| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Systematics and Phytogeography |
The Herbarium, Royal Botanic Gardens, Kew, Richmond, Surrey, TW9 3AB, UK; School of Biological and Chemical Sciences, Queen Mary, University of London, Mile End Road, London, E1 4NS, UK; Department of Botany and Plant Biotechnology, University of Johannesburg, PO Box 524, Auckland Park, 2006, Gauteng, South Africa; The Jodrell Laboratory, Royal Botanic Gardens, Kew, Richmond, Surrey, TW9 3DS, UK
Received for publication November 23, 2005. Accepted for publication January 17, 2007.
ABSTRACT
The circumscription of Coffeeae (Rubiaceae) and phylogenetic relationships within the tribe were evaluated using sequence data from four plastid regions (trnL-F intron, trnL-F intergenic spacer [IGS], rpl16 intron, and accD-psa1 IGS) and a morphological data set. Eleven candidates for inclusion in Coffeeae were examined using plastid data, and a further three were investigated using morphology alone. Based on previous phylogenetic analysis of the subfamily Ixoroideae, nine genera representing five tribes were used as outgroups. Our results support an enlarged circumscription for Coffeeae, containing 11 genera, viz. Argocoffeopsis, Belonophora, Calycosiphonia, Coffea, Diplospora, Discospermum, Nostolachma, Psilanthus, Tricalysia, Sericanthe, and Xantonnea. The inclusion of Diplospora and Tricalysia within Coffeeae, based on published molecular data, and the inclusion of Argocoffeopsis, Belonophora, Calycosiphonia, Discospermum, and Sericanthe, based on morphological evidence, are well supported. Nostolachma is newly transferred from Gardenieae subtribe Diplosporinae to Coffeeae, and Xantonnea from Octotropideae to Coffeeae. The exclusion of Bertiera from Coffeeae and placement in tribe Bertiereae is supported on the basis of molecular and morphological data. The removal of Diplospora and all other genera from Gardenieae subtribe Diplosporinae to Coffeeae and Octotropideae renders Diplosporinae superfluous. It is proposed that Xantonneopsis be transferred to Octotropideae; Petitiocodon is tentatively placed in Gardenieae. The monophyly of seven genera is supported, but Coffea is identified as paraphyletic in relation to Psilanthus on the basis of molecular and combined molecular and morphological data.
Key Words: accD-psa1 • Coffea • coffee • Coffeeae • molecular systematics • morphology • rpl16 • Rubiaceae • trnL-F
Coffea L. is the genus from which the beverage coffee is produced, with C. arabica L. (arabica coffee) the principal commercial species. The importance of coffee as an agricultural commodity is paramount: it is the world's most heavily traded commodity after oil in terms of monetary value and supports at least 20 million coffee farming families in more than 50 countries (Vega et al., 2003
). Given the commercial and social importance of the genus, it is alarming that we know so little about the genus and its species. In particular, the systematic relationships between Coffea species have not been extensively studied and the systematic position of the genus within Rubiaceae is poorly understood. The resolution of systematic relationships between Coffea and Psilanthus represents a key objective for the understanding of Coffea. Before this can happen, however, it is necessary to investigate the relationship between these two genera and other closely related groups within Rubiaceae.
Coffea comprises 96 species and occurs naturally in Africa, Madagascar, and the Mascarenes (Davis et al., 2006
). A close relationship between Coffea and Psilanthus Hook. f. is evident from studies covering several disciplines. Careful morphological study by Robbrecht and Puff (1986)
and Robbrecht (1988b
, 1994
) has shown that the two genera share several key features and on this basis were considered the component genera of tribe Coffeeae DC. (Robbrecht and Puff, 1986
). Couturon et al. (1998)
produced a fertile intergeneric hybrid via the crossing of C. arabica and P. ebracteolatus Hiern, and genetic correspondence has been further revealed by recent cytological studies (Lombello and Pinto-Maglio, 2003
, 2004
). Coffea and Psilanthus have also been the focus of several recent phylogenetic studies, using systematic data from various sources, including morphology (Stoffelen, 1998
; Davis et al., 2005
), random amplified polymorphic DNA (RAPD) (Lashermes et al., 1993
), sequences from plastid DNA (Cros, 1994
; Lashermes et al., 1996
; Cros et al., 1998
), and ITS sequences of nuclear ribosomal DNA (Lashermes et al., 1997
). At the species level, the studies of Lashermes et al. (1997)
and Cros et al. (1998)
have provided the most useful data: they were able to separate Coffea species into geographical groupings and to begin to examine the relationships between Coffea and Psilanthus species. Lashermes et al. (1997)
found that one Psilanthus species (P. travancorensis) was nested within Coffea and that there was limited sequence divergence between these two genera. They concluded that their ITS data did not support the recognition of two genera. On the basis of trnL-trnF sequence data, Cros et al. (1998)
concurred with Lashermes et al. (1997)
about the close relationship between Coffea and Psilanthus, although their tree topology shows an unresolved relationship between the two species of Psilanthus that they sampled (P. mannii and P. ebracteolatus) and Coffea. Cros et al. (1998)
and Lashermes et al. (1997)
did not include representatives of closely related genera in their studies, for example, as outgroups. However, broader studies of Rubiaceae (Ixoroideae) by Andreasen et al. (1999)
and Andreasen and Bremer (2000
, fig. 3) infer paraphyly of Coffea.
Potential relatives of Coffea should be found within Coffeeae, but at the present time there is no general agreement on the constituent genera of the tribe (Robbrecht and Puff, 1986
; Robbrecht, 1988b
, 1994
; Andreasen and Bremer, 1996
, 2000
; Persson, 2000
; Bridson and Verdcourt, 2003
) and few data on how they are related to one another (Andreasen and Bremer, 2000
).
The nomenclatural origin of tribe Coffeeae is De Candolle's Coffeaceae (De Candolle, 1807
), one of four tribes recognized by De Candolle in his early classification of Rubiaceae. In a later conspectus of the family by De Candolle (1830)
, which includes 13 tribes, it is evident that his Coffeeae (‘Coffeaceae') was very broadly circumscribed and included a large number of genera, many of which have since been placed in different tribes and subfamilies. De Candolle's characterization of Coffeeae translated and reorganized by us from the Latin diagnosis (De Candolle, 1830
, p. 472), is as follows: Trees or shrubs with opposite leaves; stipules interpetiolar, paired, adnate or free. Ovaries 1-locular, each bearing a single ovule. Fruits fleshy, containing two pyrenes, each containing a single seed with a bony-crustaceous surface (endocarp layer of pyrene), endocarp flat on the adaxial surface and usually sulcate on the abaxial surface, or fruits with one pyrene by abortion; endosperm hard/horny.
Richard (1830)
provided a similar concept for Coffeeae, although it should be noted that De Candolle (1830)
had access to Richard's unpublished manuscript and incorporated parts of the classification in order to avoid nomenclatural conflict (Stern, 1957
).
Further attempts to define Coffeeae were not made until much later. Leroy (1980)
included three genera in his informal ‘Caféiers' group, namely Coffea, Nostolachma T. Durand, and Psilanthus, although no direct reference was made to this tribe by Leroy. Robbrecht and Puff (1986
; see also Robbrecht, 1988b
, 1994
) provided the first rigorous and explicit delimitation for Coffeeae and included just two genera in the tribe, Coffea and Psilanthus. Their characterization of the tribe was based on the presence of 2-carpellate ovaries, each with a single ovule, axile placentation, a hard (horny/crustaceous) endocarp, seeds with a deep L- or T-shaped ventral groove on the adaxial surface (as seen in transverse section; endocarp and seed coat invaginated), and a seed coat exotesta consisting of thin, elongated parenchymatic cells usually containing many more or less isolated fibers (derived from Robbrecht, 1981
; Robbrecht and Puff, 1986
). In addition, Stoffelen et al. (1997)
showed that the pollen of Coffeeae are (2–)3–5-colporate (zonocolporate). A simple but efficient means of characterizing Coffeeae sensu Robbrecht and Puff (1986)
is of a tribe possessing "coffee beans," i.e., seeds with a groove on the flat side of the seed. The groove ramifies within the seed as an invagination through the endosperm into the center and is clearly evident when a coffee bean is cut in transverse section. The pyrene (horny/crustaceous endocarp) also has a deep, ventral groove that follows the invagination within the seed. Robbrecht and Puff (1986)
excluded Nostolachma (= Lachnastoma Korth.), which was associated with Coffea and Psilanthus by Leroy (1980)
. Nostolachma, together with Argocoffeopsis Lebrun, Calycosiphonia Pierre ex Robbr., Cremaspora Benth., Diplospora DC., Sericanthe Robbr., and Tricalysia A. Rich. ex DC. were transferred to Gardenieae DC. subtribe Diplosporinae Miq. by Robbrecht and Puff (1986)
. It was not explicitly stated by Robbrecht and Puff (1986)
why and how Gardenieae subtribe Diplosporinae was separated from Coffeeae, although it seems that the absence of characters found in Coffeeae (as discussed earlier) coupled with the presence of exotestal cells (with or without thickenings and lacking fibers) separated these two groups. Petitiocodon Robbr. and Xantonneopsis Pit. were added to this subtribe by Robbrecht (1988b)
, followed by Discospermum Dalz. (Ali and Robbrecht, 1991
; Robbrecht, 1994
) on the basis that they conformed to the circumscription of the subtribe (after Robbrecht and Puff, 1986
).
Molecular studies (Andreasen and Bremer, 1996
, 2000
; Persson, 2000
) have demonstrated that certain genera of Gardenieae subtribe Diplosporinae are closely related to Coffea and Psilanthus. On the basis of molecular data, Andreasen and Bremer (2000)
placed two genera of Gardenieae subtribe Diplosporinae, Diplospora and Tricalysia, with Coffea and Psilanthus, in Coffeeae. The genus Paracoffea J.-F. Leroy (Leroy, 1967
) used in that work is a synonym of Psilanthus (Davis, 2003
). Bertiera Aubl., formerly either a genus of unknown systematic position within Rubiaceae (Robbrecht, 1988b
) or placed within Gardenieae subtribe Gardeniinae (Robbrecht et al., 1994
), was also shown to be closely related to Coffea and Psilanthus (Andreasen and Bremer, 1996
, 2000
; Persson, 2000
). Andreasen and Bremer (2000)
placed Bertiera in Coffeeae, based on a moderately well-supported (BP = 76, b = 4) sister relationship to Coffea, Psilanthus, Diplospora, and Tricalysia. Andreasen and Bremer (2000)
placed two further members of Gardenieae subtribe Diplosporinae, Discospermum and Sericanthe, into Coffeeae, based on the morphological evidence (Schumann, 1891
; Ali and Robbrecht, 1991
) that the former was closely related to Diplospora and the latter to Tricalysia. Andreasen and Bremer (2000)
placed Cremaspora, formerly included in Gardenieae subtribe Diplosporinae by Robbrecht and Puff (1986)
, in its own tribe (Cremasporeae Bremek. ex S.P. Darwin).
An enlarged and modified concept of Coffeeae (see Table 1) was followed by Bridson and Verdcourt (2003
, p. 387) although further significant changes were made. Based on morphology (discussed later) and on preliminary plastid sequence data that we provided (A. Davis, unpublished data), Bridson and Verdcourt (2003
, p. 451) added the genera Argocoffeopsis, Belonophora Hook. f., and Calycosiphonia to Coffeeae. Argocoffeopsis and Calycosiphonia were formerly members of Gardenieae subtribe Diplosporinae (Robbrecht, 1988b
), and Belonophora was tentatively included in Aulacocalyceae Robbr. & Puff (Robbrecht, 1988b
, 1994
). Bridson and Verdcourt (2003
, p. 388) excluded Bertiera from Coffeeae because they believed that the extension of characters necessary to accommodate this morphologically distinct genus would greatly distort the circumscription of Coffeeae. To this end, Bertiera was placed in its own tribe, Bertiereae (K.Schum.) Bridson (Bridson and Verdcourt (2003
, p. 386). The differences between the tribes were not stated, but instead the characterizations for each tribe were given in tribal descriptions (Bridson and Verdcourt, 2003
, pp. 387–388). The salient characters of Coffeeae according to Bridson and Verdcourt (2003
, p. 388) are as follows (we selected the main characters and the comments in square brackets are ours): inflorescences axillary or less often terminal on short lateral spurs; bracts and bracteoles sometimes free but more often with opposite pairs fused together to form cup-like structures (‘cupules' [= calyculi]) and arranged in a series of usually 1–4, each cupule with (0–2–) 4 lobes; secondary pollen presentation mostly present (said to be lacking in Coffea [and Psilanthus]). Anthers exserted from corolla throat or occasionally included. Pollen presenter/stigmatic lobes divergent or not. Seeds free or wholly or partly covered in arilloid placenta tissue; embryo radicle inferior or rarely superior; testa cells isodiametric or elongated, thickenings mostly absent or more or less weak, smooth along outer tangential walls, or in Coffeeae sensu stricto [Coffea and Psilanthus] with cells crushed during the development of endosperm, composed of thin elongate parenchymatic cells usually containing many more or less isolated fibers. Pollen grains 3–4-colporate.
|
reclassified Coffeeae, splitting it into two subtribes: Coffeeae subtribe Coffeinae (DC.) Robbr. & Manen (Argocoffeopsis, Belonophora, Coffea, Discospermum, Psilanthus, Sericanthe, Tricalysia, Xantonnea Pierre ex Pitard), and Coffeeae subtribe Bertierinae (K. Schum) Robbr. & Manen (Bertiera). The inclusion of component genera was largely based (i.e., all except Belonophora) on the work presented here and that given by Andreasen and Bremer (2000)
. Diplospora, a member of Coffeeae according to Andreasen and Bremer (2000)
, was not placed in Coffeeae but instead within a clade named ‘Gardenieae 4'; Calycosiphonia, a candidate for inclusion in Coffeeae was also placed in the same clade as Diplospora. According to Robbrecht and Manen (2006)
subtribe Coffeinae and subtribe Bertierinae are separated by inflorescence position and number of ovules per locule. In subtribe Coffeinae the inflorescences are axillary and paired at nodes (in some genera accompanied by terminal inflorescences) and ovules per locule 1–few (only in some Tricalysia spp. more than 20). In subtribe Bertierinae the inflorescences are terminal and ovules per locule numerous.
Following the work of Robbrecht and Puff (1986)
, Andreasen and Bremer (1996
, 2000
), Persson (1996)
, Bridson and Verdcourt (2003)
, and Robbrecht and Manen (2006)
, the aims of this study are to further investigate the delimitation and characterization of Coffeeae and to elucidate the relationships within Coffeeae with particular attention to the position of Coffea and Psilanthus. To achieve our objectives, we used sequence data from four plastid regions (the trnL intron, the trnL-F intergenic spacer [IGS], the rpl16 intron, and the accD-psa1 IGS) and compared and combined these data with morphological data. For the purposes of this study, the trnL intron and the trnL-F IGS are treated as one region: trnL-F. All genera with potential to be placed within Coffeeae, including all members of Gardenieae subtribe Diplosporinae (based on Robbrecht, 1988b
, 1994
), were included in the sampling, although we were unable to locate DNA material for three genera from Gardenieae subtribe Diplosporinae. Morphological data alone were used to deduce the systematic position of genera not sampled in our plastid sequence analyses.
MATERIALS AND METHODS
Taxon sampling
Species representatives from all genera placed within the current enlarged concept of Coffeeae by Andreasen and Bremer (2000)
and Bridson and Verdcourt (2003)
were included in our analyses, viz. Argocoffeopsis, Belonophora, Bertiera, Calycosiphonia, Coffea, Diplospora, Discospermum, Psilanthus, Sericanthe, and Tricalysia. For the three remaining members of Gardenieae subtribe Diplosporinae (Robbrecht, 1998b
), our material of Nostolachma, Petitiocodon, and Xantonneopsis did not yield usable DNA and plastid sequences could not be obtained. Petitiocodon, Xantonneopsis, and Nostolachma were included in the morphological analysis and then in a combined morphological–molecular analysis to determine their systematic position. A review of genera within Rubiaceae (A. Davis and D. Bridson, Royal Botanic Gardens, Kew, personal observations) identified one further genus likely to be placed within Coffeeae, namely Xantonnea Pierre ex Pit. This was included in the plastid sequence and morphological analyses. Collections made in 2005 from Cameroon and the Congo (B. Sonké, University of Yaoundé, personal communication) yielded a new taxon that was likely to fall within Coffeeae, although as yet it cannot be placed within a currently accepted genus. This taxon is currently referred to (B. Sonké and A. Davis, personal communication) as Calycosiphonia cf. as it superficially resembles species within this genus. All genera, and their respective samples, with the potential to fall with the broader concept of Coffeeae were defined as the ingroup. A selection of genera from subfamily Ixoroideae sensu stricto (i.e., excluding ‘basal Ixoroideae', after Andreasen and Bremer (2000)
, were used as outgroups. In total nine genera from five tribes were included in the outgroup sample, representing the tribes Ixoreae A. Gray, Pavetteae A. Rich ex Dumort, Vanguerieae Dumort, Octotropideae Beddome, and Gardenieae DC. Following the circumscription of Ixoroideae s.s. after Andreasen and Bremer (2000)
, Cremasporeae (Verdc.) S.P. Darwin, and Alberteae Hook. f. are the only tribes not included in our analyses; these taxa are not close relatives of Coffeeae (Andreasen and Bremer, 2000
; Robbrecht and Manen, 2006
). Hereafter, we refer to Ixoroideae, less basal groups (after Andreasen and Bremer, 2000
), as Ixoroideae s.s. An overview of the genera studied in this contribution, including their current tribal placement and general distribution, is given in Table 1. Taxon names, voucher information and GenBank accession numbers are given in Appendix 1.
DNA extraction, amplification and sequencing
Most of the DNA samples were obtained from silica-dried leaf material (Chase and Hills, 1991
) and the remainder from single seeds, flowers, or leaf samples obtained from herbarium specimens. DNA extraction was performed from a maximum of 0.3 g of silica-dried flower or leaf material (or from one seed) using the 2x CTAB method described by Doyle and Doyle (1987)
. The DNA was purified on cesium chloride/ethidium bromide gradients (1.55 g·ml–1 density) and dialyzed before inclusion in the DNA Bank at the Royal Botanic Gardens, Kew (http://www.rbgkew.org.uk/data/dnaBank/homepage.html). To avoid problems of PCR inhibition, all DNA samples were further purified on QIAquick purification columns (QIAgen) following the manufacturer's protocol.
DNA sequences for seven taxa were taken from another phylogenetic study (O. Maurin et al., unpublished data). The remaining 35 taxa were sequenced according to the methods described next.
Amplification of plastid regions was performed using the primers listed in Table 2. Amplification of trnL-F was carried out using primers c and f designed by Taberlet et al. (1991)
. For many taxa, the internal primers d and e also had to be used because of difficulty in amplifying the region as a single piece. Amplification of the rpl16 region was carried out using primers 71F and 1661R of Jordan et al. (1996)
. For many taxa, the amplification of DNA using these primers was not satisfactory, so we designed internal primers (Table 2) on the basis of our first sequences in a conserved and GC-rich region suitable for amplification of the rpl16 region in two fragments for Rubiaceae. The accD-psa1 IGS region was amplified using the primer ACCD-769 forward and PSA1-75 reverse from Mendenhall (1994)
. Two internal primers (Table 2) were again designed to obtain satisfactory PCR products for our recalcitrant specimens.
|
A PCR mastermix containing 2.5 mM MgCl2 (Advanced Biotechnologies, Epsom, Surrey, UK) was used for trnL-F and rpl16 amplifications. For accD-psa1, the commercial mastermix did not give good amplifications, so we prepared our own premix using Biotaq DNA polymerase (Bioline, UK), 10x NH4 reaction buffer (Bioline, UK), 50 mM MgCl2, and dNTPs (Promega, Southampton, UK). Amplified products were purified using QIAquick purification columns (QIAgen) as described in the manufacturer's protocol. Cycle sequencing reactions were carried out using BigDye Terminator Mix (Applied Biosystems, Warrington, Cheshire, UK). The program consisted of 26 cycles of 10 s denaturation (96°C), 5 s annealing (50°C), and 4 min elongation (60°C). PCR and sequencing reactions were run using a Perkin-Elmer (Warrington, Cheshire, UK) GenAMP model 9600 or 9700 PCR system, and sequencing products were run on either an ABI 3100 Genetic Analyzer or an ABI 377 automated sequencer according to the manufacturer's protocols (Applied Biosystems). Electropherograms were edited and assembled into contigs using Sequencher version 3.2.2. (Gene Codes Corp., Ann Arbor, Michigan, USA). The sequences generated were submitted to GenBank using the Sequin application (version 5.26; available from http://www.ncbi.nlm.nih.gov/Sequin/).
Data matrix composition and parsimony analysis
All sequences were aligned manually in PAUP* (version 4.0b10; Swofford, 2002
) without difficulty due to low levels of sequence variation. Areas with ambiguous alignment were excluded from the analysis, as were regions with missing sequences, for example, the beginning and end of sequences and around the internal primer binding sites. In total, 508 characters were excluded from the molecular matrix.
Thirty-one morphological characters were scored using the available literature (given later) and new observations from herbarium specimens, samples preserved in alcohol, and in some cases living material in situ. Selected herbarium material from DSM, K, L, P and YA, and specimens preserved in alcohol at K were consulted (abbreviations after Holmgren et al., 1990
). Representatives of all African and Madagascan genera, excluding Petitiocodon, were examined in situ, in Cameroon, Tanzania, and Madagascar by A. Davis (personal observation). The literature was consulted for most genera: Bremekamp (1934)
, Bridson (1978)
, Robbrecht (1978
, 1981
, 1988a
, b
), Puff et al. (1984)
, Bridson and Verdcourt (1988
, 2003
), Robbrecht and Puff (1986)
, Leroy (1989)
, Ali and Robbrecht (1991)
, Puff et al. (1996)
, Robbrecht et al. (1994)
, Persson (1996)
, Bridson (1998)
, Andreasen and Bremer (2000)
, Cheek and Dawson (2000)
, Vinckier et al. (2000)
, Stone and Davis (2004)
, Davis et al. (2005)
, and Rakotonasolo and Davis (2006)
.
The same nine genera that were used as outgroups for the molecular analysis were scored for the morphological matrix. The total morphological variation within Ixoroideae s.s. and its tribes cannot be covered in a sample of this size, but the objective was to polarize characters for the ingroup only. Likewise, scoring of morphological characters for genera, and particularly large genera such as Ixora, by using one terminal, will not encompass all the variation within a genus. However, in the case of the genera we have studied in conjunction with the characters we have used, sampling of variation within genera is sufficiently representative and more than adequate for the purposes of the study. The morphological matrix provided here for Ixoroideae s.s. is the most comprehensive so far undertaken and represents a foundation for further investigation. The general distribution of character states for genera and tribes within Ixoroideae s.s., other than those sampled by us, are given in Appendix 2. The list of characters and character states, with explanatory notes (including potentially informative characters not scored in our analysis), is given in Appendix 2. The morphological matrix is given in Appendix 3. All morphological characters were treated as unordered.
Maximum parsimony was implemented to analyze (1) trnL-F, (2) rpl16, (3) accD-psa1, (4) combined plastid data, (5) morphological data, and (6) combined molecular and morphological data in PAUP* (Swofford, 2002
). In all analyses, gaps were treated as missing data and characters were equally weighted and unordered (Fitch, 1971
). The data sets were analyzed separately to identify topological conflicts: strict consensus trees were produced from each data set and compared carefully by eye. Almost no incongruence was found in the plastid analyses, as far as each analysis was resolved, and the only anomaly was the placement of Canephora sp. (Octotropideae) sister to all Tricalysia (Coffeeae) spp. in the rpl16 analysis (BP 69). In all other analyses Canephora is placed with Polysphaeria sp. (also Octotropideae). On the basis of this general agreement, the three plastid analyses were combined. The morphological analysis was combined with the combined plastid analysis on the basis that there was no strongly supported incongruence between the two data sets.
An additional combined molecular–morphological analysis was conducted to include the three genera (four taxa) for which we had no DNA material (Petitiocodon, Xantonneopsis, and Nostolachma [2 spp.]). This analysis was performed with the sole objective of elucidating the systematic position of these genera within Ixoroideaes s.s.
Tree searches were conducted using 10 000 replicates of random taxon sequence addition, retaining 10 trees at each step, with tree-bisection-reconnection (TBR) branch swapping, delayed transformation (DELTRAN) optimization, MulTrees in effect, and saving a maximum of 10 trees per replicate. Support for clades in all analyses was estimated using bootstrap analysis (Felsenstein, 1985
), with 10 000 replicates of full heuristic search, simple sequence addition, TBR swapping, with MulTrees in effect and saving a maximum of 10 trees per replicate. Bootstrap percentages (BP) are described as high/well-supported (85–100%), moderate (75–84%), or low (50–74%). In addition, Bremer support values (b) (Bremer, 1988
, 1994
; Källersjö et al., 1992
), otherwise known as decay values, were obtained using PAUP* (Swofford, 2002
), in conjunction with AutoDecay 4.0.2 (Eriksson, 1999
), with 100 replicates of random addition for each constraint tree.
For the morphological, combined molecular analysis, and combined molecular–morphological analyses, two representatives of Vanguerieae (Pyrostria spp.) were used as outgroups in the analyses. This decision was based on wider studies of Ixoroideae by Andreasen and Bremer (1996
, 2000
), Persson (2000)
, and Robbrecht and Manen (2006)
. The mapping of morphological characters on the morphological and morphological–molecular trees was undertaken using MacClade (version 4.07 PPC; Maddison and Maddison, 2005
). Key morphological characters were mapped onto the strict consensus tree of the morphological–molecular analysis for illustrative purposes (Fig. 3).
|
Tree data and statistics for individual analyses using (1) trnL-F, (2) rpl16, (3) accD-psa1, (4) combined molecular data, (5) morphological data, and (6) combined molecular and morphological data are given in Table 3. Individual plastid sequence analyses were topologically consistent and, for the purpose of the Results, Discussion, and Conclusions, are here combined and treated as a single analysis. We retrieved a significantly smaller number of trees for the analysis of rpl16 (Table 3), although we were unable to find the reason for this anomaly. The percentage of parsimony informative characters and the values for the consistency index (CI) and retention index (RI) were similar for all plastid data sets.
|
|
|
Combined molecular and morphological analysis, including taxa without sequence data (46 taxa)
The objective of this analysis was to deduce the systematic postion for the four taxa lacking sequence data (Nostolachma densiflora, N. khasiana, Petitiocodon parviflora, and Xantonneopsis robinsonii). The analysis resulted in 234 most parsimonious trees. The strict consensus tree of 46 taxa is consistent with the combined molecular–morphological data tree based on 42 taxa (Fig. 3) but with some slight changes in topology, slightly lower bootstrap values, and considerably reduced decay values (Appendix S1, see Supplemental Data with online version of this article). The two species of Nostolachma are placed with the Asian representatives of Coffeeae, sister to Xantonnea; Petitiocodon is placed with members of Gardenieae, and Xantonneopsis with those of Octotropideae, although these relationships are poorly supported.
Distribution of morphological characters for and within the ingroup
In the combined morphological–molecular analysis, the ingroup is supported by two morphological synapomorphies, the presence of calyculi, and a distinctly lobed style (characters 6, 23, respectively; see Appendices 2 and 3). Bertiera does not possess these characters. Within the ingroup, other suprageneric groupings are supported by morphological data. Coffea and Psilanthus are supported by the apparent loss of secondary pollen presentation, the presence of a hard (horny/crustaceous) endocarp, seeds with a deep ventral groove, and a seed coat consisting of crushed endotestal cells and more or less isolated fibers (characters 9, 26, 28, 29). Calycosiphonia, Argocoffeopsis, and Calycosiphonia cf. by the absence of a seed coat (character 29), and Diplospora, Discospermum, and Xantonnea (and Nostolachma) by the universal presence of 4-merous flowers (character 8). The monophyly of several genera is supported by morphology: Belonophora by the presence of imbricate calyx lobes, included anthers, and superior radicle (characters 10, 14, 21); Sericanthe 7–9-merous flowers, distinctly basifixed anthers, and horizontal micropyle (radicle) orientation (characters 8, 15, 21); Calycosiphonia by having anthers with thecae (character 16); Tricalysia seeds with a hilar groove or shallow excavation (character 28); and Psilanthus by included anthers, supramedifixed anthers, and included style (characters 14, 15, 22). The monophyly of Bertiera is supported by a bimorphic corolla tube, a distinctly peltate placenta, and an included style with longitudinal ridges (characters 12, 19, 22, 24). Key characters are mapped on Fig. 3.
DISCUSSION
Close relatives of Coffea and their inclusion within Coffeeae
On the basis of plastid sequence data and morphology, the genera Argocoffeopsis, Belonophora, Calycosiphonia, Calycosiphonia cf. Coffea, Diplospora, Discospermum, Psilanthus, Tricalysia, Sericanthe, and Xantonnea form a well-supported clade (Fig. 3: BP 97, b = 6). With these genera included, Coffeeae is readily distinguished by the presence of calyculi and a distinctly 2-lobed style (characters 6 and 23). None of these characters are unique synapomorphies for the ingroup, although the presence of calyculi is otherwise so far only known in Doricera and perhaps some species of Ixora. The presence of a distinctly 2-lobed style (character 23) seems to be restricted to rather few genera within Ixoroideae s.s. (e.g., Ixora, Robbrechtia De Block). Paired axillary inflorescences (character 4) are found within several Ixoroideae but serve to provide a further distinguishing character for Coffeeae: paired axillary inflorescences are consistently present in Coffeeae.
Bertiera is sister to the ingroup in the combined plastid sequence analysis (BP 40, b = 1) but sister to two representatives of Gardenieae in the combined morphological–molecular analysis (BP 37, b = 1). Bertiera is morphologically distinct from members of Coffeeae, differing in the following characters (characters for Coffeeae in parentheses): (1) terminal inflorescences (axillary), although in three species of Bertiera the inflorescence is distinctly (paired) axillary (Robbrecht et al., 1994
); (2) calyculi absent (present); (3) bimorphic corolla (corolla straight); (4) numerous (>30) ovules per locule (usually 1-several, but up to 20 in some species of Tricalysia); (5) peltate placenta (various shapes but never peltate); (6) multidirectional micropyle orientation (usually downward but always in one direction); (7) distinctly ridged style (style entire) with (8) adnate lobes (lobes free). The "axillary and then terminal" position of inflorescences in some members of Coffeeae is not homologous with the truly terminal inflorescences of most Bertiera species (Davis et al., 2005
). The distinctly peltate placenta in Bertiera seems to be unique within Ixoroideae s.s., and the genus possesses a combination of characters that is unique within this part of the subfamily, e.g. peltate placenta, numerous ovules per locule, and distinctly ridged style. In addition, most Bertiera species seem to have a type of exotestal thickening not found in other Rubiaceae (Robbrecht et al., 1994
), although further critical study is required. Bertiera is the only genus of Ixoroideae s.s. that includes herbaceous species. Our findings support the removal of Bertiera from Coffeeae and placement in its own tribe, Bertiereae, in agreement with Bridson and Verdcourt (2003)
. A description and circumscription of Bertiera and Bertiereae is given by Bridson and Verdcourt (2003
, p. 386). The placement of Bertiera within Coffeeae as subtribe Coffeinae as proposed by Robbrecht and Manen (2006)
conflicts with the results of our analyses for both molecular and morphological data. Morphological data separate Bertiera from Coffeeae, whereas there are no morphological characters supporting the grouping of Bertiera and Coffeeae. Robbrecht and Manen (2006)
provide no morphological characterization or rationale for Coffeeae including Bertiera. We believe that our new circumscription of Coffeeae, with the addition of new genera and the removal of Bertiera, is both scientifically coherent and practical.
Three generic candidates for inclusion in Coffeeae (see Table 1), viz. Nostolachma, Petitiocodon, and Xantonneopsis (all currently Gardenieae subtribe Diplosporinae) were not sampled for the plastid sequence analysis, and morphology alone was used to determine their systematic placement: observation of morphological characteristics and position within a combined molecular–morphological analysis (Appendix S1, see Supplemental Data with online version of this article). Nostolachma is placed with the Asian representatives of Coffeeae (Diplospora, Discospermum, and Xantonnea), which are characterized within the tribe in having consistently 4-merous flowers. This placement is not well supported, although Nostolachma is convincingly placed within Coffeeae. Our observations of floral morphology indicate that Nostolachma is dioecious, which is common in Diplospora and Discospermum but absent in the African genera of the ingroup, although some species of Madagascan Tricalysia are dioecious (Ali and Robbrecht, 1991
; P. De Block, National Botanic Garden of Belgium, personal communication). Further study of Coffeeae genera occurring in Asia is warranted.
Xantonneopsis consists of a single species, X. robinsonii Pit. (Pitard, 1923
). The combined molecular–morphological analysis (Appendix S1, see Supplemental Data with online version of this article) places X. robinsonii with Octotropideae (BP 69, b = 1), which is consistent with basic morphological observation: paired pedunculate, supra-axillary inflorescences and the presence of two distinctly pendulous ovules per locule arranged on a biseriate placenta. The biseriate placenta appears to be unique to Octotropideae (Robbrecht, 1988b
; Bridson and Verdcourt, 2003
) although it may not occur in all species within the tribe, and some genera (e.g., Polysphaeria Hook. f.) have a single ovule per locule. The seed coat of Octotropideae has a very distinct fingerprint-like or sometimes reticulate pattern (Robbrecht, 1988b
; Bridson and Verdcourt, 2003
; Stone and Davis, 2004
) but unfortunately seeds of X. robinsonii were not available for study.
Petitiocodon also comprises a single species, P. parviflorum (Keay) Robbr.; the morphology of this species is described in detail by Robbrecht (1988b)
. This species does not have the key characters of Coffeeae but instead shares greater morphological affinity with members of Gardenieae. It has paired pedunculate, supra-axillary inflorescences, and a rather large broad-tubed corolla, and each locule of the bilocular ovary has a placenta attached at its base to the dividing wall of the locule with two more or less immersed ovules. The size and shape of the corolla, the position of the placental attachment, and the position of the ovules within the placenta infer a systematic position close to or within Gardenieae, although further data is required to clearly resolve the tribal placement of Petitiocodon within Rubiaceae. The combined molecular–morphological analysis (Appendix S1, Supplemental Data with online version of this article) supports our morphological observations, although the support values for placement within Gardenieae are very low (BP < 50, b = 1).
Characterization of Coffeeae
The morphological characters supporting an enlarged circumscription of Coffeeae given here are consistent with the characterization of the tribe given by Bridson and Verdcourt (2003
, p. 387). One main point of difference is the nature of the apparently terminal inflorescences in Coffeeae, which has been elucidated by Davis et al. (2005)
. The position of the inflorescence on a short lateral shoot present in Coffea subgen. Baracoffea (J.-F.Leroy) J.-F.Leroy, most species of Psilanthus subgen. Afrocoffea (Moens) Bridson, and most species of Argocoffeopsis differs from a true terminal inflorescence. Each inflorescence originates as one of the paired axillary inflorescence and then becomes terminal on a short "shoot" or "spur" following renewed meristematic activity (Davis et al., 2005
, p. 410). Thus the inflorescences in Coffeeae are never strictly terminal.
The genus Doricera, a member of Ixoreae, is sometimes affiliated with Coffeeae (e.g., Leroy, 1989
), as it has paired axillary inflorescences, calyculi, and Coffea-like flowers. However, the style is club-shaped, shortly 3-lobed (not simple, and distinctly 2-lobed), the ovary is 3-locular (not 2-locular), and the seeds have a small circular excavation ("hilar cavity") on the ventral surface (not entire or with a longitudinal invagination). Our molecular and combined molecular–morphological analyses indicate the position of Doricera within Ixoreae (Fig. 1: BP 100, b = 16; Fig. 3: BP 100, b = 17).
We propose an emended characterization of Coffeeae, as given next, with key characters identified with an asterisk *.
Coffeeae DC., Ann. Mus. Hist. Nat. (Paris) 9: 217 (1807), as ‘Coffeaceae' (including numerous genera); Robbr. & Puff, Bot. Jahrb. Syst. 108: 122 (1986) (restricted to Coffea L. and Psilanthus Hook. f.); Andreasen & B. Bremer, Am. J. Bot. 87: 1742 (2000) (also including several genera from Gardenieae DC. subtribe Diplosporinae Miq. and Bertiera Aubl.); Coffeeae DC. subtribe Coffeinae (DC.) Robbr. & Manen, Syst. Geogr. Pl. 76: 133 (2006). Type: Coffea L.
Synonym: Gardenieae DC. subtribe Diplosporinae Miq., Fl. Ned. Ind. 2: 237 (1857), as ‘Diplosporeae'; Robbr. & Puff, Bot. Jahrb. Syst. 108: 114 (1986) (excluding Cremaspora Benth.). Type: Diplospora DC.
(1) Trees, shrubs, woody climbers, and woody monocauls; (2) stipule pairs adnate at the base and sometimes semi-sheathing, apex entire or apiculate; (3)* inflorescences paired, axillary or axillary and then appearing terminal (by continued meristematic activity of the inflorescence) on short shoots (typically in inflorescences from the previous year); (4)* inflorescences sessile (lacking a peduncle) with lowermost calyculus in the leaf axil; (5)* calyculi (cupule-like structures formed by the contraction of shoot tissue and the reduction and fusion of leaves and stipules) present, usually 4-lobed, but sometimes 2-lobed or lobes lacking; (6) flowers 4–8(–12)-merous; (7)* corolla tube narrow and straight and (8) with lobes overlapping to the left (Coffea-like flowers); (9) anthers exserted, rarely included; (10) stigmatic lobes usually excluded but sometimes distinctly included (Psilanthus), and included or excluded in dioecious species; (11) ovary 2-locular, placentation axile; (12) each locule with the placenta attached to the septum, at the middle or just above, rarely near the apex (only Belonophora and Sericanthe); (13)* ovules usually 1 or 2 per locule or up to 10 (rarely c. 20); (14) ovules on a small placenta, or placenta large and ovules either on or semi-embedded within the placental tissue; (15) embryo radicle usually inferior, rarely horizontal (only Sericanthe), or superior (only Belonophora); (16) style simple (lacking specialized features), glabrous, *distinctly 2-lobed; (17) fruit an indehiscent drupe, with few (1 or 2) to several seeds (rarely up to c. 10); (18) ventral surface of seed more or less entire, (sometimes with a shallow hilar grove or shallow excavation), or with a distinct longitudinal ventral invagination ("coffee-bean" morphology: only Coffea and Psilanthus); (19) endosperm usually entire but sometimes ruminate (only some species of Tricalysia and Coffea); (20) testa cells isodiametric or elongated, thickenings present or absent, smooth along outer tangential walls, or (in Coffea and Psilanthus) with cells crushed during the development of endosperm and composed of thin elongate parenchymatic cells usually containing many more or less isolated fibers; (21) pollen (2–)3–5-colporate (zonocolporate).
Genera included Argocoffeopsis, Belonophora, Calycosiphonia, Coffea, Diplospora, Discospermum, Nostolachma, Psilanthus, Sericanthe, Tricalysia, and Xantonnea. The entity Calycosiphonia cf. may represent an additional genus for the tribe; further material and morphological investigations are required.
Further details of salient characters are given in Appendix 2. In addition to the findings presented here, this morphological characterization is based on data derived from Robbrecht (1981)
, Robbrecht and Puff (1986)
, Bridson and Verdcourt (2003)
, Davis et al. (2005)
, and for pollen, Stoffelen et al. (1997)
.
Relationships within Coffeeae
Based on combined molecular data (Fig. 1; BP 100, b = 9) and combined molecular–morphological data (Fig. 3; BP 100, b = 13), there is strong support for a relationship between Coffea and Psilanthus. This is to be expected, given that they share a carpel morphology that is unique within Ixoroideae s.s. (see Fig. 3; Introduction; Appendices 2 and 3). They are also peculiar within Ixoroideae s.s. in lacking secondary (ixoroid) pollen presentation, which is present in nearly all members of the subfamily (Puff et al., 1996
), although the distribution of this character in some monoecious members of Ixoroideae (e.g., Diplospora and Discospermum) is still unclear (e.g., Ali and Robbrecht, 1991
). A close phylogenetic relationship between Coffea and Psilanthus based on molecular and morphological data is concomitant with close cytological correspondence (Lombello and Pinto-Maglio, 2003
, 2004
), including the production of a fertile intergeneric hybrid (Couturon et al., 1998
). Our results are also in agreement with the phylogenetic studies of Lashermes et al. (1997)
, Cros et al. (1998)
, Andreasen et al. (1999)
, and Andreasen and Bremer (2000
, fig. 3), who showed that Psilanthus is nested within Coffea. Detailed species-level studies are now required to assess the phylogenetic relationship between Coffea and Psilanthus.
The combined molecular–morphological analysis shows that that there is strong support (Fig. 3: BP 99, b = 9) for a clade containing Argocoffeopsis, Calycosiphonia, and Calycosiphonia cf. distinguished by the lack of a testa. Earlier studies did not identify close phylogenetic links between Argocoffeopsis and Calycosiphonia (Robbrecht, 1981
; Robbrecht and Puff, 1996), although it has been known for some time that these genera lack a testa (e.g., Robbrecht and Puff, 1996) and that this feature is all but absent within Coffeeae and Gardenieae subtribe Diplosporinae (Ali and Robbrecht, 1991
) and rare in Ixoroideae s.s. (Robbrecht and Puff, 1986
). Ali and Robbrecht (1991)
report that in Diplospora (only D. wrayi King & Gamble) the mature exotesta cells are very rarely crushed by the development of the endosperm, so that the seed coat is reduced to a thin pellicle, as in Argocoffeopsis and Calycosiphonia. The exact systematic position of Calycosiphonia cf. has not been elucidated by our study. It generally looks much more like Calycosiphonia than Argocoffeopsis, although it can be excluded from the former genus on the basis of lacking anther thecae and having 5-merous flowers (7- to 8-merous in Calycosiphonia). Calycosiphonia cf. also closely resembles one species of Argocoffeopsis, A. lemblinii (A. Chev.) Robbr. Most species of Argocoffeopsis are woody climbers, but A. lemblinii [and A. rupestris (Hiern) Robbr.] are shrubs or small trees. The only readily discernible difference between Calycosiphonia cf. and A. lemblinii is that the former has large ellipsoid fruits (seeds elliptic in outline) and the latter has small spheroid fruits (seeds circular in outline). However, further study is required.
The combined molecular–morphological analysis weakly supports a relationship between three of the four Coffeeae genera occurring in Asia: Diplospora, Discospermum, and Xantonnea (Fig. 3; BP 64, b = 1), and the fourth, Nostolachma, is placed within this group based on morphology alone (Appendix S1, see Supplemental Data with online version of this article). This group is characterized by the presence of 4-merous flowers, a character unique to these four genera within Coffeeae (discussed earlier). This is the first time a relationship between these four genera has been suggested. These results agree with Robbrecht and Puff (1986)
, who concluded that Nostolachma does not have a close relationship with Coffea and Psilanthus as postulated by Leroy (1980)
. The recognition of Diplospora and Discospermum (from Asia) as separate from Tricalysia (Africa and Madagascar) by Ali and Robbrecht (1991)
is supported by our study. According to Ali and Robbrecht (1991)
, Asian Diplospora and Discospermum can be separated from African and Madagascan Tricalysia (characters for Tricalysia in parentheses) by standard type colleters (mostly with modified colleter type); almost consistently 4-merous flowers (predominantly 5-pleiomerous); inconspicuous bracteoles/calyculi (large, sheathing bracteoles resembling the calyx); small flowers (larger flowers); strong tendency toward unisexual flowers (hermaphrodite; unisexual in Madagascan species); secondary pollen presentation possibly absent (generally or universally present); seed coat mostly with thickened exotestal cells (not thickened). Three of these characters are included in our morphological matrix (see Appendices 2 and 3). The other Asian genus, Xantonnea, was included in the tribe Octotropideae by Robbrecht (1988b
, 1994
), but our molecular data and the presence of key morphological features place it with the genera of Coffeeae restricted to Asia (Figs. 1, 3).
Given the low branch support values (BP 57, b = 1) and low level of internal resolution for the clade that includes Tricalysia, Belonophora, Sericanthe, and the Asian Coffeeae (Diplospora, Discospermum, and Xantonnea), further data are needed to fully elucidate the systematic relationships for and within this clade. In addition, the relationship of the Argocoffepsis-Calycosiphonia-Calycosiphonia cf. clade with all of the above genera needs further investigation.
The monophyly and circumscription of genera of Coffeeae
Combined molecular and morphological data (Fig. 3) are consistent with the assumed monophyly of Tricalysia (BP 100, b = 10), Belonophora (BP 100, b = 15), Sericanthe (BP 100, b = 8), Calycosiphonia (BP 85, b = 3), Argocoffeopsis (BP 81, b = 2), and Psilanthus (BP 100, b = 8). However, for some genera we have only sampled a relatively low number of species per genus. In particular, Tricalysia (98 spp.) and Coffea (95 spp.) require further sampling. Coffea is paraphyletic with respect to Psilanthus (see earlier). The monophyly of the Asian Coffeeae genera could not be tested in our study due to a lack of suitable DNA material. Molecular and morphological data are needed to test the monophyly and elucidate the systematic relationships of Diplospora, Discospermum, Nostolachma, and Xantonnea. It is likely that some generic synonymy may result from further studies.
Apart from the Asian genera, the monophyly of most genera of Coffeeae is supported by one or more synapomorphies (see earlier). The exceptions are Argocoffeopsis, which is based on a combination of characters (e.g., Bridson and Verdcourt, 2003
), and Coffea (Davis et al., 2005
).
Coffea differs from Psilanthus in its long, exserted style (short, included in Psilanthus) and submedifixed (vs. supramedifixed) excluded (vs. included) anthers. According to Davis et al. (2005)
, there are two other morphological differences between these two genera [excluding P. melanocarpus (Welw. ex Hiern) J.-F. Leroy]. Coffea has short to long anther filaments (very short to more or less absent filaments in Psilanthus) and pollen with 3 or 4, rarely 2, apertures (vs. 4 or 5, rarely 3). Despite these morphological distinctions, molecular, morphological, and combined analyses all show Psilanthus as derived within the Coffea clade (Figs. 1, 3), so that the recognition of Psilanthus results in a paraphyletic Coffea.
Conclusions
The cladistic analyses presented here provide the most inclusive systematic study of Coffeeae to date, with molecular sequence data for all but one of the component genera. The enlargement of Coffeeae by the addition of Diplospora and Tricalysia, on the basis of molecular sequence data (Andreasen and Bremer, 1996
, 2000
; Persson, 2000
) is supported by our study. The addition of Discospermum and Sericanthe on the basis of morphological inference (Andreasen and Bremer, 2000
) is also confirmed. Preliminary results from this study (A. P. Davis, unpublished data) were used to support the transfer of Argocoffeopsis, Belonophora, and Calycosiphonia to Coffeeae by Bridson and Verdcourt (2003)
, and these transfers are again confirmed by our completed study. Xantonnea is transferred from Octotropideae (Robbrecht, 1988b
, 1994
) to Coffeeae on the basis of molecular sequence data and morphology. Nostolachma is transferred from Gardenieae subtribe Diplosporinae (Robbrecht, 1988b
, 1994
) to Coffeeae on the basis of morphological data. Petitiocodon and Xantonneopsis are not members of Coffeeae; based on morphological data, the former is placed tentatively in Gardenieae and the latter Octotropideae.
We propose that Coffeeae should include 11 genera, viz. Argocoffeopsis, Belonophora, Calycosiphonia, Coffea, Diplospora, Discospermum, Nostolachma, Psilanthus, Tricalysia, Sericanthe, and Xantonnea (see Table 1), and have given an updated morphological characterization of the tribe (see earlier). A poorly known and informal taxon, Calycosiphonia cf. is also placed in Coffeeae, although its generic status has yet to be confirmed (A. P. Davis and B. Sonké, personal communication).
Coffea and Psilanthus are clearly distinct within Coffeeae, and a sister relationship to the rest of the tribe is inferred but is only weakly supported by our combined molecular–morphological analysis (Fig. 3: BP 59, b = 1). Within other Coffeeae, a well-supported phylogenetic relationship is inferred for the species lacking a discernible seed coat (Argocoffeopsis, Calycosiphonia, and Calycosiphonia cf.). The monophyly of Argocoffeopsis, Belonophora, Calycosiphonia, Psilanthus, Tricalysia, and Sericanthe is consistent with our analyses; one or more clear-cut synapomorphies are identified for Belonophora, Calycosiphonia, Psilanthus, and Sericanthe. Coffea is paraphyletic, as identified in other molecular analyses (Lashermes et al., 1997
; Cros et al., 1998
; Andreasen et al., 1999
; Andreasen and Bremer, 2000
, fig. 3). Morphological separation of Coffea and Psilanthus is possible (Davis et al., 2005
; and see Discussion), but further data, including wider sampling in these genera, are now urgently required to ascertain whether two genera can be upheld.
The exclusion of Bertiera from Coffeeae and its placement in its own tribe (Bridson and Verdcourt, 2003
, p. 386), Bertiereae, are supported by our combined molecular (Fig. 1), combined molecular–morphological data (Fig. 3), and our morphological survey. The inclusion of Bertiera within Coffeeae, as subtribe Bertierinae as proposed by Robbrecht and Manen (2006)
is not supported by our data. Coffeeae becomes a strictly Old World tribe following the exclusion of Bertiera. The decision to place Gardenieae subtribe Diplosporinae into the synonymy of Coffeeae (Bridson and Verdcourt, 2003
) is substantiated by our analyses.
APPENDIX
Voucher information and GenBank accession numbers for taxa used in this study. A dash indicates the region was not sampled. Voucher specimens are deposited in the following herbaria: K = Royal Botanic Gardens, Kew; TAN = Parc de Tsimbazaza, Antananarivo; BR = National Botanic Garden of Belgium, Meise; YA = National Herbarium of Cameroon, Yaoundé.
Taxon—Voucher (herbarium), Origin; GenBank accession nos.: accD-psaI, rpl16, trnL-F.
Argocoffeopsis eketensis (Wernham) Robbr.—Davis 3031 (K), Cameroon; DQ180497, DQ180531, DQ180566. Argocoffeopsis rupestris (Heirn) subsp.thonneri (Lebrun) Robbr.—Harris 8168 (K), Central African Republic; DQ180496, DQ180532, DQ180567. Argocoffeopsis scandens (K.Schum.) Lebrun—Davis 3016 (K), Cameroon; DQ180498, DQ180533, DQ180568. Belonophora coriacea Hoyle—Maurin 5 (K), Cameroon; DQ180499, DQ180534, DQ180569. Belonophora coriacea Hoyle—Maurin 19 (K), Cameroon; DQ180500, DQ180535, DQ180570. Belonophora sp.—TAF 480 (K), Cameroon; DQ180501, DQ180536, DQ180571. Bertiera bicarpellata (K.Schum.) N.Hallé—Davis 3051 (K), Cameroon; DQ180502, DQ180537, DQ180572. Bertiera laxissima K.Schum.—Maurin 13 (K), Cameroon; DQ180503, DQ180538, DQ180573. Bertiera sp.—Davis 3017 (K), Cameroon; DQ180504, DQ180539, DQ180574. Calycosiphonia cf.—Sonké 3783 (K, YA), Cameroon; DQ180505, DQ180540, DQ180565. Calycosiphonia macrochlamys (K.Schum.) Robbr.—Davis 3036 (K), Cameroon; DQ180506, DQ180541, DQ180575. Calycosiphonia macrochlamys (K.Schum.) Robbr.—Davis 3044 (K), Cameroon; DQ180507, DQ180542, DQ180576. Calycosiphonia spathicalyx (K.Schum.) Robbr.—Mvungi 22 (DSM, K), Tanzania; DQ180508, DQ180543, DQ180577. Calycosiphonia spathicalyx (K.Schum.) Robbr.—Davis 2925 (K), Tanzania; DQ180509, DQ180544
