Extremely low genetic diversity in a circumpolar dragonfly species, Somatochlora sahlbergi (Insecta: Odonata: Anisoptera)

We present the first empirical treatment of the northernmost breeding dragonfly, Somatochlora sahlbergi. We sequenced populations from United States, Canada, Finland, Sweden and Norway for cytochrome oxidase I (COI) and D2 region of 28s. We found that, despite geographic barriers across its vast arctic range, S. sahlbergi is a single species. Not only does it appear to interbreed across its entire range, there also seems to be almost no variation among European and North American populations in their COI gene fragment (the barcode gene), which is usually extremely variable. We further found that characters thought to be diagnostic for the larvae of S. sahlbergi were absent in our European samples. We review and re-describe the habitat of this species based on new findings from recent field observations. Finally, we report for the first time the likely presence of this species in Japan. We hope our findings will encourage further study of this species and other under-studied insect taxa that inhabit the remote Arctic.

S. sahlbergi is usually associated with peatlands in permafrost (specifically in Canada, Finland, Sweden and Norway), which have low mean annual temperatures and precipitation 3,7,8,11 . Adults fly around these peatlands where their larvae live in cold, deep 3 and flowing water 10 (pers. obs. MKK and WRK). In most of its North American and European range, S. sahlbergi occurs at the latitudinal treeline 3 , north of which trees are unable to grow due to low moisture and extreme temperature. Schröter 8 suggested that S. sahlbergi is closely associated with palsa mires in the region of Fennoscandia (i.e., the Scandinavian Peninsula, Finland, Karelia, and the Kola Peninsula). In Canada and Alaska, larvae are found in deep, cold ponds surrounded by sedges and aquatic mosses and underlain with permafrost soil 7 . These locations are also characterized by annual mean temperatures of less than 0 °C and mean annual rainfall <450 mm 7 .
Although it was described 120 years ago, S. sahlbergi is still relatively rare in collections 3,[12][13][14][15][16] and information about its evolutionary history and ecology is poor 8 . It is a fascinating species with a broad and perplexing distribution in some of the planet's most extreme habitats. In this study, we explore the population structure of this species across its range. We believe that because of its geography S. sahlbergi might be isolated into Old World and New World clades without gene flow instead of a single panmictic population. We test the hypothesis that there is distinct genetic structure among populations in North America and Eurasia. In addition, we discuss the utility of lateral spines for taxonomic identification of northern congeneric Somatochlora and speculate how the natural history of S. sahlbergi may have been influenced by Quaternary glaciation, as is the case with many other northern species.

Methods
Taxon Sampling. Specimens were collected in Scandinavia and Finland in September 2014 (Sweden, Norway, Finland, by JLW and GS) and July 2015 (Finland, by Magnus Billqvist). Samples from the Yukon Territory in Canada were collected in August 2015 (by MKK and WRK). In Scandinavia and Finland, we visited known S. sahlbergi sites, and also explored potential habitat (see Fig. 1). During the Yukon Territory sampling, we visited all the previously known localities 7 (see Fig. 1). At each location, we sampled both larvae and adults and collected 28 samples in total that we identified as S. sahlbergi from Europe and Canada. Four samples were adults from the Yukon and the rest were larvae from Europe. To supplement our North American sampling, we borrowed 6 North American specimens from the Beaty Biodiversity Museum, University of British Columbia, Vancouver, bringing our specimen count to 34. Besides S. sahlbergi, we also collected Somatochlora albicincta (Burmeister), which were included in our analysis. See Supp. Table S2 for further information on the samples. DNA isolation, gene selection and PCR amplifications. DNA was extracted from legs of the 34 S. sahlbergi samples using a Qiagen Blood and Tissue Extraction Kit (Qiagen Sciences, Germantown, MD). We followed the protocol specified by the manufacturer but extended the incubation time to 24 h. Each sample was amplified for the mitochondrial gene cytochrome oxidase I (COI) and a fragment of the nuclear ribosomal 28 S subunit, D2. We used a COI primer pair (Supp . Table S3) which amplifies approximately 516 bp fragment found 214 bp  downstream of 5′ region of COI gene. COI, universally regarded as the barcode gene, is a fast-evolving gene that has been repeatedly demonstrated to vary among individuals in a population (see [17][18][19][20]. Therefore, this gene fragment should provide enough resolution to discern the genetic structure among the populations. The nuclear gene D2 was chosen to provide an independent dataset to compare to the mitochondrial COI. It contains five hypervariable stem and loop regions, which have been shown to be useful for distinguishing inter-and intrageneric species relationships 21 . The amplified PCR product was purified and sequenced by Macrogen (Macrogen USA, Rockville, MD). We successfully amplified 28 sequences for COI and 25 for D2. All sequences were deposited in the NCBI GenBank database (see Supp. Table S2 for accession numbers).
Sequence alignment. Initial alignments were performed in ClustalX 22 and then manually checked for incongruencies in Mesquite 23 . The D2 region was largely invariant among the species in our dataset, hence secondary structural alignment was not necessary. We created two data matrices: one for each of the genes. The final COI matrix consisted of 28 samples sequenced de novo and an additional 20 samples amplified at the Environmental Protection Agency (EPA) in Cincinnati, OH (E. Pilgrim, unpublished) bringing our S. sahlbergi total to 48 (Supp . Table S2). To determine whether S. sahlbergi is a monophyletic species, we included sequences from other Somatochlora species, S. alpestris  . Table S2). These sequences were downloaded from GenBank. Helocordulia uhleri (Selys) and Cordulia amurensis Selys were used as more distant outgroups.
Our final D2 matrix included a total of 42 individuals, of which 25 were S. sahlbergi and the rest were S. albicincta (Supp . Table S2). This table includes only those specimens that were successfully sequenced (at the Ware Lab) for both genes.
To compare S. sahlbergi to other northern odonate species, we downloaded nucleotide sequences from the Barcode of Life Data System 24 for dragonflies with similar ranges to S. sahlbergi in North America and Europe. We included 10 species in addition to S. albicincta and S. sahlbergi (Supp. Phylogenetic and population analysis. We conducted a phylogenetic analysis to check the validity of S. sahlbergi as a monophyletic species within the genus Somatochlora. We based our phylogenetic reconstruction of the genus Somatochlora on COI for which we had a more exhaustive sequence sampling compared to D2. We also reconstructed the relationships among populations of S. sahlbergi using a concatenated matrix of COI and D2 data. Phylogenetic analyses for the datasets (COI and COI + D2) were done using maximum likelihood inference via iq-tree 25 . For the COI + D2 dataset, we partitioned the dataset by gene fragment. Substitution models for both the data sets were determined in iq-tree prior to tree reconstruction 26 . Bootstrap values for reconstructed trees from the two datasets were obtained using the ultrafast bootstrap approach in iq-tree, with 1000 replications 27 . The iq-tree analysis was run on the iq-tree web server 28 .
Haplotype analysis of the COI data for all species were performed in popart 29 using the minimum-spanning network algorithm 30 . Because our preliminary tree reconstruction suggested that 3 Somatochlora alpestris sequences taken from GenBank were S. sahlbergi, we included these specimens also in our haplotype reconstruction of the latter species, giving a total of 51 samples.
We used mega (v7.0.14) 31 for a comparative analysis of intraspecific genetic diversity among the 12 northern odonate species, including S. sahlbergi. We estimated the percentage of intraspecific diversity (referred to as PID here after) for each of the species using both Jukes-Cantor (JC) and Kimura-2-parameter (K2P) nucleotide evolution models.
Larval morphology. Larvae were inspected and identified under a Nikon SMZ-745T zoom stereo dissecting microscope. We counted the number of S. sahlbergi specimens with and without lateral spines on abdominal segments 8 and 9.

Results
Phylogenetic reconstruction. To test the monophyly of S. sahlbergi we reconstructed a phylogenetic tree from COI dataset, comprising a total of 19 Somatochlora species, shown in Fig. 2. Helocordulia uhleri and Cordulia amurensis were used as outgroups. Somatochlora graeseri + S. uchidai form a clade (99% bootstrap support) that is sister to the remaining Somatochlora. At the next most apical split, S. sahlbergi is supported as a valid species with a bootstrap value of 87%. It is recovered as sister to the remaining species in Somatochlora, which form several multi-species clades.
Somatochlora albicincta, S. whitehousei, S. septrentionalis, and S. hudsonica are recovered in one clade with a bootstrap support of 99%. Somatochlora minor, S. elongata and S. williamsoni are recovered as a monophyletic clade with strong support (88%), as are S. metallica and S. exuberata (82% bootstrap support). S. franklini is also recovered as a monophyletic group with 99% support.
Interestingly, three NCBI samples, identified by accession numbers AB708910.1, AB708909.1 and AB708908.1, are labelled as "S. alpestris" on GenBank, yet were consistently recovered within the monophyletic S. sahlbergi clade (Figs 2 and 3) in our analysis. These sequences originated from Karube et al. 32 , who published a total of five sequences from Hokkaido, Japan, each identified as S. alpestris. We included all 5 sequences 32 in our tree reconstruction: three were recovered within the S. sahlbergi clade while the other two were recovered as a separate clade with 100% bootstrap support (Fig. 2).
For the CO1 + D2 dataset, we included S. sahlbergi and S. albicincta sequenced de novo from the Yukon and Europe. Helocordulia uhleri was used as the outgroup in this analysis. Somatochlora albicincta was recovered as a valid species with 86% bootstrap support, while S. sahlbergi was recovered as a valid species with 74% bootstrap support (Supp. Fig. S1). Within S. sahlbergi, samples were recovered in a polytomy with short branch lengths, suggesting that these sequences have only few nucleotide differences between them. We conclude from our phylogenetic analysis that S. sahlbergi, sampled across its Holarctic range, comprises a single species from its well-supported monophyly in the tree; however, we hesitate to draw further conclusions regarding the phylogenetic relationships among other members of the genus Somatochlora. We recommend additional gene and taxon sampling among Somatochlora before drawing such conclusions.
Larval morphology. All but three S. sahlbergi larvae lacked lateral spines on segments 8 and 9 (with: N = 3, without: N = 15). This lack of spines did not appear to be associated with larval stadium.

Discussion
Here, we present the first comprehensive phylogenetic reconstruction of Somatochlora sahlbergi and double its known localities from approximately 71 8 to around 140 (Supp . Table S1). Our results support three main conclusions regarding S. sahlbergi: (1) it is a single species across its entire geographic range, with low levels of intraspecific genetic diversity; (2) the presence/absence of lateral spines is not a reliable character for distinguishing its larvae from those of other Somatochlora species; and (3) its range seems to include Japan.

Somatochlora sahlbergi is a single species across its geographic range. Despite populations in
North America and Europe being divided by oceans, our results suggest that S. sahlbergi populations may be panmictic across their Holarctic range. We recovered four different haplotypes among all the populations of S. sahlbergi, but most of the individuals, sampled from different continents, shared one single haplotype (Fig. 3). The additional haplotypes (H 1 , H 2 and H 3 ) differ from the main one (H 0 ) by a single nucleotide difference each, but this does not indicate any significant difference within the populations because a one-nucleotide difference can be captured in a sequence just by chance (e.g., due to a sequencing error). Additionally, in comparison to other northern odonates, S. sahlbergi shows the lowest amount of intraspecific divergence (PID = 0.1%) in the COI. The only other species to show intraspecific sequence divergence of 0.1% is S. albicincta; however, this value can be expected in S. albicincta because all the samples used in the analysis were collected from a small geographic location. In fact, most of the northern species, despite having smaller geographic ranges, show more intraspecific variation across their populations in comparison to S. sahlbergi (Supp . Table S5).
Along with low intraspecific diversity, S. sahlbergi also shows low haplotypic diversity. Figure 4 shows haplotype networks for some northern species in comparison to S. sahlbergi. Aeshna juncea, for example, a circumboreal dragonfly that overlaps with S. sahlbergi in its northern range and exhibits a large haplotypic diversity, indicating much higher genetic variation than that seen in S. sahlbergi (sample size of A. juncea: 57; sample size of S. sahlbergi: 51). Similarly, Libellula quadrimaculata, which is found in both Europe and North America, also shows high haplotype diversity with a complex network connecting these haplotypes. We found that the D2 region of ribosomal 28S, known for being highly variable due to the presence of loops in its secondary structure, was extremely invariable in S. sahlbergi. Below we discuss four hypotheses that either alone or in a combination may explain the low genetic variation seen in the COI and D2 genes in S. sahlbergi.
First, gene flow might be prevalent among the North American and European populations. It would follow that individuals of S. sahlbergi are moving (likely through flight) between North America and Eurasia and mating. There are two possible directions of this movement: individuals may travel from western North America, through eastern North America (where they do not appear to live), and into Europe and Russia over the Atlantic Ocean; or they may fly the shorter distance over the Bering Sea (See Fig. 1). In either case, these insects would be dispersing for thousands of kilometers. Anax junius Drury 33 and A. ephippiger (Burmeister) 34 have been recorded travelling large distances and crossing vast water-bodies; however, the longest transoceanic migration ever recorded for a dragonfly (and any insect for that matter) is accomplished by Pantala flavescens (Fabricius). This dragonfly is capable of migrating a remarkable distance of 3,500 km across Indian Ocean 35,36 . Troast et al. 17 concluded that P. flavescens is a panmictic species, with gene flow among its various populations across the globe. S. sahlbergi shows a similar trend, with little genetic variation across its global populations. However, there is a difference in the possible mode of dispersion for these dragonflies: P. flavescens is a passive flier, dependent on wind systems like intercontinental trade winds to carry it across oceans 17 , whereas S. sahlbergi, with its comparatively narrow wings would have to rely more on active flight, engaging in more flapping flight than gliding to stay airborne. There are examples of active flying, migratory dragonfly species (e.g., A. junius 37 ); however, none of them are known to fly as far as P. flavescens, or over the distances that S. sahlbergi individuals would need to fly to maintain gene flow between populations. Based on a few observations of patrolling behaviour exhibited by males, S. sahlbergi are considered to be "flyers", which spend the majority of their active time flying rather than perching 3,14,38 . Yet, due to the elusive nature of this species, there are no empirical records of the distance they are capable of travelling. Radio-transmitters have been successfully used to track dispersal distance and direction of A. junius 39 , but the smaller S. sahlbergi may not be able to support the weight of these devices. Alternatively, an isotope analysis 40 of wings may help determine whether individuals are migrating across the Northern Atlantic Ocean or Bering Sea. To test this hypothesis, we are now extending our sampling to the Russian population.
Second, the time since divergence between the populations may be relatively short, indicating that populations S. sahlbergi were connected until relatively recently. Assuming this species had the same geographic range in its past as it does now, then during the last glaciation these populations would have been connected to each other through Beringia. The Beringian land bridge, which spanned the modern-day Bering Strait, disappeared relatively recently when the sea level rose around 11,500 cal BP 41 . Perhaps members of S. sahlbergi were able to maintain migration along that route for a while after the bridge disappeared. This hypothesis can be tested by estimating the divergence time between populations using coalescence models 42,43 . However, results from coalescence analysis using our current data set will not be reliable since the lack of genetic variation in our DNA sequences will lead to recovering artificially low substitution rates.
Third, long generation times in S. sahlbergi may explain their lack of genetic diversity. Molecular evolution rates in species are thought to be correlated with generation time: the smaller the generation time, the faster the rate of molecular evolution. This correlation has been found in several groups of organisms, including invertebrates 44 . In temperate regions, odonate life cycles become increasingly long with increasing latitude as larval growth can be arrested by a reduction in photoperiod and temperature 6 . Individuals of S. sahlbergi survive in places where the water is frozen for at least eight months out of the year (observations by GS). Therefore, perhaps this species takes several years to complete its life cycle. Ulf Norling studied life cycle length in two northern odonate species, Leucorrhinia dubia (Vander Linden) and Coenagrion hastulatum Charpentier, at 67°N. He found that L. dubia had an approximately 4-year development period 45 while C. hastulatum took 3-4 years to develop 46 . Based on these observations, it is reasonable to assume that the very short summers in the north add at least one extra year to the life cycle for S. sahlbergi, suggesting 4-5 years of total larval development time, which is supported by analyses of various larval instars by Cannings and Cannings 3 . Thus, a relatively long generation time, combined with a relatively short time since divergence (as discussed above) might have led to very low levels of genetic variation in S. sahlbergi.
Fourth and finally, S. sahlbergi may have suffered a genetic bottleneck at some point in its evolutionary history. Genetic bottlenecks, resulting in a dramatic decrease in genetic diversity, can be caused by events such as the introduction of new pathogens or predators, or habitat loss (perhaps caused by a change in climate). Due to their proximity to glacial ice sheets, populations of S. sahlbergi have likely experienced many climatic fluctuations in their evolutionary past. During the last glacial termination (17.7-11.5 kBP), the Northern hemisphere saw periods of cold/warm oscillations as the glaciers were receding 47 . This was marked by an intense period (Younger Dryas: 12.9-11.7 kBP) marked by a cold and dry climate 48 . The relatively stable Holocene climate has also seen several intervals of rapid climate change (e.g., the Little Ice Age and Medieval Warm Period) that affected the Northern Hemisphere 49 . These climatic fluctuations might have affected some populations of S. sahlbergi more severely than other, perhaps more southern populations, especially in North America and Europe, leading to their local extinction. Additionally, the Holocene fluctuations after the last glaciation may have made recolonization difficult for the few remaining individuals.
Larval morphology. Lateral spines. Larvae of S. sahlbergi are described as having dense and fairly-long setae covering their abdomens, with prominent lateral spines on segments 8 and 9 2 ; however, we found that these characteristics were not always present, especially in the case of lateral spines. Out of the total 24 larvae from northern Europe that genetically group as S. sahlbergi, only 3 had lateral spines. This is in contrast to samples from Norway examined previously by Norling and Sahlén 4 , as well as the description of the larva from Canada 3 where all larvae possessed lateral spines. Absence of these lateral spines in larvae makes S. sahlbergi difficult to distinguish from other sympatric Somatochlora species (S. albicincta in the Yukon and S. alpestris in Europe, which sometimes occupy similar habitats). Color. All the 24 larval specimen that were collected by GS and JLW in Europe were dark brown in color (Fig. 5), unlike the orange-brown suggested by Cannings and Cannings 3 . The underside of these larvae were greyish/ green-brown and we did not see any specimen with a yellow underside. We suspect that the reddish/orange color observed by both Cannings and Cannings 3 and Valle 50 might be an artefact of the preservation method rather than the original color of a freshly caught larvae. Mud or mineral deposits (for example iron) can cause larvae to appear a different color than the original. Japan: A new location for Somatochlora sahlbergi. S. sahlbergi was described from a sample collected in Siberia 51 . Since then, it has been recorded around the Arctic and subarctic regions (Fig. 1). Based on our results, we propose that this species may also be present in Japan. S. sahlbergi has not been recorded in Japan before, but has been found in the Kamchatka peninsula in Russia, which is ~2,700 km from Japan. Three samples that have been identified as S. alpestris on GenBank can either be a case of misidentification or may demonstrate hybridization between S. sahlbergi and S. alpestris. S. sahlbergi has been known to hybridize with other congeneric Somatochlora species like S. hudsonica and S. albicincta 3,7 . Therefore, it is important to further study the Somatochlora that are found in northern Japan to gain further insight into the history of this group and of S. sahlbergi.
Habitat notes. Based on previous literature, S. sahlbergi has been considered a habitat specialist 5,8 , often recorded in or near a rather specialized Arctic habitat called a "palsa mire"; however, our recent observations in northwestern Canada, Sweden, Norway and Finland and records from Kamchatka indicate that S. sahlbergi may not be restricted to palsa mires 7,52 . We suggest instead that this species inhabits a variety of habitat types.
In Sweden, Norway and Finland, larvae were taken from small to large bodies of slowly-seeping water, which were covered in mosses and directly connected to deep, cold springs (pers. obs. by GW and JLW). In northwestern Canada, S. sahlbergi has been associated with pools and small lakes that have "aquatic moss as the dominant vegetation and deep, cold water" 3 .
In August 2015, MKK and WRK surveyed 70 distinct sites in the Yukon and Northwest Territories along the Dempster Highway between the Ogilvie Mountains to the south and the Mackenzie River to the northeast; these included eight localities in the Yukon Territory where S. sahlbergi had previously been spotted or collected between 1980 and 2010 (SG Cannings, pers. comm.). S. sahlbergi was found in only two of these localities, and only one those two was previously recorded by Cannings. The first S. sahlbergi site (Supp. Table S1, site code CA150812-03) was a pond, surrounded by tall trees, shrubs and sedges, with moss on its bottom. The water was 39-54 cm deep and approximately 14 °C. A single S. sahlbergi larva was collected along with several aeshnid larvae. The second site (site code CA150815-05) was a pair of connected ponds, with a shallow region of sedges between them, surrounded by trees. Three male and one female S. sahlbergi adults were collected, primarily flying over this sedge area, as well as one S. albicincta larva. It is interesting, and perhaps alarming, to note that S. sahlbergi had apparently disappeared (or at least were undetected) from seven of Cannings' previous sites.
Given this evidence, we conclude that S. sahlbergi is perhaps not as much of a specialist as previously suggested 5,8 in the literature. It is important to come to a consensus on the habitat of this species because we may not have the opportunity to study this system for long, as S. sahlbergi inhabits the edge of a climatic zone that is at immediate risk from climate change. Observations from a recent expedition to the Yukon (by MKK and WRK) and recent studies in the Scandinavian region 5 suggest that S. sahlbergi might already be facing habitat loss in these locations because of temperature increase caused by global climate change 53,54 . Additionally, there is evidence that S. sahlbergi is facing competition from S. metallica in Eurasia, which is moving north due to global warming 5 . This study is a first step toward exploring the evolutionary and biogeographical history of S. sahlbergi, a single species that shows surprisingly low genetic diversity in its global distribution and is likely under stress from warming and competition.

Conclusion
S. sahlbergi is a single, panmictic species across the North American and European section of its range. Populations of S. sahlbergi show very little genetic variation in comparison to other dragonfly species from northern latitudes. It is presently unclear whether S. sahlbergi individuals are dispersing long distances and if so, whether they cross the ocean. We also conclude that lateral spines are not an identification character for S. sahlbergi compared to other congenerics. Future work should examine material from the Asian part of the geographic range of this remarkable Arctic dragonfly.

Data Availability Statement
Sequences used in the study are available on GenBank (see Supp. Table S2 for accession numbers). Alignments created during the analysis are available on request from the corresponding author.