An evaluation of tribes and generic relationships in Melioideae ( Meliaceae ) based on nuclear ITS ribosomal DNA

1 Molecular Systematics Section, Jodrell Laboratory, Royal Botanic Gardens Kew, Richmond, Surrey TW9 3DS, U.K. 2 Present Address: Grunelius-Moellgaard Laboratory, Department of Botany and Molecular Evolution, Research Institute Senckenberg, Senckenberganlage 25, 60325 Frankfurt am Main, Germany. alexandra.muellner@senckenberg.de (author for correspondence) 3 Department of Systematic and Evolutionary Botany, Faculty Center Botany, University of Vienna, Rennweg 14, 1030 Vienna, Austria 4 Division of Biology and Medicine, Brown University, Providence, Rhode Island 02912, U.S.A.


INTRODUCTION
Meliaceae are a widely distributed subtropical and tropical angiosperm family occurring in a variety of habitats, from rain forests and mangrove swamps to semideserts (Pennington & Styles, 1975;Pennington & al., 1981;Pannell, 1992;Mabberley & al., 1995).Together with the contributions on Meliaceae in Flora Neotropica by Pennington & al. (1981) and in Flora Malesiana by Mabberley & al. (1995), the most authoritative work on the family is the generic monograph by Pennington & Styles (1975).Currently recognized are 49 to 51 genera with about 565 species (Pennington & Styles, 1975;Mabberley & al., 1995;Cheek, 1996;Chase & al., 1999;Mabberley, 2000).Pennington & Styles (1975) recognized four subfamilies, of which Melioideae and Swietenioideae consist of seven tribes with 34 to 36 genera and three tribes with 13 genera, respectively.Quivisianthoideae and Capuronianthoideae each contain a single monotypic genus (Quivisianthe Baill.and Capuronianthus Leroy) and were newly recognized by Pennington & Styles (1975).A re-cent reassessment of the circumscription of the four subfamilies by means of phylogenetic analyses of sequence data from three regions (plastid rbcL,matK,nuclear 26S rDNA) showed that the members of the two small monogeneric subfamilies, Quivisianthe and Capuronianthus, fall within Melioideae and Swietenioideae, respectively, supporting their taxonomic inclusion in these groups (Muellner & al., 2003).Pennington & Styles (1975) found a wide range of morphological variation especially in the subfamily Melioideae.To obtain an improved tribal scheme compared to that of Harms (1940), Pennington & Styles (1975) subordinated the supposed evolutionary significance of individual characters in favour of groupings based on correlations between the maximum number of characters of use at this level of classification and on detection of discontinuities in variation of these characters.Pennington & Styles (1975) argued that the most natural grouping of genera was obtained by basing classification on a large number of characters; thus, artificial assemblages resulting from the weighting of a few characters were avoided.Using these principles, Pennington & Styles (1975) recognized seven tribes within subfamily Melioideae but stated that limits of Trichilieae, Aglaiaeae and Guareeae could only be defined by overlapping morphological, anatomical and palynological characters.All tribes of Melioideae are represented in Malesia, but only two (Guareeae, Trichilieae) are pantropical and two other ones are restricted to the Old World (Turraeeae, Melieae); the remainder are restricted to Indomalesia and the western Pacific (Vavaeeae, Aglaieae, Sandoriceae;Mabberley & al., 1995).
The internal transcribed spacers (ITS) of nuclear ribosomal DNA (nrDNA), defined as the unit containing the ITS1 spacer, 5.8S rRNA gene and ITS2 spacer, are not only useful in assessing relationships at the infrageneric, but also at higher taxonomic levels (Hershkovitz & Zimmer, 1996;Soltis & Soltis, 1998).Secondary structure models of RNA transcripts, employed in the taxonomic group under investigation, allow for optimizing alignment of variable and putatively phylogenetically informative regions of ITS even across more distantly related taxa.This is due to the fact that the secondary structure of ITS is more conserved than the primary sequence (Mai & Coleman, 1997;Coleman & al., 1998).
In this study we performed maximum parsimony, maximum likelihood and Bayesian analyses of sequence data from nuclear ITS to estimate phylogenetic relationships within subfamily Melioideae for which former analyses of plastid rbcL, matK and nuclear 26S rDNA did not provide sufficient information (Muellner & al., 2003).Based on 51 species, including representatives of all currently recognized tribes, this study thus provides the first detailed reassessment of tribal and generic relationships in Melioideae.The ITS data are compared to rbcL data recently collected in the course of a survey on the biogeographic history of Meliaceae (Muellner & al., 2006).

MATERIALS AND METHODS
Plant material.-We analysed ITS sequences of 51 species of subfamily Melioideae (ingroup) and one species each of genera Swietenia Jacq., Khaya A. Jussieu, Toona (Endl.)M. Roem.and Cedrela P. Browne of subfamily Swietenioideae as outgroups (Appendix).The justification for the inclusion of the ingroup taxa in Melioideae and Swietenia, Khaya, Toona and Cedrela in Swietenioideae was based on a previous evaluation of the higher-level classification of Meliaceae (Muellner & al., 2003).
Plant material was collected during excursions to Thailand, Malaysia, Sri Lanka and Australia and from the living collections of Forestry Research Institute Malaysia (FRIM), Kebun Raya (Bogor Botanic Garden), Indonesia, and the Royal Botanic Gardens, Kew, U.K. Herbarium specimens are deposited at FHO, FR, K, NBG, NCU and WU.
Isolation of DNA and amplification.-Fieldcollected material was dried and stored in silica gel prior to DNA extraction (Chase & Hills, 1991).DNA extraction and PCR amplification followed Muellner & al. (2005).The fragment size amplified was between 627 and 664 bp for the entire ITS region.After amplification, samples were gel purified using the QIAquick gel extraction kit (QIAGEN, Margaritella, Vienna, Austria).
Sequencing. -PCR primers were also used for sequencing.Cycle-sequencing followed Muellner & al. (2005).Sequencing reactions were run on an ABI PRISM 377 DNA Sequencer and on an ABI 3100 capillary sequencer following the manufacturer's protocols.
Sequence editing and alignment.-For editing and assembly of the complementary strands, the software programs Autoassembler version 1.4.0 (Applied Biosystems) and DNA STRIDER version 1.2 (Christian Marck, CEA -Commissariat à L'ènergie Atomique/Saclay, France) were used.ITS sequences were explored for the presence of several structural motifs.Thus, in the ITS1 region we searched for the presence of the conserved angiosperm motif GGCRY-(4 to 7 n)-GYGYCAAGGAA (Liu & Schardl, 1994), which was also found in several gymnosperms (Maggini & al., 1998).We also looked for the presence of the conserved (C1-C6) and variable (V1-V6) domains determined for plant ITS2 sequences (Hershkovitz & Zimmer, 1996), as well as for the conserved angiosperm motif 5′-GAATTGCAGAATCC-3′ within the 5.8S rDNA gene, which can be used to differentiate between flowering plants, fungi and algae (Jobes & Thien, 1997).Folding predictions of secondary structures of the ITS1 and ITS2 RNA transcripts were made at the M. Zuker web server (http://www.bioinfo.rpi.edu/~zukerm/) by use of the mfold program version 3.1 (Mathews & al., 1999;Zuker & al., 1999).Foldings were conducted at 37°C.After a first rough alignment with CLUSTAL version X (Thompson & al., 1997), corrections were made manually by using secondary structure predictions of ITS1 and ITS2 RNA transcripts as a guide for alignment across genera.Secondary structure predictions were confirmed by hemi-compensatory base changes (hemi-CBCs) and full compensatory base changes (CBCs) that preserved the predicted folding pattern.First, the secondary structure used was not always the energetically most favourable, but rather the folding that was common to all genera and species and supported by CBCs and hemi-CBCs.Second, the structural motifs common to all eukaryote ITS2 (Coleman, 2007) were present there, exactly in their expected positions in the secondary structure.These were the most conserved sequences, as also expected.A total of 794 aligned positions were included in the matrices for phylogenetic analyses for ITS (including ITS1, 5.8S rDNA and ITS2).Gaps were coded as missing data.All sequences are deposited in GenBank (http://www.ncbi.nlm.nih.gov/).Phylogenetic analysis.-Individual maximum parsimony (MP) analyses of the ITS and the rbcL dataset (data for the latter region were obtained from Muellner & al., 2006) were performed using PAUP*4.0b10(Swofford, 2002).Visual inspection of the individual bootstrap consensus trees was used for determining congruence of datasets (Whitten & al., 2000).Although there were strongly supported ( > 85% bootstrap), incongruent patterns among the individual analyses, direct combination was carried out to confirm observations based on the separate analyses (trees not shown).Substitutions at each nucleotide position were treated as independent, unordered, multi-state characters of equal weight (Fitch parsimony;Fitch, 1971).Heuristic searches were performed using 1,000 random additions of taxa, tree bisection-reconnection (TBR) branch swapping and MulTrees on (keeping multiple, shortest trees).Robustness of clades was estimated using the bootstrap (Felsenstein, 1985) with 5,000 replicates with simple sequence addition, TBR branch swapping and MulTrees on.
Bayesian analyses were conducted with MrBayes version 3.01 (Huelsenbeck & Ronquist, 2001) using four Markov chains simultaneously started from random trees.Modeltest 3.06 (Posada & Crandall, 1998, 2001) was used to select the optimal substitution model (GTR, general time reversible model).One million cycles were performed, sampling a tree at every 100 generations.Trees that preceded the stabilization of the likelihood value (the burn-in) were excluded, and the remaining trees were used to construct a majority rule consensus in PAUP (version 4.0b10;Swofford, 2002).The percentages on this tree are the Bayesian posterior probabilities.
Maximum likelihood (ML) analyses were performed with PAUP*4.0b10(Swofford, 2002).The substitution model employed in the analyses was the same as for the Bayesian analyses.

RESULTS
Structure, size and composition of ITS.-Length of the entire ITS region, including ITS1, 5.8S rDNA and ITS2, varied among Melioideae accessions from 627 to 664 bp.ITS1 ranged in length from 233 to 273 bp, 5.8S rDNA from 156 to 172 bp, and ITS2 from 214 to 238 bp.The mean GC ratios of Melioideae taxa for the sequences of ITS1, 5.8S and ITS2 were 66%, 55% and 66%, respectively.The complete set of statistics for all datasets is summarized in Tables 1 and 2.
Phylogeny estimation based on ITS.-The aligned ITS matrix consisted of 794 characters (Table 1).For the entire ITS matrix, 499 (63 %) positions were variable and 403 (51 %) were potentially parsimony informative.The parsimony search produced three most parsimonious trees of 2,421 steps with consistency index (CI) = 0.38 and retention index (RI) = 0.54 for the entire ITS matrix (Fig. 1).Bayesian results derived from the entire ITS matrix are shown in Figure 2. The broad phylogenetic patterns are similar to the MP analysis: Aglaieae are monophyletic (51% bootstrap percentage, BP; 97% posterior probability, PP), Guareae are paraphyletic (Figs. 1, 2).Turraeeae are paraphyletic and appear in a clade with representatives of Trichilieae (53% BP; 94 PP; Fig. 2).Members of the latter also appear in other parts of the tree.Sandoriceae are monophyletic (100 BP; 100 PP; Figs. 1, 2), as are Melieae (87 BB; 100 PP; Figs. 1, 2).Maximum likelihood results reflect the same broad patterns (tree not shown).
Phylogeny estimation based on rbcL.-The aligned rbcL matrix consisted of 1,387 characters (Table 1).For the rbcL matrix, 186 (13%) positions were variable and 97 (7%) were potentially parsimony informative.The parsimony search produced 7,199 most parsimonious trees of 293 steps with CI = 0.59 and RI = 0.82 (Table 1).Bayesian tree topology derived from the rbcL matrix is identical to the MP results (Fig. 3), with one exception: Aglaia and Lansium formed a clade supported by 51% in the Bayesian majority rule consensus tree.Posterior probabilities are plotted on the MP tree (Fig. 3).Again, Aglaieae are monophyletic (74 PP; Fig. 3) and Guareae paraphyletic (Fig. 3).With the exception of Munronia, all other representatives of Turraeeae are members of a monophyletic group (98 BP; 71 PP; Fig. 3).All members of Turraeeae appear in a clade with representatives of Trichilieae (67 BP; 100 PP; Fig. 3).Again, members of the latter appear in other parts of the tree.Vavaea, Quivisianthe and Sandoricum are interdigitated with Trichilieae.Maximum likelihood results are almost identical to the MP and Bayesian topologies (Fig. 4).
Maximum likelihood results based on the combined ITS/rbcL matrix support Aglaieae as monophyletic; Guareae as paraphyletic; Turraeeae as paraphyletic, appearing in a clade with Trichilieae (Fig. 5).As for the single ITS and rbcL analyses, members of the latter also appear in other parts of the tree (Fig. 5).

DISCUSSION
Tribal affiliation within Melioideae.-At a glance, Pterorhachis Harms is distinct from all other Meliaceae on morphological grounds and resembles instead some members of Sapindaceae (Pennington & Styles, 1975).Placed in Meliaceae tribe Turraeeae by Harms (1940), a critical examination of morphology, wood and pollen showed that it definitely belongs in Meliaceae and is related to Trichilia L. (Pennington & Styles, 1975).This study confirms the position of Pter-orhachis in subfamily Melioideae and a close relationship to tribes Trichilieae and Turraeeae (Figs. 1, 2).Pennington & Styles (1975) demonstrated that secondary xylem provides good characters for subfamilial delimitation in Meliaceae, as well as for delimitation of tribal groups within Melioideae.They recognized two groups of tribes within the latter: (1) Sandoriceae, Turraeeae, Trichilieae (except Cipadessa) and Melieae, and (2) Aglaieae, Guareeae (except Turraeanthus) and Vavaeeae.This pattern of relationship among tribes is broadly confirmed by our study (Figs.1-5).First, Aglaieae plus Guareeae are always monophyletic.Second, Sandoriceae, Turraeeae and Trichilieae are closely interrelated (Figs.1-5).Vavaea is sister to the clade formed by Guareae and Aglaieae in the MP ITS tree (Fig. 1) and is sister to the clade uniting Aglaieae/Guareeae and most Turraeeae/ Trichilieae in the rbcL tree (Fig. 3).Tribe Aglaieae.-Aglaieae, currently including Aglaia Lour., Aphanamixis Blume, Lansium Correa, Reinwardtiodendron Koord.and Sphaerosacme Wall.ex Royle, owe their current circumscription to the work of Pennington & Styles (1975).These five genera are restricted to the Asian tropics and extend into the western Pacific.All except Sphaerosacme, of which there is one species, S. decandra, restricted to the Himalayas, are represented in Malesia.
The close morphological relationships of Aglaia, Lansium and Reinwardtiodendron collectively with Aphana-mixis and Sphaerosacme are reflected by our phylogenetic trees (Figs.1-5; for a detailed taxonomic history see Pennington & Styles, 1975;compare Mabberley &al., 1995 andMuellner &al., 2005).A detailed account on the evaluation of taxonomic concepts in the morphologically variable genus Aglaia based on DNA data and secondary metabolites was recently published by Muellner & al. (2005).
Our ITS study includes members of all three sections of Aglaia (sect.Aglaia, sect.Amoora, sect.Neoaglaia), all but one species of Aphanamixis, monospecific Sphaerosacme and all but one species each of Lansium and Reinwardtiodendron (we were unable to amplify these two species).Aglaia forms a monophyletic group with Lansium and Reinwardtiodendron (53 BP, Fig. 1; 99 PP, Fig. 2).Lansium and Reinwardtiodendron are monophyletic, Aglaia is paraphyletic; the three sections of Aglaia  -5).Tribe Guareeae.-Guareeae comprise nine genera, of which two, Cabralea A. Juss.and Ruagea Karst., are restricted to tropical America, two, Heckeldora Pierre and Turraeanthus Baill., to Africa, three, Anthocarapa Pierre, Chisocheton Blume and Dysoxylum Blume, to Indomalesia and western Pacific and one, Synoum A. Juss., to tropical Australia.
Our analysis of ITS includes representatives of all genera of Guareeae and therefore permits a detailed review of relationships within the tribe.As a whole, Guareeae are a paraphyletic group.Guarea and Ruagea (clade with 80 BP, 87 PP; Figs. 1, 2) are sister to Turraeanthus (Figs. 1, 2).The relationship to Heckeldora, Chisocheton and Dysoxylum lacks strong support; the same applies to Cabralea and Synoum.Anthocarapa nitidula and a sample collected as "Pseudocarapa" nitidula (regarded as synonym of the latter; Mabberley & al. 1995) form a clade supported by 85 BP (Fig. 1) and 100 PP (Fig. 2).Although regarded as a single species, the two samples exhibit a high number of autapomorphies (26 and 34, respectively; Fig. 1), which needs further investigation.
Tribe Vavaeeae.-Vavaeeae are a monogeneric tribe of four species distributed from Sumatra eastwards through Malesia to tropical Australia, Micronesia, Melanesia and Polynesia.Vavaea occupies a morphologically isolated position within Melioideae.It possesses most of the individual morphological, anatomical and palynological characters of the subfamily, but in a distinctive combination enabling it to be easily distinguished from all other genera.Vavaea has morphological similarities to various tribes and genera: Turraeeae (leaves), Trichilieae (fruit, seed, embryo), Sandoriceae (wood anatomy, pollen), Aglaia (pollen).The ambiguous morphological relationships are reflected in our phylogenetic trees: Vavaea occupies an isolated position sister to Aglaieae/Guareeae in the MP ITS tree (Fig. 1) and is sister to the clade uniting Aglaieae/Guareeae and most Turraeeae/Trichilieae in the rbcL and combined trees (Figs.3-5 The taxonomic history of Trichilieae is complex and closely related to that of Turraeeae (reviewed in Pennington & Styles, 1975).For morphological reasons, Pennington & Styles (1975) concluded that Pterorhachis and Cipadessa did not belong in Turraeeae, in which they were placed by Harms (1940).A critical examination of morphology, wood and pollen showed that Pterorhachis is closely related to Trichilia, from which it differs principally in having more numerous filament appendages and from most species of Trichilia in its spheroidal pollen grains (Pennington & Styles, 1975).Cipadessa is similar in these same characters to Trichilieae as well, with an hypothesized relationship to Ekebergia, and was therefore, like Pterorhachis, included in this tribe (Pennington & Styles, 1975).Pseudobersama is thought to be closely related to Trichilia (Pennington & Styles, 1975).
Our study of ITS reveals Pterorhachis as the closest relative of Nymania, a member of Turraeeae (66 BP and 84 PP;Figs. 1,2).A close relationship of Cipadessa to Ekebergia is confirmed by ITS (100 BP, 100 PP; Figs. 1, 2), though not by rbcL.In the analysis of rbcL, Pseudobersama forms a clade with Trichilia, its closest morphological relative.As for the remaining genera of Trichilieae, relationships based on ITS and rbcL are incongruent.Based on our results, it is impossible to keep Trichilieae separated from Turraeeae, .To reach a robust and well-resolved phylogenetic appreciation of Trichilieae, sampling of additional taxa on species level and the collection of much more data will be necessary.
Tribe Turraeeae.-Turraeeae comprise six or seven genera: Calodecaryia Leroy, Humbertioturraea Leroy, Munronia Wight, Nymania S.O.Lindb., Turraea L. including Naregamia Wight & Arn., and perhaps an undescribed genus ("Turraea breviflora" Ridley), all restricted to the Old World tropics.The largest is Turraea, which is the most widespread; the rest are small genera, one or two restricted to Indomalesia, two to Madagascar, one to southern Africa and one found in both India and southern Africa (Mabberley & al., 1995).Pennington & Styles (1975) used Turraeeae in the introduction of their generic monograph to illustrate that most tribes in Meliaceae can only be diagnosed by using a combination of several differential characters (as defined by White, 1962).They stated that members of Turraeeae cannot be distinguished from other Meliaceae on the basis of a single diagnostic character and that most characterstates typical of the tribe have at least a few exceptions and also occur at least occasionally in other tribes, but always in markedly different combinations.The overall pattern, however, was such that all members of Turraeeae possess many more of the tribal character-states than any excluded species.Thus, Pennington & Styles (1975) claimed Turraeeae, as well as all other tribes in the monograph, to be objectively circumscribed, being based on gaps in the pattern of variation.
Our investigation includes representatives of all genera of Turraeeae and therefore allows a detailed review of relationships within the tribe.In our analysis of ITS, Nymania is sister (as part of a clade with Pterorhachis) to the "core group" of Trichilieae, formed by Turraea, Humbertioturraea, Calodecaryia and Naregamia (Figs. 1, 2).In the rbcL tree, Nymania is again sister to this core group .Naregamia is sister to the clade formed by Turraea, Humbertioturraea and Calodecaryia in both the single ITS and rbcL, and in the combined trees (Figs.1-5).The separation of Naregamia from these three genera is well supported in ITS and rbcL trees (99 BP, Fig. 1; 100 PP, Fig. 2; 71 BP, 94 PP, Fig. 3), emphasizing that Naregamia is genetically distinguishable from Turraea.Naregamia was reduced to synonymy with Turraea by Cheek (1996; for a detailed discussion of characters and the status of Naregamia and Turraea see Cheek, 1990).Cheek (1990) stated that, as far as seed structure was concerned, Naregamia could not be separated from Turraea.Previously, Pennington & Styles (1975) had claimed Naregamia to be easily distinguished from Turraea by combined characteristics of leaves, the staminal tube and seed structure.Our data agree with these earlier findings of Pennington & Styles (1975); we propose to keep Naregamia separate from Turraea.The position of Munronia remains ambiguous, as expected by its morphological intermediacy between typical Turraeeae and the remainder of Melioideae (Figs. 1-5).Unfortunately, we were unable to amplify samples of Turraea breviflora collected from herbarium specimens located in Kepong (KEP), Malay Peninsula, and in Kew (K), U.K., due to the old age of specimens and resulting poor quality of DNA extracts (high degradation).The species, according to Mabberley & al. (1995) perhaps an undescribed genus, is known only from a few localities in the Malay Peninsula and Singapore.The fruit has never been observed; recent collections are lacking.
As for Trichilieae, an increase of sampling on species level and the collection of additional DNA data will be necessary to make final decisions about a new circumscription of Trichilieae, especially the inclusion/exclusion of Munronia in the tribe.
Tribe Sandoriceae.-Sandoriceae are monogeneric with five species, all but one (S.koetjape) restricted to western Malesia (Mabberley & al., 1995).Pennington & Styles (1975) claimed Sandoricum to be a morphologically distinct genus, without a close relationship to Dysoxylum as proposed by Harms (1940) or to Guareeae.Sandoricum is at once identifiable by trifoliate leaves, the ribbed staminal tube, characteristic style-head with divided stigma and indehiscent drupaceous fruit, presumably the reason Pennington & Styles (1975) placed the genus in its own tribe.
Our data confirm that Sandoricum has no close relationship to either Dysoxylum or .In our analysis of ITS, the two species of Sandoricum form a strongly supported clade (100 BP, Fig. 1; 100 PP, Fig. 2) and are characterized by a relatively high number of autapomorphies (29, Fig. 1).In the rbcL trees, Sandoricum is sister to Ekebergia and Quivisianthe (Figs. 3-4) and again characterized by a relatively high number of autapomorphies (11, Fig. 3).
Tribe Melieae.-Melieae comprise two genera, Melia L. (one to possibly three species) and Azadirachta A. Juss.(two species), in the wild state restricted to the Old World Tropics.Melia and Azadirachta are similar morphologically (Pennington & Styles, 1975).Both genera share a number of anatomical characters not recorded elsewhere in Meliaceae (e.g., clusters of minute vessels with spiral wall thickening).Our single and combined analyses of ITS and rbcL confirm monophyly of Melieae (Figs. 1-5).In the ITS MP and Bayesian analyses, Melieae are sister to all other Melioideae (Figs. 1, 2).The same is true for the combined analysis (Fig. 5).In the analyses of rbcL, Melieae are sister to Owenia (98 BP, 100 PP, Fig. 3; Fig. 4), and this clade is sister to all other Melioideae.
Quivisianthe (Quivisianthoideae). -Although treated in a monogeneric subfamily by Pennington & Styles (1975), the authors mentioned in their generic monograph that the genus is similar in its floral structure to some genera in Trichilieae and that the complete staminal tube without appendages and with the anthers or antherodes inserted on the margin is similar to that of Ekebergia.Our ITS and rbcL data confirm the position of Quivisianthe in Melioideae (Figs. 1-5).In the rbcL tree, Quivisianthe exhibits a close relationship to Ekebergia (clade with 74 BP, 100 PP; Fig. 3), whereas for ITS it appears as sister to Walsura (Figs. 1, 2).In the combined Bayesian (tree not shown) and ML analyses (Fig. 5), Quivisianthe occupies an isolated position, in the MP analysis the genus appears as sister to Walsura (tree not shown).
Concluding remarks.-DNA data of Melioideae and related genera contribute to a better understanding of the intricate systematic relationships of this group of trees that constitute an important component of moist tropical forests world-wide.This study is the first to assess circumscription of Melioideae and the component tribes in detail with data independent of morphology.Maximum parsimony, maximum likelihood and Bayesian analyses of nuclear ITS, compared with analyses based on plastid rbcL, confirm monophyly for Aglaieae, Sandoriceae and Melieae, an isolated position for Vavaeeae, the position of Pterorhachis and Quivisianthe in Melioideae, and close relations between Turraeeae and Trichilieae.Trichilieae are the most complex clade.Anthocarapa and Pseudocarapa, regarded as synonym of the latter, form a clade, but exhibit each a high number of autapomorphies, which needs further investigation.We propose to keep Naregamia separate from Turraea because the two are not exclusively related.These taxonomic decisions are based on DNA data as well as morphological variation.

Fig. 1 .
Fig. 1.One of the three most parsimonious trees obtained from the maximum parsimony analysis of the ITS nrDNA dataset of 55 Meliaceae accessions.Tribal names and numbers after Pennington & Styles (1975).Numbers above branches are estimated branch lengths (DELTRAN optimization), numbers below branches are bootstrap percentages (5,000 replicates); in italics.The arrow indicates a group not present in the strict consensus tree.

Fig. 3 .
Fig. 3.One of the 7,199 most parsimonious trees obtained from the maximum parsimony analysis of the plastid rbcL dataset of 37 Meliaceae accessions.Tribes after Pennington & Styles (1975).Numbers above branches are estimated branch lengths (DELTRAN optimization), and bootstrap percentages (1,000 replicates; in italics).Numbers below branches are Bayesian posterior probabilities (10,000 total trees, burn-in of 1,100 trees).Arrows indicate groups not present in the strict consensus tree.

Fig. 4 .
Fig. 4. Tree obtained from the maximum likelihood analysis of the plastid rbcL dataset of 37 Meliaceae accessions.Tribal numbers in brackets after species names.