Molecular Phylogenetics and Evolution 76 (2014) 227–240

Contents lists available at ScienceDirect

Molecular Phylogenetics and Evolution

journal homepage: www.elsevier .com/ locate /ympev

Evolution of stamen number in Ptychospermatinae (Arecaceae): Insightsfrom a new molecular phylogeny of the subtribe

http://dx.doi.org/10.1016/j.ympev.2014.02.0261055-7903/� 2014 Elsevier Inc. All rights reserved.

⇑ Corresponding author.E-mail address: sophie.nadot@u-psud.fr (S. Nadot).

Elodie Alapetite a, William J. Baker b, Sophie Nadot a,⇑a Univ Paris-Sud, Laboratoire Ecologie, Systématique et Evolution, UMR 9079, Orsay F-91405, Franceb Royal Botanic Gardens, Kew, Richmond, Surrey TW9 3AB, UK

a r t i c l e i n f o a b s t r a c t

Article history:Received 20 November 2013Revised 19 February 2014Accepted 28 February 2014Available online 12 March 2014

Keywords:AndroeciumArecaceaeArecoideaeAreceaePhylogenyPalmsPalmaePtychospermatinaeStamen numberAGAMOUSPRKLow-copy nuclear DNA

The palm subtribe Ptychospermatinae (Arecaceae: Arecoideae) is naturally distributed in the South WestPacific area and contains 12 genera and around 60 species, including numerous popular ornamentals. Likemany palms, Ptychospermatinae flowers are small, trimerous, unisexual and always grouped into inflo-rescences of various sizes. However they exhibit a wide diversity in stamen number (a few to severaldozen or even hundreds) that is poorly understood from an evolutionary point of view. Althoughadvances have been made in elucidating phylogenetic relationships within Ptychospermatinae, somerelationships among and within genera still remain to be clarified. Here we used a combination of fivenuclear markers (nrITS2, the conserved nuclear intron BRSC10 and three low copy genes, PRK, RPB2and AGAMOUS) and three chloroplast markers (matK, ndhA and rps15-ycf1) to propose a new phyloge-netic hypothesis for the subtribe. The combination of all these markers improved the resolution androbustness of phylogenetic relationships within the subtribe, allowing us to identify four major clades.This phylogenetic framework was used to examine the evolution of stamen number in the clade. Theoptimization of stamen number on the phylogeny highlighted the high level of interspecific variability,showing that the character is highly labile and raising questions about the evolutionary and functionalsignificance of this lability.

� 2014 Elsevier Inc. All rights reserved.

1. Introduction

Over the last few decades, thanks to easy access to molecularcharacters, our knowledge of phylogenetic relationships betweenplant species has increased, providing useful frameworks to studyevolutionary trends in morphological and ecological traits. In par-ticular, the evolution of floral morphology has provoked greatinterest (e.g., Endress, 2011) since the flower is considered to bethe key to the evolutionary success of Angiosperms.

In palms (Arecaceae or Palmae) flowers are usually small (gen-erally less than 1 cm in length) and have the typical trimerousground plan of most monocot flowers, with three sepals, three pet-als and three carpels (Dransfield et al., 2008). There is a wide var-iation in androecium features and especially in stamen number. Arecent study examined the evolution of the character ‘‘stamennumber’’ on the genus-level phylogeny of the palm family (Nadotet al., 2011). From an ancestral state with six stamens (twice theperianth merism), both reduction and increase have occurred.

Ten genera (of 184) include species with the number of stamensper flower reduced to 3, and 84 genera include species displayingflowers with more than 6 stamens. Polyandry (defined here asthe presence of more than 6 stamens) is thus quite common inpalms and stamen number can reach large values (several dozento several hundreds). While the transitions towards polyandryfrom the ancestral state with 6 stamens have been identifiedamong genera, the patterns of transitions among species withpolyandrous flowers within genera are still poorly understood.Within Arecoideae (the largest subfamily of palms), the subtribePtychospermatinae (tribe Areceae) includes exclusively generathat produce polyandrous flowers, with a number of stamens vary-ing between 12 and 320. This subtribe represents a good candidateto examine the patterns of variation in stamen number among spe-cies with polyandrous flowers and to test whether particulartrends can be detected.

The subtribe Ptychospermatinae (Arecaceae: Arecoideae: Are-ceae; Hooker 1883) comprises 12 genera and approximately 60species (Dransfield et al., 2008; Palmweb, 2013). They are naturallywidespread in East Malesia (Moluccas, New Guinea, SalomonIslands, Vanuatu) extending to Fiji, Tonga, Samoa and Northern

228 E. Alapetite et al. / Molecular Phylogenetics and Evolution 76 (2014) 227–240

Australia. Two genera, Adonidia and Ponapea, have members moredistant from the main distribution area. One species of Adonidia (A.merrillii) is found west of Wallace’s Line, on the island of Palawanand its offshore islands (Philippines) and on the coast of Sabah(Borneo), and three species of Ponapea are found in the CarolineIslands (Pohnpei and Palau). Ptychospermatinae palms are rarelyused as primary resources by local populations, but the stems aresometimes used as timber (e.g. Ptychoccocus), or as food (e.g.Veitchia palm heart; Dransfield et al., 2008). However many speciesof the subtribe make elegant ornamentals widely planted through-out the tropics and subtropics (e.g. Adonidia merrillii, Balakaseemannii, Carpentaria acuminata, Normanbya normanbyi, Ptychospermamacarthurii or Wodyetia bifurcata; Dransfield et al., 2008). Several speciesare classified in the IUCN Red List of Threatened Species (http://www.iucnredlist.org/), and are threatened by reduction of their naturalhabitat.

In the past decades, there has been an important effort in delim-iting genera and species of the subtribe, resulting in numerous tax-onomic revisions (Baker and Heatubun, 2012; Essig, 1978; Hodel,2010; Irvine, 1983; Zona, 1999b, 2005; Zona and Essig, 1999; Zonaand Fuller, 1999). The monophyly of the subtribe is strongly sup-ported by morphology (Zona, 1999a). Synapomorphies includeblack scales at the apex of the leaf sheath, leaf segment tip erose,peduncular bract that pierces the tip of prophyll, lag time betweenpeduncular bract fall and anthesis, bullet-shape staminate buds(more or less perpendicular to rachilla), numerous stamens and apistillode usually bottle shaped or reduced and coned-shaped(e.g. in Veitchia). Within the subtribe, the limited variation in mor-phological characters makes synapomorphies difficult to identify,resulting in low morphological support for genera (Zona, 1999a).The latter study supported however the reinstatement of twomonotypic genera, Adonidia (previously included in Veitchia) andSolfia (previously included in Drymophloeus). In the first molecularphylogenies of palms including several members of the subtribe,the monophyly of Ptychospermatinae was not retrieved (Asmussenand Chase, 2001; Hahn, 2002). In 2006, two studies based respec-tively on plastid and nuclear markers (Asmussen et al., 2006; Norupet al., 2006) increased the taxonomic sampling and identified Pty-chospermatinae as potentially monophyletic (with moderate sup-port). In the complete generic-level phylogenetic analysis ofpalms published by Baker et al. (2009), the subtribe was resolvedas monophyletic with strong support. In a large survey of phyloge-netic relationship among arecoid palms (Baker et al., 2011) basedon sequences of the two low-copy nuclear genes PRK and RPB2and including 16 species of Ptychospermatinae, the subtribe wasresolved as monophyletic with moderate support (82% maximumparsimony bootstrap support and 88% maximum likelihood boot-strap support). Zona et al. (2011) recently produced the first molec-ular phylogeny focused on the subtribe, based also on PRK and RPB2and including 37 species from 12 genera. The results confirmed themonophyly of Ptychospermatinae with strong support, and identi-fied six major clades. They also found that the genera Drymophloeus,Ponapea and Veitchia were non-monophyletic and made taxonomicchanges accordingly. Although this study provided notableimprovement in our knowledge of the evolutionary history of thesubtribe, there were several polytomies in the tree and some nodeswere poorly supported.

Choosing DNA regions with the appropriate rate of variation toaddress issues on phylogenetic relationships among taxa is crucialand depends on the taxonomic level of the study. Rapidly evolvingDNA sequences are required for studying closely related species.Until recently, the majority of species-level phylogenies in plantsrelied on a limited set of non-coding plastid DNA loci (Hugheset al., 2006) because they are potentially variable at low taxonomiclevels and easy to amplify using universal (for plants) primers. Ithas been demonstrated that plastid DNA sequences evolve slowly

in palms compared to other monocot groups (Wilson et al., 1990;Gaut et al., 1996; Asmussen and Chase, 2001). However Cuencaand Asmussen-Lange (2007) challenged this statement and sug-gested that the evolution of plastid sequences in palms is complexand may vary among groups within the family. As a consequencewe chose to use a combination of eight markers from both the nu-clear and plastid genomes. Five markers were nuclear sequences:AGAMOUS 1 (AG1) gene, beta-carotene hydroxylase gene partialsequence (BRSC10), nuclear ribosomal internal transcribed spacer(nrITS2), intron 4 of the phosphoribulokinase (PRK) gene and in-tron 23 of RNA polymerase II (RPB2) gene. Three plastid markerswere added: the matK pseudogene, and the introns of ndhA andrps15-ycf1. Some of these DNA sequences have been widely usedin phylogenetic analyses of palms, like PRK, RPB2 (e.g. Lewis andDoyle, 2002; Roncal et al., 2008; Zona et al., 2011) and matK (e.g.Asmussen et al., 2006). The other markers were chosen becausethere were considered to be potentially variable enough to be use-ful for infra-generic reconstruction in palms (see Ludena et al.(2011) for AG1, Bacon et al. (2007) for BRSC10, Jeanson et al.(2011) for nrITS2, and Scarcelli et al. (2011) for the introns of ndhAand rps15-ycf1). Phylogenetic reconstruction was carried out usingeach marker separately and in combination. It is now widelyadmitted that adding taxa and markers is beneficial (even if char-acters are missing for some taxa), in order to increase the accuracyof phylogenetic inference (Zwickl and Hillis, 2002; Wiens andMorrill, 2011).

The aims of our study were (1) to improve the resolution of thephylogenetic relationships among genera and species of Ptycho-spermatinae, (2) to use the resulting historical framework toreconstruct the evolutionary history of stamen number in thisgroup.

2. Material and methods

2.1. Taxonomic sampling

Taxa sampled, voucher and accession number information, andGenBank accessions are given in Table S1 [Supplementary mate-rial]. We obtained specimens from 47 of the 60 species recognizedin Ptychospermatinae, representing all of the 12 genera of the sub-tribe (Dransfield et al., 2008). The ingroup included all species ofAdonidia, Brassiophoenix and Ptychococcus (2 species in each genus).We sampled 17/30 species of Ptychosperma, 2/3 species of Drymo-phloeus, 6/9 species of Balaka (following Hodel, 2010), 9/11 speciesof Veitchia and 1/4 species of Ponapea. The delimitation of generafollows Dransfield et al. (2008) with subsequent modifications byZona et al. (2011). The sampling includes an unpublished taxonthat originates from Indonesia, discovered on Gag Island off thewestern tip of New Guinea (called ‘New taxon Gag Island’ in thisstudy). Based on recent phylogenies (Baker et al., 2009, 2011), 13outgroup species were chosen among various subtribes of the tribeAreceae, in order to root the trees. Plant material or DNA sampleswere obtained from DNA banks or living collections of botanicalgardens around the world (Royal Botanic Gardens, Kew; Univers-itas Negeri Papua, Indonesia; Fairchild Tropical Botanic Garden;Aarhus University; Muséum National d’Histoire Naturelle, Paris;Montgomery Botanical Center, and Singapore Botanic Gardens;see also acknowledgements).

2.2. DNA extraction, amplification and sequencing

Total genomic DNA was extracted from silica-gel dried leafmaterial using the NucleoSpin� Plant II (Machery-Nagel, Düren,Germany) extraction kit, following the manufacturer’s instructions.Primers used to amplify each DNA region are given in Table S2

E. Alapetite et al. / Molecular Phylogenetics and Evolution 76 (2014) 227–240 229

[Supplementary material]. PCR reactions (50 lL) were performedusing the following conditions: 1x PCR buffer (MP biomedicals),1.5 mM MgCl2, 200 lM each dNTP, 1 lM each primer, 1U Taq poly-merase (Taq CORE Kit 10; MP biomedicals, Illkirch, France) and 1 or2 lL of DNA template. Reactions were run using MJ-Research(PTC-100) and Applied Biosystems (Veriti 96 Well) thermal cyclersprogrammed with the temperature profiles shown in Table S3[Supplementary material]. PCR products were purified and se-quenced by Beckman Coulter Genomics (Essex, UK) or at the Geno-scope (www.genoscope.cns.fr). Forward and reverse sequenceswere assembled and edited into contigs using CodonCode Aligner4.0.3 (CodonCode Corporation, Dedham, MA, USA). Whenever pos-sible, we used DNA from the same specimen to amplify all selectedmarkers, in order to obtain a homogeneous dataset. We combinedthe sequences obtained from two different specimens for four spe-cies only. All the sequences for RPB2 and some of the sequences forthe other markers were retrieved from GenBank (specified inAppendix A).

2.3. Alignment and phylogenetic analyses

Sequences were aligned with ClustalW implemented in MEGA4(Tamura et al., 2007) with manual adjustments. A highly variabledinucleotide microsatellite (TC/GA)n is present in intron 7 of AG1(Ludena et al., 2011), and a region of nine nucleotides within therps15-ycf1 sequence also displayed high variability. These regionswere too difficult to align with confidence and were therefore ex-cluded from all analyses. Unambiguously aligned gaps (excludingmissing data) were coded according to the modified complex indelcoding (MCIC, Simmons et al., 2007) using the software SeqState(Müller, 2005).

Phylogenetic analyses were conducted on each markerseparately (single region analyses), on plastid markers combined(plastid analysis), on nuclear markers combined (nuclear analysis),and finally on all markers combined (total evidence analysis; Kluge,1989).

2.3.1. Maximum parsimony analysisMaximum parsimony (MP) analyses were conducted using

PAUP* version 4.0b10 (Swofford, 2002). All characters were unor-dered and had equal weight (Fitch, 1971). Most parsimonious treeswere found with a heuristic search with 1000 replicates using ran-dom stepwise addition of sequence. The tree-bisection-reconnec-tion (TBR) branch swapping option was chosen and multipletrees (MULTREES) were saved at each replicate. The maximumnumber of trees that can be saved was fixed to 150,000 for allsearches. The analyses were performed under the DELTRAN option.When the analyses yielded 150,000 trees before completion of theheuristic search (BRSC10, nrITS2, ndhA, RPB2 and combined data-sets), a two-step heuristic search was applied: 1000 random repli-cates searches were conducted using TBR branch-swapping andonly 10 trees per replicates were saved. A round of TBR swapping(to completion) was performed on this set of trees. When the max-imum number of trees was reached before completion (BRSC10and combined matrices) the parsimony ratchet method (Nixon,1999) was also used. This method allows performing heuristicsearch with fewer trees in memory. Twenty independent replicatesof 200 iterations (with 15% of characters reweighted each itera-tion) were generated using the software PAUPRat (Sikes and Lewis,2001). Strict consensus trees were calculated from all most parsi-monious trees using PAUP. Support values for clades were calcu-lated by conducting 1000 bootstraps replicates with one searchper replicate under TBR branch swapping, starting trees built ran-dom addition sequence and 20 trees saved at each replicate.

2.3.2. Maximum likelihood and Bayesian inference analysesModel parameters for each data set and for coding vs. non-

coding regions within data sets were determined using jModelTest2.1.1 (Darriba et al., 2012). More models were tested for the MLanalysis (40) than for the BI analysis (24), to deal with the modelsimplemented in the softwares used for phylogenetic inference(PhyML and MrBayes). The Akaike information criterion was cho-sen to select the most appropriate model of DNA substitution.The models used are presented in Table S4 [Supplementarymaterial]. Maximum likelihood (ML) analyses were performedusing the PhyML 3.0 online server (Guindon et al., 2005). Indelswere not included because PhyML 3.0 does not treat coded gaps.Tree searching was made with five random BIONJ starting treesand SPR tree improvement. Branch support was assessed by calcu-lating 200 bootstrap pseudo-replicates.

Bayesian inference analyses (BI) were run using MrBayes 3.2(Ronquist et al., 2012). In the combined analyses and total evidenceanalysis, each region was treated as an individual partition, with itsown model of evolution, the same as the one used in the single re-gion analyses (see Table S4 [Supplementary material]). Indels weretreated as additional datatype (like a morphological present/absentcharacter). Two independent Markov chain Monte Carlo (MCMC)runs, each comprising four linked chains (one cold and threeheated; as default settings), were performed for 1,000,000 genera-tions, sampling every 100 generations. The convergence of the tworuns (the two tree samples becoming similar) was assessed bystopping the analysis when the average standard deviation was be-low 0.01 (stoprule = yes and stopval = 0.01 in the mcmc com-mand). For ndhA, rps15-ycf1, the plastid-DNA combined dataset,the nuclear-DNA combined dataset and the total evidence dataset,1,000,000 generations were not enough to reach adequate averagestandard deviation (<0.01). The analysis had to be run again forrespectively 1,510,000, 2,000,000, 1,787,000, 5,081,000 and1,331,000 generations. The first 25% trees were discarded asburn-in.

2.4. Stamen number optimization

All the species included in this study have polyandrous flowers(more than six stamens). Data on stamen number at species levelwere mostly obtained from taxonomic treatments available inthe literature. Information on stamen number was also obtainedfrom direct observation on flowers from spirit and herbarium col-lections (Aarhus University; Royal Botanic Gardens, Kew). Flowerswere dissected under a Zeiss Stemi SV6 stereomicroscope (CarlZeiss AG, Göttingen, Germany) to count the stamens. Range values,references and accession numbers are indicated in Table 2. Itshould be noted that in most studies (including this one), stamennumbers are estimated from a limited number of flowers and indi-viduals. This can obscure the real number of stamens and underes-timate the range of variation within the species. We were unable tofind data on stamen number for one taxon, Ptychosperma nicolai (itwas classified as dubious by Essig (1978) due to incompletespecimen).

For character optimization, stamen number was coded as a dis-crete character. Three states were defined, namely (0) between 6and 12 stamens, (1) between 15 and 60 stamens, correspondingto moderately polyandrous flowers and (3) between 61 and 300stamens, corresponding to highly polyandrous flowers. Given thatall Ptychospermatinae have flowers with 15 or more stamens, andoutgroups had 6 to 12 stamens, we chose to combine 6 and 12 sta-mens in a single state in order to limit the number of characterstates. We checked that separating 6 and 12 in two distinct statesdid not have any impact on the inferred ancestral state for the sub-tribe. The Bayesian 50% majority rule consensus tree resulting fromthe total evidence analysis was used to optimize the character with

230 E. Alapetite et al. / Molecular Phylogenetics and Evolution 76 (2014) 227–240

the MP method implemented in Mesquite 2.75 (Maddison andMaddison, 2011). Character states were treated as unordered,allowing any transition among states.

3. Results

3.1. DNA amplification and alignments

The average lengths of the DNA regions amplified were approx-imately 450 bp for AG1, 320 bp for BRSC10, 400 bp for nrITS2,663 bp for PRK, 1104 bp for RPB2, 1700 bp for matK, 1090 bp forndhA and 600 bp for rps15-ycf1. The number of characters in singlegene and combined regions alignments, including nucleotides andcoded gaps, is presented in Table 1. Amplification failed for a fewsamples resulting in datasets of different sizes (see Table 1). Wewere unable to obtain amplification for the full sequence of nrITS2in some specimens because the sequence of the reverse primer ap-peared within the sequence. We chose to keep specimens withmissing data in our analyses, in order to maximize the amount ofinformation used for phylogenetic inference.

3.2. Phylogenetic analyses

Maximum parsimony statistics (number of parsimony informa-tive characters, number of most parsimonious trees, tree lengthsand consistency (CI) and retention (RI) indices) for each individualmarker and for the combined datasets are given in Table 1. Therewas little homoplasy in most datasets (CI > 0.80) except for nrITS2,RPB2 and the combined matrices (respectively CI = 0.69, CI = 0.77,CI = 0.79, CI = 0.79 and CI = 0.78). Trees were rooted with Carpoxy-lon as potential sister group to all other species, according to recentphylogenies of the family (Baker et al., 2009) and of the tribe Are-ceae (Baker et al., 2011). The trees resulting from the MP (data notshown), ML (data not shown) and BI analyses (Figs. S1-S8 [Supple-mentary material]) conducted on single region alignments showednearly identical topologies. Bayesian trees however tended to beslightly more resolved than the trees obtained with the two othermethods.

3.2.1. Nuclear dataAll nuclear single region analyses but one retrieved the mono-

phyly of the subtribe with good support (Figs. S1–S5 [Supplemen-tary material]). The best support values are found for nrITS2(Fig. S1 [Supplementary material]), with PBS (parsimony bootstrapsupport) = 99%, LBS (likelihood bootstrap support) = 80% and PP(posterior probabilities from Bayesian analysis) = 1.00 (clade cred-ibility of 100%). The monophyly was less strongly supported with

Table 1Main features of the datasets used and statistics related to the maximum parsimonyMP = maximum parsimony. In square brackets: number of characters resulting from indel

Data set Number of

Ingroup taxa (outgroup taxa) Characters Varia

AG1 + [coded indels] 43 (9) 410 + [5] 66 +BRSC10 + [coded indels] 46 (6) 342 + [11] 91 +nrITS2 + [coded indels] 41 (6) 420 + [20] 174 +PRK + [coded indels] 44 (13) 663 + [8] 155 +RPB2 + [coded indels] 33 (13) 1104 + [29] 310 +nuclear-DNA combined matrix 47 (13) 2939 + [73] 796 +matK + [coded indels] 41 (7) 1706 + [3] 60 +ndhA + [coded indels] 46 (10) 1141 + [12] 39 +rps15-ycf1 + [coded indels] 45 (12) 622 + [10] 33 +Plastid-DNA combined matrix 47 (12) 3469 + [25] 132 +Total evidence matrix 47 (13) 6408 + [98] 928 +

a Number of most parsimonious trees generated by parsimony ratchet analyses.

BRSC10, PRK and RPB2 (Figs. S2–S4 [Supplementary material]),with, PBS = 90%, LBS < 50% and PP = 1.00 for BRSC10, PBS = 66%,LBS = 57% and PP = 1.00 for PRK, and PBS = 51%, LBS = 50% andPP = 1.00 for RPB2. The subtribe was not monophyletic with theAG1 marker but most nodes were poorly supported (Fig. S5[Supplementary material]). Since topological incongruences werenot supported, we combined the data for further analyses.

The nuclear DNA-combined alignment included 3012 charac-ters (2939 nucleotides positions and 73 coded indels) of which324 (11%) were parsimony-informative. In the combined nuclearanalysis (Fig. 1), Ptychospermatinae appeared monophyletic withstrong support (PBS = 100%, LBS = 100%, PP = 1.00). Five mainclades emerged from this nuclear analysis, of which two werestrongly supported (PBS and LBS > 80% and PP > 0.95). Only fourgenera, Veitchia, Drymophloeus, Ptychococcus and Brassiophoenix,were resolved as monophyletic with good support. Balaka andSolfia form a strongly supported clade, as well as Carpentaria andWodyetia.

3.2.2. Plastid dataThe three single plastid regions alignments displayed very few

variable sites (Table 1), and resulted in poorly resolved trees(Figs. S6–S8 [Supplementary material]). The monophyly of Ptycho-spermatinae was poorly to moderately supported. The best supportvalues were found with rps15-ycf1 (PBS = 61%, LBS = 58%, PP = 0.82;Fig. S6 [Supplementary material]. The plastid DNA-combinedalignment included 3494 characters (3469 nucleotides positionsand 25 coded indels) of which 47 (1.4%) were parsimony-informa-tive. The 50% majority-rule consensus tree resulting from theBayesian Inference analysis, with clade support, is shown inFig. 2. The subtribe was resolved as monophyletic with low support(PBS = 66%, LBS = 74%, PP = 0.68). Only three of the five majorclades emerging from the nuclear DNA-combined analysis were re-trieved in this analysis. These markers provided little information,as shown by the lack of resolution in the resulting tree, but some ofthe terminal groupings observed in the nuclear tree were alsofound in this plastid analysis (e.g. Solfia samoensis sister group toBalaka brachychlamys, or Ptychosperma salomonense sister groupto P. gracile, P. elegans and P. caryotoides).

3.2.3. Total evidence analyses based on the global nuclear-plastidalignment

Since we detected no well supported conflicts (>80% bootstrapsupport and >0.95 posterior probabilities) between the nuclearand plastid analyses, we combined the nuclear and plastid datain order to conduct a total evidence analysis (Kluge and Wolf,1993; Nixon and Carpenter, 1996; Wiens, 1998).

analysis. CI = consistency index; RI = retention index; PI = parsimony-informative;coding. In brackets: percentage of indels in the alignments.

Tree length CI RI

ble characters (%) PI characters (%) MP trees

[5] (17.1%) 23 + [3] (6.3%) 2 83 0.91 0.93[11] (29%) 36 + [7] (12.2%) 4004a 143 0.87 0.91

[20] (44.1%) 97 + [10] (24.3%) 1939 337 0.69 0.78[8] (24.3%) 58 + [4] (9.2%) 1113 217 0.87 0.92[29] (30%) 79 + [7] (7.6%) 4020a 437 0.77 0.88[73] (28.9%) 293 + [31] (10.8%) 3715a 1290 0.79 0.82

[3] (3.7%) 18 + [3] (1.2%) 448 71 0.94 0.96[12] (4.4%) 9 + [6] (1.3%) 2328 91 0.89 0.91[10] (6.8%) 7 + [4] (1.7%) 30 49 0.94 0.93

[25] (4.5%) 34 + [13] (1.4%) 3694a 223 0.79 0.85[98] (15.8%) 327 + [44] (5.7%) 3698a 1539 0.78 0.81

Table 2Stamen numbers for the species included in the phylogeny. Data were obtained from the literature and from the observation of herbarium specimens (italics). (K) Kew herbariumand spirit collection, (AAU) Aarhus University herbarium and spirit collection. Character states: (0) 6–12 stamens per flower = oligandrous flowers, (1) 15–60 stamens = mod-erately polyandrous flowers, (2) 61–300 stamens = highly polyandrous flowers.

Species Stamen number References and/or specimens Character state

Adonidia maturbongsii 26–43 Heatubun 906 (K) 1Adonidia merrillii 43–55 54459 (K) 1Balaka diffusa 50 Hodel (2010) 1Balaka longirostris 25–44 Dowe (1989) 1Balaka macrocarpa 40 Hodel (2010) 1Balaka microcarpa 20–32 Dowe (1989)/Hodel (2010) 1Balaka seemanii 15–29 Dowe (1989)/Hodel (2010) 1Balaka tahitensis 50 Hodel (2010) 1Brassiophoenix schumannii 130–200 Essig (1975) 2Brassiophoenix drymophloeoides 50–100 Burret (1935) 2Carpentaria acuminata 30–40 Dowe (2010)/46653 (K) 1Drymophloeus litigiosus 24–32 Zona (1999) 1Drymophloeus oliviformis 30–66 Zona (1999) 1Normanbya normanbyi 24–40 Dowe (2010) 1Ponapea ledermanniana 110–120 Essig (1978)/64789 (K) 2Ptychococcus lepidotus 69–138 Zona (2005) 2Ptychococcus paradoxus 100–213 Zona (2005)/383 (AAU) 2Ptychosperma burretianum 30–35 Essig (1978) 1Ptychosperma caryotoides 14–29 Essig (1978) 1Ptychosperma cuneatum 24–30 Essig (1978) 1Ptychosperma elegans 12–30 Essig (1978)/Dowe (2010)/48132 (K) 1Ptychosperma furcatum 30 Essig (1978) 1Ptychosperma gracile 20–34 Essig (1978)/Gardiner 298 (K) 1Ptychosperma halmaherense 16 Heatubun (2011) 1Ptychosperma lauterbachii 24–44 Essig (1978) 1Ptychosperma lineare 15–32 Essig (1978)/63986 (K) 1Ptychosperma macarthurii 23–40 Essig (1978)/Dowe (2010)/41437 (AAU) 1Ptychosperma microcarpum 18–30 Essig (1978) 1Ptychosperma propinquum 33–50 Essig (1978) 1Ptychosperma pulleni 20 Essig (1978) 1Ptychosperma salomomense 20–40 Essig (1978)/Gardiner 296 (K) 1Ptychosperma sanderianum 15–25 Essig (1978) 1Ptychosperma schefferi 33–38 Essig (1978)/75021 (K) 1Ptychosperma waitianum 20–38 Essig (1978)/Gardiner 292 (K) 1Solfia samoensis 35–41 Dransfield et al. (2008)/75019 (K) 1Veitchia arecina 80–131 Zona and Fuller (1999) 2Veitchia filifera 23–46 Zona and Fuller (1999) 1Veitchia joannis 74–123 Zona and Fuller (1999) 2Veitchia metiti 106–156 Zona and Fuller (1999) 2Veitchia pachyclada 174–320 Zona (1999) 2Veitchia spiralis 85–118 Zona and Fuller (1999)/Dowe (1989) 2Veitchia subdisticha 135–219 Zona (1999) 2Veitchia vitiensis 24–36 Zona and Fuller (1999) 1Veitchia winin 42–63 Zona and Fuller (1999) 1Wodyetia bifurcata 58–72 Dowe (2010) 2New taxon Gag Island 40–46 Heatubun 1126 (K) 1

Outgroup taxaActinokentia huerlimannii 19–23 Moore (1980) 1Actinorhytis calapparia 24–33 Dransfield et al. (2008) 1Basselinia humboldtiana 6 Dransfield et al. (2008) 0Basselinia velutina 6 Dransfield et al. (2008) 0Calyptrocalyx awa 6–7 Dowe and Ferrero (2001) 0Carpoxylon macrospermum 6 Dransfield et al. (2008) 0Cyphosperma balansae 6 Moore and Uhl (1984) 0Hedyscepe canterburyana 9–12 Dransfield et al. (2008) 0Heterospathe scitula 6 Fernando (1990) 0Howea belmoreana 30–70 Dowe (2010) 1Linospadix albertisianus 10–12 Dowe (1998) 0Loxococcus rupicola 12 Dransfield et al. (2008) 0Rhopalostylis baueri 6 Dransfield et al. (2008) 0

E. Alapetite et al. / Molecular Phylogenetics and Evolution 76 (2014) 227–240 231

The alignment combining all eight markers included 6506 char-acters (6408 nucleotides positions and 98 coded indels) of which371 (5.7%) were parsimony-informative. The 50% majority-rule con-sensus tree resulting from the BI analysis, with support values fromall three analyses, is shown in Fig. 3. The monophyly of the subtribewas very strongly supported (PBS = 100%, LBS = 100%, PP = 1.00).This analysis allowed us to identify four major clades (also retrievedin the nuclear DNA-combined analysis). Two clades are stronglysupported in all three phylogenetic reconstruction methods

(Fig. 3): the ‘Veitchia clade’ including all the species of Veitchia(PBS = 100%, LBS = 100%, PP = 1.00), and the ‘Balaka clade’ includingBalaka and Solfia (PBS = 98%, LBS = 100%, PP = 1.00). Ptychococcusand Brassiophoenix are sister groups with full support (PBS = 100%,LBS = 100%, PP = 1.00), like Carpentaria and Wodyetia (PBS = 100%,LBS = 100%, PP = 1.00). Together with the two other species ofDrymophloeus and Normanbya normanbyi they form the ‘Drymophloeusclade’ with low support. The ‘Ptychosperma clade’ includes all spe-cies of the genus Ptychosperma (except P. halmaherense) and is fully

Fig. 1. 50% majority-rule consensus tree resulting from the Bayesian analysis of the nuclear DNA-combined alignment. Bootstrap supports (parsimony bootstrap support(PBS)/likelihood bootstrap support (LBS)) are indicated above branches; posterior probabilities (PP) are indicated below branches. Thick lines highlight well-supported clades(PBS and LBS > 80%, and PP > 0.95).

232 E. Alapetite et al. / Molecular Phylogenetics and Evolution 76 (2014) 227–240

supported in the ML and BI analyses (LBS = 100%, PP = 1.00), andmoderately in the MP analysis (PBS = 58%). One difference betweenthe nuclear-DNA analysis and the total evidence analysis concernsthe monophyly of Adonidia maturbongsii. In the nuclear-DNA analy-sis, Adonidia and the newly discovered species Ptychospermahalmaherense form a monophyletic group. In the total evidenceanalysis, A. maturbongsii is sister to P. halmaherense and A. merrilliiis sister to the new taxon from Gag Island, however with no supportin the MP analysis and moderate support in the ML analysis.Ponapea ledermanniana is sister to the [‘Veitchia clade’ + ‘Balakaclade’ + Adonidia + new taxon] group, however with low support.

The ‘Veitchia clade’ and the ‘Balaka clade’ are sister groups withalmost maximum support (PBS = 98%, LBS = 99%, PP = 1.00).

3.3. Stamen number optimization

The optimization of stamen number evolution on the total evi-dence analysis Bayesian tree (50% majority-rule consensus tree) isshown in Fig. 4. All the species included in the study have polyan-drous flowers (more than 6 stamens). However some are moder-ately polyandrous (between 15 and 60 stamens; in gray in Fig. 4)and some are highly polyandrous (between 61 and 300 stamens;

Fig. 2. 50% majority-rule consensus tree resulting from the Bayesian analysis of the plastid DNA-combined alignment. Bootstrap supports are indicated above branches(parsimony bootstrap support/likelihood bootstrap support); posterior probabilities are indicated below branches.

E. Alapetite et al. / Molecular Phylogenetics and Evolution 76 (2014) 227–240 233

in black in Fig. 4). The character optimization suggests that thecommon ancestor of the subtribe had probably moderately polyan-drous flowers, and that transitions towards a higher number of sta-mens occurred repeatedly during the evolutionary history of thesubtribe. One transition occurred in the branch leading to Ponapealedermanniana, one in Wodyetia bifurcata and one in the commonancestor of Ptychococcus and Brassiophoenix. Transitions betweena moderate and a high number of stamens also occurred withinthe ‘Veitchia clade’. The ancestral state in this clade is ambiguous,because two scenarios are equally parsimonious (with three steps)to account for the observed distribution of character states. Thefirst one implies three independent transitions towards a highnumber of stamens, one in the common ancestor of [Veitchiaarecina + V. joannis + V. spiralis], one in V. metiti and one in the

common ancestor of V. subdisticha and V. pachyclada. The secondimplies only one transition in the common ancestor of the wholeclade except V. filifera and two reversals in Veitchia winin andVeitchia vitiensis.

4. Discussion

4.1. Nuclear vs. plastid markers

Phylogenetic relationships within subtribe Ptychospermatinaehave been explored recently using two nuclear markers, PRK andRPB2 (Zona et al., 2011), including a sampling of 37 species (amongthe 60 recognized in the subtribe) and 1640 molecular characters

Fig. 3. 50% majority-rule consensus tree resulting from the Bayesian analysis of the total evidence dataset. Bootstrap supports are indicated above branches (parsimonybootstrap support (PBS)/likelihood bootstrap support (LBS)); posterior probabilities (PP) are indicated below branches. Thick lines represent well-supported clades (PBS andLBS > 80%, and PP > 0.95).

234 E. Alapetite et al. / Molecular Phylogenetics and Evolution 76 (2014) 227–240

plus 46 coded gaps (123 characters were parsimony-informative).In spite of this effort, the phylogeny of Zona et al. (2011) remainedpartly unresolved, in particular concerning relationships within thegenera Ptychosperma and Veitchia. In our study, all selected regionshad the high variation rate expected. The non-coding plastid mark-ers included very little informative variation, with less than 2% par-simony informative sites overall (Table 1). As an example, only onenode was strongly supported (PBS and LBS > 80% and PP > 0.95) inthe tree based on matK (Fig. S6 [Supplementary material]). Theslow evolutionary rate of plastid DNA within Ptychospermatinaeis consistent with what is found in the whole palm family

(Asmussen and Chase, 2001). The level of parsimony informativevariation in the chosen nuclear regions varied between 6.3%(AG1) and 24.3% (ITS2) and was therefore much higher than inplastid alignments. It can be noted that the variation rate of AG1observed for Ptychospermatinae was similar to the one foundpreviously in Bactridinae (Arecoideae; Ludena et al., 2011). Thenumber of well-supported nodes in trees based on individual nu-clear markers varied between two and seven, showing that nuclearmarkers does bring useful information for phylogenetic recon-struction in the case of Ptychospermatinae. The use of ITS markercan be challenging due to possible amplification difficulty and

Fig. 4. 50% majority-rule consensus tree resulting from the Bayesian analysis of the total evidence dataset (cf. Fig. 3) showing the evolution of stamen number optimizedusing Maximum Parsimony. Boxes at the tip of branches are colored according to the character state: White = 6 or 12 stamens (oligandrous flowers), Gray = 15–60 stamens(moderately polyandrous flowers), Black = 61–300 stamens (highly polyandrous flowers). Branch colors correspond to the inferred ancestral sate. Several colors on the samebranch denote ambiguity in the ancestral state.

E. Alapetite et al. / Molecular Phylogenetics and Evolution 76 (2014) 227–240 235

paralogous copies (Baker et al., 2000; Hollingsworth, 2011). In ourcase, the amplification of ITS2 was successful most of the time,except when a sequence similar to the reverse primer occurred to-wards the 3’ end of the sequence, leading to an aborted ampliconand raising the question of universality of primers at a scale as

large as seed plants (Chen et al., 2010). The presence of ITS para-logues is an important issue (Alvarez and Wendel, 2003) and canmake phylogenetic reconstruction difficult in palm (Baker et al.,2000; Lewis and Doyle, 2001). However ITS sequences recentlyattracted new interest and proved to be useful in some groups of

236 E. Alapetite et al. / Molecular Phylogenetics and Evolution 76 (2014) 227–240

palms (Eiserhardt et al., 2011; Jeanson et al., 2011). In Ptychosper-matinae, only one sequence per specimen was detected on theelectropherograms, although double nucleotide peaks occurredfor a few sites. No well-supported incongruence between phyloge-netic reconstructions based on ITS2 vs. other markers was de-tected, so we included ITS2 data in our combined analysis.However, if the use of the ITS region is planned at a large scalewithin the palm family further detailed investigations will beneeded, as suggested by Feliner and Rosselló (2007).

In spite of the little informativeness displayed by our plastidmarkers, we chose to include them in our analyses, because theplastid genome is inherited uniparentally and is not subject torecombination. It can be thus interesting to compare the evolution-ary history of plastid vs. nuclear markers, in order to identify pos-sible effect of recombination on the accuracy of species treeestimation (Posada and Crandall, 2002). In our case however, therewere no major incongruences among trees based on individualmarkers. The tree resulting from the total evidence analysis ofour seven markers is the best resolved one, with 17 strongly sup-ported nodes (Fig. 3), stressing the utility of combining varioussources of molecular characters for phylogenetic reconstruction.

4.2. Phylogenetic reconstruction

In almost all single marker and combined analyses (nuclear,plastid and total evidence), the subtribe Ptychospermatinae wasresolved as monophyletic with good support. This confirms theconclusions of previous studies, based on morphological andmolecular data (Asmussen et al., 2006; Baker et al., 2009, 2011;Norup et al., 2006; Zona, 1999b; Zona et al., 2011). Compared tothese studies, our phylogenetic hypothesis based on the total evi-dence analysis (Fig. 3) provides improved resolution of the rela-tionships among and within genera, allowing us to identify fivemajor clades within the subtribe: the ‘Veitchia clade’, the ‘Balakaclade’, the ‘Adonidia clade’, the ‘Drymophloeus clade’ and the‘Ptychosperma clade’. The names were chosen as to match asclosely as possible those used in Zona et al. (2011), in order toallow easy comparison. Each clade is discussed below in light ofthe morphological characteristics and geographical range of thespecies included. It should be noted that phylogenetic relation-ships among the outgroups (all belonging to the tribe Areceae)were not fully resolved.

4.2.1. The Veitchia cladeThis well-supported clade comprises all nine Veitchia species of

our sampling (out of eleven recognized in the genus according toDransfield et al., 2008 and Zona et al., 2011). These species arecharacterized by an endocarp bearing a single flattened ridge onone side (Zona, 1999b; Zona and Fuller, 1999). Within Veitchia,two well-supported clades emerge. One includes the four endemicspecies from Vanuatu (V. arecina, V. metiti, V. spiralis, V. winin) andone species from Fiji and Tonga (V. joannis). The second includestwo species from Solomon Islands (V. pachyclada and V. subdisti-cha). Veitchia filifera appears as sister group to all other Veitchiaspecies. The placement of this Fijian species is consistent with ear-lier inferences (Zona et al., 2011) that Fiji is the ancestral area ofthe genus. Both species of the Solomon Islands, V. pachyclada andV. subdisticha were formerly classified in the genus Drymophloeus(and prior to that in the genus Rehderophoenix). The monophylyof Veitchia is strongly supported in our MP, ML and BI analysesas in the previous study (Zona et al., 2011). Several shared morpho-logical features are consistent with this molecular result. All thesespecies are emergent palms, moderate to tall, they have inflores-cences branched to 3 or 4 orders, caducous prophyll and peduncu-lar bract, and pistillode as long as the stamens (Zona, 1999b;

Dransfield et al., 2008). Moreover, the Solomon Islands are closeto the geographical range of other species of Veitchia (Vanuatuand Fiji). However some features characterize V. pachyclada andV. subdisticha, especially the shape of leaf segments and prophyllsplitting (Zona, 1999b).

Unfortunately we were not able to include V. lepidota in ouranalysis (the material was not available), a species formerly knownas Drymophloeus lepidotus and recombined in Veitchia by Zona et al.(2011). It would have been interesting to verify its phylogeneticposition using our set of markers. However the geographical range,the morphology and the ecology of this species support its rela-tionship with V. pachyclada and V. subdisticha (Zona, 1999b).

4.2.2. The Balaka cladeThis clade is resolved as sister to the Veitchia clade with full sup-

port. These two clades comprise species with the easternmost dis-tribution within Ptychospermatinae, reaching the Tonga andSamoa islands. Our analysis confirms that Solfia samoensis is re-lated to Balaka, as found by Zona et al. (2011), and suggests fur-thermore that the species is embedded with the latter genus. TheBayesian Inference analysis strongly supported a sister group rela-tionship between the monotypic genus Solfia and Balaka tahitensis,which is congruent with the distribution of these two endemic taxain the Samoa Islands. Balaka and Solfia are usually distinguished bythe endocarp shape, angular with a rostrum for Balaka vs. tereteand rounded for Solfia (Zona, 1999a; Dransfield et al., 2008). How-ever, in Balaka tahitensis ridges on the endocarp are only slightlyangled and rostrum is absent (Hodel, 2010), making it more similarto the endocarp of Solfia samoensis and supporting the close rela-tionship between both species. The other species of Balaka, all fromthe Fiji Islands (Viti Levu or Vanua Levu), form a well-supportedclade. Balaka diffusa was formerly considered as a subspecies ofBalaka macrocarpa (Moore, 1980; Fuller, 1998) but Hodel (2010) in-stated the species on the basis of differences in geographical range(Balaka macrocarpaBalaka diffusa is from Viti Levu and Balakamacrocarpa is from Vanua Levu) and length of inflorescence(has smaller and more compact inflorescences).

4.2.3. The Adonidia cladeAdonidia was reinstated as a separate genus (from Veitchia) by

Zona (1999a) and this was confirmed by Zona et al. (2011). Bothspecies of Adonidia were related in the study of Zona et al.(2011). They display strong similarities in their inflorescencesand infructescences (white inflorescences branched up to fourorders and red fruits; Baker and Heatubun, 2012). In contrast,our total evidence analysis suggests that Adonidia is paraphyletic,due to the placement of A. maturbongsii as sister (together withPtychosperma halmaherense), to the Balaka and Veitchia clades.The relationship is however poorly supported. The newly discoveredtaxon from Gag Island (Indonesia) appears related to A. merrillii,but with little support. Interestingly, all species of Adonidia andPtychosperma halmaherense form a monophyletic group in the nu-clear DNA-combined analysis, although with low support. Adonidiamerrillii is the only species of Ptychospermatinae occurring to thewest of Wallace’s Line, an important biogeographic interface(Baker and Couvreur, 2012). Adonidia maturbongsii, Ptychospermahalmaherense and the new Gag island taxon are geographicallyclose to each other, occurring on islands at the western end ofNew Guinea and in North Moluccas. Heatubun (2011) includedP. halmaherense from Halmahera (North Moluccas) in the genusPtychosperma (subgenus Ptychosperma) based on floral characters,such as the pistillode equalling or exceeding stamens in length,small fruit, seed with ruminate endosperm, solitary stem and inflo-rescence with upper peduncular bract reduced to inconspicuousstubs. However this species differs from all other species of the

E. Alapetite et al. / Molecular Phylogenetics and Evolution 76 (2014) 227–240 237

genus Ptychosperma by its seeds rounded in cross-section (vs.angular in the rest of Ptychosperma), lacking grooves or angles.Based on our molecular phylogeny and the fact that thisP. halmaherense is geographically distant from most other speciesof Ptychosperma, morphology should be re-examined and theaffiliation to Ptychosperma or Adonidia tested again usingadditional molecular markers.

In our analysis, Ponapea ledermanniana is sister to the [‘Veitchiaclade’ + ‘Balaka clade’ + Adonidia + new taxon] group. This speciesbelongs to the reinstated genus Ponapea (Dransfield et al., 2008;Zona et al., 2011), which includes three species formerly consid-ered as belonging to Ptychosperma and one to Drymophloeus. InZona et al. (2011), Ponapea was sister group to Ptychosperma. Inour study, the position of Ponapea ledermanniana is highly variableaccording to the marker. AG1 and RPB2 are in favor of the inclusionof Ponapea within a [‘Veitchia clade’ + ‘Balaka clade’ + Adoni-dia + new taxa] group (Supplementary Fig. 4 and 5) whereas PRKplaces Ponapea, together with Normanbya, as sister to Ptychosperma(Fig. S3 [Supplementary material]). In both cases, these relation-ships are poorly supported.

4.2.4. The Drymophloeus cladeThe two species of Drymophloeus, D. oliviformis and D. litigiosus,

are grouped together in the total evidence tree with strong sup-port. They are distributed in New Guinea and the Moluccas (Zona,1999b). They are both understory palms and share severalmorphological features like stilt roots, persistent prophyll andpeduncular bract (Zona, 1999b). Previous molecular studies hadsuggested a relationship (although moderately supported)between Ptychococcus and Brassiophoenix (Asmussen et al., 2006;Baker et al., 2011; Norup et al., 2006; Zona et al., 2011). Our studyconfirms that these two genera are monophyletic and closely re-lated with strong support. After taxonomic revision, the numberof species in Ptychococcus was reduced to two and the geographicalrange restricted to New Guinea including the Bismarck Archipela-gos (Zona, 2005). Brassiophoenix includes two species also endemicto New Guinea (Zona and Essig, 1999). Both genera share a stronglyridged endocarp (they differ by the color, black or brown forPtychococcus, straw-colored for Brassiophoenix) and have the samedistribution area. Our analysis suggests with moderate supportthat Drymophloeus is related to Ptychococcus and Brassiophoenix(the relationship is however fully supported in the Bayesian anal-ysis). All these species share a common geographical range (NewGuinea and surrounding islands). They are understorey to mid-storey palms of small to moderate size. Brassiophoenix andDrymophloeus have similar leaflet forms (Dransfield et al., 2008).It should be noted that plastid markers support an alternativehypothesis, with Brassiophoenix and Ptychococcus forming a cladewith Adonidia merrillii and the new taxon from Gag. Our analysisconfirms with strong support the sister group relationshipbetween the two monotypic genera Carpentaria and Wodyetiasuggested by Zona et al. (2011). Both genera are endemic toNorthern Australia (Northern Territory and north-east Queenslandrespectively; Dowe, 2010). They have flat, strongly forking fibro-vascular bundles in the endocarp, a ring of fibrovascular bundlesin the mesocarp (Irvine, 1983), and greenish inflorescence axes(Zona, 1999a), contrary to the white color found in all othermembers of the subtribe. In other morphological features, Wodye-tia is more similar to Normanbya, another monotypic generaendemic to north-east Queensland, but in Normanbya the bundlesof the fruit are purely fibrous (Irvine, 1983), unlike the fruit ofWodyetia and Carpenteria. In our total evidence analysis, Normanbyais sister to the rest of the Drymophloeus clade, but it should be notedthat its position within the phylogeny varies among single markeranalyses (see Figs. S2, S3, S4 and S7 [Supplementary material]).

4.2.5. The Ptychosperma cladePtychosperma, the largest genus of Ptychospermatinae, is re-

solved as monophyletic (with the exception of the recently de-scribed species P. halmaherense, discussed above) with strongsupport in the ML and Bayesian analysis (LBS = 100% andPP = 1.00) but low support in the MP analysis (PBS = 60%). Histori-cally the genus was divided in four subgenera: Ponapea, Korora(both included in the reinstated genus Ponapea; Dransfield,2008), Ptychosperma and Actinophloeus. This latter subgenus hadbeen first considered as a subgenus of Drymophloeus in 1877 byOdoardo Beccari, later as a separate genus, named Actinophloeusand finally as a subgenus of Ptychosperma in 1935 by the sameauthor (Essig, 1978). The genus Ptychosperma is mostly distributedin Papua New Guinea, although there are several species distrib-uted elsewhere, like P. salomonense, found in the Solomon Islands,P. elegans and P. macarthurii occurring in northern Australia,P. gracile from the Bismarck Archipelago, and P. propinquumdistributed in the Moluccas (Essig, 1978). Our results are in agree-ment with the two subgenera defined by Essig (1978). Ptychospermasalomonense, P. gracile, P. elegans and P. caryotoides, the four membersof the subgenus Ptychosperma form a well supported clade. Theremaining species are grouped in another clade (but with lowsupport) corresponding to the subgenus Actinophloeus. These twosubgenera are morphologically differentiated by a number of subtlecharacters in fruit and seed (Essig, 1978). For example, the subgenusPtychosperma has seeds with ruminate endosperm, angular orbroadly rounded seed lobes, and fruits with tanniferous mesocarpwhereas the subgenus Actinophloeus has homogeneous endosperm,squarish seed lobes and non-tanniniferous mesocarp. The subgenusActinophloeus was divided by Essig (1978) into two sections, Actino-phloeus (stems always solitary) and Caespitosa (stems usually caespi-tose). Our results do not support this division, suggesting that thecaespitose vs. solitary habit may be homoplasious.

4.3. Future prospects for the phylogeny of Ptychospermatinae

In spite of the improved resolution of phylogenetic relationshipswithin Ptychospermatinae in our study, some relationships remainunresolved. It is notably the case within the genus Ptychosperma.The 50% majority-rule tree resulting from the total evidenceBayesian analysis (Fig. 5) shows that within this genus, branchesare very short, revealing a very low rate of variation. Generic andspecies limits in subtribe Ptychospermatinae are often problematicas evidenced by the frequent taxonomic changes that have occurredin the past (see Dransfield et al., 2008 for an overview). This raisesthe question of how species should be delimited within somegenera, Ptychosperma in particular. The pattern observed withinthe genus could be due to different factors, like recent and rapidspeciation events, or very slow evolution (generation time, evolu-tionary history, selection; Felsenstein, 2004), strong hybridizationthat led to homogenous sequences or confusion between popula-tions and species. Population genetics approaches could bringnew insights into phylogenetic relationships, species limits andcongruence between markers (Linder and Rieseberg, 2004) withinthe different clades identified in Ptychospermatinae (inPtychosperma in particular). Microsatellite loci have been isolatedand characterized for several palm species (e.g. Couvreur et al.,2006; Martinez et al., 2002; Rodrigues et al., 2004), providingpotential markers for phylogenetic reconstruction among closelyrelated species of Ptychospermatinae, together with the use ofcoalescent methods.

4.4. Stamen number optimization

Within Ptychospermatinae, variation in stamen number isamazingly high (Table 2) at all levels: interspecific, intraspecific

Fig. 5. Maximum-likelihood topology obtained from the total evidence analysis, showing branch lengths.

238 E. Alapetite et al. / Molecular Phylogenetics and Evolution 76 (2014) 227–240

and even intra-individual. This makes stamen number optimiza-tion a challenge. We chose to use discrete optimization in orderto avoid calculating a mean needed in continuous optimization,and thus to avoid losing information on variation range. Moreoverwe tried to reduce as much as possible overlapping between cate-gories. According to our coding, the ancestor of the subtribe hadprobably moderately polyandrous flowers (15–60 stamens) andseveral transitions toward a higher number of stamens occurredrepeatedly within the subtribe (Fig. 4). Transitions from moderatepolyandry towards high polyandry occurred at least four times. Intwo cases, within the ‘Veitchia clade’ and the ‘Drymophloeus clade’,high polyandry is a synapomorphy that supports the monophyly ofgroups. Androecium development was studied in details inPtychosperma gracile (Uhl, 1976). Like in most palms species withpolyandrous flowers, the androecium is basically trimerous (Uhland Moore, 1980). When stamens initiation starts, the apex isdivided in six primary primordia, three opposite each sepal andthree opposite each petal. Stamens arise as secondary primordia

in alternating antesepalous and antepetalous whorls. The numberof secondary primordia is always one in antesepalous whorls butcan vary between one and three (or more) in antepetalous whorls.Variation in stamen number between flowers of the same speciesoccurs in two ways. The lower antepetalous whorls may have morestamens opposite one petal or more frequently, the number of pri-mordia in innermost whorls is slightly irregular. The ancestral statefor the palm flower is six stamens, three in an antesepalous whorland three in an antepetalous whorl. Thus in Ptychospermatinae,polyandry results from the increase of whorl number and the in-crease of primordia number within each antepetalous whorl. Inpalms polyandry mostly coincides with unisexual flowers (Nadotet al., 2011) which is the condition found in all Ptychospermatinaespecies. Transition to unisexuality, with reduced and unfertile fe-male organs (pistillode), could have allowed space for increasingthe number of stamen whorls and thus the number of stamens.

The reason why the number of stamens progressively increasedwithin the subtribe is unclear. Like many other flower features, for

E. Alapetite et al. / Molecular Phylogenetics and Evolution 76 (2014) 227–240 239

example symmetry (Jabbour et al., 2009) or nectar production(Barrera and Nobel, 2004), stamen number may have evolved ina plant–pollinator interaction context. At the angiosperm level,polyandry is often observed in beetle-pollinated species(Bernhardt, 2000). Pollination biology has been well studied insome members of the palm family (see Barfod et al., 2011 for areview), but in Ptychospermatinae pollination data are scarce.Ptychosperma is believed to be visited by various insects such asflies, syrphids (Diptera) and bees (Hymenoptera; Essig, 1973),and Normanbya normanbyi is visited by thrips (Thysanoptera), fliesand weevils (Coleoptera; Kitching et al., 2007). Polyandrous flow-ers in palms are thus not only visited by beetles but by many otherinsects. It has been suggested that more stamens is a way to producemore pollen in response to the feeding of the pollinators (Uhl andMoore, 1977). A potential correlation between stamen number andpollen production has been demonstrated within the Arecoideae,and especially within Areceae (Alapetite, unpublished). Whetherthe evolutionary pattern observed for stamen number results fromadaptation to pollination mode or results from a simple relaxationof developmental constraints remains to be tested.

5. Conclusions

We investigated the evolution of stamen number within thesubtribe Ptychospermatinae using a new molecular phylogeny ofthe subtribe. The combined use of five nuclear and three plastidmarkers improved the resolution of phylogenetic relationships be-tween genera and species compared to previous studies. Furtherphylogenetic or population genetics studies that incorporate morevariable molecular markers (e.g. microsatellite loci) would be help-ful to further clarify the circumscription of taxonomic subdivisions.Nevertheless, our work provides a useful framework for studies onmorphological and ecological characters. In this study, we exam-ined the evolution of stamen number, a strikingly variable floralfeature of Ptychospermatinae. We identified several transitions be-tween moderate and high polyandry, raising the issue of the adap-tive significance of polyandry on the one hand, and variation in thedegree of polyandry on the other hand. Studies of visitors and poll-inators should provide information on a possible correlation be-tween stamen number and a specific kind of insects (taxonomicor functional group). Studying the evolutionary history of stamennumber within palms could help us to investigate if the co-evolu-tion with insect pollinators could have driven the modification ofthis character at a larger scale.

Acknowledgements

The authors are grateful to Lauren Gardiner, Felix Forest, EdithKapinos and Laszlo Csiba (Royal Botanic Gardens, Kew), Charlie D.Heatubun (Universitas Negeri Papua, Indonesia), Carl E. Lewis andHillary Burgess (Fairchild Tropical Botanic Garden), Anders Barfod(Aarhus University herbarium), Eric Joly and Denis Larpin (MuséumNational d’Histoire Naturelle, Paris, France), Larry Noblick (Mont-gomery Botanical Center), Nura Abdul Karim (Singapore BotanicGardens) and Dick Watling (Fiji) all of whom provided plant mate-rial. SN received a research grant from Aarhus University to conductpart of this research. EA received financial support from the Centrede Recherche sur la Paléobiodiversité et les Paléoenvironnements(UMR 7207 CNRS – MNHN – UPMC) to sample material in Kew,for which Jean-Yves Dubuisson is gratefully acknowledged.

Appendix A. Supplementary material

Supplementary data associated with this article can be found, inthe online version, at http://dx.doi.org/10.1016/j.ympev.2014.02.026.

References

Alvarez, I., Wendel, J.F., 2003. Ribosomal ITS sequences and plant phylogeneticinference. Mol. Phylogenet. Evol. 29, 417–434.

Asmussen, C.B., Chase, M.W., 2001. Coding and noncoding plastid DNA in palmsystematics. Am. J. Bot. 88, 1103–1117.

Asmussen, C.B., Dransfield, J., Deickmann, V., Barfod, A.S., Pintaud, J.-C., Baker, W.J.,2006. A new subfamily classification of the palm family (Arecaceae): evidencefrom plastid DNA phylogeny. Bot. J. Linn. Soc. 151, 15–38.

Bacon, C.D., Feltus, F.A., Paterson, A.H., Bailey, C.D., 2007. Novel nuclear intron-spanning primers for Arecaceae evolutionary biology. Mol. Ecol. Notes 8, 211–214.

Baker, W.J., Couvreur, T.L.P., 2012. Global biogeography and diversification of palmssheds light on the evolution of tropical lineages. II. Diversification history andorigin of regional assemblages. J. Biogeogr. 40, 286–298.

Baker, W.J., Heatubun, C.D., 2012. New palms from Biak and Supiori, Western NewGuinea. Palms 56, 131–150.

Baker, W.J., Hedderson, T.A., Dransfield, J., 2000. Molecular phylogenetics ofsubfamily Calamoideae (Palmae) based on nrDNA ITS and cpDNA rps16 intronsequence data. Mol. Phylogenet. Evol. 14, 195–217.

Baker, W.J., Norup, M.V., Clarkson, J.J., Couvreur, T.L.P., Dowe, J.L., Lewis, C.E.,Pintaud, J.-C., Savolainen, V., Wilmot, T., Chase, M.W., 2011. Phylogeneticrelationships among arecoid palms (Arecaceae: Arecoideae). Ann. Bot. 108,1417–1432.

Baker, W.J., Savolainen, V., Asmussen-Lange, C.B., Chase, M.W., Dransfield, J., Forest,F., Harley, M.M., Uhl, N.W., Wilkinson, M., 2009. Complete generic-levelphylogenetic analyses of palms (Arecaceae) with comparisons of supertreeand supermatrix approaches. Syst. Biol. 58, 240–256.

Barfod, A.S., Hagen, M., Borchsenius, F., 2011. Twenty-five years of progress inunderstanding pollination mechanims in palms (Arecaceae). Ann. Bot. 108,1503–1516.

Barrera, E., Nobel, P.S., 2004. Nectar: properties, floral aspects, and speculation onorigin. TRENDS Plant Sci. 9, 65–69.

Bernhardt, P., 2000. Convergent evolution and adaptive radiation of beetle-pollinated angiosperms. Plant Syst. Evol. 222, 293–320.

Burret, M., 1935. Neue Palmen aus Neuguinea II. Notizblatt des Königl. botanischenGartens und Museums zu Berlin 12, 309–348.

Chen, S., Yao, H., Han, J., Liu, C., Song, J., Shi, L., et al., 2010. Validation of the ITS2region as a novel DNA barcode for identifying medicinal plant species. PLoS ONE5, e8613.

Couvreur, T.L.P., Billote, N., Risterucci, A.-M., Lara, C., et al., 2006. Close geneticproximity between cultivated and wild Bactris gasipaes Kunth revealed bymicrosatellite markers in Western Ecuador. Genet. Resour. Crop Evol. 53, 1361–1373.

Cuenca, A., Asmussen-Lange, C.B., 2007. Phylogeny of the palm tribe Chamaedoreeae(Arecaceae) based on plastid DNA sequences. Syst. Bot. 32, 250–263.

Darriba, D., Taboada, G.L., Doallo, R., Posada, D., 2012. JModelTest 2: more models,new heuristics and parallel computing. Nat. Methods 9, 772.

Dowe, J.L., 1989. Palms of the South-west Pacific: Their Origin Distribution andDescription. Palm and Cycad Societies of Australia, 198p.

Dowe, J.L., 1998. Palms and Cycads. Palm and Cycad Societies of Australia, p. 58.Dowe, J.L., 2010. Australian Palms: Biogeography, Ecology and Systematics. CSIRO

PUBLISHING.Dowe, J.L., Ferrero, M.D., 2001. Revision of Calyptrocalyx and the New Guinea

species of Linospadix (Linospadicinae: Arecoideae: Arecaceae). Blumea 46, 207–251.

Dransfield, J., Uhl, N.W., Asmussen, C.B., Baker, W.J., Harley, M.M., Lewis, C.E., 2008.Genera Palmarum: The Evolution and Classification of Palms. Kew Publishing,Royal Botanic Gardens, Kew.

Eiserhardt, W.L., Pintaud, J., Asmussen-Lange, C., Hahn, W., Bernal, R., Balslev, H.,Borchsenius, F., 2011. A phylogeny of Bactridinae (Arecaceae) based on plastid05 and nuclear DNA sequences. Taxon 60, 485–498.

Endress, P.K., 2011. Evolutionary diversification of the flowers in angiosperms. Am.J. Bot. 98, 370–396.

Essig, F.B., 1973. Pollination of some New Guinea palms. Principes 17, 75–83.Essig, F.B., 1975. Brassiophoenix schumannii (Palmae). Principes 19, 100–103.Essig, F.B., 1978. A revision of the genus Ptychosperma Labill. (Arecaceae). Allertonia

1, 415–478.Feliner, G.N., Rosselló, J., 2007. Better the devil you know? Guidelines for insightful

utilization of nrDNA ITS in species-level evolutionary studies in plants. Mol.Phylogenet. Evol. 44, 911–919.

Felsenstein, J., 2004. Inferring Phylogenies. Sinauer Associates Inc., U.S., 580p.Fernando, E.S., 1990. The genus Heterospathe (Palmae: Arecoideae) in the

Philippines. Kew. Bull. 45, 219–234.Fitch, W.M., 1971. Toward defining the course of evolution: minimum change for a

specific topology. Syst. Zool. 20, 406–416.Fuller, D., 1998. Balaka macrocarpa. In: IUCN 2012. IUCN Red List of Threatened

Species. Version 2012.2. <www.iucnredlist.org>.Gaut, B.S., Morton, B.R., McCaig, B.C., Clegg, M.T., 1996. Substitution rate

comparisons between grasses and palms: synonymous rate differences at thenuclear gene Adh parallel rate differences at the plastid gene rbcL. Proc. Natl.Acad. Sci. USA 93, 10274–10279.

Guindon, S., Lethiec, F., Duroux, P., Gascuel, O., 2005. PHYML Online-a web serverfor fast maximum likelihood-based phylogenetic inference. Nucl. Acids Res. 33,W557–W559.

240 E. Alapetite et al. / Molecular Phylogenetics and Evolution 76 (2014) 227–240

Hahn, W.J., 2002. A molecular phylogenetic study of the palmae (Arecaceae) basedon atpB, rbcL, and 18S nrDNA sequences. Syst. Biol. 51, 92–112.

Heatubun, C.D., 2011. A new species of Ptychosperma from Halmahera, NorthMoluccas. Palms 55, 183–189.

Hodel, D.R., 2010. A synopsis of the genus Balaka. Palms 54, 161–188.Hollingsworth, P.M., 2011. Refining the DNA barcode for land plants. Proc. Natl.

Acad. Sci. USA 108, 19451–19452.Hughes, C.E., Eastwood, R.J., Bailey, C.D., 2006. From famine to feast? Selecting

nuclear DNA sequence loci for plant species-level phylogeny reconstruction.Philos. Trans. Roy. Soc. B 361, 211–225.

Irvine, A., 1983. Wodyetia, a new arecoid genus from Australia. Principes 27, 158–167.

Jabbour, F., Nadot, S., Damerval, C., 2009. Evolution of floral symmetry: a state of theart. Comptes-Rendus Biol. 332, 219–231.

Jeanson, M.L., Labat, J.-N., Little, D.P., 2011. DNA barcoding: a new tool for palmtaxonomists? Ann. Bot. 108, 1445–1451.

Kitching, R.L., Boulter, S.L., Howlett, B.G., Goodall, K., 2007. Visitor assemblages atflowers in a tropical rain forest canopy. Austral Ecol. 32, 29–42.

Kluge, A.G., 1989. A concern for evidence and a phylogenetic hypothesis ofrelationships among Epicrates (Boidae, Serpentes). Syst. Biol. 38, 7–25.

Kluge, A.G., Wolf, J., 1993. Cladistics: what’s in a word? Cladistics 9, 183–199.Lewis, C.E., Doyle, J.J., 2001. Phylogenetic utility of the nuclear gene malate synthase

in the palm family (Arecaceae). Mol. Phylogenet. Evol. 19, 409–420.Lewis, C.E., Doyle, J.J., 2002. A phylogenetic analysis of tribe Areceae (Arecaceae)

using two low-copy nuclear genes. Plant Syst. Evol. 236, 1–17.Linder, C.R., Rieseberg, L.H., 2004. Reconstructing patterns of reticulate evolution of

plants. Am. J. Bot. 91, 1700–1708.Ludena, B., Chabrillange, N., Aberlenc-Bertossi, F., Adam, H., Tregear, J.W., Pintaud,

J.-C., 2011. Phylogenetic utility of the nuclear genes AGAMOUS 1 andPHYTOCHROME B in palms (Arecaceae): an example within Bactridinae. Ann.Bot. 108, 1433–1444.

Maddison, W.P., Maddison, D.R., 2011. Mesquite: a modular system for evolutionaryanalysis. Version 2.75 <http://mesquiteproject.org>.

Martinez, A.K., Gaitan-Solis, E., Duque, M.C., Bernal, R., Tohme, J., 2002.Microsatellite loci in Batris gasipaes (Arecaceae): their isolation andcharacterization. Mol. Ecol. Notes 2, 408–410.

Moore, H.E., 1980. New genera and species of Palmae from New Caledonia, II.Gentes Herbarum 12, 17–24.

Moore, H.E., Uhl, N.W., 1984. The indigenous palms of New Caledonia. Allertonia 3,313–402.

Müller, K., 2005. SeqState – primer design and sequence statistics for phylogeneticDNA data sets. Appl. Bioinform. 4, 65–69.

Nadot, S., Sannier, J., Barfod, A.S., Baker, W.J., 2011. Evolution of the palmandroecium as revealed by character mapping on a supertree. In: Wanntorp,L., Ronce de Craene, L.P. (Eds.), Flowers on the Tree of Life. Cambridge UniversityPress, pp. 156–180.

Nixon, K.C., 1999. The parsimony ratchet, a new method for rapid parsimonyanalysis. Cladistics 15, 407–414.

Nixon, K.C., Carpenter, J.M., 1996. On simultaneous analysis. Cladistics 12, 221–241.Norup, M.V., Dransfield, J., Chase, M.W., Barfod, A.S., Fernando, E.S., Baker, W.B.,

2006. Homoplasious character combinations and generic delimitations: a casestudy from the Indo-Pacific arecoid palms (Arecaceae: Areceae). Am. J. Bot. 93,1065–1080.

Palmweb, 2013. Palmweb: Palms of the World Online. <http://www.palmweb.org/>.Posada, D., Crandall, K.A., 2002. The effect of recombination on the accuracy of

phylogeny estimation. J. Mol. Evol. 54, 396–402.Rodrigues, D.P., Vinson, C., Ciampi, A.Y., Farias, I.P., Lemes, M.R., et al., 2004. Novel

microsatellite markers for Bactris gasipaes (Palmae). Mol. Ecol. Notes 4, 575–576.

Roncal, J., Zona, S., Lewis, C.E., 2008. Molecular phylogenetic studies of carribeanpalms (Arecaceae) and their relationships to biogeography and conservation.Bot. Rev. 74, 78–102.

Ronquist, F., Teslenko, M., van der Mark, P., Ayres, D.L., Darling, A., Höhna, S., Larget,B., Liu, L., Suchard, M.A., Huelsenbeck, J.P., 2012. MrBayes 3.2: Efficient Bayesianphylogenetic inference and model choice across a large model space. Syst. Biol.61, 539–542.

Scarcelli, N., Barnaud, A., Eiserhardt, W., Treier, U.A., Seveno, M., d’Anfray, A.,Vigouroux, Y., Pintaud, J.-C., 2011. A set of 100 chloroplast DNA primer pairs tostudy population genetics and phyogeny in Monocotyledons. PLoS ONE 6,e19954.

Sikes, D.S., Lewis, P.O., PAUPRat: PAUP implementation of the parsimony ratchet. In:Beta software, v., distributed by the authors, (Ed.), Department of Ecology andEvolutionary Biology, University of Connecticut, Storrs, USA, 2001.

Simmons, M.P., Müller, K., Norton, A.P., 2007. The relative performance of indel-coding methods in simulations. Mol. Phylogenet. Evol. 44, 724–740.

Swofford, D.L., PAUP*: Phylogenetic analysis using parsimony (⁄and other methods),beta version 4.0. In: Sinauer, (Ed.), Sunderland, MA, USA, 2002.

Tamura, K., Dudley, J., Nei, M., Kumar, S., 2007. MEGA4: molecular evolutionarygenetics analysis (MEGA) software version 4.0. Mol. Biol. Evol. 24, 1596–1599.

Uhl, N., Moore, H.E., 1980. Androecial development in six polyandrous generarepresenting five major groups of palms. Ann. Bot. 45, 57–75.

Uhl, N.W., 1976. Developmental studies in Ptychosmerpa (Palmae). II. The staminateand pistillate flowers. Am. J. Bot. 63, 97–109.

Uhl, N.W., Moore, H.E., 1977. Correlation of inflorescence, flower structure, andfloral anatomy with pollination in some palms. Biotropica 9, 170–190.

Wiens, J., Morrill, M., 2011. Missing data in phylogenetic analysis: reconcilingresults from simulations and empirical data. Syst. Biol. 60, 719–731.

Wiens, J.J., 1998. Combining data sets with different phylogenetic histories. Syst.Biol. 47, 568–581.

Wilson, M.A., Gaut, B.S., Clegg, M.T., 1990. Chloroplast DNA evolves slowly in thepalm family (Arecaceae). Mol. Biol. Evol. 7, 303–314.

Zona, S., 1999a. New perspective on generic limits and relationships in thePtychospermatinae (Palmae: Arecoideae). In: Henderson, A., Borchsenius, F.(Eds.), Evolution, Variation and Classification of Palms. The New York BotanicalGarden Press, Bronx New York, pp. 255–263.

Zona, S., 1999b. Revision of Drymophloeus (Arecaceae: Arecoideae). Blumea 44, 1–24.

Zona, S., 2005. A revision of Ptychococcus (Arecaceae). Syst. Bot. 30, 520–529.Zona, S., Essig, F.B., 1999. How many species of Brassiophoenix? Palms 43, 45–48.Zona, S., Francisco-Ortega, J., Jestrow, B., Baker, W.J., Lewis, C.E., 2011. Molecular

phylogenetics of the palm subtribe Ptychospermatinae (Arecaceae). Am. J. Bot.98, 1–11.

Zona, S., Fuller, D., 1999. A revision of Veitchia (Arecaceae–Arecoideae). Harvard Pap.Bot. 4, 543–560.

Zwickl, D.J., Hillis, D.M., 2002. Increased taxon sampling greatly reducesphylogenetic error. Syst. Biol. 51, 588–598.