How to kill two genera with one tree: clarifying generic

How to kill two genera with one tree: clarifying genericcircumscriptions in an endemic Malagasy cladeof Sapindaceae

SVEN BUERKI1,2*, PORTER P. LOWRY II3,4, SYLVIE ANDRIAMBOLOLONERA5,PETER B. PHILLIPSON3,4, LAURA VARY6 and MARTIN W. CALLMANDER3,7

1Department of Biodiversity and Conservation, Real Jardin Botánico, CSIC, Plaza de Murillo 2,28014 Madrid, Spain2Molecular Systematics Section, Jodrell Laboratory, Royal Botanic Gardens, Kew, Richmond, SurreyTW9 3DS, UK3Missouri Botanical Garden, P.O. Box 299, St Louis, MO 63166-0299, USA4Muséum National d’Histoire Naturelle, Case Postale 39, 57 rue Cuvier, 75231 05 CEDEX, Paris,France5Missouri Botanical Garden, Madagascar Research and Conservation Program, B.P. 3391,Antananarivo 101, Madagascar6Department of Ecology and Evolutionary Biology, University of California-Irvine, 321 SteinhausHall, Irvine, CA 92697, USA7Conservatoire et Jardin botaniques de la Ville de Genève, ch. de l’Impératrice 1, CH-1292Chambésy, Switzerland

Received 11 August 2010; revision 28 October 2010; accepted for publication 8 November 2010

Phylogenetic relationships in a Malagasy clade of Sapindaceae, encompassing Molinaea (with members also inthe Mascarene Islands), Neotina, Tina and Tinopsis, were inferred by expanding a previous nuclear and plastidDNA data set for the family. The circumscription of these morphologically similar genera has remained prob-lematic since the first family-wide treatment. To investigate this situation, representative taxa were analysed to:(1) test the monophyly of the genera; (2) investigate their phylogenetic relationships; and (3) explore alternativecircumscriptions that reflect phylogeny and yield genera that are morphologically coherent and easily charac-terized. Phylogenetic inferences supported the monophyly of the group and its subdivision into three clades. Allspecies of Molinaea sampled belong to a clade (Clade I) that is sister to a clade comprising Neotina, Tina andTinopsis, within which one clade (Clade II) encompasses Tinopsis and Neotina (with the latter nested within theformer) and another (Clade III) comprises all taxa of Tina. These three genera can be easily distinguished fromMolinaea by having two rather than three carpels, which represents an unambiguous synapomorphy. Given theparaphyly of Tinopsis with regard to Neotina and the strong support for the monophyly of Tina, two potentiallyviable options are available for the generic delimitation of the taxa in this clade: (1) to recognize two generacorresponding, respectively, to Clades II and III; or (2) to place all of the taxa in a single genus encompassingboth clades. Based on a review of morphological evidence the second option is favoured and consequently a broadgeneric concept is applied. © 2011 The Linnean Society of London, Botanical Journal of the Linnean Society,2011, 165, 223–234.

ADDITIONAL KEYWORDS: Cupania group – Molinaea – Neotina – phylogenetic inference – taxonomy –Tina – Tinopsis.

*Corresponding author. E-mail: [email protected]

Botanical Journal of the Linnean Society, 2011, 165, 223–234. With 2 figures

© 2011 The Linnean Society of London, Botanical Journal of the Linnean Society, 2011, 165, 223–234 223

INTRODUCTION

Recent phylogenetic analyses of Sapindaceae inferredfrom nuclear and plastid sequence data have revealeda high level of para- and polyphyly at the subfamilial,tribal and even generic levels (Harrington et al., 2005;Buerki et al., 2009a). The worldwide Sapindaceae, ascircumscribed by the Angiosperm Phylogeny Group(APG II, 2003; APG III, 2009), include c. 1900 speciesin 142 genera and four subfamilies (Xanthoceroideae,Hippocastanoideae, Dodonaeoideae and Sapindoideae)(Buerki et al., 2009a), although Buerki et al. (2010a)recently adopted a narrower family circumscriptionbased on molecular, biogeographical, dating and mor-phological evidence. To accommodate the high level oftribal para/polyphyly, a new informal infra-familialclassification was proposed by Buerki et al. (2009a),mainly based on molecular evidence, and additionalwork has been conducted to develop new genericcircumscriptions (see Buerki et al., 2010b). Thesestudies have clarified relationships at the family leveland made important advances towards an improvedclassification of Sapindaceae. They have also broughtinto focus the need for further investigations to iden-tify new synapomorphies that support the groupsdefined in the molecular analyses and that can providethe basis for developing a formal tribal classification(see Buerki et al., 2009a).

Madagascar is home to a remarkable array of mor-phological and genetic diversity within Sapindaceaeand an exceptional level of endemism (Capuron, 1969).In total, c. 100 species in 27 genera are currentlyrecognized in Madagascar, with 11 genera endemic tothe island; namely, Beguea Capuron, ChouxiaCapuron, Conchopetalum Radlk., Gereaua Buerki &Callm., Neotina Capuron, Plagioscyphus Radlk.,Pseudopteris Baill., Tina Schult., Tinopsis Radlk. andTsingya Capuron (Capuron, 1969; Buerki et al.,2010b). In the context of preparing a treatment of thefamily for the Catalogue of the Vascular Plants ofMadagascar (http://www.efloras.org/madagascar), theaim of which is to provide an authoritative taxonomicsynthesis of the Malagasy flora, an initial set ofrevisions has been conducted on several Malagasygenera (e.g. Schatz, Gereau & Lowry, 1999; Buerkiet al., 2009b, 2010b) and others are in progress. Phy-logenetic analyses have shown that most Sapindaceaepresent on this large Indian Ocean island (especiallythose in endemic genera) belong to one of two well-supported clades, referred to as the Macphersonia andCupania groups (Buerki et al., 2009a). Relationshipswithin the first of these clades, members of which aredistributed mainly in Madagascar with some taxa ineastern Africa, were recently investigated by Buerkiet al. (2010b), whereas the second group, which has awider distribution, with taxa occurring in Australasia,

South America and Madagascar (Buerki et al., 2009a),has not yet been examined in detail. Within theCupania group, four genera form a strongly supportedclade: Molinaea Comm. ex Juss. (with eight species inthe Malagasy region: five in Madagascar and three inthe Mascarene Islands), Neotina (two species), Tina(six species) and Tinopsis (11 species), the latter threeall endemic to Madagascar. These genera are closelyrelated to several South American genera, notablyCupania and Matayba Aubl. (Buerki et al., 2009a).Although the four Malagasy genera form a monophyl-etic group, their circumscription and defining charac-ters have been problematic ever since Radlkofer (1933)published the first comprehensive classification for thefamily and they have posed problems for taxonomistssince then (Capuron, 1969; Andriambololonera, 1999).

There has been considerable confusion regarding thedefinition and circumscription of these four genera (seeCapuron, 1969; Acevedo-Rodríguez, 2003), but Moli-naea can be easily distinguished morphologically frommembers of the other genera by its three-carpellategynoecium (vs. two carpels in Neotina, Tina and Tinop-sis) (Fig. 1). When Radlkofer (in Durand, 1888)described Tinopsis, based on T. apiculata Radlk., hedistinguished it from Tina on the basis of the numberof stamens (five vs. eight, respectively), whichprompted Choux (1925) to transfer Tina isoneuraRadlk. to Tinopsis as it also has five stamens, althoughit has a dehiscent fruit characteristic of Tina (Tinopsisis the only genus with indehiscent fruit; see below andTable 1). Choux (1927) and Radlkofer (1933) subse-quently changed their minds and chose to recognise asingle genus, Tina, because of the absence of strongdiscriminating morphological characters. In contrast,in his monograph of Malagasy Sapindaceae, Capuron(1969) resurrected Tinopsis (in which he describedeight new species) and described Neotina as a newgenus to accommodate Tina isoneura because of itsunique combination of fruit morphology, number ofstamens and lomatorrhizal embryo (a character sharedwith Tinopsis), whereas the embryo of Tina is notor-rhizal (see Table 1). Although these genera are mor-phologically similar in many respects, Capuron (1969)assigned them to two different tribes: Neotina and Tina(and Molinaea) were placed in Cupanieae, character-ized by a dehiscent fruit with a ceraceous (waxy),coloured (generally orange to pale red) arillode thatpartially surrounds the seed (in some cases the arillodeis somewhat obscure), whereas Tinopsis was assignedto Schleichereae, members of which have an indehis-cent fruit with a fleshy, translucent arillode surround-ing the entire seed (similar to that of the widelycultivated Litchi chinensis Sonn.; see Fig. 1 for asummary of fruit morphology).

As mentioned above, the phylogenetic analyses ofBuerki et al. (2009a, 2010a, b) were in agreement

224 S. BUERKI ET AL.

© 2011 The Linnean Society of London, Botanical Journal of the Linnean Society, 2011, 165, 223–234

A B

C D

Figure 1. A survey of fruit morphology in representative members of Molinaea, Neotina, Tina and Tinopsis. A, Tinastriata Radlk. ssp. striata (Buerki 75; photograph: S. Buerki); B, Neotina coursii Capuron (Malcomber 1293; photograph:G.E. Schatz); C, Tinopsis macrocarpa Capuron (Buerki 134; photograph: S. Buerki); D, Molinaea retusa Radlk. (Call-mander 572; photograph: M.W. Callmander).

GENERIC LIMITS IN MALAGASY SAPINDACEAE 225


with the views of Choux (1927) and Radlkofer (1933)with regard to considering Molinaea, Neotina, Tinaand Tinopsis as closely related genera. In thesemolecular studies, Molinaea was shown to be thesister lineage of the remaining genera (with Tina inturn being sister to Neotina + Tinopsis; Buerki et al.,2009a, 2010a, b). However, because these phyloge-netic analyses were based on limited sampling (justone or two exemplars per genus), they do not providea robust understanding of relationships within thisclade. In an attempt to address this deficiency, wehave expanded the data set of Buerki et al. (2010b) bysignificantly augmenting the number of taxa withinthis clade (hereafter referred to as the ingroup) inorder to: (1) test the monophyly of the four genera ascurrently defined; (2) investigate phylogenetic rela-tionships among the members of the ingroup; and (3)explore alternative circumscriptions that reflect phy-logeny and yield genera that are morphologicallycoherent and easily characterized.

MATERIAL AND METHODSSAMPLING, SEQUENCE DATA AND

PHYLOGENETIC ANALYSES

Species names, voucher information and GenBankaccession numbers for all sequences are provided inBuerki et al. (2010b) except for the taxa added for thisstudy (see Appendix), which include representativespecies of the Malagasy genera Molinaea, Neotina,Tina and Tinopsis, all of which are members of the

Cupania group (Buerki et al., 2009a). The outgroupsampling included one species of Anacardiaceae(species of Sorindeia Thou.; defined as the outgroupin all analyses; Buerki et al., 2009a) and one speciesof Simaroubaceae (Harrisonia abyssinica Oliv.). TheDNA extraction, amplification and sequencing proto-cols for the nuclear and plastid regions studied areprovided in Buerki et al. (2009a). The nuclearsequences include the entire internal transcribedspacer (ITS) region (ITS1, 5.8S and ITS2) and theplastid markers include both coding (matK and rpoB)and non-coding regions (the trnL intron and the inter-genic spacers trnD-trnT, trnK-matK, trnL-trnF andtrnS-trnG).

In earlier phylogenetic studies of Sapindaceae(Buerki et al., 2009a, 2010a, b, c), none of the moder-ately to strongly supported relationships recovered(i.e. with bootstrap support > 75%) showed incongru-ence between the single-gene analyses performed anda total evidence approach was therefore adopted. Asthe present study employs an expanded version of thesame basic data set, we have again chosen to presentthe result of our analyses based on a combined dataset, using maximum likelihood (ML) and maximumparsimony (MP) criteria, following the same procedureas in Buerki et al. (2009a). The parsimony ratchet(Nixon, 1999) was performed using PAUPrat (Sikes &Lewis, 2001). Ten independent searches were per-formed with 200 iterations and 15% of the parsimonyinformative characters perturbed. A strict consensustree was constructed based on the shortest equally

Table 1. Comparison of the Malagasy genera Tina, Neotina and Tinopsis

Tina Roemer & Schult. Neotina Capuron Tinopsis Radlk.

Tribe Cupanieae Cupanieae SchleichereaePhylogenetic

groupingCupania group Cupania group Cupania group

Leaflet Denticulate (at least in part) Entire EntirePetal scale Free Free or united Free or unitedStamens (5 or) 6–8 (or 9) 5 (6 or 7) 5 (6 or 7)Anther Subcordiform, apiculous and

glandular at the apexOblong, emarginate and

eglandular at the apexOblong, emarginate and

eglandular at the apexStigmatic line Short Well developed along the style Well developed along the styleFruit Dehiscent, splitting into two

valves that become widelyseparated

Dehiscent, splitting into twovalves that become widelyseparated

Indehiscent or incompletelysplitting into two erectvalves

Endocarp Glabrous or pubescent Glabrous GlabrousArillode Not surrounding the entire seed

(sometimes reduced orabsent), ceraceous, coloured(usually orange or pale red)

Not surrounding the entire seed(sometimes reduced orabsent), ceraceous, coloured(usually orange or pale red)

Surrounding the entire seed,fleshy, translucent

Embryo type Notorrhizal Lomatorrhizal Lomatorrhizal

The definition of tribes follows Radlkofer (1933) and the phylogenetic groupings are those of Buerki et al. (2009a).Morphological characters were adapted from Capuron (1969).



parsimonious trees. To assess support at each node,non-parametric bootstrap analyses (Felsenstein, 1985)were performed using PAUP* (Swofford, 2002) follow-ing the same procedure as in Buerki et al. (2009a). AnML analysis was performed using RAxML version7.0.0 (Stamatakis, 2006; Stamatakis, Hoover & Rouge-mont, 2008) with 1000 rapid bootstrap analyses fol-lowed by the search of the best-scoring tree in onesingle run based on the GTR + G + I model (see Buerkiet al., 2009a). These analyses were conducted using thefacilities made available by the Vital-IT portal at theSwiss Institute of Bioinformatics (Lausanne, Switzer-land; http://www.vital-it.ch/about/).

RESULTS

The ML and MP total evidence trees were highly con-gruent and revealed the same major groups of Sapin-daceae as presented in Buerki et al. (2009a, 2010a, b).The most parsimonious tree for the combined analysiswas 10 376 steps in length [consistency index (CI) =0.497 and retention index (RI) = 0.750] and the con-sensus tree was based on 1269 trees. The best ML treehad a log likelihood of -73 138.44. Statistics for eachmarker within the ingroup are provided in Table 2(statistics for the full data set are given in Buerki et al.,2009a, 2010b). As our results are congruent with thoseof earlier studies of Sapindaceae, i.e. phylogeneticrelationships and bootstrap support values (BS) aresimilar (Fig. 2A), only the ML total evidence tree isdiscussed below because it contains the maximumamount of phylogenetic information (Fig. 2). Both MLand MP analyses strongly support the monophyly ofthe ingroup and its position within the Cupania group(BS: 100; Fig. 2). Within the ingroup, three clades canbe recognized: Clade I (BS: 100) is sister to Clade II(BS: 65) + Clade III (BS: 98) (Fig. 2B). Clade I exclu-sively comprises taxa of Molinaea. The composition ofClade II suggests that Tinopsis is paraphyletic withrespect to Neotina (Fig. 2B), although relationshipsare weakly supported and might better be regarded asunresolved. All species belonging to Tina are placed inClade III, within which accessions belonging to T.striata appear to be paraphyletic with respect tocertain other members of the genus, although somenodes are weakly supported and further analyses willbe needed to confirm this finding (Fig. 2B).

DISCUSSIONRELATIONSHIPS WITHIN THE INGROUP

The phylogenetic analyses conducted in this studyusing significantly expanded ingroup sampling(including 47 specimens representative of ingroupdiversity) confirm: (1) the monophyly of the ingroup(BS: 100); (2) its placement within the Cupania group T

able

2.C

har

acte

rist

ics

ofpa

rtit

ion

su

sed

inth

eph

ylog

enet

ican

alys

es

ITS

mat

Krp

oBtr

nD

-trn

TIG

Str

nK

-mat

KIG

Str

nL

intr

ontr

nL

-trn

FIG

Str

nS

-trn

GIG

SA

llei

ght

regi

ons

Nu

mbe

rof

sequ

ence

s36

3243

3132

4646

2847

Ali

gnm

ent

len

gth

(in

clu

din

gou

tgro

up)

783

1089

357

1498

723

529

431

1374

6784

Nu

mbe

rof

con

stan

tch

arac

ters

(%)

653

1034

349

1436

691

499

413

1319

6394

(83.

4)(9

4.95

)(9

7.76

)(9

5.86

)(9

5.57

)(9

4.33

)(9

5.82

)(9

6.0)

(94.

25)

Nu

mbe

rof

vari

able

char

acte

rs(%

)13

055

862

3230

1855

390

(16.

6)(5

.05)

(2.2

4)(4

.14)

(4.4

3)(5

.67)

(4.1

8)(4

.0)

(5.7

5)N

um

ber

ofpo

ten

tial

lypa

rsim

ony-

info

rmat

ive

char

acte

rs(%

)75

204

2115

1011

2718

3(9

.58)

(1.8

4)(1

.12)

(1.4

)(2

.07)

(1.8

9)(2

.55)

(1.9

7)(2

.7)

Th

eva

lues

repo

rted

corr

espo

nd

toth

ein

grou

psa

mpl

ing

only

.S

eeB

uer

kiet

al.

(200

9a)

for

valu

esre

late

dto

the

enti

reda

tase

t.IG

S,

inte

rgen

icsp

acer

;IT

S,

inte

rnal

tran

scri

bed

spac

er.



A

Figure 2. A, best maximum likelihood phylogenetic tree inferred from eight plastid and nuclear markers summarizingrelationships within Sapindaceae.



(Fig. 2A); (3) the sister position of Molinaea (Clade I;Fig. 2B) to the other genera within the ingroup, assuggested earlier based on much more limited sam-pling (Buerki et al., 2009a); (4) the monophyly of Tina(Clade III); and (5) the close relationship betweenTinopsis and Neotina (Clade II BS: 65; previouslysuggested by Buerki et al., 2009a; Fig. 2B). Basedexclusively on phylogenetic evidence, the monophylyof both Molinaea and Tina is supported, whereas thestatus of the two other genera remains problematic.In the following discussion, we will explore the taxo-nomic implications of the paraphyly of Tinopsis(based on morphological evidence) and will attemptto propose a coherent generic treatment.

GENERIC CIRCUMSCRIPTION: RADLKOFER

VS. CAPURON

One of the goals of the present study is to improve onthe current generic level taxonomy for this group ofMalagasy Sapindaceae by proposing an alternativethat reflects evolutionary relationships and is sup-ported by easily discernable morphological features.Although Capuron (1969) resurrected Tinopsis anddescribed Neotina in an attempt to characterize thediversity exhibited by members of this group, heacknowledged that no vegetative or floral featuresunambiguously distinguished Tina from Tinopsisand/or Neotina. As a consequence, hundreds of sterile

B

Figure 2. B, phylogenetic relationships within the Malagasy Sapindaceae clade (ingroup). The South American sisterclade is also represented and used as the outgroup. Bootstrap supports are indicated above each branch. The classificationfollows Buerki et al. (2009a).



and flowering herbarium specimens remain unas-signed to genus, serving as a clear indication of theinadequacies of the current generic framework andproviding strong motivation for our efforts to improveon it. Using the phylogenetic hypotheses resultingfrom our molecular analyses, we have consideredalternatives for generic circumscriptions with particu-lar regard to potentially diagnostic morphologicalcharacters (see Table 1). Such an approach consti-tutes the first step towards understanding the evolu-tion of this Malagasy clade and will provide a basisfor further investigations focusing, for example, onspecies delimitations and patterns of morphologicalcharacter evolution.

Molinaea, which forms a clade sister to all othermembers of the ingroup (Clade I; Fig. 2B), is charac-terized by an ovary with three carpels, a feature foundin most Sapindaceae, including the South Americangenera (such as Cupania L.; Fig. 2B) that are sister tothe clade comprising the genera being studied here (i.e.the ingroup). This suggests that a reduction in thenumber of carpels took place in a common ancestor ofthe clade comprising Neotina, Tina and Tinopsis andthat this feature thus constitutes a synapomorphy forthem (Fig. 2B). Capuron (1969) hypothesized just sucha trend and he also argued that, although the Mala-gasy genera are morphologically similar to those fromSouth America (especially Cupania), they are never-theless sufficiently distinctive and geographicallyseparate to be retained. He further suggested thatTina, Tinopsis and Neotina shared a common ancestorwith Molinaea, a hypothesis that is strongly supportedby our results (Fig. 2B). Among the four ingroupgenera, Molinaea is also the only one to occur outsideMadagascar, with three of the nine described speciesfound in the Mascarene Islands (Capuron, 1969). Ataxonomic revision of this genus will soon be completed(M. W. Callmander, P. B. Phillipson and S. Buerki,unpubl. data).

Based on the evidence presented here, Molinaeacan be comfortably maintained as a well-supportedand easily recognized genus, but the status of thethree other genera is less clear. Given the paraphylyof Tinopsis with regard to Neotina in Clade II and thestrong support for the monophyly of Tina (comprisingClade III; Fig. 2), two potentially viable alternativesare available for the generic delimitation of the taxain this clade: (1) to recognize two genera correspond-ing to Clades II and III, respectively; and (2) to placeall of the taxa in a single genus encompassing bothclades (Fig. 2B; Table 1). Below we will consider theadvantages and drawbacks of these alternative clas-sifications.

Option 1 would result in the circumscription of twogenera, Tinopsis (including Neotina) and Tina, only thelatter of which is well supported by molecular data

(BS: 98; Fig. 2B). However, adopting this generic align-ment would present practical problems with regard tomorphology. A broadened circumscription of Tinopsiswould include some members with indehiscent fruitsand fleshy, translucent arillodes and others (thosecurrently assigned to Neotina) with dehiscent fruitsand ceraceous, coloured arillodes, a combination ofcharacters also found in Tina. Moreover, recent fieldobservations made by the authors have shown that thefresh fruits of the species currently referred to asTinopsis macrocarpa Capuron exhibit a well-definedline of dehiscence that is initiated early in develop-ment (Fig. 1), despite the fact that this taxon is nestedwell within a subclade that otherwise comprisesspecies with indehiscent fruits (Fig. 2B). This findingof homoplasy in fruit dehiscence further calls intoquestion the taxonomic utility of this character, whichCapuron (1969) regarded as important for distinguish-ing genera within the group. Our results also lendsupport to studies that have revealed a trend ofhomoplasy in fruit morphology more broadly withinthe family (Harrington et al., 2005; Buerki et al.,2009a) and they suggest that the importance attachedto fruit structure in the past (e.g. by Radlkofer, 1933;Capuron, 1969) may have been misplaced. Furtherfield investigations may show that the true mode ofdehiscence of some members of the family is not fullyreflected in what can be observed on dried herbariumspecimen, as in the case of Tinopsis macrocarpa.

If fruit characters prove to be less informative thanonce supposed, it will be necessary to identify otherattributes that are more reliable for distinguishingmajor groups within the family. In the case of theMalagasy clade being examined here, several fea-tures, including details of the margin of the leaflet,and the number of stamens and stigmatic lines,appear to corroborate the close relationship betweenNeotina and Tinopsis revealed by molecular evidence(Table 1) and might thus lend support to option 1mentioned above. According to Capuron (1969),species of Tina have denticulate leaflets, whereasthose of the two other genera he recognized haveentire leaflets (Table 1). Although this characterappears to have diagnostic value for species, it doesnot correlate well with the current circumscription ofgenera, contrary to Capuron’s assertion. For example,Tina thouarsiana (Cambess.) Capuron has entire leaf-lets, whereas those of Tina dasycarpa Radlk. mayeither be entire or have evident teeth (as observedrecently by the authors). With regard to the numberof stamens, Tina has long been seen as distinctive inhaving six to eight stamens vs. five in the two othergenera (Table 1). However, field observations reportedby Capuron (1969), and confirmed by the authors ofthe present study, clearly show a high level of vari-ability in this character, especially within Tina (for a



detailed review see Capuron, 1969), casting doubtover its value for circumscribing genera. Alterna-tively, the presence and shape of a gland at the apexof the anther appears to have potential for distin-guishing Tinopsis (including Neotina), which hasoblong, emarginate anthers without a gland at theapex, from Tina, members of which have subcordi-form, apiculous anthers with an apical gland(Table 1). The presence of a short stigmatic line inTina, vs. a well-developed stigmatic line that extendsalong the style in the other genera (Table 1), alsoseems to support option 1. Finally, Capuron (1969),like Radlkofer (1933) before him, attributed greatimportance to embryo type, and on this basis hedistinguished Neotina, with its lomatorrhizal embryo(a character shared with Tinopsis), from Tina, whichhas a notorrhizal embryo (Table 1). We have not foundany exceptions to this pattern (based on limited obser-vations), but embryo type is not easily observed in thefield or on dried specimens, a fact that significantlylimits its utility as a generic level character.

Option 2, which would involve placing all the taxacurrently assigned to Neotina, Tina and Tinopsis in asingle, significantly expanded genus, would be fullyconsistent with our molecular findings, which providestrong support for this clade (BS: 100). If this optionwere adopted, the name Tina would have nomencla-tural priority (Acevedo-Rodríguez, 2003). Circum-scribed in this manner, the genus would comprise c. 20species, encompassing all members of the ingroup witha bicarpellate gynoecium, which would constitute arobust and easily observed synapomorphy, unambigu-ously enabling fertile material of Tina to be distin-guished from specimens of Molinaea. This option thushas a clear practical advantage over the alternativeoutlined above by avoiding problems involving charac-ters related to the anthers and stigmas, which may beof value for distinguishing the three traditionallyrecognized genera, but are often particularly difficultto observe in the small flowers produced by theseplants, frequently requiring magnification to becertain of which character state is expressed.

Our observations suggest that the floral and veg-etative characters mentioned above will prove to bemore suitable for distinguishing species within anewly expanded Tina. Moreover, they may be of valuefor clarifying taxonomic limits within the most prob-lematic member of the genus, T. striata Radlk., whichexhibits a high level of morphological polymorphism(with five subspecies recognized by Capuron, 1969)and appears to be polyphyletic as currently defined(Fig. 2B). An evaluation of the potential utility ofthese morphological features will be facilitated bymaking use of the microsatellites recently developedby Vary et al. (2009) to investigate species limitswithin the T. striata complex.

It should be noted that the phylogenetic distancesbetween taxa within the Cupania group are quite low,especially compared with other groups of Sapindaceae(Fig. 2A), despite the high level of generic diversityobserved in the group (c. 32 genera; Buerki et al.,2009a, 2010a, b). The significant non-monophyly foundin some of the currently recognized genera (Cupaniop-sis Radlk., Guioa Cav., Matayba, Sarcotoechia Radlk.and Tinopsis) might reflect a historical tendencytoward taxonomic over-splitting. If this proves to bethe case, then the option we have proposed here forexpanding Tina to include Neotina and Tinopsis maybe the first of several such modifications to the currentgeneric level taxonomy within the group. Also, we notethat further studies (along the same lines as the onepresented here) will be needed to test the utility androbustness of these genera. From an evolutionarypoint of view, the phylogenetic information presentedhere is consistent with a rapid diversification withinthe widely distributed Cupania group, the varioussubgroups of which have members throughout theTropics (with the sole exception of Africa).

LET’S KILL TWO GENERA WITH ONE TREE:RADLKOFER WON!

Based on the results presented above, we recommendthat an expanded circumscription of Tina be adopted toencompass all bicarpellate members of the Malagasyclade under consideration here (the required taxo-nomic and nomenclatural changes will be formallypublished in a forthcoming paper). This approachavoids the problems associated with homoplasy foundin fruit features historically emphasized and the diffi-culty of accurately observing minute and often crypticfloral characters. Moreover, placing Tinopsis andNeotina in synonymy under Tina provides a morepractical and easily applied taxonomic framework thatis fully consistent with our molecular findings andmore easily accommodates material the morphology ofwhich does not conform to historical generic limits.

In view of the rapidly increasing affordability andefficiency of using molecular techniques to elucidateevolutionary relationships, generic circumscriptionsand taxonomic revisions should, whenever possible,be based on a strong phylogenetic framework. At thesame time, when taxa are circumscribed, they mustbe supported by unambiguous and easily observablemorphological synapomorphies, insofar as possible.This is the approach we have used in proposing ourrevised circumscription of Tina in a way that enablesit to be easily recognized and unambiguously distin-guished from Molinaea, returning to the genericalignment adopted more than 80 years ago by Choux(1925, 1927) and Radlkofer (1933) based solely onmorphological evidence.



ACKNOWLEDGEMENTS

The authors wish to thank the Malagasy staff of theMissouri Botanical Garden (MBG) based in Antanan-arivo for local assistance and the curators of thefollowing herbaria for making material available forthis study: G, K, MO, P, TAN and TEF. Work inMadagascar was conducted under collaborative agree-ments between MBG and the Parc Botanique etZoologique de Tsimbazaza and the Département ofBiologie végétale at the Université d’ Antananarivo(Madagascar). We gratefully acknowledge courtesiesextended by the Government of Madagascar (DirectionGénérale de la Gestion des Ressources Forestières)and by Madagascar National Parks (formerly theAssociation Nationale pour la Gestion des Aires Pro-tégées). The authors are grateful to the editor, MikePole, and an anonymous reviewer for their valuablecomments that improved our manuscript. The authorswould like to thank G. E. Schatz for helping with thecollections at MO. The first author is also grateful toIsabel Sanmartín and Nadir Alvarez for their support.The work of the first author was supported financiallyby a Prospective Researchers Fellowship from theSwiss National Science Foundation (no. PBNEP3-129903). The participation of P.P.L., P.B.P. and M.W.C.was supported by a grant from the US NationalScience Foundation (0743355; P.P.L., co-principalinvestigator) and the Andrew W. Mellon Foundation(P.B.P. and P.P.L., co-principal investigators).

REFERENCES

Acevedo-Rodríguez P. 2003. (1592) Proposal to conserve thename Tina (Sapindaceae) with a conserved type. Taxon 52:373–374.

Andriambololonera S. 1999. Contribution à l’étude du genreendémique Tina Roem. & Schultes (Sapindaceae). Mémoiresde Diplôme d’Etudes Approfondies en Sciences BiologiquesAppliquées. Université d’Antananarivo.

APG II. 2003. An update of the Angiosperm Phylogeny Groupclassification for the orders and families of flowering plants:APG II. Botanical Journal of the Linnean Society 141:399–436.

APG III. 2009. An update of the Angiosperm PhylogenyGroup classification for the orders and families of floweringplants: APG III. Botanical Journal of the Linnean Society161: 105–121.

Buerki S, Forest F, Acevedo-Rodríguez P, CallmanderMW, Nylander JAA, Harrington M, Sanmartín I,Küpfer P, Alvarez P. 2009a. Plastid and nuclear DNAmarkers reveal intricate relationships at subfamilial andtribal levels in the soapberry family (Sapindaceae). Molecu-lar Phylogenetics and Evolution 51: 238–258.

Buerki S, Callmander MW, Lowry PP II, Phillipson PB.2009b. A synoptic revision of the genus Lepisanthes Blume

(Sapindaceae) in Madagascar. Adansonia, Série 3 31: 301–309.

Buerki S, Lowry PP II, Alvarez N, RazafimandimbisonSG, Küpfer P, Callmander MW. 2010a. Phylogeny andcircumscription of Sapindaceae revisited: molecularsequence data, morphology and biogeography support rec-ognition of a new family, Xanthoceraceae. Plant Ecology andEvolution 143: 148–161.

Buerki S, Phillipson PB, Lowry PP II, Callmander MW.2010b. Molecular phylogenetics and morphological evidencesupport recognition of Gereaua, a new endemic genus ofSapindaceae from Madagascar. Systematic Botany 35: 172–180.

Buerki S, Forest F, Alvarez N, Nylander JAA, Arrigo N,Sanmartín I. 2010c. An evaluation of new parsimony-based versus parametric inference methods in biogeogra-phy: a case study using the globally distributed plant familySapindaceae. Journal of Biogeography 38: 531–550.

Capuron R. 1969. Révision des Sapindacées de Madagascaret des Comores. Mémoires du Muséum National d’HistoireNaturelle, Série B Botanique 19: 1–189.

Choux P. 1925. Les Cupaniées malgaches. Comptes RendusHebdomadaires des Séances de l’Académie des Sciences 181:71–72.

Choux P. 1927. Les Sapindacées de Madagascar. Mémoiresde l’Académie Malgache 4: 1–118.

Durand TH. 1888. Index generum phanerogamorum. Brus-sels: Sumptibus Auctoris.

Felsenstein J. 1985. Confidence-limits on phylogenies – anapproach using the bootstrap. Evolution 39: 783–791.

Harrington MG, Edwards KJ, Johnson SA, Chase MW,Gadek PA. 2005. Phylogenetic inference in Sapindaceaesensu lato using plastid matK and rbcL DNA sequences.Systematic Botany 30: 366–382.

Nixon KC. 1999. The parsimony ratchet, a new method forrapid parsimony analysis. Cladistics 15: 407–414.

Radlkofer L. 1933. Sapindaceae. In: Engler A, ed. Das Pflan-zenreich IV, 165 (Heft 98a,h). Leipzig: Verlag von WilhelmEngelmann, 1–1539.

Schatz GE, Gereau RE, Lowry PP II. 1999. A revision ofthe Malagasy endemic genus Chouxia Capuron (Sapin-daceae). Adansonia, série 3 21: 51–62.

Sikes DS, Lewis PO. 2001. Pauprat. Storrs: University ofConnecticut.

Stamatakis A. 2006. RAxML-VI-HPC: maximum likelihood-based phylogenetic analyses with thousands of taxa andmixed models. Bioinformatics 22: 2688–2690.

Stamatakis A, Hoover P, Rougemont J. 2008. A rapidbootstrap algorithm for the RAxML web-servers. SystematicBiology 57: 758–771.

Swofford DL. 2002. PAUP*: phylogenetic analysis using par-simony (*and other methods), version 4. Sunderland:Sinauer Associates.

Vary LB, Lance SL, Hagen C, Tsyusko O, Glenn TC,Sakai AK, Weller SG. 2009. Characterization of microsat-ellite loci from the Malagasy endemic, Tina striata Radlk.(Sapindaceae). Conservation Genetics 10: 1113–1115.



AP

PE

ND

IX

Ingr

oup

vou

cher

info

rmat

ion

and

Gen

Ban

kac

cess

ion

nu

mbe

rsfo

rta

xau

sed

inth

eph

ylog

enet

ican

alys

is(s

eeB

uer

kiet

al.,

2010

bfo

rde

tail

son

the

oth

erta

xa).

G,

Con

serv

atoi

reet

Jard

inB

otan

iqu

esde

laV

ille

deG

enèv

e,S

wit

zerl

and;

MO

,M

isso

uri

Bot

anic

alG

arde

n,

US

A;

P,M

usé

um

Nat

ion

ald’

His

toir

eN

atu

rell

e,F

ran

ce;

IGS

,in

trag

enic

spac

er;

ITS

,in

tern

altr

ansc

ribe

dsp

acer

.

Taxo

nA

uth

orV

ouch

erH

erba

riu

mIT

Sm

atK

rpoB

trn

D-t

rnT

IGS

trn

K-m

atK

IGS

trn

Lin

tron

trn

L-t

rnF

IGS

trn

S-t

rnG

IGS

N mar

kers

Mol

inae

are

tusa

Rad

lk.

Cal

lman

der

572

G–

HQ

3992

43H

Q39

9278

HQ

3993

06H

Q39

9332

HQ

3994

10H

Q39

9370

HQ

3994

367

Mol

inae

asp

.in

det.

Gau

tier

4783

GH

Q39

9202

HQ

3992

27H

Q39

9257

–H

Q39

9316

HQ

3993

87H

Q39

9347

–6

Mol

ina

easp

.n

ov.

1A

nti

lah

imen

a43

01M

OE

U72

0510

EU

7206

62E

U72

0854

EU

7209

83E

U72

1099

EU

7212

80E

U72

1468

EU

7215

788

Mol

ina

easp

.n

ov.

2R

avel

onar

ivo

1784

MO

–H

Q39

9240

HQ

3992

75H

Q39

9303

HQ

3993

29H

Q39

9407

HQ

3993

67–

6M

olin

aea

tola

mbi

tou

(Cam

b.)

Rad

lk.

Rab

enan

toan

dro

1448

MO

EU

7205

54E

U72

0700

EU

7209

02E

U72

1007

EU

7211

38E

U72

1324

EU

7215

12–

7N

eoti

na

cou

rsii

Cap

uro

nR

azafi

nd

raib

e10

9M

O–

HQ

3992

37H

Q39

9272

HQ

3993

00H

Q39

9326

HQ

3994

04H

Q39

9364

HQ

3994

317

Neo

tin

aco

urs

iiC

apu

ron

Raz

afin

dra

ibe

119

MO

EU

7205

43E

U72

0690

EU

7208

91E

U72

1002

EU

7211

28E

U72

1313

EU

7215

01E

U72

1594

8N

eoti

na

cou

rsii

Cap

uro

nVa

ry29

PH

Q39

9203

–H

Q39

9258

––

HQ

3993

88H

Q39

9348

–4

Neo

tin

aco

urs

iiC

apu

ron

Vary

35P

–H

Q39

9244

HQ

3992

79–

HQ

3993

33H

Q39

9411

HQ

3993

71–

5N

eoti

na

ison

eura

(Rad

lk.)

Cap

uro

nR

azak

amal

ala

3004

MO

HQ

3991

88–

HQ

3992

45H

Q39

9282

–H

Q39

9374

HQ

3993

34H

Q39

9414

6T

ina

chap

elie

rian

a(C

amb.

)K

alk.

Mil

ler

8759

MO

HQ

3991

97H

Q39

9224

HQ

3992

53H

Q39

9289

HQ

3993

13H

Q39

9383

HQ

3993

43H

Q39

9420

8T

ina

chap

elie

rian

a(C

amb.

)K

alk.

Ran

iris

on82

7M

OE

U72

0520

EU

7206

67E

U72

0864

EU

7209

86E

U72

1104

EU

7212

86E

U72

1474

EU

7215

798

Tin

afu

lvin

ervi

sR

adlk

.B

uer

ki13

6G

–H

Q39

9239

HQ

3992

74H

Q39

9302

HQ

3993

28H

Q39

9406

HQ

3993

66H

Q39

9433

7T

ina

stri

ata

subs

p.m

ult

ifol

iata

Cap

uro

nC

allm

and

er58

4G

HQ

3992

04H

Q39

9228

HQ

3992

59–

HQ

3993

17H

Q39

9389

HQ

3993

49–

6

Tin

ast

riat

asu

bsp.

mu

ltif

olia

taC

apu

ron

Cal

lman

der

618

MO

HQ

3992

09H

Q39

9232

HQ

3992

64H

Q39

9296

HQ

3993

21H

Q39

9394

HQ

3993

54H

Q39

9427

8

Tin

ast

riat

asu

bsp.

mu

ltif

olia

taC

apu

ron

Sch

atz

3746

MO

HQ

3992

07–

HQ

3992

62H

Q39

9294

–H

Q39

9392

HQ

3993

52H

Q39

9425

6

Tin

ast

riat

asu

bsp.

parv

ifol

iaC

apu

ron

An

tila

him

ena

4789

MO

–H

Q39

9241

HQ

3992

76H

Q39

9304

HQ

3993

30H

Q39

9408

HQ

3993

68H

Q39

9434

7

Tin

ast

riat

asu

bsp.

parv

ifol

iaC

apu

ron

Cal

lman

der

647

MO

HQ

3992

12H

Q39

9235

HQ

3992

67H

Q39

9299

HQ

3993

24H

Q39

9397

HQ

3993

57H

Q39

9430

8

Tin

ast

riat

asu

bsp.

parv

ifol

iaC

apu

ron

Raz

afits

alam

a11

31M

OH

Q39

9210

HQ

3992

33H

Q39

9265

HQ

3992

97H

Q39

9322

HQ

3993

95H

Q39

9355

HQ

3994

288

Tin

ast

riat

asu

bsp.

parv

ifol

iaC

apu

ron

Raz

afits

alam

a11

32M

OH

Q39

9211

HQ

3992

34H

Q39

9266

HQ

3992

98H

Q39

9323

HQ

3993

96H

Q39

9356

HQ

3994

298

Tin

ast

riat

asu

bsp.

stri

ata

Rad

lk.

Bu

erki

75G

–H

Q39

9238

HQ

3992

73H

Q39

9301

HQ

3993

27H

Q39

9405

HQ

3993

65H

Q39

9432

7

Tin

ast

riat

asu

bsp.

stri

ata

Rad

lk.

Ran

dri

anar

ivel

o37

8M

OH

Q39

9206

HQ

3992

30H

Q39

9261

HQ

3992

93H

Q39

9319

HQ

3993

91H

Q39

9351

HQ

3994

248



AP

PE

ND

IXC

onti

nu

ed

Taxo

nA

uth

orV

ouch

erH

erba

riu

mIT

Sm

atK

rpoB

trn

D-t

rnT

IGS

trn

K-m

atK

IGS

trn

Lin

tron

trn

L-t

rnF

IGS

trn

S-t

rnG

IGS

N mar

kers

Tin

ast

riat

asu

bsp.

stri

ata

Rad

lk.

Rav

elon

ariv

o19

04M

O–

HQ

3992

42H

Q39

9277

HQ

3993

05H

Q39

9331

HQ

3994

09H

Q39

9369

HQ

3994

357

Tin

ast

riat

aR

adlk

.su

bsp.

stri

ata

Sch

atz

4024

MO

HQ

3992

08H

Q39

9231

HQ

3992

63H

Q39

9295

HQ

3993

20H

Q39

9393

HQ

3993

53H

Q39

9426

8

Tin

ast

riat

aR

adlk

.su

bsp.

stri

ata

Vary

26P

HQ

3992

14–

––

–H

Q39

9399

HQ

3993

59–

3

Tin

ast

riat

aR

adlk

.su

bsp.

stri

ata

Vary

27P

HQ

3992

15–

––

–H

Q39

9400

HQ

3993

60–

3

Tin

ast

riat

aR

adlk

.su

bsp.

stri

ata

Vary

30P

HQ

3992

13–

HQ

3992

68–

–H

Q39

9398

HQ

3993

58–

4

Tin

ast

riat

aR

adlk

.su

bsp.

stri

ata

Vary

31M

OH

Q39

9216

–H

Q39

9269

––

HQ

3994

01H

Q39

9361

–4

Tin

ast

riat

aR

adlk

.su

bsp.

stri

ata

Vary

43P

HQ

3992

17H

Q39

9236

HQ

3992

70–

HQ

3993

25H

Q39

9402

HQ

3993

62–

6

Tin

ast

riat

aR

adlk

.su

bsp.

stri

ata

Vary

45P

EU

7205

09E

U72

0661

EU

7208

53–

EU

7210

98E

U72

1279

EU

7214

67–

6

Tin

ast

riat

aR

adlk

.su

bsp.

stri

ata

Vary

52M

OH

Q39

9218

–H

Q39

9271

––

HQ

3994

03H

Q39

9363

–4

Tin

ath

ouar

sian

a(C

amb.

)C

apu

ron

Low

ry60

21M

OH

Q39

9205

HQ

3992

29H

Q39

9260

–H

Q39

9318

HQ

3993

90H

Q39

9350

–6

Tin

ath

ouar

sian

a(C

amb.

)C

apu

ron

Rab

evoh

itra

4445

MO

––

HQ

3992

80–

–H

Q39

9412

HQ

3993

72–

3T

inop

sis

anto

ngi

len

sis

Cap

uro

nA

nti

lah

imen

a46

14M

OH

Q39

9195

–H

Q39

9251

HQ

3992

88–

HQ

3993

81H

Q39

9341

HQ

3994

196

Tin

opsi

san

ton

gile

nsi

sC

apu

ron

An

tila

him

ena

5493

MO

HQ

3991

99–

––

––

––

1T

inop

sis

anto

ngi

len

sis

Cap

uro

nC

allm

and

er38

8G

HQ

3991

98–

HQ

3992

54H

Q39

9290

–H

Q39

9384

HQ

3993

44H

Q39

9421

6T

inop

sis

anto

ngi

len

sis

Cap

uro

nR

avel

onar

ivo

1664

MO

HQ

3991

94–

HQ

3992

50H

Q39

9287

–H

Q39

9380

HQ

3993

40H

Q39

9418

6T

inop

sis

apic

ula

taR

adlk

.B

uer

ki13

1G

EU

7204

22E

U72

0589

EU

7207

44E

U72

0936

EU

7210

34E

U72

1180

EU

7213

68E

U72

1540

8T

inop

sis

con

juga

ta(R

adlk

.)C

apu

ron

Mil

ler

8757

MO

HQ

3991

96H

Q39

9223

HQ

3992

52–

HQ

3993

12H

Q39

9382

HQ

3993

42–

6T

inop

sis

con

juga

ta(R

adlk

.)C

apu

ron

Rab

enan

toan

dro

1216

MO

––

HQ

3992

81H

Q39

9307

–H

Q39

9413

HQ

3993

73H

Q39

9437

5T

inop

sis

mac

roca

rpa

Cap

uro

nB

uer

ki13

4G

HQ

3992

01H

Q39

9226

HQ

3992

56H

Q39

9292

HQ

3993

15H

Q39

9386

HQ

3993

46H

Q39

9423

8T

inop

sis

phel

loca

rpa

Cap

uro

nA

nti

lah

imen

a43

28M

OH

Q39

9200

HQ

3992

25H

Q39

9255

HQ

3992

91H

Q39

9314

HQ

3993

85H

Q39

9345

HQ

3994

228

Tin

opsi

sta

mat

aven

sis

Cap

uro

nB

uer

ki13

3G

HQ

3991

91H

Q39

9221

HQ

3992

47H

Q39

9284

HQ

3993

10H

Q39

9377

HQ

3993

37H

Q39

9416

8T

inop

sis

tam

atav

ensi

sC

apu

ron

Bu

erki

135

GH

Q39

9192

–H

Q39

9248

HQ

3992

85–

HQ

3993

78H

Q39

9338

HQ

3994

176

Tin

opsi

sta

mat

aven

sis

Cap

uro

nB

uer

ki14

0G

HQ

3991

93H

Q39

9222

HQ

3992

49H

Q39

9286

HQ

3993

11H

Q39

9379

HQ

3993

39–

7T

inop

sis

urs

chii

Cap

uro

nA

nti

lah

imen

a39

51M

OH

Q39

9190

HQ

3992

20H

Q39

9246

HQ

3992

83H

Q39

9309

HQ

3993

76H

Q39

9336

HQ

3994

158

Tin

opsi

su

rsch

iiC

apu

ron

An

tila

him

ena

5198

MO

HQ

3991

89H

Q39

9219

––

HQ

3993

08H

Q39

9375

HQ

3993

35–

5



Documents

How to kill two genera with one tree: clarifying generic