Society, 2004, 146, 257–283 PHYLOGENY OF VANGUERIEAE 271 Mussaenda erythrophylla Ixora coccinea Anyclanthos rubiginosus a Ancylanthos rubiginosus b Lagynias dryadum Lagynias lasiantha Pachystigma pygmaeum Tapiphyllum cinerascens Vangueria apiculata Vangueria infausta Vangueria madagascariensis Vangueria parvifolia Rytigynia fuscosetulosa Tapiphyllum velutinum Vangueriopsis cf. longiflora Pygmaeothamnus zeyheri Robynsia glabrata Cuviera angolensis Canthium oligocarpum Multidentia concrescens Multidentia fanshawei Fadogia ancylantha Fadogia verdcourtii Fadogia cienkowskii Fadogia tetraquetra Fadogia elskensii Fadogiella stigmatoloba Rytigynia senegalensis Hutchinsonia barbata Rytigynia bagshawei Canthium ciliatum Plectroniella armata Canthium coromandelicum Canthium inerme Canthium glaucum Meyna tetraphylla Rytigynia mrimaensis Rytigynia bugoyensis Vangueriella spinosa Pygmaeothamnus cf. chamaedendrum Canthium burttii Canthium lactescens Canthium keniense Canthium pseudoverticillatum Canthium parasiebenslistii Canthium siebenlistii Canthium gilfillanii Canthium mundianum Keetia gueinzii Keetia lukei Keetia venosa Keetia zanzibarica Canthium pseudosetiflorum Leroya cf. richardiae Neoleroya verdcourtii Pseudopeponidium ampijoroense Pyrostria hystrix Pyrostria phyllantoidea Pseudopeponidum asosa Pyrostria bibracteata Scyphochlamys revoluta Peponidium horridum Peponidium sp. Cyclophyllum barbatum Psydrax kraussioides Psydrax livida Psydrax locuples Psydrax parviflora Psydrax nitidum Psydrax obovata Psydrax schimperiana Psydrax sp. A of F.T.E.A. Alberta magna Number of locules Fruit size 2 Up to 1 cm 3-5 1 cm or more >5 Equivocal Polymorphic Figure 4. Number of locules in the ovary and size of fruits (characters 29 and 30; Appendix 1) mapped onto one of the most parsimonious trees using MacClade v.4.01 with ACCTRAN optimization. An absence of a square at the end of the branches indicates that the character is either not present or not known for the species. © 2004 The Linnean Society of London, Botanical Journal of the Linnean Society, 2004, 146, 257–283 272 H. LANTZ and B. BREMER Vanguerieae is homogeneous and of little use for delimiting genera, we code no pollen characters here. The basic type of pollen in Vanguerieae can be described as 3-zonoaperturate, suboblate to oblate spheroidal, pororate or porate (rarely colporate), and with a perforate to sometimes reticulate sexine. There is a tendency for the taxa in the large-flowered group (Fig. 1), to have both larger and more spheroidal pollen (Igersheim, 1989). There are some interesting pollen structures found in the tribe worth mentioning. Tilney & van Wyk (1997) discuss the presence of intine and/or protoplasm protruding from the apertures (disappears in acetolysis), which they considered homologous to the oncus (a lens-shaped structure occurring just inside of the aperture). They suggest the term ‘protruding oncus’ for this structure and note that it is present in Keetia, Psydrax and most subgenera of Canthium (subgenus Canthium was not investigated). From pictures in Verdcourt (1987) and Igersheim (1989) it is also noticeable on pollen of Pyrostria, Fadogia and Vangueria. It is known from several other rubiaceous groups, e.g. Myrmecodia (Robbrecht, 1988) and Ophiorrhiza (Weber & Igersheim, 1994). Further studies on nonacetolysed pollen would very likely reveal the structure from more taxa. Also of interest is the presence of sporopollenin bridges that unite the pollen into large masses consisting of hundreds of pollen in the genera Keetia and Psydrax (Igersheim, 1989). It is not known if this is a regular feature of these genera or if it is only a product of faulty development; it was not reported in other studies of these genera (Tilney & van Wyk, 1997; Lens et al., 2000b). The presence of baculae in several taxa (Igersheim, 1989; Tilney & van Wyk, 1997; Lens et al., 2000b), especially in the large-flowered group, is potentially useful. It is known from Canthium subgenus Afrocanthium, Cuviera, Fadogia, Fadogiella, Lagynias, Pygmaeothamnus, Vangueria and Vangueriopsis. PHYLOGENY As discussed above, the morphological variation in Vanguerieae is in many cases continuous, with the consequence that there are few characters easily divided into discrete states. This follows a similar pattern to many other species-rich tropical groups, e.g. Strobilanthineae in the Acanthaceae (Carine & Scotland, 2002) and Andira in the Fabaceae (Pennington, 1996). This lack of useful morphological characters complicates the delimitation of genera. In some cases single, easily seen and measured characters have been used to delimit genera to the exclusion of other ‘more difficult’ characters. When this ‘cardinal character’ has been subject to parallel evolution, the resulting genera have not been monophyletic. An example of this is the polyphyletic genus Tapiphyllum (Fig. 1), mostly defined by the presence of velvety indumentum obscuring the leaf surface. For Rytigynia, also a polyphyletic genus, several characters have been used to delimit the genus, but the characters have been too general and shared by many other genera, i.e. a shrubby habit and few-flowered inflorescences. The lack of a unique synapomorphy for Rytigynia has resulted in a polyphyletic genus. In the following discussion we focus on groups that receive strong support from the phylogenetic analysis, and discuss how these groups compare to the traditionally recognized genera. We also indicate morphological characters that could be used to delimit these groups. Spiny group The spiny group is a strongly supported (100; 100) suprageneric group. It forms a strongly supported clade (A; Fig. 1) together with the large-flowered group. Apart from Pygmaeothamnus cf. chamaedendrum, all species in the spiny group are trees or shrubs, have paired leaves, domatia and paired supraaxillary spines that are, with the exception of Canthium inerme, in combination with brachyblasts. Spines are rare in the tribe outside this group, so it was therefore given an informal name based on this structure (Lantz et al., 2002). Plectroniella and Meyna are plurilocular; the other taxa have two locules. Pygmaeothamnus cf. chamaedendrum is very different from the other species in the group. It is a geofrutex that lacks spines, brachyblasts and domatia. Six genera are represented in the spiny group. The clade is largely unresolved and presents several taxonomic problems. Three Canthium species representing two different subgenera belong in the group; Canthium coromandelicum (the type species of Canthium) and C. glaucum in subgenus Canthium and C. inerme, the type of subgenus Lycioserissa. Species from the other subgenera of Canthium are found in distantly related clades, and the genus needs to be restricted. Rytigynia bugoyensis and R. mrimaensis are not close to the type species of the genus (R. senegalensis), and should be removed from Rytigynia. Only the clade consisting of Rytigynia bugoyensis and Vangueriella spinosa has strong support (92; 96). This confirms the suggestion of Bridson (1992) that Rytigynia bugoyensis possibly should be transferred to Vangueriella sect. Stenosepala. The two South African species Canthium ciliatum and Plectroniella armata also form a clade, but with weaker support (63; 77), an affinity suspected also from morphology (Bridson, 1992). Based on morphology, Meyna has been suggested as a likely candidate for inclusion in Canthium subgen. Canthium (Bridson, 1992). Meyna, together © 2004 The Linnean Society of London, Botanical Journal of the Linnean Society, 2004, 146, 257–283 PHYLOGENY OF VANGUERIEAE with Canthium subgen. Canthium, share a distribution with representatives in both Asia and on the African mainland. No other genus in the group is found in Asia, and this could suggest a common origin for Meyna and Canthium subgen. Canthium. The inclusion of Rytigynia mrimaensis in Rytigynia has earlier been questioned and it has been suggested to be close to some Indian species of Canthium (Group 1; Bridson, 1992). None of these species are sampled here, but they all have paired supra-axillary spines in combination with brachyblasts and thus most likely belong in the spiny group. The unresolved nature of the clade makes it difficult to suggest how to deal with the polyphyletic genera. One solution would be to sink all taxa in the spiny group into Canthium. Canthium is one of the first genera described in Vanguerieae (Lamarck, 1785) and C. coromandelicum is the type species of the genus. The group is strongly supported (100; 100), but would be very heterogeneous morphologically. This is mainly due to Pygmaeothamnus, which differs in a number of characters from the other taxa in the tribe. A generic description of this group would include most of the morphological variation present in the whole tribe. If Pygmaeothamnus were excluded, the resulting group would be much better delimited in a morphological sense. This group would be characterized by the presence of paired supra-axillary spines, domatia and usually brachyblasts and thus be morphologically homogeneous. We suggest that Canthium is restricted to the spiny group with the exception of Pygmaeothamnus cf. chamaedendrum. In view of the very different morphology of this species compared to the other species in the clade, we await more material, preferable flowering, before we draw any conclusions about the taxonomy. We make combinations at a generic level and also make new combinations for the species included in the analysis, but avoid making combinations for any species not sampled. No modern revision exists for Meyna and it is has been suggested that the species are poorly delimited (Verdcourt & Bridson, 1991). It should be noted that we only sample one Asiatic species (C. coromandelicum) but the majority of the species congruent with our view of Canthium are Asiatic (see Bridson, 1992). In morphology these species are entirely in accordance with our definition of Canthium, i.e. they have paired supra-axillary spines in combination with brachyblasts and have, apart from the species in Meyna, already names in Canthium. Thus, our circumscription of Canthium disrupts the current taxonomy of these species to only a minor extent. It has been suggested that smaller groups within the Asiatic Canthium may be recognized (Bridson, 1992), but we consider the differences between these groups to be small and that the advantages of a strongly delimited 273 Canthium far outweigh the fact that these groups are not formally recognized. If their monophyly is eventually confirmed, they may be recognized at the subgeneric level. There remains also a number of Asiatic and Pacific species currently in Canthium that need to be transferred to Cyclophyllum, Psydrax or Pyrostria (Bridson, 1985, 1987b; Smith & Darwin, 1988). These genera are already strongly delimited in a morphological sense, and with the new circumscription of Canthium suggested here, this is true also for Canthium. What is needed now is that all non-African species of Canthium should be examined and formally assigned to one of these genera. Large-flowered group Multidentia, a genus with a wide distribution in tropical Africa, receives strong support (91; 95) in the analysis. A combination of conspicuous tertiary nerves, a well-developed calyx limb tube and thick, woody pyrenes makes the genus morphologically distinct. Multidentia, Canthium oligocarpum and Cuviera angolensis are at unresolved basalmost positions in a moderately supported (73; 82) clade (E; Fig. 1) whose members, with few exceptions, have large fruits and smooth, retrorse corolla tube hairs. Canthium oligocarpum, a member of Canthium subgen. Lycioserissa, is restricted to high-altitude evergreen forests in eastern and southern Africa. It is not close to either the type of Canthium or to C. inerme, the type of subgenus Lycioserissa, both in the spiny group, and while it is certain it cannot remain in Canthium, the unresolved position does not enable us to suggest a generic placement for the species. Cuviera is a genus of about 20 species centred in West Africa, exhibiting many features rare in the tribe such as myrmecophyly and swollen hairy styles (e.g. Hallé, 1959). A moderately supported (76; 81) clade (clade C; Fig. 1) positions Pygmaeothamnus zeyheri and Robynsia glabrata at unresolved basal positions. Pygmaeothamnus zeyheri is a geofrutex with a widespread distribution in central, eastern, and southern Africa. The other member of the genus (P. chamaedendrum; see spiny group) has a restricted distribution in north-eastern South Africa. Robynsia is a monotypic West African genus showing some morphological affinity with Cuviera, such as large many-flowered inflorescences with large bracts and bracteoles. Vangueriopsis cf. longiflora, forms a weekly supported (58; 68) clade together with the strongly supported (87; 95) Vangueria group. Vangueriopsis (as circumscribed by Verdcourt, 1987) is a small genus of four species with a mostly Guineo-Congolian distribution (Robbrecht, 1996) but also occurring in East Africa (Verdcourt & Bridson, 1991). A member of the genus, V. lanciflora, was included in an earlier analy- © 2004 The Linnean Society of London, Botanical Journal of the Linnean Society, 2004, 146, 257–283 274 H. LANTZ and B. BREMER sis (Lantz et al., 2002), where it grouped with Vangueria madagascariensis and Vangueria infausta with strong support. We are uncertain about the identity of the specimen used in the earlier analysis, and it was therefore not included in the present study. Vangueriopsis exhibits a number of floral features rare in the tribe. Corolla lobes are long and linear, usually > 15 mm, anthers are exserted on long filaments, and hairs are present on the anthers. These characters are not shared by the Vangueria group, an assemblage of several genera with especially unclear generic limits (Verdcourt, 1981; Bridson, 1996, 1998). All taxa in the Vangueria group lack domatia (except a rare occurrence in V. apiculata) and have three to five locules. Several nodes are collapsed in the strict consensus tree and the support for several clades is low, but Tapiphyllum is polyphyletic and the monophyly of Vangueria is uncertain. Lagynias is monophyletic with strong support (100; 100), but the two included species are close and the monophyly of Lagynias is not really tested until more aberrant species such as L. monteiroi (Oliv.) Bridson are included. Sister to Lagynias is a moderately supported (78; 83) group that includes several Vangueria species of which the type species, V. madagascariensis, is one. The Fadogia–Rytigynia group is a strongly supported (100; 100) group sister to the other taxa in the large-flowered clade. Fadogia is paraphyletic with Fadogiella nested within it; the genera also share several morphological features. Synapomorphic characters for Fadogia including Fadogiella are geofrutescent habit, whorled leaves (only sometimes in Fadogiella) and an absence of domatia. Rytigynia is highly polyphyletic and is, in addition to the Fadogia– Rytigynia group, also represented in the spiny group and in the Vangueria group. The genus does not form a monophyletic group even within the Fadogia– Rytigynia group. Rytigynia senegalensis, the type species of Rytigynia, has an isolated position as sister to Fadogia/Fadogiella. Rytigynia bagshawei is sister to all other taxa in the Fadogia–Rytigynia group. It is within the large-flowered group that we encounter some of the most difficult issues concerning the classification of Vanguerieae. At least 15 genera are represented, there are cases of incongruence between the nuclear and plastid datasets, several nodes are weekly supported or collapsed, and several genera are para- or polyphyletic. These issues will be investigated in greater detail in another study (work in progress) in which more species are sampled and more data are added to better resolve the phylogeny of the group. This will hopefully enable us to elucidate the taxonomy of the group. Canthium subgen. Afrocanthium, Keetia and Psydrax. Canthium received a lot of attention during the pro- duction of the Vanguerieae part of the Flora of Tropical East Africa (Verdcourt & Bridson, 1991). The transfer of many species from the large genus Canthium created several smaller, better-delimited genera. Psydrax and Keetia are two such genera (Bridson, 1985, 1986). The remaining species of Canthium were divided into subgenera (Verdcourt & Bridson, 1991; Bridson, 1992) to reflect the fact that there were still large morphological differences present in the genus and to avoid premature name changes. Canthium subgen. Afrocanthium, Keetia, and Psydrax are all supported as monophyletic in the phylogeny (Fig. 1). Keetia has been thought to be related to Psydrax (Bridson, 1985, 1986) and shares many morphological characters with that genus, including style at least twice the length of the corolla tube (except P. kraussioides of subgenus Phallaria), style deeply recessed into pollen presenter, hairy disks (not all taxa), cotyledons orientated parallel to the ventral face of seed, and the occurrence of sporopollenin bridges in the pollen. Keetia receives strong support (100; 100) in the phylogeny and is well delimited in a morphological sense. All species are climbers, have a lid-like area on the pyrenes and probably have smooth retrorse hairs in the corolla tube, although this last character has not been confirmed for species not included here. Two subclades are supported, but apart from the slightly longer calyx lobes in the strongly supported clade of K. gueinzii and K. lukei, morphological support is lacking for this division. The support for Psydrax is lower (70; 77), which is surprising considering the number of morphological synapomorphies shared by the species. Unicellular hairs, no hairs inside the stipules, and anthers exserted on reflexed filaments are all characters that are synapomorphic for the genus. The species also share a region of 40 bases in trnT-F (except possibly P. kraussioides; see Results) that is not alignable with other taxa. It does not appear to be an inversion and is probably best explained by intramolecular recombination (Kelchner, 2000). This synapomorphic molecular character further supports the monophyly of Psydrax. The genus is the largest in Vanguerieae comprising over 100 species and has a distribution equal to that of the tribe. The Asiatic species P. nitidum is nested within a clade of African species, suggesting an African origin of Psydrax, but the support for this relationship is weak and we only include a single species from Asia. Keetia and Canthium subgen. Afrocanthium are strongly supported (89; 94) as sister taxa. However, apart from a lack of hairs within the stipules that is shared by K. gueinzii and C. mundianum, the morphology gives little indication for this relationship. We know of no morphological synapomorphy for this group, and the two groups should remain separate. © 2004 The Linnean Society of London, Botanical Journal of the Linnean Society, 2004, 146, 257–283 PHYLOGENY OF VANGUERIEAE Canthium subgen. Afrocanthium receives strong support from the phylogeny (100; 100). The subgenus is restricted to eastern and southern Africa and in the most recent revision 20 species are recognized (Bridson, 1992). Eight species are included in this analysis sampled from Kenya (C. keniense, C. lactescens, C. pseudoverticillatum), Tanzania (C. burttii, C. siebenlistii), Malawi (C. parasiebenlistii), and South Africa (C. gilfillanii, C. mundianum) and thus represent a broad geographical range. The clade is divided into two clades; one strongly supported (100; 100) consisting of the two South African species C. mundianum and C. gilfillanii (perhaps conspecific and best differentiated at the infraspecific level) and one more moderately supported (69; 80) in which the East African and Zambezian species are included. Canthium siebenlistii and C. parasiebenlistii, considered close (Verdcourt & Bridson, 1991), also form a strongly supported clade (97; 98), as does C. keniense and C. pseudoverticillatum (96; 98), but the relationships within the Afrocanthium clade are otherwise weakly supported or unresolved. The subgenus is recognized by a combination of absence of dark connective on the stamens, inflorescences borne at nodes from which the leaves have fallen, usually evident lenticels, and very short calyx limb tubes. A style that usually only slightly exceeds the corolla tube sets it apart from Keetia and Psydrax. Within the subgenus there is variation especially in the size of leaves, type of stipules and shape of fruits, and the majority of this spectrum of variation is exhibited by the species sampled. Canthium subgen. Afrocanthium is a morphologically distinct and strongly supported subgenus clearly separate from the type species C. coromandelicum and should be given generic rank. The type of the subgenus, C. lactescens, is strongly supported as a member of the clade and the name Afrocanthium is therefore applicable to the clade. Cyclophyllum and the dioecious group The clade consisting of Pyrostria, Cyclophyllum and closely related genera is strongly supported (94; 97). Leroya, Neoleroya, Peponidium and Pseudopeponidium are restricted to Madagascar, Canthium subgen. Bullockia to the African mainland, and Pyrostria is represented both in Africa, on Madagascar, on the Mascarenes and the Comoro Islands, and in southeast Asia (Bridson, 1987b). Scyphochlamys is a monotypic genus restricted to the small island of Rodriguez located east of Madagascar. Cyclophyllum is a Pacific and Australian genus (Reynolds & Henderson, 2001) and is thus geographically separated from the other genera in the clade. Pseudopeponidium (Bridson, 1987b), Leroya, Neoleroya and Peponidium (Schatz, 2001) have been formally included in Pyrostria, but no individual combinations have been made and we 275 therefore use the old generic names in the following discussion. The main morphological characters shared by the genera are fleshy corollas and a large amount of moniliform hairs in the corolla throat congesting the opening. These characters are not shared by Canthium subgen. Bullockia (here represented by C. pseudosetiflorum) and Peponidium, but are otherwise present for all species in the clade. At the base of the clade is the hermaphrodite Cyclophyllum barbatum, a member of a genus that has been suspected to have an affinity with Pyrostria (Bridson, 1987b), and this is supported here. A strongly supported clade (99; 100) that includes all dioecious Vanguerieae taxa (Pyrostria bibracteata is a hermaphrodite) is sister to C. barbatum. This clade will be referred to as the dioecious group (Fig. 1). Apart from Canthium pseudosetiflorum, a member of the dioecious subgenus Bullockia, all species in the clade also have large paired bracts, or homologous structures, on the peduncles. The clade is largely unresolved but a few clades receive moderate to strong support. Two species from the African mainland, Pyrostria hystrix and P. phyllantoidea, form a moderately supported (75; 82) clade but a third species, P. bibracteata with a distribution in both Africa and on Madagascar, groups with the Malagasy species Pseudopeponidium asosa (90; 95). The two species of Peponidium also group together (71; 77), but at an unresolved basal position. Although strongly supported as a group, the dioecious clade is lacking internal support and we refrain from making any new combinations pending further data (work in progress). The only solution that receives any support is the inclusion of the whole dioecious clade in Pyrostria, which would necessitate an inclusion also of Scyphochlamys and Canthium subgen. Bullockia. PHYLOGENETIC AFFINITIES OF GENERA NOT INCLUDED IN THE PHYLOGENETIC ANALYSIS The following conclusions are based both on comparisons of actual material and on data taken from literature. Eriosemopsis was described by Robyns (1928) to accommodate a single species from South Africa, E. subanisophylla. It is a geofrutex with elongated calyx lobes, two locules in the ovary, pilose hairs at the opening of the corolla tube and a ring of retrorse hairs further down in the tube. All of these characters agree with Pygmaeothamnus zeyheri, that also occurs in South Africa. Eriosemopsis subanisophylla was probably described in a genus of its own and not included in Pygmaeothamnus owing to its thick indumentum and raised venation on the leaves, a unique combination in the tribe, but otherwise the diagnostic charac- © 2004 The Linnean Society of London, Botanical Journal of the Linnean Society, 2004, 146, 257–283 276 H. LANTZ and B. BREMER ters of both genera overlap and the Robyns (1928: fig. II) considered them closely related. Everistia was erected for a single Australian species, E. vaccinifolia (Reynolds & Henderson, 1999), based on the different morphology of this species compared to other genera occurring in Australia, i.e. Psydrax and Cyclophyllum. The genus is characterized by a multi-branched habit, young branches usually resembling spines, obscure nerves in the leaves and a deeply two-lobed pollen presenter. Of the two other genera occurring in Australia, it approaches Psydrax the most. Both genera have long, exserted filaments and delicate corollas (compared to the fleshy corollas of Cyclophyllum), but in Psydrax the filaments are usually reflexed. We find it likely that Everistia is related to Psydrax, but it remains to be seen whether or not Everistia is nested within Psydrax. Perakanthus is a monotypic genus endemic to peninsular Malaysia (Wong, 1989). Only a few other Vanguerieae genera, i.e. Psydrax (Bridson, 1985), Pyrostria (Bridson, 1987b), Meyna, and several unrevised groups currently with names in Canthium (Wong, 1989; Bridson, 1992) occur in south-east Asia, and none of them are similar to Perakanthus. The flowers of Perakanthus have slightly curved corolla tubes £12 mm long with external indumentum, some hairs present at the opening of the tube, and a ring of retrorse hairs close to the disk. Similar flowers are found in Ancylanthos, Lagynias and Tapiphyllum (Bridson, 1996), but these are African genera and are without exception plurilocular; Perakanthus has two locules. It remains to be settled whether Perakanthus is a rare representative of the Vangueria group in south-east Asia or whether it is related to any of the other genera occurring in the same area with which it has acquired several autapomorphies. Temnocalyx is a monotypic genus with a restricted distribution in south-west Tanzania (Verdcourt, 1981). Diagnostic features of the genus are a long and robust corolla tube which is glabrous both internally and externally, and an obconic pollen presenter which is continuous with the style (Verdcourt & Bridson, 1991). Similar long and externally glabrous corolla tubes are only found in Fadogia, but these species usually have hairs in the corolla tube. The only exception is F. fuchsoides, which sometimes lacks internal hairs. Furthermore, Fadogia is a morphologically welldelimited genus characterized by a number of features (see above) and Temnocalyx does not share these features. Based on these differences it is unlikely that Temnocalyx is close to Fadogia. The plurilocular ovary together with inflorescences borne at nodes from which the leaves have fallen, suggest an affinity with the Vangueria group. However, Temnocalyx has solitary flowers and domatia, and this is not known from the Vangueria group. Solitary flowers are common in Rytigynia, and this genus (i.e. excluding the species known not to belong in the Fadogia–Rytigynia group) also has short calyx lobes, a feature shared with Temnocalyx, but large flowers are unknown in the genus and the inflorescences are usually subtended by mature leaves. A few species in the dioecious group also have a pollen presenter continuous with the style, but these pollen presenters are usually small and globose, not obconic and deeply divided by the stigmatic lobes as in Temnocalyx. In a tribe diagnosed by the presence of a usually globose or cylindrical pollen presenter with the style recessed into the presenter, this is of special interest, and could possibly imply a plesiomorphic state. The otherwise unknown combination of characters of Temnocalyx and the unique nature of the pollen presenter makes it hard to draw any conclusions concerning the phylogenetic affinities of the genus. At the moment, it is uncertain whether Temnocalyx is a genus that has acquired several autapomorphic characters or if it perhaps has retained some plesiomorphic features. CONCLUSIONS The well resolved and in many cases also strongly supported phylogeny presented here makes it possible for us to make several statements concerning the classification of Vanguerieae. The phylogenetic affinities of most Vanguerieae genera are now rather clear. Only four currently accepted genera are not included in our analysis. For two of these, i.e. Eriosemopsis and Everistia, there is morphological evidence that enables us to suggest possible relationships, but for Perakanthus and Temnocalyx molecular data are needed to ascertain their phylogenetic relationships. It is clear that the taxonomy now in use is in bad agreement with the phylogeny. Canthium, Rytigynia, Tapiphyllum and possibly Pygmaeothamnus are polyphyletic, Fadogia is paraphyletic. Some genera are monophyletic with strong support, i.e. Keetia, Lagynias (poorly sampled) and Multidentia, as is Canthium subgen. Afrocanthium, but most of the strongly supported clades are not identical to currently accepted genera. Four such clades (apart from the very inclusive large-flowered group) are discussed; the spiny group, the Vangueria group, the Fadogia–Rytigynia group, and the dioecious group. We suggest several changes to the classification of Vanguerieae (see Fig. 5), but also refrain from making new combinations when we consider data are lacking. We restrict Canthium to the spiny group and make new combinations in Canthium for the species sampled with the exception of Pygmaeothamnus cf. chamaedendrum. The affinity between this nonspiny geofrutex and a group which without exception © 2004 The Linnean Society of London, Botanical Journal of the Linnean Society, 2004, 146, 257–283 PHYLOGENY OF VANGUERIEAE 95 94 87 95 100 100 64 70 73 79 80 85 78 83 90 94 76 81 73 82 C E 91 95 99 100 100 100 A B 88 94 81 89 100 100 92 97 96 98 D 86 92 58 64 72 78 63 77 100 100 92 96 96 98 69 80 100 100 100 100 89 94 100 100 100 100 <50 <50 99 100 94 97 51 64 <50 <50 <50 <50 90 95 97 98 96 98 69 73 75 82 55 71 77 70 77 71 78 53 67 58 63 53 51 Outgroup Vangueria group Multidentia Large-flowered group 58 68 Mussaenda erythrophylla Ixora coccinea Alberta magna Ancylanthos rubiginosus a Ancylanthos rubiginosus b Lagynias dryadum Lagynias lasiantha Pachystigma pygmaeum Tapiphyllum cinerascens Vangueria apiculata Vangueria infausta Vangueria madagascariensis Vangueria parvifolia Rytigynia fuscosetulosa Tapiphyllum velutinum Vangueriopsis cf. longiflora Pygmaeothamnus zeyheri Robynsia glabrata Canthium oligocarpum Cuviera angolensis Multidentia concrescens Multidentia fanshawei Fadogia ancylantha Fadogia verdcourtii Fadogia cienkowskii Fadogia tetraquetra Fadogia elskensii Fadogiella stigmatoloba Rytigynia senegalensis Hutchinsonia barbata Rytigynia bagshawei Canthium ciliatum Canthium armatum Canthium coromandelicum Canthium glaucum Canthium inerme Canthium tetraphyllum Canthium mrimaense Canthium bugoyense Canthium stenosepalum Pygmaeothamnus cf. chamaedendrum Afrocanthium burttii Afrocanthium keniense Afrocanthium pseudoverticillatum Afrocanthium lactescens Afrocanthium parasiebenlistii Afroanthium siebenlistii Afroanthium gilfillanii Afrocanthium mundianum Keetia gueinzii Keetia lukei Keetia venosa Keetia zanzibarica Canthium pseudosetiflorum Leroya cf. richardiae Neoleroya verdcourtii Pseudopeponidium ampijoroense Pyrostria hystrix Pyrostria phyllantoidea Pseudopeponidium asosa Pyrostria bibracteata Scyphochlamys revoluta Peponidium horridum Peponidium sp. Cyclophyllum barbatum Psydrax kraussioides Psydrax livida Psydrax locuples Psydrax parviflora Psydrax nitidum Psydrax obovata Psydrax schimperiana Psydrax sp. A of F.T.E.A. 277 FadogiaRytigynia group Canthium sensu stricto Afrocanthium Keetia Dioecious group Psydrax Figure 5. The strict consensus tree from the combined analysis (ITS + trnT-F + morphology) with the new classification implemented. Bootstrap support values are above the branches, jack-knife below. A, B, C, etc. refer to clades mentioned in the text. Type species of currently recognized genera in bold. © 2004 The Linnean Society of London, Botanical Journal of the Linnean Society, 2004, 146, 257–283 278 H. LANTZ and B. BREMER constitutes spiny shrubs or trees needs to be investigated further. We raise the status of Canthium subgen. Afrocanthium to genus Afrocanthium and make new combinations in this genus for 17 species. The Vangueria group, the Fadogia–Rytigynia group, and the dioecious group all contain more than 50 species, and these groups should be better sampled before any new classification is suggested (work in progress). With this study we now have a better understanding of how the genera of Vanguerieae are related. What is needed now is better sampled studies within the monophyletic groups identified, and revisional work in which the results of phylogenetic studies are utilized. KILIFIENSE (BRIDSON) LANTZ COMB. Canthium kilifiense Bridson in Fl. Trop. E. Africa, Rubiaceae 3: 874. 1991. AFROCANTHIUM LACTESCENS (HIERN) LANTZ COMB. NOV. Canthium lactescens Hiern in Cat. Afr. Pl. 1: 511. 1898. AFROCANTHIUM MUNDIANUM (CHAM. & SCHLTDL.) LANTZ COMB. NOV. Canthium mundianum Cham. & Schltdl. in Linnaea 4: 131. 1829. AFROCANTHIUM NGONII (BRIDSON) LANTZ COMB. NOV. Canthium ngonii Bridson in Kew Bull. 47: 371, fig. 4. 1992. TAXONOMIC TREATMENT AFROCANTHIUM Canthium subgen. Afrocanthium is strongly supported as monophyletic and is also morphologically distinct. It is distanced from the type of Canthium and we therefore change the status of the subgenus to genus. That the subgenus was distinct from the type of Canthium was also known by the author of the subgenus (D. Bridson, pers. comm.). We accept the revision of Afrocanthium (Bridson in Verdcourt & Bridson, 1991; see also Bridson, 1992) without modifications to the circumscriptions of any species but make new combinations under Afrocanthium. We do not provide names for the species that in the revision were described as species A, B and C, but suggest that these species are formally described under Afrocanthium when complete material is available. See these publications also for a list of full synonyms. Afrocanthium (Bridson) Lantz & Bremer stat. nov. = Canthium subgen. Afrocanthium Bridson, Fl. Trop. E. Africa, Rubiaceae 3: 864. 1991. Description: See Bridson (1992). Type: Afrocanthium lactescens (Hiern) Lantz NEW AFROCANTHIUM NOV. AFROCANTHIUM (BRIDSON) LANTZ & BREMER COMBINATIONS IN AFROCANTHIUM PARASIEBENLISTII (BRIDSON) LANTZ COMB. NOV. Canthium parasiebenlistii Bridson in Fl. Trop. E. Africa, Rubiaceae 3: 870. 1991. AFROCANTHIUM PETERI (BRIDSON) LANTZ COMB. NOV. Canthium peteri Bridson in Fl. Trop. E. Africa, Rubiaceae 3: 873. 1991. AFROCANTHIUM PSEUDORANDII (BRIDSON) LANTZ COMB. NOV. Canthium pseudorandii Bridson in Kew Bull. 47: 385, fig. 12. 1992. AFROCANTHIUM PSEUDOVERTICILLATUM (S. MOORE) LANTZ COMB. NOV. Canthium pseudoverticillatum S. Moore, J. Bot. 43: 352. 1905. AFROCANTHIUM RACEMULOSUM (S. MOORE) LANTZ COMB. NOV. Canthium racemulosum S. Moore in J. Linn. Soc., Bot. 40: 87. 1911. AFROCANTHIUM RONDOENSE (BRIDSON) LANTZ COMB. NOV. Canthium rondoense in Fl. Trop. E. Africa, Rubiaceae 3: 868. 1991. AFROCANTHIUM BURTTII (BULLOCK) LANTZ COMB. NOV. Canthium burtii Bullock in Kew Bull. 1933: 146. 1933. AFROCANTHIUM SALUBENII (BRIDSON) LANTZ COMB. NOV. Canthium salubenii Bridson in Kew Bull. 47: 377, fig. 7. 1992. AFROCANTHIUM GILFILLANII (N.E. BR.) LANTZ COMB. NOV. Plectronia gilfillanii N.E. Br. in Bull. Misc. Inform. Kew 1906: 105. 1906. AFROCANTHIUM SHABANII (BRIDSON) LANTZ COMB. NOV. Canthium shabanii Bridson in Fl. Trop. E. Africa, Rubiaceae 3: 873. 1991. AFROCANTHIUM KENIENSE (BULLOCK) LANTZ COMB. NOV. Canthium keniense Bullock in Kew Bull. 1932: 377. 1932. AFROCANTHIUM SIEBENLISTII (K. KRAUSE) LANTZ COMB. NOV. Plectronia siebenlistii K. Krause in Bot. Jahrb. Syst. 57: 35. 1920. © 2004 The Linnean Society of London, Botanical Journal of the Linnean Society, 2004, 146, 257–283 PHYLOGENY OF VANGUERIEAE AFROCANTHIUM VOLLESENII (BRIDSON) LANTZ COMB. NOV. Canthium vollesenii Bridson in Fl. Trop. E. Africa, Rubiaceae 3: 875. 1991. CANTHIUM Canthium is polyphyletic and the circumscription of the genus needs to be changed. We suggest that the genus is restricted to the spiny group with the exception of Pygmaeothamnus cf. chamaedendrum, a taxon that needs to be investigated further. New combinations at the generic level are made and we also make combinations for the species sampled in this study. It should be mentioned that the genus Plectronia was for a long time thought to be an earlier synonym of Canthium and many species now in Canthium were therefore described in this genus. This was made in error and the name Plectronia should not be used. See Ross (1975) and Verdcourt (1987) for more details. Note that even with these combinations and the change of status for subgenus Afrocanthium suggested here, Canthium is not monophyletic. The issue of Canthium oligocarpum and Canthium subgen. Bullockia will be addressed in later publications (work in progress). The description included below is for Canthium s.s.; Afrocanthium, Canthium subgen. Bullockia, and C. oligocarpum are excluded. CANTHIUM Lam., ENCYCL. 1: 602. 1785. Lycioserissa Roem. & Schultes, Syst. Veg. 4: 353. 1819. Dondisia DC., Prodr. 4: 469. 1830. Psilostoma Klotzsch ex. Eckl. & Zeyher, Enum. pl. afric. austral. 361. 1837. Meyna Link, Jahrb. Gewächsk. 1(3): 32. 1820, syn. nov. Plectroniella Robyns, Bull. Jard. Bot. État. 11: 243. 1928, syn. nov. Vangueriopsis sect. Stenosepalae Robyns, Bull. Jard. Bot. État. 11: 243. 1928; Vangueriella sect. Stenosepalae (Robyns) Verdc., Kew Bull. 42: 189. 1987, syn. nov. Canthium subgen. Lyciocerissa (Roem. & Schultes) Bridson, Fl. Trop. E. Africa, Rubiaceae 3: 876. 1991, syn. nov. Description: Shrubs, small trees, scandent shrubs, or (rarely) lianas. Paired supra-axillary spines and brachyblasts usually present. LEAVES deciduous, paired on brachyblasts or sometimes, leaves paired on main shoots, glabrous to sparsely hairy, rarely with thick indumentum; domatia present as tufts of hair, as shallow hair-covered depressions, or (rarely) lacking; stipules small, triangular or with a broad base and a subulate to filiform appendage, pubescent within. FLOWERS 4- or 5-merous, one to twenty, in fasciculate to long pedunculate inflorescences; 279 bracts and bracteoles inconspicuous. Calyx limb tube short, not equalling disk, or sometimes more developed, repand, shortly dentate or lobes triangular to linear-lanceolate. Corolla white to green, externally glabrous or with a few scattered hairs, internally with tangled hairs at opening of tube and/or with retrorse hairs; lobes acute to acuminate, sometimes shortly apiculate. Anthers inserted at throat or clearly exserted, with darkened connective or not. OVARY 2–5-locular. Style slightly to long exserted. Pollen presenter coroniform to globose, rarely longer than wide, hollow at base. FRUIT a fleshy drupe with 1–5 pyrenes. New combinations in Canthium CANTHIUM ARMATUM (K. SCHUM.) LANTZ COMB. NOV. Vangueria armata K. Schum. in Bot. Jahrb. Syst. 28: 69. 1899. Plectroniella armata (K. Schum.) Robyns in Bull. Jard. Bot. État. 11: 243. 1928. CANTHIUM MRIMAENSE (VERDC.) LANTZ COMB. NOV. Rytigynia mrimaensis Verdc., in Kew Bull. 42: 170. 1987. CANTHIUM BUGOYENSE (K. KRAUSE) LANTZ COMB. NOV. Plectronia bugoyensis K. Krause in Wiss. Erg. Deutch. Zentr.-Afr. Exped., Bot. 1907–8: 327. 1911. Rytigynia bugoyensis (K. Krause) Verdc. in Bull. Jard. Bot. État. 50: 515. 1980. CANTHIUM STENOSEPALUM LANTZ NOM. NOV. Phallaria spinosa Schumach. & Thonn. in Beskr. Guin. Pl. 113. 1827. Vangueriella spinosa (Schumach. & Thonn.) Verdc. in Kew Bull. 42: 10. 1987. Canthium spinosum (Klotzsch ex Ecklon & Zeyher) Kuntze already exists. The species epithet is derived from Vangueriella sect. Stenosepalae Verdc. for which Vangueriella spinosa is the type. CANTHIUM TETRAPHYLLUM (SCHWEINF. EX. HIERN) BAILL. in Adansonia 12: 192. 1878. Vangueria tetraphylla Schweinf. ex. Hiern in Fl. Trop. Afr. 3: 152. 1877. Meyna tetraphylla (Schweinf. EX. Hiern) Robyns in Bull. Jard. Bot. État. 11: 233. 1928. ACKNOWLEDGEMENTS A number of people helped during the production of this manuscript, and we would like to thank K. Andreasen, K. Bremer, J. Kårehed, B. Oxelman and L. Tibell for suggestions and constructive criticisms and especially M. Thulin for help with nomenclature. We thank Diane Bridson and an anonymous reviewer for valuable comments on the manuscript. The authors © 2004 The Linnean Society of London, Botanical Journal of the Linnean Society, 2004, 146, 257–283 280 H. LANTZ and B. BREMER would like to thank the directors of the BR, K, LISC, MO, NYBG, S, TEF, and UPS for supplying plant material. DNA material was also received from M. Chase, R. Gereau and D. Lorence; their help is greatly appreciated. Special thanks go to the wonderful people at the National Botanical Garden in Zomba, Malawi, for kind assistance given to the first author during fieldwork. We also thank A. Bauer and N. Heidari for help with sequencing. Financial support was received from the Swedish Research Council (to B. B.), Letterstedts resestipendier (Uppsala University), the Royal Swedish Academy of Sciences, and Helge Ax:son Johnsons stiftelse (to H. L.), which is gratefully acknowledged. REFERENCES Anderson LE. 1954. Hoyer’s solution as a rapid mounting medium for bryophytes. Bryologist 57: 242–247. Andreasen K, Bremer B. 2000. Combined phylogenetic analysis in the Rubiaceae-Ixoroideae: morphology, nuclear and chloroplast DNA data. American Journal of Botany 87: 1731–1748. Arènes J. 1960. A propos de quelques genres Malgaches de Rubiacées (Vanguériées et Gardenieés). Notulae Systematicae, Herbier du Muséum de Paris Phanérogamie 16: 6–41. Balkwill K, Sebola JR, Robinson ER. 1996. Sex expression in Canthium mundianum (Rubiaceae). In: van der Maesen LJG, van der Burgt XM, van Medenbach de Rooy JM, eds. The biodiversity of African plants. Dordrecht: Kluwer Academic Publishers, 650–655. Bremer B, Bremer K, Heidari N, Erixon P, Olmstead RG, Anderberg AA, Källersjö M, Barkhordarian E. 2002. Phylogenetics of asterids based on 3 coding and 3 non-coding chloroplast DNA markers and the utility of non-coding DNA at higher taxonomic levels. Molecular Phylogenetics and Evolution 24: 274–301. Bridson DM. 1985. The reinstatement of Psydrax (Rubiaceae, subfam. Cinchonoideae tribe Vanguerieae) and a revision of the African species. Kew Bulletin 40: 687–725. Bridson DM. 1986. The reinstatement of the African genus Keetia (Rubiaceae subfam. Cinchonoideae, tribe Vanguerieae). Kew Bulletin 41: 965–994. Bridson DM. 1987a. The recognition and recircumscription of the African genus Multidentia (Rubiaceae-Vanguerieae). Kew Bulletin 42: 641–654. Bridson DM. 1987b. Studies in African RubiaceaeVanguerieae: a new circumscription of Pyrostria and a new subgenus, Canthium subgen. bullockia. Kew Bulletin 42: 611–639. Bridson DM. 1992. The genus Canthium (RubiaceaeVanguerieae) in tropical Africa. Kew Bulletin 47: 353– 401. Bridson DM. 1996. The tropical African genus Ancylanthos (Rubiaceae-Vanguerieae) reconsidered. Kew Bulletin 51: 343–352. Bridson DM. 1998. Rubiaceae (tribe Vanguerieae). In: Pope GV, ed. Flora Zambesiaca. London: Royal Botanic Gardens, Kew, 211–377. Buckler IVES, Ippolito A, Holtsford TP. 1997. The evolution of ribosomal DNA: divergent paralogues and phylogenetic implications. Genetics 145: 821–832. Capuron R. 1969. A propos des Rubiacées-Vangueriées de Madagascar. Adansonia 2 (9): 47–55. Carine MA, Scotland RW. 2002. Classification of the Strobilanthinae (Acanthaceae): trying to classify the unclassifiable? Taxon 51: 259–279. Coates Palgrave M. 2002. Keith Coates Palgrave trees of Southern Africa, 3rd edn. Cape Town: Struik Publishers. Goloboff PA. 1991. Homoplasy and the choice among cladograms. Cladistics 7: 215–232. Hallé N. 1959. Sur les Cuviera (Rubiacées) d’Afrique intertropicale et description pour ce genre de deux espèces et de deux variétés nouvelles. Bulletin de la Société Botanique de France 7–8: 342–348. Havard A, Verdcourt B. 1987. A pollen survey of Tapiphyllum (Rubiaceae-Vanguerieae). Kew Bulletin 42: 605–609. Holmgren PK, Holmgren NH, Barnett LC. 1990. Index herbariorum, Part 1: the herbaria of the world, 8th edn. New York: New York Botanical Garden, Bronx. Igersheim AF. 1989. Beiträge zur Klärung der Gattungsabgrenzungsprobleme innerhalb der Rubiaceae-Vanguerieae. D.Phil. thesis, University of Vienna. Igersheim AF. 1993. Gynoecium development in RubiaceaeVanguerieae, with particular reference to the ‘stylar-head’complex and secondary pollen presentation. Plant Systematics and Evolution 187: 175–190. Kelchner SA. 2000. The evolution of non-coding chloroplast DNA and its application in plant systematics. Annals of the Missouri Botanical Garden 87: 482–498. Kluge A, Farris JS. 1969. Quantitative phyletics and the evolution of the anurans. Systematic Zoology 18: 1–32. Lamarck JBAPM de. 1785. Encyclopédie méthodique 1(2). Paris. Lantz H, Andreasen K, Bremer B. 2002. Nuclear rDNA ITS data used to construct the first phylogeny of Vanguerieae (Rubiaceae). Plant Systematics and Evolution 230: 173–187. Lee D-C, Bryant HN. 1999. A reconsideration of the coding of inapplicable characters: assumptions and problems. Cladistics 15: 373–378. Lens F, Jansen S, Huysmans S, Robbrecht E, Smets E. 2000b. Pollen morphological variation in Vanguerieae (Ixoroideae-Rubiaceae). Grana 39: 90–102. Lens F, Jansen S, Robbrecht E, Smets E. 2000a. Wood anatomy of the Vanguerieae (Ixoroideae-Rubiaceae), with special emphasis on some geofrutices. IAWA Journal 21: 443–455. Lidén M, Fukuhara T, Axberg T. 1995. Phylogeny of Corydalis, ITS and morphology. Plant Systematics and Evolution 9 (suppl.): 183–188. Maddison DR, Maddison WP. 2001. MacClade 4: analysis of phylogeny and character evolution. Sunderland, MA: Sinauer Associates. Nepokroeff M, Bremer B, Sytsma KJ. 1999. Reorganization © 2004 The Linnean Society of London, Botanical Journal of the Linnean Society, 2004, 146, 257–283 PHYLOGENY OF VANGUERIEAE of the genus Psychotria and tribe Psychotrieae (Rubiaceae) inferred from ITS and rbcL sequence data. Systematic Botany 24: 5–27. Patterson C. 1988. Homology in classical and molecular biology. Molecular Biology and Evolution 5: 603–625. Pennington RT. 1996. Molecular and morphological data provide phylogenetic resolution at different hierarchical levels in Andira. Systematic Biology 45: 496–515. Popp M, Oxelman B. 2001. Inferring the history of the polyploid Silene aegaea (Caryophyllaceae) using plastid and homoeologous nuclear DNA sequences. Molecular Phylogenetics and Evolution 20: 474–481. Puff C, Robbrecht E, Buchner R, De Block P. 1996. A survey of secondary pollen presentation in the Rubiaceae. Opera Botanica Belgica 7: 369–402. Retief E. 2003. Rubiaceae. In: Germishuizen G, Meyer NL, eds. Plants of southern Africa: an annotated checklist. Strelizia 14, 825–841. Pretoria: National Botanical Institute. Reynolds ST, Henderson RJF. 1999. Vanguerieae A. Rich. ex Dum. (Rubiaceae) in Australia, 1. Everistia. Austrobaileya 5: 353–361. Reynolds ST, Henderson RJF. 2001. Vanguerieae A. Rich. ex Dum. (Rubiaceae) in Australia, 2. Cyclophyllum. Austrobaileya 6: 41–66. Robbrecht E. 1988. Tropical woody Rubiaceae. Opera Botanica Belgica 1: 1–271. Robbrecht E. 1996. Generic distribution patterns in subsaharan African Rubiaceae (Angiospermae). Journal of Biogeography 23: 311–328. Robyns W. 1928. Tentamen Monographiae Vanguerieae Generumque Affinium. Bulletin du Jardin Botanique de l’état Bruxelles 11: 1–359. Ross JH. 1975. The typification of Lycium inerme. Bothalia 11: 491–493. Schatz GE. 2001. Generic tree flora of Madagascar. London: Royal Botanic Gardens, Kew/ St. Louis, MO: Missouri Botanical Garden. Skottsberg C. 1945. The flower of Canthium. Arkiv för Botanik 32A: 1–12. Smith AC, Darwin SP. 1988. Rubiaceae. In: Smith AC, eds. Flora Vitiensis Nova 4. Lawai, Kauai, Hawaii: The Pacific Tropical Botanical Garden, 143–376. Swofford DL. 2002. PAUP*: phylogenetic analysis using parsimony (*and other methods), v.4. Sunderland, MA: Sinauer Associates. Taberlet P, Gielly L, Pautou G, Bouvet J. 1991. Universal primers for amplification of three non-coding regions of chloroplast DNA. Plant Molecular Biology 17: 1105–1109. Terrell EE. 2001. Taxonomic review of Houstonia acerosa and H. palmeri, with notes on Hedyotis and Oldenlandia (Rubiaceae). SIDA Contributions to Botany 19: 913–922. Tilney PM. 1986. The taxonomic significance of anatomical and morphological characters in the southern African species of Canthium Lam. (Rubiaceae). D.Phil. thesis, University of Pretoria. Tilney PM, Kok PDF, van Wyk AE. 1990. The taxonomic significance of anatomical characters of the leaf in the south- 281 ern African species of Canthium s.l. (Rubiaceae). South African Journal of Botany 56: 363–382. Tilney PM, van Wyk AE. 1997. Pollen morphology of Canthium, Keetia and Psydrax (Rubiaceae: Vanguerieae) in southern Africa. Grana 36: 249–260. Tilney PM, van Wyk AE, Kok PDF. 1988. The taxonomic significance of anatomical characters of the stem in the southern African species of Canthium s.l. (Rubiaceae). South African Journal of Botany 54: 585–595. Urbatsch LE, Baldwin BG, Donoghue MJ. 2000. Phylogeny of the coneflowers and relatives (Heliantheae: Asteraceae) based on nuclear rDNA internal transcribed spacer (ITS) sequences and chloroplast DNA restriction site data. Systematic Botany 25: 539–565. Verdcourt B. 1958. Remarks on the classification of the Rubiaceae. Bulletin du Jardin Botanique de l’état Bruxelles 28: 209–281. Verdcourt B. 1981. Notes on African Rubiaceae. Kew Bulletin 36: 493–557. Verdcourt B. 1983. Notes on Mascarene Rubiaceae. Kew Bulletin 37: 563–570. Verdcourt B. 1987. Notes on African Rubiaceae-Vanguerieae. Kew Bulletin 42: 123–199. Verdcourt B, Bridson DM. 1991. Rubiaceae (Part 3). In: Polhill RM, ed. Flora of tropical East Africa. Rotterdam/ Brookfield: A.A. Balkema, 749–956. Weber M, Igersheim A. 1994. ‘Pollen buds’ in Ophiorrhiza (Rubiaceae) and their role in Pollenkitt release. Botanica Acta 107: 257–262. Weberling F. 1965. Typology of inflorescences. Journal of the Linnean Society. Botany 59: 215–221. White TJ, Bruns T, Lee S, Taylor J. 1990. Amplification and direct sequencing of fungal ribosomal RNA genes for phylogenetics. In: Innis M, Gelfand D, Sninsky J, White T, eds. PCR protocols: a guide to methods and applications. San Diego: Academic Press, 315–322. Wiens JJ. 1999. Character analysis in morphological phylogenetics: problems and solutions. Systematic Biology 50: 689–699. Wong KM. 1989. Rubiaceae. In: Ng FSP, ed. Tree flora of Malaya 4. Petaling Jaya: Longman Malaysia Sdn. Berhad, 324–425. APPENDIX 1 Morphological characters and character states. All multistate characters are unordered. Characters 11 and 27 are used to define the ingroup. All characters are discussed in more detail in the text apart for numbers 11, 25 and 27, which are only mentioned here. 1. 2. 3. 4. Habit (0) geofrutex; (1) shrub or tree; (2) climber Supra-axillary spines (0) absent; (1) present Reduced lateral branches (0) absent; (1) present Sex distribution (0) hermaphrodite; (1) dioecious; (2) gynodioecious. © 2004 The Linnean Society of London, Botanical Journal of the Linnean Society, 2004, 146, 257–283 282 H. LANTZ and B. BREMER Leaves 5. Leaf arrangement (0) paired; (1) in whorls of three or more 6. Indumentum (0) absent; (1) present 7. Amount of indumentum (0) not obscuring the surface; (1) obscuring the surface 8. Domatia (0) absent; (1) present 9. Hairtuft inside stipules (0) absent; (1) present 10. Stipule apex widest above midpoint (0) no; (1) yes, at least sometimes. Inflorescence 11. Inflorescence position (0) terminal; (1) axillary 12. Axillary inflorescences borne at leafless node (0) no; (1) yes 13. Bracts present on peduncle, enclosing the young inflorescence (0) absent; (1) present 14. Number of flowers per inflorescence (0) 1–2; (1) 3–10; (2) 11–20; (3) > 20 Calyx 15. Limb tube short, disk visible from side (0) no; (1) yes 16. Lobe length (0) < 1 mm, length/width ratio < 1; (1) > 1 mm, length/width ratio > 1 Corolla 17. Number of lobes (0) 4; (1) 5; (2) 6 18. Tube length (0) 8 mm or less; (1) more than 8 mm 19. Lobe apex tailed (0) no; (1) yes 20. Smooth hairs in the corolla tube (0) absent; (1) present 21. Type of smooth hairs (0) retrorse, straight, in a well defined ring; (1) diffusely spreading, rarely retrorse 22. Verrucose hairs in corolla tube (0) absent; (1) present Androecium 23. Filament reflexed (0) no; (1) yes 24. Dark connective (0) absent; (1) present Disk 25. Disk (0) glabrous; (1) indumentum present Gynoecium 26. Style length (0) less than twice as long as tube; (1) at least twice as long as tube 27. Pollen presenter (0) absent; (1) present 28. Pollen presenter type (0) hollow, style recessed into presenter; (1) solid, style not recessed into presenter 29. Number of locules (0) 2; (1) 3–5; (2) > 5 Fruit 30. Size (0) £ 1 cm long; (1) > 1 cm long APPENDIX 2 Morphological matrix. Polymorphic characters are abbreviated as follows: A = 0,1; B = 1,2; C = 2,3; D = 0,1,2, E = 1,2,3, ? = unknown state, – = inapplicable. Character number Species 1 Alberta magna Ixora coccinea Mussaenda erythrophylla Ancylanthos rubiginosus Canthium burttii Canthium ciliatum Canthium coromandelicum Canthium gilfillanii Canthium glaucum Canthium inerme Canthium keniense Canthium lactescens Canthium mundianum Canthium oligocarpum Canthium parasiebenlistii Canthium pseudosetiflorum Canthium pseudoverticillatum Canthium siebenlistii Cuviera angolensis Cyclophyllum barbatum Fadogia ancylantha Fadogia cienkowskii 100000-0100-03011100-000000-00 100000-0100-03010100-0?1000–00 20000100100–0301110010?1000–01 0000A1A?1011010111110101001011 10?00A?1?0110B10100???0??01?00 11100101101100101010–101001001 111000–11011011000011001001000 100?010110110B101000-100001000 B11000–110110100A000-101001001 1A0000-11011011010011001001001 100000–110110B10100???0?001001 1010010A10110310A000-100001000 10020A0100110B1A1000-100001000 100000–110100E0A1010-001001001 10100A0110110B10A00???000?1000 1011A10010100?011000–101001001 B0100A0110110B10A000-101001000 10000A0110110B100000-100011000 1?0000–01110020110110001001011 100000-A101001101000-101001100 0000AA001010010011011000001011 0000111?1010010110111101001011 10 20 30 © 2004 The Linnean Society of London, Botanical Journal of the Linnean Society, 2004, 146, 257–283 PHYLOGENY OF VANGUERIEAE 283 APPENDIX 2 Continued Character number Species 1 Fadogia elskensii Fadogia tetraquetra Fadogia verdcourtii Fadogiella stigmatoloba Hutchinsonia barbata Keetia gueinzii Keetia lukei Keetia venosa Keetia zanzibarica Lagynias dryadum Lagynias lasiantha Leroya cf. richardiae Meyna tetraphylla Multidentia concrescens Multidentia fanshawei Neoleroya verdcourtii Pachystigma pygmaeum Peponidium horridum Peponidium sp. Plectroniella armata Pseudopeponidium ampijoroense Pseudopeponidium asosa Psydrax kraussioides Psydrax livida Psydrax locuples Psydrax nitidum Psydrax obovata Psydrax parviflora Psydrax schimperiana Psydrax spA Pygmaeothamnus cf. chamaedendrum Pygmaeothamnus zeyheri Pyrostria bibracteata Pyrostria hystrix Pyrostria phyllantoidea Robynsia glabrata Rytigynia bagshawei Rytigynia bugoyensis Rytigynia fuscosetulosa Rytigynia mrimaensis Rytigynia senegalensis Scyphochlamys revoluta Tapiphyllum cinerascens Tapiphyllum velutinum Vangueria apiculata Vangueria infausta Vangueria madagascariensis Vangueria parvifolia Vangueriella spinosa Vangueriopsis cf. longiflora 0000111?1010010110111001001010 00001A001010010110111101001010 00001AA010100100111???0??01010 0010A11?1011010010011001001010 B?0001011010000111111001001010 200001010010030010010101111000 200000–111100200100???01111001 2000010110100300D0010101111000 B000010110100300A0010101111000 100001001011010110110?01011011 1?00010010110B0110110101011011 100?010110111?01??????????1?21 B1100A0A1011011010011100011011 0000A0-01010020110010101001001 B00000-11010030011010101001001 100100–010101?00001???01001?01 0000AA001011010110110000001011 1001010010101?01101???01?010B1 100?00–110101?01100111000010?? 11100101101102000000–101011011 100100-A00101?11?000–10100102? 100100–110101?11?000–101001120 200000-A0?100D1010110110001000 1000010100110E0000011011011000 100000–100100E0000011011011000 100000–10010010010011011111000 100000–001100300A0011011011000 100000–10010030000011011111000 100000-A01100C0010011011011000 100000–100100B0000011010011000 000?010010110B01?0??????0?1?01 000011001011020110110101001001 100000-A10101E100000-101001100 101100-A10101?10001???010?1?00 101100–010101?11001???010?1?00 1?0001001011030111110001001011 10000101101000101010–101001000 1110010110110111101???00001000 1000010010100101100???00001011 11100101101?00?0?01???0??01?01 101000–11110000010011001011010 100100–000101?00A00???0?0?1?21 A000A11?1011010110110000001010 1010011?10110B0110110100001011 100000-A1011031110110100001011 100001001011031110A10100001011 100000-0101103111000-100001011 10A0011?1011011110A10100101011 111000–110110C111010-100011001 100?010?10110?01?0????????1?01 10 © 2004 The Linnean Society of London, Botanical Journal of the Linnean Society, 2004, 146, 257–283 20 30