Geographic pattern of genetic variation in the European ... - ISYEB

MEC_1618.fm Page 2337 Tuesday, October 22, 2002 6:09 PM

Molecular Ecology (2002) 11, 2337–2347

Geographic pattern of genetic variation in the European globeflower Trollius europaeus L. (Ranunculaceae) inferred from amplified fragment length polymorphism markers Blackwell Science, Ltd

L A U R E N C E D E S P R E S , S A N D R I N E L O R I O T and M Y R I A M G A U D E U L Laboratoire de Biologie des Populations d’Altitude, CNRS-UMR 5553, Université J. Fourier, BP 53–38041 Grenoble Cedex 09, France

Abstract The distribution of genetic variation and the phylogenetic relationships between 18 populations of the arctic-alpine plant Trollius europaeus were analysed in three main regions (Alps, Pyrenees and Fennoscandia) by using dominant AFLP markers. Analysis of molecular variance revealed that most of the genetic variability was found within populations (64%), although variation among regions (17%) and among populations within regions (19%) was highly significant (P < 0.001). Accordingly, the global fixation index FST averaged over loci was high (0.39). The among-population differentiation indicates restricted gene flow, congruent with limited dispersal of specific globeflower’s pollinating flies (Chiastocheta spp.). Within-population diversity levels were significantly higher in the Alps (mean Nei’s expected heterozygosity HE = 0.229) than in the Pyrenees (HE = 0.197) or in Fennoscandia (HE = 0.158). This finding is congruent with the speciesrichness of the associated flies, which is maximum in the Alps. We discuss the processes involved in shaping observed patterns of genetic diversity within and among T. europaeus populations. Genetic drift is the major factor acting on the small Pyrenean populations at the southern edge of T. europaeus distribution, while large Fennoscandian populations result probably from a founder effect followed by demographic expansion. The Alpine populations represent moderately fragmented relics of large southern ancestral populations. The patterns of genetic variability observed in the host plant support the hypothesis of sympatric speciation in associated flies, rather than recurrent allopatric speciations. Keywords: amplified fragment length polymorphism markers (AFLP); genetic variation; habitat fragmentation; population differentiation; postglacial recolonization; sympatric speciation, Trollius europaeus Received 5 March 2002; revision received 24 June 2002; accepted 19 July 2002

Introduction The genetic structure of a species is both influenced by its past history and by current gene flow. The Quaternary cold periods in Europe have heavily influenced the distribution of plant species with many range contractions–expansions in direct relation with climatic variations (Hewitt 1996; Comes & Kadereit 1998). The ice sheets of the Northern hemisphere began to grow about 2.5 My ago and the major climatic oscillations occurred during the last 700 ky (Webb & Bartlein 1992). During glaciations, many temperate Correspondence: Laurence Després. Fax: + 33 4 76 51 42 79; Email: [email protected] © 2002 Blackwell Science Ltd

European plant species were shown to be restricted to three main southern ice-free refugia: one in Portugal– Spain, one in Italy and one in the Balkans (Taberlet et al. 1998; Hewitt 2001). Most present plant species distributions in Europe result from a northward recolonization from those southern refugia after the last glaciation, about 12 000 years ago (the ‘tabula rasa hypothesis’, Nordal 1987; Birks 1996). Under this hypothesis, the southern part of Europe should present the highest genetic diversity by contrast with the northern part recently recolonized (founder effect). This picture is likely to be quite different for cold-tolerant plants which may have survived in icefree northern, alpine or eastern refugia during Pleistocene glaciations (Lagercrantz & Ryman 1990; Abbott et al. 1995;


2338 L . D E S P R E S , S . L O R I O T and M . G A U D E U L Tremblay & Schoen 1999; Vendramin et al. 2000), resulting in the maintenance of high genetic diversity in nordic and/ or alpine populations. Much fewer phylogeographical analyses had been undertaken on these taxa, so that their glacial refugia are still poorly known, but are likely to be found at higher latitude/altitude than those described for temperate taxa. Moreover, not only past history, but also current balance between gene flow and genetic drift shapes the observed genetic patterns of species: genetic drift in small populations at the edge of the species range will lead to reduced within population genetic diversity and high between population differentiation, while populations admixture can lead to increased genetic diversity (Walter & Epperson 2001). Discriminating between patterns of genetic variation caused by gene flow/drift balance from those caused by common ancestry requires analysing genetic variability at both spatial and temporal scales: within vs. among populations and regions throughout Europe, and past history (phylogeography) vs. recent history (drift and gene flow). The European globeflower Trollius europaeus L. (Ranunculaceae) is a perennial arctic-alpine species growing in moist and cold meadows. The species occurs at low altitude in northern Europe (i.e Fennoscandia and Russia) where populations are usually large and more or less continuously distributed. By contrast, in southern Europe, T. europaeus is restricted to mountains above 700 m and forms much more fragmented populations, especially at the southern edges of its distribution, where populations are usually small and patchily distributed. During the last glaciation, glaciers covered most of the present range of the species. The pollination ecology of T. europaeus has been studied extensively as it is one of the rare reported cases of a plant being pollinated exclusively by a seed parasite (Pellmyr 1989; Jaeger & Després 1998; Jaeger et al. 2000, 2001). Pollinators are small anthomyiid flies (genus Chiastocheta) whose larvae develop by eating a fraction of the plant’s seeds. Several Chiastocheta species may coexist in T. europaeus populations with up to six species observed in many Alpine populations (Després & Jaeger 1999). These species differ in the date of oviposition. By contrast with Alpine populations, a maximum of five species were described throughout the whole northern range (Pellmyr 1992; Després et al. 2002), and only three have been observed so far in the five Pyrenean populations sampled in the present study (L.D., unpublished data). Moreover, the highest level of mtDNA sequence polymorphism within Chiastocheta species was observed in the Alps (Després et al. 2002). This high level of biodiversity (both in terms of species number and within species variability) found in the globeflower flies in the Alps may reflect the history of T. europaeus populations in Europe. Indeed, because of the intimacy of the relationship linking the plant and the insect, their evolutionary history is likely to be

the same; our expectation is therefore to find the highest plant genetic diversity in the Alps. Our main objective is to determine the evolutionary processes responsible for the observed pattern of genetic variation in T. europaeus, and to infer those driving Chiastocheta spp. radiation. The timescales involved in Chiastocheta radiation and in T. europaeus phylogeography are different: Chiastocheta speciation events occurred several times during the last 2 million years (Després et al. 2002), whereas the present study of T. europaeus genetic structure considers only the plant history since the end of the last glacial period (12 000 years ago). However, the last 2 million years have been characterized by several range contraction–expansions comparable to the last one, in concordance with cyclical climatic changes of similar amplitude. Recurrent allopatric speciations may have occurred in the Chiastocheta genus during fragmentation periods of the host-plant range, followed by remixture of allopatrically differentiated species. Alternatively, sympatric speciation may have occurred in ancestral large host-plant southern populations throughout disruptive selection on the date of oviposition (Després & Jaeger 1999; Ferdy et al. 2002), followed by local species extinction during fragmentation of the host-plant populations and/ or species loss during range expansion. The finding of well-differentiated host-plant lineages originating from different geographical regions and coexisting in the Alps would support the first hypothesis, whereas the finding that Pyrenean and Alpine populations represent fragmented relics of large southern, low altitudinal, interconnected populations would support the second hypothesis. In this study, we analyse the genetic diversity of the globeflower from the north to the south of its geographical distribution. Nuclear and chloroplast sequencing failed to reveal nucleotide variability among T. europaeus populations (Després et al. submitted). For this study, we therefore chose the amplified fragment length polymorphism (AFLP) technique, validated as efficient for biogeographical questions (Weising et al. 1995; Mueller & Wolfenbarger 1999). The AFLP technique, described first by Vos et al. (1995), offers the advantage of generating a large number of markers, spanning the whole genome without requiring any prior knowledge. The major flaw of this marker is that it is a dominant marker, therefore precluding any inference of the within-population genetic structure. We first studied the within- and among-population genetic diversity in three main geographical regions: Fennoscandia (i.e northern part of the T. europaeus distribution area), Alps and the Pyrenees (i.e southern part of the species distribution area). We then inferred the phylogenetic relationships between these populations in relation to their geographical location (phylogeography), and finally proposed a scenario describing the history of T. europaeus populations during the last postglaciation period. © 2002 Blackwell Science Ltd, Molecular Ecology, 11, 2337–2347


G E N E T I C D I V E R S I T Y I N T H E E U R O P E A N G L O B E F L O W E R 2339 Fig. 1 Geographic location of the 18 studied populations of Trollius europaeus.

Materials and methods Plant material Trollius europaeus L. is an arctic-alpine, early-flowering perennial. This diploid plant (2n = 16) is an obligate outcrosser depending on Chiastocheta flies for its pollination throughout its geographical range (Pellmyr 1989; Jaeger & Després 1998; Hemborg & Després 1999). The plant is 0.2 –1 m high and usually produces a single yellow globose flower composed of several multi-ovulate carpels. The closed shape of the flower does not allow visitors other than Chiastocheta flies. Males and females feed, rest and mate inside the flowers, pollinating them passively. Females deposit eggs on carpels and larvae develop on seeds. Mature follicules open at the end of the summer, liberating seeds and pupae that fall down in the soil to germinate/emerge the following spring. Only the underground taproot persists over winter, and there is no evidence of asexual multiplication. Adult Chiastocheta flies do not survive during winter. Eighteen populations of T. europaeus were chosen in order to cover a broad geographical range (Fig. 1). Five populations in each of three main regions were sampled: Fennoscandia, Alps and the Pyrenees. In addition, three more isolated populations were sampled in Germany, Poland and Romanian Carpathians. The population sizes and elevations were distributed widely ranging, respectively, from 30 to thousands of individuals, and from 0 to 2500 m (Table 1). Thirty individuals per population were sampled randomly and for each individual, leaf material © 2002 Blackwell Science Ltd, Molecular Ecology, 11, 2337–2347

was collected and dried immediately in silica gel. Genetic analyses were performed on 10 individuals per population, resulting in a total of 180 individuals genotyped.

AFLP protocol DNA extraction was performed with the Dneasy Plant Mini Kit (Qiagen, Hilden, Germany) according to the manufacturer’s protocol, using about 20 mg of dried leaf material. DNA concentration was determined by fluorometry with the Picogreen® DNA quantification Kit (Molecular Probes, Leiden, Netherlands) and ranged from 15 to 25 ng/µL. The AFLP method was performed as described by Gaudeul et al. (2000): EcoRI adapters were 5′-CTCGTAGACTGCGTACC-3′ and 5′-AATTGGTACGCAGTCTAC-3′, and MseI adapters 5′-GACGATGAGTCCCTGAG-3′ and 5′TACTCAGGACTCAT-3′. For the preselective amplification by PCR, parameters were as follows: 2 min at 72 °C, 30 cycles of 30 s denaturing at 94 °C, 30 s annealing at 56 °C, and 2 min extension at 72 °C, ending with 10 min at 72 °C for complete extension, using EcoRI primer E.A (5′GACTGCGTACCAATTCA-3′) and MseI primer M.C (5′-GATGAGTCCTGAGTAAC-3′). For the selective amplification, PCR parameters were: 10 min at 95 °C followed by 36 cycles of 30 s denaturing at 94 °C, 30 s annealing, and 1 min extension at 72 °C, ending with 10 min at 72 °C for complete extension. Annealing was initiated at a temperature of 65 °C, which was then reduced by 0.7 °C for the next 12 cycles and maintained at 56 °C for the subsequent 23 cycles. Three selective PCR primer pairs were selected over 16 tested for the quality of the produced


2340 L . D E S P R E S , S . L O R I O T and M . G A U D E U L Table 1 Name, location, altitude, size and collector name of the 18 studied populations

Population

Region–country

Longitude/latitude

Elevation (m)

Approximate number of plants

Collector

1 Areche 2 Menee 3 Fournel 4 Cherlieu 5 Galibier 6 Fulda 7 Pradella 8 Eynes 9 Gazies 10 Gabardères 11 Puymorens 12 Kvaloeya 13 Ringassoeya 14 Rhindunjira 15 Slattatjakka 16 Oulanka 17 Biebrza 18 Piatra Fontanele

Alps — France Alps — France Alps — France Alps — France Alps — France Germany Pyrenees — France Pyrenees — France Pyrenees — France Pyrenees — France Pyrenees — France Fennoscandia — Norway Fennoscandia — Norway Fennoscandia — Sweden Fennoscandia — Sweden Fennoscandia — Finland Poland Romania

6°34′ E/45°40′ N 5°31′ E/44°42′ N 6°53′ E/44°79′ N 5°46′ E/45°18′ N 6°24′ E/45°04′ N 9°39′ E/51°25′ N 2°01′ E/42°32′ N 2°10′ E/42°30′ N 2°79′ W/42°91′ N 2°79′ W/42°85′ N 1°49′ E/42°32′ N 18°43′ E/69°39′ N 19°00′ E/69°50′ N 18°49′ E/68°21′ N 18°46′ E/68°20′ N 28°59′ E/66°26′ N 22°28′ E/53°25′ N 24°50′ E/47°12′ N

1700 1400 1500 950 2300 800 1850 2000 1500 1680 1850 180 0 400 670 200 150 1000

> 1000 > 1000 500 > 1000 > 1000 < 500 50 500 100 50 500 500 500 > 1000 > 1000 > 1000 < 500 500

L. Després L. Després M. Gaudeul L. Després L. Després J. Johannesen L. Després L. Després F. d’Amico F. d’Amico L. Després G. Yoccoz G. Yoccoz A. Hemborg A. Hemborg P. Siikamäki A. Wrodlewska I. Chintauan

bands (i.e. even distribution of bands with relatively homogeneous intensity): E.ATC/M.CAG, E.ATC/M.CAT and E.AGA/M.CTC. Products were separated by electrophoresis for 6 h on a 5% polyacrylamide gel (automated sequencer ABI 377TM Perkin-Elmer). AFLP patterns were then visualized with Genescan Analysis® 3.1 (Perkin-Elmer): a fluorescent peak corresponds to the presence of an amplified restriction fragment. Polymorphic peaks were checked individually and a presence/absence (i.e. 1/0) matrix was manually constructed. Reproducibility of each primer pair was checked by carrying out two times the whole AFLP protocol for three individuals chosen randomly.

Data analysis Independence of markers was assessed first by calculating their pairwise linkage index DA,B = 1/n Σi|V(A,i) − V(B,i)| where V(A,i) is the allele value of individual i for the marker A and n, the total number of individuals analysed. If marker pairs had values of D = 0.01 or D = 0.99, one was discarded from the data set to avoid redundancy. D-values were calculated using the program DDM (disequilibrium between dominant markers) version 0.1 (P. Berthier, pers. comm.). Statistical analyses of AFLP patterns were based on the following assumptions: (i) AFLP markers behave as diploid, dominant markers with alleles either present (amplified) or absent (nonamplified); (ii) comigrating fragments represent homologous loci; and (iii) populations are at the Hardy–Weinberg equilibrium (HWE). This assumption appears justified because the globeflower

is an obligate outcrosser throughout its geographical distribution. Only polymorphic markers were taken into account in all calculations. Mean genetic diversity within populations was estimated using popgene Version 1.31 (Yeh et al. 1997) in three ways: (i) the percentage of polymorphic loci out of all polymorphic loci (P%) (ii) Nei’s (1978) unbiased expected heterozygosity (HE) and (iii) Shannon’s index of phenotypic diversity (IS & Lewontin 1972). Estimates of HE and IS were obtained by averaging across loci. Genetic differentiation among populations and regions was calculated as the unbiased FST estimator of Weir & Cockerham (1984), and its 95% confidence interval was obtained by bootstrapping 1000 replicates over loci using tfpga Version 1.3 (Miller 1997). To evaluate among-population or among-region differentiation, total genetic diversity was also partitioned among regions, among populations within regions, and within populations by carrying out a hierarchical analysis of molecular variance (amova) on euclidian pairwise distances (ΦST) using arlequin Version 2.000 (Schneider et al. 2000). Finally, Fisher’s exact tests were performed on marker frequencies at each locus between all pairs of populations/regions (Raymond & Rousset 1995): these pairwise tests determine if significant differences in marker frequencies exist between groups of individuals. In order to test for a correlation between genetic (Nei 1978) and geographical distances (in km) among populations, Mantel tests (Mantel 1967) were performed using tfpga (1000 permutations). To determine the phylogenetic relationships among the populations, a neighbour-joining dendrogram based on Nei’s distance was constructed, and bootstrap © 2002 Blackwell Science Ltd, Molecular Ecology, 11, 2337–2347


G E N E T I C D I V E R S I T Y I N T H E E U R O P E A N G L O B E F L O W E R 2341 values were obtained by resampling with replacement over loci (2000 replicates) using the program njbs (J.-M. Cornuet, pers. comm.).

Results AFLP polymorphism Using three primer pairs, 128 scorable polymorphic fragments were generated. All the 180 individuals genotyped presented different profiles. When all the 180 genotypes were considered, a complete or nearly complete linkage disequilibrium (D ≤ 0.01) was found for 57 pairs of markers, due probably to physical linkage among these markers. To avoid redundant information, 11 markers were discarded before subsequent analyses, leading to a total of 117 independent locus markers: 46 for the primer pair E.ATC/M.CAG, 47 for E.ATC/M.CAT and 24 for E.AGA/ M.CTC. The mean linkage disequilibrium over all pairs of loci was D = 0.44.

Genetic distance analysis The neighbour-joining dendrogram based on Nei’s (1978) unbiased genetic distances between all pairwise combinations of populations (Table 2) revealed three population clusters: one group includes all the Pyrenean populations, another groups the five Alpine populations with the German population, and the third groups Fennoscandian populations with the Polish and Romanian populations in a basal position (Fig. 2). This latter group is

supported by generally high bootstrap values (range 72 – 99%), except for the Romanian population, different from all other populations. The two other groups present less robust nodes (range 24– 89%). We therefore defined three groups of populations corresponding to three geographical regions: the Pyrenees, the Alps (including the German population Fulda) and Fennoscandia. These three groups are comparable both in sample size (respectively five, six and five populations sampled) and in geographical area (populations within a region are less than 700 km apart). Because the sampling was too restricted in eastern Europe (only two populations sampled), the Polish and the Romanian populations were excluded from the grouped populations analyses.

Within-population and within-region genetic diversity The percentage of polymorphic loci (P%) and the Nei (HE) and Shannon (I) indices for each population were strongly correlated (Table 3, pairwise Spearman’s rank correlation coefficient P% vs. HE: r s = 0.812; HE vs. I: r s = 0.998; P% vs. I: r s = 0.814; all P < 0.001). The mean percentage of polymorphic loci was significantly higher within the Alpine populations (60.78 ± 2.21%) compared to the Pyrenean (46.15 ± 5.39%) and Fennoscandian populations (43.25 ± 2.22%), whereas these two latter groups did not significantly differ (Mann–Whitney U-test; Alps vs. Pyrenees: P = 0.035; Alps vs. Fennoscandia: P = 0.008; Pyrenees vs. Fennoscandia: P = 0.6). Nei’s and Shannon diversity indices were also higher in the Alps (0.229 ± 0.024 and 0.338 ± 0.033, respectively) but it was significant only when

Fig. 2 Neighbour-joining dendrogram based on Nei’s distances for the 18 studied populations, and the corresponding regions. Bootstrap values over loci (based on 2000 replicates) are indicated for each node.

© 2002 Blackwell Science Ltd, Molecular Ecology, 11, 2337–2347

2

3

4

5

6

7

8

9

10

11

2 3 4 5 6

0.0608ns 0.089ns 0.0964ns 0.0687ns 0.0952ns

0.1166ns 0.1210ns 0.0916ns 0.1401ns

0.1501ns 0.1096ns 0.1830‡

0.0739ns 0.0925ns

0.0829ns

7 8 9 10 11

0.0969ns 0.0809ns 0.3158‡ 0.1162ns 0.1269ns

0.1249ns 0.1141ns 0.3176‡ 0.1541† 0.1406ns

0.1158ns 0.1147ns 0.2601‡ 0.1443* 0.1386ns

0.1059ns 0.1223ns 0.3464‡ 0.1544‡ 0.1648†

12 13 14 15 16

0.1589‡ 0.1688‡ 0.1529‡ 0.1672‡ 0.1622‡

0.1850‡ 0.1979‡ 0.1761‡ 0.1874‡ 0.1672‡

0.2200‡ 0.2119‡ 0.1997‡ 0.1816‡ 0.1941‡

17 18

0.1196ns 0.1768‡

0.1473† 0.1994‡

0.2228‡ 0.1954‡

12

13

14

15

16

0.0751ns 0.1022ns 0.3243‡ 0.1551† 0.1414ns

0.1211ns 0.1131ns 0.3588‡ 0.1692‡ 0.1646†

0.0630ns 0.2940‡ 0.0886ns 0.1157ns

0.2717‡ 0.0958 0.0923ns

0.2549‡ 0.1981‡

0.0624ns

0.2098‡ 0.1896‡ 0.2206‡ 0.2230‡ 0.2233‡

0.1931‡ 0.1713‡ 0.1932‡ 0.1683‡ 0.1696‡

0.1699‡ 0.1594‡ 0.2107‡ 0.2364‡ 0.1881‡

0.2066‡ 0.2095‡ 0.2178‡ 0.1964‡ 0.1837‡

0.1670‡ 0.1645‡ 0.1980‡ 0.1982‡ 0.1627‡

0.3054‡ 0.3006‡ 0.3447‡ 0.3291‡ 0.3628‡

0.2183‡ 0.2109‡ 0.2292‡ 0.2251‡ 0.2049‡

0.1646‡ 0.1385‡ 0.1842‡ 0.1934‡ 0.1939‡

0.0513ns 0.0718ns 0.1265ns 0.1533‡

0.0785ns 0.1262ns 0.1720

0.0715ns 0.1102ns

0.0917ns

0.1679‡ 0.2224‡

0.1890 0.1998‡

0.1570† 0.2073‡

0.1811‡ 0.1662‡

0.1588‡ 0.1606‡

0.3074‡ 0.3047‡

0.1602‡ 0.1900‡

0.1421‡ 0.1710‡

0.1567 0.2169‡

0.1445 0.2181‡

0.1344† 0.2176‡

0.1434* 0.2223‡

Populations numbered as in Fig. 1. Populations are grouped into regions. Pairwise comparisons within regions are in italic type. Significance levels of exact tests of differentiation are shown by: ns: non significant. *0.01 < P < 0.05; †0.001 < P < 0.01; ‡P < 0.001.

17

18

0.1580‡ 0.1855‡

0.1867‡



1

2342 L . D E S P R E S , S . L O R I O T and M . G A U D E U L

Table 2 Pairwise Nei’s unbiased genetic distance among populations of European globeflowers and exact tests of population differentiation


G E N E T I C D I V E R S I T Y I N T H E E U R O P E A N G L O B E F L O W E R 2343 Table 3 Within-population genetic diversity indices of the 18 populations studied, and total (T) or mean (M) values for the three geographical regions defined. Standard deviations are in parentheses

Regions

Populations

Alps

1 Arêche 2 Menée 3 Fournel 4 Cherlieu 5 Galibier 6 Fulda

Pyrenees

7 Pradella 8 Eynes 9 Gazies 10 Gabardères 11 Puymorens

Fennoscandia

12 Kvaloeya 13 Ringassoeya 14 Rhindunjira 15 Slattatjakka 16 Oulanka

Poland Romania

17 Biebrza 18 Piatra Fontanele

Number of polymorphic loci

Percentage of polymorphic loci (P%)

Nei’s diversity index (HE)

Shannon index (I)

75 80 67 66 66 76 T: 107 54 70 33 50 63 T: 100 44 50 57 46 56 T: 84 53 52

64.10% 67.52% 57.26% 54.41% 56.41% 64.96% M: 60.78% (5.40%) 46.15% 59.83% 28.21% 42.73% 53.84% M: 46.15% (12.04%) 37.61% 42.73% 48.72% 39.32% 47.86% M: 43.25% (4.97%) 42.30% 44.44%

0.239 (0.205) 0.270 (0.207) 0.230 (0.218) 0.202 (0.206) 0.213 (0.213) 0.222 (0.201) M: 0.229 (0.024) 0.177 (0.210) 0.240 (0.216) 0.220 (0.213) 0.148 (0.193) 0.200 (0.210) M: 0.197 (0.036) 0.138 (0.198) 0.154 (0.198) 0.189 (0.212) 0.144 (0.201) 0.167 (0.202) M: 0.158 (0.020) 0.161 (0.200) 0.173 (0.209)

0.353 (0.290) 0.395 (0.291) 0.335 (0.308) 0.301 (0.293) 0.313 (0.301) 0.333 (0.283) M: 0.338 (0.033) 0.259 (0.300) 0.349 (0.305) 0.321 (0.304) 0.222 (0.279) 0.295 (0.298) M: 0.289 (0.050) 0.203 (0.283) 0.229 (0.285) 0.277 (0.302) 0.212 (0.287) 0.249 (0.288) M: 0.234 (0.029) 0.241 (0.287) 0.254 (0.299)

compared to Fennoscandian populations (0.158 ± 0.020 and 0.234 ± 0.029, respectively), the Pyrenean populations being intermediate and not significantly different from the two other regions for these two indices (0.197 ± 0.036 and 0.289 ± 0.050, respectively). The Polish and Romanian populations exhibited levels of diversity comparable to that of Pyrenean or Fennoscandian populations. At the regional level, 107 markers were polymorphic in the Alps, 100 in the Pyrenees and only 84 in Fennoscandia, indicating that although the average within population percentage of polymorphic markers was not higher in the Pyrenees as in Fennoscandia, different markers were polymorphic in different Pyrenean populations, resulting in a total number of polymorphic markers in the Pyrenees close to that observed in the Alps. The Pyrenees were also characterized by higher heterogeneity for all withinpopulation diversity indices. For example, the standard deviation of the percentage of polymorphic loci was 12.04% in the Pyrenees, whereas it was only 5.40% and 4.97% in the Alps and in Fennoscandia, respectively.

Among-population and among-region differentiation The global FST value among all populations was 0.39 (95% confidence interval (CI) 0.37– 0.42), indicating a strong among-population differentiation. At a regional scale (Polish and Romanian populations excluded), the Pyre© 2002 Blackwell Science Ltd, Molecular Ecology, 11, 2337–2347

nees appeared to be the more differentiated region (FST = 0.39, 95% CI 0.33 – 0.44), followed closely by Fennoscandia (FST = 0.36, 95% CI 0.30–0.42), whereas the Alps had significantly lower among-population differentiation (FST = 0.24, 95% CI 0.20–0.27). Although the major part of genetic variation was found within populations (64.0%), with only 19.5% variation among populations within regions and 16.6% among regions (amova, Table 4), exact tests showed a strong genetic differentiation among the three regions (all three comparisons P < 0.001). When each region was analysed separately, most of the variation was again detected within populations, with up to 84.8% of the variance found within populations in the Alps. Pairwise exact tests of population differentiation indicate that most pairs of populations from different regions significantly differed for their marker frequencies (Table 2). However, the three populations from the oriental Pyrenees (seven, eight and 11) were not significantly differentiated from some Alpine populations (Table 2), indicating a close relationship between these two regions. To test whether the Pyrenees and the Alps really represent two differentiated regions, we excluded Fennoscandia from the amova and partitioned total variance among the Pyrenees and the Alps (11 populations): variance among regions fell down to 7%, but this was still significant (Table 4). Within regions, most pairwise comparisons showed no significant difference for marker


2344 L . D E S P R E S , S . L O R I O T and M . G A U D E U L Table 4 Analysis of molecular variance (amova) based on 117 AFLP loci in three regions: Alps, Pyrenees and Fennoscandia. Global analysis (three regions), the group Alps + Pyrenees, and each region analysed separately. The Polish and Romanian populations were excluded from the analysis because they cannot be assigned to one of the three regions

Source of variation

d.f.

Sum of squares (SSD)

Variance components

% of the total variance

Global analysis Among regions Among populations within regions Within populations Total

2 13 143 158

410.919 606.200 1660.567 2677.686

3.01004 3.52397 11.61235 18.14636

16.59 19.42 63.99

< 0.001 < 0.001

Alps + Pyrenees Among regions Among populations within regions Within populations Total

1 9 98 108

112.495 422.800 1212.467 1747.761

1.21520 3.49224 12.37211 17.07954

7.11 20.45 72.44

< 0.005 < 0.001

Alps Among populations in the Alps Within populations in the Alps Total

5 54 59

190.950 739.500 930.450

2.44956 13.69444 16.14400

15.17 84.83

< 0.001

Pyrenees Among populations in the Pyrenees Within populations in the Pyrenees Total

4 45 49

231.850 472.967 704.816

4.81968 10.74924 15.56892

30.96 69.04

< 0.001

Fennoscandia Among populations in Fennoscandia Within populations in Fennoscandia Total

4 45 49

183.400 448.100 631.500

3.58922 9.95778 13.5470

26.49 73.51

< 0.001

frequencies, except for the German population differing from one south-alpine population (Fournel), one Pyrenean population (Gazies) differing from all the other populations in the Pyrenees, and the Finnish population differing from the Norwegian populations.

Genetic vs. geographical structure The overall Mantel test based on the 18 populations was significant (r = 0.4904; P = 0.001). At a regional scale, there was no correlation between genetic and geographical distances in the Pyrenees (r = 0. 1882; P = 0.191), nor in the Alps (r = 0.3201; P = 0.225), nor in a region grouping Pyrenean and Alpine populations (r = 0.3050; P = 0.09). By contrast, a significant positive correlation between the two distance matrices was found in Fennoscandia (r = 0.7494; P = 0.03).

Discussion Within-population diversity Whatever the region, most of the genetic diversity was found within populations of the European globeflower.

P-value

Such a finding of higher genetic diversity within rather than among populations was reported commonly in outcrossing and/or perennial plants, by contrast to selfing and annual species which tend to exhibit the opposite pattern (Hamrick et al. 1991). The highest mean within-population diversity was found in the Alps, and the lowest in Fennoscandia. This finding of lowest variability in northern populations is probably explained by the past history of these populations, and is in accordance with the ‘tabula rasa hypothesis’: only a small sample of the southern genetic variability was represented in the migrants that recolonized the newly available space in the north (founder effect). However, the level of variability within Fennoscandian populations was still quite high (ranging from 37.6 to 48.7% of polymorphic loci), and not significantly different from the variability observed within Pyrenean populations (range 28.2–59.8%). In the Pyrenees, the lowest within-population diversity was observed in Gazies, which is also one of the smallest populations sampled in this region. By contrast, another population in the Pyrenees, Eynes, exhibited similar level of genetic diversity as Alpine populations. This population is one of the largest populations sampled in the Pyrenees. This suggests that the comparatively low genetic variability © 2002 Blackwell Science Ltd, Molecular Ecology, 11, 2337–2347


G E N E T I C D I V E R S I T Y I N T H E E U R O P E A N G L O B E F L O W E R 2345 observed in the Pyrenees could be the result of genetic drift (random fixation of alleles) in small populations, especially strong at the southern edge of their distribution. A positive correlation between population size and genetic variability has been found in many plant species (Fischer & Matthies 1998; Gaudeul et al. 2000), in agreement with the hypothesis that small populations cannot maintain diversities as high as those found in larger populations. No such relationship between population size and genetic diversity can be evidenced in the present study, because large populations in Fennoscandia do not exhibit the highest level of genetic variability, probably due to the founder effect when demographic expansion occurred. Genetic drift in the Pyrenees is also supported by the fact that different markers are polymorphic in different populations, indicating random allele fixation, and by a higher heterogeneity between within-population diversity indices. Therefore, relatively low and similar levels of withinpopulation variability in the Pyrenees and in Fennoscandia result from two very different processes: genetic drift due to current small population sizes in the Pyrenees, and a founder effect during postglacial recolonization in Fennoscandian populations, followed by demographic expansion. The high genetic diversity observed within Alpine populations reflects the proximity of a glacial refugium, and the fact that populations are larger than in the Pyrenees. The presence of a glacial refugium in southwestern Alps was evidenced in Picea abies (Scotti et al. 2000). The high genetic variability found within Alpine populations is concordant with the richness of its associated pollinators community.

Among-population differentiation Despite the fact that most variability was observed withinpopulations, overall among-population differentiation was high (FST = 0.39) suggesting that there is little gene flow among populations. This is not very surprising given the wide geographical range surveyed (distances between two populations range from 2 to 3252 km). Amongpopulations gene flow is limited by pollen and seed dispersal. T. europaeus being an insect-pollinated plant, pollen dispersal is limited by the flying capacity of its pollinator, a small short-living fly (Chiastocheta genus). An allozyme analysis of the genetic structure of Chiastocheta populations at a fine geographical scale (in Denmark) evidenced strong local differentiation and low dispersal (Johannesen & Loeschcke 1996). Moreover, seed dispersal is not likely to be very efficient, given the weight of T. europaeus seed (approximately 0.5 mg, excluding wind transport) and its lack of dispersal structures. The Pyrenees was the region with the highest amongpopulations differentiation (FST = 0.39), despite being the smallest region surveyed (maximum distance between © 2002 Blackwell Science Ltd, Molecular Ecology, 11, 2337–2347

populations = 206 km). There was no correlation between genetic and geographical distances in the Pyrenees, as expected if genetic drift is the major factor of differentiation. For example, Gazies was genetically distant from all the other Pyrenean populations, although it was only 10 km apart from Gabardères. The high among-population differentiation observed in the Pyrenees is the result of the random fixation of different alleles in different populations (up to 84 alleles were fixed in Gazies), regardless of their geographical location: this indicates that populations evolve independently from each other, and that gene flow is very limited among them. By contrast, genetic distances between Fennoscandian populations increased with geographical distances, in agreement with an isolation by distance model (Ellstrand & Elam 1993). This pattern may reflect a stepwise recolonization process following glacier retreat, and/or present day short-distance gene flow among more or less continuous large populations. The lack of correlation between genetic and geographical distances in the Alps indicates that genetic drift is acting in the Alps, but to a lesser extent than in the Pyrenees: gene flow among populations is higher in the Alps than in the Pyrenees as showed by lower among-population differentiation. Thus, gene flow seems to counteract the effect of genetic drift, and limits genetic erosion.

Phylogenetic relationships among populations and colonization routes The 18 globeflower populations studied form three clusters corresponding to three present-day southern refugia for T. europaeus: the Alpine, Pyrenean and Carpathian refugia. The lack of differentiation between some Alpine and Pyrenean populations, together with the low bootstrap values obtained for the two clades ‘Pyrenees’ and ‘Alps’ which are clustered together, suggests that the Alpine and the Pyrenean regions may have exchanged migrants during the last glacial period. Depending on the intensity and regularity of these past exchanges, the Alpine and Pyrenean populations could even be considered as presentday relics of a single large T. europaeus population, as the French Mediterranean region was not covered with ice during the last ice-age. However, we evidenced significant genetic differentiation between these two regions, and nonsignificant correlation between geographical and genetic distances, indicating that even if some connections did occur at some time between the Alpine and the Pyrenean ancestral populations, genetic drift is more important than gene flow in shaping present-day genetic diversity patterns. We found no evidence of contribution of the more eastern refuge (Carpathians) to the recolonization of the Alps, but our sampling included only populations from the French Alps, i.e. western Alps. A recent genetic study of the Alpine populations of Picea abies evidenced the


2346 L . D E S P R E S , S . L O R I O T and M . G A U D E U L presence of two lineages in the Alps, with the western populations being well differentiated from the eastern populations (Scotti et al. 2000). The population from Germany, Fulda, is grouped with the Alpine populations. This suggests that there is a route northward from the Alpine refugium. All the Fennoscandian populations sampled nest inside a cluster with Polish and Romanian population in a basal position, indicating that north Arctic was probably recolonized from an eastern Carpathian refugia, with no admixture of populations from the Pyrenean or Alpine refugia. Several recent molecular phylogeographies involving animals and plants evidenced the postglacial recolonization of northern Fennoscandia from the Balkanic refugia, together with the recolonization of southern Scandinavia from the Iberian or Italian refuges (Taberlet et al. 1998; Hewitt 2001). These expansion routes meet in central Scandinavia where the ice-cap melted some 9000 years ago and form an hybrid zone. Our sampling did not include globeflowers from southern Scandinavia, but such an east–west colonization pattern for T. europaeus is suggested by the German population expanding northward from the Alps. Although a number of studies demonstrated recolonization by coldtolerant plant species of boreal regions from high arctic refugia (Abbott et al. 1995; Lagercrantz & Ryman 1990; Tremblay & Schoen 1999), our study shows no evidence of a northern refugium for T. europaeus. This cold-tolerant plant is very sensitive to dryness, and is totally dependent on Chiastocheta fly activity for its reproduction, so that its survival during dry glacial periods in high arctic ice-free refugia was probably precluded. The lack of northern refugia for the host-plant precludes the possibility for Chiastocheta allopatric speciation in differentiated southern/ northern refugia. Furthermore, the lack of strong genetic differentiation between the Alpine and Pyrenean globeflower populations suggests that these two regions were connected during Pleistocene, presumably preventing allopatric fly speciation. No Chiastocheta species is specific to a geographical region. This suggests that sympatric speciation may have occurred in large, ancestral, southern host-plant populations throughout disruptive selection on the date of oviposition. In this perspective, the absence of some Chiastocheta species in the Pyrenees and Fennoscandia could result from species extinction during fragmentation of the host-plant populations (in the Pyrenees), and species loss during northward expansion (in Fennoscandia).

Conclusion Our study shows that T. europaeus probably survived the last glaciation in more or less interconnected southern populations, represented nowadays by the relictual populations found at high elevation in the Pyrenees, Alps and Carpathians. The high arctic populations originated

from one single southeastern lineage, supporting the ‘tabula rasa’ hypothesis. Moreover, the analysis and interpretation of our data at several spatial and temporal scales suggests that distinct evolutionary processes are responsible for the patterns of within and among population genetic diversity observed in the Alps, Pyrenees and Fennoscandia. Whereas genetic drift seems to be the major factor acting in the small Pyrenean populations, leading to decreased within-population variability and increased among-population differentiation with no correlation between genetic and geographical distances (random fixation of alleles), the low diversity in Fennoscandia would rather be due to a past founder effect during northward recolonization. This past recolonization wave and/or the present gene exchange between adjacent large populations are concordant with the isolation-by-distance model. Last, the Alpine populations retain most of the ancestral genetic variability, in a moderately fragmented habitat, and therefore exhibit the maximum withinpopulation diversity. The geographical pattern of genetic variation observed in T. europaeus supports sympatric speciation of associated flies, rather than allopatric speciations in well-isolated refugial populations. Local species extinction could have occurred latter, during host-plant range fragmentation in the south, and during northern expansion.

Acknowledgements We thank all the T. europaeus collectors: Jes Johannesen (Germany), Frank d’Amico (France), Ada Wrodlewska (Bierbza National Park, Poland), Ioan Chintauan (Carpathians, Romania), Gilles Yoccoz (Norway), Asa Hemborg (Abisko Biological station, Sweden) and Pirkko Siikamaki (Oulanka Biological Station, Finland). We aknowledge Ludovic Gielly for his help with the automated sequencer, Jean-Marie Cornuet and Pierre Berthier for providing the computer programs NJBS and DDM, respectively, and Pierre Taberlet for helpful comments on the manuscript.

References Abbott RJ, Chapman HM, Crawford RMM, Forbes DG (1995) Molecular diversity and derivations of populations of Silene acaulis and saxifraga oppositifolia from the high Arctic and southerly latitudes. Molecular Ecology, 4, 199 –207. Birks HJ (1996) Statistical approaches to interpreting diversity patterns in the Norwegian mountain flora. Ecography, 19, 332–340. Comes HP, Kadereit JW (1998) The effects of Quaternary climatic changes on plant distribution and evolution. Trends in Plant Science, 3, 432–438. Després L, Jaeger N (1999) Evolution of oviposition strategies and speciation in the globeflower flies Chiastocheta spp. (Anthomyiidae). Journal of Evolutionary Biology, 12, 822 – 831. Després L, Pettex E, Plaisance V, Pompanon F (2002) Speciation in the globeflower flies Chiastocheta ssp. (Diptera: Anthomyiidae) in relation to host plant species, biogeography and morphology. Molecular Phylogenetics and Evolution, 22, 258 – 268. © 2002 Blackwell Science Ltd, Molecular Ecology, 11, 2337–2347


G E N E T I C D I V E R S I T Y I N T H E E U R O P E A N G L O B E F L O W E R 2347 Ellstrand NC, Elam DR (1993) Population genetic consequences of small population size: implications for plant conservation. Annual Review of Ecology and Systematics, 24, 217–242. Ferdy JB, Després L, Godelle B (2002) Evolution of mutualism between globeflowers and their pollinating flies. Journal of Theoretical Biology, 217, 219– 234. Fischer M, Matthies D (1998) RAPD variation in relation to population size and plant fitness in the rare Gentianella germanica (Gentianaceae). American Journal of Botany, 85, 811–819. Gaudeul M, Taberlet P, Till-Bottraud I (2000) Genetic diversity in an endangered Alpine plant, Eryngium alpinum L. (Apiaceae), inferred from amplified fragment length polymorphism markers. Molecular Ecology, 9, 1625 –1637. Hamrick JL, Godt MJW, Murawski DA, Loveless MD (1991) Correlations betwwen species traits and allozyme diversity: implications for conservation biology. In: Genetics and Conservation of Rare Plants (eds Falk DA, Holsinger KE), pp. 75–86. Oxford University Press, Oxford. Hemborg A, Després L (1999) Oviposition by mutualistic seed parasitic pollinators and its effects on annual fitness of singleand multi-flowered host plants. Oecologia, 120, 427–436. Hewitt GM (1996) Some genetic consequences of ice ages, and their role in divergence and speciation. Biological Journal of the Linnaean Society, 58, 247– 276. Hewitt GM (2001) Speciation, hybrid zones and phylogeographyseeing genes in space and time. Molecular Ecology, 10, 537– 549. Jaeger N, Després L (1998) Obligate mutualism between Trollius europaeus and its seed-parasite pollinators Chiastocheta flies in the Alps. Comptes Rendus de l’Academie des Sciences, Paris, Séries III, 321, 789 –796. Jaeger N, Pompanon F, Després L (2001) Variation in predation costs with Chiastocheta egg number on Trollius europaeus: how many seeds to pay for pollination? Ecological Entomology, 26, 56 – 62. Jaeger N, Till-Bottraud I, Després L (2000) Evolutionnary conflict between Trollius europaeus and its seed-parasite pollinator Chiastocheta flies. Evolutionary Ecology Research, 2, 885–896. Johannesen J, Loeschcke V (1996) A hierarchical analysis of genetic structure and variability in patchily distributed coexisting Chiastocheta species (Diptera: Anthomyiidae). Heredity, 76, 437–448. Lagercrantz U, Ryman N (1990) Genetic structure of Norway spruce (Picea abies): concordance of morphological and allozymic variation. Evolution, 44, 38 – 53. Lewontin RC (1972) The apportionment of human diversity. Evolutionary Biology, 6, 381– 394. Mantel N (1967) The detection of disease clustering and a generalized regression approach. Cancer Research, 27, 209–220. Miller MP (1997) Tools for Population Genetic Analysis (TFPGA), Version 1.3. Department of Biological Sciences, Northern Arizona University, AZ, USA. Mueller UG, Wolfenbarger LL (1999) AFLP genotyping and fingerprinting. Trends in Ecology and Evolution, 14, 389–395. Nei M (1978) Estimation of average heterozygosity and genetic distance from a small number of individuals. Genetics, 89, 583 – 590.


Nordal I (1987) Tabula rasa after all? Botanical evidence for ice-free refugia in Scandinavia reviewed. Journal of Biogeography, 14, 377–388. Pellmyr O (1989) The cost of mutualism between Trollius europaeus and its pollinating parasites. Oecologia, 78, 53 – 59. Pellmyr O (1992) The phylogeny of a mutualism: evolution and coadaptation between Trollius and its seed-parasitic pollinators. Biological Journal of the Linnean Society, 47, 337– 365. Raymond M, Rousset F (1995) An exact test for population differentiation. Evolution, 49, 1280–1283. Schneider S, Küffer JM, Rössli D, Excoffier L (2000) Arlequin, Version 2.0: a Software for Population Genetic Data Analysis. Genetics and Biometry Laboratory, University of Geneva, Geneva. Scotti I, Vendramin GG, Matteotti LS, Scarponi C, Sari-Gorla M, Binelli G (2000) Postglacial recolonization routes for Picea abies in Italy as suggested by the analysis of sequencecharacterized amplified region (SCAR) markers. Molecular Ecology, 9, 699–708. Taberlet P, Fumagalli L, Wust-Saucy A-G, Cosson JF (1998) Comparative phylogeography and post-glacial colonization routes in Europe. Molecular Ecology, 7, 453 – 464. Tremblay NO, Schoen DJ (1999) Molecular phylogeography of Dryas integrifolia: glacial refugia and post-glacial recolonization. Molecular Ecology, 8, 1187–1198. Vendramin GG, Anzidei M, Madaghiele A, Sperisen C, Bucci G (2000) Chloroplast microsatellite analysis reveals the presence of population subdivision in Norway spruce (Picea abies K.). Genome, 43, 68–78. Vos P, Hogers R, Bleeker M et al. (1995) AFLP: a new technique for DNA fingerprinting. Nucleic Acids Research, 23, 4407– 4414. Walter R, Epperson BK (2001) Geographic pattern of genetic variation in Pinus resinosa: area of greatest diversity is not the origin of postglacial populations. Molecular Ecology, 10, 103 –111. Webb T, Bartlein PJ (1992) Global changes during the last 3 million years: climatic controls and biotic response. Annual Review of Ecological Systematic, 23, 141–173. Weir BS, Cockerham CC (1984) Estimating F-statistics for the analysis of population structure. Evolution, 38, 1358 –1370. Weising K, Nybom H, Wolff K, Meyer W (1995) DNA-fingerprinting in Plants and Fungi. CRC Press, Boca Raton. Yeh F, Yang RC, Boyle T (1997) POPGENE. The User-Friendly Shareware for Population Genetic Analysis. Molecular and Biotechnology Centre, University of Alberta, Edmonton.

Laurence Després teaches evolutionary ecology, population genetics and phylogeny at the University of Grenoble. Her main research interests are in speciation patterns and underlying evolutionary processes in host-parasites and plant–insects interactions. The present work is a contribution to a broad ongoing study on the evolutionary ecology of the Trollius/Chiastocheta interaction, and is part of Sandrine Loriot’s DEA (Master’s degree) on the population genetics of Trollius europaeus. Myriam Gaudeul is a PhD student interested in the evolutionary biology and conservation genetics and ecology of plant species.