Stationary Distributions of Microsatellite Loci Between Divergent ...

Stationary Distributions of Microsatellite Loci Between Divergent Population Groups of the European Rabbit (Oryctolagus cuniculus) Guillaume Queney, * Nuno Fewand,?$ Steven Weiss,"florence Mougel, *l and Monique Monnerot * *Centre de GCnCtique MolCculaire (CGM), CNRS, Gif sur Yvette cedex, France; lCentro de Estudos de CiCncia Animal (CECA), Campus Agrhio de VairBo, Vila do Conde, Portugal; and $Departamento de Zoologia e Antropologia, Faculdade de CiCncias do Porto, Praqa Gomes Teixeira, Porto, Portugal Previous analysis of rnitochondrial DNA polymorphism in the native range of the European rabbit (Oryctolagus cuniculus) demonstrated the occurrence of two highly divergent (2 Myr) maternal lineages with a well-defined geographical distribution. Analysis of both protein and immunoglobulin polymorphisms are highly concordant with this pattern of differentiation. However, the present analysis of nine polymorphic microsatellite loci (with a total of 169 alleles) in 24 wild populations reveals severe allele-size homoplasy which vastly underestimates divergence between the main groups of populations in Iberia. Nonetheless, when applied to more recent historical phenomena, this same data set not only confirms the occurrence of a strong bottleneck associated with the colonization of Mediterranean France but also suggests a two-step dispersal scenario that began with gene flow from northern Spain through the Pyrenean barrier and subsequent range expansion into northern France. The strength and appropriateness of applying microsatellites to more recent evolutionary questions is highlighted by the fact that both mtDNA and protein markers lacked the allelic diversity necessary to properly evaluate the colonization of France. The welldocumented natural history of European rabbit populations provides an unusually comprehensive framework within which one can appraise the advantages and limitations of microsatellite markers in revealing patterns of genetic differentiation that have occurred across varying degrees of evolutionary time. The degree of size homoplasy presented in our data should serve as a warning to those drawing conclusions from microsatellite data sets which lack a set of complementary comparative markers, or involve long periods of evolutionary history, even within a single species.

Introduction The well-documented history of the European rabbit (Oryctolagus cuniculus) offers an excellent phylogeographic framework within which one can test the efficacy of using different genetic markers (mtDNA, proteins, microsatellites) to uncover intraspecific evolutionary history. Fossil data and archaeological remains suggest that the species arose in the southern half of the Iberian Peninsula at least 1 MYA (Lopez-Martinez 1989; Callou 1995) and was able to extend its range through the Pyrenean bamer into Mediterranean France ca. 500,000 years ago (Pages 1980). However, Donard (1982) suggested that the Mediterranean region has undergone several episodes of extinction and recolonization and present-day populations of France have a much more recent origin. This hypothesis is concordant with the characterization of a genetic bottleneck using mtDNA (Monnerot et al. 1994; Hardy et al. 1995; BranCO, Ferrand, and Monnerot 2000) as well as nuclear markers, namely immunoglobulins (van der Loo, Ferrand, and Soriguer 1991; van der Loo et al. 1999) and protein polymorphism (Ferrand 1995; Ferrand and BranCO,unpublished data). Under this scenario, it was not until the Middle Ages that the rabbit's geographic distribution expanded from southern to northern France and

' Present address: Population GtnCtique et Evolution (PGE), CNRS, Gif sur Yvette cedex, France. Key words: European rabbit, phylogeography, microsatellite, homoplasy, allele size constraints. Address for correspondence and reprints: Guillaume Queney, Centre de GtnCtique Moltculaire, CNRS, 91198 Gif sur Yvette cedex, France. E-mail: queney @cgm.cnrs-gif.fr. Mol. Biol. Evol. 18(12):2169-2178. 2001 O 2001 by the Society for Molecular Biology and Evolution. ISSN 0737-4038

then on to central and northern Europe (Callou 1995). While introductions in northern Africa and Mediterranean islands occurred as early as 3,000 years ago (Vigne 1988; Dobson 1998) the Middle Ages represent the most important period of expansion for a species that has subsequently shown a remarkable ability to colonize new territories, being now found in most of Australia and New Zealand, portions of South America, and in more than 800 islands throughout the world (Flux and Fullagar 1983). Several studies based on mitochondrial DNA polymorphism within the rabbit's native range reveal two highly divergent (at least 2 Myr) maternal lineages, each with a well-defined geographical distribution: one lineage occurs in southwestern Iberia and the other in northeastern Spain, France, and in domestic breeds (Biju-Duval et al. 1991; Monnerot et al. 1994). Recently, Branco, Ferrand, and Monnerot (2000) provided a more comprehensive picture of mtDNA variation within the Iberian Peninsula, reporting the existence of a relatively narrow contact zone for the two maternal lineages that bisects the peninsula along a northwest-southeast axis. Analysis of 20 polymorphic protein loci exhibiting more than 100 alleles also reveals two major groups of population~coincident with the mtDNA subdivision (Ferrand 1995; Ferrand and Branco, unpublished data) as does the analysis of immunoglobulin polymorphism (van der Loo, Ferrand. and Soriguer 1991; van der LOO et al. 1999). Collectively, these data suggest that these population groups evolved separately for a significant period of time before a hybrid zone was formed following more recent secondary contact. Additionally, significant loss of genetic variability in populations north of

2170

Queney et al.

Sac

ern Spain and Caparosso (Cap), Peralta (Per), Tudela (Tud), Tarragona (Tar), and LCrida (Ler) in northern Spain; and (3) France (FR) containing Perpignan (Prp), Estagnol (Est), Villeneuve (Vil), Fos sur mer (Fos), la Tour du Valat (Tdv), Carlucet (Car), Abadia (Aba), Arjuzanx (Arj), Vaulx-en-Velin (Vau), Ferribre (Fer), Saclay (Sac), Versailles (Ver) and Gerstheim (Ger). Microsatellite Typing

Iberian Peninsula

\_/ Southwestern

Iberian Peninsula (SWIP) FIG. 1.-Geographical sampled in this study.

locations of the wild rabbit populations

the Pyrenees relative to those in the south, seen at both mtDNA and polymorphic protein loci, indicates a genetic bottleneck associated with the postglacial expansion of the rabbit from its pan-Iberian distribution area. Within this phylogeographic framework, we screened a set of microsatellite markers to evaluate, on a finer scale, current hypotheses concerning both the rabbit's evolutionary past and its initial stages of geographic expansion. We are particularly interested in addressing two questions: (1) how informative are microsatellites in revealing the deep genetic divergence between groups of populations, and (2) can microsatellites reveal the pattern of a recent population expansion across France, a phenomenon that allozymes and mtDNA have failed to elucidate with any degree of explanatory resolution.

Whole genomic DNA was extracted with either a standard phenol-chloroform protocol or in Qiagen columns (QIAamp kit). A total of eight dinucleotide and one tetranucleotide loci was chosen (Mougel, Mounolou, and Monnerot 1997) from gene banks (sat2, sat3, and sat4) or isolated from a genomic library (sat5, sat7, sat8, satl2, sat 13, and satl6). Some microsatellite typing was achieved with radioactive-labeling, single locus PCR, and 6% polyacrylamide gels (Mougel, Mounolou, and Monnerot 1997), while most was accomplished with multiplex PCR, fluorescently labeled primers, and an AB1 310 (Applied Biosystems) automated sequencer (Queney 2000, p. 202). To ensure that there was no bias in allele detection or sizing, three populations (Saclay, SantarCm and Las Lomas) were scored with both methods. Statistical Analysis

Comparative measures of genetic diversity for each population were calculated in the form of allelic diversity (total number of alleles, mean number of alleles per locus, and private alleles), observed heterozygosity, and nonbiased expected heterozygosity (Nei 1987) using the program GENETIX (Belkhir et al. 1996). Hardy-Weinberg equilibrium (HWE) was evaluated for all loci across all populations, and linkage disequilibrium between pairs of loci was evaluated using GENEPOP software (Raymond and Rousset 1995). Statistical significance was determined using Bonferroni correction (Rice 1989). To enable large-scale inferences on the relation of major groups of rabbit populations and their expansion outside Iberia, diversity indices were averaged across all loci, and a mean value was calculated for each of the three geographical groups of populations (SWIP, NEIP, and FR). A Wilcoxon-Mann-Whitney test was Materials and Methods used to test for significant differences in allelic diversity Sampling or heterozygosity across these three groups. All tests Blood or tissue samples were collected from 24 were conducted separately for each measure of diversity, populations (N = 829) located across the Iberian Pen- using STATVIEW (Abacus Concepts Inc., Berkeley, insula and France. Sampling was designed to represent Calif.). To estimate gene flow within and among groups of the original native range of the rabbit (i.e., Iberia), and France, the initial area of expansion outside the Iberian populations, estimators of FsT(0) and their 95% confiPeninsula (Callou 1995). The contact zone between the dence intervals (bootstrapping over loci) were calculated two ancestral maternal lineages in Iberia, reported by using FSTAT (Goudet 1995). Estimates of FsTfor miBranco, Ferrand, and Monnerot (2000), was excluded. crosatellite data were compared to available data on The sampled populations correspond to three geographic these populations based on mtDNA RFLPs and polyregions (fig. 1): (1) the southwestern Iberian Peninsula morphic protein loci. To depict the genetic relationships among all pop(SWIP) consisting of SantarCm (San) and Idanha (Ida) in Portugal, and Huelva (Hue), DoHana (Don) and Las ulation~,networks were generated using the NeighborLomas (Llo) in southern Spain; (2) the northeastern Ibe- Joining (NJ) algorithm with the program NEIGHBOR rian Peninsula (NEIP) containing Alicante (Ali) in east- in the PHYLIP package. Because there is still consid-

Stationary Distributions of Microsatellite Loci

2171

Table 1 Genetic Diversity Indices Calculated from 9 Microsatellite Loci for All 24 Rabbit Populations. Shown Are the Sample Sizes for Each Population (n), the Number of Alleles, the Observed (Ho) and Expected (He) Heterozygosities, the Mean and Variance in Allele Sizes Averaged Across All Loci. Standard Deviations (SD) Are Given in Brackets NUMBER OF ALLELES POPULATIONS

n

Total

per Locus

Private

ALLELESIZE Ho

He (SD)

Mean (SD)

Variance (SD)

(FR) France (n = 508) Gertsheim . . . . . . . . . . . 22 33 3.7 Saclay.. . . . . . . . . . . . . . 98 47 5.2 Versailles . . . . . . . . . . . . 47 42 4.7 40 4.4 Abadia . . . . . . . . . . . . . . 30 Arjuzanx . . . . . . . . . . . . 50 50 5.6 48 5.3 Carlucet.. . . . . . . . . . . . 36 Estagnol.. . . . . . . . . . . . 44 50 5.6 49 5.4 Ferrikre.. . . . . . . . . . . . . 24 Fos sur m e r . . . . . . . . . . 10 46 5.1 56 6.2 Perpignan.. . . . . . . . . . . 26 Tour du Valat . . . . . . . . 32 41 4.6 53 5.9 Vaulx-en-Velin. . . . . . . . 7 1 Villeneuve . . . . . . . . . . . 18 53 5.9 (NEIP) Northeastern Iberian Peninsula (n = 164) 59 6.6 Alicante . . . . . . . . . . . . . 18 Caparosso.. . . . . . . . . . . 24 75 8.3 55 6.1 L 6 i d a . . . . . . . . . . . . . . . 12 Peralta . . . . . . . . . . . . . . 37 81 9.0 69 7.7 Tarragona.. . . . . . . . . . . 24 Tudela . . . . . . . . . . . . . . 49 84 9.3 (SWIP) Southwestern Iberian Peninsula (n = 157) Dofiana . . . . . . . . . . . . . . 22 82 9.1 Huelva.. . . . . . . . . . . . . 24 100 11.1 Idanha . . . . . . . . . . . . . . 30 73 8.1 92 10.2 Las Lomas. . . . . . . . . . . 58 Santarkm . . . . . . . . . . . . 23 72 8.0 Mean values FR . . . . . . . . . . . . . . . . . 39.1 46.8 5.2 70.5 7.8 NEIP . . . . . . . . . . . . . . . 27.3 SWIP . . . . . . . . . . . . . . . 31.4 83.8 9.3 60.4 6.7 Global mean . . . . . . . . 34.5

probable. Few departures from HWE were detected at individual loci, and none for populations across all loci (using a Bonferroni correction) when the sat16 locus was removed. This locus had a suspected null allele, which resulted in strong departures from equilibrium in some populations (Queney 2000, p. 202). The mean total number of alleles (a) and expected heterozygosity (H,) were highest in SWIP populations (a = 83.8, H, = 0.823) slightly lower in NEIP (a = 70.5, H,= 0.777), and lowest in populations of France (a = 46.8, H, = 0.644), conforming to our expectations of reduced genetic diversity in the area of initial geographic expansion compared to Iberian refugia (table 1). However, there were no significant differences in allelic diversity ( P = 0.144) between SWIP and NEIP, whereas Iberian populations were significantly more diverse than populations in France ( P < 0.01, table 1). There was a large difference in the number of private alleles (defined Results here as alleles found in a single population throughout Genetic Diversity the study region) between SWIP (27) and both NEIP (4) One pair of loci (sat2 and sat4) revealed significant and FR (5); the similar numbers for NEIP and FR problinkage disequilibrium for most populations ( P < ably reflect the much higher sampling effort in FR. Populations in France exhibited a total of 93 alleles 0.001). These two loci are known to occur within the mammalian casein gene cluster (Threadgill and Womack which, except for six alleles not found in Iberia, ap1990; Archibald 1994) and thus physical linkage is most peared to represent a subset (ca. 55%) of the allelic di-

erable debate over the merits and drawbacks of various microsatellite-based genetic distances, we used five different matrices of genetic distances as input for the NJ algorithm. Nei's standard distance (Nei 1987) was chosen as it assumes an infinite allele model (Kimura and Crow 1964), whereas three microsatellite-specific measures: dk2 (Goldstein et al. 1995), Dsw (Shriver et al. 1995), and RsT (Slatkin 1995) all assume a stepwise mutation model (Ohta and Kimura 1973) but may differ in how they reflect varying amounts of drift and mutation. Finally, the simple allele-sharing statistic DAs (Bowcock et al. 1994) was used to represent a measure which makes no evolutionary assumptions. Nei's distances were calculated between all populations using GENDIST in the PHYLIP package (Felsenstein 1993) while all other distances were calculated using MICROSAT (Minch 1996).

2172 Queney et al.

Table 2 F-Statistics Between and Within Geographical Groups of Populations Within groups FR.. . . . . . . . . . . . . NEIP . . . . . . . . . . . . SWIP . . . . . . . . . . . All populations . . . Between groups NEIPXSWIP . . . . FRXNEIP . . . . . . FR X SWIP. . . . . .

sat2

sat3

sat4

sat5

sat7

sat8

sat12

sat13

sat16

Allloci

95% C1

0.173 0.075 0.055 0.195

0.116 0.127 0.042 0.1 15

0.133 0.074 0.060 0.159

0.071 0.042 0.048 0.092

0.165 0.052 0.059 0.161

0.210 0.121 0.083 0.229

0.113 0.041 0.031 0.090

0.166 0.046 0.052 0.153

0.127 0.056 0.089 0.136

0.140 0.074 0.055 0.146

[0.116; 0.1651 [0.055;0.094] [0.046;0.064] [0.120; 0.1751

0.047 0.093 0.200

0.012 0.035 0.035

0.071 0.053 0.142

0.012 0.051 0.084

0.027 0.078 0.130

0.178 0.049 0.209

0.002 0.017 0.027

0.047 0.084 0.057

0.008 0.039 0.026

0.047 0.056 0.104

[0.018;0.086] [0.041;0.071] 10.049; 0.1621

versity found in all populations. The Iberian regions were considerably more diverse, with NEIP containing 74% and SWIP 86% of the total number of alleles found in the study. Allelic frequencies at each locus summed over all populations are given in figure 2 and are available on the web for each of the 24 populations (ftp.cgm.cnrs-gif.fr/pub/genevoVrabbit~pop~freq.xls).

wise differences in allele sizes, both Nei's distance (fig. 3a) and the allele sharing statistic (data not shown) displayed relatively homogeneous branch lengths and additionally distinguished three specific geographical regions within FR (southwest, southeast, and north of France).

Genetic Structuring

Discussion Patterns of Genetic Diversity and Differentiation

Population differentiation is presented in table 2. FsTvalues within the Iberian regions NEIP (0.074) and SWIP (0.055) were low compared to previous estimates obtained using mtDNA (Branco, Ferrand, and Monnerot 2000) and protein data (Ferrand 1995; Ferrand and Branco, unpublished data). However, in France, a region of low genetic diversity, the overall FsTestimate was actually higher (0.140) than that calculated for either Iberian region. FR populations exhibited more genetic affinity with the more proximate NEIP (FsT = 0.056) than the more distant SWIP (FsT = 0.104), supporting that the former region was the source for expansion into France. The FsTvalue between NEIP and SWIP (0.047) was significantly different from zero but this value was considerably lower than that obtained for either allozymes (0.16) or mtDNA (0.81) (Branco, Ferrand, and Monnerot 2000; Branco and Ferrand, unpublished data). Population Relationships The overall pattern of population relationships displayed with NJ networks was highly dependent on the distance matrix used. For two of the three microsatellitespecific distances (dp2 and R,,), the network topology showed little concordance with the three well-defined geographic regions (fig. 3c, RsT cladogram not shown). The remaining distance measured all generated topologies that were basically concordant with large-scale geography with nodes dividing NEIP, SWIP, and FR (fig. 3a and b, other cladograms not shown). There were minor inconsistencies in the branching patterns among all networks and no clear geographic pattern was seen within NEIP or SWIP for any distance measure. Compared to Dsw (fig. 3b), which is more influenced by nonstep-

FIG. 2.-Allelic on the x-axis.

Genetic variation in Iberian rabbits was found to be significantly higher than rabbits of France, and allelic distribution profiles depicted in figure 2 clearly show that FR populations are a subset of IP populations. However, the two Iberian regions which are highly differentiated at the mtDNA level displayed similar allelic profiles (fig. 2), mean variances in allele size across all 1.9), and loci (NEIP = 5.7 -t 2.2 and SWIP = 6.2 approximately the same mean number of alleles per population (table l). The most striking feature distinguishing these two regions was in the number of private alleles. Based on the diversity and distribution of mtDNA haplotypes, Branco, Ferrand, and Monnerot (2000) suggested that large and stable effective population sizes in SWIP have promoted the maintenance of low-frequency haplotypes as opposed to more fluctuating population sizes in NEIP Whereas rare alleles may be used as a rough measure of gene flow in some situations (qualitatively or quantitatively), different effective population sizes and homoplasy can both result in bias (Slatkin 1985). We suggest that lower effective population sizes in NEIP have resulted in the loss of many rare alleles in comparison to SWIl? Based on the nearly identical overlapping distribution in allele sizes in these two regions, it is apparent that drift has promoted the loss of alleles in NEIP that correspond to those that are now private in SWIl? Thus, these two regions, which are thought to have supported rabbit populations for equal periods of history (based on equal spans of mtDNA networks), differ at highly variable microsatellite loci not because of new mutations, but because of differential loss of low-frequency alleles.

+

frequencies summed over all populations for each geographical group (FR, NEIP, SWIP). Allele sizes (in bp) are shown


FR

FR

NEIP

NEIP

SWIP

SWB

NEIP

FR

NEE

NEIP

swn

SWIP

sat l 3 FR

NEIP

@

FR

NEP

@

swa

allele 195 at locus sat4 is due to a deletion of 13 dinucleotide repeats (Mougel 1997)

2173

2174 Queney et al.

(a)

Iberian Peninsula


2175

(c

Iberian Peninsula

FIG. 3 (Continued)

High andlor differential rates of drift should be reflected in at least some microsatellite-specific distance measures such as Goldstein's d p 2 where the mean variance across loci is expected to remain equal in magnitude but undergo a modal shift between two groups of populations over time. However, given allele size range overlap, we must conclude that the mutation spectrum has become saturated during 2 million years of divergence, through a combination of mutation constraints and back mutations that have homogenized allele size distributions as predicted by Nauta and Weissing (1996). This form of homoplasy, which we refer to as size homoplasy, simply means that alleles are identical in state but have different mutational histories that have led to their present state. This definition neither necessitates nor excludes the possibility that alleles also differ at the sequence level. A mechanistic model for such allele size homogenization for tetranucleotide repeats in humans has recently been shown in Xu et al. (2000). The only alternative scenario that could explain our observations would be a pattern of long-term, sexbiased dispersal in which females remain in their breeding groups and gene flow is essentially male mediated. However, Ferrand (1995) and Ferrand and Branco (unpublished data) were able to describe a relatively strong phylogenetic signal (compatible with mtDNA) between southwestern and northeastern Iberian rabbits as well as

strong population substructure within regions using protein polymorphisms. Likewise, in our study, the large numbers of region-specific private alleles (as opposed to private for the study area) for both NEIP (18) and SWIP (38) suggest fine-scaled population structure in Iberian rabbits. Further evidence for the lack of gene flow over time is seen at the locus sat13 where intermediately sized private alleles (sizes 113, 115, and 127) in NEIP suggest that there has been a sufficient period of isolation for a combination of point mutations andlor a point mutation and subsequent slippage to occur without spreading to SWIP. Such population structure is incompatible with strong gene flow across the Iberian Peninsula. Thus, to our knowledge, we provide the first example of empirically stationary allele distributions across a set of microsatellite loci applied to intraspecific populations in a known phylogeographic context. Despite stationary allele distributions, it is illuminating to evaluate the effectiveness of different genetic distance measures in assessing the genetic relation among populations. Goldstein's dp2 and RsT failed to reveal the three main geographic regions as they depend heavily on detecting differences in allele size variance, a parameter which can become wholly obscured by homoplasy. However, both Nei's distance and DAs are not affected by allele size variances, being more weighted toward demonstrating differences in allele frequencies

Frc. 3.-Neighbor-joining network for the 24 rabbit populations based on three different distance measures. Branches are grouped by geographical location. Bootstrap support values (100 replications) are shown when greater than 50%. a, Nei's distance. b, Dsw distance. c, JpZ distance.

2176 Queney et al.

Microsatellite Homoplasy and Complex Evolutionary Histories

FIG. 4.-Colonization scenario of the rabbit into France. Numbers refer to three successive stages of expansion.

and the presence or absence of alleles. These measures clearly identified the three main geographic areas, and additionally were concordant with subregion division within France. D,, which incorporates allelic variance to some extent, also supported differentiation of NEIP, SWIP, and FR, but was discordant with Nei's distance and D,,-based trees in depicting the pattern of differentiation within FR regions. Colonization of France and Expansion to the North Our microsatellite data on FR populations clearly reflects depleted levels of genetic diversity when compared with Iberian populations (table 1). The disjunct allelic size distributions in contrast with those displayed by SWIP and NEIP are especially informative in supporting a founder effect resulting from colonization from Iberia. Furthermore, as allelic diversity in FR populations has not been restored, the lack of new mutations supports a very recent (i.e., postglacial) founding event. The derivation of French populations from NEIP is strongly suggested by several microsatellite loci, and especially by the occurrence of alleles 195 at sat4 (totally absent in SWIP), 247 at sat2, 140 at sat8 and 184 at sat7. Within France, the overall results show significant differences in allele frequencies (data not shown) allowing the definition of three population assemblages: southwest, southeast, and the north of France (fig. 3a). This pattern suggests a two-step colonization of France by the rabbit. In the first step, rabbits may have expanded following two main geographical routes, colonizing the southwest and the southeast (geographically separated by the mountains of Massif Central) from the Mediterranean region immediately adjacent to the eastern Pyrenees, after a single colonization event. In a second step, the recent colonization of the north of France may have resulted from an expansion of the southwestern group, as indicated by the phylogenetic reconstruction depicted in figure 3a and illustrated in the map of figure 4.

Microsatellites are the most popular genetic markers for answering a wide range of biological questions at the intraspecific level, despite continued dispute concerning the mode and mechanisms of their evolution (Goldstein and Schlotterer 1999), and several theoretically sound arguments warning of the misleading results that excessive homoplasy will generate (Garza, Slatkin, and Freimer 1995; Nauta and Weissing 1996). Empirical evaluations of some of these concerns are emerging in studies that compare the congruency of results with other genetic markers, or explore the efficacy of microsatellite data in revealing population subdivisions in welldescribed phylogeographic contexts. For example, Allendorf and Seeb (2000) compared gene flow estimates among microsatellites, mtDNA, allozymes, and RAPD markers and concluded that there was little difference in F,,-type estimates provided that they are corrected for differing numbers of alleles and heterozygosity. This study was conducted on a set of geographically proximate populations with a shallow evolutionary history, a situation thought to be most appropriate for the application of microsatellites (Takezaki and Nei 1996; Angers and Bernatchez 1998). However, several studies have successfully applied microsatellite markers in a deeper phylogeographic context as well as across species boundaries. Estoup et al. (1995) found basic concordance between microsatellites and mtDNA for honeybee subspecies, but suggested that allele size homoplasy resulted in underestimation of divergence among major lineages. Similarly, Harr et al. (1998) reported an unambiguous phenetic relation of four closely related species of Drosophila based on 39 microsatellite loci, but these same data provided divergence estimates an order of magnitude or more lower than those based on DNA sequence data. Thus, despite claims that some microsatellite distance measures can be linear with time (Takezaki and Nei 1996), it is clear that empirical studies often reveal the contrary. Most recently, Balloux et al. (2000) reported severe underestimation of divergence between two chromosomal races of a common shrew based on microsatellites. However, this example is somewhat unique in that there were sex-biased viability differences between the races, for which the evolutionary implications are not yet entirely clear. Thus, while there is ample evidence of homoplasy affecting divergence estimates, and most recently several mechanistic explanations of the dynamics of homoplasy (Ham and Schlotterer 2000; Xu et al. 2000), there is no empirical study yet that has revealed the stationary distributions predicted by Nauta and Weissing (1996) that will result from even moderate population sizes and some level of divergence. Our data provide a clear example of these predictions, where the pattern of homoplasy is most heuristically explained by considering the mutational spectrum as being filled between two boundaries, one representing the minimal repeat size below which no more slippage occurs, and the other representing a constraint on the maximum size of an allele. The evolutionary his-


tory of rabbits in Iberia has been sufficiently long, and population sizes sufficiently large to produce such a phenomenon. Despite extensive homoplasy, we were nonetheless able to distinguish between major geographic regions, because of fluctuating population sizes in one geographic unit (NEIP) which promoted sufficient drift of some intermediately sized alleles. While our pattern of private allele distributions between areas of refuge (SWIP and NEIP) and expansion (FR) can be seen as being analogous to those obtained for human populations (PerezLezaun et al. 1997) it may be dangerous to draw conclusions concerning private alleles when the nature of their evolution in terms of drift, mutation, and shifts in allele size distributions over time is not known. The clearest example of microsatellites effectively differentiating populations in a phylogeographic context involves a shallow history and constant and relatively small population sizes with N,'s on the order of hundreds of individuals (Goldstein and Schlotterer 1999). In our study, microsatellites were also most effective in the periphery of the rabbit's native range, where a hypothesized expansion and colonization scenario through the Pyrenees into southern France was well supported. However, even at the intraspecific level, an increasing number of studies are revealing complex evolutionary histories and population dynamics across broad temporal scales (Avise 2000). In such contexts, the sole use of microsatellites may mislead as often as inform on patterns of genetic differentiation and gene flow among populations. In our study, as in that of Calafell et al. (1998), there were large differences in the ability of various genetic distance measures to distinguish known phylogeographic pattern. We suggest that without some a priori understanding of historical complexity, in terms of fluctuating population sizes and divergence among populations, the application of various microsatellitebased distance measures may become arbitrary, especially in supporting explicit interpretations as to why populations d o or d o not appear differentiated.

2177

ANGERS,B., and L. BERNATCHEZ. 1998. Combined use of SMM and non-SMM methods to infer fine structure and evolutionary history of closely related brook charr populations from microsatellites. Mol. Biol. Evol. 15(2):143-159. ARCHIBALD, A. L. 1994. Mapping the pig genome. Cun: Opin. Genet. Dev. 4:395-400. AVISE,J. 2000. Phylogeography. The history and formation of species. Harvard University Press, Cambridge, Mass. BALLOUX, E, H. BRUNNER, N. LUGON-MOULIN, J. HAUSSER, and J. GOUDET. 2000. Microsatellites can be misleading: an empirical and simulation study. Evol. Int. J. Org. Evol. 54(4): 1414-1422. BELKHIR, K., l? BORSA,J. GOUDET,L. CHIKHI,and E BONHOMME. 1996. Genetix, logiciel sous Windows pour la gCnCtique des populations. Laboratoire GCnome et Population~,Montpellier, France. BIJU-DUVAL, C., H. ENNAFAA, N. DENNEBOUY, M. MONNEROT, E MIGNOTTE, R. SORIGUER, A. E. GAA~ED, A. E. HILI,and J. MOUNOLOU. 1991. Mitochondrial DNA evolution in lagomorphs: origin of systematic heteroplasmy and organization of diversity in european rabbits. J. Mol. Evol. 33:92-102. BOWCOCK, A. M., A. RUIZ-LINARES, J. TOMFOHRDE, E. MINCH, J. R. KIDD,and L. L. CAVALLI-SFORZA. 1994. High resolution of human evolutionary trees with polymorphic microsatellites. Nature 368:455-457. BRANCO, M., N. FERRAND, and M. MONNEROT. 2000. Phylogeography of the European rabbit (Oryctolagus cuniculus) on the Iberian Peninsula inferred from RFLP analysis of the cytochrome b gene. Heredity 85307-317. CALAFELL, E, A. SHUSTER, W. C. SPEED,J. R. KIDD,and K. K. KIDD.1998. Short tandem repeat polymorphism evolution in humans. Eur. J. Hum. Genet. 6(1):38-49. CALLOU, C. 1995. Modifications de I'aire de repartition du lapin (Oryctolagus cuniculus) en France et en Espagne, du PlCistocbne k I'Cpoque actuelle. Etat de la question. Anthropozoologica 21:95-L 14. DOBSON, M. 1998. Mammal distributions in the western mediterranean: the role of human intervention. Mammal Rev. 28(2):77-88. DONARD, E. 1982. Recherches sur les LCporinCs quaternaires (PlCistocbne moyen et superieur, Holocbne). PhD thesis, UniversitC Bordeaux I. M. SOLIGNAC, and J.-M. CORNUET. DTOUP,A., L. GARNERY, 1995. Microsatellite variation in honey bee (Apis mellifera L.) populations: hierarchical genetic structure and test of the Acknowledgments infinite allele and stepwise mutation models. Genetics 140: 679-695. N.E was supported by a grant from Direc~iioGeral J. 1993. PHYLIP (phylogeny inference package). das Florestas. We thank Paulo CClio Alves, J. Arques, FELSENSTEIN, Version 3.5. Distributed by the author, Department of GeSerge Avignon, Enrique Castien, Bruno Degrange, Gilnetics, University of Washington, Seattle. les Delacour, Jean-SCbastien Dorier, Olivier Galaup, Pa- FERRAND, N. 1995. Varia@o genCtica de proteinas em poputrice Galvand, Raquel Godinho, StCphane Griffe, JCr6me laqi5es de coelho (Oryctolagus cuniculus). PhD thesis, UnivLetty, StCphane Marchandeau, Sacrament0 Moreno, V. ersidade do Porto. Peiro, Fernando Queiros, Jean-Claude Ricci, Ignacio R u x , J., and F.' FULLAGAR. 1983. World distribution of the Rodriguez, JosC Luis Rosa, Maria Sanchez, Ramon Sorabbit (Oryctolagus cuniculus). Acta Zool. Fennica 174:7577. riguer, Rafael Villafuerte and Philippe van d e Walle for and N. FREIMER. 1995. Microsatellite capturing the rabbits, collecting field data, and taking GARZA,J., M. SLATKIN, allele frequencies in humans and chimpanzees with impliblood or tissue samples. Thanks are also due to one cations for constraints on allele size. Mol. Biol. Evol. 12(4): anonymous reviewer for his helpful comments. 594-603. and M. GOLDSTEIN, D., A. LINARES,L. CAVALLI-SFORZA, LITERATURE CITED FELDMAN. 1995. An evaluation of genetic distances for use with microsatellite loci. Genetics 139:463-471. ALLENDORF, E, and L. SEEB.2000. Concordance of genetic D., and C. SCHL~TTERER. 1999. Microsatellites. divergence among Sockeye salmon populations at allozyme, GOLDSTEIN, Evolution and applications. Oxford University Press, nuclear DNA, and mitochondrial DNA markers. Evolution Oxford. 54(2):640-65 1.

2178 Queney et al.

GOUDET,J. 1995. F-STAT version 1.2: a computer program to variation and the differentiation of modern humans. Hum. calculate F-statistic. J. Hered. 86(6):485-486. Genet. 99(1): 1-7. HARDY,C., C. CALLOU,J. VIGNE,D. CASANE, N. DENNEBOUY,QUENEY,G. 2000. Histoire des populations et organisation soJ. MOUNOLOU, and M. MONNEROT. 1995. Rabbit mitochonciale du lapin europCen (Oryctolagus cuniculus) B travers drial DNA diversity from prehistoric to modern times. J. 1'Ctude de marqueurs microsatellites. Denis Diderot University, Paris. Mol. Evol. 40:227-237. HARR,B., and C. SCHLOTTERER. 2000. Long microsatellite al- RAYMOND,M., and F. ROUSSET.1995. GENEPOP (version leles in Drosophila melanogaster have a downward muta1.2): a population genetics software for exact tests and ection bias and short persistence times, which cause their geumenicism. J. Hered. 86:248-249. nome-wide underrepresentation. Genetics 155(3): 1213- RICE, W. 1989. Analyzing tables of statistical tests. Evolution 43(1):223-225. 1220. HARR,B., S. WEISS,J. DAVID,G. BREM,and C. SCHLOTTERER.SHRIVER, M., L. JIN, E. BOERWINKLE, R. DEKA, R. FERREL, 1998. A microsatellite-based multilocus phylogeny of the and R. CHAKRABORTY. 1995. A novel measure of genetic Drosophila melanogaster genome. Curr. Biol. 8:1183-1 186. distance for highly polymorphic tandem repeat loci. Mol. Biol. Evol. 12(5):914-920. KIMURA,M., and J. CROW.1964. The number of alleles that SLATKIN,M. 1985. Rare alleles as indicators of gene flow. can be maintained in a finite population. Genetics 49:725Evolution 39(1):53-65. 738. . 1995. A measure of population subdivision based on LOPEZ-MARTINEZ, N. 1989. Revision sistematica y biostratimicrosatellite allele frequencies. Genetics 139:457-462. grafica de 10s lagomorphos (Marnmalia) del terciaro U cuaN., and M. NEI. 1996. Genetic distances and reconternario de Espana. Memorias del MusCo Paleontol6gico de TAKEZAKI, struction of phylogenetic trees from microsatellite DNA. la Universidad de Zaragoza. Diputaci6n general de Arag6n. Genetics 144:389-399. MINCH,E. 1996. Microsatellite distance program, http://lotka. THREADGILL, D., and J. WOMACK.1990. Genomic analysis of stanford.edu/microsat/. the major bovine milk protein. NAR 18:6,935-6,942. MONNEROT, M., J.-D. VIGNE,C. BIJU-DUVAL, D. CASANE,C. and R. SORIGUER. 1991. EsCALLOU,C. HARDY,E MOUGEL,R. C. SORIGUER, N. DEN- VAN DER LOO, W., N. FERRAND, timation of gene diversity at the b locus of the constant NEBOUY, and J.-C. MOUNOLOU. 1994. Rabbit and man: geregion of the immunoglobulin light chain in natural popunetic and historic approach. Genet. Select. Evol. 26(Suppl. lation~of european rabbit (Oryctolagus cuniculus) in Por1): 167s-182s. tugal, Andalusia and on the Azorean Islands. Genetics 127: MOUGEL,E, J. MOUNOLOU, and M. MONNEROT.1997. Nine 789-799. polymorphic microsatellite loci in the rabbit, Oryctolagus VAN DER LOO, W., E MOUGEL,C. BOUTON, M. SANCHEZ, and cuniculus. Anim. Genet. 2858-7 1. M. MONNEROT.1999. The allotypic patchwork pattern of NAUTA,M., and E WEISSING.1996. Constraints on allele size the rabbit IGKCl allele b5wf: genic exchange or common at microsatellite loci: implications for genetic differentiaancestry? Immunogenetics 49:7-14. tion. Genetics 143:1,021-1,032. NEI, M. 1987. Molecular evolutionary genetics. Columbia Uni- VIGNE,J.-D. 1988. DonnCes prCliminaires sur l'histoire du peuplement mammalien de 1'Plot de Zembra (Tunisie). Mamversity Press, New York. OHTA,T., and M. KIMURA.1973. A model of mutation appromalia 52(4):567-574. priate to estimate the number of electrophoretically detect- Xu, X., M. PENG,Z. FANG,and X. Xu. 2000. The direction of able alleles in a finite population. Genet. Res., Cambridge microsatellite mutations is dependant upon allele length. 22:201-204. Nat. Genet. 24:396-399. PAGES,M.-V. 1980. Essai de reconstitution de l'histoire du lapin de garenne en Europe. Bull. Mens. Off. Natl. Chasse, PIERRECAPY,reviewing editor Sp. Scien. Techn., DCcembre 1980:13-21. PEREZ-LEZAUN, A., E CALAFELL,E. MATEU,D. COMAS,R. RUIZ-PACHECO, and J. BERTRANPETIT. 1997. Microsatellite Accepted June 26, 2001

v