Multidisciplinary perspectives on banana(Musa spp.) domesticationXavier Perriera, Edmond De Langheb, Mark Donohuec, Carol Lentferd, Luc Vrydaghse, Frédéric Bakrya, Françoise Carreelf,Isabelle Hippolytea, Jean-Pierre Horrya, Christophe Jennyg, Vincent Leboth, Ange-Marie Risteruccia, Kodjo Tomekpea,Hugues Doutreleponte, Terry Balli, Jason Manwaringi, Pierre de Maretj, and Tim Denhamk,1

aCentre de Coopération Internationale en Recherche Agronomique pour le Développement, Unité Mixte de Recherche AméliorationGénétique et Adaptation des Plantes, F-34398 Montpellier, France; bLaboratory of Tropical Crop Improvement, Katholieke Universiteit, 3001Leuven, Belgium; cDepartment of Linguistics, Australian National University, Canberra 0200, Australia; dSchool of Social Science, Universityof Queensland, St. Lucia 4072, Australia; eResearch Team in Archaeo- and Palaeosciences, 1160 Brussels, Belgium; fCentre de CoopérationInternationale en Recherche Agronomique pour le Développement, Unité Mixte de Recherche Biologie et Génétique des InteractionsPlantes-Parasites, F-34398 Montpellier, France; gCentre de Coopération Internationale en Recherche Agronomique pour le Développement,Unité Mixte de Recherche Amélioration Génétique et Adaptation des Plantes, F-97130 Capesterre-Belle-Eau, Guadeloupe, France; hCentrede Coopération Internationale en Recherche Agronomique pour le Développement, Unité Mixte de Recherche Amélioration Génétique etAdaptation des Plantes, Department of Agriculture, Centre Agronomique de Recherche et de Formation du Vanuatu, 946 Port Vila,Vanuatu; iDepartment of Ancient Scripture, Brigham Young University, Provo, UT 84602; jSecrétariat, Centre d’Anthropologie Culturelle,Université Libre de Bruxelles, 1000 Brussels, Belgium; and kSchool of Geography and Environmental Science, Monash University, Victoria3800, Australia

Edited by Dolores R. Piperno, Smithsonian National Museum of Natural History and Smithsonian Tropical Research Institute, Panama City, Panama, andapproved May 25, 2011 (received for review March 5, 2011)

Original multidisciplinary research hereby clarifies the complex geodomestication pathways that generated the vast range of bananacultivars (cvs). Genetic analyses identify thewild ancestors ofmodern-day cvs and elucidate several key stages of domestication for differentcv groups. Archaeology and linguistics shed light on the historical roles of people in the movement and cultivation of bananas from NewGuinea to West Africa during the Holocene. The historical reconstruction of domestication processes is essential for breeding programsseeking to diversify and improve banana cvs for the future.

plant genetics | historical linguistics | archaeobotany | diploid banana cultivars | triploid banana cultivars

New multidisciplinary findingsfrom archaeology, genetics,and linguistics clarify thecomplex geodomestication

pathways—the geographical config-urations of hybridization and dispersal—that generated the range of modern ba-nana cultivars (cvs). Although recentmolecular research, combined with theoutcomes of previous genetic studies, elu-cidates major stages of banana domesti-cation, such as the generation of ediblediploids and triploids, it sheds only partiallight on the historical and sociospatialcontexts of domestication. The geographicdistributions of genotypes involved in ba-nana domestication require human trans-locations of plants, most likely undervegetative forms of cultivation, across vastregions. Linguistic analyses of (traditional)local terms for bananas reveal severalstriking regional-scale correspondencesbetween genetic and linguistic patterns.These multidisciplinary findings enable therelative dating of the principal events inbanana geodomestication and situate ba-nana cultivation within broader socio-spatial contexts. Archaeological findingsprovide a timeline to anchor and calibratethe relative chronology.Banana ranks next to rice, wheat, and

maize in terms of its importance as a foodcrop. In addition to being a major cash croparound the world, more than 85% of ba-nanas are grown for local consumption in

tropical and subtropical regions (1). Thehundreds of cultivated varieties are prod-ucts of centuries—in some cases millennia—of clonal (vegetative) propagation.Banana fruits can be cooked, roasted, oreven brewed (e.g., plantains; East AfricanHighland cvs) or eaten raw (e.g., theyellow Cavendish banana sold in super-markets globally).

Musa Genetics and DomesticationThresholdsThe monocotyledon Musaceae familyincludes the Asian and African genusEnsete, the genetically proximal AsianMusella genus, and the East Asian genusMusa, which is divided into sections with22 (Eumusa, Rhodochlamys) and 20chromosomes (Australimusa, Callimusa).Most edible bananas belong to the Eu-musa section, and are diploid or triploidhybrids from Musa acuminata (A-genome)alone or from hybridization with Musabalbisiana (B-genome). A minor cv group,including Fe’i bananas, is confined tothe Pacific region and is derived fromAustralimusa species.The evolution from wild to edible ba-

nanas involved seed suppression and par-thenocarpy development. Simmonds (2)proposed the first reconstruction of thiscomplex process in the 1960s. Since thenthe spectrum of available genetic re-sources has been greatly expanded, nota-bly with collections from New Guinea

(NG), the native area of the M. acuminatasubsp. banksii (3), and through the appli-cation of molecular methods to bananadiversity analysis (4–7).In the present study, we present the

conclusions of two recent analyses by usingsimple sequence repeats (8) and DarT (9)markers on the international collection inGuadeloupe (Centre de Coopération In-ternationale en Recherche Agronomiquepour le Développement, France) whichassembles more than 400 wild and culti-vated accessions covering much of globalMusa diversity. Analyses here include ad-ditional accessions from Cameroon andNigeria to better represent African cv di-versity (Table S1). The results of theseanalyses were integrated with previousresults, including ploidy level and DNAcontent (10), anthocyanin patterns (11),isozymes (12), nuclear restriction (13)and amplified fragment length poly-morphism (14), and chloroplastic and mi-tochondrial restriction fragment length

Author contributions: P.d.M. designed research; E.D.L.,M.D., C.L., L.V., F.B., F.C., I.H., J.-P.H., C.J., L.V., A.-M.R.,K.T., H.D., T.B., and J.M. performed research; X.P. analyzeddata; and X.P., E.D.L., M.D., and T.D. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.1To whom correspondence should be addressed. E-mail:tim.denham@arts.monash.edu.au.

This article contains supporting information online atwww.pnas.org/lookup/suppl/doi:10.1073/pnas.1102001108/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1102001108 PNAS | July 12, 2011 | vol. 108 | no. 28 | 11311–11318

PERSPECTIV

ED

ownl

oade

d by

gue

st o

n M

ay 1

2, 2

020

polymorphism (15). The synthesis of theseresults sheds light upon the two foundingevents during banana domestication: thetransition from wild to edible diploids andthe formation of triploids from ediblediploids (16) (Fig. S1).Wild M. acuminata species are diverse

and have been differentiated into severalsubspecies (13, 16). This differentiation isclarified here with the identification ofspecific nuclear alleles, chromosomal re-arrangements, and chloroplast and mito-chondrial patterns. The exclusive naturalranges of the subspecies are also con-firmed (Fig. 1).The key result relevant to domestication

history is that edible diploid cv subgroups(AAcvs) derive from hybridizations be-tween different M. acuminata subspeciesin island Southeast Asia (SEA; ISEA)and western Melanesia (Fig. 2 and Fig.S2), which could only be brought intocontact by human movement and propa-gation, most probably during the Holo-cene (i.e., the past 11,700 y). The

structural heterozygosity of these hybridAAcvs, caused by chromosomal re-arrangements between parental subspeciesof M. acuminata (17), contributed to ga-metic sterility (18). This sterility, in asso-ciation with human selection for pulpenhancement, led to parthenocarpic fruitsand edibility.A further consequence of hybrid status

was erratic meiosis in edible AAcvs,thereby occasionally producing diploidgametes (2). The fusion of diploid gameteswith haploid gametes generated steriletriploid genotypes. Spontaneous triploid-izations involved almost all diploid cvsleading, under human selection via vege-tative propagation, to the diversity ofmodern cultivated triploids, including pureM. acuminata varieties (AAA) and in-terspecific M. acuminata × M. balbisianavarieties (AAB, ABB) (19).Molecular phylogenetic analyses have

identified the diploid subgroups contrib-uting to modern triploid cvs. Of signifi-cance, NG-derived banksii cvs played a

major role in the development of severalimportant diploid and triploid cv groupsthat are now widely dispersed. For exam-ple, M. acuminata subsp. zebrina- andsubsp. banksii-derived AAcvs contributedto AAA Highland bananas of East Africa.Likewise, subsp. banksii-derived AAcvswith the BB genome contributed to AABplantains of West Africa and the Pacific.Moreover, in several triploid subgroups,specific accessions within AAcv subgroupshave been identified as potential ances-tors, suggesting more recent formation(16, 20). For example, the 2N and N pa-rents of the common dessert banana, theAAA “Cavendish,” are genetically close to“Akondro Mainty” of the AAcv “Mlali”subgroup and “Pisang Pipit” of the AAcv“Khai” subgroup, respectively (Fig. 2).Strict vegetative propagation (i.e., clon-

ing) over long periods of the most populardiploid and triploid varieties led to soma-clonal variants, thus amplifying phenotypicdiversity (14). The geographical ranges ofdiploid parents and triploid hybrids pointto numerous long-distance movements ofeither one of the parents before hybrid-ization, or of triploid cvs after hybridiza-tion. The geographical dimension of thebanana domestication process requiresa reconstruction of relevant humanmovements and interactions—the driversof the geodomestication process.

Anchoring Banana Domestication inTime and PlaceMicrofossil (phytoliths, pollen) and mac-rofossil (seeds, pseudostem impressions)remains of bananas have been documentedat numerous archaeological and paleo-ecological sites from Oceania to Africa(21), although their significance variesgreatly for understanding the historyof banana domestication and cultivation.Within the natural range of Musa,archaeobotanical and paleoecologicalfinds require differentiation to a species’or subspecies’ level to indicate potentialhuman movement, and also require asso-ciated evidence of human exploitation orcultivation. Most identifications are to thegenus level and not directly associatedwith archaeological remains; conse-quently, they are of limited value forunderstanding banana domestication be-cause they may solely record wild bananas(21). Outside the natural range of Musa,bananas were exotic, dispersed under cul-tivation, predominantly sterile, and vege-tatively propagated; consequently, anyarchaeobotanical find of Musa signifiesthe presence of agriculture.Predominantly, the Musa genus has

been identified in the archaeological re-cord by using phytoliths (21). As a result ofparthenocarpy in most cvs, seeds are un-likely to be found in prehistoric agricul-tural contexts, and usually leaf phytoliths

Fig. 1. Geographical distribution ofM. balbisiana and subspecies ofM. acuminata, the wild ancestors ofcultivated bananas. Nuclear and cytoplasmic markers (8, 13, 15) differentiate M. acuminata wild diploidsinto clusters thatfitwith subspecies previously defined frommorphological characters: banksii cluster fromNG;malaccensis cluster fromMalayan Peninsula; and burmannica, burmannicoides, and siamea cannot bediscriminated by their nuclear genome and form a complex that covers South China, Thailand, Myanmar,Bangladesh, northeast India with sporadic populations southward to Sri Lanka. This complex is geneticallycloser to the malaccensis cluster than to other clusters, they are geographically overlapping, and severalwild accessions from Thailand are identified as hybrids. Zebrina cluster from Java and its genome size issignificantly (10%) higher (8). Its anthocyanin pathway is the most primeval (19), banksii is at an in-termediate stage, and the other subspecies are at a more advanced step of evolution. Microcarpa wasidentified frommorphological characters, isozymes, and chloroplastic genome; however, nuclear genomerevealed similarities with other subspecies, particularly zebrina, instigating consideration of a zebrine/microcarpa complex. Truncata is endemic to the highlands of Peninsular Malaysia. Errans from the Phil-ippines was given the status of a subspecies of M. acuminata (22), although only one accession has beenstudied. It is the only wildM. acuminatawith the α-mitochondrial type, an important type found in a lot ofdiploid and triploid cultivars. It shares similarities with banksii nuclear genome. M. balbisiana has a morenortherly distribution and, although not domesticated, it has been widely translocated for its many uses,thereby founding small free-growing populations from NG in the east to Sri Lanka in the west.

11312 | www.pnas.org/cgi/doi/10.1073/pnas.1102001108 Perrier et al.

Dow

nloa

ded

by g

uest

on

May

12,

202

0

have been used for identification to thegenus level, although they are less readilydifferentiable to the species level thanseed phytoliths (22–24). Here, discrimi-nating morphological criteria of leaf andseed phytoliths are reexamined and aug-mented for reliable interspecific and in-traspecific differentiation within Musa andbetween Musa and Ensete (Figs. S3 and S4and Tables S2 and S3). These reviseddiscriminatory criteria are applied to as-semblages at two key archaeological sitesthat anchor thresholds of prehistoric Musageodomestication: cultivation of M. acu-minata at 6,950 to 6,440 calibrated y (cal)BP at Kuk Swamp in the highlands of NGand the dispersal under cultivation ofsterile AAB plantains by 2,750 to 2,300 calBP to Nkang, Cameroon.AbundantMusaceae phytoliths in feature

fills associated with agricultural practicesand dated to 6,950 to 6,440 cal BP at Kukare indicative of banana cultivation (25, 26).Identifiable morphotypes with attributesmatching the comparative reference sam-ples of M. acuminata subsp. banksii fromNG (Fig. 3 A–C), the only naturally oc-curring M. acuminata subspecies on the is-

land, can be differentiated from otherMusaceae grown there (Fig. S3 and TableS2), although archaeobotanical occurrencesare rare (Fig. 3D). So, the occurrence ofM.acuminata morphotypes in an early agri-cultural context is highly significant andprovides a minimum date for the cultiva-tion of subsp. banksii derivatives in NG.Volcaniform leaf phytoliths ofMusa spp.

were reported from two secure archaeo-logical contexts at Nkang in Cameroon:from two horizons within pit 9 (n = 20,dated to 2,790–2,300 cal BP) and froma charred deposit adhered to a potterysherd from pit 7 (n = 5, dated 2,750–2,100cal BP; Fig. 3 E–H) (22, 23, 27). Theoriginal morphological criteria used todistinguish Musa and Ensete phytoliths atNkang have been revised (Fig. S4 andTable S3) and confirm the phytoliths atNkang to be Musa, as opposed to the na-tive African genus Ensete. So, this reeval-uation of the Nkang assemblages providesa secure archaeobotanical foundation andminimum date for the cultivation of exoticplantains in West Africa.This verification of the archaeobotany of

Musa at Kuk and Nkang is essential to the

development of a chronological frame-work for the domestication phylogeny(from genetics) and dispersal of cvs (fromlinguistics). Other significant archaeo-botanical finds of Musa are discussed withreference to other lines of evidence inthe subsequent sections. However, differ-entiation below the genus level has notbeen possible at most archaeological orpaleoecological sites; for example, Musaphytoliths date to 4,000 y ago at Kot Dijiin Pakistan (28) and have also been asso-ciated with the post–3,500-y-old Lapitaphenomenon in the western Pacific (21,29). Exceptions include the identificationsof M. balbisiana and cf M. acuminata seedsfrom Terminal Pleistocene contexts atBeli-Lana on Sri Lanka (30) and the dis-crimination of Ensete,M. acuminata subsp.banksii, and Australimusa bananas inmid-Holocene contexts on central NewBritain (31). The Sri Lanka evidence isconsistent with modern, wildM. acuminatasubsp. burmannica there, reflecting conti-nuity with populations in East Indiaand Myanmar. However, no significantcv (AA, AAA, or AAB) based on subsp.burmannica has been identified; conse-quently, this subspecies and this regionare not significant for understandingbanana domestication.

Tracing Banana Dispersal Through His-torical LinguisticsA cultivated plant often migrates with itsname, and when this plant is culturallyinnovative, its name is often retained inthe receiving language. Successive dis-persals accumulate terminological changesfrom the original forms that enable thehistorical path of successive trans-formations to be reconstructed. Unlikenames that frequently change when mov-ing, the associated genotype, in case ofbanana, is relatively fixed by vegetativepropagation. A geographical reading of thelinguistic paths illustrates the movementsof these terms and associated genotypes,providing rich evidence for the spreadof Musa cvs.More than 1,100 terms related to banana

varieties were collected during a largesurvey focused on Melanesia and SEA(Table S4). The great variety of modernterms for bananas have been describedin terms of a smaller number of cognatesets representing common origins withdistinctive, although overlapping, dis-tributions. Four of these cognate groupsprovide valuable cross-regional insightsinto ancient processes of banana domes-tication and dispersal (Fig. 4 and Fig. S5),those reflecting *muku, *punti, *qaRutay,and *baRat (32–34). Phylograms for thesekey terms have been reconstructed usingthe standard comparative method (35),implemented with bioinformatic softwareto determine likely developmental paths

malaccensis

Mlali

Buaya

FigueKhai

siamea burmannica

zebrina

banksiiSpiralVudu Beo

Katual

Bagul

Mala

Galeo

Beram

microcarpa

AAA PNGPagatau

AAA PNGWandia

AAA PNGMedja

AAA Gros Michel

AAACavendish

AABPome

M. balbisiana

errans

Fig. 2. Phylogenetic relations between AA cvs and wild acuminata subspecies. NJtree on genetic dis-similarities from 22 simple sequence repeat markers, on 41 AAw and 131 AAcv (also refer to Fig. S2): M.acuminata subsp. are in color (in gray for unclassified AAw); clusters of AAcv, identified by the name ofa representative accession, are in black. AAcv appeared as hybrids between M. acuminata subspecies asillustrated by the clusters from Spiral to Beram collected in PNG, the native area of subsp. banksii. If thecontribution of the banksii genome was still found dominant for the first clusters, it decreased rapidly,balanced by an increasing contribution of zebrina/microcarpa genome, in parallel with an increasingheterozygosity. The frequency of banksii cytoplasmic type Vф (12) decreased to the benefit of hybridforms Vα (Vф × IIα) or specifically to the Mala cluster, IIф (IIα × Vф). Contributions of these AAcv totriploids, as 2N donor (red arrows) and N donor (blue arrows), are illustrated for some AAA and AAB.

Perrier et al. PNAS | July 12, 2011 | vol. 108 | no. 28 | 11313

Dow

nloa

ded

by g

uest

on

May

12,

202

0

(36). Each distribution is suggestive ofdifferent historic processes of dispersal;local diversity of other terminologicalsets, primarily in NG and near the Hima-layas, is suggestive of long-term use ofMusaceae.The distribution of more than 40 sur-

viving reflexes of the protoform *muku is

concentrated in Wallacea and western NG(34) (Fig. 4A). This “cognate” is attestedacross a number of unrelated languagefamilies. It spreads west from NG acrossa range of modern Papuan languages inWallacea, suggesting an early distributionbefore the purported Austronesian dispersalca. 4,000 to 3,500 cal BP (34, 37). Reflexes

of *muku are found in areas (38) where pre-Austronesian linguistic ecologies survive. Inareas now dominated by Austronesian lan-guages, *muku represents terminologicalsurvival and strong resistance to change inthe face of radical language shift.The protoform *qaRutay is evident in

the Philippines with forms “*agutay” and“kelutay” for wild bananas. Hence, it islikely that the initial reflexes of *qaRutaywere generated in the Philippines region(Fig. 4C). The linguistic data suggest aninitial southward dispersal of *qaRutayfrom the Philippines into Indonesia, withlater “ripples” west into mainland Asiaand east as far as NG. The initial south-ward dispersal to Indonesia saw the*qaRutay term come into the attestedrange of *muku. Early branches of the*qaRutay dendrogram, representing re-flexes of *kaluay and *kalu, are distributedeast to Melanesia and west to the Asianmainland. An intermediate form is *kelo,which is currently attested in multiplelanguages as far west as the Indian sub-continent and east on NG, but is now al-most absent from ISEA. The last of theseterminological transformations, *loka(from *kalo via metathesis), has only re-cently begun its spread and is restrictedto locales across Sulawesi and easternIndonesia.The analysis of *baRat indicates

a probable origin in the Philippines andlater spread toward Borneo and SEA(Fig. 4D). Although not widespread,*baRat dominates in those areas inwhich it is found.In contrast to the phylogenetic depth of

*qaRutay, the lack of a hierarchy associ-ated with *punti reflects a relatively recentand rapid spread with Austronesian lan-guages in ISEA (Fig. 4B) and attestsmultiple processes of local differentiation.

Multidisciplinary Consilience onBanana DomesticationThe integrated analysis of genetic, lin-guistic, and archaeological data enablea coherent reconstruction of the majorevents in banana domestication, includinggeographical occurrence, timeline, andcultural associations.

Generation of Diploid cv Subgroups. The firstcrucial step in banana domestication wasthe hybridization between geographicallyisolated subspecies ofM. acuminata, whichwere brought into contact by people withinISEA and western Melanesia. Ethnobot-any supports the human transport of thesefertile genotypes because useful parts ofbanana plants were potentially exploitedover millennia for food, fodder, medicine,fiber, domestic uses, or construction ma-terials (39). The term “cultiwild” has beencoined to denote these potentially pre-domesticated banana plants, namely, “[a]ny

Fig. 3. Ancient banana phytoliths recovered from archaeological excavations at Kuk Swamp, PapuaNew Guinea (SEM images, A–D), and at Nkang, Cameroon (optical images, E–H). For Kuk: (A and B)dorsal and lateral view of Eumusa seed phytolith recovered from sample 5, Kuk dated to 6,990 to 6,440cal BP; (C) lateral view of another Eumusa seed phytolith from sample 5, Kuk; and (D) dorsal view ofEumusa seed phytolith recovered from sample 28, Kuk predating 3,000 cal BP. Morphotypes shown in A–C are specific to M. acuminata (Fig. S3 A, B and D). The morphotype with lobate margins shown in Doccurs in M. acuminata (Fig. S3A) and a similar morphotype occurs in Musa schizocarpa (Fig. S3 E and G).For Nkang: (E–H) multidimensional diagnostics of Musa-type volcaniform phytolith from Pit F9, Horizon7 at Nkang, dated to 2,750 to 2,100 cal BP; (E) small indentation on left side of crater rim; (F) rectangularbase, psilate surface, eccentric cone, and continuous rim; (G) processes along the edge of the base; and(H) processes along the base.

11314 | www.pnas.org/cgi/doi/10.1073/pnas.1102001108 Perrier et al.

Dow

nloa

ded

by g

uest

on

May

12,

202

0

member of a wild population, or its directfertile derivatives, whether cloned or not,growing outside of the natural habitat, orrange, of the species [or subspecies] towhich it belongs” (40). Diverse wild M.acuminata subspecies were probably sub-ject to predomesticated exploitationin their native regions. In particular, culti-wilds derived from subsp. banksii in the NGregion contributed genetically to numerouscv groups; their antiquity is attested by ar-chaeobotanical evidence at Kuk by at least6,950 to 6,440 cal BP.Domesticated parthenocarpic diploids

resulted from anthropic translocation ofcultiwilds outside of their natural range andsubsequent hybridization with local sub-species in at least three contact regions(Fig. 5A): between NG and Java (South-ern), between NG and the Philippines(Eastern), and among the Philippines,Borneo, and mainland SEA (Northern).Each of these hybridization areas has aclear linguistic association (Fig. 5B).In the Southern contact area, between

NG and Java, initial hybridizations oc-curred between subspecies of banksii andthe zebrina/microcarpa complex. This earlydispersal can be traced linguistically by thedistribution of *muku reflexes in Wallaceaand in western NG (34). This dispersalpredates the purported arrival of Austro-nesian languages in the region at approx-

imately 3,500 cal BP. This antiquity wouldbe consistent with the distribution of de-rived cvs AAcvs or AAAcvs that are notnow found in the area of origin but arewidely cultivated in Africa.The Eastern contact area is genetically

evidenced by the contribution of M. acu-minata subsp. errans from Philippines tothe cytoplasm of numerous AAcvs. ThisEastern contact area is linguistically at-tested by the Philippines’ derived Negritoterm *qaRutay, which has widely dis-persed modern reflexes. This contact areaaccords with a prehistoric linguistic trailfrom the Philippines to northwestern NG(41). M. balbisiana probably followed thistrail from South China before hybridi-zing with populations of M. acuminatasubsp. banksii.Within the Eastern contact area, the

southward translocation of M. balbisiana isattested by small populations growingalong the trail from South China to NG.Although not domesticated for edibility,M. balbisiana has been used and moved bypeople for multiple purposes (e.g., malebud, immature fruit, leaves). Moreover,several AAB and ABB triploid subgroups,with banksii genome as an A component,are absent from the northern end of thistrail, thereby indicating a southern originfor the contact area and availability of theB genome in this region. The reverse hy-

pothesis, a northward translocation of M.acuminata subsp. banksii and subsequenthybridization with M. balbisiana popula-tions, is discounted because no genetictrace of the subspecies is found in thePhilippines and further north.The Northern contact area is hypothe-

sized for the formation of several AAcvs,which contributed to important commer-cial AAA crops. Their genomes representcontributions of M. acuminata subsp.malaccensis or microcarpa, derived frommainland SEA and Borneo, and of sub-species errans from the Philippines. Thedispersal of the reflex *baRak in NorthBorneo from the prototerm *baRat in thePhilippines clearly corresponds to thegenetic hypothesis.These hybrid AAcvs were widely adop-

ted within the primary area of M. acumi-nata subspecies diversity, where they arestill commonly cultivated. Very few dis-persed outside this primary area. A strik-ing exception is the AA Mlali subgroup,which provided the parents of severalpopular triploid subgroups. Allelic fre-quencies point to this subgroup being M.acuminata subsp. banksii × zebrina hybrids,with the Southern contact area as proba-ble origin. Interestingly, this AAcv sub-group is now only grown along the eastcoast of Africa and proximal islands (i.e.,Madagascar, Zanzibar, Comoros). Theisolated occurrence of these cultivated AAdiploids far away from the region of pre-sumed origin, in which they no longeroccur, suggests an ancient transfer acrossthe Indian Ocean.

Emergence of Triploid cv Subgroups.A directconsequence of the perturbed gameteformation in hybrid AAcvs was the emer-gence of triploid genotypes AAA and AABor ABB by interspecific hybridizations.Triploidization occurred independentlyin various contact areas between diploidsand from different parental combinations(Fig. 6). Some of the triploids selectedfor cultivation underwent somaclonal var-iation under vegetative propagation toproduce the modern phenotypically char-acterized triploid subgroups. Triploidiza-tion probably started early after the AAcvemergence and is certainly still ongoing,as attested by the recent origin inferredfor the genome of several AAA in NG.Of the numerous triploid subgroups,

three are remarkable because they arelargely cultivated far from their region ofgeneration: the African AAA “Mutika Lu-jugira,” AAB “African Plantains,” andAAB “Pacific Plantains.” These triploidsare not cultivated in Asia, with the excep-tion of a few African Plantain cvs that wereprobably introduced to South Indiawith theEast African slave trade. The antiquity ofeach subgroup is attested by the extraordi-narily large number of cultivated

Fig. 4. Dendrograms and maps indicating the development and dispersal of the major banana cognatesets referred to here.Maps indicate different color-coded branches of each dendrogram; the identificationof the four cognate sets was moderated by linguistically informed judgments on the likelihood of similardevelopments being independent. *muku (A) is restricted to areas near NG, with the center of diversityalong thewest,whereas the *punti terms (B) arewidespreadwith little indicationof a“homeland.”*baRat(D) is ambiguous between a homeland in the Philippines, Borneo, or on mainland SEA. *qaRutay (C) is thetermwith themost divergent dendrogram,with anunambiguous origin in the Philippines andmany cladesbeing attested to both the east and the west, including NG and South Asia. Likely homelands for the dif-ferent terminologies are shown in yellow shading, and two locations for *qaRutay are similarly shaded forthe subclade containing*kela, found in theextremewest andextremeeast of the range. Phylogeographiesof individual clades for the different cognate sets are presented in Fig. S5.

Perrier et al. PNAS | July 12, 2011 | vol. 108 | no. 28 | 11315

Dow

nloa

ded

by g

uest

on

May

12,

202

0

morphotypes (42, 43) from only very fewimported introductions, which indicatesextremely long periods of somaclonal mu-tation within their regions of cultivation.

The AAA Mutika Lujugira subgroupis widely cultivated in the East Africanhighlands for cooking or brewing beer.The constitutive banksii and zebrina

genomes indicate the eastern side ofthe Southern contact area as origin,where genetically close AAAs are stillcultivated.The AAB African Plantains are a staple

crop in the rainforest zone of Westand Central Africa. The constitutive Agenomes were identified as banksii. AnAAB similar to African Plantains has beendiscovered under traditional cultivationby Negrito tribes in the Philippines (44).Another AAB subgroup, Laknau, withanalogous morphology and genetic simi-larity has been documented in the Philip-pines’ vicinity. These findings suggesta probable origin for the African Plantainsin the Eastern contact area between thePhilippines and NG, possibly around theCelebes Sea (45).Two independent introduction events

are proposed for these triploid subgroupsto Africa: AAA Mutika Lujugira and as-sociated AAcv to East Africa from theSouth contact area between Java and NGand AAB African Plantains from theEast contact area between the Philippinesand NG. The arrival of bananas to EastAfrica is not documented archaeologically;the reported finds at Munsa, Uganda,are problematic, primarily in terms ofchronology (46, 47). The generic namesfor African Plantain are mostly reflexes ofthe proto-Bantu term *kondo (48), in-dicating that they arrived in Africa beforethe expansion of Bantu-associated culturesfrom ca. 3,000 cal BP (49). The archaeo-logical finds at Nkang (Cameroon) 2,750to 2,100 BP conform to this timeframeand predate the spread of Austronesianlanguages from Borneo to Madagascar.Consequently, the mode of dispersal fromAsia to Africa is uncertain; terrestrial(42) and maritime routes (48) have beenproposed. It is currently unclear how, orif, the banana phytoliths dating to ca.4,000 y ago at Kot Diji in Pakistan (28)are relevant to understanding the dispersalof bananas to Africa.AAB Pacific Plantains are spread

over the whole Pacific and include theclosely relatedMaia Maoli, Popoulou, andIholena. Like African Plantains, theyare comprised of M. balbisiana and M.acuminata subsp. banksii genomes. Theabsence of genetic relationships withPhilippine cvs and the persistence inNG of genetically similar AAB formsindicate an origin in the Eastern contactarea, but east of and distinct from thatof AAB African Plantains. AAB PacificPlantains potentially originate along thenorth coast of NG or in the BismarckArchipelago (29) and were dispersedacross the Pacific by Austronesian-speaking colonists (30).The continual relocation of AAcvs by

people contributed to the developmentof numerous other triploid subgroups

South contact area

Eastcontact area

North contact area

3

2

1

A

B

Fig. 5. Genetically derived contact areas betweenM. acuminata subsp. at the origin of cultivated diploidsand corresponding linguistic evidence. (A) The three main contact areas: north among malaccensis, mi-crocarpa, and errans; east between errans and banksii; and south among banksii, zebrina, andmicrocarpa.(B) Linguistic extension for terms *muku (“1”), *qaRutay (“2”), *baRat (“3”), and their derivatives (Fig. 4).

AAA Mutika

AAB Pome

AAB others

ABB West

AAA others

AAA Cavendish

AA Mlali

AAB Plantain

ABB East

AAB Popoulou

BB

Fig. 6. Origins andmigrations of themain triploid subgroups. Plain arrows indicate long-term prehistoricmigrations of triploid cvs toAfrica andPacific islands.Graydottedarrows indicate (i) themigrationsofMlaliAAcv subgroup, which is not found in ISEA today, tomainland southeast Asia, where it contributed toAAACavendish, then to India, where it met M. balbisiana to give AAB Pome; and (ii) migrations of the Mlalisubgroup to the East African coast. Black dotted arrows indicate the route of M. balbisiana from southChina toNGover Taiwan and the Philippines, if Austronesian speakerswere instrumental in thedispersal ofthis species.

11316 | www.pnas.org/cgi/doi/10.1073/pnas.1102001108 Perrier et al.

Dow

nloa

ded

by g

uest

on

May

12,

202

0

through hybridization with local diploids.For example, the globally distributedcommercial AAAs, “Gros Michel” and“Cavendish,” derive genetically from a2N gamete of the AAcv Mlali subgroupand an N gamete of the AAcv Khai. Thenorthern origin of these AAA—aroundthe Gulf of Thailand or the South ChinaSea—implies an earlier northwestwarddispersal of Mlali from its original South-ern contact zone (Fig. 6). Further west-ward, the Mlali genome reached Indiaand is found in the Indian AAB cv sub-groups “Pome” and the geneticallyproximal “Nendra Padathi” and “Nadan”(Fig. 6). The long dispersal route of theseAA cvs from Wallacea to India is sup-ported linguistically by the continuous seriesof *qaRutay reflexes from Indonesia overmainland SEA to India (Fig. 5B).

ConclusionsThe integration of archaeology, genetics,and linguistics provides robust insightsto the history of banana domestication.Foremost, it allows us to clearly identifyseveral major stages in the domesticationhistory of Musa bananas. Early contactsbetween different Musa genotypes werecreated by human interactions, whether

by migration or exchange. The resultantinterspecific and intersubspecific hybrid-izations generated both parthenocarpicdiploids and triploids. Some of these hy-brid cvs, whether by preference or bychance, were widely adopted and dis-persed. A small number of triploids,perhaps because of the environmentaladaptability conferred by triploidy, havebeen dispersed by clonal propagationacross vast areas.Research on banana cultivation and do-

mestication is a window on often poorlyknown human/environment interactionswithin tropicalandsubtropical rainforests inthe past. Additionally, the dispersals ofbananas through wet tropical and sub-tropical regions, including around or acrossthe Indian Ocean, are indicators of inter-linked, yet predominantly local, social net-works extending from NG to West Africa;these networks are at least 2,500 y old.Research on banana domestication dem-onstrates how multidisciplinary teams,as well as the application of new meth-ods within each discipline, are essentialfor unraveling complex social processes inthe past.Bananas have contributed over several

millennia to the staple diets of numerous

peoples in the tropics and subtropics.Current global production of more than100 million tons is based on large-scalevegetative propagation of a small numberof genotypes, which derive from onlya few ancient sexual recombination events.These genetically restricted and inflexibleclones are particularly susceptible todiseases, pests, and current ecologicalchanges. The challenge for banana im-provement is to produce resistant andsterile polyploid hybrids through geneticrecombinations of fertile diploids thatmeet consumer expectations for eachcv type. The required breeding strategywill need to reproduce the sequence ofcrossings and selections that occurredminimally during the past 6,500 y, whilesubstituting punctually some genitorsfrom closely related genomes selectedfor their level of resistance to biotic andabiotic stresses. Hence, a prerequisitefor banana improvement is to re-construct as precisely as possible thedomestication pathways of the majorcv groups.

ACKNOWLEDGMENTS. The authors thank TobyWood (Monash University) and Kay Dancey (Aus-tralian National University) for their assistancewith figure production.

1. Frison E, Sharrock S (1998) Musa production around

the world - trends, varieties and regional importance.

Focus Paper 2, INIBAP Annual Report (INIBAP, Montpellier,

France), pp 42–47.2. Simmonds NW (1962) The Evolution of the Bananas

(Longmans, London).3. Sharrock S (1990) Collecting Musa in Papua New

Guinea. Identification of Genetic Diversity in the

Genus Musa, ed Jarret RL (International Network for

the Improvement of Banana and Plantain, Montpellier,

France), pp 140–157.4. Jarret RL, Gawel N, Whittemore A, Sharrock S (1992)

RFLP based phylogeny of Musa species in Papua New

Guinea. Theor Appl Genet 84:579–584.5. Ude G, Pillay M, Nwakanma D, Tenkouano A (2002)

Analysis of genetic diversity and sectional relationships

in Musa using AFLP markers. Theor Appl Genet 104:

1239–1245.6. Creste S, Tulmann NA, De Oliveira SS, Figueira A (2003)

Genetic characterization of banana cultivars (Musa

spp.) from Brazil using microsatellite markers. Euphytica

132:259–268.7. Heslop-HarrisonJS, Schwarzacher T (2007)Domestication,

genomics and the future for banana.AnnBot (Lond) 100:

1073–1084.8. Hippolyte I, et al. (2010) A saturated SSR/DArT linkage

map of Musa acuminata addressing genome re-

arrangements among banana. BMC Plant Biol 10:65–

83.9. Risterucci A-M, et al. (2009) Development and

assessment of Diversity Arrays Technology for high-

throughput DNA analyses in Musa. Theor Appl Genet

119:1093–1103.10. Jenny C, et al. (2003) Tracking the banana: Its

significance in Early Agriculture. Banana in Genetic

Diversity of Cultivated Tropical Plants (Enfield

Science, Montpellier, France), pp 99–124.11. Horry JP, Jay M (1988) Distribution of anthocyanins in

wild and cultivated banana varieties. Phytochemistry

27:2667–2672.12. Lebot V, Aradhya KM, Manshardt R, Meilleur B (1993)

Genetic relationships among cultivated bananas and

plantains from Asia and the Pacific. Euphytica 67:163–175.

13. Carreel F, et al. (1994) Evaluation de la diversitégénétique chez les bananiers diploïdes (Musa sp.).Genet Sel Evol 26:125–136.

14. Noyer JL, Causse S, Tomekpe K, Bouet A, Baurens FC(2005) A new image of plantain diversity assessed bySSR, AFLP and MSAP markers. Genetica 124:61–69.

15. Carreel F, et al. (2002) Ascertaining maternal andpaternal lineage within Musa by chloroplast andmitochondrial DNA RFLP analyses. Genome 45:679–692.

16. Perrier X, et al. (2009) Combining biological ap-proaches to shed light on the evolution of ediblebananas. Ethnobot Res Appl 7:199–216.

17. Shepherd K (1999) Cytogenetics of the genus Musa.160 (International Network for the Improvement ofBanana and Plantain, Montpellier, France).

18. Dessauw D (1988) Etude des facteurs de la stérilité dubananier (Musa spp.) et des relations cytotaxinomiquesentre M. acuminata Colla et M. balbisiana Colla. Fruits27:539–558.

19. D’Hont A, Paget-Goy A, Escoute J, Carreel F (2000) Theinterspecific genome structure of cultivated banana,Musa spp. revealed by genomic DNA in situ hybridi-zation. Theor Appl Genet 100:177–183.

20. Raboin LM, et al. (2005) Diploid ancestors of triploidexport banana cultivars: Molecular identification of 2nrestitution gamete donors and n gamete donors. MolBreed 16:333–341.

21. Horrocks M, Bedford S, Spriggs M (2009) A short noteon banana phytoliths in Lapita, immediately post-Lapita and modern period archaeological depositsfrom Vanuatu. J Archaeol Sci 36:2048–2054.

22. Ball T, Vrydaghs L, van den Houwe I, Manwaring J, DeLanghe E (2006) Differentiating banana phytoliths,wild and edible: Musa acuminata and Musabalbisiana. J Archaeol Sci 33:1228–1236.

23. Vrydaghs L, et al. (2009) Differentiating the volcaniformphytoliths of bananas: Musa acuminata. Ethnobot ResAppl 7:239–246.

24. Lentfer C (2009) Tracing domestication and cultivationof bananas from phytoliths: An update from PapuaNew Guinea. Ethnobot Res Appl 7:247–270.

25. Denham TP, et al. (2003) Origins of agriculture at KukSwamp in the highlands of New Guinea. Science 301:189–193.

26. Denham TP, Haberle S, Lentfer C (2004) New evidenceand interpretations for early agriculture in HighlandNew Guinea. Antiquity 78:839–857.

27. Mbida CM, et al. (2001) First archaeological evidenceof banana cultivation in central Africa during the thirdmillennium before present. Veget Hist Archaeobot 10:1–6.

28. Fuller D, Madella M (2009) Banana cultivation inSouth Asia and East Asia. Ethnobot. Res. Appl. 7:333–351.

29. Lentfer C, Green R (2004) Phytoliths and the evidencefor banana cultivation at the Lapita Reber-RakivalSite on Watom Island, Papua New Guinea. A PacificOdyssey: Archaeology and Anthropology in theWestern Pacific: Records of the Australian Museum29, eds Attenbrow V, Fullagar R (Australian MuseumSydney), pp 75–88.

30. Kajale MD (1989) Mesolithic exploitation of wild plantsin Sri Lanka: Archaeobotanical study at the cave site ofBeli-Lena. Foraging and Farming: The Evolution of PlantExploitation, eds Harris D, Hillman G (Unwin Hyman,London), pp 269–281.

31. Lentfer C, Pavlides C, Specht J (2010) Natural andhuman impacts in a 35 000-year vegetation history incentral New Britain, Papua New Guinea. Quat Sci Rev29:3750–3767.

32. Donohue M, Denham TP (2009) Banana (Musa spp.)domestication in the Asia-Pacific region: Linguisticand archaeobotanical perspectives. Ethnobot ResAppl 7:293–332.

33. Zorc RD(1995)AglossaryofAustronesian reconstructions.Comparative Austronesian Dictionary, ed Tryon DT(Mouton de Gruyter, Berlin), Part I, pp 1105–1197.

34. Denham TP, Donohue M (2009) Pre-Austronesiandispersal of banana cultivars west from New Guinea.Archaeol Ocean 44:18–28.

35. Joseph BD, Janda RD, eds. (2003) Handbook of His-torical Linguistics (Blackwell, London).

36. Saitou N, Nei M (1987) The neighbor-joining method:A new method for reconstructing phylogenetic trees.Mol Biol Evol 4:406–425.

37. Blust R (2009) The Austronesian Languages (PacificLinguistics, Canberra, Australia).

38. Nichols J (1992) Linguistic Diversity in Space and Time(Univ Chicago, Chicago).

Perrier et al. PNAS | July 12, 2011 | vol. 108 | no. 28 | 11317

Dow

nloa

ded

by g

uest

on

May

12,

202

0

39. Kennedy J (2009) Bananas and people in the homelandof genusMusa: Not just pretty fruit. Ethnobot Res Appl7:179–197.

40. De Langhe E, Vrydaghs L, de Maret P, Perrier X,Denham TP (2009) Why bananas matter. EthnobotRes Appl 7:165–177.

41. Mahdi W (1994) Some Austronesian maverick proto-forms with culture-historical implications: I. OceanLinguist 33:167–229.

42. DeLangheE(2007)Theestablishmentoftraditionalplantaincultivation in the African rain forest: A working hypothesis.

RethinkingAgriculture, eds DenhamTP, Irarte J, Vrydaghs L(Left Coast Press, Walnut Creek, CA), pp 361–370.

43. Ploetz RC, Kepler AK, Daniells J, Nelson SC (2007)Banana and plantain – an overview with an emphasison Pacific island cultivars [Musacea (banana family)].Species Profiles for Pacific Island Agroforestry (Per-manent Agriculture Resources, Ho-lualoa, Hawaii).

44. De Langhe E, Valmayor RV (1979) French plantains inSoutheast Asia. IBPGR Newsletter for SE Asia 3–4.

45. De Langhe E (2009) Relevance of banana seeds inarchaeology. Ethnobot. Res. Appl. 7:271–281.

46. Lejju BJ, Robertshaw P, Taylor D (2006) Africa’s earliest

bananas? J Archaeol Sci 33:102–113.47. Neumann K, Hildebrand E (2009) Early bananas in

Africa: The state of the art. Ethnobot. Res. Appl. 7:

353–362.48. Blench R (2009) Bananas and plantains in Africa: Re-

interpreting the linguistic evidence. Ethnobot. Res.

Appl. 7:363–380.49. Vansina J (1990) Paths in theRainforests (UnivWisconsin,

Madison, WI).

11318 | www.pnas.org/cgi/doi/10.1073/pnas.1102001108 Perrier et al.

Dow

nloa

ded

by g

uest

on

May

12,

202

0