Vous êtes sur la page 1sur 16

R. LORENZINI and S.

LOVARI

Biological Journal of the Linnean Society, 2006, 88, 85100. With 4 figures

Genetic diversity and phylogeography of the European roe deer: the refuge area theory revisited
R. LORENZINI1* and S. LOVARI2
1

Istituto Zooprolattico Sperimentale dellAbruzzo e del Molise G. Caporale, Campo Boario, I-64100 Teramo, Italy 2 Dipartimento di Scienze Ambientali, Sezione di Ecologia Comportamentale, Etologia e Gestione della Fauna, Universit di Siena, Via P. A. Mattioli 4, I-53100 Siena, Italy

Received 5 November 2004; accepted for publication 1 July 2005

The extant taxa of central and northern Europe are commonly believed to derive from Pleistocene ancestors, who moved to the north from three separate glacial refugia: the Iberian and Italian peninsulae, as well as the southern Balkans. The issue of postglacial dispersal patterns was addressed through the investigation of population structure and phylogeography of the European roe deer, Capreolus capreolus. The genetic diversity in 376 individuals representing 14 allegedly native populations across their European range was assessed, using ten autosomal microsatellite loci and restriction fragment length polymorphisms of the mitochondrial D-loop and NADH dehydrogenase 1 gene segments. Our results suggest the existence of three major genetic lineages of roe deer in Europe. One comprises populations in the south-western limit of the species distribution (i.e. Iberia), where an internal substructure splits a northern from a southern sublineage. A second lineage includes populations of southern and eastern Europe, as well as a separate sublineage sampled in central-southern Italy, where the existence of the subspecies Capreolus c. italicus was supported. In central-northern Europe, a third lineage is present, which appeared genetically rather homogeneous, although admixed, and equally divergent from both the eastern and western lineages. Current patterns of intraspecic genetic variation suggest that postglacial recolonization routes of this cervid to northern Europe could be due to range expansion from one or more refugia in central-eastern Europe, rather than proceeding from the Mediterranean areas. 2006 The Linnean Society of London, Biological Journal of the Linnean Society, 2006, 88, 85100.

ADDITIONAL KEYWORDS: Capreolus capreolus colonization glacial refugium microsatellites


mitochondrial DNA subspecies Pleistocene.

INTRODUCTION
Consistency of temporal and spatial movements because of climatic changes should not be expected for taxa with different ecological requirements. Accordingly, a small degree of congruence between phylogeographical patterns during the Quaternary cold periods has been suggested for a sample of heterogeneous species across Europe (Taberlet et al., 1998). Similar trends of dispersal and recolonization of former range have been detected for taxa with comparable ecological requirements. During the ice ages, warm-

*Corresponding author. E-mail: r.lorenzini@izs.it

temperate adapted taxa may have survived in refugium areas with appropriate climatic and ecological features. It has been suggested (Hewitt, 1996; Taberlet et al., 1998) that the northern regions of Europe were colonized generally from the Iberian and Balkanic refugia. The Italian peninsula was considered as another Mediterranean refugium for temperate species expanding their ranges northwards in the late glacial and early postglacial (Hewitt, 1996). In particular, Iberia and the Balkans could have been the sources of recolonization routes to central and northern Europe for some mammals; for example, the western lineage of the brown bear (Taberlet & Bouvet, 1994; Kohn et al., 1995). Three to four recognized refuge areas have been identied during the Late

2006 The Linnean Society of London, Biological Journal of the Linnean Society, 2006, 88, 85100

85

86

R. LORENZINI and S. LOVARI relatives of modern roe deer appeared all over Europe as early as in the beginning of the Middle Pleistocene (900800 000 years ago), while forms closely related to the extant species may have been present in southern Europe already close to the end of the Middle Pleistocene, approximately 250 000 years ago (Kurtn, 1968; Lister, Grubb & Sumner, 1998). Roe deer occurrence is documented in both interglacial and glacial phases. Nevertheless, during the greatest expansions of continental ice caps that covered Scandinavia and the northern belt of central Europe, as well as the highest mountain chains, its range was limited to southern latitudes. With regard to the last glacial cycle (from 120 000 to 10 000 years ago), the roe deer appears to have recolonized central and northern Europe probably after the de-glaciation (starting 16 15 000 years ago). No fossils of roe deer have been found older than 1211 000 years ago (younger Dryas) in central and North Europe (Hufthammer & AarisSrensen, 1998). On the other hand, in South Europe, roe deer have always been present during the Late Pleistocene and the Holocene (Gurin & PatouMathis, 1996). For at least 400 years, the roe deer has been a popular game animal and has undergone frequent local extinctions, reintroductions, and translocations. These man-induced movements may greatly mislead phylogeographical conclusions, unless a strict selection of sampling localities is ensured, considering only autochthonous stock. We analysed mitochondrial and nuclear variation in 14 autochthonous populations of roe deer across their whole European range to address several complementary questions concerning the phylogeography of this cervid. Mitochondrial DNA has been widely used to assess intraspecic variation (Feulner et al., 2004; Ludt et al., 2004). However, because mtDNA is maternally inherited, data on autosomal biparental markers (i.e. the highly variable microsatellites) are needed to avoid a female-biased description of population structure. In the present study, we attempt to discuss our results in a phylogeographical perspective by inferring the patterns of population dispersal and evolution of roe deer in the Pleistocene.

Pleniglacial (2318 000 years ago), where deciduous vegetation and conifer forests occurred: one included three-quarters of the Iberian peninsula; one extended from the northern Apennines down to the Pollino massif, in the central and southern parts of the Italian peninsula; a third large area was located approximately from the Drava river down to include nearly all the Balkans; a fourth refuge area, adjacent to the last one, extended from south of the Sudety, through the High Tatra to the Carpathian mountains (Zagwijn, 1992). These areas were surrounded by steppe-tundra and boreal meadows in the north, as well as steppe in the south, interspersed with Alpine meadows and glaciers on mountain tops: all highly unsuitable habitat for a warm-temperate adapted species such as the roe deer, Capreolus capreolus (Linnaeus, 1758), although it may also survive in alpine and steppe habitats. From data of mitochondrial DNA sequence variation, Wiehler & Tiedemann (1998) suggested that two different genetic lineages of roe deer might reect the former existence of two refuge areas, in south-western and south-eastern Europe, respectively, which were the origins of postglacial recolonization waves in the Late Pleistocene. These waves may have met in central Europe and determined a mixture of haplotypes of these two lineages. Alternatively, the same authors proposed that just one refuge area could have occurred, in western or southern Europe, from which roe deer could have moved eastwards. Recently, Randi et al. (2004) identied three clades of haplotypes, one of which was widespread throughout Europe. The other two clades, sampled mainly in northern Spain, Portugal, and in southern Europe (Greece, Serbia), respectively, might have originated in the Iberian and Balkanic refugia during the last cold episodes (Randi et al., 2004). Genetic distinction was not observed for roe deer in central and southern Spain. According to their molecular results, these authors suggested an admixed origin (i.e. polyphyly), for those peripheral populations. By contrast, a unique clade of haplotypes and signicant differences at microsatellite loci were found in the roe deer sampled from central and southern Apennines, thus supporting the existence of the Italian subspecies Capreolus c. italicus (Festa, 1925; Lorenzini, Lovari & Masseti, 2002; Randi et al., 2004). Presently, the roe deer occurs throughout the West Palaearctic region, from the Iberian peninsula to South Scandinavia, and eastwards to the Caucasus. This small ungulate (Artiodactyla: Cervidae) is a generalist browser, which can also make use of grasses and sedges (Duncan et al., 1998), that thrives in deciduous wooded areas with glades and clearings, as well as in elds interspersed with wood patches, mostly in temperate climates (Andersen, Duncan & Linnell, 1998). The origin of this species is unclear. The oldest

MATERIAL AND METHODS


SAMPLING
AND

DNA

ISOLATION

A total of 376 roe deer samples were collected between 1999 and 2003 from the following sites across Europe (Fig. 1): central-southern Italy (Castelporziano Presidential Estate, Gargano and Pollino National Parks, Maremma Regional Park, Siena and Grosseto provinces), n = 127; eastern Italian Alps (Val Rendena), n = 13; northern Spain (Asturias, Lugo, Burgos,

2006 The Linnean Society of London, Biological Journal of the Linnean Society, 2006, 88, 85100

GENETIC DIVERSITY AND PHYLOGEOGRAPHY OF ROE DEER

87

Figure 1. Collection sites of roe deer samples. 1, France; 2, Sweden; 3, Denmark; 4, Austria; 5, central-southern Italy; 6, eastern Italian Alps; 7, northern Spain; 8, southern Spain; 9, Portugal; 10, Greece; 11, Romania; 12, Poland; 13, Lithuania; 14, Turkey. Multiple sampling locations for one population are indicated by the same number. The shaded areas show the traditional Pleistocene refugia (see text for details).

Basque Country, La Rioja, Huesca, Zaragoza, Soria), n = 66; southern Spain (Toledo, Cceres, Ciudad Real, Cdiz), n = 55; Portugal (Peneda-Gers National Park, Douro International Natural Park), n = 9; France (Aquitaine, Chiz), n = 25; Denmark (Kal), n = 10; Sweden, n = 6; Austria (Styria, Lower Austria), n = 23, Poland (Bia l owieza, Bieszczady Mountains), n = 8; Romania, n = 10; Greece (Fokida, Serres), n = 6; Turkey (Ankara zoo, stock from Turkey), n = 3; and Lithuania (Panevezys, Jurbarkas, Moletai, Sirvintos, Kaunas), n = 15. We ensured that specimens were collected from allegedly native populations, or from locations where (re)introductions have never been documented, to avoid any complexity of population structure due to human interference (with the exception of the Swedish roe deer, which belonged to a reintroduced population from a local stock). Tissue samples were obtained from animals killed during the hunting season. Samples of roe deer from protected areas consisted of blood, ear disks (5 mm in diameter) or hair gathered from live-caught animals. Only scats

were available for the endangered roe deer from Turkey. Total genomic DNA was isolated from either skeletal muscle and ear cartilage preserved in 75% ethanol, or blood collected in EDTA-coated tubes, by standard methods (Sambrook & Russel, 2001). DNA from hair roots and scats was isolated using the Qiagen DNeasy Tissue kit and the Qiamp Stool kit, respectively. DNA quality and amount of these samples allowed only microsatellite analysis.

RESTRICTION FRAGMENT LENGTH POLYMORPHISMS (RFLPS) OF MITOCHONDRIAL DNA AND


MICROSATELLITE ANALYSIS

Two mitochondrial segments, the D-loop (D-loop, 1 kb in length) and NADH dehydrogenase 1 (ND1, 1.7 kb in length), were polymerase chain reaction (PCR)-amplied using the primers characterized by Jger, Hecht & Herzog (1992) and Cronin et al. (1994). Endonucleases with tetranucleotide recognition sites ( AluI, BstUI, DdeI, HaeIII, HhaI, Hsp92II, MboI, MspI, RsaI, ScrFI,

2006 The Linnean Society of London, Biological Journal of the Linnean Society, 2006, 88, 85100

88

R. LORENZINI and S. LOVARI per locus because it yields a measure of allelic diversity corrected for differences in sample sizes. Measures of alleles per locus are adjusted to the smallest number of individuals typed for a locus in a sample. Fishers exact test was performed to check for genotypic linkage disequilibrium for all pairs of loci by employing the Markov chain method. More than two alleles were present at each locus, and therefore, deviations from HW expectations were evaluated with a probability test using the Markov chain of Guo & Thompson (1992). Standard Bonferroni correction for multiple tests was employed to adjust the signicance level. The sample of Turkish roe deer was excluded from this test, due to small size (n = 3). Fixation index FST was calculated both as an overall value and for pairs of populations, and signicance was assessed by permutation tests. FST values are obtained from variance in allele frequencies, and provide reliable measures of genetic distance when divergence between populations is caused by drift (Reynolds, Weir & Cockerham, 1983) under the Innite Alleles Model (IAM) (Kimura & Crow, 1964). When microsatellites are concerned, other genetic distance measures may be more appropriate (for a review, see Balloux & Lugon-Moulin, 2002). Therefore, we also calculated RST (Slatkin, 1995), an analogue to FST adapted to microsatellite loci, which accounts for variance in allele sizes under the Stepwise Mutation Model (SMM) (Kimura & Otha, 1978). To test for congruence between pairwise values of FST and RST, a Mantel correlation test was applied with 1000 iterations using the program MANTEL version 2.0 (Liedloff, 1999). Besides FST and RST, other measures of genetic distance were calculated based on allele frequencies under the IAM (chord distance Dc, Cavalli-Sforza & Edwards, 1967; Neis standard genetic distance D, Nei, 1972), and based on allele size under the SMM [()2, Goldstein et al., 1995], using the programme MICROSAT (Minch et al., 1995). A drift-based measure, the proportion of shared alleles (Dps, Bowcock et al., 1994), which is not based upon any underlying mutation model, was also calculated. The programme NEIGHBOUR in PHYLIP version 3.6 (alpha3) (Felsenstein, 1993) was used to construct Neighbour-joining trees from the distance matrices. Statistical support at the internodes on the trees was assessed by 1000 bootstrap replications employing the CONSENSE programme included in the PHYLIP package. The software FSTAT, version 2.9.3 (Goudet, 1995), GENEPOP, version 3.3 (Raymond & Rousset, 1995), and the package ARLEQUIN, version 2.0 (Schneider, Roessli & Excofer, 2000) were also used to conduct the analyses. As an alternative approach to explore differentiation of populations, a Factorial Correspondence Analysis (FCA) (Benzecri, 1973) was employed to cluster individual microsatellite proles in a multidimen-

TaqI, Tru9I, Tsp509I, BfaI, AciI) and pentanucleotide sites (HinfI) were used to produce restriction digests for each segment. The specimens were genotyped at ten polymorphic microsatellite loci using the primer pairs RT1, RT7, NVHRT16, NVHRT21, NVHRT24, NVHRT30, NVHRT48, ROE06, ROE08, and ROE09, according to Wilson et al. (1997), Red & Midthjell (1998), and Fickel & Reinsch (2000). Amplied products were multiplex pooled in individual tubes into groups of ve loci, depending on product size and uorescent dye label. Pools were heated at 95 C for 2 min, placed on ice and loaded onto an ABI Prism 310 Genetic Analyser. Methods for PCR amplications of mtDNA segments, microsatellites, and for digestions followed Lorenzini et al. (2002).

STATISTICAL

METHODS

Mitochondrial DNA The proportion of shared restriction fragments was used to estimate genetic distance (Nei, 1987) in terms of nucleotide sequence divergence ( p, base substitutions per nucleotide) between all pairs of haplotypes and populations following the fragment method of Nei & Miller (1990) as implemented in the programme RESTSITE (Miller, 1991). For population comparisons, the values of genetic distance were corrected for within-population diversity. All fragments, variant and invariant, were used in the calculations. Populations and haplotypes were clustered on unrooted Neighbour-joining trees (Saitou & Nei, 1987) constructed from the matrices of the p-values derived by RESTSITE. Cervus elaphus was used as an outgroup to dene the phylogenetic relationships of Capreolus capreolus populations. Robustness of tree topologies was assessed through 1000 bootstrap replications. Sample haplotype frequencies were used to calculate the overall xation index FST as an estimator of genetic subdivision of populations. Differentiation between pairs of populations was evaluated by pairwise FST values, with signicance at the 5% nominal level being tested by a non-parametric permutation procedure (Excofer, Smouse & Quattro, 1992), after standard Bonferroni correction for multiple testing (Rice, 1989). Genetic diversity within populations was assessed by estimating nucleotide () and haplotype (h) diversity according to Nei (1987). Microsatellite data Genetic polymorphism for each population was measured (sample size of at least six) as the mean values of observed heterozygosity ( Ho), heterozygosity expected (He) from HardyWeinberg (HW) assumptions (Nei, 1987), and allelic richness. Allelic richness was calculated instead of the mean number of alleles

2006 The Linnean Society of London, Biological Journal of the Linnean Society, 2006, 88, 85100

GENETIC DIVERSITY AND PHYLOGEOGRAPHY OF ROE DEER sional space, using the algorithm implemented in GENETIX (Belkhir et al., 1999). We applied a Bayesian method, through the modelbased clustering algorithm as used in the programme STRUCTURE, to investigate further the pattern of population structure (Pritchard, Stephens & Donnelly, 2000). This method assumes that loci are at HW equilibrium, and linkage equilibrium, and uses multilocus genotypes to identify clusters of genetically similar individuals, which are assigned to K theoretical populations. The probability of the number of populations equalling K was calculated from the estimated posterior log-likelihood of the data, which assumes the highest estimate with the most probable value of K. The proportional membership ( qi) of individual genotypes was derived for each of the K theoretical populations. An individual was arbitrarily assigned to one inferred cluster if the proportion of its genotype assigned to that cluster was greater than 75%. We ran the program for 100 000 iterations after a burn-in period of 10 000.

89

Table 1. Mitochondrial DNA composite haplotypes observed in roe deer populations from Europe Haplotype G E PP P H L I M Q O R N D ST Y W T QQ BB S A AA B C Z ZA TT SS MM UU VV YY
a

D-loopa AAABABA AAAAAAB AAABADA AAAADAB ABABAAB BAABBBA ABCCACA AAABAAB ABAAACA ABBCACA ABABADB AABCABA AAAAAAA AAABAAA AAAEAAA AAAAABB AABCAAB BBAAACA ABAAACB ABBDACA AAAACCB AAAAECB ABAAABA AACACBB AAAACBB AACACCB ABAFAEC ABAGAEC AACHCFB DCAAACA CDAIABC CEALDFB

ND1 AAAA AAAA AAAA AAAA ABAB AAAA AAAA AAAA AAAA AAAA AAAB AAAA AAAA AAAA AAAA CAAA ACAA AAAA AAAA AAAA BABA BABA AAAA BABA BABA BABA DDCA DDCA BABA AAAA DDCA EABA

RESULTS
GENETIC
VARIABILITY WITHIN POPULATIONS

Thirty-two composite haplotypes were dened in the sampled roe deer populations by digestions of two mtDNA fragments (Table 1). Eleven out of 32 enzymesegment combinations produced variable patterns of restriction fragments. Different RFLP proles were detected with seven and four of 16 restriction enzymes in the D-loop and ND1, respectively. The D-loop was highly polymorphic, producing 32 different restriction patterns in the 358 individuals surveyed. ND1 proved less informative, with eight RFLP proles identied. With the exception of Swedish roe deer, all the populations were polymorphic, showing from two (Greece, Denmark) to six (northern Spain) different haplotypes. Values of haplotype diversity (Table 2) ranged from h = 0.333 in roe deer from Greece, to h = 0.822 in roe deer from Romania. In samples where haplotypic variation occurred, values of nucleotide diversity of haplotypes ranged from = 0.021% in roe deer from Greece to = 0.188% in roe deer from Lithuania. On average, mitochondrial DNA variation was higher in the easternmost populations (Romania, Poland) and eastern Italian Alps. Microsatellite genotypes at ten loci were determined for 368 individuals. A total of 114 distinct alleles were detected over the complete data set, yielding a mean number of 11.4 alleles per locus. Loci displayed high levels of polymorphism across samples, showing from three (ROE09) to 17 (NVHRT21) alleles, and values of mean observed (expected) heterozygosity which ranged from 0.335 (0.381) to 0.720 (0.810) for ROE09

From left to right, letters refer to: D-loop: DdeI, Sau3A, HinfI, Hsp92II, Tru9I, Tsp509I, AciI; ND1: DdeI, HinfI, RsaI, Hsp92II.

and RT1, respectively. Two pairs of loci (NVHRT21/ RT1, NVHRT48/RT7) were linked (tests for linkage disequilibrium across populations, P < 0.001, Bonferroni corrected). However, the exclusion of two of these loci (RT1 and RT7) did not affect the results of the multilocus calculations employed in the analyses. Therefore, each of ten microsatellite markers was considered as providing independent genetic information, and was entered in the computations. Measures of genetic diversity for each population were calculated from the observed distributions of allele frequencies (Table 3). Allelic richness within most populations ranged between 3.3 and 4.7, with roe deer from Poland showing the greatest value (5.0). All populations revealed on average high levels of variability in terms

2006 The Linnean Society of London, Biological Journal of the Linnean Society, 2006, 88, 85100

90

R. LORENZINI and S. LOVARI


*Haplotype diversity (h) was calculated as 2n(1 fi2)/(2n 1), where fi is the frequency of the ith haplotype and n is the number of individuals sampled; nucleotide diversity () was calculated as xi xj ij, where xi is the population frequency of the ith sequence, xj is the population frequency of the jth sequence and ij is the number of nucleotide differences per nucleotide site between the ith and jth sequences. Standard deviations are given in parentheses.

C Z ZA MM Y T S ST TT SS QQ BB UU VV YY h*

of He, which ranged from 0.448 for roe deer from Sweden to 0.765 for roe deer from France. The Monte Carlo approximation of the Fishers exact test for all loci combined showed signicant deviations from HW expectations (P < 0.05, after Bonferroni adjustment) in roe deer populations from central-southern Italy, France, Romania, northern Spain, southern Spain, and departures were recorded at 7, 3, 2, 4, and 6 loci, respectively. HardyWeinberg proportions at single loci deviated due to signicant decits of heterozygotes (P < 0.05 for all FIS values within populations). Disequilibrium at many loci within populations suggests a strong Wahlund effect (i.e. the existence at lower geographical scale of internal subpopulations), possibly with restricted gene ow (as in our case), rather than indicating the presence of nonamplifying alleles (Pemberton et al., 1995). Null alleles or allelic dropout, which are minor, although real, sources of errors, can be excluded, given that they should be present at high frequency at many loci to produce such deviations, and DNA quality was good for all samples, with the exception of scat samples of Turkish roe deer. Sex-linkage was also excluded for all loci.

(0.013) (0.059) (0.036) (0.015) (0.055) (0.031) (0.023) (%) (0.043) (0.059) (0.054) (0.073) (0.160) (0.095) (0.013)

10

0.366 0.795 0.614 0.367 0.667 0.703 0.467 0.0 0.576 0.822 0.750 0.333 0.533

(0.108) (0.097) (0.139) (0.215) (0.126)

0.023 0.107 0.066 0.024 0.090 0.052 0.036 0.0 0.045 0.094 0.124 0.021 0.188

(0.026) (0.054) (0.072) (0.017) (0.099)

POPULATION

GENETIC STRUCTURE

The distribution of mitochondrial haplotypes showed a clear segregation according to their occurrence in populations (Table 2). Apart from haplotypes E and D, which are widely distributed across populations, most haplotypes occurred at high frequencies in, or were restricted to, single populations. For example, haplotypes G and L were frequent and conned to the roe deer of central-southern Italy, while haplotypes H and O were common in those from the Italian alpine area. B was found as a highly frequent haplotype in northern Spain, whereas haplotype A was the most widespread in southern Spain, and AA was limited to Portugal. The same holds for roe deer in Romania, Poland, Greece, and Lithuania, where most of their haplotypes are private. By contrast, populations from Austria, France, Denmark, Sweden and, to a lesser extent, northern Spain, shared their most common haplotypes (D, E), thus showing uniformity at the mitochondrial level. A Neighbour-joining tree of mitochondrial haplotypes (Fig. 2) was built on the values of pairwise nucleotide divergence (not shown). Haplotypes grouped with high statistical support ( 80%) into three main clusters, but only two of them, both supported by high bootstrap values at all internodes, showed a geographical pattern. One cluster (Haplogroup I) included haplotypes SS, TT and VV, YY found in Poland and Lithuania, respectively; the other cluster (Haplogroup II) comprises the haplotypes present in southern Spain and Portugal (Z, A, AA, ZA, C, with the exception of MM which was found in France). A

AA B

W A

15 1

4 43 7

38 5 9 1 2

6 4 1 1 2 4 2

M H P Q O R D

Table 2. Distribution of roe deer mtDNA haplotypes

PP N L

Haplotypes

Central-southern Italy (126) 97 2 Eastern Italian Alps (13) 2 Northern Spain (66) 3 Southern Spain (55) Portugal (7) France (14) 3 Denmark (10) 7 Sweden (6) Austria (22) 14 Romania (10) Poland (8) 1 Greece (6) Lithuania (15)

Population (N)

20 2

2006 The Linnean Society of London, Biological Journal of the Linnean Society, 2006, 88, 85100

GENETIC DIVERSITY AND PHYLOGEOGRAPHY OF ROE DEER

91

Table 3. Measures of genetic variability in 13 European populations of roe deer, as estimated by 10 microsatellite loci Population (N) France (25) Poland (8) Lithuania (15) Romania (10) Northern Spain (66) Denmark (10) Austria (23) Central-southern Italy (127) Eastern Italian Alps (13) Greece (6) Portugal (9) Southern Spain (47) Sweden (6) Ho 0.639 0.686 0.664 0.517 0.556 0.620 0.649 0.451 0.549 0.450 0.479 0.391 0.483 (0.044) (0.089) (0.067) (0.092) (0.051) (0.062) (0.086) (0.050) (0.076) (0.093) (0.083) (0.072) (0.068) He 0.765 0.694 0.680 0.679 0.673 0.662 0.659 0.626 0.607 0.561 0.538 0.520 0.448 (0.021) (0.071) (0.056) (0.074) (0.056) (0.048) (0.075) (0.070) (0.078) (0.104) (0.081) (0.088) (0.059) Allelic richness 4.7 5.0 4.4 4.6 4.3 3.8 4.6 4.0 3.5 3.8 3.3 3.5 2.7

The sample from Turkey is not included because of small size. He (Ho), mean expected (observed) heterozygosity. Standard errors are given in parentheses.

third cluster (Haplogroup III) groups all other haplotypes, which are widespread throughout all populations, including Portugal and the eastern ones (Lithuania and Poland), but not southern Spain. When looking at haplotype frequencies, they were signicantly distributed between populations, as suggested by the overall FST value = 0.637 (P < 0.001, 10000 permutations), which indicates that 64% of total variation was due to differences among populations. Signicant pairwise FST values (P < 0.05, after standard Bonferroni correction, 1000 permutations) were obtained in all but ve population comparisons, involving France/Austria, France/Denmark, France/ Sweden, Austria/Denmark, and Romania/eastern Italian Alps pairs, respectively. Roe deer populations were assorted on a Neighbourjoining tree (Fig. 3A) derived from a matrix of pairwise nucleotide divergence between populations, with Cervus elaphus used as an outgroup. Roe deer from Poland are located at the root of the tree with 74% bootstrap support and, on average, they are the most divergent ones, showing the highest mean SD distance (0.45 0.069%). Populations from Lithuania, southern Spain, and Portugal were strikingly divergent as well, and occurred on separate branches supported by consistent bootstrap values. A major cluster can be identied with 65% statistical support at the most internal internode which encompasses all the remaining populations. However, only the subcluster formed by roe deer from Austria, France, Denmark, and Sweden is consistently located within the cluster (bootstrap values > 74%). By contrast, Greece, Romania and eastern Italian Alps, central-southern Italy, and northern Spain occur on separate branches, but the low bootstrap values at most internodes make

their genetic relationships less clear. Genetic distances ranged from 0.001% between France and Denmark (the negative values indicating that withinpopulation diversity exceeded the difference between populations), to 0.604% between Poland and southern Spain, with a mean SD value of 0.239 0.156%. Populations differed signicantly in allele frequencies at microsatellite loci, as suggested by the overall values of FST = 0.182 (P < 0.001, 10000 permutations). Mean RST was 0.235. These values suggest that approximately 20% of the total variation at microsatellite loci is distributed among populations. Seventeen percent of all alleles were private (1.4 alleles per population; SD 1.4). Private alleles, which ranged in number from zero in Portugal, Greece, France, Austria, and Sweden, to four in central-southern Italy, were present at very low frequencies ( < 0.17), and therefore they were neither representative of their respective populations, nor contributed much to the distinctiveness of single populations. The only exceptions are the Danish roe deer, which show one allele at locus ROE08 occurring at the appreciable frequency of 0.250, and those from Turkey, where one private allele at locus NVH21 was present with a frequency of 0.5. FST values were signicant through 1000 permutations in all pairwise population comparisons according to the 5% criterion, except for comparisons involving Romanian roe deer vs. those from Poland, Lithuania, and Turkey, and Polish roe deer vs. those from Lithuania. A Mantel correlation analysis between estimates of pairwise FST and RST revealed the two matrices to be signicantly correlated (g = 5.11, Z = 8.79, r = 0.76, P < 0.001), indicating that both the distribution of allele frequencies and allele size accounted for distinction among pairs of populations.

2006 The Linnean Society of London, Biological Journal of the Linnean Society, 2006, 88, 85100

92

R. LORENZINI and S. LOVARI


PP G
57

L ST
67

Y M H
80

Haplogroup III
(Northern Spain, Portugal, North, South, Central and East Europe)

R B BB
81 60

Q I
58

O
73

QQ UU

W
84

D E P N
58

Haplogroup II
(Southern Spain, Portugal)

Z
80 79

A AA

51 71 63

ZA C MM

Haplogroup I
(Eastern Europe) 68 72 85

SS TT VV YY Cervus elaphus

0.001

Figure 2. Neighbour-joining tree based on pairwise nucleotide divergence of mtDNA haplotypes. Branch lengths reect genetic distances according to scale, and numbers at the nodes show support from 1000 bootstrap replicates when values are over 50%.
2006 The Linnean Society of London, Biological Journal of the Linnean Society, 2006, 88, 85100

GENETIC DIVERSITY AND PHYLOGEOGRAPHY OF ROE DEER

93

A
Greece Eastern Italian Alps
65 71

Romania Central Southern Italy

Northern Spain
74

Austria
75

Sweden
69 81

Denmark
79

France
74

85

Portugal Lithuania

Southern Spain

Poland

Cervus elaphus
0.001

B
Southern Spain
75

Northern Spain Denmark


85

France Sweden Eastern Italian Alps Portugal

Groups II III

74

Austria
81

Lithuania Romania
65

Poland Central Southern Italy

Group I

74

Greece

Turkey

Cervus elaphus

Figure 3. Neighbour-joining trees of populations. A, topology based on pairwise nucleotide divergence of mtDNA haplotypes. B, consensus microsatellite tree based on pairwise [( )2] values. Bootstrap support at the nodes (1000 replicates) are indicated only when values exceed 50%. Both trees were arbitrarily rooted at the outgroup mtDNA haplotypes and microsatellite genotypes, respectively, of the red deer Cervus elaphus. Group I comprises central-southern Italy, Austria, Turkey, Greece, Poland, Lithuania, Romania; Group II comprises northern Spain, southern Spain, Portugal; Group III comprises France, Sweden, Denmark, eastern Italian Alps.

2006 The Linnean Society of London, Biological Journal of the Linnean Society, 2006, 88, 85100

94

R. LORENZINI and S. LOVARI

GroupI I Group

GroupIIII Group

Axis 2 (25%)

-5

- 10

GroupIII III Group


-5 0

Axis 1 (75% )

10

Figure 4. Factorial correspondence analysis of individual roe deer. The proportion of total variation explained by the rst two factorial dimensions are indicated at the axes. Group I comprises central-southern Italy, Austria, Turkey, Greece, Poland, Lithuania, Romania; Group II comprises northern Spain, southern Spain, Portugal; Group III comprises France, Sweden, Denmark, eastern Italian Alps.

The overall pattern of genetic relationships between populations was examined by deriving Neighbourjoining trees constructed from the matrices of genetic distances. The distances based on allele size variation under the SMM model [ RST ()2] produced identical topologies; therefore, only the consensus tree built on the ()2-values is shown (Fig. 3B). In the tree, populations were grouped into two major clusters. One cluster (indicated as Groups II and III in Fig. 3B) comprises populations occurring in the Iberian peninsula, in central-northern Europe (France, Denmark, Sweden) and eastern Italian Alps, while the other cluster (indicated as Group I) encompasses populations from southern and eastern Europe (central-southern Italy, Greece, Turkey, Poland, Lithuania, Romania) and Austria. Both groups are well supported (bootstrap values = 74%) at the most internal internodes of the tree. Using the IAM-based distances calculated from allele frequency (FST, Dc, Neis D) and the negative logarithm of the proportion of shared alleles (log Dps), the general topology of the trees (not shown) remained unchanged with respect to that shown in Figure 3B. However, for distantly related populations, the IAMbased distances increase less than the SMM-based distances. Consequently, the divergence between the two clusters is smaller. The plot of results from a FCA (Fig. 4) of individual multilocus genotypes conrmed the roe deer populations from southern and eastern Europe as a distinct

group (Group I) and, furthermore, separated roe deer of the Iberian peninsula (Group II) from the populations of central-northern Europe (Group III), which, by contrast, appeared as a single group in the Neighbour-joining tree (Fig. 3B). Groups I and II were completely separated on the rst factorial component (Axis 1), which explained 75% of the total genetic variation, whereas Group III had intermediate component scores and was only partly distinct (mainly on Axis 2, which represented 25% of the total variation), given that a slight overlap with both the other groups is apparent. Population structure was further investigated using a likelihood-based method of clustering. Individual multilocus genotypes were grouped into K panmictic clusters by the Bayesian method. Setting K as unknown a priori, the highest log-probability of our data was obtained for K = 5 (Table 4), indicating that approximately ve genetically distinct populations are appropriate to account for the actual data set. The values of (qi), the proportional membership of individual roe deer in each of the ve theoretical populations, allowed separation of the group of populations from southern and eastern Europe (Group I in Figs 3B, 4) into two further groups. One group (PI in Table 4) included the eastern roe deer, with 91% of the total assigned individuals (64%) showing ( qi) values falling in cluster 3. The other group (PIV in Table 4) encompassed the populations of central-southern Italy, with

2006 The Linnean Society of London, Biological Journal of the Linnean Society, 2006, 88, 85100

GENETIC DIVERSITY AND PHYLOGEOGRAPHY OF ROE DEER


Table 4. Bayesian analysis of the roe deer sampled Percent of assigned individuals 64 82 51 86

95

Populations PI Group I PII Group II PIII Group III PIV Group IV

Cluster 1 0.078 0.009 0.127 0.527

Cluster 2 0.038 0.008 0.037 0.399

Cluster 3 0.782 0.006 0.429 0.038

Cluster 4 0.062 0.325 0.271 0.019

Cluster 5 0.039 0.652 0.136 0.017

Percent of individuals assigned to not-shared clusters 91 100 0 100

Mean proportion of membership ( qi) of populations in ve clusters inferred by using STRUCTURE. PI: Greece, Turkey, Romania, Poland, Austria, Lithuania; PII: northern Spain, southern Spain, Portugal; PIII: eastern Italian Alps, France, Sweden, Denmark; PIV: central-southern Italy. Individuals with (qi) values < 0.75 were not assigned to any cluster.

100% of the total assigned individuals (86%) which showed (qi) values falling in clusters 1 and 2. The assignment of individuals of one group to two clusters is due to the presence of internal subpopulations (see also group PII). Clusters 4 and 5 grouped 100% of the total assigned individuals (82%) from the Iberian peninsula (PII in Table 4 and Group II in Fig. 4). Roe deer of central-northern Europe (PIII in Table 4 and Group III in Fig. 4) were admixed and portions of their ancestry fell partially into all but one cluster, a gure consistent with results from the FCA analysis. Globally, 25% of all individuals showed ( qi) values < 75% and were left unassigned.

DISCUSSION
RELATIONSHIPS
BETWEEN POPULATIONS

Analysis of RFLPs in two mitochondrial gene segments, the D-loop and NDH-1, and of ten autosomal microsatellite loci yielded high levels of intraspecic variability in the European roe deer, in accordance with other studies based on biochemical and DNA markers (Hartl et al., 1991, 1993; Lorenzini et al., 1993; Hewison, 1995; Wiehler & Tiedemann, 1998). High variability has been observed within all populations studied, except for the roe deer of Sweden, which exhibit the lowest values of expected heterozygosity and allelic richness at microsatellite loci and no mitochondrial variation. This low level of variability is reasonably due to a founder effect, rather than to a bias because of small sample size. A local, bottlenecked stock of around 100 individuals was the source population for all the re-colonizations of roe deer in Scandinavia (R. Andersen and M. Hewison, ex verbis), after the species went nearly extinct from most parts of the peninsula in the last century (Lehmann & Sgesser, 1986). Because of difculties in getting large enough sample sizes for all collection sites, some populations are under-represented in the present survey. Although aware of potential bias due to limited sampling, we

observed high levels of genetic variation also in small sized samples (see Poland for microsatellites, or Portugal for mtDNA). We feel rather condent that, even though some genetic information may have gone lost, this does not affect seriously the general trend of genetic relationships among populations. Most of the variation in haplotype frequencies is signicantly partitioned among populations, revealing a remarkable geographical pattern. Mitochondrial differentiation, in terms of both haplotype frequency and sequence divergence, concerns particularly roe deer populations from southern Spain and Portugal on the one hand, as well as Poland and Lithuania on the other, in the south-western- and north-easternmost areas of the species range. In the mtDNA tree, populations (northern Spain included) geographically intermediate between the extremes of the distribution join into the same cluster, although well separated from each other. This pattern indicates genetic distinctiveness of single populations. By contrast, the most external cluster, including roe deer of Austria, Sweden, Denmark, and France, suggests the presence of genetically rather homogeneous roe deer populations in central and northern Europe (cf. Wang & Schreiber, 2001). Wang & Schreiber (2001) undertook a comprehensive investigation, using ten allozyme markers and nine microsatellite loci, on roe deer across the Netherlands, Germany, and France. Their study highlighted a low level of population subdivision throughout central Europe, mainly at a local scale, probably due to the ecology and spatial behaviour of individual populations, rather than to isolation-by-distance or any regional groupings. By contrast to mtDNA, the analysis of microsatellite loci revealed that most of the nuclear variation is distributed within populations. Only 20% variation depends on signicant differences between populations. These mild differences made the pattern of genetic relationships difcult to establish. Moreover, a high mutation rate and mode of variation for microsatellites (increasedecrease in tandem repeats, allele

2006 The Linnean Society of London, Biological Journal of the Linnean Society, 2006, 88, 85100

96

R. LORENZINI and S. LOVARI tions in central and eastern Europe (where these haplotypes are the most widespread), and those living in the south of the peninsula, which identies a sublineage of haplotypes not found outside Iberia. This nding suggests a long-term isolation for the latter. By contrast, an admixed origin for roe deer in Spain has been proposed (Randi et al., 2004). Haplotypes derived from a clade which was widespread throughout Europe were sampled in the central and southern areas of the Iberian peninsula. Randi et al. (2004) suggested that, if this clade originated in eastern Europe, then the monophyly of the Spanish putative subspecies, C. c. garganta (Meunier, 1983), would not be supported by molecular data. In Portugal, the extant native populations derive from the north-western areas of Spain by natural spreading. The majority of our samples came from the Peneda-Gers National Park, north of the Douro River, where a relict nucleus from a panmictic Pleistocene population is supposed to have survived, after the species restricted its distribution in the peninsula, rst because of climatic changes and later because of human activities (J. Vingada, pers. comm.). If this is the case, it is not surprising that haplotypes shared with both the northern and southern lineages of Spain have been found in the Portuguese samples. Furthermore, their private haplotype AA clusters with the southern haplogroup. By contrast, all microsatellite alleles are shared with the northern populations, but only 70% are in common with the southern ones. Thus, maternal DNA data support the view of an ancient native origin of this population, because recent introductions from southern Spain to Portugal appear to be very unlikely (C. San Jos, ex verbis). On the other hand, biparental markers indicate that admixture with their northern neighbours occurred recently, possibly through malemediated gene ow. Phylogenetic and management implications for the Iberian roe deer were previously discussed in detail by Lorenzini et al. (2003). The eastern and southern populations of Romania, Poland, Lithuania, Turkey, and Greece represent the second genetic lineage. The most divergent haplogroup was sampled within this lineage, in the roe deer from Poland and Lithuania. These populations are highly differentiated according to mtDNA, but not to microsatellites, which yielded the highest divergence for the populations of Greece and Turkey, probably as a consequence of their recent genetic drift, or, alternatively, because of a sample bias. Within the eastern and southern group, a sublineage is represented by populations from central-southern Italy, where the existence of the subspecies C. c. italicus was proposed at the beginning of the last century based on mild divergence in morphologic traits (Festa, 1925), and as later supported by molecular evidence (Lorenzini, Lovari & Masseti, 2002; Randi et al., 2004). By con-

size constrains) may rapidly lead to coalescence. Thus, underestimation of genetic distances between more divergent populations can occur. Consequently, different statistical approaches for analysing microsatellite data were needed to achieve a reasonable picture of population subdivision. Classical clustering methods (Neighbour-joining tree) led to a partition of European roe deer into two main groups: one included populations of southern and eastern Europe (centralsouthern Italy, Greece, Turkey, Austria, Romania, Lithuania, Poland), while the other one united populations of western, central, and northern Europe (Iberian peninsula, eastern Italian Alps, Sweden, Denmark, France). Randi et al. (2004) warned against the use of pairwise genetic distances to infer mtDNA or microsatellite trees of populations, if populations are not in mutation-drift equilibrium. As a game species, the roe deer has been intensively managed since historical times. Founder events, bottlenecks, and isolation due to human manipulation may have led to a rapid differentiation of populations through random sampling of genetic variants. However, we deliberately collected samples exclusively from allegedly autochthonous populations, which were assumed not to have undergone severe natural or anthropogenic bottlenecks in the recent past (Swedish sample excluded), nor had they been reinforced by releasing imported deer. This should have minimized the risk of misleading phylogeographical inferences. Moreover, in addition to genetic distance-based clustering methods, different approaches to microsatellite data were used to derive population structure, which supported basically the same view. Compared to the Neighbourjoining tree, FCA allowed a further separation of the Iberian roe deer (northern and southern Spain, Portugal) from populations of the former western-centralnorthern cluster, while maintaining all the southern and eastern populations as a single separate group. A likelihood-based Bayesian approach showed a ner resolution, while supporting the results derived from traditional clustering methods and FCA. Probabilistic assignment of the individuals to distinct genetic clusters allowed the separation of roe deer of centralsouthern Italy from those of the southern and eastern group. In summary, mitochondrial, and microsatellite analyses supplied partly overlapping and partly complementary information, leading to an exhaustive picture of the intraspecic genetic differentiation within C. capreolus. Three major genetic lineages of roe deer are apparently present in Europe. One includes populations in the Iberian peninsula, which represent the south-western limit of the distribution and where an internal substructure of populations is also present. Signicant genetic differentiation was observed between roe deer populations in northern Spain, which share haplotypes D and E with other popula-

2006 The Linnean Society of London, Biological Journal of the Linnean Society, 2006, 88, 85100

GENETIC DIVERSITY AND PHYLOGEOGRAPHY OF ROE DEER trast, both mitochondrial and nuclear data suggest the presence of genetically rather homogeneous, although admixed roe deer in central and northern Europe (France, Denmark, Sweden) (cf. Randi et al., 2004), representing a third major lineage, somewhat less separated and almost equally divergent from both the eastern and western lineages. Roe deer from the Italian Alps and Austria appear to be intermediate between the central and eastern lineages. In particular, according to nuclear markers, but not to mtDNA, roe deer of Austria appear more similar to those of the eastern lineage than to those of the central-northern group. The opposite happens for roe deer of the Italian Alps, included in the central-northern lineage according to the microsatellite data, but not to the mtDNA data.

97

PHYLOGEOGRAPHY

OF THE ROE DEER AND

POSTGLACIAL DISPERSAL PATTERNS

The roe deer was present in the Iberian and Balkan peninsulae, as well as in southern Italy during the last glacial (e.g. Gliozzi et al., 1997; Hufthammer & AarisSrensen, 1998; Petit-Maire & Vrielynck, 2004), probably surviving the coldest climatic waves in relatively dry and warm pockets. In the last stadial of Wrm, the climate was one of the coldest ever and quite arid. Thus, the direct ancestors of the extant roe deer probably reached North-central Europe only after the retreat of the last Ice Cap (cf. Hufthammer & AarisSrensen, 1998). We may expect that this cervid recently moved from South to North. In the Late Pleniglacial, the Iberian group (Spain, Portugal) and the South-eastern group (Turkey, Greece, Romania, Poland, Lithuania) of roe deer could have differentiated in the refuge areas of South and eastern Europe, respectively. The geographical distance between these refugia must have prevented gene ow of roe deer surviving in these areas, whereas a genetic exchange probably occurred between populations of the Alpine regions of Italy and East Europe through the low passes of the eastern Alps, shortly after the last glacial event. The central-northern group of populations (France, Denmark, Sweden), as well as those from Austria and the eastern Italian Alps, show admixed characters, as would be expected in a group formed recently through migration from the Pleniglacial refuge areas, probably to different extents (i.e. the eastern refuge areas may have contributed more than the southern ones) (cf. Wiehler & Tiedemann, 1998). Roe deer arrived very late in the Baltics (63000 years ago; Hufthammer & Aaris-Srensen, 1998): some genetic differentiation may be expected to have occurred since the dispersal movements started no less than 106000 years ago (probably much earlier, at the end of the Pleniglacial).

In summary, our results suggest an evolutionary history that is partly different from the traditional models of Quaternary radiations of populations. In the Iberian peninsula, one of the main refugia in the Mediterranean region (Bennet, Tzedakis & Willis, 1991), the ice sheets of glacial periods probably covered the northern and central mountainous chains, leaving the low altitudes and the southern regions free from ice. Thus, the southern regions functioned as an effective refuge, but the roe deer population, rather than spreading northwards later, remained as a geographical isolate, failing to act as a source for postglacial recolonizations. The Pyrenees partly limited the movements of populations, preventing their massive advance to the north. A comparable scenario holds for the Italian refuge: colonization of roe deer from the central-southern regions northwards was prevented by the Alpine barrier and, partly, by the Northern Apennines too. Furthermore, in both peninsulae, endemisms of small and large mammals are rather common (Nascetti et al., 1985; Bilton et al., 1998; Frati et al., 1998; Branco, Ferrand & Monnerot, 2000). Consequently, recolonization routes to northern Europe were probably due to range expansions from one or more refugia in central and eastern Europe, rather than proceeding from the Mediterranean areas. This could indicate that the easternmost roe deer hardly got involved in postglacial colonizations, and contributed little to current intraspecic variability in northern Europe. In this scenario, only the Balkans, with mountain chains extending in a northsouth direction, appear as a suitable corridor for northward postglacial expansions. Recolonization waves may have proceeded mainly from south to the North-East. Accordingly, our samples of roe deer collected in Greece joined the eastern group with both mitochondrial and microsatellite markers. Divergence times were derived using the values of nucleotide sequence differentiation among populations and 2% substitution per Myr as the mean mutation rate for the entire mtDNA molecule (Wilson et al., 1985). Given an average sequence divergence of 0.24 for the mitochondrial molecule, we obtained that the differentiation between populations of C. capreolus may be traced back to some 120 000 years ago (with a maximum of 265 000 years). These estimates are consistent with previous results based on sequencing of the mitochondrial control region, which suggest that population differentiation predates the last part of pleistocene glaciation, dating back possibly to the Middle Pleistocene, approximately from 135 000 (Vernesi et al., 2002) to 150 000300 000 (Randi, Pierpaoli & Danilkin, 1998) years ago. A genetic legacy of the Quaternary climatic changes has been proposed to explain current intraspecic variation (Hewitt, 2000). A central tenet in phylogeog-

2006 The Linnean Society of London, Biological Journal of the Linnean Society, 2006, 88, 85100

98

R. LORENZINI and S. LOVARI

raphy dictates that inferences of pleistocene colonizations can be drawn from the distribution of contemporary genetic lineages (Avise et al., 1987). However, the general validity of this assumption has recently been questioned (Taberlet et al., 1998; Consuegra et al., 2002). Postglacial population differences would result from lineage sorting (i.e. sorting out of pre-existing haplotypes) (Avise, 2000), rather than from new relevant mutations. In other words, splitting of the populations may be too recent to be reected by differences of the mitochondrial sequences themselves. Consequently, the extant populations may not necessarily be representative of their ancestors. However, the present structure is unlikely to develop under nonequilibrium processes. Thus, the population structure of the European roe deer derives from ancient timescale events rather than from drift alone in the contemporary populations. Furthermore, the proposed pattern of dispersal appears to be supported by fossil evidence and data from other species. If mammalian taxa with different ecological requirements respond independently to climate oscillations, the general trends of colonizations during the Pleistocene can mainly be inferred through further genetic studies on a number of species with a wide European range. One further aspect that should not be dismissed comprises the recent manipulation of wildlife, especially game species, carried out by man through local extinctions, reintroductions, introductions, and translocations. Over the last few hundred years, these operations have greatly increased. No sensible phylogeographical pattern may be determined unless a severe selection is exerted by limiting sampling areas to allegedly natural populations, in a representative part of the present distribution range of a taxon. If this aspect is neglected, one may fail to detect the local existence of a key-genotype or even a taxon (e.g. Vernesi et al., 2002; for C. c. italicus), thus shadowing the reliability of phylogeographical reconstructions.

REFERENCES
Andersen RP, Duncan P, Linnell JDC (eds). 1998. The European roe deer: the biology of success. Oslo: Scandinavian University Press. Avise JC. 2000. Phylogeography: the history and formation of species. London: Harvard University Press. Avise JC, Arnold J, Ball RM, Bermingham E, Lamb T, Neigel JE, Reeb CA, Saunders NC. 1987. Intraspecic phylogeography: the mitochondrial bridge between population genetics and systematics. Annual Review of Ecology, Evolution and Systematics 18: 489522. Balloux F, Lugon-Moulin N. 2002. The estimation of population differentiation with microsatellite markers. Molecular Ecology 11: 155165. Belkhir K, Borsa P, Goudet J, Chikhi L, Bonhomme F. 1999. GENETIX, logiciel sous WindowsTM pour la gntique des populations. Laboratoire Gnome, Populations, Interactions. Montpellier: CNRS UMR 5000, Universit de Montpellier II. Bennet K, Tzedakis PC, Willis K. 1991. Quaternary refugia on north European trees. Journal of Biogeography 18: 103 115. Benzcri JP. 1973. LAnalyse des Donnes: Tome 2, lAnalyse Des Correspondances. Paris: Dunod Press. Bilton DT, Mirol PM, Mascheretti S, Fredga K, Zima J, Searle JB. 1998. Mediterranean Europe as an area of endemism for small mammals rather than a source of northwards postglacial colonization. Proceedings of the Royal Society of London, Series B 265: 12191226. Bowcock AM, Ruiz-Linares A, Tomfohrde J, Minch E, Kidd JR, Cavalli-Sforza LL. 1994. High resolution human evolutionary trees with polymorphic microsatellites. Nature 368: 455457. Branco M, Ferrand N, Monnerot M. 2000. Phylogeography of the European rabbit (Oryctolagus cuniculus) in the Iberian Peninsula inferred by RFLP analysis of the cytocrome b gene. Heredity 85: 307317. Cavalli-Sforza LL, Edwards AWF. 1967. Phylogenetic analysis: models and estimation procedures. American Journal of Human Genetics 19: 233257. Consuegra S, Garca De Leniz C, Serdio A, Gonzlez Morales M, Straus LG, Knox D, Verspoor E. 2002. Mitochondrial DNA variation in Pleistocene and modern Atlantic salmon from the Iberian glacial refugium. Molecular Ecology 11: 20372048. Cronin MA, Hills S, Born EW, Patton JC. 1994. Mitochondrial DNA variation in Atlantic and Pacic walruses. Canadian Journal of Zoology 72: 10351043. Duncan P, Tixier H, Hofmann RR, Lechner-Doll M. 1998. Feeding strategies and the physiology of digestion in roe deer. In: Andersen R, Duncan P, Linnell JDC, eds. The European roe deer: the biology of success. Oslo: Scandinavian University Press, 91116. Excofer L, Smouse P, Quattro J. 1992. Analysis of molecular variance inferred from metric distances among DNA haplotypes: application to human mitochondrial DNA restriction data. Genetics 131: 479491. Felsenstein J. 1993. PHYLIP. Phylogeny inference package,

ACKNOWLEDGEMENTS
We are greatly indebted to F. Masini for his valuable comments on our rst draft. We also thank all our colleagues who kindly provided the samples used in this study: R. Andersen, L. Balciauskas, L. Borger, F. Braza, M. Caldarella, C. M. Cal, N. Franconi, M. Gioiosa, R. Gula, Y. Iliopoulos, O. Ionescu, A. Kence, R. Mazzoni Della Stella, C. R. Olesen, A. Sangiuliano, C. San Jos, A. Sforzi, F. Suchentrunk, and J. Vingada. Without their cooperation, this paper could not have been written. F. Frati, M. Hewison, and an anonymous reviewer greatly improved an earlier draft of this manuscript. G. Plantamura is thanked for graphical support of Figure 1.

2006 The Linnean Society of London, Biological Journal of the Linnean Society, 2006, 88, 85100

GENETIC DIVERSITY AND PHYLOGEOGRAPHY OF ROE DEER


Version 3.6; Alpha2. Seattle, WA: Department of Genetics. University of Washington. Festa E. 1925. Il capriolo dellItalia Centrale. Bollettino del Museo di Zoologia Ed Anatomia Comparata dellUniversit di Torino 7: 12. Feulner PGD, Bielfeldt W, Zachos FE, Bradvarovic J, Eckert I, Hartl GB. 2004. Mitochondrial DNA and microsatellite analyses of the genetic status of the presumed subspecies Cervus elaphus montanus (Carpathian red deer). Heredity 93: 299306. Fickel J, Reinsch A. 2000. Microsatellite markers for the European roe deer (Capreolus capreolus). Molecular Ecology 9: 994995. Frati F, Hartl GB, Lovari S, Delibes M, Markov G. 1998. Quaternary radiation and genetic structure of the red fox Vulpes vulpes in the Mediterranean basin, as revealed by allozyme and mitochondrial DNA. Journal of Zoology of London 245: 4352. Gliozzi E, Abbazzi L, Argenti P, Azzaroli A, Caloi L, Capasso Barbato L, di Stefano G, Esu D, Ficcarelli G, Girotti O, Kotsakis T, Masini F, Mazza P, Mezzabotta C, Palombo MR, Petronio C, Rook L, Sala B, Sardella R, Zanalda E, Torre D. 1997. Biochronology of selected mammals, molluscs and ostracods from the Middle Pliocene to the Late Pleistocene in Italy. The state of the art. Rivista Italiana di Paleontologia e Stratigraa 103: 369388. Goldstein DB, Linares AR, Cavalli-Sforza LL, Feldman MW. 1995. An evaluation of genetic distances for use with microsatellite loci. Genetics 139: 463471. Goudet J. 1995. FSTAT version 1.2. A computer program to calculate F-statistics. Journal of Heredity 86: 485486. Gurin C, Patou-Mathis M. 1996. Les Grands Mammifres Plio-Plistocnes dEurope. Paris: Masson. Guo S, Thompson. 1992. Performing the exact test of Hardy Weinberg proportion for multiple alleles. Biometrics 48: 361 372. Hartl GB, Reimoser F, Willing R, Kller J. 1991. Genetic variability and differentiation in roe deer (Capreolus capreolus L.) of Central Europe. Genetics, Selection, Evolution 23: 281299. Hartl GB, Markov G, Rubin A, Findo S, Willing R. 1993. Allozyme diversity within and among populations of three ungulate species (Cervus elaphus, Capreolus capreolus, Sus scrofa) of Southeastern and Central Europe. Zeitschrift fr Sugetierkunde 58: 352361. Hewison AJM. 1995. Isozyme variation in roe deer in relation to their population history in Britain. Journal of Zoology of London 235: 279288. Hewitt G. 1996. Some genetic consequences of ice ages, and their role in divergence and speciation. Biological Journal of the Linnean Society 58: 247276. Hewitt G. 2000. The genetic legacy of the Quaternary ice ages. Nature 405: 907913. Hufthammer AK, Aaris-Srensen K. 1998. Late- and postglacial European roe deer. In: Andersen R, Duncan P, Linnell JDC, eds. The European roe deer: the biology of success. Oslo: Scandinavian University Press, 4769.

99

Jger F, Hecht W, Herzog A. 1992. Untersuchungen an mitochondrialer DNS (mtDNS) von hessischem Rehwild (C. capreolus). Zeitschrift fr Jagdwissenschaft 38: 2633. Kimura M, Crow JF. 1964. The number of alleles that can be maintained in a nite population. Genetics 143: 549555. Kimura M, Otha T. 1978. Stepwise mutation model and the distribution of allelic frequencies in nite population. Proceedings of the National Academy of Sciences of the USA 75: 28682872. Kohn M, Knauer F, Stoffella A, Schrder W, Pbo S. 1995. Conservation genetics of the European brown bear a study using excremental PCR of nuclear and mitochondrial sequences. Molecular Ecology 4: 95103. Kurtn B. 1968. Pleistocene Mammals of Europe. London: Weidenfeld & Nicolson. Liedloff A. 1999. MANTEL Nonparametric Test Calculator, Version 2.0.Brisbane: School of Natural Resource Sciences, Queensland University of Technology. Lister AM, Grubb P, Sumner SRM. 1998. Taxonomy, morphology and evolution of European roe deer. In: Andersen R, Duncan P, Linnell JDC, eds. The European Roe Deer: The Biology of Success. Oslo: Scandinavian University Press, 23 46. Lorenzini R, Patalano M, Apollonio M, Mazzarone V. 1993. Genetic variability of roe deer Capreolus capreolus in Italy: electrophoretic survey on populations of different origin. In: Hartl GB, Markowski J, eds. Ecological genetics in mammals. Acta Theriologica 38 (Suppl. 2): 141151. Lorenzini R, Lovari S, Masseti M. 2002. The rediscovery of the Italian roe deer: genetic differentiation and management implications. Italian Journal of Zoology 69: 367379. Lorenzini R, San Jos C, Braza C, Aragn S. 2003. Genetic differentiation and phylogeography of roe deer in Spain, as suggested by mitochondrial DNA and microsatellite analysis. Italian Journal of Zoology 70: 8999. Ludt CJ, Schroeder W, Rottmann O, Kuehn R. 2004. Mitochondrial DANN phylogeography of red deer (Cervus elaphus). Molecular Phylogenetics and Evolution 31: 1064 1083. Meunier K. 1983. Das Spanische Reh. In: Hofmann RR, ed. Wildbiologische Informationen fr Den Jger. Jagd+Hege Ausbildungsbuch VI. Berlin, 149153. Miller JC. 1991. RESTSITE: a phylogenetic program that sorts raw restriction data. Journal of Heredity 82: 262263. Minch E, Ruiz-Linares A, Goldstein D, Feldman M, Cavalli-Sforza LL. 1995. MICROSAT: a computer program for calculating various statistics on microsatellite allele data, Version 1.5b. Stanford, CA: Stanford University Press. Nascetti G, Lovari S, Lanfranchi P, Berducou C, Mattiucci S, Rossi L, Bullini L. 1985. Revision of Rupicapra genus. III. Electrophoretic studies demonstrating species distinction of chamois populations of the Alps from those of the Apennines and Pyrenees. In: Lovari S, ed. Biology and management of mountain ungulates. London: Croom-Helm, 5762. Nei M. 1972. Genetic distance between populations. American Naturalist 106: 283292.

2006 The Linnean Society of London, Biological Journal of the Linnean Society, 2006, 88, 85100

100

R. LORENZINI and S. LOVARI


2.000: a software for population genetics data analysis. Geneve: Genetics and Biometry Laboratory, University of Geneve. Slatkin M. 1995. A measure of population subdivision based on microsatellite allele frequencies. Genetics 139: 457462. Taberlet P, Bouvet J. 1994. Mitochondrial DNA polymorphism, phylogeography, and conservation genetics of the brown bear (Ursus arctos) in Europe. Proceedings of the Royal Society of London, Series B 255: 195200. Taberlet P, Fumagalli L, Wust-Saucy A, Cossons J. 1998. Comparative phylogeography and postglacial colonization routes in Europe. Molecular Ecology 7: 453464. Vernesi C, Pecchioli E, Caramelli D, Tiedemann R, Randi E, Bertorelle G. 2002. The genetic structure of natural and reintroduced roe deer (Capreolus capreolus) populations in the Alps and central Italy, with reference to the mitochondrial DNA phylogeography of Europe. Molecular Ecology 11: 12851297. von Lehmann E, Sgesser H. 1986. Capreolus capreolus Linnaeus, 1758. Reh. In: Niethammer j Krapp F, ed. Handbuch der Sugetiere Europas, Vol. 2. Wiesbaden: AulaVerlag, 233268. Wang M, Schreiber A. 2001. The impact of habitat fragmentation and social structure on the population genetics of roe deer (Capreolus capreolus L.) in Central Europe. Heredity 86: 703715. Wiehler J, Tiedemann R. 1998. Phylogeography of the European roe deer Capreolus capreolus as revealed by sequence analysis of the mitochondrial control region. In: Hartl GB, Markowski J, eds. Ecological genetics in mammals. Acta Theriologica (Suppl. 5): 187197. Wilson AC, Cann RE, Carr SM, George M, Gyllensten UB, Helm-Bychowski KM, Higuch RG, Palumbi SR, Prager EM, Sage RD, Stoneking M. 1985. Mitochondrial DNA and two perspectives on evolutionary genetics. Biological Journal of the Linnean Society 26: 375400. Wilson GA, Strobeck L, Wu L, Cofn J. 1997. Characterization of microsatellite loci in caribou Rangifer tarandus and their use in other artiodactyls. Molecular Ecology 6: 697699. Zagwijn WH. 1992. Migration of vegetation during the Quaternary in Europe. Frankfurt Courier Forschung Senckenberg 153: 920.

Nei M. 1987. Molecular evolutionary genetics. New York, NY: Columbia University Press. Nei M, Miller JC. 1990. A simple method for estimating average number of nucleotide substitutions within and between populations from restriction data. Genetics 125: 873879. Pemberton JM, Slate J, Bancroft DR, Barret JA. 1995. Nonamplifying alleles at microsatellite loci: a caution for parentage and population studies. Molecular Ecology 4: 249 252. Petit-Maire N, Vrielynck B. 2004. Maps of the Mediterranean environments during the last two climatic extremes. Map 1: the Last Glacial Maximum (19.500 16.100 14C yr BP = 24.000 18.000 cal-yr bp). Paris: Commission for the Geological Map of the World. Pritchard JK, Stephens M, Donnelly P. 2000. Inference of population structure using multilocus genotype data. Genetics 155: 945959. Randi E, Alves PC, Carranza J, Milotevic-Zlatanovic S, Sfougaris A, Mucci N. 2004. Phylogeography of roe deer (Capreolus. capreolus) populations: the effects of historical genetic subdivisions and recent nonequilibrium dynamics. Molecular Ecology 13: 30713083. Randi E, Pierpaoli M, Danilkin A. 1998. Mitochondrial DNA polymorphism in populations of Siberian and European roe deer (Capreolus pygargus and C. capreolus). Heredity 80: 429437. Raymond M, Rousset F. 1995. GENEPOP (version 1.2): population genetics software for exact tests and ecumenism. Journal of Heredity 86: 248249. Reynolds J, Weir BS, Cockerham CC. 1983. Estimation of the coancestry coefcient: basis for a short-term genetic distance. Genetics 105: 767779. Rice WR. 1989. Analyzing tables of statistical tests. Evolution 43: 223225. Red KH, Midthjell L. 1998. Microsatellites in reindeer Rangifer tarandus and their use in other cervids. Molecular Ecology 7: 17731776. Saitou N, Nei M. 1987. The neighbor-joining method: a new method for reconstructing phylogenetic trees. Molecular Biology and Evolution 4: 406425. Sambrook J, Russel DV. 2001. Molecular cloning: a laboratory manual. New York, NY: Cold Spring Harbour Press. Schneider S, Roessli D, Excofer L. 2000. Arlequin ver

2006 The Linnean Society of London, Biological Journal of the Linnean Society, 2006, 88, 85100

Vous aimerez peut-être aussi