Biological Journal of the Linnean Society, 2017, 121, 294–304. With 4 figures.

Genetic differentiation in the boreal dragonfly Leucorrhinia dubia in the Palearctic region F. JOHANSSON1*, P. HALVARSSON1, D.J. MIKOLAJEWSKI2 and J. HÖGLUND1 Department of Ecology and Genetics, Uppsala University, SE-751 05 Uppsala, Sweden Institut für Biologie, Freie Universität Berlin, Königin-Luise-Str. 1–3, 14195 Berlin, Germany

1 2

Received 12 August 2016; revised 8 November 2016; accepted for publication 2 December 2016

The last glacial period had a strong influence on the population genetic structure of boreal species in southern and central Europe. In addition, recent and current human impact on the boreal environment has led to habitat loss, which also has a large influence on population genetic structure of species. Here we present the spatial genetic structure of the boreal dragonfly Leucorrhinia dubia using ddRAD sequencing. We sampled individuals from nine locations in Europe, three in Asia (Russia and Japan) and one location of L. intermedia in Japan. Results showed three distinct genetic clusters in Europe. One genetic cluster consisted of individuals sampled from the locations in the Swiss Alps, a second consisted of individuals sampled in the United Kingdom, and a third cluster consisted of individuals from the rest of the seven sampled locations in Europe covering a latitudinal gradient from the French Pyrenees to the north of Finland. There was also a week support that the French Pyrenees and Austrian Alps samples differentiated from the cluster of the five samples from central and north Europe. We suggest that these clusters reflect historical recolonization patterns since the last glaciation. The L. dubia individuals sampled from locations in Asia formed one cluster referring to L. dubia orientalis separated from the individuals sampled in European and from the L. intermedia locality sampled. Our result suggests that aquatic insects in the fragmented boreal landscape in south central Europe and United Kingdom need conservation consideration.

ADDITIONAL KEYWORDS:  boreal – ddRAD – genetic differentiation – Leucorrhinia – population structure – postglacial.

INTRODUCTION The boreal zone is an ecosystem in the northern hemisphere, today roughly being located between the 60° and 70° latitude north. Over the past, the distribution of species occurring in boreal habitats has been shifted southwards in response to ice ages and glaciers, with many boreal species having had their southern range limit much further south than currently (Hewitt, 1999). After retraction of the ice, these species moved their southern range limit to the north, where they now have their core distribution. However, in suitable habitats like at high altitude with corresponding environmental conditions, boreal species still are able to establish remaining populations even outside their current core distribution, often reflecting former species distribution during glacial periods (Gentili et al., 2015). The populations in such small isolated refuge *Corresponding author. E-mail: [email protected]

areas are sometimes interpreted as relict populations (Spitzer & Danks, 2006; Hájek et al., 2009), or species confined to cryptic southern refugia (Stewart et al., 2010). During the last ice age 18 000 BP, Northern Europe and the mountain areas of Europe were covered by ice, and south of this ice there was tundra with permafrost that extended to the borders of northern Spain, Italy and Greece (Hewitt, 1999). The Iberian Peninsula, Central Italy and the Balkan Peninsula thus served as separate refuges for temperate species that now have their core range in central and northern Europe. Re-colonization of Central and Northern Europe after retraction from ice consequently occurred independently from these three sources, with current genetic differentiation of high altitude relict and northern latitudes populations reflecting this past population structure (Hewitt, 1999). In contrast, cold-adapted species might have had a larger distribution during the ice age, since large area of cold climate covered

© 2017 The Linnean Society of London, Biological Journal of the Linnean Society, 2017, 121, 294–304

294

GENETIC DIFFERENTIATION IN L. DUBIA  central and southern Europe (Stewart et al., 2010). Nevertheless, the current distribution of species in the northern hemisphere as well as their genetic structure has been influenced heavily by the last glacial periods and the subsequent warming that occurred after the retraction of the ice (Hewitt, 2000; Stewart et al., 2010). In addition to natural changes affecting species distributions caused by, for instance, ice ages, more recent human actions tend to affect the range size of many species. Major impacts like the heavy deforestation that started about 5000 years ago and still continues today have altered the distribution of many boreal species (Wallenius et al., 2010). Thus, many boreal species currently show strong decreases in range size and have become critically endangered due to human activities (Samways, 1994). Along with deforestation, boreal wetlands such as bogs and mires have significantly declined due to loss of forest and peat land by draining (Spitzer & Danks, 2006). Also this habitat loss has strong consequences for boreal species. Especially vulnerable are populations located at the edge of species distribution. For instance, numerous boreal peat bog ponds in central Europe harbour cold-adapted species that are common in the northern part of Europe (Spitzer & Danks, 2006). Conservation of species requires knowledge of their current distribution, genetic diversity and past and present factors that affect these two variables. Such knowledge is especially important for relict populations in isolated refugia, since these isolated populations are very dependent on immigration and sensitive to demographic events (Höglund, 2009). In addition, conserving populations in isolated refugia is important because such populations might harbour genotypes and phenotypes that are important for adaptation in a changing environment (Forsman, 2014). Hence, by gaining distinct knowledge about population structures within a species, we can pinpoint areas that need special emphasis with regard to conservation, by devoting our restricted resources towards these populations. Molecular data provide an excellent tool to study population structures and migration among populations and enable us to identify the distribution of genotypes as well as pattern of isolation (Höglund, 2009). Despite the fact that we have good knowledge on the current status and change in distribution since the last ice age of many vertebrate, plants and terrestrial insect (Hewitt, 1999; Svennig, Normand & Kageyama, 2008), we have less knowledge of insects restricted to boreal forest and very little on insects associated to bog ponds in the boreal environment (Spitzer & Danks, 2006; Bernard et al., 2011). In this study, we attempt to fill this gap by providing information on genetic differentiation in a dragonfly species associated with bog

295

ponds in the boreal zone. Such knowledge should be important for conservation biologist, because it provides information about population connectivity across the entire species range and reveals levels of isolations for relict populations. The dragonfly Leucorrhinia dubia (Vander Linden, 1825) is a boreal dragonfly species which has a distribution from Western Europe to Japan and Kamchatka (Boudot & Kalkman, 2015). It occurs in oligotrophic to mesotrophic pools, ponds and lakes on moors and mosses with sphagnum and is confined to waters that are generally acidic (Askew, 2004; Boudot & Kalkman, 2015). The species is common in northern Europe and central and eastern Europe mountain areas, but rare throughout the rest of Europe. For example, it is rare and has scattered populations in the Pyrenees, lowlands in France and the southern United Kingdom (Boudot & Kalkman, 2015). In addition, it is red-listed in several European countries, for example Germany, Austria, Switzerland and United Kingdom (Gonseth & Monnerat, 2002; Daguet, French, & Taylor, 2008; Ott et al., 2015). Leucorrhinia dubia can thus be regarded as a relict species in its southern range (Fig. 1), with habitat fragmentation and peat land draining being a major threat in south and central Europe via fragmentation of populations and lack of dispersal among populations. Hence, strong isolation of populations might cause inbreeding and stochastic demographic effects, ultimately leading to local extinction. Knowledge about the genetic structure of Leucorrhinia dubia across Europe will provide insights about the past colonization of European habitats after retraction of glacial ice, but also about population structures and frequencies of genotypes as well as hint at connectivity of populations and potential threats to relict populations. The aim of this study is to evaluate the population structure and population differentiation of L. dubia across Europe comparing it to localities from Eastern Russia and Japan. For this, we analysed L. dubia samples from nine localities distributed across Europe with the addition of samples from three localities from Eastern Russia and Japan.

MATERIAL AND METHODS Molecular methods To examine genetic structure and population differentiation, we sampled adults and larvae from 12 L. dubia locations (Table S1). Nine of these sampled locations were from Europe, two from Russia (Asia) and one from Japan (Asia). The Asian samples were determined by the collectors as the subspecies L. dubia orientalis (Selys, 1887). Leucorrhinia dubia orientalis males lack red spots on segment four and five (red spots in L. dubia), and the spot on segment seven is

© 2017 The Linnean Society of London, Biological Journal of the Linnean Society, 2017, 121, 294–304

296  F. JOHANSSON ET AL. Biolabs, Massachusetts, USA), see Mastretta-Yanes et al. (2015) for full details. PCRs were replicated four times for the same sample and the individual positions on the PCR plate were randomly selected. After PCR, the samples were pooled and size selections were performed in the same gel. The library was then sequenced in a single separate lane on an Illumina HiSeq200 from both directions (2 × 125 bp) in high throughput mode at SciLifeLab, Uppsala, Sweden.

Bioinformatics pipeline

Figure 1.  (A) Map showing the distribution of Leucorrhinia dubia in Europe. The distribution east of Poland is unknown and scattered and therefore not shown. Dense shading denotes high abundance and light shading denotes low abundance. Grey dots indicate the sampled locations. The map is based on information from Askew (2004) and Boudot & Kalkman (2015). (B) Map showing the location of the sampled localities. The triangle symbol in Japan denotes Leucorrhinia intermedia.

yellow in mature individuals (red in L. dubia). The two species also differ in their shape of the dark basal spot on the hind wing. In female L. dubia orientalis, the spot part above vein CuA is missing or vestigial, which is diagnostic for orientalis vs. dubia. For more details see Belyshev (1973). We also extracted DNA from Leucorrhinia intermedia ssp. ijimai (Asahina, 1961) collected in Japan (Table S1) and included this sampled locality in the analyses as a reference location. After collection, samples were stored in ethanol until DNA extraction. DNA was extracted either from flight muscle in adults or from the abdominal muscles of last instar larvae, using a modified high salt extraction protocol (Paxton et al., 1996). Libraries for double digest restriction associated DNA (ddRAD) data were created using a modified version of protocols from Parchman et al. (2012), Peterson et al. (2012) and Mastretta-Yanes et al. (2015). In short, DNA was cut by the enzymes EcoRI-HF and MseI, individual tags and primer sites ligated with T4 DNA ligase, and PCR was conducted with Q5 DNA polymerase (New England

The raw reads were quality filtered and demultiplexed using STACKS v1.34 (Catchen et al., 2013) software process_radtags. Demultiplexed data were assembled de novo by denovo_map.pl in STACKS with the values (m = 3, M = 4, n = 2). The single nucleotide polymorphism (SNP) locus for each stack was extracted using POPULATIONS for further analysis. Since different species or distant populations are expected to share fewer alleles, four data sets were extracted (Table S1). The first contained individuals from L. dubia (i.e. all 12 sampled L. dubia location including L. dubia orientalis) as well as L. intermedia (925 SNPs in 104 individuals from 13 locations sampled). The second contained all sampled L. dubia locations, including L. dubia orientalis (1315 SNPs in 98 individuals from 12 locations), the third only contained European L. dubia (1674 SNPs in 79 individuals from nine ­locations) and the fourth included only European locations with exception of the United Kingdom location (1674 SNPs in 65 individuals from eight locations). The computations were performed on resources provided by SNIC through Uppsala Multidisciplinary Center for Advanced Computational Science (UPPMAX) under Project b2015095 (Lampa et al., 2013).

Population structure and natural population genetic differentiation indices

To find genetically homogeneous groups of individuals, (K), we used STRUCTURE v2.3.4 (Pritchard, Stephens & Donnelly, 2000). Individuals were assigned to these clusters under the assumption of admixture using sample location as prior information. All data sets were run ten times per K = 1–7 with 100 000 burn-in followed by 1 000 000 MCMC iterations. To adjust for minor allele frequencies, lambda was set to 0.3 after letting STRUCTURE estimating it for K = 1. Averaging of all STRUCTURE runs for each K was done using CLUMPP v1.1.2 (Jakobsson & Rosenberg, 2007) and visualized using the R package POPHELPER v1.1.6 (Francis, 2017). STRUCTURE was run on the four data sets given above. To investigate population structure in individuals sampled from the European locations only, the R package ADEGENET (Jombart, 2008;

© 2017 The Linnean Society of London, Biological Journal of the Linnean Society, 2017, 121, 294–304

GENETIC DIFFERENTIATION IN L. DUBIA  Jombart & Ahmed, 2011) was used for discriminant analysis of principle components (DAPC) to visualize multidimensional PCA relationships between each individual in a two-dimensional plot. To investigate population differentiation between the sampled locations, pair-wise FST and Nei’s genetic distance were calculated using SpaGeDi v1.5 (Hardy & Vekemans, 2002). Increasing distance between locations restricts gene flow, and therefore, FST and differentiation should correlate positively with distance (Wright, 1943) and deviations from linearity can indicate gene flow, isolation, or change in range. We also estimated isolation by distance (IBD) using the same package, SpaGeDi v1.5. This program uses FST values to visualize and compare distances classes of the sampled localities. Therefore, Straight line distances between the locations sampled were measured manually using Google Earth v7.1.4.1529 (Google Inc., California, USA) ‘Map Length’ measurement.

RESULTS Population structure When all sampled Leucorrhinia locations where included in the STRUCTURE analysis, the Evanno method (Evanno et al., 2005) gave the highest support for K = 3 (delta-K) (Figs 2 and S1). Leucorrhinia intermedia corresponded to one cluster, which is not surprising given it is a different species. Leucorrhinia dubia orientalis from Japan and Russia formed a second cluster and all the sampled L. dubia locations from Europe a third cluster. Hence, L. dubia dubia and L. dubia orientalis formed two separate clusters when all sampled locations are considered. If L. intermedia is excluded, three clusters were supported (Figs S1 and S2): one with L. dubia orientalis from Japan, one with L. dubia orientalis from Russia and a third one consisting of L. dubia orientalis from all the European locations. Thus, this analysis suggests that the sampled L. dubia orientalis locations from Russia and Japan do not cluster together. STRUCTURE analyses considering only European sampled locations of L. dubia suggest three clusters (Figs 2 and S1). One cluster consists of the United Kingdom locations, and a second cluster consists of the French location from the Pyrenees which cluster together with the Swiss Alp location. The third cluster consists of the rest of the six European locations. Interestingly the sampled Austrian Alps location does not cluster with the sampled Swiss Alps location. This pattern of clustering was also found when we run the STRUCTURE analysis on the mainland European locations (excluding the sampled United Kingdom location) (Figs S1 and S2). The discriminant analysis of principle components (DAPC) supported three main clusters among the

297

European samples based on the Bayesian Information Criterion. The United Kingdom separated from the rest of the sampled European locations along the first discriminant axis, suggesting a distinct United Kingdom population (Fig. 3). The sampled Swiss Alp location separated from the majority of the sampled locations along the second discriminant axis, suggesting a distinct population in this part of the Alps. The third cluster consisted of the remaining seven sampled locations. However, the sampled French Pyrenees and Austrian Alps locations were somewhat separated from the majority of the remaining seven sampled locations (Fig. 3). This pattern can also be seen in the STRUCURE analysis at K = 6 (Fig. 2).

Population differentiation Pair-wise F ST values ranged between 0.0001 and 0.32 when all samples were included in the analyses (Table 1). The lowest values were found between sampled locations from Finland, Sweden and Poland. The highest were found between L. intermedia location and the L. dubia locations. High FST values were also found between sampled Japanese L. dubia orientalis, Russian L. dubia orientalis locations and the sampled European locations of L. dubia dubia. Nei’s genetic distance values mirrored those of the FST values (Table 1). When pair-wise FST analyses were run only on the sampled L. dubia locations from Europe, the values ranged between 0.0002 and 0.055 (Table 2). The lowest values were found between Polish and German locations and among Polish, Swedish and Finnish locations. The highest values were found between locations sampled from United Kingdom and the rest of Europe, but also between the locations sampled in the French Pyrenees and Swiss alp and the rest of the locations in Europe (Table 2). The IBD analyses found ten distance classes which are visualized, and the results suggest that the Japanese and Russian locations show an IBD pattern: compare mean and error bars (Fig. 4). However, when only Europe was included in the analyses, there was no evidence of an IBD (Fig. 4).

DISCUSSION When we analysed all the 13 sampled locations in STRUCTURE, three distinct clusters were evident: one consisted of L. intermedia, another of L. dubia orientalis from Japan and Russia, and a third including all European locations. We will discuss these clusters in relation to environmental factors shaping patterns of population divergence as well as its consequences from a conservational point of view, and we start with the European cluster.

© 2017 The Linnean Society of London, Biological Journal of the Linnean Society, 2017, 121, 294–304

298  F. JOHANSSON ET AL.

Figure 2.  Bayesian clustering results. A assignment of individuals using program STRUCTURE. Each plot from aligning ten STRUCTURE runs using CLUMPP for K = 2–7. Each vertical bar represents one individual and the colour the probability for each individual to belong to a certain population. Left column: 925 SNPs from 104 individuals of L. dubia and L. intermedia. Right column: European population of L. dubia, 1674 SNPs from 79 individuals.

Europe The structure analysis and the discriminant analysis suggested three subclusters of L. dubia in Europe: one consisting of animals from the United Kingdom, one from the Swiss Alps and a third consisting of the remaining seven locations. In addition, within the

seven locations, the French Pyrenees and the Austrian Alps locations were slightly separated from the others. A similar structure was found with regard to the FST values, because these were high between United Kingdom locations and the rest of the European locations and between the Swiss location and the other

© 2017 The Linnean Society of London, Biological Journal of the Linnean Society, 2017, 121, 294–304

GENETIC DIFFERENTIATION IN L. DUBIA 

Figure 3.  Discriminant analysis of principle components (DAPC) scatterplot of L. dubia genotypes. Twenty-six principle components (PC) and two discriminant functions (dimensions) were retained to describe the relationships between clusters. DAPC is based on 1674 SNPs from 78 individuals in European L. dubia locations. Each individual is represented by a symbol and each symbol and colour combination indicate a population.

eight European locations. Such genetic isolated clusters between Pyrenean, Central European and British populations are common in other animals confined to boreal forest (Hewitt, 1999; Segelbacher et al., 2003). Here we present one of the first examples from an aquatic insect species that is associated with bog ponds in the boreal area of Europe. From a conservation perspective, our result suggests that preserving boreal forest for the purpose of more charismatic species such as capercaillie and black grouse (Storch, 2000) might also benefit aquatic insects confined to bog ponds in the boreal forest. The structure analysis and to some degree the discriminate analysis suggest a genetic differentiation between the Swiss Alps and French Pyrenees and the rest of Europe. This differentiation may reflect past colonization history since the last glacial period. Probably the ancestral populations of the French and the Swiss populations had their glacial refuge in the Iberian Peninsula, while the other European population had their ancestral origin in Italy and/or the Balkans. Such a re-colonization pattern following the recent ice age is common in plant and animal species in Europe (Hewitt, 1999; Segelbacher et al., 2003; Bhagwat & Willis, 2008). For example, Segelbacher et al. (2003) found low genetic differentiation within the continuous

299

range in the boreal forest and significant differences between isolated central Europe populations and the Alpine and boreal populations for capercaillie, another boreal species (see also Klinga et al., 2015). Interestingly, L. dubia individuals sampled the Swiss Alps and the Austrian Alps locations did not cluster together even though the distance between them is only 42 km with an altitude difference of 106 m. This is in contrast to the study by Segelbacher et al. (2003) who found low genetic differentiation in the Alps in their boreal species. Our data suggest that L. dubia individuals in these two locations have different origin since the last ice age. Alternatively, these two localities were colonized with the same gene pool thousands of years ago and have since then been separated genetically because of low gene flow caused by high mountain tops in between them. It would be interesting to sample more locations in the Alps to explore whether there actually is a contact zone separating the two invasion routes or whether the isolated populations in the Alps are random colonization events reflecting haphazardly distributed populations originating from the east and west postglacial expansion. The British location formed an isolated cluster, suggesting that they are genetically isolated from the mainland Europe. If the British islands were colonized by few individuals, such a founder effect might have contributed to the genetic differentiation between the British population and the mainland Europe. It would be interesting to analyses which of the mainland population that are most related to the British ones, and we will answer this in a forthcoming article that is under progress. We found no IBD pattern among the European locations. This suggests that L. dubia has good dispersal abilities despite the three clusters found in Europe. This is what is predicted if Europe was colonized from two distinct glacial refugia. Under such a scenario, populations with distinct demographic histories and ancestry (genetically distinct lineages) might be geographically close (like in the Alps), thus obscuring any pattern of IBD. Also strong genetic drift in small and isolated populations may have affected the allele frequency distributions further obscuring any geographic signals (Rogell et al., 2010; Segelbacher et al., 2014). The structure and discriminant analyses suggest a high connectivity between the central and north European locations. This was somewhat surprising given that the more recently decline in suitable habitats of L. dubia in central Europe has resulted in a noncontinuous population distribution. For example, Belgium and Germany populations are very scattered and do not hold high population densities (De Knijf, 2006; Daguet, French &Taylor, 2008) compared to, for example, Swedish populations (Billqvist, Smallshire, & Swash, 2012). Perhaps a larger sample size would

© 2017 The Linnean Society of London, Biological Journal of the Linnean Society, 2017, 121, 294–304

0.3640 0.1689 0.1414 0.1983 0.1725 0.1045 0.1467 0.3676 0.6840 0.7135 0.8610 0.8794

France Switzerland Austria Belgum Germany Poland Sweden Finland UK Russia (Blagoves.) Russia (Tuva) Japan (Akan.) Japan (Yamahana) 0.1478 0.1121 0.1094 0.1007 0.1638 0.1059 0.3307 0.6981 0.7300 0.8745 0.8795

0.0047

Switzerland

0.0553 0.0672 0.0624 0.0827 0.0611 0.2412 0.7057 0.7387 0.8707 0.8764

0.0021 0.0018

Austria

0.0513 0.0048 0.0158 0.0165 0.1492 0.7468 0.7790 0.8845 0.8894

0.0028 0.0022 0.0010

Belgum

0.0118 0.0113 0.0149 0.1012 0.7651 0.7975 0.8941 0.9029

0.0029 0.0018 0.0009 0.0008

Germany

0.0052 0.0152 0.1303 0.7582 0.7931 0.8973 0.9019

0.0024 0.0016 0.0008 0.0002 0.0002

Poland

0.0058 0.1043 0.7232 0.7561 0.8687 0.8875

0.0022 0.0033 0.0015 0.0004 0.0002 0.0003

Sweden

0.1146 0.7685 0.7998 0.8912 0.9006

0.0026 0.0021 0.0011 0.0003 0.0002 0.0003 0.0002

Finland

0.7999 0.8353 0.9227 0.9208

0.0042 0.0036 0.0023 0.0017 0.0010 0.0013 0.0011 0.0014

UK

0.1015 0.3910 0.7992

0.1705 0.1700 0.1686 0.1662 0.1679 0.1670 0.1660 0.1651 0.1687

Russia (Blagoves.)

0.4052 0.8012

0.1699 0.1694 0.1680 0.1656 0.1670 0.1663 0.1651 0.1644 0.1681 0.0162

Russia (Tuva)

France

0.2318 0.1806 0.1459 0.1382 0.1130 0.1208 0.1205 0.3620

ALL LOCI

France Switzerland Austria Belgium Germany Poland Sweden Finland United Kingdom 0.2044 0.1754 0.1595 0.1592 0.1566 0.1550 0.3954

0.0328

Switzerland

0.1281 0.1136 0.1008 0.1250 0.1096 0.3411

0.0281 0.0318

Austria

0.0316 0.0245 0.0492 0.0345 0.2570

0.0241 0.0288 0.0196

Belgum

0.0021 0.0262 0.0154 0.2399

0.0220 0.0258 0.0172 0.0047

Germany

0.0193 0.0159 0.2353

0.0188 0.0260 0.0153 0.0039 0.0010

Poland

–0.0001 0.2412

0.0199 0.0261 0.0198 0.0077 0.0043 0.0035

Sweden

0.2240

0.0216 0.0272 0.0179 0.0055 0.0027 0.0030 0.0009

Finland

Table 2.  Pairwise F ST values (below the diagonal) and Nei’s genetic distance (Ds) (above the diagonal) from European L. dubia locations

France

ALL LOCI

0.3208 0.3225 0.3190 0.3159 0.3170 0.3167 0.3179 0.3154 0.3172 0.2753 0.2732 0.3020

Japan (Yamahana)

0.0458 0.0554 0.0459 0.0342 0.0315 0.0304 0.0302 0.0309

United Kingdom

0.8267

0.2069 0.2057 0.2042 0.2037 0.2040 0.2035 0.2021 0.2015 0.2043 0.0651 0.0654

Japan (Akan.)

Table 1.   Location pairwise FST values (below the diagonal) and Nei’s genetic distance (Ds) (above the diagonal) for L. dubia and L. intermedia (last row/column)

300  F. JOHANSSON ET AL.

© 2017 The Linnean Society of London, Biological Journal of the Linnean Society, 2017, 121, 294–304

GENETIC DIFFERENTIATION IN L. DUBIA 

301

reveal some genetic differentiation between these sampled localities. On the other hand, L. dubia do disperse from their native waters (Pajunen, 1962), and dragonflies (Anisoptera) in general are strong flyers, which are reflected in low genetic differentiation in several studies (Freeland et al., 2003, Damm & Hadrys, 2012; Troast et al., 2016), but see Keller et al. (2010), for example in the endangered dragonfly Leucorrhinia caudalis. Hence, despite the scattered distribution across many countries in central Europe, there seems to be a quite strong gene flow among these localities, suggesting that given enough suitable habitats the species should be able to maintain stable populations in these areas.

Asia

Figure 4.  Isolation by distance plots (IBD) for SNP data. FST as a function of distance (km). Upper plot: IBD for 104 individuals of L. dubia and L. intermedia using 925 SNPs. Middle plot: IBD for 89 individuals of all L. dubia using 1315 SNPs. Lower plot: IBD for 78 individuals of European L. dubia using 1674 SNPs. Error bars indicate ±SE obtained by permutations of each spatial group.

Not surprisingly the STRUCTURE and the FST analysis, at K = 3, showed that the L. intermedia species was distinct from the rest of the localities which belong to a different species, L. dubia. In addition, L. intermedia is also morphologically distinct from L. dubia (Belyshev, 1973). Further, a phylogeny based on nuclear rDNA also found a clear distinction between L. dubia and L. intermedia (Hovmöller & Johansson, 2004). Interestingly, at K = 2, the Asian/Japanese L. dubia orientalis did cluster together with L. intermedia and not with the European L. dubia. This suggests that, at least for the loci used in this study, European L. dubia are distinct from both L. orientalis and intermedia in the east. This may be explained by some past and/or recent gene flow between L. intermedia and L. dubia orientalis. This would suggest incomplete speciation among L. dubia and L. intermedia. But we note that our sample size was based on three individuals of L. dubia orientalis in Japan and therefore more data are needed to confirm our suggestion. Since dubia and L. intermedia are considered separate species in the east (Belyshev, 1973), our analyses suggest western L. dubia is a distinct species separated from both the eastern taxa. This also suggests that L. dubia orientalis and L. dubia from Europe could be considered as distinct species, despite of the opinion by Kosterin & Zaika (2010), who treated them as subspecies. Further sampling across the Palearctic would be needed to resolve this issue. Leucorrhinia dubia orientalis from Blagovesghchensk in Russia clustered together with L. dubia orientalis from Tuva in Russia, rather than together with L. dubia orientalis from Japan. The individuals from the Japanese L. dubia orientalis locality came out as a separate cluster at K = 5. Our analysis is based on three localities only, and a more thorough material covering the eastern distribution is needed to resolve the status of the taxon L. dubia orientalis.

© 2017 The Linnean Society of London, Biological Journal of the Linnean Society, 2017, 121, 294–304

302  F. JOHANSSON ET AL. Conservation management recommendations In order to maintain healthy populations of L. dubia in Southern and Central Europe, it is important to preserve areas with bog ponds and also create new such habitats since this species is dependent on such environments. The absence of an IBD pattern suggests that given enough suitable habitats available within a reasonable distance, L. dubia should be able to colonize such environments. Furthermore, it is important to preserve the current habitats in United Kingdom, the Pyrenees and the Alps, because these population are somewhat genetically distinct from other populations in Europe at the level of neutral markers and might thus also harbour important quantitative traits that are important for local adaptation to these environments and act as an important source for adaptation to changing environmental conditions. Finally, there is an urgent need to study the French populations in the Massif Central and other localities in UK that we did not include in this study, to better understand their value in terms of conservation.

ACKNOWLEDGEMENTS We thank Tim Beynon, Pawel Bucynski, Merja Elo, Elena Malikova, Oleg Kosterin, Robby Stoks, Hidenori Ubukata, Hansruedi Wildermuth for collecting samples. Thanks to Alicia Mastretta-Yanes for sharing a version of her ddRAD protocol prior to publication. Thanks also to Oleg Kosterin and two anonymous reviewers for valuable comments that improved this article.

SHARED DATA Data is available from the Dryad Digital Repository. http://dx.doi.org/10.5061/dryad.c9j53 (Johansson et al., 2016).

REFERENCES Askew RR. 2004. The Dragonflies of Europe. Colchester: Harley Books (B. H. & A. Harley Ltd.), 291 pp. Belyshev BF. 1973. Dragonflies (Odonata) of Siberia, Vol. 1, Part 1. Novosibirsk: Nauka [in Russian]. Bernard R, Heiser M, Hochkirch A, Schmidt T. 2011. Genetic homogenity of hedgling Nehalennia speciosa (Odonata: Coenagrionidae) indicates a single Würm glacial refugium and trans-Palearctic postglacial expansion. Journal of Zoological Systematics and Evolutionary Research 49: 292–297. Bhagwat SA, Willis KJ. 2008. Species persistence in northerly glacial refugia of Europe: a matter of chance or biogeographical traits? Journal of Biogeography 35: 464–482. Billqvist M, Smallshire D, Swash A. 2012. Svenska Trollsländeguiden. Värnamo: Elanders, Fählt & Hässler.

Boudot J-P, Kalkman VJ. (eds.) 2015. Atlas of European dragonflies and damselflies. The Netherlands: KNNV Publishing. Catchen J, Hohenlohe PA, Bassham S, Amores A, Cresko WA. 2013. Stacks: an analysis tool set for population genomics. Molecular Ecology 22: 3124–3140. Daguet CA, French GC, Taylor P. 2008. The Odonata red data list for Great Britain. Species status, Vol. 11. Peterborough: Joint Nature Conservation Committee, 1–34. Damm S, Hadrys H. 2012. A dragonfly in the desert: genetic pathways of the widespread Trithemis arteriosa (Odonata: Libellulidae) suggest male-biased dispersal. Organism Diversity and Evolution 12: 267–279. De Knijf G. 2006. The red list of the dragonflies of Flanders. In: De Knijf G, Anselin A, Goffart P, Tailly M, eds. The Dragonflies of Belgium: distribution – evolution – habitats. Dragonfly Society of Belgium Gomphus & Research Institute for Nature and Forest, Omslagtitel: De libellen van België, 241–257. ISBN: 90-403-0249Evanno G, Regnaut S, Goudet J. 2005. Detecting the number of clusters of individuals using the software STRUCTURE: a simulation study. Molecular Ecology 14: 2611–2620. Forsman A. 2014. Effects of genotypic and phenotypic variation on establishment are important for conservation, invasion, and infection biology. Proceedings of the National Academy of Sciences of the United States of America 11: 302–307. Francis RM. 2017. POPHELPER: an R package and web app to analyse and visualise population structure. Molecular Ecology Resources,17: 27–32. Freeland JR, May M, Lodge R, Conrad KF. 2003. Genetic diversity and widespread haplotypes in a migratory dragonfly, the common green darner Anax junius. Ecological Entomology 28: 413–421. Gentili R, Bacchetta G, Fenu G, Cogoni D, Abeli T, Rossi G, Salvatore MC, Baroni C, Citterio S. 2015. From cold to warmstage refugia for boreo-alpine plants in southern European and Mediterranean mountains: the last chance to survive or an opportunity for speciation? Biodiversity 16: 247–261. Gonseth Y, Monnerat C. 2002. Rote Liste der gefährdeten Libellen der Schweiz. Hrsg. Bundesamt für Umwelt, Wald und Landschaft, Bern, und Schweizer Zentrum für die Kartografie der Fauna, Neuenburg. BUWAL-Reihe Vollzug Umwelt. 46 S. Hájek M, Hájková P, Apostolova I, Horsák M, Plásĕk V, Shaw B, Lazarova M. 2009. Disjunct occurrences of plant species in the refugial mires of Bulgaria. Folia Geobotanica 44: 365–386. Hardy O, Vekemans X. 2002. SPAGeDi: a versatile computer program to analyse spatial genetic structure at the individual or population levels. Molecular Ecology Notes 2: 618–620. Hewitt GM. 1999. Post-glacial re-colonization of European biota. Biological Journal of the Linnean Society 68: 87–112. Hewitt G. 2000. The genetic legacy of the Quaternary ice ages. Nature 405: 907–913. Hovmöller R, Johansson F. 2004. A phylogenetic perspective on larval spine morphology in Leucorrhinia (Odonata: Libellulidae) based on ITS1, 5.8S, and ITS2 rDNA sequences. Molecular Phylogenetics and Evolution 30: 653–662.

© 2017 The Linnean Society of London, Biological Journal of the Linnean Society, 2017, 121, 294–304

GENETIC DIFFERENTIATION IN L. DUBIA  Höglund J. 2009. Evolutionary conservation genetics. Oxford: Oxford University Press, 189. Jakobsson M, Rosenberg NA. 2007. CLUMPP: a cluster matching and permutation program for dealing with label switching and multimodality in analysis of population structure. Bioinformatics 23: 1801–1806. Johansson F, Halvarsson P, Mikolajewski DJ, Höglund J. 2016. Data from: Genetic differentiation in the boreal dragonfly Leucorrhinia dubia in the Palearctic region. Jombart T. 2008. adegenet: a R package for the multivariate analysis of genetic markers. Bioinformatics 24: 1403–1405. Jombart T, Ahmed I. 2011. adegenet 1.3-1: new tools for the analysis of genome-wide SNP data. Bioinformatics 27: 3070–3071. Keller D, Brodbeck S, Flöss I, Vonwil G, Holderegger R. 2010. Ecological and genetic measurements of dispersal in a threatened dragonfly. Biological Conservation 143: 2658–2663. Klinga P, Mikolas M, Zhelev P, Höglund J, Paule L. 2015. Genetic differentiation of western capercaillie in the Carpathian mountains: the importance of post glacial expansions and habitat connectivity. Biological Journal of the Linnean Society 116: 873–889. Kosterin OE, Zaika VV. 2010. Odonata of Tuva Russia. International Journal of Odonatology 13: 277–327. Lampa S, Dahlö M, Olason PI, Hagberg J, Spjuth O. 2013. Lessons learned from implementing a national infrastructure in Sweden for storage and analysis of next-generation sequencing data. GigaScience 2: 9. Mastretta-Yanes A, Arrigo N, Alvarez N, Jorgensen TH, Piñero D, Emerson BC. 2015. Restriction site-associated DNA sequencing, genotyping error estimation and de novo assembly optimization for population genetic inference. Molecular Ecology Resources 15: 28–41. Ott J, Conze K.-J, Günther A, Lohr M, Mauersberger R, Roland H-J, Suhling F. 2015. Erschienen in Libellula, Rote Liste der Libellen Deutschlands, Supplement 14, Atlas der Libellen Deutschlands, GdO e.V. 2015, ISSN 0723-6514. Pajunen VI. 1962. Studies on the population ecology of Leucorrhinia dubia V. d. Lind. (Odon., Libellulidae). Annales Zoologici Societatis Zoologicae-botanicae Fennicae ‘Vanamo’ 24: 1–79. Parchman TL, Gompert Z, Mudge J, Schilkey FD, Benkman CW, Buerkle CA. 2012. Genome-wide association genetics of an adaptive trait in lodgepole pine. Molecular Ecology 21: 2991–3005.

303

Paxton RJ, Thorén PA, Tengö J, Estoup A, Pamilo P. 1996. Mating structure and nestmate relatedness in a communal bee, Andrena jacobi (Hymenoptera, Andrenidae), using microsatellites. Molecular Ecology 5: 511–519. Peterson BK, Weber JN, Kay EH, Fisher HS, Hoekstra HE. 2012. Double digest RADseq: an inexpensive method for de novo SNP discovery and genotyping in model and nonmodel species. PLoS One 7: e37135. Pritchard JK, Stephens M, Donnelly P. 2000. Inference of population structure using multilocus genotype data. Genetics 155: 945–959. Rogell B, Thörngren H, Laurila A, Höglund J. 2010. Genetic structure in peripheral populations of the natterjack toad, Bufo calamita, as revealed by AFLP. Conservation Genetics 11: 173–181. Samways MJ. 1994. Insect conservation biology. London: Chapman and Hall, 358 pp. Segelbacher G, Höglund J, Storch I. 2003. From connectivity to isolation: genetic consequences of population fragmentation in capercaillie across Europe. Molecular Ecology 12: 1773–1780. Segelbacher G, Strand T, Quintela M, Axelsson T, Jansman HAH, Koelewijn P, Höglund J. 2014. Historical and current populations of black grouse in Central Europe – evidence for rapid loss of genetic diversity. Conservation Genetics 15: 1183–1195. Spitzer K, Danks HV. 2006. Insect biodiversity of boreal peat bogs. Annual Review of Entomology 51: 137–161. Stewart JR, Lister AM, Barnes I, Dalén L. 2010. Refugia revisited: individualistic responses of species in space and time. Proceedings of the Royal Society B: Biological Sciences 277: 661–671. Storch I. 2000. Grouse status survey and conservation action plan 2000–2004. WPA/Bird Life/SSC Grouse specialist group, IUCN, Gland, Switzerland/World Pheasant Association, Reading, UK. Svennig J-C, Normand S, Kageyama, M. 2008. Glacial refugia of temperate trees in Europe: insights from species distribution modelling. Journal of Ecology 96: 1117–1127. Troast D, Suhling F, Jinguji H, Sahlén G, Ware J. 2016. A global population genetic study of Pantala flavescens. PloS One 11: e0148949. Wallenius T, Niskanen L, Virtanen T, Hottola J, Brumelis G, Angervuori A, Julkunen J, Pihlström M. 2010. Loss of habitats, naturalness and species diversity in Eurasian forest landscapes. Ecolological Indicators 10: 1093–1101. Wright S. 1943. Isolation by distance. Genetics 28: 114–138.

SUPPORTING INFORMATION Additional Supporting Information may be found in the online version of this article at the publisher’s website: Table S1. Name and coordinates of locations sampled. N denotes number of individuals sampled. The samples from Poland is a pooled sample from four localities treated as one location (54°02’45’’ N 17°52’45’’E; 53°54.37 N, 16°41.65’E; 54°02’18 N, 17°51’03 E; 54°23’14’’ N 17°58’00’’E).

© 2017 The Linnean Society of London, Biological Journal of the Linnean Society, 2017, 121, 294–304

304  F. JOHANSSON ET AL. Figure S1. Evanno method plots for different combinations of data. In each panel different estimates of the likelihood of the number of clusters is given in panels A, B and C (A=mean L(K), B=L’(K) and C=L’’(K)) and D plots the difference in likelihood (K) given K ( see Evanno et al. 2005 for a detailed explanation). Figure S2. Bayesian clustering results. Individual assignment of individuals using program STRUCTURE. Each plot from aligning 10 STRUCTURE runs using CLUMPP for K=2 to K=7. Each vertical bar represent one individual and the color the probability for each individual to belong to a certain population. Left column: 1315 SNPs from 98 individuals of L. dubia and L. dubia orientalis. Right column: European population of L. dubia, excluding United Kingdom 1674 SNPs from 65 individuals.

© 2017 The Linnean Society of London, Biological Journal of the Linnean Society, 2017, 121, 294–304