Molecular Phylogenetics and Evolution 60 (2011) 428–444 Contents lists available at ScienceDirect Molecular Phylogenetics and Evolution journal homepage: www.elsevier.com/locate/ympev A non-coding plastid DNA phylogeny of Asian Begonia (Begoniaceae): Evidence for morphological homoplasy and sectional polyphyly D.C. Thomas a,b,⇑, M. Hughes a, T. Phutthai c, S. Rajbhandary d, R. Rubite e, W.H. Ardi f, J.E. Richardson a a Royal Botanic Garden Edinburgh, 20A Inverleith Row, Edinburgh EH3 5LR, UK School of Biological Sciences, University of Hong Kong, Pok Fu Lam Road, Hong Kong, PR China c Herbarium, Princess Maha Chakri Sirindhorn Natural History Museum & Centre for Biodiversity of Peninsular Thailand (CBiPT), Department of Biology, Faculty of Science, Prince of Songkla University, Hat Yai, Songkhla 90112, Thailand d Central Department of Botany, Institute of Science and Technology, Tribhuvan University, Kirtipur, Kathmandu, Nepal e Department of Biology, College of Arts and Sciences, University of the Philippines Manila, Padre Faura, Manila, Philippines f Bogor Botanic Gardens, Jl. Ir. H. Juanda No. 13, Bogor 16003, Indonesia b a r t i c l e i n f o Article history: Received 10 January 2011 Revised 4 April 2011 Accepted 6 May 2011 Available online 17 May 2011 Keywords: Asia Begoniaceae Begonia Phylogeny Character evolution a b s t r a c t Maximum likelihood and Bayesian analyses of non-coding plastid DNA sequence data based on a broad sampling of all major Asian Begonia sections (ndhA intron, ndhF-rpl32 spacer, rpl32-trnL spacer, 3977 aligned characters, 84 species) were used to reconstruct the phylogeny of Asian Begonia and to test the monophyly of major Asian Begonia sections. Ovary and fruit characters which are crucial in current sectional circumscriptions were mapped on the phylogeny to assess their utility in infrageneric classifications. The results indicate that the strong systematic emphasis placed on single, homoplasious characters such as undivided placenta lamellae (section Reichenheimia) and fleshy pericarps (section Sphenanthera), and the recognition of sections primarily based on a suite of plesiomorphic characters including three-locular ovaries with axillary, bilamellate placentae and dry, dehiscent pericarps (section Diploclinium), has resulted in the circumscription of several polyphyletic sections. Moreover, sections Platycentrum and Petermannia were recovered as paraphyletic. Because of the homoplasy of systematically important characters, current classifications have a certain diagnostic, but only poor predictive value. The presented phylogeny provides for the first time a reasonably resolved and supported phylogenetic framework for Asian Begonia which has the power to inform future taxonomic, biogeographic and evolutionary studies. Ó 2011 Elsevier Inc. All rights reserved. 1. Introduction The pantropically distributed genus Begonia L. is with more than 1500 species among the ten largest genera of vascular plants (Frodin, 2004; Hughes, 2008). One hotspot of species diversity lies in Southeast Asia, whose Begonia flora comprises more than 550 species (Hughes, 2008; Hughes & Pullan, 2007). This exceptional species diversity and the wide distribution of the genus in tropical rainforests offers the opportunity to address biogeographical questions and to investigate the processes which underlie modern patterns of biodiversity, but also poses major taxonomic challenges. Although Asian Begonia species are morphologically diverse, especially with regard to growth habit, perennation organ, leaf shape, ⇑ Corresponding author at: School of Biological Sciences, University of Hong Kong, Pok Fu Lam Road, Hong Kong, PR China. E-mail addresses: dthomas@hku.hk (D.C. Thomas), m.hughes@rgbe.ac.uk (M. Hughes), tputthai@yahoo.com (T. Phutthai), imogine3@gmail.com (S. Rajbhandary), rosariorubite@yahoo.com (R. Rubite), wisn001@lipi.go.id (W.H. Ardi), j.richardson@rbge.ac.uk (J.E. Richardson). 1055-7903/$ - see front matter Ó 2011 Elsevier Inc. All rights reserved. doi:10.1016/j.ympev.2011.05.006 inflorescence architecture and fruit type, few apomorphies characterizing infrageneric taxa have been identified and delimitation of Asian sections is often highly problematic (Doorenbos et al., 1998). A robust phylogenetic framework of Asian Begonia is currently lacking. A reliable infrageneric classification and the subdivision of the otherwise unwieldy number of species in mega-diverse genera such as Begonia are important to facilitate scientific communication and inform taxonomic monographs as well as biogeographical and evolutionary studies. The revision of section circumscriptions by Doorenbos et al. (1998) provided a crucial foundation for the infrageneric classification of Begonia. In this revision 18 Asian sections were recognized (Doorenbos et al., 1998), and another four Asian sections were subsequently proposed (Forrest and Hollingsworth, 2003; Gu, 2007; Ku, 1999; Shui et al., 2002). These 22 sections are highly unbalanced with regard to species numbers with the largest eight, sections Petermannia (Klotzsch) A.DC., Platycentrum (Klotzsch) A.DC., Diploclinium (Lindl.) A.DC., Reichenheimia (Klotzsch) A.DC., Coelocentrum Irmsch., Parvibegonia A.DC., Sphenanthera (Hassk.) Warb. and Symbegonia (Warb.) L.L. Forrest D.C. Thomas et al. / Molecular Phylogenetics and Evolution 60 (2011) 428–444 & P.M. Hollingsworth, accounting for more than 95% of the ca. 750 species in Asia. Morphological and anatomical fruit and ovary characters have traditionally played an essential role as diagnostic characters and for infrageneric taxon delimitation in Begonia (Irmscher, 1925; Warburg, 1894), and modern classifications still strongly rely on these characters (Doorenbos et al., 1998). One example is the recent Begonia treatise in the Flora of China, in which a new monotypic Begonia section was proposed and sections present in China were re-circumscribed solely based on character combinations of carpel number, ovary locule number, placentation type, and the type of placenta divisions (Gu, 2007). This allowed clear sectional placement for all of the 173 described Chinese species. To a lesser extent, vegetative characters such as the presence and type of perennating organs like tubers and rhizomes, and floral and inflorescence characters such as the numbers of tepals, the number of style branches and the distribution of male and female flowers in the inflorescences have contributed to the delimitation of infrageneric taxa in Begonia (Doorenbos et al., 1998). However, recent molecular phylogenetic studies challenge the strong emphasis on few easily observable morphological and anatomical characters in infrageneric Begonia classifications, and several crucial characters used in sectional circumscription were identified as highly homoplasious in phylogenies based on lowdensity, world-wide taxon sampling of the genus (Forrest and Hollingsworth, 2003; Forrest et al., 2005). The results of a study by Tebbitt et al. (2006), who analysed sequence data of the nuclear ribosomal DNA (nrDNA) internal transcribed spacer region (ITS) of 46 Asian Begonia species, indicate that fruit syndromes associated with animal dispersal evolved multiple times within Asian Begonia, and the sections Sphenanthera, Platycentrum and Leprosae (T.C. Ku) Y.M. Shui were identified as polyphyletic. Forrest and Hollingsworth (2003) showed, based on nrDNA ITS and 26S sequence data, that the New Guinean genus Symbegonia Warb., which was traditionally separated from Begonia based on floral characters (a syntepalous perianth and a characteristic monadelphous androecium), is nested within Begonia section Petermannia. They proposed to recognize Symbegonia at sectional level rendering the large section Petermannia paraphyletic, but retaining a morphologically easily recognizable taxon (Forrest and Hollingsworth, 2003). Moreover, recent phylogenetic analyses of ITS data of between 83 and 125 accessions of Asian taxa in the framework of unpublished dissertations on the systematics and phylogenetics of Himalayan Begonia (Rajbhandary, 2010), Philippine Begonia (Rubite, 2010), and Southeast Asian Begonia (Thomas, 2010) indicate that (i) the monotypic sections Baryandra A.DC. is nested within a clade of Philippine species assigned to the polyphyletic section Diploclinium (Rajbhandary, 2010; Rubite, 2010; Thomas, 2010), (ii) section Putzeysia (Klotzsch) A.DC. is closely related to Himalayan species assigned to the polyphyletic section Diploclinium (Rajbhandary, 2010; Thomas, 2010), and (iii) section Monopteron (A.DC.) Warb. is nested within a clade of species placed in sections Platycentrum and Sphenanthera (Rajbhandary, 2010; Rubite, 2010; Thomas, 2010). Thus, the monophyly of some sections was tested and their phylogenetic positions were clarified in molecular phylogenetic studies including a wider sampling of Asian Begonia. However, phylogenetic analyses of nrDNA sequence data have failed to resolve deeper relationships within Asian Begonia. Despite high percentages of potentially parsimony informative characters, ITS phylogenies of Asian Begonia show only poorly resolved or poorly supported backbones. Because of limited taxon sampling and the lack of resolution and support of phylogenies presented in previous studies, intersectional relationships within Asian Begonia are still only very fragmentarily understood. The extent to which parallel evolution of crucial characters has hampered the recognition of evolutionary lineages in Asian Begonia requires further investigation. The aims of this study, therefore, are to reconstruct the phy- 429 logeny of Asian Begonia; to determine whether major Asian sections (sections Coelocentrum, Diploclinium, Parvibegonia, Petermannia, Platycentrum, Reichenheimia, Sphenanthera, Symbegonia) are monophyletic; and to map morphological characters which are crucial in current sectional delimitations on the phylogeny to determine their degree of homoplasy and to assess their suitability in infrageneric classifications. For this purpose we used data of three non-coding chloroplast DNA (cpDNA) regions, the ndhA intron and the ndhF-rpl32 and rpl32-trnL spacers, which were shown to exhibit high levels of variability and to have potentially high phylogenetic utility at the interspecific level in comparison to other cpDNA coding and non-coding regions (Shaw et al., 2007). 2. Materials and methods 2.1. Taxon sampling The dataset comprized 67 taxa sampled broadly from all major Asian Begonia sections (sections Coelocentrum, Diploclinium, Parvibegonia, Petermannia, Platycentrum, Reichenheimia, Sphenanthera, Symbegonia). Moreover, samples of the small Asian sections Alicida C.B. Clarke, Bracteibegonia A.DC., Haagea (Klotzsch) A.DC. and Ridleyella Irmsch., and one species currently unplaced to section (Begonia malabarica Lam.), as well as samples of the only two Begonia species known from the Socotra archipelago (Yemen), Begonia socotrana Hook. f. and Begonia samhaensis M. Hughes & A.G. Mill (section Peltaugustia [Warb.] Barkley), were included. Six African species and three American species were chosen as outgroup based on molecular phylogenetic studies by Goodall-Copestake et al. (2010) and Plana et al. (2004). The results of these studies indicate that Begonia initially diversified in Africa, that a monophyletic Socotran–Asian Begonia lineage is derived from African Begonia, and that South African and American taxa form the sister clade to the Socotran–Asian clade (Goodall-Copestake et al., 2010; Plana et al., 2004). All sequences of the cpDNA regions were newly generated for this study. Voucher information and Genbank accession numbers are listed in Appendix A. 2.2. DNA extraction, amplification and sequencing Total genomic data was extracted from living material or silica gel dried material using the DNeasy Plant Mini Kit (Qiagen) according to the manufacturer’s protocols. For amplification of the cpDNA regions each 25 ll PCR contained 15.25 ll of ddH2O, 2.5 ll of 10 reaction buffer, 1.25 ll of 25 mM MgCl2, 2.5 ll dNTPs (2 mM), 0.75 ll of each forward and reverse primer (10 lM), 0.8 ll bovine serum albumin (BSA, 0.4%), 0.2 ll of Biotaq DNA polymerase (Bioline) and 1 ll of DNA template. Table 1 shows all primers used in this study. The amplification of the ndhF-rpl32 spacer of some samples in Begonia section Reichenheimia, Coelocentrum, and Petermannia failed with the universal primers designed by Shaw et al. (2007) and required the design of specific internal primers, ndhFBeg-F and rpl32Beg-R (Table 1). The PCR temperature profile used was the same as in Shaw et al. (2007): template denaturation at 80 °C for 5 min followed by 30 cycles of denaturation at 95 °C for 1 min, primer annealing at 50 °C for 1 min, followed by a ramp of 0.3 °C/s to 65 °C, and primer extension at 65 °C for 4 min; followed by a final extension step at 65 °C for 5 min. Poly A/T homonucleotide strands composed of eight or more nucleotides, which were present in the sequences of most accessions, can cause PCR artefacts by slippedstrand mis-pairing (Shinde et al., 2003). To mitigate this problem an alternative amplification protocol was applied for problematic samples using Phusion polymerase (Finnzymes). Phusion polymerase was shown to reduce slipped-strand mis-pairing, at least for homonucleotide strands of up to 15 bp length, possibly because 430 D.C. Thomas et al. / Molecular Phylogenetics and Evolution 60 (2011) 428–444 Table 1 Primers used in this study. DNA region Primer Primer sequence (50 –30 ) Source ndhA intron ndhAx1 GCYCAATCWATTAGTTATGAAATACC ndhAx2 GGTTGACGCCAMARATTCCA Shaw et al. (2007) Shaw et al. (2007) rpL32-R CCAATATCCCTTYYTTTTCCAA ndhF GAAAGGTATKATCCAYGMATATT ndhFBegF trnLBeg-R TGGATGTGAAAGACATATTTTGCT TTTGAAAAGGGTCAGTTAATAACAA This study trnL CTGCTTCCTAAGAGCAGCGT rpL32-F CAGTTCCAAAAAAACGTACTTC Shaw et al. (2007) Shaw et al. (2007) ndhF-rpl32 rpl32-trnL of increased contact surface between enzyme and DNA in comparison to Taq polymerases (Fazekas et al., 2010). Each 25 ll PCR contained 13.0–14.0 ll of ddH2O, 5.0 ll of 5  Phusion HF buffer, 2.5 ll dNTPs (2 mM), 1.25 ll of each forward and reverse primer (10 lM), 0.25 ll of Phusion DNA Polymerase, 1.0–2.0 ll of DNA template. The PCR temperature profile included template denaturation at 98 °C for 30 s followed by 32 cycles of denaturation at 98 °C for 10 s, primer annealing at 62 °C (ndhF-rpl32) or 63 °C (ndhA intron, rpl32-trnL) for 30 s, primer extension at 72 °C for 30 s; followed by a final extension step at 72 °C for 10 min. Amplification products were visualized under UV light after electrophoretic separation on a 1% agarose TBE gel stained with SYBR Safe gel stain (Invitrogen). To remove superfluous dNTPs and primers, PCR products were subsequently purified using ExoSAP-IT (Affymetrix) according to the manufacturer’s protocols. Sequencing PCRs used the same primers as for amplification, and were quarter reactions using the BigDye Terminator Cycle Sequencing Kit (Applied Biosystems). The protocol included an initial denaturation at 95 °C for 1 min, followed by 25 cycles of denaturation at 96 °C for 10 s, primer annealing at 60 °C for 5 s, and primer extension at 60 °C for 4 min. Sequencing PCR products were purified and sequenced at the GenePool facilities at the University of Edinburgh (GenePool) using an AB 3730 DNA Analyser (Applied Biosystems). 2.3. Alignment Sequences were assembled and edited using Geneious v5.1.4 (Drummond et al., 2010). The sequences were pre-aligned using the multiple sequence alignment software MUSCLE (Edgar, 2004) implemented in Geneious using default settings, and subsequently manually checked and optimized in Geneious. An autapomorphic inversion of 48 bp was identified in the ndhA intron sequence of the Neotropical Begonia radicans Vell. Inversions of 355 bp, or due to deletion of 309 bp, flanked on both sides by A/T homonucleotide strands, were identified in the ndhF-rpl32 spacer region of all Philippine samples of Begonia section Diploclinium. Sequences of the rpl32-trnL spacer of Begonia pendula Ridl. showed an inversion of 37 bp, which is flanked by complementary regions indicating hairpin secondary structures. A fourth inversion, flanked by homonucleotide repeats of four G/Cs on each side, was identified in the rpl32-trnL spacer. This 11 bp inversion was present in three Bornean species in Begonia section Petermannia (Begonia amphioxus Sands, Begonia burbidgei Stapf, B. pendula) as well as in Begonia masoniana Irmsch. ex Ziesenh. (section Coelocentrum), and Begonia roxburghii A.DC. (section Sphenanthera). These species are only distantly related in phylogenies resulting from the analysis of the Shaw et al. (2007) Shaw et al. (2007) This study cpDNA nucleotide data, indicating a homoplasious origin of this inversion in the three sections. All inversions were reverse-complemented, thereby retaining substitution information in the fragments in the matrix (Borsch and Quandt, 2009; Graham et al., 2000; Löhne and Borsch, 2005). Nineteen mutational hotspots, most of which were length differences of homonucleotide strands, together ca. 5.1% of the aligned positions, were excluded from the final matrix because of uncertain homology (Borsch and Quandt, 2009; Kelchner, 2000). All alignments are available from the corresponding author upon request. 2.4. Phylogenetic analyses Bayesian phylogenetic reconstructions were performed in MrBayes v3.1.2 (Huelsenbeck and Ronquist, 2001; Ronquist and Huelsenbeck, 2003) using the Oslo Bioportal at the University of Oslo (www.bioportal.uio.no). Three partitions based on spacer and intron identity (ndhA intron, ndhF-rpl32 spacer, rpl32-trnL spacer) were defined a priori, or alternatively, the three cpDNA regions were concatenated and analysed without partitioning. Models of sequence evolution for the concatenated matrix and each nucleotide sequence partition were determined using jModelTest (Posada, 2008). Maximum likelihood topologies were used to estimate the optimal evolutionary model comparing 88 distinct models (11 substitution schemes, with equal or unequal base frequencies, a proportion of invariable sites, and rate variation among sites). Log-likelihoods of different models under maximum likelihood tree topologies were compared using the Akaike Information Criterion (AIC) and its corrected version for small samples (AICc) as model selection criteria (Posada and Buckley, 2004). The AICc converges towards the AIC, when larger sampling sizes are used, and should therefore always be used regardless of the sample size (Burnham and Anderson, 2004). Selected models which are not implemented in MrBayes were substituted by the closest overparameterized implemented model (Huelsenbeck and Rannala, 2004). Overall performance of analyses of unpartitioned, concatenated nucleotide datasets and partitioned nucleotide datasets were assessed with comparison of the mean lnL of all trees sampled from the posterior distribution at stationarity for each strategy, and with Bayes Factor comparison implemented in Tracer v1.5 (Rambaut and Drummond, 2009), which is based on smoothed estimates of marginal likelihoods (Newton and Raftery, 1994; Suchard et al., 2001). The criterion of 2ln Bayes Factor of P10 was used as a benchmark indicating very strong evidence in favour of one strategy over another (Kass and Raftery, 1995). Four independent Metropolis-coupled MCMC analyses were run. Each search D.C. Thomas et al. / Molecular Phylogenetics and Evolution 60 (2011) 428–444 used three incrementally heated and one cold Markov chain, a temperature parameter setting of 0.8, and was run for 1  107 generations and sampled every 1000 generations. The parameters for character state frequencies, the substitution rates of the nucleotide substitution models, and the rate variation among sites were unlinked across partitions. Convergence was assessed by using the standard deviation of split frequencies as convergence index with values <0.005 interpreted as indicating good convergence. Tracer v1.5 (Rambaut and Drummond, 2009) was used to determine whether the MCMC parameter samples were drawn from a stationary, unimodal distribution, and whether adequate effective sample sizes for each parameter (ESS > 200) were reached. Topological convergence and convergence of posterior probabilities of splits from different runs were visually checked using the Compare and Cumulative functions of AWTY (Nylander et al., 2008). The initial 25% of samples of each Metropolis-coupled MCMC run were discarded as burnin, and the post burnin samples were summarized as 50% majority rule consensus phylograms with nodal support expressed as posterior probabilities. Maximum likelihood (ML) analyses were performed using RAxML-HPC2 v7.2.7 (Stamatakis, 2006) using the CIPRES Portals (www.phylo.org/sub_sections/portal). The dataset was divided into three partitions based on spacer and intron identity. One thousand inferences were run from distinct random stepwise addition sequence maximum parsimony starting trees under the general time reversible nucleotide substitution model (GTR, Tavaré, 1986) with among-site rate variation modelled with a gamma distribution. Subsequently, 1000 non-parametric bootstraps were performed under the partition data mode. 2.5. Ancestral character state reconstructions Ancestral character states of five characters that have traditionally been used to define infrageneric taxa in Begonia including the presence and type of perennation organs, fruit types, locule numbers, placentation types, and placenta divisions were reconstructed. Fruit and ovary characters are illustrated in Fig. 1. Definitions and discussions of character states can be found in Phutthai et al. (2009) and Thomas (2010) for perennation organs and stem metamorphoses, and in Kiew (2005), Tebbitt et al. (2006), and Thomas (2010) for ovary and fruit characters. The character matrix can be found in Appendix B. (A) Stem metamorphoses: Four character states were differentiated for the ancestral character reconstructions: 1. No specialization with regard to storage and perennation function; 2. Rhizome; 3. Tuber; 4. Thickened stem base. (B) Fruit types: Three character states are differentiated for the ancestral character reconstructions: 1. Dry capsule (Fig. 1A–C); 2. Rain-ballist capsule (Fig. 1D); 3. Baccate fruit (Fig. 1E and F). (C) Ovary locule number: Four character states are differentiated for the ancestral character reconstructions based on locule numbers (one to four locules) (Fig. 1G–L). (D) Placenta configuration: Three character states are differentiated for the ancestral character reconstructions: 1. Axillary (Fig. 1G, H, J, and L); 2. Septal; 3. Parietal (Fig. 1I). (E) Placenta division: Three character states are differentiated for the ancestral character reconstructions: 1. Unilamellate (Fig. 1G); 2. Bilamellate (Fig. 1H–L); 3. More than two primary placenta lamellae. Ancestral character states were reconstructed using parsimony and likelihood methods implemented in Mesquite v2.7.4 (Maddison and Maddison, 2010). To generate the input trees a Bayesian analysis using MrBayes was run using three data partitions and the same specifications and taxa as outlined above, except for the exclusion of three Bornean taxa, B. amphioxus, B. burbidgei, and B. pendula. These taxa were excluded because of putative, ancient hybridization involving the ancestor of this lineage (see Sec- 431 tion 4.6). Ancestral character reconstructions were mapped on the majority rule consensus tree obtained from the Bayesian analysis. Parsimony reconstructions search for the ancestral character states which minimize the number of required steps of character change given a tree and observed extant character distributions at the terminals. Character-state changes were modelled as unordered for all characters. Likelihood reconstructions optimize the character states at each node which maximize the probability of arriving at the observed extant character states of the terminals, given a model of evolution. To account for phylogenetic uncertainty the ‘‘Trace over trees’’ option was selected, and 10,000 trees from the stabilized part of the MCMCMC analysis were included as input trees. The Mk1 model (Markov k-state 1 parameter model) (Lewis, 2001) was selected. Under this model any particular change is equally probable, and the rate of change is the only parameter. The ‘‘trace over trees’’ option summarizes for every node the proportion of the average likelihood received by each character state as the ancestral character of a given clade. 3. Results 3.1. Phylogeny The concatenated alignment of the 84-taxon dataset consisted of 3977 aligned positions. Descriptive statistics for the concatenated dataset and its nucleotide partitions based on spacer and intron identity including the number of aligned positions, the length of the analysed fragments, the number and percentage of variable sites, the number and percentage of parsimony informative sites, and the number and percentage of excluded sites are given in Table 2. Of the three non-coding chloroplast regions used for this study, the ndhA intron was the least variable region, exhibiting distinctly lower percentages of variable and potentially parsimony informative sites than the ndhF-rpl32 and rpl32-trnL spacers. Nucleotide model selection under the AIC and its corrected version for small sample sizes (AICc) did not differ for the partitions, and the transversional model including rate variation among sites (TVM + G, Posada 2003), was selected for all partitions. As the TVM + G is not implemented in MrBayes, the closest overparameterized implemented model, GTR + G, was selected for the analyses (Huelsenbeck and Rannala, 2004). Partitioning improved mean lnL values in the Bayesian analyses (mean lnLunpartitioned = 20,514; mean lnLpartitioned = 20,413) and the analyses using partitioning provided distinctly better explanations of the data than analyses of unpartitioned datasets according to Bayes Factor comparison (ln Bayes Factor = 100). The subsequent presentation of the results of the Bayesian analyses will be limited to the trees derived from the partitioned analyses. For the results of the Bayesian analyses a 95% posterior probability lower threshold was considered to indicate well supported relationships. For the results of the ML analyses 70–79% bootstrap support were considered to indicate moderate support, 80–90% bootstrap support to indicate well supported relationships, and 90–100% bootstrap support to indicate strongly supported relationships. Sectional placement of taxa shown in Fig. 2 primarily follows Doorenbos et al. (1998) and more recent species descriptions and sectional placement corrections (Kiew, 2001, 2005; Hughes, 2008; Tebbitt and Dickson, 2000). Accessions unidentified at species level are placed to section based on morphological sectional circumscriptions in Doorenbos et al. (1998). Fig. 2 shows the majority rule consensus trees of the Bayesian analyses, with posterior clade probabilities (PP) and bootstrap support values (BS) of corresponding clades of the best-scoring tree 432 D.C. Thomas et al. / Molecular Phylogenetics and Evolution 60 (2011) 428–444 Fig. 1. Overview of fruit and ovary morphology and anatomy in Asian Begonia. A: Begonia dipetala: capsule, dry pericarp, equal wings. Scale bar = 16 mm. B: Begonia varipeltata: capsule, dry pericarp, equal wings. Scale bar = 12 mm. C: Begonia masoniana: capsule, dry pericarp, unequal wings. Scale bar = 10 mm. D: Begonia pavonina: rainballist capsule, coriaceous pericarp, unequal wings. Scale bar = 8 mm. E: Begonia aptera: berry, fleshy pericarp, wings reduced. Scale bar = 10 mm. F: Begonia obovoidea: berry, fleshy pericarp, wings absent. Scale bar = 15 mm. G: Begonia dipetala: ovary cross-section, three-locular ovary with axillary, undivided placentae. Scale bar = 3 mm. H: Begonia varipeltata: ovary cross-section, three-locular ovary with axillary, bilamellate placentae. Scale bar = 3 mm. I: Begonia masoniana: ovary cross-section, unilocular ovary with parietal, bilamellate placentae. Scale bar = 3 mm. J: Begonia pavonina: ovary cross-section, two-locular ovary with axillary, bilamellate placentae. Scale bar = 3 mm. K: Begonia aptera: ovary cross-section, three-locular ovary with axillary, bilamellate placentae. Scale bar = 3 mm. L: Begonia obovoidea: ovary cross-section, four-locular ovary with axillary, bilamellate placentae. Scale bar = 3 mm. Table 2 Descriptive statistics of analysed non-coding cpDNA sequence data. Partition Aligned sites (#) Analysed fragment length (bp) Variable sites (# (%)) Parsimony informative sites (# (%)) Excluded alignment sites (# (%)) ndhA intron ndhF-rpl32 rpl32-trnL Combined 1424 1167 1386 3977 1078–1201 694–969 569–1113 2467–3246 349 (24.5) 429 (36.8) 467 (33.7) 1245 (31.3) 161 228 234 623 37 (2.5) 72 (5.8) 106 (7.1) 215 (5.1) (11.3) (19.5) (16.9) (15.7) 433 D.C. Thomas et al. / Molecular Phylogenetics and Evolution 60 (2011) 428–444 B. oxyloba MEZ B. goudotii QUA B. poculifera SQU 1/100 B. polygonoides TET B. sutherlandii AUG 1/100 B. dregei AUG 1/99 B. nelumbiifolia GIR B. radicans SOL .95/54 B. boliviensis BAR B. dipetala HAA 30 .96/77 B. socotrana PEL 28 1/100 B. samhaensis PEL .89/56 B. floccifera REI 28, 30, 32 1/100 B. malabarica IGN 60 B. smithiae PLA B. hymenophylla REI .87/63 1/86 B. tenuifolia PAR 22 1/100 B. spec. Vietnam1 PAR 1/96 B. spec. Thailand1 PAR PAR .92/53 1/71 1/100 B. elisabethae PAR B. grandis DIP 24, 26 B. alicida ALI 1/95 B. puttii DIP B. spec. China1 DIP 1/100 1/98 B. rabilii DIP 1/100 B. brandisiana REI .88/81 1/98 B. aceroides DIP .99/96 B. demissa PAR 1/100 B. flagellaris DIP B. versicolor PLA 22 --/55 B. venusta PLA 44 1/100 B. decora PLA 22 1/96 .86/50 1/100 B. pavonina PLA 22 B. sikkimensis PLA 22 1/98 B. palmata PLA 22 1/100 1/98 B. sizemoreae PLA B. hatacoa PLA 22 .97/73 B. spec. China2 PLA --/52 1/95 B. spec. Sulawesi1 SPH PLA-SPH B. areolata PLA 1/100 .91/64 B. multangula SPH .99/74 B. robusta SPH 88 B. roxburghii SPH 22 1/94 B. obovoidea SPH 1/98 B. acetosella SPH B. silletensis SPH 1/100 B. aptera SPH 1/90 B. longifolia SPH 22 B. morsei COE 30 1/100 B. masoniana COE COE B. pendula PET B. burbidgei PET PET I 1/100 1/98 B. amphioxus PET B. kingiana RID B. spec. Sumbawa1 REI 1/100 1/86 B. goegoensis REI 34 1/89 REI B. muricata REI .93/66 B. sudjanae REI .84/55 --/51 B. cleopatrae DIP B. nigritarum DIP 44 1/100 B. fenicis DIP 26, 56 DIP II B. chloroneura DIP 1/100 1/100 B. hernandioides DIP B. verecunda BRA 1/97 BRA B. lepida BRA .88/68 B. corrugata PET B. laruei PET 1/96 B. wrayi PET 1/98 .80/54 B. multijugata PET .89/78 B. chlorosticta PET B. varipeltata PET 1/98 1/90 B. masarangensis PET B. siccacaudata PET 1/83 B. watuwilensis PET .98/75 PET II B. guttapila PET .93/74 .99/83 B. pseudolateralis PET 1/85 B. koordersii PET B. symsanguinea SYM .99/75 B. strigosa SYM 1/100 B. argenteomarginata SYM B. polilloensis PET .99/74 1/98 B. negrosensis PET .97/64 B. brevirimosa PET 44 1/100 B. serratipetala PET 44 A B C 1 DIP I grade C 2 D Fig. 2. Bayesian majority rule consensus tree (cpDNA data: ndhA intron, ndhF-rpl32, rpl32-trnL; 3 data partitions; 84 taxa). Bayesian posterior probability (PP) support values P0.8 and bootstrap (BS) support values P50 of corresponding clades of the best-scoring tree derived from the ML analysis are indicated at each node: PP/BS. Broken lines indicate branches which lead to nodes with a PP < 0.95 and/or BS < 70. Somatic chromosome counts (Doorenbos et al., 1998; Gu et al., 2007; Legro and Doorenbos, 1969, 1971, 1973; Oginuma and Peng, 2002) are indicated as bold numbers after the taxon names and section abbreviations. Sectional placement of taxa is indicated by the following abbreviations: ALI: Alicida, AUG: Augustia, BAR: Barya, BRA: Bracteibegonia, COE: Coelocentrum, DIP: Diploclinium, GIR: Gireoudia, HAA: Haagea, MEZ: Mezierea, PAR: Parvibegonia, PEL: Peltaugustia, PET: Petermannia, PLA: Platycentrum, QUA: Quadrilobaria, REI: Reichenheimia, RID: Ridleyella, SOL: Solananthera, SPH: Sphenanthera, SQU: Squamibegonia, SYM: Symbegonia, TET: Tetraphila. 434 D.C. Thomas et al. / Molecular Phylogenetics and Evolution 60 (2011) 428–444 derived from the maximum likelihood analysis indicated at the nodes. Asian and Socotran taxa form a strongly supported clade (PP: 1, BS: 100). The relationships of two clades, Clades A and B, which diverge at two of the deepest nodes within the Asian–Socotran crown group are only poorly supported. Clade A (PP: 0.96, BS: 77) includes Begonia dipetala Graham (section Haagea) as sister to a strongly supported clade including two Socotran species of section Peltaugustia. Clade B (PP: 1, BS:100) is composed of Begonia floccifera Bedd. (section Reichenheimia) and B. malabarica (unplaced to section). The rest of the Asian species fall into two major clades: Clades C and D. Clade C is strongly supported in the Bayesian analyses, but only moderately supported in the ML analysis (PP: 1, BS: 71). Within Clade C, a clade consisting of species placed in section Parvibegonia, as well as Begonia hymenophylla Gagnep. (section Reichenheimia) and Begonia smithiae Geddes (section Platycentrum) (subclade C1, PP: 1, BS: 86) is the sister clade to the rest of the clade (subclade C2, PP: 1, BS: 95). Within subclade C2 lineages predominantly consisting of species placed in section Diploclinium, but also including Begonia alicida C.B. Clarke (section Alicida), Begonia brandisiana Kurz (section Reichenheimia) and Begonia demissa Craib (section Parvibegonia), form a grade, and species placed in sections Platycentrum and Sphenanthera form a strongly supported subclade (PP: 1, BS: 96). Within this Platycentrum–Sphenanthera clade species placed in section Sphenanthera fall into two clades, one of which is an intermixed assemblage of species placed in sections Sphenanthera and Platycentrum. Clade D is strongly supported in both BI and ML analyses (PP: 1, BS: 98). Section Coelocentrum is sister to the well to strongly supported rest of the clade (PP: 1, BS: 86) which comprises five strongly supported subclades: Subclade PET I includes three species of section Petermannia; subclade REI comprises species of section Reichenheimia; subclade DIP II comprises species of section Diploclinium; subclade BRA comprises species of section Bracteibegonia and forms the sister clade to subclade PET II, which contains species in section Petermannia as well as a subclade of New Guinean species placed in section Symbegonia nested within. The relationships among the PET I, REI, DIP II, and BRA/PET II clades and the section Ridleyella are only poorly supported. Most major Asian sections are not supported as monophyletic. Section Diploclinium is polyphyletic with Asian mainland species found in Clade C, while Philippine Diploclinium species fall into Clade D. Section Reichenheimia is polyphyletic with Malesian species forming a strongly supported subclade in clade D, while other species assigned to this section are found nested within a clade of mainland taxa of section Diploclinium in Clade C2 and closely associated with section Parvibegonia in Clade C1. One Indian species, B. floccifera, which has also been assigned to section Reichenheimia, falls into an early divergent clade (Clade B) of unclear affinity. Species of section Petermannia are found in two apparently only distantly related subclades in Clade D (subclades PET I and PET II), and species in section Symbegonia are nested within a subclade of the larger Petermannia clade (PET II). Sections Platycentrum and Sphenanthera are also not monophyletic and form interdigitated assemblages in Clade C2. B. smithiae, which has been assigned to section Platycentrum in the past, is only distantly related to other species in this sections and exhibits close relationships to section Parvibegonia (Clade C1). B. demissa (section Parvibegonia) is not retrieved in a clade with other species assigned to section Parvibegonia (PAR), but falls into the DIP I grade in Clade C2. 3.2. Ancestral character reconstructions Parsimony and likelihood ancestral character state reconstructions of five characters including the presence and type of specialized perennating organs, fruit types, locule numbers, placentation types, and placenta divisions are presented in Figs. 3–7. Presence of specialized perennation organs (Fig. 3): Reconstructions of the ancestral character states at the deepest nodes of the phylogeny are equivocal. Tubers are reconstructed as most likely character state at the crown node of Clade C (proportional likelihood [PL]: 0.88). The reconstructions indicate a character transition from the tuberous habit found in species placed in section Diploclinium to the rhizomatous habit exhibited by species in the Platycentrum–Sphenanthera clade (PL: 1). Within the rhizomatous Platycentrum–Sphenanthera clade, rhizomes were lost in a clade which comprises Begonia acetosella Craib, Begonia longifolia Blume and Begonia aptera Blume, which exhibit erect stems and fibrous root systems. Rhizomes are reconstructed as most likely character state at the crown node of Clade D (PL: 0.96). The reconstruction indicates the loss of rhizomes in the lineage comprizing sections Bracteibegonia, Petermannia and section Symbegonia (PL: 0.99). Within this lineage, only Begonia siccacaudata J. Door. exhibits tuberous organs, which evolved independently from the tubers of other lineages of Asian Begonia. Fruit type, ovary locule numbers and placentation type (Figs. 4– 6): Character state reconstructions indicate that three-locular ovaries developing into dry capsules at maturity are ancestral in Asian Begonia. Two-locular ovaries and rainballist fruits likely evolved independently in sections Parvibegonia (Clade C1), Platycentrum (Clade C2) and Malesian Diploclinium (Clade D). Two-locular ovaries also evolved independently in section Ridleyella (Clade D). The reconstructions indicate at least two independent character transitions from two-locular, rain-ballist fruits to fleshy fruits within the Platycentrum–Sphenanthera clade in Clade C2. Independently, fleshy fruits evolved in some species of section Petermannia (Clade D). The unilocular ovaries with parietal placentation in section Coelocentrum (Clade D) are likely derived from three-locular ovaries with axile placentation, which are indicated to be ancestral in Asian Begonia. Placenta division (Fig. 6): Bilamellate placentae are reconstructed as most likely character state at the crown node of Asian Begonia (PL: 1). The reconstructions indicate that placentae with undivided lamellae are homoplasious and evolved in sections Ridleyella and Reichenheimia, which fall in Clade D, and two taxa which fall in Clade C (B. hymenophylla in clade C1 and in B. brandisiana in Clade C2). Undivided placentae can also be found in early divergent lineages comprizing species assigned to sections Haagea, Reichenheimia, Peltaugustia and species unplaced to section (Clades A and B). 4. Discussion The results indicate that Asian and Socotran Begonia species form a well supported clade. This confirms, with greater taxon sampling, the results of former phylogenetic studies (Forrest and Hollingsworth, 2003; Forrest et al., 2005; Goodall-Copestake et al., 2010). 4.1. Major subclades of Asian Begonia The majority of species within the Socotran–Asian clade fall into two well supported major subclades: Clades C and D. Clade C exhibits subclades of species of taxa which are most diverse on the Asian mainland including sections Parvibegonia, continental Asian species placed in section Diploclinium, and species of section Platycentrum s.l. (inclusive section Sphenanthera). Chromosome counts of species in this clade indicate a predominant somatic chromosome number of 2n = 22 (Fig. 2), but there are also polyploid series and numerous aneuploid derivates (Doorenbos et al., 1998; Gu et al., 2007; Legro and Doorenbos, 1969, 1971, 1973; Oginuma and Peng, 2002). Further chromosome counts in sections 435 D.C. Thomas et al. / Molecular Phylogenetics and Evolution 60 (2011) 428–444 Perennation organs Not specialised Rhizomatous Thickened stem base Tuberous Equivocal Node absent B. oxyloba MEZ B. goudotii QUA B. polygonoides TET B. poculifera SQU B. sutherlandii AUG B. dregei AUG B. nelumbiifolia GIR B. radicans SOL B. boliviensis BAR B. dipetala HAA B. samhaensis PEL B. socotrana PEL B. malabarica IGN B. floccifera REI B. hymenophylla REI B. smithiae PLA B. tenuifolia PAR B. spec. Vietnam1 PAR B. spec. Thailand1 PAR B. elisabethae PAR B. grandis DIP B. alicida ALI B. spec. China1 DIP B. puttii DIP B. rabilii DIP B. brandisiana REI B. aceroides DIP B. demissa PAR B. flagellaris DIP B. versicolor PLA B. venusta PLA B. decora PLA B. pavonina PLA B. sikkimensis PLA B. palmata PLA B. sizemoreae PLA B. roxburghii SPH B. obovoidea SPH B. silletensis SPH B. acetosella SPH B. longifolia SPH B. aptera SPH B. hatacoa PLA B. spec. China2 PLA B. spec. Sulawesi1 SPH B. areolata PLA B. robusta SPH B. multangula SPH B. masoniana COE B. morsei COE B. kingiana RID B. spec. Sumbawa1 REI B. goegoensis REI B. sudjanae REI B. muricata REI B. cleopatrae DIP B. nigritarum DIP B. fenicis DIP B. hernandioides DIP B. chloroneura DIP B. verecunda BRA B. lepida BRA B. corrugata PET B. laruei PET B. wrayi PET B. chlorosticta PET B. multijugata PET B. masarangensis PET B. varipeltata PET B. siccacaudata PET B. watuwilensis PET B. guttapila PET B. koordersii PET B. pseudolateralis PET B. symsanguinea SYM B. strigosa SYM B. argenteomarginata SYM B. negrosensis PET B. polilloensis PET B. serratipetala PET B. brevirimosa PET A B C 1 C 2 D Fig. 3. Parsimony and likelihood ancestral character reconstruction: Perennation organs. Character reconstructions across 10,000 Bayesian input trees are shown on the Bayesian majority rule consensus tree (ndhA intron, ndhF-rpl32, rpl32-trnL; 3 data partitions; 81 taxa). Branch colour indicates parsimony optimization of ancestral character states. Pie charts at each node illustrate the likelihood reconstructions and show the proportion of the average likelihood received by each character state as the ancestral character of a given clade. Sectional placement of taxa is indicated by three-letter abbreviations (see captions of Fig. 2). Parvibegonia and mainland Diploclinium are needed to test whether a base chromosome number of n = 11 was likely present in the most recent common ancestor of Clade C or whether it is characteristic for a subclade including sections Platycentrum s.l. and closely related, continental Asian species in section Diploclinium. Clade D includes species of the predominantly Chinese section Coelocentrum, species in the predominantly or exclusively Malesian sections Ridleyella, Bracteibegonia, Petermannia, Symbegonia, and Malesian species placed in the polyphyletic sections Diploclinium and Reichenheimia. Chromosome counts are sparse for the Malesian taxa, but the vast majority of species in section Coelocentrum, which is the sister to the rest of the clade, have a chromosome number of 2n = 30 (Gu et al., 2007; Ku, 2006), and somatic chromosome numbers of 30 and 44 have been reported from some Malesian species in sections Reichenheimia, Diploclinium and Petermannia (see Fig. 2; Doorenbos et al., 1998; Legro and Doorenbos, 1969, 1971, 1973). Legro and Doorenbos (1971, 1973) hypothesized that the somatic chromosome numbers of 44 in some species in sections Reichenheimia and Petermannia likely arose from triploids of species with 30 somatic chromosomes. This indicates 436 D.C. Thomas et al. / Molecular Phylogenetics and Evolution 60 (2011) 428–444 Fruit type Dry capsules Rain-ballist capsules Fleshy Equivocal Node absent B. oxyloba MEZ B. goudotii QUA B. polygonoides TET B. poculifera SQU B. sutherlandii AUG B. dregei AUG B. nelumbiifolia GIR B. radicans SOL B. boliviensis BAR B. dipetala HAA B. samhaensis PEL B. socotrana PEL B. malabarica IGN B. floccifera REI B. hymenophylla REI B. smithiae PLA B. tenuifolia PAR B. spec. Vietnam1 PAR B. spec. Thailand1 PAR B. elisabethae PAR B. grandis DIP B. alicida ALI B. spec. China1 DIP B. puttii DIP B. rabilii DIP B. brandisiana REI B. aceroides DIP B. demissa PAR B. flagellaris DIP B. versicolor PLA B. venusta PLA B. decora PLA B. pavonina PLA B. sikkimensis PLA B. palmata PLA B. sizemoreae PLA B. roxburghii SPH B. obovoidea SPH B. silletensis SPH B. acetosella SPH B. longifolia SPH B. aptera SPH B. hatacoa PLA B. spec. China2 PLA B. spec. Sulawesi1 SPH B. areolata PLA B. robusta SPH B. multangula SPH B. masoniana COE B. morsei COE B. kingiana RID B. spec. Sumbawa1 REI B. goegoensis REI B. sudjanae REI B. muricata REI B. cleopatrae DIP B. nigritarum DIP B. fenicis DIP B. hernandioides DIP B. chloroneura DIP B. verecunda BRA B. lepida BRA B. corrugata PET B. laruei PET B. wrayi PET B. chlorosticta PET B. multijugata PET B. masarangensis PET B. varipeltata PET B. siccacaudata PET B. watuwilensis PET B. guttapila PET B. koordersii PET B. pseudolateralis PET B. symsanguinea SYM B. strigosa SYM B. argenteomarginata SYM B. negrosensis PET B. polilloensis PET B. serratipetala PET B. brevirimosa PET A B C 1 C 2 D Fig. 4. Parsimony and likelihood ancestral character reconstruction: Fruit type. Character reconstructions across 10,000 Bayesian input trees are shown on the Bayesian majority rule consensus tree (ndhA intron, ndhF-rpl32, rpl32-trnL; 3 data partitions; 81 taxa). Branch colour indicates parsimony optimization of ancestral character states. Pie charts at each node illustrate the likelihood reconstructions and show the proportion of the average likelihood received by each character state as the ancestral character of a given clade. Sectional placement of taxa is indicated by three-letter abbreviations (see captions of Fig. 2). that a somatic chromosome number of 2n = 30 may be ancestral for Clade D. The relationships of two early divergent clades which comprise Indian, Sri Lankan and Socotran taxa assigned to sections Reichenheimia (B. floccifera), Haagea (B. dipetala), Peltaugustia (B. samhaensis, B. socotrana) and one species unplaced to section (B. malabarica) are only poorly supported and their affinities to the main clades C and D remain obscure. The presence of a somatic chromosome numbers of 30 and 60 for three Indian and Sri Lankan species, and 28 for one of the two Socotran species may indicate that a primary base chromosome number of n = 15 is ancestral in Asian Begonia. 4.2. Polyphyly of section Reichenheimia and homoplasy of undivided placenta lamellae in Asian Begonia Since Klotzsch (1854) erected the genus Reichenheimia to accommodate two species from Sri Lanka and South India, which are characterized by a tuberous, acaulescent habit, and three-locular ovaries with undivided placentae, almost all Asian Begonia species which exhibit ovaries with undivided placenta lamellae were placed in section Reichenheimia. Exceptions are two species placed in section Ridleyella, which Irmscher (1929) described to accommodate species with two-locular ovaries and undivided placentae, 437 D.C. Thomas et al. / Molecular Phylogenetics and Evolution 60 (2011) 428–444 Ovary locule number 1-locular 2-locular 3-locular 4-locular Equivocal Node absent B. oxyloba MEZ B. goudotii QUA B. polygonoides TET B. poculifera SQU B. sutherlandii AUG B. dregei AUG B. nelumbiifolia GIR B. radicans SOL B. boliviensis BAR B. dipetala HAA B. samhaensis PEL B. socotrana PEL B. malabarica IGN B. floccifera REI B. hymenophylla REI B. smithiae PLA B. tenuifolia PAR B. spec. Vietnam1 PAR B. spec. Thailand1 PAR B. elisabethae PAR B. grandis DIP B. alicida ALI B. spec. China1 DIP B. puttii DIP B. rabilii DIP B. brandisiana REI B. aceroides DIP B. demissa PAR B. flagellaris DIP B. versicolor PLA B. venusta PLA B. decora PLA B. pavonina PLA B. sikkimensis PLA B. palmata PLA B. sizemoreae PLA B. roxburghii SPH B. obovoidea SPH B. silletensis SPH B. acetosella SPH B. longifolia SPH B. aptera SPH B. hatacoa PLA B. spec. China2 PLA B. spec. Sulawesi1 SPH B. areolata PLA B. robusta SPH B. multangula SPH B. masoniana COE B. morsei COE B. kingiana RID B. spec. Sumbawa1 REI B. goegoensis REI B. sudjanae REI B. muricata REI B. cleopatrae DIP B. nigritarum DIP B. fenicis DIP B. hernandioides DIP B. chloroneura DIP B. verecunda BRA B. lepida BRA B. corrugata PET B. laruei PET B. wrayi PET B. chlorosticta PET B. multijugata PET B. masarangensis PET B. varipeltata PET B. siccacaudata PET B. watuwilensis PET B. guttapila PET B. koordersii PET B. pseudolateralis PET B. symsanguinea SYM B. strigosa SYM B. argenteomarginata SYM B. negrosensis PET B. polilloensis PET B. serratipetala PET B. brevirimosa PET A B C 1 C 2 D Fig. 5. Parsimony and likelihood ancestral character reconstruction: Ovary locule number. Character reconstructions across 10,000 Bayesian input trees are shown on the Bayesian majority rule consensus tree (ndhA intron, ndhF-rpl32, rpl32-trnL; 3 data partitions; 81 taxa). Branch colour indicates parsimony optimization of ancestral character states. Pie charts at each node illustrate the likelihood reconstructions and show the proportion of the average likelihood received by each character state as the ancestral character of a given clade. Sectional placement of taxa is indicated by three-letter abbreviations (see captions of Fig. 2). and the erect, non-tuberous B. dipetala, which Klotzsch (1854) placed in a separate monotypic genus Haagea Klotzsch. Clarke (1879) placed Indian and Indo-Chinese species with undivided placenta lamellae, including the Indian and Sri Lankan B. floccifera and B. malabarica, in section Uniplacentales C.B. Clarke, and subsequently most of the species in Clarke’s section Uniplacentales were placed in section Reichenheimia (Doorenbos et al., 1998). However, Irmscher (1939) already emphasized that the practise of pooling all Asian species with three-locular ovaries and undivided placentae in section Reichenheimia had resulted in a morphologically heterogeneous group. The analyses of the cpDNA sequence data indicate that this morphological heterogeneity is correlated with the polyphyly of the section. Twenty-eight Malesian species are currently placed in section Reichenheimia (Hughes, 2008; Hughes et al., 2009; Kiew and Sang, 2009). The four species of this section from Sumatra, Java and Sumbawa which were included in the molecular analyses form a clade which is strongly supported as monophyletic in Clade D of the cpDNA phylogeny. There is no indication for a close relationship of these species to the Indian–Sri Lankan species B. floccifera (Clade B), or mainland Southeast Asian species also placed in section Reichenheimia which are found in Clades C1 and C2. The well developed rhizomes separate the Malesian species 438 D.C. Thomas et al. / Molecular Phylogenetics and Evolution 60 (2011) 428–444 Placenta division Unilamellate Bilamellate > 2 placenta lamellae Equivocal Node absent B. oxyloba MEZ B. goudotii QUA B. polygonoides TET B. poculifera SQU B. sutherlandii AUG B. dregei AUG B. nelumbiifolia GIR B. radicans SOL B. boliviensis BAR B. dipetala HAA B. samhaensis PEL B. socotrana PEL B. malabarica IGN B. floccifera REI B. hymenophylla REI B. smithiae PLA B. tenuifolia PAR B. spec. Vietnam1 PAR B. spec. Thailand1 PAR B. elisabethae PAR B. grandis DIP B. alicida ALI B. spec. China1 DIP B. puttii DIP B. rabilii DIP B. brandisiana REI B. aceroides DIP B. demissa PAR B. flagellaris DIP B. versicolor PLA B. venusta PLA B. decora PLA B. pavonina PLA B. sikkimensis PLA B. palmata PLA B. sizemoreae PLA B. roxburghii SPH B. obovoidea SPH B. silletensis SPH B. acetosella SPH B. longifolia SPH B. aptera SPH B. hatacoa PLA B. spec. China2 PLA B. spec. Sulawesi1 SPH B. areolata PLA B. robusta SPH B. multangula SPH B. masoniana COE B. morsei COE B. kingiana RID B. spec. Sumbawa1 REI B. goegoensis REI B. sudjanae REI B. muricata REI B. cleopatrae DIP B. nigritarum DIP B. fenicis DIP B. hernandioides DIP B. chloroneura DIP B. verecunda BRA B. lepida BRA B. corrugata PET B. laruei PET B. wrayi PET B. chlorosticta PET B. multijugata PET B. masarangensis PET B. varipeltata PET B. siccacaudata PET B. watuwilensis PET B. guttapila PET B. koordersii PET B. pseudolateralis PET B. symsanguinea SYM B. strigosa SYM B. argenteomarginata SYM B. negrosensis PET B. polilloensis PET B. serratipetala PET B. brevirimosa PET A B C 1 C 2 D Fig. 6. Parsimony and likelihood ancestral character reconstruction: Placentation type. Character reconstructions across 10,000 Bayesian input trees are shown on the Bayesian majority rule consensus tree (ndhA intron, ndhF-rpl32, rpl32-trnL; 3 data partitions; 81 taxa). Branch colour indicates parsimony optimization of ancestral character states. Pie charts at each node illustrate the likelihood reconstructions and show the proportion of the average likelihood received by each character state as the ancestral character of a given clade. Sectional placement of taxa is indicated by three-letter abbreviations (see captions of Fig. 2). from most continental Asian species assigned to section Reichenheimia, which are predominantly tuberous and either acaulescent or erect. This Malesian group is mainly distributed in the predominantly everwet Sunda Shelf region, and has its centre of diversity on Sumatra and the Malay Peninsula. Only few species extend the distributional range to eastern Malesia including the Lesser Sunda Islands, Southeast Sulawesi and the Malukku Islands (Hughes, 2008), and it is apparently absent from continental Asian regions north of the Thai–Malay Peninsula, which show a monsoonal seasonal climate with pronounced dry seasons. This group has been revised for the Malay Peninsula (Kiew, 2005) and an ongoing revi- sion of Sumatran species in section Reichenheimia at the Royal Botanic Garden Edinburgh (Mark Hughes, unpublished data) will provide the necessary morphological detail for the circumscription and formal description of this taxon at sectional rank. Ten species which are currently placed in section Reichenheimia have been described from Burma, Thailand, Laos, and Vietnam (Hughes, 2008), and eight Chinese species have been placed in this section (Gu, 2007). Most of the mainland Southeast Asian species placed in section Reichenheimia exhibit tubers and an acaulescent habit similar to continental Asian Dipoclinium lineages. Of this group, B. brandisiana and B. hymenophylla were included in the 439 D.C. Thomas et al. / Molecular Phylogenetics and Evolution 60 (2011) 428–444 Placentation type Axillary Septate Parietal Node absent B. oxyloba MEZ B. goudotii QUA B. polygonoides TET B. poculifera SQU B. sutherlandii AUG B. dregei AUG B. nelumbiifolia GIR B. radicans SOL B. boliviensis BAR B. dipetala HAA B. samhaensis PEL B. socotrana PEL B. malabarica IGN B. floccifera REI B. hymenophylla REI B. smithiae PLA B. tenuifolia PAR B. spec. Vietnam1 PAR B. spec. Thailand1 PAR B. elisabethae PAR B. grandis DIP B. alicida ALI B. spec. China1 DIP B. puttii DIP B. rabilii DIP B. brandisiana REI B. aceroides DIP B. demissa PAR B. flagellaris DIP B. versicolor PLA B. venusta PLA B. decora PLA B. pavonina PLA B. sikkimensis PLA B. palmata PLA B. sizemoreae PLA B. roxburghii SPH B. obovoidea SPH B. silletensis SPH B. acetosella SPH B. longifolia SPH B. aptera SPH B. hatacoa PLA B. spec. China2 PLA B. spec. Sulawesi1 SPH B. areolata PLA B. robusta SPH B. multangula SPH B. masoniana COE B. morsei COE B. kingiana RID B. spec. Sumbawa1 REI B. goegoensis REI B. sudjanae REI B. muricata REI B. cleopatrae DIP B. nigritarum DIP B. fenicis DIP B. hernandioides DIP B. chloroneura DIP B. verecunda BRA B. lepida BRA B. corrugata PET B. laruei PET B. wrayi PET B. chlorosticta PET B. multijugata PET B. masarangensis PET B. varipeltata PET B. siccacaudata PET B. watuwilensis PET B. guttapila PET B. koordersii PET B. pseudolateralis PET B. symsanguinea SYM B. strigosa SYM B. argenteomarginata SYM B. negrosensis PET B. polilloensis PET B. serratipetala PET B. brevirimosa PET A B C 1 C 2 D Fig. 7. Parsimony and likelihood ancestral character reconstruction: Placenta division. Character reconstructions across 10,000 Bayesian input trees are shown on the Bayesian majority rule consensus tree (ndhA intron, ndhF-rpl32, rpl32-trnL; 3 data partitions; 81 taxa). Branch colour indicates parsimony optimization of ancestral character states. Pie charts at each node illustrate the likelihood reconstructions and show the proportion of the average likelihood received by each character state as the ancestral character of a given clade. Sectional placement of taxa is indicated by three-letter abbreviations (see captions of Fig. 2). analyses, which indicate a close relationship of these species to mainland species in sections Diploclinium and to section Parvibegonia, respectively. Irmscher (1939) placed four Chinese species in section Reichenheimia, and the recent Flora of China treatise followed these placements, and another four species were added to the list (Gu, 2007). These eight species form a morphologically heterogeneous assemblage. Two Chinese species assigned to Reichenheimia, Begonia cylindrica D.R. Liang & X.X. Chen and Begonia filiformis Irmsch., are not tuberous, but exhibit well-developed rhizomes. B. cylindrica is morphologically aberrant for the section and exhibits rhizomes, fleshy, wingless fruits and uni- or bilamellate placentae (Gu et al., 2007; Tebbitt, 2005; Tebbitt et al., 2006). This species was placed in section Leprosae by Shui et al. (2002). Tebbitt et al. (2006) pointed out that B. cylindrica is morphologically most similar and maybe conspecific with B. leprosa Hance, which, based on ITS data, falls into a well supported clade with species of section Coelocentrum (Tebbitt et al., 2006; Thomas, 2010). B. filiformis also shows close morphological affinities to section Coelocentrum, in which it was placed by Shui et al. (2002). The character combination of rhizomes, uniloculate ovaries and the yellowish–greenish tepals strongly support this placement. The other six species exhibit a tuberous, acaulescent habit similar to Chinese species placed 440 D.C. Thomas et al. / Molecular Phylogenetics and Evolution 60 (2011) 428–444 in section Diploclinium. The development of variably undivided and bifid placenta lamellae has been described for some Chinese species placed in section Diploclinium (Begonia labordei H. Lév., Begonia fimbristipula Hance, Begonia wilsonii Gagnep.), and these species seem to morphologically link the tuberous Chinese Reichenheimia species with tuberous Chinese species assigned to section Diploclinium (Irmscher, 1939; Shui et al., 2002). However, a chromosome count of 2n = 30 for the Chinese Begonia henryi Hemsl. (Gu et al., 2007) in combination with undivided placenta lamellae may be indicative of a closer relationship to Indian–Sri Lankan species in the early divergent lineages of uncertain affinity (Clades A and B). Of the seventeen Begonia species known from the Ghats of India and from Sri Lanka (Jayasuriya, 1983; Uddin, 2007), eight, including the lectotype species of section Reichenheimia (Begonia thwaitesii Hook.), exhibit undivided placentae. Of these, B. floccifera (section Reichenheimia), B. malabarica (unplaced to section) and B. dipetala (section Haagea) were included in the analyses. These taxa, together with the only two species known from Socotra, fall into two early divergent clades, whose affinities to the main Clades C and D remain elusive. Phylogenetic analyses and detailed morphological analyses including a wider sampling of the South Indian–Sri Lankan taxa and the Chinese taxa are clearly needed to identify major lineages and apomorphies of Indian–Sri Lankan and continental Asian taxa currently placed in sections Reichenheimia and Diploclinium and to further clarify their phylogenetic relationships. Begonia section Reichenheimia, in its current circumscription, includes almost all Asian species with undivided placentae (Doorenbos et al., 1998), and this section is a prime example illustrating how the strong systematic importance associated with a single, homoplasious character resulted in the circumscription of a polyphyletic, morphologically heterogeneous taxon. 4.3. Polyphyly of section Diploclinium Doorenbos et al. (1998) lumped Asian species previously variously placed in sections Diploclinium, Knesebeckia, Begonia and Begoniastrum in section Diploclinium. Thereby, they restricted sections Begonia and Knesebeckia to the New World, and the separation of Asian taxa from these New World sections has subsequently been supported by morphological and molecular data (Badcock, 1998; Forrest, 2001; Goodall-Copestake et al., 2010). Doorenbos et al. (1998) emphasized that Diploclinium in their wide circumscription is a morphologically heterogeneous taxon. Previous molecular phylogenetic studies indicated the polyphyly of section Diploclinium, and showed that Philippine species form a well supported clade, while some Asian mainland species like Begonia grandis, Begonia alveolata Yu and Begonia wenshanensis C.M. Hu ex C.Y. Wu & T.C. Ku, may be more closely related to sections Platycentrum and Sphenanthera than to other species in section Diploclinium (Badcock, 1998; Forrest, 2001; Tebbitt et al., 2006). However, the phylogenetic relationships between the clades which include taxa placed in section Diploclinium have remained unclear, because of unresolved or poorly supported backbones of published phylogenies. The results of phylogenetic analyses of the cpDNA sequence dataset further clarify the relationships of this morphologically heterogeneous, polyphyletic taxon. Species assigned to section Diploclinium fall into a well supported clade of rhizomatous, Malesian species in Clade D, and form a grade of tuberous, erect or acaulescent continental Asian species in Clade C of the cpDNA phylogeny. The inclusion of B. alicida (sect. Alicida) in the latter group is not surprizing as it exhibits a tuberous, erect habit, three-locular capsules with bifid placentae and monadelphous androecia similar to many Asian mainland species in section Diploclinium. B. demissa Craib., which has been placed in section Parvibegonia by Doorenbos et al. (1998), also falls in the DIP I grade in Clade C. Hughes (2008) already pointed out that this species markedly differs from other species placed in section Parvibegonia and that it is probably conspecific with Begonia burmensis L.B. Sm. & Wassh., which is placed in the poorly known, small section Lauchea (Klotzsch) A.DC. The limited available molecular data indicates that continental Asian taxa placed in section Diploclinium do not form a monophyletic group, but instead form a grade within Clade C of the cpDNA phylogeny. Some taxa seem to be more closely related to species in sections Sphenanthera and Platycentrum than to other mainland Diploclinium species, and Begonia flagellaris Hara, which is endemic to Nepal, is the sister to the Platycentrum–Sphenanthera clade in the cpDNA phylogeny. The type species of section Diploclinium, B. grandis Otto ex A.DC., is retrieved as the earliest divergent lineage in the Diploclinium grade in Clade C2. Although it is obvious that more sampling of continental Asian lineages is needed, a decision will have to be made as to whether to expand the concept of section Diploclinium to cover the entire grade and accept a paraphyletic taxon, or to form a number of new monophyletic sections. A precedent has been set regarding the former case for section Symbegonia (Forrest and Hollingsworth, 2003). Malesian species placed in section Diploclinium include a large radiation on the Philippines (>40 spp.), but also five species described from Borneo and seven species from New Guinea (Hughes, 2008; Hughes et al., 2010). Most Malesian species in section Diploclinium exhibit a rhizomatous habit, and can thus be differentiated from continental Asian species assigned to the section which predominantly exhibit tubers and erect leafy stems or an acaulescent habit. Moreover, the five Philippine Diploclinium species included in the analyses of the cpDNA dataset are well supported as monophyletic and show a synapomorphic moderately large inversion of 345 bp or, due to deletion, 309 bp in the ndhF-rpl32 spacer. An ongoing revision of Philippine Diploclinium (Rubite, 2010) will provide the morphological detail to separate this predominantly Philippine group at sectional level. The current circumscription of section Diploclinium is not based on synapomorphic characters, but is primarily based on a plesiomorphic ovary and fruit syndrome with dry, three-locular capsules with bilamellate placentae, and the absence of easily observable morphological characters such as unilamellate placentae, fleshy fruits, and rain-ballist fruits, which are characteristic for other infrageneric taxa. Further molecular phylogenetic analyses based on a geographically robust taxon sampling of the predominantly tuberous Indian, Sri Lankan and continental Asian species currently placed in section Diploclinium are needed to identify major clades and apomorphic characters within this morphologically heterogeneous assemblage, and to clarify the relationships with section Platycentrum s.l. 4.4. Paraphyly of section Platycentrum, polyphyly of section Sphenanthera, and homoplasy of fleshy fruits in Asian Begonia Species of sections Sphenanthera and Platycentrum form a well supported clade in the molecular analyses of the cpDNA dataset. Within this clade, species of section Sphenanthera are found in two derived subclades nested within a paraphyletic section Platycentrum. This is largely congruent with findings of former morphological and molecular studies (Doorenbos et al., 1998; Forrest, 2001; Tebbitt et al., 2006). Species of these two sections are predominantly rhizomatous, although rhizomes were secondarily lost in a lineage in the B. longifolia complex (Tebbitt, 2003), and they exhibit androecia with characteristic anthers which dehisce via lateral slits and exhibit apically elongated connectives (Tebbitt et al., 2006). Some authors have emphasized that the two sections also share a somatic chromosome number of 2n = 22 (Legro and Doorenbos, 1973; Forrest, 2001; Tebbitt et al., 2006). However, the phylogenetic framework provided by the cpDNA phylogeny D.C. Thomas et al. / Molecular Phylogenetics and Evolution 60 (2011) 428–444 indicates that a primary base chromosome number of n = 11 might be a synapomorphic character for a wider taxon as somatic chromosome numbers of 2n = 22 have also been reported for the continental Asian Begonia picta Sm., Begonia rubella Buch.-Ham. ex D. Don, and B. fimbristipula which have been placed in section Diploclinium (Doorenbos et al., 1998; Forrest, 2001; Gu et al., 2007; Legro and Doorenbos, 1969, 1971, 1973). Moreover, somatic chromosome counts of 2n = 22 have been reported for Begonia tenuifolia Dryand. in section Parvibegonia (Legro and Doorenbos, 1971), which is part of the sister clade to the rest of the Clade C in the cpDNA phylogeny (Fig. 2). This might indicate that a primary base chromosome number of n = 11 is synapomorphic for taxa in Clade C, but further chromosome counts from species in sections Parvibegonia and Diploclinium are needed to test this hypothesis. Within the Platycentrum–Sphenanthera clade, three and four-locular, fleshy, indehiscent fruits, which are characteristic of section Sphenanthera, apparently evolved multiple times independently from ancestors which had two-locular, rain-ballist capsules, which characterize section Platycentrum. Doorenbos et al. (1998) emphasized that the fleshy-fruited Begonia robusta Blume, which is the type species of section Sphenanthera, is morphologically quite divergent from most species in the section, and could be easily accommodated in section Platycentrum, if it did not have three-locular ovaries. In the cpDNA phylogeny, B. robusta and the closely related and morphologically similar species B. multangula Blume, fall into a well supported clade which also includes taxa which have been placed in section Platycentrum and exhibit two-locular rain-ballist fruits. A second Sphenanthera clade is formed by other three- or four-locular fleshy-fruited species placed in section Sphenanthera including species of the B. longifolia complex. The analyses by Tebbitt et al. (2006) indicate that fleshy fruits likely evolved not just twice, but multiple times from rain-ballist ancestors within the Platycentrum–Sphenanthera clade. While these results can be partially explained by a different taxon sampling, it also has to be noted that the position of B. robusta within the Platycentrum–Sphenanthera clade in their ITS phylogeny is incongruent with its position in the cpDNA phylogeny presented here. This might indicate that hybridization and/or other biological processes like incomplete or differential homogenization of the ITS may have obscured phylogenetic relationships and character evolution within this group. A considerable number of putative hybrids between species in the Sphenanthera–Platycentrum group have been observed in the field (Kiew et al., 2003; Kiew, 2005; Teo and Kiew, 1999), and hybridization between and allopolyploidy in species of sections Platycentrum and Sphenanthera seem to be common in some regional Begonia floras such as the extensively studied Begonia flora of Taiwan (Chiang et al., 2001; Oginuma and Peng, 2002; Peng and Chen, 1991; Peng and Chiang, 2000; Peng and Ku, 2009; Peng and Sue, 2000). 4.5. Phylogenetic relationships of section Parvibegonia and homoplasy of rainballist capsules in Asian Begonia Sections Parvibegonia and Platycentrum share a character syndrome of two-locular fruits with bifid placenta lamellae, and De Candolle (1864) and Clarke (1879) lumped several species currently placed in section Parvibegonia together with species currently placed in section Platycentrum. However, Irmscher (1929) and Doorenbos et al. (1998) treated section Parvibegonia as distinct from section Platycentrum. Species in section Parvibegonia are predominantly tuberous, small plants, while species in section Platycentrum usually exhibit well developed rhizomes and a much more robust growth habit. Moreover, species in section Platycentrum exhibit characteristic anthers with elongated connectives, which are not present in section Parvibegonia. The phylogenetic relationships of section Parvibegonia remained unclear in a recent 441 phylogenetic study based on ITS data by Tebbitt et al. (2006) who pointed out that the possibility of two-locular rainballist capsules in section Platycentrum having arisen from similarly dispersed taxa in section Parvibegonia needed further investigation. The results of the analyses presented here indicate that section Parvibegonia, which is well supported as monophyletic, falls in a clade with B. hymenophylla and B. smithiae (Clade C1). This clade is the sister to a clade (Clade C2) comprizing a grade of continental Asian species assigned to section Diploclinium and a well supported subclade of species placed in section Platycentrum s.l. Hence we conclude that the two-locular, rain-ballist fruit syndrome evolved independently in sections Parvibegonia (clade C1) and Platycentrum (Clade C2). In addition, the placement of Begonia cleopatrae Coyle in our phylogeny (Clade D) confirms a third origin of two-locular rain-ballist fruit, following the recent description of the syndrome in this species and in some allied Philippine species in sect. Diploclinium (Hughes et al., 2010). 4.6. Paraphyly of section Petermannia and phylogenetic relationships of sections Bracteibegonia and Symbegonia Begonia section Petermannia comprises, with ca. 270 species, almost half of the Begonia species diversity in Southeast Asia, and the results of recent expeditions to under-collected areas on Borneo, Sulawesi, and New Guinea as well as studies of available herbarium material indicate that there are numerous morphologically distinct species awaiting description (e.g., Hughes et al., 2009; Kiew and Sang, 2009; Thomas et al., 2011). Section Petermannia has an almost exclusively Malesian distribution with only a few species extending the geographic range of the section to Thailand, Vietnam (Hughes, 2008) and Hainan and continental Southeast China (Shui & Chen, 2004). The Chinese species Begonia sinofloribunda Dorr., which exhibits peltate leaves, two- or sometimes three-tepaled female flowers and apparently protandrous inflorescences, is morphologically aberrant for section Petermannia, and the placement of this and two other Chinese species in section Petermannia (Shui & Chen, 2004) needs further investigation. The majority of species assigned to section Petermannia fall into Clade D (subclade PET II) of the cpDNA phylogeny, but a small, well-supported clade (PET I) of Bornean species placed in section Petermannia including B. amphioxus, B. burbidgei, and B. pendula Ridl. is not included in this clade and has unresolved or poorly supported relationships within Clade D. Species in the two separated Petermannia clades in the cpDNA phylogeny show conspicuous morphologically similarities. They lack rhizomes, they share characteristic protogynous, two- or sometimes one-flowered female inflorescences or partial inflorescences, which are basal to or separated from the male inflorescences. Species in both groups exhibit characteristic perforate cell wall thickenings in the endothecium layer of the anthers, while the endothecium cells of species of other Asian section predominantly show U-shaped wall thickenings (Tebbitt and MacIver, 1999). Moreover, species of both groups fall into a well supported, but poorly resolved clade in phylogenies based on nrDNA and mitochondrial DNA sequence data (Forrest, 2001; Goodall-Copestake et al., 2010; Tebbitt et al., 2006; Thomas, 2010). The inclusion of the three orphan species in the main Bracteibegonia–Petermannia–Symbegonia clade, as indicated by the ITS data, is concordant with the morphological and anatomical data, which supports a close relationship with other species placed in section Petermannia. The plastid and ITS gene trees seem to reflect different evolutionary histories, and the observed patterns are consistent with an evolutionary scenario which involves the transfer of foreign plastids into a Bornean Petermannia lineage via hybridization, introgression and plastid capture, and subsequent diversification on Borneo. Consequently, the three Bornean species 442 D.C. Thomas et al. / Molecular Phylogenetics and Evolution 60 (2011) 428–444 retrieved in Clade PET I were excluded from the ancestral character state reconstructions (see Section 2.5). Previous analyses of nrDNA data (26S, ITS) indicated that the erstwhile genus Symbegonia is nested within a Philippine and New Guinean clade of species assigned to Begonia section Petermannia (Forrest and Hollingsworth, 2003). This is corroborated by the analyses of non-coding plastid regions presented here. The two sections share a suite of generative characters; most of the species exhibit protogynous, two- or sometimes one-flowered female inflorescences or partial inflorescences, which are basal to or not directly associated with the male inflorescences. In addition, species in section Symbegonia lack rhizomes or tubers, as do the vast majority of species in section Petermannia. Forrest and Hollingsworth (2003) proposed the recognition of Symbegonia at sectional level rendering the large section Petermannia paraphyletic, but retaining a morphologically distinct taxon. Species placed in section Symbegonia can be easily identified by a floral syndrome with a syntepalous perianth and a characteristic monadelphous androecium, and most species in this section exhibit characteristic endothecial cells of the anthers with faint or lacking endothecial thickenings (Tebbitt and MacIver, 1999). However, the presence of basally fused tepals in some species assigned to section Petermannia (Forrest and Hollingsworth, 2003; Sands, 2009) as well as transitions between the endothecial types found in Symbegonia and Petermannia (Tebbitt and MacIver, 1999) indicate that there are morphologically and anatomically transitional species between the two taxa. Section Bracteibegonia includes only four species from Sumatra and Java, but there are several more species to be described from Sumatra (Hughes, 2008; Hughes et al., 2009). Species assigned to section Bracteibegonia form a well supported sister clade to the major Petermannia clade (PET II) in the cpDNA phylogeny. Some of the species assigned to section Bracteibegonia exhibit prostrate or weak stems, but rhizomes are not developed in this section. The lack of rhizomes is a derived character within Clade D of the cpDNA phylogeny, which supports a close relationship of the two sections. Doorenbos et al. (1998) emphasized the general morphological similarities of the two sections, but they retained section Bracteibegonia as distinct, because species of this section exhibit bisexual inflorescences which consist of one to few-flowered cymes and lack the clear separation of a basal female part and a distal male part, or separate female and male inflorescences, which are characteristic for the vast majority of species in section Petermannia (Irmscher, 1914; Doorenbos et al., 1998; Thomas et al., 2009). Our phylogeny clearly confirms the distinctness of section Bracteibegonia. 5. Conclusion The species in the mega-diverse genus Begonia exhibit an enormous vegetative diversity and even closely related species often show conspicuous differences in growth habits, indumentum characters and leaf morphologies. It is likely that this morphological diversity in vegetative characters arose due to both genetic drift and natural selection for adaptations to specific habitat conditions (Kidner and Umbreen, 2010; Matolweni et al., 2000; Neale et al., 2006). In contrast to this, generative organs provide easily observable, qualitative and quantitative, and apparently relatively complex characters such as differences in carpel and ovary locule numbers, placentation types, and pericarp types, which have been crucial for the circumscription of sections in Asian Begonia. The results of the phylogenetic analyses and ancestral character reconstructions presented here indicate that apparently complex fruit syndromes like three- or four-locular indehiscent fruits with fleshy pericarps, and two-locular rain-ballist capsules evolved multiple times independently in Asian Begonia. The genetic-developmental complexity, an essential criterion in character homology assessments, of the gain or loss of carpels, the inhibition of the development of locules in a three-locular ovary, the development of unilamellate or bilamellate placentae, and the development of a fleshy pericarp seem to have been overestimated in the past. The homoplasy of these characters has hampered the recognition of evolutionary lineages in Asian Begonia and most major Asian sections are not supported as monophyletic in the phylogenetic analyses. The strong systematic emphasis placed on single, homoplasious characters like undivided placenta lamellae (section Reichenheimia), fleshy pericarps (section Sphenanthera), and the recognition of sections primarily based on a plesiomorphic fruit syndrome and the absence of characteristic features of other taxa (section Diploclinium) has resulted in the circumscription of several highly polyphyletic taxa. The results indicate further that the presence and absence and type of stem metamorphoses and perennating organs like tubers and rhizomes and correlated growth habits are of greater systematic value in Asian Begonia than has been assumed in the past. These vegetative characters allow us to differentiate monophyletic predominantly Malesian species groups in both sections Reichenheimia and Diploclinium from distantly related Indian, Sri Lankan and continental Asian species placed in these polyphyletic sections. However, detailed comparative morphological and anatomical studies are needed to further investigate the homology and systematic importance of these organs in Begonia. Moreover, their ecological significance as perennation organs, anchor organs, and in habitat occupation and clonal reproduction are only poorly understood and studies are needed to establish whether there are correlations between precipitation, seasonality, substrate types and other environmental factors and the development of tubers, clusters of tubers and rhizomes. The current artificial infrageneric classification of Asian Begonia has a certain diagnostic, but only poor predictive value, which has hampered the understanding of the evolution of morphological and anatomical characters, karyotypes and the ecology and biogeography of Southeast Asian Begonia. The phylogeny derived from non-coding plastid data provides for the first time a reasonably resolved, and reasonably supported phylogenetic framework for Asian Begonia, which has the power to inform taxonomic work, and evolutionary and biogeographical studies, as well as clarifying aspects of character evolution and karyotype evolution. However, some problems remain and will need additional sampling to be solved, in particular how to deal with the classification of the ‘Diploclinium grade’ and the placement of the type species of some sections, particularly with respect to sect. Reichenheimia. Nevertheless, the results presented here are a large step towards facilitating a natural and stable infra-generic classification of Asian members of the mega-diverse genus Begonia. Conflict of interest None declare. Acknowledgments We are grateful to E.B. Walujo (LIPI, Cibinong Science Centre, BO), and M. Siregar and Hartutiningsih (LIPI, Bogor Botanic Gardens) for their support of our expeditions in Indonesia; to H. Wiriadinata, D. Girmansyah, A. Poulsen and K. Armstrong for providing silica-dried plant material; to the horticulture staff at Bali Botanic Garden, Bogor Botanic Gardens, the Royal Botanic Garden Edinburgh and the Royal Botanic Gardens Glasgow for their expert care of the Begonia collections; to the curators of A, B, BM, BO, CEB, E, L, K, SING and WAG for allowing us access to herbarium material and living collections. This research would not have been possible D.C. Thomas et al. / Molecular Phylogenetics and Evolution 60 (2011) 428–444 without the support of the Indonesian Ministry of Research and Technology (RISTEK) and Direktorat Jenderal Perlindungan Hutan dan Konservasi (DITJET PHKA). The funding of the first author’s PhD project by the M.L. MacIntyre Trust, and support of two expeditions to Sulawesi in 2008 and 2009 by the Royal Horticultural Society, the Stanley Smith (UK) Horticultural Trust and the Systematics Association is gratefully acknowledged. Appendix A. Supplementary material Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.ympev.2011.05.006. References Badcock, Z., 1998. A phylogenetic investigation of Begonia L. section Knesebeckia (Klotzsch) A.D.C. PhD thesis. University of Glasgow, Glasgow. Borsch, T., Quandt, D., 2009. Mutational dynamics and phylogenetic utility of noncoding chloroplast DNA. Plant Syst. Evol. 282, 169–199. Burnham, K.P., Anderson, D.R., 2004. Multimodel inference: understanding AIC and BIC in model selection. Sociol. Method Res. 33, 261–304. Chiang, T.Y., Hong, K.H., Peng, C.I., 2001. Experimental hybridization reveals biased inheritance of the internal transcribed spacer of the nuclear ribosomal DNA in Begonia x taipieiensis. J. Plant Res. 114, 343–351. Clarke, C.B., 1879. Begoniaceae. In: Hooker, J.D. (Ed.), Flora of British India. L. Reeve & Co., London, pp. 635–656. De Candolle, A., 1864. Begoniaceae. In: De Candolle, A. (Ed.), Prodromus Systematis Naturalis Regni Vegetabilis. Masson and Sons, Paris, pp. 266–408. Doorenbos, J., Sosef, M.S.M., de Wilde, J.J.F.E., 1998. The Sections of Begonia: Including Descriptions, Keys and Species Lists. Agricultural University, Wageningen. Drummond, A.J., Ashton, B., Buxton, S., Cheung, M., Cooper, A., Heled, J., Kearse, M., Moir, R., Stones-Havas, S., Sturrock, S., Thierer, T., Wilson, A., 2010. Geneious v5.1. <http://www.geneious.com/>. Edgar, R.C., 2004. MUSCLE: A multiple sequence alignment method with reduced time and space complexity. BMC Bioinformatics 5, 113. Fazekas, A.J., Steeves, R., Newmaster, S.G., 2010. Improving sequencing quality from PCR products containing long mononucleotide repeats. Biotechniques 48, 277– 285. Forrest, L.L., 2001. Phylogeny of Begoniaceae. PhD thesis. University of Glasgow, Glasgow. Forrest, L.L., Hollingsworth, P.M., 2003. A recircumscription of Begonia based on nuclear ribosomal sequences. Plant Syst. Evol. 241, 193–211. Forrest, L.L., Hughes, M., Hollingsworth, P.M., 2005. A phylogeny of Begonia using nuclear ribosomal sequence data and morphological characters. Syst. Bot. 30, 671–682. Frodin, D.G., 2004. History and concepts of big plant genera. Taxon 53, 753–776. Goodall-Copestake, W.P., Perez-Espona, S., Harris, D.J., Hollingsworth, P.M., 2010. The early evolution of the mega-diverse genus Begonia (Begoniaceae) inferred from organelle DNA phylogenies. Biol. J. Linn. Soc. 101, 243–250. Graham, S.W., Reeves, P.A., Burns, A.C.E., Olmstead, R.G., 2000. Microstructural changes in noncoding chloroplast DNA: interpretation, evolution, and utility of indels and inversions in basal angiosperm phylogenetic inference. Int. J. Plant Sci. 161, S83–S96. Gu, C.Z., 2007. Infrageneric classification of Begonia. In: Wu, Z.Y., Raven, P.H., Hong, D.Y. (Eds.), Flora of China. Science Press, Beijing, pp. 205–207. Gu, C.Z., Peng, C.I., Turland, N.J., 2007. Begoniaceae. In: Wu, Z.Y., Raven, P.H., Hong, D.Y. (Eds.), Flora of China. Science Press, Beijing, pp. 153–207. Huelsenbeck, J.P., Rannala, B., 2004. Frequentist properties of Bayesian posterior probabilities of phylogenetic trees under simple and complex substitution models. Syst. Biol. 53, 904–913. Huelsenbeck, J.P., Ronquist, F., 2001. MRBAYES: Bayesian inference of phylogenetic trees. Bioinformatics 17, 754–755. Hughes, M., 2008. An Annotated Checklist of Southeast Asian Begonia. Royal Botanic Garden Edinburgh, Edinburgh, U.K. Hughes, M., Pullan, M., 2007. Southeast Asian Begonia database. <http:// elmer.rbge.org.uk/begonia/> Hughes, M., Coyle, C., Rubite, R.R., 2010. A revision of Begonia section Diploclinium (Begoniceae) on the Philippine island of Palawan, including five new species. Edinburgh J. Bot. 67, 123–140. Hughes, M., Girmansyah, D., Ardi, W.H., Nurainas, 2009. Seven new species of Begonia from Sumatra. Gard. Bull. Singapore 61, 29–44. Irmscher, E., 1914. Die Verteilung der Geschlechter in den Inflorescenzen der Begoniaceen unter Berücksichtigung der morphologischen Verhältnisse. Bot. Jahrb. Syst. 50, 556–577. Irmscher, E., 1925. Begoniaceae. In: Engler, A., Prantl, K. (Eds.), Naturlichen Pflanzenfamilien. Wilhelm Engelmann, Leipzig, pp. 548–588. Irmscher, E., 1929. Die Begoniaceen der Malaiischen Halbinsel. Mitt. Inst. Allg. Bot. Hamburg 8, 86–160. 443 Irmscher, E., 1939. Die Begonien Chinas und ihre Bedeutung für die Frage der Formbildung in polymorphen Sippen. Mitt. Inst. Allg. Bot. Hamburg 10, 431– 557. Jayasuriya, A.H.M., 1983. Begoniaceae. In: Dassanayake, M.D., Fosberg, F.R. (Eds.), Flora of Ceylon. Amerind Publishing, New Delhi, pp. 137–152. Kass, R.E., Raftery, A.E., 1995. Bayes factors. J. Am. Stat. Assoc. 90, 773–795. Kelchner, S.A., 2000. The evolution of non-coding chloroplast DNA and its application in plant systematics. Ann. Missouri Bot. Gard. 87, 482–498. Kidner, C.A., Umbreen, S., 2010. Why is leaf shape so variable? Int. J. Dev. Biol. 4, 64– 75. Kiew, R., 2001. The limestone Begonias of Sabah, Borneo: flagship species for conservation. Gard. Bull. Singapore 53, 241–286. Kiew, R., 2005. Begonias of Peninsular Malaysia. Natural History Publications (Borneo), Kota Kinabalu. Kiew, R., Sang, J., 2009. Seven new species of Begonia (Begoniaceae) from the Ulu Merirai and Bukit Sarang limestone areas in Sarawak, Borneo. Gard. Bull. Singapore 60, 351–372. Kiew, R., Teo, L.L., Gan, Y.Y., 2003. Assessment of the hybrid status of some Malesian plants using Amplified Fragment Length Polymorphism. Telopea 10, 225–233. Klotzsch, J.F., 1854. Begoniaceen: Gattungen und Arten. Abhandlungen der Königlichen Akademie der Wissenschaften zu Berlin, Berlin. Ku, S.M., 2006. Systematics of Begonia sect. Coelocentrum (Begoniaceae) of China. MSc thesis. National Cheng-Kung University, Tainan. Ku, T.C., 1999. Begoniaceae. In: Ku, T.C. (Ed.), Flora Reipublicae Popularis Sinica. Science, Press, Beijing, pp. 126–269, 401–402. Legro, R.A.H., Doorenbos, J., 1969. Chromosome numbers in Begonia. Netherlands J. Agric. Sci. 17, 189–202. Legro, R.A.H., Doorenbos, J., 1971. Chromosome numbers in Begonia 2. Netherlands J. Agric. Sci. 19, 176–183. Legro, R.A.H., Doorenbos, J., 1973. Chromosome numbers in Begonia 3. Netherlands J. Agric. Sci. 21, 167–170. Lewis, P.O., 2001. A likelihood approach to estimating phylogeny from discrete morphological character data. Syst. Biol. 50, 913–925. Löhne, C., Borsch, T., 2005. Molecular evolution and phylogenetic utility of the petD group II intron: a case study in basal angiosperms. Mol. Biol. Evol. 22, 317–332. Maddison, W.P., Maddison, D.R., 2010. Mesquite: A Modular System for Evolutionary Analysis. Version 2.7.4. Matolweni, L.O., Balkwill, K., McLellan, T., 2000. Genetic diversity and gene flow in the morphologically variable, rare endemics Begonia dregei and Begonia homonyma (Begoniaceae). Am. J. Bot. 87, 431–439. Neale, S., Goodall-Copestake, W., Kidner, C., 2006. The evolution of diversity in Begonia. In: Teixeira da Silva, J.A. (Ed.), Floriculture, Ornamental and Plant Biotechnology: Advances and Topical Issues. Global Science Books, Isleworth, UK, pp. 606–611. Newton, M.A., Raftery, A.E., 1994. Approximate Bayesian inference with the weighted likelihood bootstrap. J. Roy. Stat. Soc. B Met. 56, 3–48. Nylander, J.A.A., Wilgenbusch, J.C., Warren, D.L., Swofford, D.L., 2008. AWTY (Are we there yet?): a system for graphical exploration of MCMC convergence in Bayesian phylogenetics. Bioinformatics 24, 581–583. Oginuma, K., Peng, C.I., 2002. Karyomorphology of Taiwanese Begonia (Begoniaceae): Taxonomic implications. J. Plant Res. 115, 225–235. Peng, C.I., Chen, Y.K., 1991. Hybridity and parentage of Begonia buimontana Yamamoto (Begoniaceae) from Taiwan. Ann. Missouri Bot. Gard. 78, 995–1001. Peng, C.I., Chiang, T.Y., 2000. Molecular confirmation of unidirectional hybridization in Begonia x taipeiensis Peng (Begoniaceae) from Taiwan. Ann. Missouri Bot. Gard. 87, 273–285. Peng, C.I., Ku, S.M., 2009. Begonia x chungii (Begoniaceae), a new natural hybrid in Taiwan. Bot. Stud. (Taipei) 50, 241–250. Peng, C.I., Sue, C.Y., 2000. Begonia x taipeiensis (Begoniaceae), a new natural hybrid in Taiwan. Bot. Bull. Acad. Sin. 41, 151–158. Phutthai, T., Sands, M., Sridith, K., 2009. Field surveys of natural populations of Begonia L. in Thailand. Thai Forest Bull. Bot., 186–196 (special issue). Plana, V., Gascoigne, A., Forrest, L.L., Harris, D., Pennington, R.T., 2004. Pleistocene and pre-Pleistocene Begonia speciation in Africa. Mol. Phylogen. Evol. 31, 449– 461. Posada, D., 2008. JModelTest: phylogenetic model averaging. Mol. Biol. Evol. 25, 1253–1256. Posada, D., Buckley, T.R., 2004. Model selection and model averaging in phylogenetics: advantages of akaike information criterion and Bayesian approaches over likelihood ratio tests. Syst. Biol. 53, 793–808. Rajbhandary, S., 2010. Systematic revision of the genus Begonia L. (Begoniaceae) in the Himalayas. PhD thesis. Tribhuvan University, Kathmandu, Nepal. Rambaut, A., Drummond, A.J., 2009. Tracer v1.5. <http://beast.bio.ed.ac.uk/Tracer>. Ronquist, F., Huelsenbeck, J.P., 2003. MrBayes 3: Bayesian phylogenetic inference under mixed models. Bioinformatics 19, 1572–1574. Rubite, R.R., 2010. Systematic studies on Philippine Begonia L. secion Diploclinium (Lindl.) A. D.C. (Begoniaceae). PhD thesis. De La Salle University, Manila. Sands, M.J.S., 2009. The Begonias of New Guinea – an overview. Blumea 54, 272– 277. Shaw, J., Lickey, E.B., Schilling, E.E., Small, R.L., 2007. Comparison of whole chloroplast genome sequences to choose noncoding regions for phylogenetic studies in angiosperms: the tortoise and the hare III. Am. J. Bot. 94, 275–288. Shinde, D., Lai, Y.L., Sun, F.Z., Arnheim, N., 2003. Taq DNA polymerase slippage mutation rates measured by PCR and quasi-likelihood analysis: (CA/GT)(n) and (A/T)(n) microsatellites. Nucleic Acids Res. 31, 974–980. 444 D.C. Thomas et al. / Molecular Phylogenetics and Evolution 60 (2011) 428–444 Shui, Y.M., Peng, C.I., Wu, C.Y., 2002. Synopsis of the Chinese species of Begonia (Begoniaceae), with a reappraisal of sectional delimitation. Bot. Bull. Acad. Sin. 43, 313–327. Shui, Y.-M., Chen, W.-H., 2004. Revision to sect. Petermannia of Begonia (Begoniaceae) in China. Acta Bot. Yunnan. 26, 482–486. Stamatakis, A., 2006. RAxML-VI-HPC: maximum likelihood-based phylogenetic analyses with thousands of taxa and mixed models. Bioinformatics 22, 2688– 2690. Suchard, M.A., Weiss, R.E., Sinsheimer, J.S., 2001. Bayesian selection of continuoustime Markov chain evolutionary models. Mol. Biol. Evol. 18, 1001–1013. Tavaré, S., 1986. Some probabilistic and statistical problems in the analysis of DNA sequences. In: Miura, R.M. (Ed.), Some Mathematical Questions in Biology – DNA Sequence Analysis. American Mathematical Society, Providence, RI, pp. 57–86. Tebbitt, M.C., 2003. Taxonomy of Begonia longifolia Blume (Begoniaceae) and related species. Brittonia 55, 19–29. Tebbitt, M.C., 2005. Begonias: Cultivation, Identification, and Natural History. Timber Press, Portland, Or. Tebbitt, M.C., Dickson, J.H., 2000. Amended descriptions and revised sectional assignment of some Asian Begonias (Begoniaceae). Brittonia 52, 112–117. View publication stats Tebbitt, M.C., Lowe-Forrest, L., Santoriello, A., Clement, W.L., Swensen, S.M., 2006. Phylogenetic relationships of Asian Begonia, with an emphasis on the evolution of rain-ballist and animal dispersal mechanisms in sections Platycentrum, Sphenanthera and Leprosae. Syst. Bot. 31, 327–336. Tebbitt, M.C., MacIver, C.M., 1999. The systematic significance of the endothecium in Begoniaceae. Bot. J. Linn. Soc. 131, 203–221. Teo, L.-L., Kiew, R., 1999. First record of a natural Begonia hybrid in Malaysia. Gard. Bull. Singapore 51, 103–118. Thomas, D.C., 2010. Phylogenetics and historical biogeography of Southeast Asian Begonia L. (Begoniaceae). PhD thesis. University of Glasgow, Glasgow. Thomas, D.C., Ardi, W.H., Hughes, M., 2009. Two new species of Begonia (Begoniaceae) from Central Sulawesi, Indonesia. Edinburgh J. Bot. 66, 103–114. Thomas, D.C., Ardi, W.H., Hughes, M., 2011. Nine new species of Begonia (Begoniaceae) from South and West Sulawesi, Indonesia. Edinburgh J. Bot. 68, 1–31. Uddin, A., 2007. Distribution and status of Indian Begonia L. species. J. Econ. Taxon. Bot. 31, 591–597. Warburg, O., 1894. Begoniaceae. In: Engler, A., Prantl, K. (Eds.), Naturlichen Pflanzenfamilien. Wilhelm Engelmann, Leipzig, pp. 121–150.