| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Systematics and Phytogeography |
Department of Systematic and Evolutionary Botany, Faculty of Life Sciences, University of Vienna, Rennweg 14, A-1030, Vienna, Austria; 3Department of Forest Biology, Faculty of Forestry, Kasetsart University, Chatuchak, Bangkok, Thailand; 4Natural History Museum, Burgring 7, A-1010, Vienna, Austria; 5Laboratorie de Botanique, Centre IRD de Nouméa, BP. A5 Noumea Cedex, New Caledonia; and 6Jodrell Laboratory, Royal Botanic Gardens, Kew, Richmond, Surrey TW9 3DS, UK
Received for publication May 29, 2006. Accepted for publication October 9, 2006.
ABSTRACT
Phylogenetic relationships of the pantropical family Ebenaceae s.l. were investigated using plastid DNA sequence data from six regions: atpB, matK, ndhF, trnK intron, trnL intron, and trnL-trnF spacer. Sampling included representatives of all currently recognized genera of Ebenaceae, Diospyros, Euclea, and Lissocarpa, and nearly all taxa that were previously recognized at the generic level, e.g., Cargillia, Gunisanthus, Maba, Macreightia, Royena, and Tetraclis. Our results strongly support monophyly of Ebenaceae s.l. and demonstrate that the previous infrafamilar classifications of the family do not circumscribe monophyletic groups. A new infrafamilial classification based on a phylogenetic approach is proposed here and consists of two subfamilies, Lissocarpoideae and Ebenoideae, and four genera, Lissocarpa, Euclea, Royena, and Diospyros. Relationships and potential synapomorphic characters are discussed and summarized. This study supports a western Gondwanan origin of family and indicates that both vicariant and long-distance dispersal events played an important role in attaining current distributions.
Key Words: atpB • Ebenaceae • infrafamilial classification • matK • ndhF • phylogenetic relationships • trnK intron • trnL-trnF region
Ebenaceae sensu lato (s.l.) consist of Ebenaceae sensu stricto (s.s.) and Lissocarpaceae; they are a medium-sized pantropical family with the greatest number of species in Asia and the Indo-Pacific region, although the greatest morphological diversity is in Africa and Madagascar (White, 1983
; Wallnöfer, 2001
, 2004
). The family is the source of several economically important products; the most valuable are timber (ebony) and fruits (persimmons). They also are a conspicuous forest component of Africa and Asia (Heywood, 1978
; Judd et al., 2002
).
Ebenaceae s.s. is a well-known family (Hiern, 1873
; Bakhuizen, 1936–1955
; White, 1980
, 1983
, 1993
; Singh, 2005
), but molecular studies (Berry et al., 2001
; Anderberg et al., 2002
; Bremer et al., 2002
) have recently shown the family could also include Lissocarpa (formerly assigned to the monogeneric family Lissocarpaceae). Based on this molecular result, Lissocarpa has been transferred to Ebenaceae s.l. as subfamily Lissocarpoideae (Wallnöfer, 2004
). Unless noted otherwise, we will use the broader circumscription throughout this paper (i.e., including Lissocarpa). Previous placements of both Ebenaceae s.s. and Lissocarpaceae were in order Ebenales (sensu Cronquist, 1981
, 1988
; Dahlgren, 1989
) or Styracales (sensu Thorne, 1992
; Takhtajan, 1997
). According to the APG system, Ebenaceae and other families of Ebenales and Styracales have been placed within an expanded order Ericales (Chase et al., 1993
; Soltis et al., 2000
; Berry et al., 2001
; Anderberg et al., 2002
; Bremer et al., 2002
; APG II, 2003
; Schönenberger et al., 2005
).
Infrafamilial classifications of Ebenaceae s.s. have been proposed by de Candolle
(1844), Hiern (1873)
, Bakhuizen (1936–1955
), White (1980
, 1983
, 1993
), and Singh (2005). The previous classifications based on morphological and anatomical characters have been considered to be problematic; generic and infrageneric boundaries of each system have been different and much debated. The number of genera recognized in Ebenaceae s.s. has varied from two to eight (Table 1), and some of these circumscriptions are contradictory, so at least some of them do not circumscribe monophyletic groups. The earliest infrafamilial classification for Ebenaceae s.s. on a worldwide scale was that of de Candolle
(1844), who recognized eight genera: Cargillia, Diospyros, Euclea, Gunisanthus, Maba, Macreightia, Rospidios, and Royena. Hiern (1873)
placed the 249 then-recognized species in five genera, Diospyros, Euclea, Maba, Royena, and Tetraclis. In contrast to de Candolle's system, he proposed a new Madagascan endemic, Tetraclis, and lumped Cargillia, Gunisanthus, and Rospidios with Diospyros and Macreightia with Maba. Bakhuizen (1936–1955
) in his regional revision united Maba with Diospyros, thus recognizing only four genera, Diospyros, Euclea, Royena, and Tetraclis (the last three genera only according to the literature; they were not in the area he covered). Bakhuizen also divided Diospyros into five subgenera (Table 1). White (1980
, 1983
) had a much broader generic concept, reducing Royena and Tetraclis to synonymy with Diospyros; he thus recognized only two genera, Diospyros and Euclea, in Ebenaceae s.s. Singh (2005)
, while dealing with Indian Ebenaceae, followed the generic concepts of White (1980
, 1983
). He divided Indian Diospyros species into 27 sections. Recently, the monogeneric family, Lissocarpaceae has been formally included in Ebenaceae (Wallnöfer, 2004
); Lissocarpa was divided into two sections, Lissocarpa and Enho. Molecular studies previously suggested that Diospyros sensu White (1980)
is paraphyletic because species of Diospyros section Royena form a clade with Euclea (Morton et al., 1996
; Berry et al., 2001
), and this relationship is also supported by morphological and anatomical features (club-shaped trichomes and ingrowth of testa wall) that are shared between Diospyros section Royena and Euclea (Morton et al., 1996
; Wallnöfer, 2001
). In fact, it is still unclear how many genera of Ebenaceae reflect monophyletic groups. A molecular phylogenetic approach should help to resolve infrafamilial relationships of Ebenaceae and aid in assessments of generic boundaries within the family as it had done in the other families, e.g., Rhamnaceae (Richardson et al., 2000a
, b
), Zingiberaceae (Kress et al., 2002
) and Phyllanthaceae (Kathriarachchi et al., 2005
).
|
; Berry et al., 2001
; Anderberg et al., 2002
; Bremer et al., 2002
; Geeraerts, 2003
). Many phylogenetic questions about the family raised by Berry et al. (2001)
still need to be investigated, such as (1) the position of Lissocarpa when more species are included in the analysis, (2) the position of Diospyros section Royena as sister to Euclea, and (3) the relationships of this sister pair with respect to the remaining species of Diospyros. To answer these questions, we have to develop clearer phylogenetic hypotheses, for which we need more taxon sampling, especially for the problematic genera, and additional DNA markers.
In this study, we assess phylogenetic relationships in Ebenaceae s.l. using plastid nucleotide sequences from three coding regions (atpB, ndhF, and matK) and three noncoding regions (trnK intron, trnL intron, and trnL-trnF intergenic spacer). Plastid coding regions, such as atpB, ndhF, and matK, have proven useful in addressing phylogenetic relationships at higher taxonomic categories, whereas noncoding regions (introns and intergenic spacers) are viewed as being more useful at lower taxonomic categories (see Soltis and Soltis, 1998
), but there are exceptions to this general pattern (e.g., Chase et al., 2000
; Pires et al., 2001
). Many studies (Reeves et al., 2001
; Kress et al., 2002
; Wang et al., 2004
; Barfuss et al., 2005
; Kathriarachchi et al., 2005
; Wurdack et al., 2005
; Soejima and Wen, 2006
) have used combined plastid DNA data from several regions, coding and noncoding, to provide improved resolution compared to single-region analyses. Our specific objectives were to (1) investigate monophyly of Ebenaceae s.l. when more taxa and more data are included in the analysis, (2) assess generic boundaries by evaluating relationships, and (3) identify characteristic features that may be putative synapomorphies useful for classification within Ebenaceae.
MATERIALS AND METHODS
Taxon sampling and plant material
Accession name, voucher citations, and GenBank numbers for all sequences are provided in Appendix 1. Ingroup sampling comprised 99 species, including all currently recognized genera of Ebenaceae, Diospyros, Euclea, and Lissocarpa (Wallnöfer, 2001
, 2004
) and representatives of taxa that were previously recognized at the generic level, e.g., Cargillia, Gunisanthus, Maba, Macreightia, Royena, and Tetraclis (de Candolle,
1844; Hiern, 1873
). In the case of the largest genus, Diospyros, our sampling represents all parts of its distribution (South, Central and North America, Africa, Madagascar, Asia, Australia and New Caledonia). Outgroup sampling included Lecythidaceae, Maesaceae, Marcgraviaceae, Pentaphylacaceae, Sapotaceae, Styracaceae, and Theaceae, as found in previous studies (Anderberg et al., 2002
; Bremer et al., 2002
; Schönenberger et al., 2005
).
Silica-gel-dried collections (Chase and Hills, 1991
) were obtained during fieldtrips in Brunei, Madagascar, New Caledonia, South America, and Thailand, as well as from the DNA bank of the Missouri Botanical Garden (http://mobot.mobot.org/W3T/Search/dna/projsdna.html) and the Royal Botanic Gardens, Kew (http://www.rbgkew.org.uk/data/dnaBank/homepage.html). The remaining DNA extractions were obtained from herbarium material at the Royal Botanic Gardens, Kew (K) and the Natural History Museum Wien (W).
DNA extraction, amplification, and sequencing
Total DNA was extracted following the 2x cetyltrimethyl ammonium bromide (CTAB) procedure of Doyle and Doyle (1987)
with minor modifications (see Muellner et al., 2005
); we washed ground plant material with sorbitol buffer (Tel-zur et al., 1999
) followed by centrifugation before incubating with CTAB buffer. Samples from Royal Botanic Gardens, Kew, and the DNA bank of Missouri Botanical Garden were extracted at Jodrell Laboratory, Royal Botanic Gardens, Kew, using the same procedure but with cleaning on a cesium chloride/ethidium bromide gradient (1.55 g·ml–1).
Amplifications of selected regions used 50-µL reactions containing 45 µL 1.1x ReddyMix PCR Master Mix (Advanced Biotechnologies, ABgene House, UK), 2 µL 1.0% bovine serum albumin (BSA), 1 µL each primer (20 mmol/L) and 1 µL template DNA. The PCR profile consisted of an initial 3-min premelt at 94°C and 36 cycles of 1-min denaturation at 94°C, 1-min annealing at 48°C, and a 1-min (or 2-min, depending on length of the amplified fragment) extension at 72°C, followed by final extension of 10 min at 72°C.
Primers used are those described in Hoot et al. (1995)
for atpB, Samuel et al. (2005)
for matK and the rest of the trnK intron, Olmstead and Sweere (1994)
and Oxelman et al. (1999)
for ndhF (3' portion), and Taberlet et al. (1991)
for trnL intron and trnL-trnF intergenic spacer. Products were purified with the Invisorb Spin DNA Extraction kit (Invitek GmbH, Berlin, Germany) or ExoSAP-IT kit (USB Corp., Cleveland, Ohio, USA). Cycle sequencing was performed using the ABI PRISM BigDye Terminator Cycle Sequence kit, version 3.1 (Applied Biosystems, ABI, Vienna, Austria) using the same primers that we used for amplification and the manufacturer's protocols. Sequences were initially edited using Sequence Navigator (ABI), and complementary sequences were assembled using AutoAssember version 1.4.0 (ABI).
A total of 107 atpB, 107 ndhF, 110 matK-trnK intron, and 109 trnL intron and trnL-trnF intergenic spacer sequences were newly generated for this study for the ingroup and outgroup taxa. The remaining 25 sequences of seven outgroup taxa were obtained from GenBank.
Sequencing alignment and phylogenetic analysis
Sequences were initially aligned with ClustalX (Thompson et al., 1997
), and the alignment was adjusted visually following the guidelines of Kelchner (2000)
. No indels were found in alignment of atpB, whereas length variation (in multiples of three) was observed in matK and ndhF. Most of the data matrices contained 99 ingroup taxa except the atpB and ndhF matrices, which contained 97 ingroup taxa. In the matK and trnK matrices, there are 15 outgroup taxa, but the rest of the data matrices contained 16 outgroup taxa (Table 2). For some of the outgroup taxa, data from different species within the same genus were combined into a single terminal, i.e., (1) for Marcgravia, M. rectiflora was available for atpB, ndhF, trnL intron, and trnL-trnF spacer and Marcgravia sp. for matK and trnK intron; (2) for Norantea, N. peduncularis was available for atpB and ndhF and N. guianensis for trnL intron and trnL-trnF spacer; and (3) for Maesa, M. myrsinoides was available for atpB and M. tenera for ndhF, matK, trnK intron, trnL intron, and trnL-trnF spacer.
|
; equal weight for all changes) were performed as implemented in PAUP* version 4.0b10 (Swofford, 2003
) with a two-stage search strategy. Nine data matrices were analyzed (see Table 2): (1) partial atpB, (2) partial ndhF, (3) matK, (4) trnK intron, (5) trnL intron (including trnL 3' exon), (6) trnL-trnF intergenic spacer, (7) combined coding regions, (8) combined noncoding regions, and (9) all six regions combined. Gaps were treated as missing data and not scored for inclusion because they did not provide any additional support for clades, i.e., the clades they characterized already had strong support from the sequence data; however, some indels are mapped onto Fig. 3 to demonstrate their congruence with the results found without their inclusion. Only the combined coding, noncoding, and all six regions results are shown. For each data set, heuristic searches were conducted using 1000 replicates random taxon addition, tree-bisection-reconnection (TBR) branch swapping, and "keeping multiple trees" (MulTrees) in effect, but saving only 10 trees per replicate. All trees found in the 1000 replicates were then used as starting trees for another search with a restriction of 15 000 trees, which were allowed to be swapped to completion. To assess support for each clade, bootstrap analyses (Felsenstein, 1985
) were performed with 1000 bootstrap replicates, TBR branch swapping, and simple sequence addition. Bootstrap percentages are described as high (85–100%), moderate (75–84%), or low (50–74%).
|
) as implemented in PAUP* version 4.0b10. We used 1000 replicates on parsimony-informative characters using TBR branch swapping with simple sequence addition and the MulTrees option in effect.
Bayesian inference (Huelsenbeck and Ronquist, 2001
; Lewis, 2001
) was performed with MrBayes 3.0b4 (Huelsenbeck and Ronquist, 2001
) for a matrix of the six regions combined. Modeltest 3.7 (Posada and Crandall, 1998
) was used to find the best-fitting substitution model; the GTR + I + G model was selected for the combined data matrix. Parameters set were nst = 6 and rate = invgamma; the others used were the default settings. The Markov chains, three heated and one cold, ran simultaneously starting from a random tree for one million generations, and trees were sampled every 100 generations. Trees that preceded stabilization of the likelihood value were discarded as the burn-in (1300 trees). The majority-rule consensus tree containing posterior probabilities (PP; Larget and Simon, 1999
) was built from the remaining trees sampled. Bayesian analyses were repeated four times to confirm results.
RESULTS
The data characteristics and statistics from the maximum parsimony analyses of the nine matrices are given in Table 2. The results of the six individual region analyses support for monophyly of Ebenaceae s.l. but only partially resolved relationships within the family (trees not shown). These relationships are much more resolved for matrices of the combined noncoding, combined coding, and the six regions combined (Figs. 1–4). Because of poor congruence of our molecular results with previous classifications of Ebenaceae s.l. and a need for clade names, we named the major clades (Figs. 1–4) recovered from our DNA analyses as A–Q (Table 3). There is no significant conflict (p value of ILD test = 0.31) for tree topologies among the six (atpB, ndhF, matK, trnK intron, trnL intron, and trnL-trnF intergeneric spacer) individual data sets.
|
|
|
Analysis of the six regions combined
Phylogenetic trees generated from maximum parsimony analysis of the combined matrix are shown as strict consensus tree in Fig. 2 and an individual tree (randomly selected) with branch lengths (DELTRAN optimization) in Fig. 3. Results show strong support for monophyly of all seven outgroup families, but interfamilial relationships between Ebenaceae s.l. and outgroups are not well resolved. Monophyly of Ebenaceae s.l. is strongly supported (BP 100) as well as that of the two major clades (labeled A and B, BP 100). The first major clade (A) contains only members of Lissocarpa (subfamily Lissocarpoideae sensu Wallnöfer) and is divided into two subclades that are congruent with the infrageneric classification of Wallnöfer (2004)
. The second major clade (B) contains all members of Ebenaceae s.s. and has two well-supported subclades: C/D (BP 100) and E (BP 100). Subclade C/D consists of two well-supported sister taxa, Euclea and Royena (BP 100; the latter previously recognized as Diospyros section Royena sensu White, 1980
, 1983
). Within the diverse clade E, which comprises all other taxa of Diospyros, Madagascan Maba and Tetraclis, there are two groups (F, BP 100; and G, BP 100, respectively) that are successive sisters to a large internally unresolved clade (H BP 83). Clade F contains two species of Diospyros subgenus Hierniodendron sensu Bakhuizen and clade G consists of Diospyros subgenus Cargillia sensu Bakhuizen and other two Diospyros species from New Caledonia. Although relationships within the large clade (H) that contains the remaining species of Diospyros and genus Tetraclis are unresolved, there are within it nine well-supported clades: I, J (BP 100), K (BP 93), L (BP 95), M (BP 96), N (BP 100), O (BP 100), P (BP 99), and Q (BP 73).
|
In the Bayesian analyses of the combined matrix, there is no major conflict of tree topologies among the four independent analyses. The majority-rule consensus tree is shown in Fig. 4. The Bayesian tree is generally congruent with that inferred from maximum parsimony (Fig. 2), except for the sister-group relationship of clade N and clade O to clade P/Q; these three clades appear as a polytomy in the parsimony results.
DISCUSSION
Utility of the plastid regions used
The combined analysis of all regions provides clearer phylogenetic relationships within the Ebenaceae s.l. than any of the individual analyses; this was observed in other studies as well and is now an expected phenomenon (Reeves et al., 2001
; Kress et al., 2002
; Wang et al., 2004
; Barfuss et al., 2005
; Kathriarachchi et al., 2005
; Wurdack et al., 2005
; Soejima and Wen, 2006
). Each of the six individual regions possessed different amounts of phylogenetic data (Table 2). Overall, matK and 3' ndhF are more useful than the other four regions (atpB, trnK intron, trnL intron, and trnL-trnF spacer) because these two regions have more parsimony informative characters; furthermore, variable sites within 3' ndhF change more often, which if sampled extensively provides more phylogenetic structure (Table 2; Asmussen and Chase, 2001
). The strict consensus trees resulting from the individual analyses of matK and ndhF displayed more consistently resolved clades (49 and 40, respectively) than those of the other four regions (34, 34, 21, and 14 for atpB, trnK intron, trnL intron and trnL-trnF spacer, respectively), and topologies of the matK and ndhF trees are most similar to that from the combined matrix (Fig. 2). Overall, ndhF resolved the deeper nodes better than the more terminal ones, which is in contrast with the results from analysis of matK. As expected from previous studies, atpB has low numbers of phylogenetically informative sites within Ebenaceae s.l.; it has only 192 parsimony-informative sites (13.75% of the total number of characters). The three noncoding regions (trnK intron, trnL intron, trnL-trnF spacer) and even the combined analysis of these three regions (Fig. 1B) did not clearly resolve phylogenetic relationships within Ebenaceae s.l., especially within the large and diverse genus Diospyros (clade E). Nonetheless, the addition of data from these three regions resulted in improved resolution, increasing the number of supported clades from 61 (Fig. 1A) to 71 (Fig. 2) and also increasing the bootstrap percentages.
This study is one of only a few examples illustrating the phylogenetic utility of ndhF at the infrageneric level (e.g., Clarkson et al. [2004
] used it within Nicotiana, Solanaceae). Generally, this marker has been used for phylogenetic analyses at the generic level and above (Olmstead and Sweere, 1994
; Olmstead and Reeves, 1995
; Davis et al., 2001
; Kim et al., 2001
; Levin et al., 2003
; Datwyler and Weiblen, 2004
; Lohmann, 2006
). In our study, ndhF helps resolve relationships within Diospyros (clade E, tree not shown). As mentioned, 3' ndhF provided more parsimony-informative characters (32.70%) than any of the other regions (13.75–31.04%), and also each individual site within ndhF changed more often than those in the other regions. Not surprisingly, matK was also found to be useful at infrageneric levels; many previous studies have confirmed its utility at this level (e.g., Miller and Bayer, 2001
; Wang et al., 2004
; Rønsted et al., 2005
; Kathriarachchi et al., 2006
).
Comparison between molecular evidence and previous classifications
This extensive study, including approximately 20% of the species of Ebenaceae, supports monophyly of Ebenaceae s.l. and is consistent with the recent inclusion of Lissocarpa in Ebenaceae (Berry et al., 2001
; APG II, 2003
) as the monogeneric subfamily Lissocarpoideae (Wallnöfer, 2004
). Putative synapomorphic characters of Ebenaceae s.l. are the presence of naphthoquinones, lack of stipules, extrafloral nectaries on abaxial leaf surfaces, a persistent calyx, usually unisexual flowers, and pendulous ovules (Wallnöfer, 2001
, 2004
). The similarity of wood anatomy of Lissocarpa and Diospyros was already pointed out in previous studies (Ng, 1971
, 1991
; Berry et al., 2001
; Lens, 2005
). Our results also provide a clear estimate of phylogenetic relationships within Ebenaceae s.l. and clarify generic boundaries that have been considered to be problematic in previous classifications. The infrafamilial classifications of de Candolle
(1844), Hiern (1873)
, Bakhuizen (1936–1955
), White (1980
, 1983
, 1993
), and Singh (2005) (see Table 1) are only partially congruent with our molecular results (Fig. 2); at least some of the generic or infrageneric taxa in each system are not monophyletic.
De Candolle (1844) recognized the following eight genera within Ebenaceae s.s.: Cargillia, Diospyros, Euclea, Gunisanthus, Maba, Macreightia, Rospidios, and Royena (Table 1). Compared with our molecular results, only two genera, Euclea and Royena, are monophyletic and independent from Diospyros, and the other four genera, Cargillia, Gunisanthus, Maba, and Macreightia, are embedded within Diospyros. According to our results, two species of Cargillia (represented here by D. australis and D. pentamera) are grouped together with two species of Diospyros from New Caledonia (clade G) and are the sister group of Diospyros clade H (Fig. 2). The monotypic genus Gunisanthus with only D. pilosula does not warrant generic status on morphological grounds and is closely related to two Asian Diospyros species in clade Q (BP 94). All of them share long, slender pedicels and tetralocular ovaries. Maba sensu de Candolle
forms a monophyletic group together with some Madagascan Diospyros and Tetraclis species. Only one species of the genus Macreightia, represented here by D. crassinervis (occurring in the Carribean islands), was included in our study. It is strongly supported as a sister to D. tetrasperma (occurring in Central America and the Carribean islands) in clade Q. In addition the infrageneric classification of Diospyros sensu de Candolle
is artificial because the four sections of Diospyros that he recognized are not monophyletic.
In the last worldwide revision of Ebenaceae s.s., five genera were recognized by Hiern (1873)
: Diospyros, Euclea, Maba, Royena, and Tetraclis (Table 1). He recognized Tetraclis as an independent genus because of its valvate corolla, and Perrier de la Bâthie (1952)
used the same feature to distinguish three species of Tetraclis. As shown by our results (Figs. 2–4), Tetraclis (T. baroni and T. cf. clusiaeflora) does not deserve generic status because it groups with Madagascan species of Diospyros and other members of the D. ferrea complex (Maba sensu de Candolle) in clade L. The genus Maba sensu Hiern, which is broader than Maba sensu de Candolle,
is a group of species with usually trimerous flowers. It consists of the following six sections, Ferreola, Macreightia, Holochilus, Rhipidostigma, Barberia, and Trichanthera. Maba as circumscribed by Hiern is polyphyletic in our results. Species of this concept appear in many different positions within Diospyros, such as in clade F (Maba section Barberia, represented here by D. maingayi and D. puncticulosa), clade L (Maba sections Ferreola and Holochilus, represented here by the D. ferrea complex and D. natalensis, respectively), clade O (Maba section Holochilus, with D. abyssinica), and clade Q (Maba sections Macreightia and Rhipidostigma, with D. crassinervis, D. confertiflora, D. fasciculosa, and D. venosa). Tetraclis and Maba are thus embedded within Diospyros sensu Hiern. As is the case in de Candolle's system, only two genera, Euclea and Royena, are strongly supported as monophyletic. Fifteen sections were recognized in the infrageneric classification of Diospyros sensu Hiern (Table 1). However, members of section Melonia turned up in three clades, J (D. fulvopilosa = D. kurzii), K (D. mespiliformis), and Q (D. venosa = D. rotundiflora). The species of section Gunisanthus are separated into two clades, L and Q (D. gracilipes and D. pilosula, respectively). The members of section Patonia are also placed in two clades: K (D. philippinensis) and Q (D. maritima and D. undulata). Section Danzleria is polyphyletic; the first group of this section occurs in clade N (D. lotus, D. kaki, and D. virginiana) and the second in clade Q (D. montana). Section Paralea is polyphyletic and embedded in three clades, G (D. macrocarpa and D. pentamera), K (D. cauliflora and D. oblonga), and Q (D. diepenhorstii, D. olen, and D. texana); even in those clades the representatives of section Paralea do not fall together.
Bakhuizen (1936–1955) studied Ebenaceae s.s. extensively in Southeast Asia; only Diospyros occurs there, and he basically adopted the system of Hiern (1873)
for species occurring outside his area of study (Table 1). Bakhuizen pointed out that the distinction of Maba from Diospyros was unclear and often arbitrary, and so he recognized Maba as a subgenus of Diospyros. Although he included Maba within Diospyros, his concept of Diospyros is still not monophyletic because Tetraclis is embedded within it. Bakhuizen divided Diospyros into five subgenera: Cargillia, Diospyros (his "Eudiospyros"), Hierniodendron, Maba, and Mabacea (Table 1). The first four genera are included in our study. According to our results, only two subgenera, Hierniodendron (F) and Cargillia (G), are confirmed as monophyletic and both are small. Neither subgenus Maba nor Diospyros forms an independent clade (Fig. 2). Three sections of subgenus Maba are embedded within Diospyros and fall in two different clades. Section Ferreola, represented here by D. ferrea, D. flavocarpa, and D. labillardierei, is placed in clade L. Both sections Miquelia (represented here by D. andamanica and D. tahanensis) and Rhipidiostigma (with D. confertiflora, D. fasciculosa, and D. venosa) are placed in clade Q. Our study, which includes 16 of 32 sections of subgenus Diospyros sensu Bakhuizen finds the same problem at this level. Sections Podophora (D. dasyphylla, D. diepenhorstii, and D. insidiosa), Kurzella (D. fulvopilosa and D. pubicalix), and Nesindica (D. macrocarpa, D. maritima, D. styraciformis, and D. undulata) are polyphyletic. Although the remaining sections are not split, most of them do not form clades of their own. Thus, members of section Brachycylix (D. philippinensis) group with sections Campanulata (D. bejaudii and D. ridleyi), Confertiflora (D. curranii), Ebenaster (D. celebica), Ptychocylix (D. ferruginescens and D. oblonga), and Stelechantha (D. cauliflora); together, they make up most of clade K. Section Rigidophylla (D. rigida) forms a clade (P) with sections Glutinosa (D. malabarica var. atrata and D. malabarica var. malabarica) and Saccocalix (D. mindanaensis). Section Acanthebenus (D. montana) is placed in clade Q together with three other sections, Basithrix (D. ferox), Ebenus (D. olen), and Podophora (D. dasyphylla, D. diepenhorstii, and D. insidiosa), in which they are polyphyletic.
Only two genera, Diospyros and Euclea, were recognized in Ebenaceae s.s. by White (1980, 1983). He lumped Royena and Tetraclis within an enlarged Diospyros due to the lack of distinguishing characters (White and Barnes, 1958
). Diospyros in its broader circumscription is paraphyletic. In our study, which included four species each of both Euclea and Royena (= White's Diospyros section Royena), clearly showed a close relationship of these two groups (clades C and D). Our results also do not support White's (1980, 1993) infrageneric classification of Diospyros. The African Diospyros taxa are not monophyletic as White (1980)
suggested. Most of them do not form a clade of their own, but instead group with species from other areas, e.g., D. mespiliformis with taxa from Asia (clade K), D. natalensis and D. consolatae with Asian–Madagascan–New Caledonian taxa (L), whereas six more African species group with South American taxa in clade M. Only clade N consists entirely of African species (Fig. 3). Except for the problematic section Royena (as mentioned), there are eight other sections in his system, Brevistyla, Brevituba, Calvitiella, Dodonium, Forbesia, Latibulum, Maba, and Tabonaca that are included in our study, and most of them are not monophyletic. Diospyros consolatae of section Forbesia grouped with members of section Maba (D. ferrea and D. natalensis) in clade L. Two species from section Tabonaca are embedded in different places: D. fragrans groups with D. mannii (section Dodonium) in clade M, but D. pseudomespilus falls with D. abyssinica (section Brevituba) in clade O. Diospyros bipindensis (section Latibulum) is embedded within section Calvitella (represented here by D. cooperi and D. melocarpa; clade M), which makes the latter paraphyletic.
The recent monograph of Singh (2005)
covering the Indian region follows the generic concepts of White (1980
, 1983
). For an infrageneric classification, he divided the Indian Diospyros into 27 sections, which were mainly adopted from the system of Bakhuizen (1936–1955
). The 12 sections that he adopted from Bakhuizen are not monophyletic. He also established 10 new sections to accommodate other Indian taxa. Our results do not support his sectional concepts. The members of his section Lotus occur in clade N (Diospyros lotus) and Q (D. montana) (Figs. 2–4). Additionally, D. lotus is grouped in clade N together with members of three of his other sections: Kaki (D. kaki), Sylvatica (D. glandulosa), and Basithrix (D. virginiana).
Circumscription and a new classification of Ebenaceae s.l
The differing circumscriptions of the five previous systems just described suggest that there are problems of generic delimitation in Ebenaceae s.s.; furthermore, Lissocarpa has newly been included in Ebenaceae. Here, we present a new classification for Ebenaceae s.l. based on a phylogenetic approach. Our results clearly resolve relationships at the subfamily and generic levels. A classification of Ebenaceae compatible with our results (Figs. 2–4) is the following: two subfamilies, Lissocarpoideae and Ebenoideae, with four genera, Lissocarpa, Euclea, Royena, and Diospyros. Synapomorphic characters of this family were described in the first paragraph of section comparison between molecular evidence and previous classification. A key to the subfamilies and genera is presented in Table 4, and relationships among the subfamilies and genera are shown in a schematic diagram (Fig. 5). All clades, with the exception of Diospyros clade Q (BP 73) received high bootstrap support. The characters of the four genera of Ebenaceae are presented in Table 5.
|
|
|
). Based on previous molecular results with limited sampling (Berry et al., 2001
; Anderberg et al., 2002
; Bremer et al., 2002
; APG II, 2003
), Lissocarpa was formally transferred to Ebenaceae as the monogeneric subfamily Lissocarpoideae (Wallnöfer, 2004
). Our much more extensive study is consistent with inclusion of Lissocarpa in Ebenaceae and placement in a distinct subfamily. Lissocarpa shares various characters with other members of Ebenaceae, e.g., the black color of roots and bark, extrafloral nectaries (flachnektarien) on abaxial leaf surfaces, a persistent calyx, unisexual flowers, biovulate carpels with pendulous ovules, and a similar wood anatomy (Ng, 1991
; Berry et al., 2001
; Wallnöfer, 2001
, 2004
; Lens, 2005
). However, Lissocarpa is distinguished by several morphological characters from genera in Ebenoideae, e.g., absence of an indumentum, vessels with both scalariform and simple perforation plates, subopposite bracteoles, triporate pollen grains, and an inferior ovary (Table 5). Although Lissocarpa displays some advanced (derived) characters (inferior ovary and corona), it lacks a 12-bp deletion in matK (Fig. 4) that is present in all members of Ebenoideae. This molecular feature is plesiomorphic and distinguishes it from the rest of Ebenaceae.
Arrangement of the eight known species of Lissocarpa into two sections, Lissocarpa and Enho (Wallnöfer, 2004
), is also compatible with the results of our study. Section Lissocarpa (represented here by L. guianensis, L. benthamii, and L. kating) is characterized by a corona that originated from staminodia and a strongly elevated midvein on the adaxial leaf surface. On the other hand, flowers of species belonging to section Enho (represented here by L. ronliesneri and L. tetramera) do not possess a corona, and the midvein on the adaxial leaf surface is strongly sunken.
Subfamily Ebenoideae
This subfamily consists of three genera, Euclea, Royena, and Diospyros. The shared features of this subfamily are the presence of indumentum, vessels with only simple perforation plates, alternate bracteoles, tricolporate pollen grains, superior ovary, and absence of a corona. This subfamily can be further subdivided into two major lineages. The first consists of two genera, Euclea and Royena, which are mainly restricted to southern Africa (except for a few species of Euclea that occur northward to the Arabian Peninsula, Socotra, and the Comoro Islands). This pattern of relationships was also observed in previous molecular studies (Morton et al., 1997; Berry et al., 2001
). A relationship between Euclea and Royena is well supported by various characters such as seed anatomy and a deletion of 69 bp in the 3' trnK intron (Fig. 3). On a morphological basis, species belonging to the first clade can be distinguished from the second by the presence of an ingrowth of the testa surrounding the radicle of the embryo. Royena and Euclea also usually possess a pericyclic phellogen, in contrast to Diospyros, which has a subepidermal phellogen (Hiern, 1906
). The second lineage contains only the genus Diospyros, which is pantropical. It is more diverse in its morphology (Table 5) and DNA divergence (Fig. 3) than the other three genera. Our results allow us to lump all other previously recognized genera (Cargillia, Gunisanthus, Maba, Macreightia, and Tetraclis) within an expanded Diospyros. This genus is distinguished from the other two genera in Ebenoideae by the lack of an invagination surrounding the radicle.
Euclea
The 20 species of Euclea are confined to Africa, Arabia, Socotra, and the Comoro Islands. This genus can be separated from Royena and Diospyros (including Maba) by the following characters: calyx not accrescent in fruit, staminodes usually absent from female flowers, inflorescence usually pseudoracemose, seed usually subglobose, radicle completely surrounded by invagination of testa, and an embryo with cotyledons flexed at right angles to the radicle (Hiern, 1873
; Ng, 1971
; White, 1983
). Euclea female flowers are distinctly smaller than male flowers, but in Diospyros female flowers are often larger (White, 1983
). With our limited sampling of species, we cannot evaluate de Candolle's (1844) infrageneic classification of Euclea.
Royena
This genus has been recognized for a long time as distinct (de Candolle,
1844; Hiern, 1873
; Bakhuizen, 1936–1955
). According to Hiern (1873
, 1906
), it can be distinguished from Diospyros and Euclea in having structurally hermaphroditic flowers (pistillodes and antherodes well developed so that the flowers appear bisexual) and stamens in one row (vs. dioecious or rarely polygamous; stamens usually in two or more rows, often in pairs). However, this genus was lumped in Diospyros only a few decades ago (de Winter and White, 1961
; de Winter, 1963
; White, 1980
). The main reason for this action was the discovery of Salter (1953)
that the flowers of Royena glabra are only structurally but not functionally hermaphroditic, and thus it is dioecious, as is (with some exceptions) usually the case in Diospyros. According to our results, which are basically congruent with those of previous studies (Morton et al., 1997; Berry et al., 2001
; APG II, 2003
), it is preferable to resurrect Royena. All species of this genus examined in our study share a 6-bp deletion in matK. Royena is distinguished from Euclea by its invariably structurally hermaphroditic flowers, usually accrescent calyx (vs. not accrescent), and many-seeded (vs. one-seeded) fruits (see Table 5). Royena has a center of distribution in southern Africa (White, 1988
).
Diospyros
Diospyros contains more than 500 species distributed in the tropics and subtropics. As circumscribed here, it includes the other previously recognized genera, Cargillia, Gunisanthus, Maba, Macreightia, and Tetraclis (de Candolle,
1844; Hiern, 1873
) and excludes members of section Royena sensu White (1980)
. Nomenclatural changes will be necessary for the taxa from Madagascar previously placed in Maba and Tetraclis. Our results support two successive sister clades (F and G) and a large internally unresolved clade within this genus. Although relationships within the last remain unclear, nine well-supported clades (I–Q) were found in our study (Fig. 2–4). As mentioned, earlier infrageneric classifications of Diospyros (i.e., de Candolle,
1844; Hiern, 1873
; Bakhuizen, 1936–1955
; White, 1980
, 1993
; Singh, 2005
) are not compatible with our results. However, due to the lack of a clear estimate of relationships, we could not construct a new infrageneric classification. Further work is necessary and will not only require more data, both molecular and morphological, but also more extensive taxon sampling.
Diospyros clade F
This clade corresponds to Bakhuizen's subgenus Hierniodendron and is placed as sister to the rest of Diospyros. Both D. maingayi and D. puncticulosa are restricted to Southeast Asia where they grow in peat swamps and true tropical rain forest. Their filaments are fused together to form a hollow tube, which has not been observed in any other Diospyros species examined in our study. This group is also supported by a 6-bp indel in ndhF (Fig. 3).
Diospyros clade G
This group contains D. subgenus Cargillia sensu Bakhuizen (D. australis and D. pentamera) plus two New Caledonian species, D. brassica and D. macrocarpa. They share a 49-bp deletion in the trnK intron (Fig. 3). Their close relationship is supported by a combination of biogeographic and morphological evidence. All of them are confined to Australia and New Caledonia (eastern Gondwana) and have subsessile, axillary inflorescences, a deeply lobed corolla, and exserted anthers.
Diospyros clades I, J, and K
All species of these three clades share a 24-bp deletion in the trnL intron (Fig. 3). Diospyros borneensis (I) may be sister to the rest. It is distinguished from the other two clades by a cylindrical or funnel-shaped calyx and woody pericarp. Clade J has two Asian species, D. fulvopilosa and D. mollis, both of them restricted to Southeast Asia. Clade K consists of nine Asian species, D. bejaudii, D. cauliflora, D. celebica, D. curranii, D. ferruginescens, D. oblonga, D. brandisiana, D. philippinensis, and D. ridleyi and also the widespread African species, D. mespiliformis (White, 1988
). Up to now, we have identified no unique morphological features for this clade as a whole. However, a ruminate endosperm is a unifying character for D. mespiliformis and five closely related species (D. bejaudii, D. celebica, D. ferruginescens, D. philippinensis, and D. ridleyi). Long-distance dispersal perhaps played a role for this Africa–Asia connection. The fruits of both African (White, 1983
) and Asian (Ng, 1978b
; S. Duangjai, personal observation) species are edible. In Thailand their fruits are eaten by mammals (S. Duangjai, personal observation), which may be dispersing seeds of this group.
Diospyros clade L
This clade contains the D. ferrea complex (including Madagascan Maba) and all Madagascan Diospyros species (including Tetraclis). The D. ferrea complex has been referred to genus Maba sensu de Candolle
or D. sections Ferreola plus Cupulifera sensu Bakhuizen, and it is represented here by D. ferrea, D. flavocarpa, D. labillardierei, D. natalensis, Maba magnifolia, and M. myriophylla. This group has trimerous flowers with a trilocular ovary that is biovulate. It occurs throughout the Old World tropics: Africa, Madagascar, Asia, Polynesia, Australia, the Pacific islands, and Hawaii. Most have small fruits, which may be relatively easily dispersed across water barriers by migratory birds (Pannell and White, 1988
; White, 1993
). Our results strongly support a close relationship between the D. ferrea complex and the Madagascan Diospyros species, but within the clade, resolution is poor. The question of the origin of Madagascan Diospyros from the African mainland is still open. More intensive study will provide robust estimate of phylogenetic relationships, and this should enable us to evaluate better this hypothesis.
Diospyros clade M
This clade consists of African and South American species, and it is further divided into three subclades. The first two subclades comprise only African species (D. bipindensis, D. cooperi, D. fragrans, D. melocarpa, D. mannii, and an unidentified species), all of them are distributed in the western side of Africa, mainly in the Guineo-Congolian region (White, 1978
). The third subclade contains the South American species, all of which have a 42-bp deletion in the trnL intron (Fig. 3).
Diospyros clade N
This group contains five temperate or subtemperate species, D. glandulosa, D. glaucifolia, D. kaki, D. lotus, and D. virginiana. A relationship of these species is supported by a suite of morphological characteristics and was reported in a previous study (Morton et al., 1996
). Fruits of this group are edible, and three species have been brought into cultivation: persimmon (D. kaki), date plum (D. lotus), and common persimmon (D. virginiana). Polyploidy is reported in D. kaki (2n = 60, 90, and 135) and D. virginiana (2n = 60 and 90). The other three species, D. glandulosa, D. glaucifolia, and D. lotus, are diploids (2n = 30). Based on morphological characters, D. glandulosa was proposed as the progenitor of D. kaki (Ng, 1978a
). This hypothesis is consistent with our results: D. glandulosa is sister to D. kaki.
Diospyros clade O
This clade contains three African species, D. abyssinica, D. pseudomespilus, and an unidentified species. Although this clade is well supported by molecular data, shared morphological characters are still unclear. Further investigation with more extensive sampling is necessary to clarify synapomorphic characters for this African group.
Diospyros clades P and Q
The close relationship of these two clades has low support. Clade P consists of D. malabarica, D. mindanaensis, and D. rigida, all of which are distributed in southern and Southeast Asia. Species of this clade are characterized by a combination of the reddish inner bark and ruminate endosperm. Clade Q is formed by American, Asian, and New Caledonian species, and we could not find any unifying morphological features for this clade. According to the combined coding region tree (Fig. 1A), there are four subclades, which seem to be correlated with biogeography. The first subclade is confined to Asia and New Caledonia, the second contains American species, and the other two are restricted to Asia.
Character evolution
Our results (Fig. 4) indicate that a superior ovary is plesiomorphic in Ebenaceae s.l., with a shift to an inferior ovary only one time in Lissocarpa, because a superior ovary also appears in the outgroup families. Tetramerous and pentamerous flowers are plesiomorphic in Ebenaceae s.l., and they are more common than trimerous and hexamerous flowers (Fig. 4). Tetramerous and pentamerous flowers are present in all four genera, Lissocarpa, Euclea, Diospyros, and Royena, but trimerous and hexamerous flowers appear only in Diospyros. Trimerous and hexamerous flowers might not have evolved from the same ancestral condition; our results (Fig. 4) indicate that the trimerous flowers have evolved from the tetramerous flowers, whereas the hexamerous flowers are derived from the pentamerous flowers.
Mapping of the number of floral parts on the phylogenetic tree (Fig. 4) provides evidence that the trimery has evolved from tetramery more than one time. This is important to note because previous taxonomists (Hiern, 1873
; Bakhuizen, 1936–1955
) always used this character to separate the Maba (Table 1) from the other taxa. The first lineage of trimerous flowers is the D. ferrea complex in clade L. The others occur in clade M (D. melocarpa), O (D. abyssinica), and Q (D. andamanica, D. tahanensis, D. castanea, D. confertiflora, D. crassinervis, D. fasciculosa, and D. venosa). These clades differ in character of ovary: the first clade has three biovulate locules, but the others have an ovary with six uniovulate locules (Bakhuizen, 1936–1955
). In addition, D. venosa has both trimerous and tetramerous flowers on the same tree (S. Duangjai, personal observation).
Hexamerous flowers are sometime found in D. artanthaefolia, D. ropourea, and D. vestita in clade M, whereas the members of this clade usually have tetramerous or pentamerous flowers (Fig. 4). Thus, hexamerous flowers seem to be derived from pentamerous flowers.
Biogeographical implications
The hypothesis of a western Gondwanan origin of Ebenceae (Raven and Axelrod, 1974
) is supported by our molecular results (Fig. 3). The basal nodes can be assigned to only this region: Lissocarpa is found in the northern half of South America, while Euclea and Royena are centered in southern Africa. The distribution of the family perhaps is thus a result of both vicariant and long-distance dispersal events. As suggested earlier in our discussion, there are several intercontinentally disjunct clades: Asian–African (clade K), African–South American (clade M), Asian–North American (clade N), and American–Asian–Australian–New Caledonian–Pacific Island (clade Q). Furthermore, some members of D. ferrea complex (clade L) also occur on volcanic islands, e.g., Hawaii (Pannell and White, 1988
) and Mauritius (Richardson, 1981
). However, due to unresolved relationships within Diospyros, we are unable to identify the direction of migration. The biogeography of this family needs to be further explored with a phylogenetic hypothesis based on more extensive sampling in Diospyros.
Conclusions
The monophyly of Ebenaceae s.l. is strongly supported by analyses of data from six plastid DNA regions. Four genera, Lissocarpa, Euclea, Royena, and Diospyros, should be recognized within the family. Our results support inclusion in Diospyros of the other previously recognized genera, e.g., Cargillia, Gunisanthus, Maba, Macreightia, and Tetraclis. The family is split into two major clades corresponding to subfamilies Lissocarpoideae (only Lissocarpa) and Ebenoideae (the other three). Molecular evidence confirms the two sections (Lissocarpa and Enho) proposed by Wallnöfer (2004)
within Lissocarpa. Royena is more closely related to Euclea than Diospyros. It also is clear that the previous infrageneric classifications of Diospyros are artificial and need to be revised. Further study will require more molecular and morphological data as well as more extensive sampling of species. Additionally, the plastid markers provide a powerful framework for identification of characters that represent synapomorphies useful in classification of this family, and by and large morphology confirms the molecular-based groups identified here.
Editor's note (3 Jan 07): This online article differs from the print version; see press erratum for print journal in January vol. 94(1).
|
1 For providing plant material that made this study possible, the authors thank the Royal Botanical Gardens at Kew, the Missouri Botanical Garden, the Botanical Garden of University of Vienna, and the following individuals: A. Sinbumrung, J. H. Ali Ahmad, M. Ariffin, A. Sieder, G. Fisher, F. Rakotonasolo, A. Britt, M. Rakotoarinivo, M.-F. Prévost, J.-F. Molino, D. Sabatier, E. J. M. M. Arets, P. E. Berry, H. Kurzweil, M. D. Pirie, R. O. Frisch, S. A. Mori, H. Rainer, P. J. M. Maas, J. R. Abbott, R. Kutalek, A. Prinz, A. Anderberg, L. W. Chatrou, J. Stone, R. Niangadouma, M. Merello, H. Schmidt, J. Amponsah, A. Welsing, K. Baah, G. McPherson, P. Lowry II, F. Carriconde, D. I. Letocart, G. Walters, A. Bradley, G. N. Essouma, A. Mbaniboua, M. Chintoh, J. Rabenantoandro, N. B. Zimba, B. Luwiika, D. K. Harder, R. Rabevohitra, R. Vasquez, and R. Ortiz-Gentry. They also thank R. De Kok, J. Gregson, L. Csiba, E. Kapinos, V. Klejna, M. H. J. Barfuss, H. Kathriarachchi, O. P
un, and H. Voglmayr for their help in various ways. Fieldwork was conducted in Brunei in collaboration with BRUN, the Brunei Forestry Centre; the authors are grateful to the entire staff of BRUN for their assistance. R.S. received financial support from FWF (Fonds zur Förderung der Wissenschaftlichen Forschung, project no. P 17094 B03). 
2 Author for correspondence (e-mail: mary.rosabella.samuel{at}univie.ac.at
)
LITERATURE CITED
Anderberg A. A. Rydin C. Källersjö M.. 2002. Phylogenetic relationships in order Ericales s.l.: analyses of molecular data from five genes from the plastid and mitochondrial genomes. American Journal of Botany 89: 677-687.
APG II.. 2003. An update of the Angiosperm Phylogeny Group classification for the orders and families of flowering plants: APG II. Botanical Journal of the Linnean Society 141: 399-436.[CrossRef]
Asmussen C. B. Chase M. W.. 2001. Coding and noncoding plastid DNA in palm systematics. American Journal of Botany 88: 1103-1117.
Bakhuizen van den Brink R. C.. 1936–1955. Revisio Ebenacearum Malayensium. Bulletin du Jardin Botanique de Buitenzorg, sèrie 3 15: 1-515.
Barfuss M. H. J. Samuel R. Till W. Stuessy T. F.. 2005. Phylogenetic relationships in subfamily Tillandsioideae (Bromeliaceae) based on DNA sequence data from seven plastid regions. American Journal of Botany 92: 337-351.
Berry P. E. Savolainen V. Sytsma K. J. Hall J. C. Chase M. W.. 2001. Lissocarpa is sister to Diospyros (Ebenaceae). Kew Bulletin 56: 725-729.
), tetramerous flowers (
), pentamerous flowers (*), and hexamerous flowers (
)
