Genetic differentiation of Rubus chamaemorus populations in the Czech Republic and Norway after the last glacial period

Abstract The population structure of cloudberry (Rubus chamaemorus L.), collected from Krkonose Mountains (the Czech Republic), continental Norway and Spitsbergen, was examined using microsatellite analyses (SSR). Among 184 individuals, 162 different genotypes were identified. The overall unbiased gene diversity was high (h^=0.463). A high level of genetic differentiation among populations (FST = 0.45; p < .01) indicated restricted gene flow between populations. Using a Bayesian approach, six clusters were found which represented the genetic structure of the studied cloudberry populations. The value of correlation index between genetic and geographical distances (r = .44) indicates that gene flow, even over a long distance, could exist. An exact test of population differentiation showed that Rubus chamaemorus populations from regions (Krkonose Mountains, continental Norway and Spitsbergen) are differentiated although some individuals within populations share common alleles even among regions. These results were confirmed by AMOVA, where the highest level of diversity was found within populations (70.8%). There was no difference between 87 pairs of populations (18.7%) mostly within cloudberry populations from continental Norway and from Spitsbergen. Based on obtained results, it is possible to conclude that Czech and Norwegian cloudberry populations are undergoing differentiation, which preserves unique allele compositions most likely from original populations during the last glaciation period. This knowledge will be important for the creation and continuation of in situ and ex situ conservation of cloudberry populations within these areas.

flavone, quercetin, and naringenin). In particular, the Alaskan Inuit and the Norwegian Sami use cloudberry as an important contribution to their diet (reviewed by Nilsen, 2005).
The opinions on cloudberry population diversity are not clear. Korpelainen, Antonius-Klemola, and Werlemark (1999) published the results of a diversity study of three Norwegian populations based on RAPD, SSR, and hybridization methods. Although cloudberry expressed clear variation in morphology, the level of genetic variability appeared to be low. This is also true for detected allozyme variability of the same cloudberry populations (Korpelainen, 1994). Debnath (2007) used intersimple sequence repeat (ISSR) PCR analysis to study genetic variability of 48 cloudberry clones from four Canadian Provinces. They found a substantial degree of genetic diversity, but only 8% of the total variation could be explained by geographical distribution (Debnath, 2007).
Interesting results were gained by Ehrich, Alsos, and Brochmann (2008) who studied 45 cloudberry populations through their main distribution area and two populations from Scotland. Based on AFLP analysis, they found a high level of genetic diversity among all populations, and more than one clone was found in nearly every local population. The phylogeographical pattern was assessed to be shallow. The authors concluded that the present circumpolar cloudberry distribution area has been colonized at least twice and possibly several times. The highest level of genetic diversity was found in the Taimyr Peninsula, Russia (Ehrich et al., 2008).
As a glacial relic, Rubus chamaemorus occurs in Alaska, British Columbia, SW Greenland, Siberia, Kamchatka, Kuril Islands, Sakhalin, North Korea, Scotland, and Poland (Hultén, 1968). The species also occurs in the Czech Republic (CR; Holub, 1995;Taylor, 1971), where it is the southernmost distribution in Europe. Here, it is recognized as a critically endangered species (Grulich, 2012) and is only found in two localities; both in the Krkonose Mountains (Kubát, 2002).
There is no information about the diversity of these populations and whether these populations have genetically diverged from populations of the closest main distribution area, such as the Scandinavian region. The results of such a study would be very useful for conservation management of cloudberry in the Czech Republic or in other places of its marginal occurrence. Therefore, the main aim of this study was to assess the genetic diversity, differentiation, and structure of isolated populations of Rubus chamaemorus from the Czech Republic with the comparison of cloudberry populations collected throughout Norway from the south coast to Spitsbergen. These results were then used to address the following questions: (1)  Fresh leaves were dried (Staats et al., 2011) and stored in ziplock plastic bags with silica gel until DNA extraction. Genomic DNA was extracted from silica gel dried material using a CTAB protocol (Doyle & Doyle, 1987;Drabkova, Kirschner, & Vlček, 2002), and the quality of the extracted DNA was checked on 0.7% agarose gels.

| Data analysis
Clone identity was determined using multilocus matches for codominant data. The probability of identity (i.e., estimating the probability of randomly matching two unrelated (PI) or related (PIsib) individuals by a particular set of loci) were calculated based on the distribution of allele frequencies in population samples using software GENECAP (Wilberg & Dreher, 2004).
A matrix of distances between all of the samples was calculated using the simple matching dissimilarity coefficient in the DARwin software (http://darwin.cirad.fr/darwin; Perrier & Jacquemoud-Collet, 2006). For clustering, an unweighted Neighbour-joining method (UNJ) was used as its cophenetic coefficient r showed the highest value (0.943). The support for the phenogram branches was obtained using 2,000 bootstrap resamplings.
The diversity statistics for each population included the percentage of polymorphic loci, the average diversity of the loci using Nei's unbiased gene diversity ĥ (Nei, 1973), and the Shannon information index (Lewontin, 1972;Shannon & Weaver, 1949). All of these statistics were calculated using the POPGENE software, version 1.32 (Yeh, Boyle, Rongcai, Ye, & Xiyan, 1999).
The divergence statistics were estimated using the hierarchical analysis of molecular variance (AMOVA; Excoffier, Smouse, & Quatro, 1992) which was performed using Arlequin version 3.5 (Excoffier & Lischer, 2010). It was used to partition the total genetic variation into three specific hierarchical levels: among the genotypes collected within localities, among the different localities within three "regions" (CR, continental Norway, and Spitsbergen), and between the "regions." The significance levels for the resultant molecular variance components were computed by default 1,023 nonparametric permutation procedures (Excoffier et al., 1992). Shannon information index as a measure of gene diversity (Shannon & Weaver, 1949).
runs of one-20 groups (K = 1-20) were performed using locprior model with admixture and correlated allele frequency (Falush, Stephens, & Pritchard, 2003;Hubisz, Falush, Stephens, & Pritchard, 2009) with the recommended 20,000 Markov chain iterations after a burning period of 10,000 iterations. The optimal value of K was estimated based on ln (K) and on the ΔK calculation, which considers the rate of change in the ln P(D) values among successive K runs to account for patterns of dispersal that are not homogeneous among populations (Evanno, Regnaut, & Goudet, 2005). The number (K) of clusters into which the sample data (X) were fitted with F I G U R E 2 (a) Sample localities of Rubus chamaemorus populations with pie charts describing the proportions of individuals classified into one of the six clusters defined using the Bayesian approach (Pritchard et al., 2000). Each color represents one of six clusters. (b) Direct output from Structure software for all populations for K = 6 posterior probability Pr (X|K) was estimated using the same model with 1,000,000 Markov chain iterations after a burning period of 100,000 iterations (Evanno et al., 2005).

| RE SULTS
A total of 180 alleles in 28 microsatellite loci were detected when analyzing 184 Rubus chamaemorus samples with 24 primer pairs. We identified a total of 162 multilocus genotypes. Forty matches were found with PI sib < 0.05; therefore, they were excluded from further analyses ( Table 1).     (Engel, Braucher, Traczyk, Laetitia, & Team, 2014;Hultén, 1968). It is known that the level of genetic variability in small isolated populations reduces as a consequence of genetic drift and bottlenecks. The possible explanation for the high diversity of cloudberry populations in this previously glaciated area is, according to Alsos, Engelskjon, Gielly, Taberlet, and Brochman (2005), efficient and broad fronted recolonization from large and diverse populations on the tundra surrounding the glaciers, as well as from more distant populations. Rubus chamaemorus occurrence in Central Europe was probably widespread in the areas in the Late-Glacial periods and has subsequently retreated and survived only on higher altitude peat boggy places with enough water and colder climates (Dostál, 1989).
According to our results, samples from the Krkonose Mountains are three genetically distinct populations (Figures 2 and 3). Russia, as the level of glaciation changed in the Quaternary period (Engel et al., 2014). Therefore, population EK2 clustered within continental Norwegian cloudberry populations SN1, SN3, SN4, NN4, NN11, and NN12. Furthermore, WK1 and WK4 populations slightly differed from other populations in Krkonose Mountains (39% and 32%, respectively) and they are closer to Norwegian populations NN1-NN12 (24% and 9%, resp.) and S1-S7 (15% and 23%, resp.; Vectors of gene flow could be birds living on cloudberry fruits and being able to migrate through long distances, for example, gray lag goose which nests in the Hebrides, Scandinavia, and Russia, winters in the British Isles and has a flight speed and metabolism which seems to permit dispersal of seed between land masses in the North Atlantic (Löve, 1963). Gene flow over long distance could therefore be possible.
Rubus chamaemorus largely reproduces itself vegetatively. Few seeds are produced in most areas but in a restricted number of localities, seed is produced regularly in fair quantity, although viability is poor (Taylor, 1971). In the Krkonose Mountains, cloudberry flowers very rarely. The last time cloudberry flowered was in a cool spring in 2005 after a long winter with a good amount of snow (Dvorak, 2005). It seems that the limiting factor for flower development is the humidity of cloudberry habitats (Ehrich et al., 2008). In spite of these difficulties, random hybridization between individuals even from different populations can occur. This is supported by an individual it is necessary to conserve the widest range of its genetic diversity.
It will be necessary to work closely with national experts and conservation managers in the Czech Republic, continental Norway, and Spitsbergen to ensure the range of diversity illustrated here is conserved both in situ and ex situ.

This research was supported by project of Czech-Norwegian
Research Programme ID number 7F14122 and RO0414. The authors would like to thank Josef Harčařík from KRNAP allowing to collect the samples and Hana Udavská for excellent technical assistance.

CO N FLI C T O F I NTE R E S T
None declared.

AUTH O R CO NTR I B UTI O N S
L.L.S., J.P., I.M., and V.H. conceived the ideas and collected samples; L.L.S. conducted genotyping and analyzed data; and all authors contributed to interpretation and writing which was led by L.L.S.

DATA ACCE SS I B I LIT Y
All data are in supporting information, and they will be available at time of publication.