The Structure of Morphological and Genetic Diversity in Natural Populations of Caricapapaya (Caricaceae) in Costa Rica

Jennifer E. Brown1, Jenise M. Bauman1, Joseph F. Lawrie1, Oscar J. Rocha2, and Richard C. Moore1,3

1 Department of Botany, Miami University, 316 Pearson Hall, Oxford, Ohio 45056, U.S.A.

2 Department of Ecological Sciences, Kent State University, PO Box 5190, Kent, Ohio 44242, U.S.A.

ABSTRACT

In Costa Rica, dioecious Carica papaya has been observed growing in disturbed areas and within secondary lowland forests. Such popula-tions can serve as a reservoir of genetic and morphological diversity for this tropical fruit crop. We quantify the levels and patterns ofthe diversity of naturally occurring populations of C. papaya and address the demographic history of these populations. We measured 29vegetative and reproductive morphological traits in situ from 252 plants and found significant heterogeneity among regional populationsin the majority of these traits. Significant variation was found among regional populations with respect to fruit size and shape, withplants in the Nicoya Peninsula possessing smaller, less fleshy fruit, a characteristic of previously described wild populations of papaya.We then assessed the levels and patterns of genetic diversity in 164 plants from natural populations and 20 cultivars. Natural populationsexhibit a deficiency of heterozygotes; however, this is much more pronounced within the cultivars. Although there is little genetic differ-entiation among natural populations, we did find evidence of cryptic genetic population structure. Analyses of population demographyindicate that these natural populations have undergone a recent genetic bottleneck, followed by recent population expansion, possiblypromoted by the transformation of the Costa Rican landscape for agricultural use.

Abstract in Spanish is available in the online version of this article.

Key words: dioecy; microsatellite; population expansion; population structure; tropical fruit crop.

THE MESOAMERICAN LANDSCAPE, INCLUDING THE REGIONS OF

SOUTHERN MEXICO TO EASTERN PANAMA, have undergone a dra-matic transformation due to fragmentation of the landscape andthe conversion of natural areas for agricultural use, leading to theintroduction of various conservation strategies geared toward pre-serving and/or restoring the biodiversity of this region (see DeCl-erck et al. 2010). In addition to such efforts, secondary forestgrowth following the abandonment of agricultural lands can alsocontribute to the restoration of native biodiversity (Chazdon et al.2009). In Costa Rica, the last 20 yr has seen a resurgence ofsecondary forest growth following the abandonment of cattle pas-tures caused by declining beef prices and increased urbanization(Arroyo‐Mora et al. 2005, Calvo‐Alvarado et al. 2008). As part ofthis natural restoration, pioneer species will be among the first toestablish themselves along the borders of these abandoned lands.One of these early successional species, papaya (Carica papaya), iscommonly found in the evolving Costa Rican landscape.

In Costa Rica, small‐fruiting papayas have been observedgrowing in secondary lowland forests, abandoned fields andregions along the sides of roads. These populations are primarilydioecious, unlike cultivars that are mostly gynodioecious. Manyquestions exist as to whether these papayas are naturally occur-ring wild plants or are feral escapees derived from cultivars, withsome scientists distinguishing between the two (Badillo 1993,

Manshardt & Zee 1994). The morphological descriptions of wildpopulations from Mexico through Guatemala (Manshardt & Zee1994) are shared with those of Costa Rican natural populations:dioecious populations with females that produce small, seedyfruits with a thin mesocarp. Feral populations have beenobserved farther south into Central and South America and aredescribed as having similar morphological traits to those of culti-vated plants, with populations containing hermaphroditic individ-uals and producing larger fruits with fewer seeds (Manshardt &Zee 1994). Based on morphological observations, Manshardt andZee (1994) conclude that the farther west and south of the geo-graphic range of wild papaya, the more introgression hasoccurred among wild and domesticated papaya.

Previous research has focused upon characterizing thegenetic (Pérez et al. 2007) and/or morphological diversity ofpapaya growing within natural areas (d'Eeckenbrugge et al. 2007).Pérez et al. (2007) analyzed genetic data from 15 microsatellitemarkers using papaya from 12 countries, many of those from theCaribbean and Central America, and observed a high frequencyof rare alleles within Costa Rican germplasm, although littlegenetic differentiation existed among samples collected from theCaribbean and Pacific coasts. Only one study (d'Eeckenbrugge etal. 2007) has compared genetic data (isozyme data) with that ofmorphological data from plants within natural populations fromCosta Rica. Although a wide range of morphological diversitywas observed among the 32 feral individuals sampled, very littlegenetic diversity existed among them (d'Eeckenbrugge et al.2007).

Received 26 July 2010; revision accepted 17 January 2011.3Corresponding author; e‐mail: moorerc@muohio.edu

ª 2011 The Author(s) 179

Journal compilation ª 2011 by The Association for Tropical Biology and Conservation

BIOTROPICA 44(2): 179–188 2012 10.1111/j.1744-7429.2011.00779.x

The history of the Costa Rican landscape may haveimpacted the demographic history and genetic diversity of papaya.The southern Costa Rican landscape has been altered by humansfor at least 1400 yr before present for agricultural purposes(Anchukaitis & Horn 2005). More recently, a large majority ofCosta Rica has been deforested since the 1920s due to the gov-ernment policies that promoted virgin areas to be used for agri-culture (Nygren 1995). These activities may have altered thelandscape to promote the expansion of papaya due to its succes-sional nature.

The goal of this study is to characterize the genetic andmorphological diversity within and among natural populations ofpapaya in Costa Rica and to address possible demographic sce-narios underlying this diversity. We sampled morphological andgenetic diversity from over 180 papaya individuals growing in fiveregions of Costa Rica using 20 microsatellite markers andaddressed the following questions: (1) to what degree do naturalpapaya populations vary morphologically? (2) What are the levelsand structure of genetic diversity within and among naturalpapaya populations and cultivars? And (3), to what extent has thegenetic diversity of natural papaya populations been shaped bypast demographic processes?

MATERIALS AND METHODS

PLANT MATERIAL AND STUDY SITE.—Young leaf tissue was collectedfrom 252 papaya plants consisting of 116 female, 126 male and 7hermaphroditic individuals, in the lowlands of five regions ofCosta Rica: Caribbean coast, northwest Pacific coast, Nicoya Pen-insula, central Pacific coast and southwest Pacific coast (Fig. 1;Table S1). Collections from different regions were subdivided intoten operational populations, defined as groups of papaya plantsseparated by at least 10 km (Fig. 1). Plants were found growingwithin 100 m of a road (most were visible from the road) andplants that were found growing on or adjacent to a dwelling werenot collected. These groups of plants were considered to belongto natural populations. Samples were collected and preserved insilica gel before DNA extraction. Cultivar samples were donatedby Ray Ming of the University of Illinois, Qingyi Yu of theHawaii Agriculture Research Center and the U.S. National PlantGermplasm System (NPGS; Table S2).

MORPHOLOGICAL DATA COLLECTION.—A combined total of 29 veg-etative and reproductive characters were measured in situ. Werestricted our collection to those plants that were in flower and/

FIGURE 1. Map of Costa Rica showing location of individuals within operational populations indicated by numbers and regions indicated by shapes (square,

Caribbean [Populations 1, 2]; circle, northwest Pacific [Populations 3, 4, 5]; triangle, Nicoya Peninsula [Population 6]; diamond, central Pacific [Populations 7, 8];

oval, southwest Pacific [Populations 9, 10]).

180 Brown, Bauman, Lawrie, Rocha, and Moore

or fruiting to ensure an accurate assignment of gender. Thesemorphological traits consisted of 20 qualitative (e.g., shape andcolor) and nine quantitative traits (International Board for PlantGenetic Resources [IBPGR] 1988; Table S3). Quantitative mea-surements were obtained from one mature leaf, flower or fruitper individual. Ripe fruit were chosen for measuring if available,otherwise, the most mature (oldest) fruit was chosen.

MORPHOLOGICAL DATA ANALYSIS.—Morphological data were statis-tically analyzed using JMP 7.0 software (SAS Institute Inc., Cary,North Carolina, U.S.A.). For qualitative traits, a contingency anal-ysis was performed for each trait to test the null hypothesis thatphenotypes are independent of region and significance assessedby the likelihood ratio v2 and Pearson v2 tests. A correspondenceanalysis was performed for each qualitative trait that showed sig-nificant departure from homogeneity.

Quantitative data were transformed on a log10 +1 scale tonormalize distribution and equalize variances and then used in alinear discriminant analysis. We tested the null hypothesis thatthere were no differences in morphological data among regionusing Wilks’ k test statistic. An analysis of variance (ANOVA)followed by Tukey’s HSD post‐hoc test was conducted on fruit(diameter, mesocarp thickness and diameter: mesocarp ratio) andleaf (length: width ratio) characteristics.

MICROSATELLITE ANALYSIS.—Genomic DNA was extracted usingQIAGEN DNeasy Plant Mini Kits (QIAGEN, Germantown,Maryland, U.S.A.). Twenty microsatellite markers distributedacross all nine papaya chromosomes were selected from thosepreviously described by Chen et al. (2007; Table S4). Productswere amplified using the conditions previously described by Chenet al. (2007). Fluorescently labeled primers were designed basedon the Applied Biosystems Inc. Dye Set G5 (Applied BiosystemsInc., Foster City, California, U.S.A.).

Following amplification, PCR products from four differentfluorophore‐labeled reactions were combined and size‐fractionedby capillary electrophoresis on an ABI 3730 DNA Analyzer(Applied Biosystems Inc.). GeneMapper 3.7 (Applied BiosystemsInc.) was used to identify alleles and compile the marker data.Alleles were binned using FlexiBin (Amos et al. 2006).

GENETIC DIVERSITY ANALYSES.—Genetic diversity within the tenpopulations and cultivars for each of 20 microsatellite loci wasmeasured using Arlequin 3.11 (Excoffier et al. 2005). Geneticdiversity was described by the number of alleles per locus (A),the expected heterozygosity (HE; Levene 1949) and the observedheterozygosity (HO) per locus. Deviations from Hardy–Weinbergequilibrium (HWE) were calculated using the exact test of Guoand Thompson (1992). HWE probabilities for each of the 20 lociper population were adjusted using the Bonferroni correction formultiple comparisons (Olejnik et al. 1997). A log‐likelihood ratiotest of genetic disequilibrium among all markers within eachregional population (a total of 759 pairwise comparisons) wasperformed using GENEPOP (Raymond & Rousset 1995, Rous-set 2008). We found only two significant associations (Bonferroni

P<0.05): marker P6K946CC (linkage group [LG] 8) andP3k7483A (LG 2) in the northwest Pacific region; and markerP8K209CC (LG 9) and ctg27CC (LG 3) in the southwest Pacificregion. With these two exceptions, our markers meet the criterionof linkage equilibrium for genetic analyses in the regional popula-tions.

POPULATION STRUCTURE ANALYSES.—The distribution of geneticvariation was calculated using an analysis of molecular variance(AMOVA) implemented in Arlequin 3.11 (Excoffier et al. 2005).Fixation indices were derived from the analysis of individual dataand tested for statistical significance through the use of nonpara-metric permutations (Weir & Cockerham 1984, Schneider et al.1997). Pairwise FST values were calculated using nonparametricpermutations to calculate significance among populations and cul-tivars (Excoffier et al. 2005). A permutation test of alleles sizes inthe program SPAGeDI 1.3 (Hardy & Vekemans 2002, Hardyet al. 2003) indicated allele size differences resulting from stepwisemutations do not significantly contribute to population differenti-ation (10,000 permutations; P=0.11); thus, we only report FSTvalues, a measure of population differentiation independent ofmicrosatellite allele size.

Genetic structure was further assessed through the use ofBayesian model‐based clustering in the program STRUCTURE2.3.1 (Pritchard et al. 2000). All STRUCTURE runs used a burn‐in length of 20,000, followed by 100,000 repetitions after burn‐in.In order to determine the number of subpopulation clusters (K),the methods of Evanno et al. (2005) were used. The modal valueof the DK was used as an indicator of the number of ancestralpopulation clusters (Evanno et al. 2005).

COMBINED GENETIC AND MORPHOLOGICAL STRUCTURE ANALYSES.—We conducted a discriminant analysis and an ANOVA in JMP7.0 (SAS Institute Inc.) using the ancestry assignments from theSTRUCTURE analysis of natural populations. Each individualwas assigned to one of three population clusters (K=3; see resultsfor choice of K) according to the highest ancestry proportionassigned by STRUCTURE. For the discriminant analysis, we usedthe canonical variables created during the previous discriminantanalyses of both male and female quantitative traits. An ANOVAwas conducted on fruit and leaf measurements in order to deter-mine whether statistical differences existed among the STRUC-TURE clusters. The ANOVA analyses were followed by aTukey's HSD post‐hoc test.

POPULATION DEMOGRAPHY ANALYSES.—Two indices of populationdemography based on microsatellite data were calculated: theGarza–Williamson index M (Garza & Williamson 2001), and theimbalance index b (Kimmel et al. 1998, King et al. 2000). M, themean ratio of the number of alleles to the range of allele sizes,was used to measure the possibility of a past genetic bottleneckand was calculated using Arlequin 3.11 (Garza & Williamson2001, Excoffier et al. 2005). The index b measures the imbalanceof variance‐based estimates and homozygosity‐based estimates ofh, the population mutation parameter (Kimmel et al. 1998). The

Morphological and Genetic Diversity of Natural Papaya 181

log of the index b (ln b) was estimated as the log ratio of themeans, ln b1 (equation (7) in King et al. 2000) and the mean ofthe log ratios, ln b2 (equation (8) in King et al. 2000). A permu-tation test of allele sizes (see ‘Population structure analyses’) wasinsignificant, suggesting that our microsatellite loci do not followa strict stepwise mutation model (SMM). However, M assumes atwo‐phase SMM of microsatellite evolution, allowing for a pro-portion of non‐one‐step mutation (Garza & Williamson 2001),while b utilizes a generalized SMM, which allows for asymmetricand arbitrary changes in allele size (Kimmel & Chakraborty 1996,Kimmel et al. 1998). As it is generally accepted that microsatelliteloci deviate from the strict SMM, with new alleles generated bythe addition and/or subtraction of single and, less frequently,multiple repeat units, the nonstrict SMMs assumed for these testsseem reasonable (Estoup et al. 2002, Buschiazzo & Gemmell2006). We also tested for past population expansion using thewithin‐locus variability statistic, k, and the between‐locus variabil-ity statistic, g (Reich & Goldstein 1998, Reich et al. 1999) asimplemented in the Microsoft Excel macro application by Bìlgìn(2007).

RESULTS

SIGNIFICANT VARIATION IN QUALITATIVE AND QUANTITATIVE

MORPHOLOGICAL TRAITS EXISTS AMONG REGIONS WITHIN COSTA

RICA.—We assessed the levels of morphological variation in eightqualitative vegetative traits from 241 papaya plants within fiveregions of Costa Rica (Fig. 1; Tables S1 and S3). Six of the eightvegetative traits were significantly heterogeneous among theregions (Table S5): petiole color (P<0.001), leaf shape (P<0.001),mature leaf teeth shape (P<0.01), petiole sinus shape (P<0.001),stem color (P<0.001) and stem pigmentation (P<0.001). Infemales, significant levels of heterogeneity were observed for fourof nine qualitative reproductive traits (Table S5): flower color(P<0.05), fruit shape (P<0.05), fruit central cavity shape (P<0.05[log‐likelihood only]) and seed color (P<0.001). In males, signifi-cant heterogeneity existed for two of three reproductive traits(Table S5): inflorescence size (P<0.001) and corolla lobe color(P<0.05 [log‐likelihood only]).

We constructed a correspondence analysis plot for each sig-nificantly heterogeneous trait to visualize differences in the distri-butions of phenotypes for those traits among regions (Fig. S1).In these plots, the distance among regional points indicates thedegree of divergence in the phenotypic distributions among theregions. Some regions are clearly differentiated from others withrespect to certain traits. For example, in the Caribbean (Region1), plants tended to have broader and less dissected leaves andfruits with black seeds (Figs. S1B and J, respectively). Plants inthe northwest Pacific (Region 2) had more curved leaf teeth (Fig.S1C). Several characters differentiate the Nicoya Peninsula(Region 3) from other regions: female plants had a higher fre-quency of yellow flowers and produced fruit with a more irregu-lar fruit cavity shape; male plants had more yellow flowercorollas than other regions (Figs. S1G, I and L, respectively).Plants in the central Pacific (Region 4) had higher frequencies of

red–purple stems and oblong/lengthened fruit (Figs. S1E and H,respectively). The southwest Pacific (Region 5) had a greater pro-portion of plants with stem pigmentation distributed in the lowerpart of the stem as well as small male inflorescences (Figs. S1Fand K, respectively).

We performed a discriminant analysis to gauge morphologi-cal differentiation for quantitative traits among the five CostaRican regions (Fig. S2). For females, the traits that contributedmost to the discrimination included petiole length, fruit diameterand fruit skin thickness (Tables S6 and S7). Plants from the Ni-coya Peninsula (Region 3) differentiate from all other regionalpopulations, with the exception of the northwest Pacific (Region2; Fig. S2A; P=0.002). Male traits that contributed most to thisdiscriminant analysis were leaf length, leaf width and flower size(Tables S6 and S8). Based on this analysis, the morphology ofmale plants in the northwest Pacific (Region 2) and central Pacific(Region 4) was distinguished from that of the southwest Pacific(Region 5; Fig S2B; P<0.001).

ANOVA revealed significant variation existed among regionsfor the quantitative measurements of fruit and leaf size (Table 1).Fruit diameter, mesocarp thickness and the ratio of these twotraits varied significantly across regions (P<0.001 to P<0.01).Fruit within the Nicoya Peninsula had a significantly smallerdiameter, thinner mesocarp and a larger ratio of diameter tomesocarp thickness than fruit from the other regions. Fruit inthe southwest Pacific had significantly thicker mesocarp than allother regions excepting those of the central Pacific.

A DEFICIENCY OF HETEROZYGOSITY EXISTS IN NATURAL POPULATIONS

AND CULTIVARS.—From the 192 accessions tested, 184 accessionsconsisting of 164 Costa Rican samples and all 20 cultivars, ampli-fied successfully with the 20 microsatellite loci. Microsatellite lociwere relatively diverse across the natural papaya populations andcultivars; the number of observed alleles per locus ranged from 6to 25, with a mean of 11.6 alleles per locus (Table S9). Mean HE

per locus and cultivars ranged from 0.51 (Population 3) to 0.64(cultivars; Table S9). However, the mean HO per locus only ran-ged from 0.14 (cultivars) to 0.45 (Population 1; Table S9). Thedeficiency of observed to expected heterozygotes, F, ranged from0.25 to 0.42 for the operational populations, and 0.75 for the cul-tivars (Table S9). Although we observed an overall deficiency inheterozygosity within natural populations, only a minority of theloci within these populations deviated significantly from HWE (0–45% of loci deviating from HWE at a Bonferroni P value<0.05;Table S9). In contrast, the cultivars showed significant deviationfrom HWE (P<0.05) in 85 percent of the loci.

LIMITED GENETIC VARIATION OR DIFFERENTIATION EXISTS AMONG

THE NATURAL PAPAYA POPULATIONS.—We used an AMOVA toassess the distribution of genetic variation within and amongpopulations, as well as among regions (Table 2). The AMOVAresults indicated that little of the genetic variation exists amongpopulations within regions (FSC=0.022) or among regions(FCT=0.020). The inbreeding coefficient, FIS, was calculated as0.312 (P<0.005). Pairwise estimates of genetic differentiation

182 Brown, Bauman, Lawrie, Rocha, and Moore

(FST) were low to moderate among natural populations, rangingbetween �0.008 and 0.107, and moderately high among naturalpopulations and the cultivars, ranging from 0.095 to 0.180 (Table3). The majority of the pairwise FST comparisons were significant(P<0.05 to P<0.001).

LIMITED GENETIC STRUCTURE EXISTS AMONG THE NATURAL

POPULATIONS.—We detected three genetically different subpopula-tions, or clusters, within natural populations using STRUCTURE(K=3; Figs. 2A and B). Many individuals within the natural popu-lations were considered to have admixed genetic structure, withsome proportions of each genome resulting from all three of thesubpopulations. Although the genetic clusters did not correspondperfectly with our geographic regions, some trends did emerge.Cluster 1 was comprised mostly of individuals from northwestPacific (Populations 3, 4 and 5), with a number of individualsfrom the central Pacific (Populations 7 and 8). Cluster 2 wascomprised mostly of individuals from the Nicoya Peninsula andsurrounding regions (Populations 4, 6 and 7), and Cluster 3 wasprimarily composed of plants from the Caribbean and southwestPacific (Populations 1, 2, 8, 9 and 10; Fig. 4B; Table S10).

We also analyzed the genetic structure of the natural acces-sions with the cultivars using STRUCTURE and identified two

subpopulations (K=2; Figs. 2A and C). At this level, the naturalpopulations were assigned to primarily Cluster 1 and the cultivarswere assigned to Cluster 2 (Fig. 2C). We observed that many ofthe natural population, individuals had different degrees of sharedgenetic ancestry with the cultivars, and may represent feral plantsor introgression of feral plants in natural populations throughhybridization. In particular, some individuals from Populations 3,5, 7, 8 and 9 had >80 percent of their ancestry attributed toCluster 2 (Fig. 2C; Table S11). Conversely, two cultivar accessionsshare >90 percent ancestry with Cluster 1; one of these is anunimproved cultivar from Rio Terraba, Costa Rica (928_M),while the other is an Australian dioecious cultivar (DREW).

GENETIC AND MORPHOLOGICAL STRUCTURES ARE RELATED WITHIN

THE NATURAL PAPAYA POPULATIONS.—We performed a discriminantanalysis of the canonical variables for female and male quantita-tive traits using the genetic ancestry assignments from theSTRUCTURE analysis of natural populations in order to assessthe association between genetic and morphological structure (Fig.S3). Wilks' k was significant for both male (P=0.022) and female(P=0.038) traits, suggesting significant morphological differencesexist among the STRUCTURE population clusters. For femaleplants, Clusters 2 and 3 were the most distinct (Fig. S3A; Table

TABLE 1. Mean values of fruit diameter, mesocarp thickness, fruit diameter to skin thickness ratio and leaf length to width ratio (± SE) for natural populations with analysis of

variance (ANOVA) and Tukey's HSD post‐hoc significance levels indicated.

Region

Mean fruit diameter

(cm)a**Mean mesocarp thickness

(cm)***Mean ratio fruit diameter: mesocarp

thickness***Mean ratio leaf blade width:

length

Caribbean 7.496 ± 0.401 a 1.577 ± 0.112 b 4.946 ± 0.358 b 0.967 ± 0.010 a

Central Pacific 7.570 ± 0.458 a 1.700 ± 0.128 ab 4.897 ± 0.408 b 0.947 ± 0.011 a

Nicoya 5.192 ± 0.568 b 0.731 ± 0.158 c 8.289 ± 0.506 a 0.927 ± 0.011 a

Northest Pacific 6.760 ± 0.528 ab 1.407 ± 0.147 b 4.948 ± 0.471 b 0.932 ± 0.010 a

Southwest

Pacific

9.044 ± 0.682 a 2.256 ± 0.190 a 4.019 ± 0.608 b 0.952 ± 0.017 a

aLetters following means indicate significance levels based on Tukey's HSD post‐hoc test of significance.

ANOVA significance level: *P<0.05, **P<0.01, ***P<0.001.

TABLE 2. Analysis of molecular variance results assessing natural papaya populations using individual data. FCT, genetic correlation among regions; FSC, genetic correlation among

populations within regions; FIS, genetic correlation among individuals within populations; FIT, genetic correlation of individuals within the total population.

Source of variation df Sum of squares Variance components Percentage of variation F‐statistics

Among regions 4 72.758 0.10832 1.99 FCT=0.020*

Among populations within regions 5 50.459 0.11561 2.11 FSC=0.022***

Among individuals within populations 154 1054.60 1.62831 29.89 FIS=0.312***

Within individuals 164 589 3.59146 66.01 FIT=0.340***

Total 372 1766.82 5.44369

***P<0.001,**P<0.01,*P<0.05.

Morphological and Genetic Diversity of Natural Papaya 183

S12). For male plants, Clusters 1 and 3 were most differentiated(Fig. S3B; Table S13).

There was significant variation in fruit characteristics amongthe STRUCTURE clusters (ANOVA, P<0.05; Table 4). Specifi-cally, fruit diameter within Cluster 3 (southwest Pacific and Carib-bean) was significantly greater than that of Clusters 1 and 2.Cluster 3 was also differentiated from Cluster 2 (primarily Ni-coya) by having a thicker mesocarp and lower ratio of diameterto mesocarp thickness.

NATURAL POPULATIONS HAVE UNDERGONE A GENETIC BOTTLENECK

FOLLOWED BY A RECENT POPULATION EXPANSION.—We used theGarza–Williamson index, M, to assess the possibility of whethera past genetic bottleneck had occurred within the natural papayapopulations. The analyses of Garza and Williamson (2001) indi-cate that M values <0.68 support the occurrence of a past reduc-tion in population size. The M value for all of the Costa Ricanregions combined was 0.54, supporting a past genetic bottleneck.Within the five Costa Rican regions, the M values ranged from0.52 to 0.58 (Table S14). The cultivars also had an M value of0.52 (Table S14).

Population expansion was assessed through the use of theimbalance index, b (Kimmel et al. 1998, King et al. 2000), thebetween‐locus variability statistic, g (Reich & Goldstein 1998,Reich et al. 1999) and the within‐locus variability statistic, k(Reich & Goldstein 1998, Reich et al. 1999). Estimates of theimbalance index (ln b1 and ln b2) that are >1 indicate arecent reduction in population size followed by population expan-sion (Kimmel et al. 1998, King et al. 2000). Consistent with thisprediction, estimates of the imbalance index ranged from 2.11 to2.37 for ln b1 and 2.58 to 2.78 for ln b2 for all of the fiveCosta Rican regions (Table S14). The values of ln b1 andln b2 for the cultivars were also >1, being 2.22 and 3.04,respectively (Table S14).

We were unable to detect population expansion in naturalpapaya populations using the g and k statistics (Reich & Gold-stein 1998, Reich et al. 1999). The g values ranged from 1.26 to1.56 for the natural populations, whereas values <1 are consid-ered indicative of population expansion (Table S14; Reich et al.1999, King et al. 2000). In contrast, the cultivars had a g valueof 0.93 (Table S14). Neither the natural papaya populations northe cultivars produced significant k values (0.10<P<0.70; TableS14). However, unlike the imbalance index, both of these testshave low power to detect very recent population expansions(Reich et al. 1999, King et al. 2000). Thus, this disparity sup-ports a more recent expansion of natural papaya populations inCosta Rica.

DISCUSSION

TO WHAT DEGREE DO NATURAL PAPAYA POPULATIONS IN COSTA RICA

VARY MORPHOLOGICALLY?.—We observed a wide range of morpho-logical diversity throughout the natural populations of papaya inCosta Rica similar to previous reports (d'Eeckenbrugge et al.2007). Most vegetative traits varied among regions, while repro-ductive variation was limited primarily to seed and flower colorand fruit shape. Some clear regional differences existed for mostof these traits; in particular, the Nicoya Peninsula region is differ-entiated from other regions with respect to male and female flo-ral color and fruit central cavity shape. Likewise, our analyses ofthe female quantitative traits indicated that plants from the Ni-coya Peninsula were the smallest of all regions, with a thin meso-carp. Our observations corroborate those of d'Eeckenbrugge etal. (2007), who noted that plants within the Nicoya Peninsula, aswell as populations near Playa Herradura and Puente Tarcoles,which correspond with our operational population 7, resembledthe wild papaya descriptions of Manshardt and Zee (1994) fromsouthern Mexico.

TABLE 3. Pairwise FST values and significance of all operational populations and cultivars. Shadings indicate degree of genetic differentiation: white, low (FST<0.05); light gray,

moderate (0.05<FST<0.10); dark gray, high (FST>0.10).

Population 1 2 3 4 5 6 7 8 9 10 Cultivar

1 — ** *** *** *** *** *** *** *** ***

2 0.037 — ** *** *** *** *** *** ***

3 0.046 0.007 —

4 0.041 0.040 �0.008 — * *** *** *** ***

5 0.044 0.031 0.028 0.014 — *** ** *** *** *** ***

6 0.078 0.072 0.041 0.012 0.042 — * *** *** *** ***

7 0.057 0.050 0.003 0.025 0.044 0.022 — * *** *** ***

8 0.040 0.043 0.014 0.040 0.040 0.064 0.026 — ** * ***

9 0.073 0.073 0.070 0.091 0.088 0.107 0.068 0.046 — *** ***

10 0.040 0.063 0.035 0.060 0.051 0.090 0.054 0.029 0.065 — ***

Cultivar 0.129 0.133 0.132 0.139 0.128 0.180 0.147 0.115 0.095 0.109 —

***P<0.001,**P<0.01,*P<0.05.

184 Brown, Bauman, Lawrie, Rocha, and Moore

The significant morphological trends among geographicregions are in contrast to the relatively little genetic variationobserved among regions based on FST estimates. Nonetheless,when we grouped individuals according to one of three geneti-cally distinct populations as defined by our STRUCTURE analy-sis, we did find significant morphological differences in femaleand male quantitative traits. In particular, fruits from Clusters 1(northwest Pacific) and 2 (Nicoya Peninsula and nearby regions)had generally smaller fruits and thinner mesocarps than thosefrom Cluster 3 (Caribbean/southwest Pacific). Thus, many of thereproductive traits we measured appear to segregate with geneticstructure, suggesting there may be an underlying genetic compo-nent to these differences.

WHAT ARE THE LEVELS AND STRUCTURE OF GENETIC DIVERSITY

WITHIN AND AMONG NATURAL PAPAYA POPULATIONS AND

CULTIVARS?.—We estimate higher levels of genetic diversity in thenatural populations than previously reported based on isozymeanalysis (Morshidi et al. 1995, d'Eeckenbrugge et al. 2007). Forexample, d'Eeckenbrugge et al. (2007) report an average HE of0.44 and 0.37 for Caribbean and Pacific papaya populations,respectively, whereas we find an average HE of 0.59 and 0.57 forthese regions. Furthermore, we find relatively high allelic diversityamong all populations. While we generally observed a deficiencyof heterozygotes in natural populations, the majority of loci forany one population did not deviate significantly from HWE. Incontrast, cultivars had a much more pronounced heterozygotedeficiency and the majority of loci did significantly deviate fromHWE, consistent with its history as a domesticated crop. Six-teenth century Spanish explorers were responsible for the initialspread of this crop beyond its native Mesoamerican distribution(Storey 1969). It is possible that 500 yr of selective breeding,particularly for fruit size, shape and color, inbreeding of predomi-nantly gynodioecious cultivars and a past domestication bottle-neck have contributed to this departure from HWE.

We observed low to moderate levels of genetic differentia-tion among natural papaya populations and among regions withinCosta Rica. While our STRUCTURE analysis identified threegenetically distinct population clusters, there was genetic admix-ture both within populations and within individuals. Nevertheless,some populations were more genetically divergent based on ourpairwise FST analyses. In particular, the Nicoya Peninsula popula-tion (Population 6) was more highly differentiated from theCaribbean populations than both the northwest Pacific (Popula-tions 3, 4 and 5) and some southwest populations (Population 8),consistent with the more geographically isolated location of thispopulation. Also, the Nicoya population had the highest degreeof genetic differentiation from the cultivars (FST=0.18). Thisresult is consistent with our observation that the papaya growingin the Nicoya Peninsula had more ‘wild’ characteristics than otherpopulations.

In contrast, the southwest Pacific populations (9 and 10)exhibited the highest degree of genetic differentiation relative toother natural populations (pairwise FST values between 0.03 and0.11) and the lowest amount of genetic differentiation from culti-vars (FST=0.095–0.11). Plants in these regions tended to have lar-ger fruits with more domestication‐like traits (i.e., thick mesocarp,and a large regularly shaped central cavity), perhaps indicative ofa more feral origin of plants in this population.

One apparently conflicting observation is the combinationof low genetic correlation among populations (FST) and relativelyhigh genetic correlation among individuals (FIS). Low genetic dif-ferentiation among populations is consistent with the dioeciousmating system and perennial life history of papaya; however, therelatively high inbreeding coefficient (FIS=0.31) is expected forpopulations with a mixed mating system or selfing (Loveless &Hamrick 1984, Hamrick & Godt 1996, Duminil et al. 2007). Asnatural papaya populations are almost exclusively dioecious, non-

FIGURE 2. STRUCTURE estimates of cryptic population structure of nat-

ural populations and cultivars. (A) Calculation of the second‐order rate of

change (DK), determined by the modal peak. The modal peak for natural

populations (open circles) is at K=3 whereas the modal peak for natural pop-

ulations with cultivars (filled circles) is at K=2. (B) STRUCTURE plot of

ancestral subpopulations from the natural populations, with shading represent-

ing the three population clusters and each line representing a single individual,

with the percent of its genome identified by the y‐axis (Cluster 1: dark gray;

Cluster 2: medium gray; Cluster 3: light gray). (C) STRUCTURE plot of two

ancestral subpopulations for the natural populations and cultivars (Cluster 1:

medium gray; Cluster 2: light gray).

Morphological and Genetic Diversity of Natural Papaya 185

random mating in these populations is most likely due to bipa-rental inbreeding. Biparental inbreeding could occur if seed and/or pollen dispersal is limited, creating family structure within apopulation and increasing the chances of mating between relatedindividuals (Loveless & Hamrick 1984, Duminil et al. 2009). Dur-ing the establishment of a population, seeds dispersed by gravity,as fruits fall from the parent to the ground, may germinate andlead to populations comprised of interrelated individuals. Subse-quently, seed dispersers such as birds and bats may introducedivergent germplasm to the population. In contrast, pollen couldbe exchanged between populations via hawkmoth (Sphingidae)pollinators, which are the primary pollinators for C. papaya and amajor pollinator in the Costa Rican tropical dry forest (Haber &Frankie 1989, Westerkamp & Gottsberger 2000). Despite the lit-tle knowledge of the foraging behavior of tropical Sphingidmoths, it has been reported that they can travel long distancebetween plants and that marked individuals visit the same plantsfor several days in succession (Haber & Frankie 1983a, b).These findings suggest that they can maintain genetic exchangebetween isolated small groups of plants. Secondary dispersal byanimals, in conjunction with increased pollen dispersal, would actto decrease genetic differentiation among populations.

In general, however, it is difficult to interpret these patternsof population differentiation given the ephemeral nature of natu-ral papaya populations; papaya colonizes disturbed areas and islater outgrown by other trees. Because of this, colonization andlocal extinctions are part of the dynamics of these species. Thecurrent pattern of land use, with lots of abandoned pastures, mayfacilitate gene exchange between clusters of plants, as the distancethat separates them may be shorter now than in the past. Eventhen, local extinction and recolonization events followed by popu-lation expansion (see below) should be part of the biology of thisspecies and should play a major role in determining the geneticstructure of these populations.

TO WHAT EXTENT HAS THE GENETIC DIVERSITY OF NATURAL PAPAYA

POPULATIONS BEEN SHAPED BY PAST DEMOGRAPHIC PROCESSES?.—Our demographic analysis suggests that Costa Rican papaya popu-lations have undergone a past genetic bottleneck followed by recentexpansion of the population. One possible explanation for thisobserved demographic history is that the natural papaya popula-tions may represent descendents from papaya that were cultivatedin the region of Costa Rica in the pre‐Columbian era. Upon discov-

ery of the New World, papaya was found from Mexico to Panama,suggesting a broader cultivation of papaya by Native Americansthroughout Central America (Storey 1976). If these plants under-went a domestication bottleneck, this would have resulted in anoverall loss of genetic diversity. Following the decline of pre‐Columbian cultures, these plants would have potentially becomeferal and subsequently spread throughout the region naturally, intheir ecological role as a pioneer species. Alternatively, the popula-tion expansion could have occurred much more recently with theloss of native lowland habitats within Costa Rica. A majority ofCosta Rica has been deforested since the 1920s, although someregions, in particular in the Guanacaste province, have seen a resur-gence of secondary forest growth (Nygren 1995, Arroyo‐Mora et al.2005, Calvo‐Alvarado et al. 2008). Since papaya thrive in disturbedhabitats, pasture abandonment within Costa Rica may have led torecent population expansion.

Papaya serves as an example of how the changing Mesoameri-can landscape can affect plant population dynamics, as well as howagricultural weeds could potentially spread to recovering areas. Thislater aspect is more complicated in papaya, as it is not only culti-vated in Mesoamerica, but is native to this region as well, thuspotentially blurring the distinction between feral and natural popu-lations. In order to further assess the introgression of cultivatedpapaya in the natural landscape, the genetic diversity of papayawithin and among other Mesoamerican regions should be com-pared with each other and to native cultivars. Ultimately, such anal-yses should also help us understand the domestication history ofthis ecologically and commercially important tropical fruit species.

ACKNOWLEDGMENTS

The authors thank the Costa Jose A. Hernandez and Marta L.Jimenez, Comisión Nacional para la Gestión de la Biodiversidad (CON-AGEBIO), for granting permission to collect papaya samplesfrom Costa Rica (permit #1‐0416‐0941). We also thank RayMing (University of Illinois) and QingYi Yu (formally of theHawaii Agricultural Research Station, currently Texas A&M Uni-versity) and Francis Zee (USDA) for their donation of cultivarleaf tissue. We also thank the editor and two anonymous review-ers for helpful suggestions for improving the manuscript. Thiswork was supported by a Miami University Department of Bot-any Academic Challenge Grant and a Miami University Depart-ment of Botany Summer Workshop in Botany Grant to JEB,

TABLE 4. Mean values of fruit diameter, mesocarp thickness, fruit diameter to mesocarp thickness ratio and leaf length to width ratio (± SE) for clusters identified by STRUCTURE

with analysis of variance (ANOVA) and Tukey's HSD post‐hoc significance levels indicated.

Structure cluster

identity

Mean fruit diameter

(cm)a**Mean mesocarp thickness

(cm)***Mean ratio fruit diameter: mesocarp

thickness***Mean ratio leaf blade width:

length

1 6.828 ± 0.518 b 1.483 ± 0.162 ab 4.859 ± 0.552 ab 0.933 ± 0.011 a

2 6.347 ± 0.568 b 1.260 ± 0.177 b 6.687 ± 0.605 a 0.930 ± 0.010 a

3 8.605 ± 0.492 a 1.930 ± 0.153 a 4.614 ± 0.524 b 0.957 ± 0.010 a

aLetters following means indicate significance levels based on Tukey's HSD post‐hoc test of significance.

ANOVA significance level: *P<0.05, **P<0.01, ***P<0.001.

186 Brown, Bauman, Lawrie, Rocha, and Moore

Miami University College of Arts and Sciences funds and NSFGrant DBI‐0922545 to RCM, and the Department of BiologicalSciences, Kent State University funds to OJR.

SUPPORTING INFORMATION

Additional Supporting Information may be found in the onlineversion of this article:

TABLE S1. Geographic positions of individual papaya sampledfromnatural populations.TABLE S2. Cultivar accessions and origins.TABLE S3. Morphological traits derived from the papayadescriptors of

IBPGR (1988).TABLE S4. List of the 20 microsatellite loci with primerdesign.TABLE S5. Chi‐square test for heterogeneity among regions forqualita-

tive traits.TABLE S6. Correlation coefficients (loadings) ofmorphological traits

with canonical variables (CANs). 1 and2TABLE S7. Canonical variable scoring coefficients fordiscriminant analy-

sis of females grouped by region.TABLE S8. Canonical variable scoring coefficients fordiscriminant analy-

sis of males grouped by region.TABLE S9. Allelic variability and deviation from HWE perlocus in

the ten operational populations of C. papaya.TABLE S10. STRUCTURE identity percentages for K=3 (natural-

populations only).TABLE S11. STRUCTURE identity percentages for K=2 (natural-

populations and cultivars).TABLE S12. Canonical variable scoring coefficients fordiscriminant

analysis of females grouped by STRUCTUREidentity.TABLE S13. Canonical variable scoring coefficients fordiscriminant

analysis of males grouped by STRUCTUREidentity.TABLE S14. Demographic indices.FIGURE S1. Correspondence analysis plots of the first (c1)

and second (c2) correspondence coefficients of significantly heter-ogeneous qualitative morphological traits.FIGURE S2. Discriminant analysis of (A) female and (B) male

quantitative morphological traits, with 95 percent confidence ellip-ses of regions indicated.FIGURE S3. Discriminant analysis of (A) female and (B) male

quantitative trait variation among STRUCTURE‐assigned popula-tion clusters with 95 percent confidence ellipses of regions indi-cated.

Please note: Wiley‐Blackwell is not responsible for the con-tent or functionality of any supporting materials supplied by theauthors. Any queries (other than missing material) should bedirected to the corresponding author for the article.

LITERATURE CITED

AMOS, W., J. I. HOFFMAN, A. FRODSHAM, L. ZHANG, S. BEST, AND A. V. S. HILL.2006. Automated binning of microsatellite alleles: Problems and solu-tions. Mol. Ecol. Notes 7: 10–14.

ANCHUKAITIS, K. J., AND S. P. HORN. 2005. A 2000‐year reconstruction of for-est disturbance from southern Pacific Costa Rica. Paleogeogr. Paleocli-mateol. Paleoecol. 221: 35–54.

ARROYO‐MORA, J. P., G. A. SÁNCHEZ‐AZOFEIFA, B. RIVARD, J. C. CALVO, AND D.H. JANZEN. 2005. Dynamics in landscape structure and compositionfor the Chorotega region, Costa Rica from 1960 to 2000. Agric. Eco-syst. Environ. 106: 27–39.

BADILLO, V. M. 1993. Caricaceae, Segundo esquema. Revista de la Facultad deAgronomía Universidad Central. Venezuela. Alcance 43.

BÌLGÌN, R. 2007. Kgtests: A simple Excel macro program to detect signaturesof population expansion using microsatellites. Mol. Ecol. Notes 7: 416–417.

BUSCHIAZZO, E., AND N. J. GEMMELL. 2006. The rise, fall and renaissance ofmicrosatellites in eukaryotic genomes. BioEssays 28: 1040–1050.

CALVO‐ALVARADO, J., B. MCLENNAN, G. A. SÁNCHEZ‐AZOFEIFA, AND T. GARVIN.2008. Deforestation and forest restoration in Guanacaste, Costa Rica:Putting conservation policies in context. For. Ecol. Manage. 258: 931–940.

CHAZDON, R. L., C. A. PERES, D. DENT, D. SHEIL, A. E. LUGO, D. LAMB, N.E. STORK, AND S. E. MILLER. 2009. The potential for species conserva-tion in tropical secondary forests. Conserv. Biol. 23: 1406–1417.

CHEN, C., Q. YU, S. HOU, Y. LI, M. EUSTICE, R. L. SKELTON, O. VEATCH, R. E.HERDES, L. DIEBOLD, J. SAW, Y. FENG, W. QIAN, L. BYNUM, P. H.MOORE, R. E. PAULL, M. ALAM, AND R. MING. 2007. Construction ofa sequence‐tagged high‐density genetic map of papaya for comparativestructural and evolutionary genomics in Brassicales. Genetics 177:2481–2491.

DECLERCK, F. A. J., R. CHAZDON, K. D. HOLL, J. C. MILDER, B. FINEGAN, A.MARINEZ‐SALINAS, P. IMBACH, L. CANET, AND Z. RAMOS. 2010. Biodi-versity conservation in human‐modified landscapes of Mesoamerica:Past, present and future. Biol. Conserv. 143: 2301–2313.

DUMINIL, J., S. FINESCHI, A. HAMPE, P. JORDANO, D. SALVINI, G. G. VENDRAM-

IN, AND R. J. PETIT. 2007. Can population genetic structure be pre-dicted by life‐history traits? Am. Nat. 169: 662–672.

DUMINIL, J., O. J. HARDY, AND R. J. PETIT. 2009. Plant traits correlated withgeneration time directly affect inbreeding depression and mating sys-tem and indirectly genetic structure. BMC Evol. Biol. 9: 177–190.

D'EECKENBRUGGE, G. C., M. T. RESTREPO, D. JIMÉNEZ, AND E. MORA. 2007.Morphological and isozyme characterization of common papaya inCosta Rica. In Y. K. Chan, E. Paull, and A. R. Shukor (Eds.). Pro-ceedings of the first international symposium on papaya genting High-lands, Malaysia, November 22–24, 2005, pp. 109–120. ISHS[Belgique], Louvain, Belgium.

ESTOUP, A., P. JARNE, AND J‐M. CORNUET. 2002. Homoplasy and mutationmodel at microsatellite loci and their consequences for populationgenetics analysis. Mol. Ecol. 11: 1591–1604.

EVANNO, G., S. REGNAUT, AND J. GOUDET. 2005. Detecting the number ofclusters of individuals using the software STRUCTURE: A simulationstudy. Mol. Ecol. 14: 2611–2620.

EXCOFFIER, L., G. LAVAL, AND S. SCHNEIDER. 2005. Arlequin ver 3.0: An inte-grated software package for population genetics data analysis. Evol.Bioinform. Online 1: 47–50.

GARZA, J. C., AND E. G. WILLIAMSON. 2001. Detection of reduction in pop-ulation size using data from microsatellite loci. Mol. Ecol. 10: 305–318.

GUO, S. W., AND E. A. THOMPSON. 1992. Performing the exact test of Hardy‐Weinberg proportion for multiple alleles. Biometrics 48: 361–372.

HABER, W. A., AND G. W. FRANKIE. 1983a. Aellopos titan (Cinta Blanca, white‐banded Sphinxlet). In D. H. Janzen (Ed.). Natural history of CostaRica, pp. 704–705. Chicago University Press, Chicago, Illinois.

HABER, W. A., AND G. W. FRANKIE. 1983b. Callionima falcifera (Mancha de Pla-ta, silver‐spotted Sphinx). In D. H. Janzen (Ed.). Natural history ofCosta Rica, pp. 680–681. Chicago University Press, Chicago, Illinois.

HABER, W. A., AND G. W. FRANKIE. 1989. A tropical hawkmoth community:Costa Rican dry forest Sphingidae. Biotropica 21: 155–172.

Morphological and Genetic Diversity of Natural Papaya 187

HAMRICK, J. L., AND M. J. W. GODT. 1996. Effects of life history on genetic diver-sity in plant species. Philos. Trans. R. Soc. Lond. B 351: 1291–1298.

HARDY, O. J., N. CHARBONNEL, H. FREVILLE, AND M. HEUERTZ. 2003. Micro-satellite allele sizes: A simple test to assess their significance on geneticdifferentiation. Genetics 163: 1467–1482.

HARDY, O. J., AND X. VEKEMANS. 2002. SPAGeDI: A versatile computer pro-gram to analyse spatial genetic structure at the individual or popula-tion levels. Mol. Ecol. Notes 2: 618–620.

INTERNATIONAL BOARD FOR PLANT GENETIC RESOURCES (IBPGR). 1988. De-scriptors for papaya. International Board for Plant Genetic Resources,Rome, Italy.

KIMMEL, M., AND R. CHAKRABORTY. 1996. Measures of variation at DNArepeat loci under a general stepwise mutation model. Theor. Popul.Biol. 50: 345–367.

KIMMEL, M., R. CHAKRABORTY, J. P. KING, M. BAMSHAD, W. S. WATKINS, AND

L. B. JORDE. 1998. Signatures of population expansion in microsatelliterepeat data. Genetics 148: 1921–1930.

KING, J. P., M. KIMMEL, AND R. CHAKRABORTY. 2000. A power analysis of mi-crosatellite‐based statistics for inferring past population growth. Mol.Biol. Evol. 17: 1859–1868.

LEVENE, H. 1949. On a matching problem in genetics. Ann. Math. Stat. 20:91–94.

LOVELESS, M. D., AND J. L. HAMRICK. 1984. Ecological determinants of geneticstructure in plant populations. Ann. Rev. Ecol. Syst. 15: 65–95.

MANSHARDT, R. M., AND F. ZEE. 1994. Papaya germplasm and breeding inHawaii. Fruit Varieties J. 48: 146–152.

MORSHIDI, M., R. M. MANSHARDT, AND F. ZEE. 1995. Isozyme variability inwild and cultivated Carica papaya. HortScience 30: 809.

NYGREN, A. 1995. Deforestation in Costa Rica: An examination of social andhistorical factors. For. Conserv. Hist. 39: 27–35.

OLEJNIK, S., J. LI, AND C. J. HUBERTY. 1997. Multiple testing and statisticalpower with modified Bonferroni procedures. J. Educ. Behav. Stat. 22:389–406.

PÉREZ, J. A. O., G. C. D'EECKENBRUGGE, A. M. RISTERUCCI, D. DAMBLER, AND

P. OLLIRAULT. 2007. Papaya genetic diversity assessed with microsatel-lite markers in germplasm from the Caribbean region. In Y. K. Chan,R. E. Paull, and A. R. Shukor (Eds.). Proceedings of the first interna-tional symposium on papaya genting highlands, Malaysia, November22–24, 2005, pp. 93–101. ISHS [Belgique], Louvain, Belgium.

PRITCHARD, J. K., M. STEPHENS, AND P. DONNELLY. 2000. Inference of popu-lation structure using multilocus genotype data. Genetics 155: 945–959.

RAYMOND, M., AND F. ROUSSET. 1995. GENEPOP (version 1.2): Populationgenetics software for exact tests and ecumenicism. J. Hered. 86: 248–249.

REICH, D. E., M. W. FELDMAN, AND D. B. GOLDSTEIN. 1999. Statistical proper-ties of two tests that use multilocus data sets to detect populationexpansions. Mol. Biol. Evol. 16: 453–466.

REICH, D. E., AND D. B. GOLDSTEIN. 1998. Genetic evidence for Paleolithichuman population expansion in Africa. Proc. Natl. Acad. Sci. USA95: 8119–8123.

ROUSSET, F. 2008. Genepop'007: A complete reimplementation of the Gene-pop software for Windows and Linux. Mol. Ecol. Resour. 8: 103–106.

SCHNEIDER, S., J. M. KUEFFER, D. ROESSLI, AND L. EXCOFFIER. 1997. Arlequin:A software for population genetic data analysis. Genetics and Biome-try Laboratory, University of Geneva, Switzerland.

STOREY, W. B. 1969. Papaya (Carica papaya L.). In P. Ferwerda and F. Wit(Eds.). Outline of perennial crop breeding in the tropics, pp. 389–407.H. Veenman and Zonen, Wageningen, The Netherlands.

STOREY, W. B. 1976. Papaya Carica papaya (Caricaceae). In N. W. Simmonds(Ed.). Evolution of crop plants, pp. 21–24. Longman Group Ltd.,London, U.K.

WEIR, B. S., AND C. C. COCKERHAM. 1984. Estimating F‐statistics for the anal-ysis of population structure. Evolution 38: 1358–1370.

WESTERKAMP, C., AND G. GOTTSBERGER. 2000. Diversity pays in crop pollina-tion. Crop Sci. 40: 1209–1222.

188 Brown, Bauman, Lawrie, Rocha, and Moore