Genetic diversity and differentiation of isolated rear‐edge populations of a cold adapted butterfly, Erebia aethiops, in Britain

Rear‐edge populations of cold‐adapted species are highly vulnerable to climate change, their adaptations maybe critical for the persistence of the species as a whole. Using AFLP, we investigated population genetic diversity and differentiation of Scotch argus (Erebia aethiops), a cold‐adapted butterfly, comparing isolated rear‐edge and core populations in Britain, a relict part of the species range. We also examined genetic differences among sub‐populations and dispersal distances conducting a mark‐release‐recapture (MRR) study at the rear‐edge. Genetic diversity was higher in the isolated rear‐edge populations than populations at the core of the range and declined with latitude, supporting the idea that current differences in genetic diversity levels across Britain are likely the result of historical range changes after the last glaciation. Populations were genetically differentiated among regions, meaning that losing the isolated rear‐edge populations may prove detrimental for the survival of the species in Britain, as these populations are likely to be better adapted to warmer climates. We found that the largest population at the rear‐edge is genetically robust, with gene flow among patches, likely maintained by males as indicated by the higher dispersal distances recorded for males in the MRR, but colonisation of empty patches is constrained by females' low mobility. Our results highlight that isolated populations at the rear‐edge of cold‐adapted species should be considered of high conservation priority, as they hold higher levels of genetic diversity and differentiation which may prove to be key for the survival of these species under global warming.


INTRODUCTION
Many insect species have experienced a dramatic decline, resulting from habitat loss and fragmentation (Sánchez-Bayo & Wyckhuys, 2019;Wagner, 2020). This has intensified in recent decades for cold adapted species, with population extinctions reported at the warm edge of species ranges, both latitudinal and elevational, in connection with climate change (Gottfried et al., 2012;Tayleur et al., 2016). Therefore, the combined effects of habitat loss, fragmentation and climate change are posing severe threats to cold adapted species and their conservation (Bellard et al., 2012;Turlure et al., 2010). Most cold adapted species in Europe show disjunct distribution today, with relict populations that have been isolated from the rest for thousands of years, commonly located at the periphery of the species range (Varga & Schmitt, 2008). These relict populations of former larger ranges when climatic conditions were more favourable , may provide clues to the adaptative potential of the species in the face of climate change. Relict parts of the range are also going to contain well connected populations in the core area and more isolated populations at the periphery, which may have been under different selection pressures in more recent times (Cassel-Lundhagen, 2010). Therefore, conservation of coldadapted species is going to require a good understanding of the genetic make-up of peripheral populations, but this information is still lacking for many species, as traditionally these populations have been considered of little value for conservation (Channell, 2004).
Genetic diversity within a species is generally regarded as being higher at the core of the species range and decreasing towards the range limits, the so-called Centre-periphery hypothesis (Eckert et al., 2008;Pironon et al., 2017). This is based on the idea that core populations are likely to be larger and better connected than edge populations. Long term isolation or restricted connectivity of edge populations can reduce or eliminate gene exchange and prevent the addition of new genetic material to the population, increasing the risk of inbreeding depression (Keyghobadi, 2007); a major cause of fitness reduction which is especially damaging to small populations (Frankham, 1995). Restricted connectivity and gene flow also make the population more susceptible to deleterious alleles persisting in the homozygous form (Zachos et al., 2007).
However, the opposite can also be seen where isolated populations are protected from purged deleterious alleles which would otherwise have been reintroduced via immigration (Gibson et al., 2009;Keller & Waller, 2002). Genetic diversity across a species' whole range is also strongly influenced by range dynamics driven by past climate (Hampe & Petit, 2005). In the northern hemisphere, current species ranges reflect past expansions from glaciation refugia that were located further south than their current range centres (Coope, 1994;Hewitt, 2000;Petit et al., 2003). This is particularly important for cold adapted species, in which populations at the current rear-edge represent, or are closer to, the past range centre, so more likely to hold higher genetic diversity (Hampe & Petit, 2005). Moreover, current core populations of northern species are likely to be even more adapted to colder climates than populations from the current warm edge of species range (Han et al., 2019;Mathiasen & Premoli, 2016). Populations at periphery may have undergone a process of adaptation due to environmental conditions differing substantially from those experienced by core populations (Nadeau & Urban, 2019). Isolation could lead to beneficial adaptations, when a small population may adapt more quickly to pressures such as climate change (Keller & Seehausen, 2012) and become fitter as a result; more able to survive in a changing environment (Channell, 2004). Moreover, peripheral populations tend to show higher levels of differentiation, so losing these populations could reduce the adaptative potential and the survival of the species as a whole (Channell, 2004). Therefore, a good understanding of the genetic diversity and differentiation of isolated rear-edge populations is crucial for the success of conservation strategies for cold-adapted species.
The Scotch argus butterfly, Erebia aethiopss (Esper, 1777), is one of these species showing a clear disjunct distribution. The species occurs across the western Palearctic from Scotland to western Siberia and the Altai Mountains but is absent from Scandinavia, and the populations in Britain represent a relict part of the range ( Figure 1). In Britain, Scotch argus shows a northerly distribution, being widespread in Scotland but restricted to only a few remaining populations in the northwest of England (Figure 1). The species has suffered decline and extinctions of populations at its warmer margin (Franco et al., 2006) and it has recently been classified as 'vulnerable' in the UK butterfly red list (Fox et al., 2022), as the species is considered at higher risk from future climate change . The few natural populations of the species in England are isolated populations of what was likely a more continuous range across Britain following the last ice age and a more fragmented range from the early to mid-20th century (Thomas & Lewington, 2011). Therefore, these populations have been isolated from the core area of the current distribution in Britain, located in Scotland, for several hundred years ( Figure 1). The characteristics of Scotch argus' range in Britain, with isolated populations at the rear-edge of species distribution, provide an excellent study system in which to investigate the conservation value of relict populations that are threatened by climate change. Here we used AFLP genetic markers to assess the genetic diversity (He) and differentiation (Fst) of populations across Britain, studying all the isolated rear-edge exiting populations and compering them with populations further north, representing the continuous core range in Britain. We specifically answer two main questions: (1) are the isolated rear-edge populations genetically depauperate? and (2) are they genetically differentiated from those within the continuous core range in Britain?
Finally, at the largest rear-edge population, we also collected individuals from separated patched (sub-populations) to assess gene flow among sub-populations and carry out a mark-release-recapture (MRR) field study to measure dispersal distances moved by individuals, to answer our last question: (3) is dispersal and gene flow within the rear-edge population enough to ensure long term persistence?

Study species
The Scotch argus butterfly occupies damp grasslands (acid and neutral), woodland clearings, and young plantations up to 500 m in elevation in Scotland, while in the northwest of England, occurs in a mosaic of sheltered limestone grasslands, scrubs and open woodlands (Asher et al., 2001). The butterfly prefers ungrazed or light grazing areas, where grasses are tall, and its caterpillar's food plants (mainly Molinea caerulea in Scotland and Sesleria caerulea in England) are abundant (Eeles, 2019). The species has a single generation a year, with adults flying from late July to early September and caterpillars overwintering in second instar (Eeles, 2019).

Sample collection
Specimens for genetic analysis were collected from 10 populations across the species range in Britain ( Figure 1 and Table S1). At the rear edge, we collected specimens from the only three remaining isolated natural populations in England and from one population resulted from an unauthorised release (Bastow Wood, Figure 1). In Scotland, representing the continuous core range of the species in Britain, specimens were collected from three populations in mainland Scotland and from three western Scottish islands (Figure 1). F I G U R E 1 Distribution of Erebia aethiops in Europe (inset) and across Britain in a 10 km resolution (white dots: 1970-2010 and black dots: 2010-2014). The collection sites of specimens used in the genetic analysis are indicated with stars.
We collected a total of 86 individuals ranging from 4 to 7 individuals per site, except for one of the isolated populations at the rearedge (Smardale Gill) for which a total of 42 individuals were collected in order to assess genetic structure at the local level (Table S1). The species is considered threatened in Britain (Fox et al., 2022) and most populations were in areas with some level of protection (nature reserves), which meant the small sample size was determined by permission restrictions. Although the sample size was small for all population (4-7 individuals) except for one (Table 1), we took several steps to reduce the impact of sample size on our results. We used AFLP which have been shown to be less affected by small sample sizes and in particular when the number of loci is high (Gorman & Renzi Jr, 1979;Nei, 1978), as is our case in which we identified 655 loci. For the population for which we collected more individuals we repeated all analyses using a random selection of 5 individuals to test the effect of sample size on our results. Finally, as the number of individuals was similar for all populations, hopefully, this makes genetic estimates more comparable.
For the genetic analysis at a local scale, specimens were collected from the largest isolated population at the rear-end, Smardale Gill, across the whole site, from six areas where the butterfly concentrated (henceforth referred to as patches). Patches correspond with those used for the mark-release-recapture study (see Figure S1 and specific details for each patch in Mark-release-recapture section).
Specimens were collected in late August (in 2015 for the Scottish populations and in 2017 for the English populations). In all populations, we collected old males in poor condition to avoid removing egg-carrying females and impacting the effective population. Individuals which were found already dead were collected regardless of sex and condition. In some populations there were not enough males for collection at the time of visiting the site and in those cases, a single leg was removed from females. Butterflies were captured with a net and euthanisation was via a quick pinch to the abdomen, following protocols recommended by Feinstein (2004). All specimens were then kept in a freezer until DNA extraction. AFLP products were sent to DBS Genomics laboratory in Durham (UK) for fragment analysis using an Applied Biosystems 3730 DNA Analyser with a DS-30 filter set and ROX500 size standard. All three products for each sample were multiplexed in a single well to achieve higher throughput. background noise may have generated false secondary peaks. The upper fluorescence threshold was set to the mid-upper limit of technical noise at 50 Relative Fluorescent Units (RFU), with the intention of filtering out missed false-presences during binning. A table of peak locations and size was created and converted to a presence/absence binary matrix using the RawGeno automated scoring R package (Arrigo et al., 2012).

Genetic analysis
Parameters were set to calculate 1%-99% quantiles of detected AFLP peaks and retain only individuals which fell within those bounds for all 3 primer combinations. Three samples were removed at this stage (one from Crosby Garret, one from Tomnavoulin and one from Mull) along with negative controls.
To allow very minor bp location differences to be disregarded and avoid over splitting peaks according to recommendations made by Holland et al. (2008), a maximal bin width of 2 bp was set. A minimal bin width of 1 bp was specified to avoid technical homoplasy (false assignment of multiple peaks from an individual into the same bin).
In total, 655 loci were retained in a binary matrix.
Genetic diversity estimates were calculated with AFLP-SURV (Vekemans, 2002) using the binary matrix of 655 loci. Assuming Hardy-Weinberg genotypic proportions, a Bayesian method with nonuniform prior distribution of allele frequencies was selected to calculate allelic distribution for each population separately (Zhivotovsky, 1999).
This generated estimates for: Proportion of polymorphic loci at the 5% The relationship between genetic distance and geographic distance was tested with a Mantel test using GenAlEx, excluding the re-introduced population at the rear-edge (Bastow Wood) from this analysis, though we calculated genetic differentiation (Fst) between this population and the others in order to shed light on its potential origin.
To visualise overall genetic structure among the E. aethiops study populations, a principal component analysis (PCA), based on the 655 loci presence/absence binary matrix, was conducted using the FactorMiner (Lê et al., 2008) and factoextra packages in R (Kassambara & Mundt, 2020).
For the genetic analysis within the Smardale Gill site, individuals were collected from different patches that were considered subpopulations. We calculated the proportion of polymorphic loci at the 5% level, expected heterozygosity (He) and Wright's fixation index (Fst) with 500 permutations for each patch. Pairwise Fst estimates between patches were used to infer interbreeding history, gene flow and to test for significant differentiation among sub-populations. The relationship between genetic distance and geographic distance for all sub-populations was tested with a Mantel test, with a PCA used to visualise overall genetic structure among patches as described above.

Mark-release-recapture study
A mark-release-recapture study was carried out at the Smardale Gill site to assess the dispersal ability of the species and to estimate population size. The study was carried out from 25th of July to 6th of September 2017 with a total of 20 days surveyed. Scotch argus is active during warm, sunny periods, so the site was visited every day with favourable weather conditions. Two recorders visited the site following a standardised route ensuring that all suitable habitat patches were surveyed, and the entire area was covered every 2 days. Habitat patches were defined as areas with abundant host plant (Sesleria caerulea) and separated by other habitats including grassy areas with no host plant, encroached scrub and woodland areas. We identified a total of nine separate patches ( Figure S1), with a maximum distance of 3.5 km between the two most separated patches, which determined the maximum movements that could be recorded.
Butterflies were marked on the underside of the second pair of wings by writing a number, using an indelible fine-line marker pen.
Tens and singles digits were marked on the left wing while hundreds and thousands were marked on the right wing. For every butterfly captured we recorded: individual number, sex, GPS location, time and date.
An estimate of population size for the whole area was generated using the RCapture package in R (Baillargeon & Rivest, 2007

Genetic diversity and differentiation of populations across Britain
We genotype a total of 86 individuals across the 10 populations (   (Table S3).
Results remained the same when groups of five randomly selected individuals were included in the analysis for the largest isolated rearedge population (Table S4).  Table S6). The patches at the centre of the site (patch 3 and 4) showed higher He values than patches at both ends of the site (patch 1 and 2 to the north and patch 8 to the south Table S6, Figure S1).
Pairwise Fst values between patches showed that the most genetically differentiated patches were patch 1, 2 and 3 and the less differentiated was patch 4 (Table S5). The overall observed Fst value was 0.044 (SE = 0.002), which is higher than the upper 95% confident limits (0.014 Fst value under the null hypothesis of no differentiation, p < 0.001), indicatingsignificantgeneticdifferentiationamongpatches.
The first two PCA components explained 10% and 6.9% of the variation in genetic composition among patches, respectively. Patch F I G U R E 2 Expected Heterozygosity (He) in populations of Erebia aethiops in different parts within the species range in Britain. Boxplots displaying the median, the first and third quartile and the maximum and minimum values.
3 at the centre of the site showed lower overlap with those patches situated at both ends of the nature reserve (patch 1 and 2 to the north and patch 8 to the south, Figure S1 and Figure S2). A Mantel test comparing pair-wise genetic distance (Fst) to geographic distance among patches, showed a positive but non-significant correlation between the two variables (r = 0.13, p = 0.300, n = 6). Of the 1697 marked individuals, 232 recaptures on different days were recorded, of which 11 were inter-patch movements, all completed by males ( Figure S1). Most movements occurred between the largest central patch (patch 3) and the surrounding patches ( Figure S1). The    (Crawford et al., 2012;Nybom, 2004).

Population size and movement at the largest isolated rear-edge population
Despite the prediction that peripheral populations are genetically depauperate, we found the opposite, with isolated rear-edge populations showing the higher diversity, as well as a significantly decline in genetic diversity with increasing latitude. This result supports the idea that these differences in genetic diversity levels are the result of historical range changes (Hewitt, 2000) and is in accordance with recent research showing that for many species in the Palaearctic, peripheral rather than central populations are more genetically diverse (Hewitt, 2000;Petit et al., 2003). Although information on the colonisation route by Scotch argus into Britain after the Last Glacial Maximum is limited, studies of populations in continental Europe indicated that the glacial refugia for the species was situated in the Alps, from which it likely expanded both west and eastwards, with the populations in Britain more similar to those currently present in the northern Alps (Wendt et al., 2021). This is also consistent with the origin of Erebia epiphron, the only other Erebia species present in Britain, in which individuals in populations at the rear-edge (northwest England) appear related to those from populations in the French Alps (Hinojosa et al., 2019;Minter et al., 2020). Thus, the observed decrease in genetic diversity with latitude likely reflects an erosion in genetic diversity along the postglacial expansion due to successive bottlenecks (Hewitt, 2000) as has been reported for other boreo-montane butterfly species in Europe .
The isolated rear-edge populations studied here also appear to be more genetically diverse than those at the connected rear-edge in Scotland reported by Harper (2011), using AFLP markers (0.187 ± 0.012 in connected rear-edge compared to 0.215 ± 0.003 for our isolated rear-edge populations). Harper (2011) also found that marginal and central populations within mainland Scotland did not differ in genetic diversity and rather than genetic diversity decreasing with latitude, there was a tendency for genetic diversity to decrease towards the east. This is consistent with our results of a high genetic diversity in the Scottish island populations, despite their potential isolation. This suggests that the expansion of Scotch argus across Britain could have predominantly followed a western route, with higher genetic diversity in the western populations compared to those in the east. Low genetic diversity may result in reduced fitness (Reed & Frankham, 2003), increasing the negative effects of inbreeding when populations are small and limiting the adaptative potential of populations to environmental change (Kahilainen et al., 2014;Lanfear et al., 2014). Inbreeding can lead to reduced fitness in butterflies (Keller & Waller, 2002;Saccheri et al., 1998), but our results suggest that most Scotch argus population in Britain are currently unaffected by strong diversity loss. Moreover, our results suggest that more isolated populations currently present both at the rear-edge in England and in the west Scottish islands are not necessarily at risk of inbreeding, even if populations are declining. This adds to the uncertainty whether small butterfly populations are automatically susceptible to extinction due to inbreeding (Harper, 2011;Schmitt et al., 2005) and suggests that local extinctions may be the result of other factors, such as climate change and habitat degradation (Franco et al., 2006). This Are isolated rear-edge populations more genetically differentiated?
We found significant genetic differentiation among regions and among populations, with the isolated rear-edge populations showing the higher differentiation. This is consistent with other studies showing that isolated peripheral populations have undergone a different evolutionary history and selection pressures than connected populations within the core of the species range and so represent distinctive genetic units (Habel, Drees, et al., 2010;Minter et al., 2020).
Moreover, the degree of differentiation was related to geographic distance, suggesting that Scotch argus colonised Britain from the continent in a single event through England and subsequently expanded towards Scotland. This is in contrast to the pattern observed for the other two species of cold-adapted butterfly species in Britain, Erebia epiphron (Minter et al., 2020) and Coenonympha tulla (Joyce et al., 2009) for which populations in England and Scotland appear to represent two separate events of colonisation from continental Europe. Moreover, the Scottish islands populations of Scotch argus showed some connection to those on the Scottish mainland, but not to each other, probably due to the length of time for which they have been separated.
Our results of the MRR showed that although males can fly long distances, most individuals moved shorter distances resulting in significant genetic differentiation between individuals in relatively close habitat patches that are surrounded by a hostile matrix. Therefore, genetic differentiation among populations in our study species is likely explained by a combination of factors including post glacial colonisation (Hewitt, 2000), limited dispersal together with habitat fragmentation and retraction due to climate change (Hampe & Petit, 2005). From a regional perspective, it is possible that the differentiation contributed to the survival of the isolated rear-edge populations. If these populations are uniquely warm adapted (Dahlhoff & Rank, 2000), it may have allowed them to persist despite the warming effect, which has driven population extinctions in both the east of England and southern Scotland (Franco et al., 2006). Therefore, assessing the potential for these isolated rear-edge populations to be better suited for warmer conditions than those further north at the current core of the species range, appear crucial for the survival of the species, at least in Britain. Although the re-introduced population showed a lower level of genetic diversity than the founder population, its persistence for the last 10 years provides some support that the area may still climatically suitable for the species, providing habitat quality is maintained and the right genotypes are preserved.
Is dispersal and gene flow within the largest isolated rear-edge population enough to ensure long-term persistence?
The  (Table S2, Figure S1). We found that genetic differentiation between patches was significant for those at the periphery of the site, but central patches showed higher genetic similarity (Table S5), indicating that the site supports a typical metapopulation, with some level of gene flow between patches maintained by occasional movement of individuals (Hanski & Gaggiotti, 2004). This is confirmed by the MRR results showing that all patches are within the maximum observed flight distance ($2 km) of males from at least one other patch. Although most individuals remained within their original patch, 6.7% of those recaptured were after inter-patch movements. If we assume that this is representative of emigration levels for the whole metapopulation, then we estimate that there were approximately 527 inter-patch movements from an estimated total population size of 7869 individuals. All inter-patch movements were performed by males, which agrees with the observed dispersal for populations in the Czech Republic (Slamova et al., 2013), in which males were more likely to leave their natal patch. However, Slamova et al. (2013) also found that females were capable of carrying out long distance flights (up to 2.1 km), which is not consistent with our maximum distance recorded from females' movements of only 90 m. These differences may be explained by differences in patch configuration and matrix permeability between the two regions, which have been shown to influence dispersal propensity (Baguette & Van Dyck, 2007;Ricketts, 2001). The highly intensively managed landscape at our study site has resulted in natural habitat patches being surrounded by a hostile matrix of heavily grazed improved grasslands, which could have imposed selection pressures against dispersal (Baguette et al., 2003;Baguette & Van Dyck, 2007).
It should be also mentioned that MRR studies normally suffer from a male-bias in capture and recapture rates, as males are more easily detected due to their patrolling behaviour. For this reason, it is possible that long-distance female movements did occur during the study period but were not recorded, although the fact that we did not record intermediate distance movements for females (100-200 m) despite continuous habitat being available within those distances (e.g., in patch 3), supports the idea of truly low female mobility.
It is surprising that patch 3 at the centre of the site showed high level of differentiation with other patches that were relatively closer (e.g., with patch 2, Table S5), as individuals were recorded moving to and from patch 3 during the mark-release-recapture study. A possible reason is that male butterflies are less likely to leave a patch with high female density (Baguette et al., 1998), so emigration may be lower overall. Another possibility is that current genetic differentiation among patches reflects past rather than present spatial configuration (Holzhauer et al., 2006), with patch 3 isolated for longer than the surrounding patches, leading to a lower level of historic mixing. However, this is unlikely given the central position occupied by this patch in the metapopulation ( Figure S1) and the very short distances to the surrounding patches, instead suggesting it represented the centre of the historical continuous population. Moreover, the population in patch 8, the most isolated from the others, was only discovered in 2015 (T. Dunbar personal communication), despite being on the roadside verge of a well transited road. Although we detected one male moving from patch 3 to patch 8, it is likely that this population is a relatively recent colonisation by a few females coming from patch 4, as reflected by the low genetic differentiation (Fst = 0.005). The distance between patch 4 and 8 is within the maximum distance moved by males but well beyond the maximum distance recorded by females so steppingstone movements, using patch 6, 7 and 9, appears to be the most likely scenario for the natural colonisation of patch 8 ( Figure S1). Moreover, although we did not record any butterflies in patch 9 during our MRR study, there have been occasional records of individuals in that patch in other years. Overall, the results show that at least males of Scotch argus are moving freely throughout the metapopulation and that current levels of dispersal are sufficient for gene flow.

Conservation implications
Our results show that the British populations of Scotch argus maintain a high level of genetic diversity representing an important reservoir of genetic diversity for the species. Our results also show that despite decline and extinction of populations at the rear-edge, the remaining isolated populations are more genetically diverse than, and differentiated from those within the continuous core rage in Scotland. This suggests that the decline and extinction of populations recorded in the last three decades are likely to be related to demographic and environmental effects rather than genetic constraints. These results highlight that isolated populations at the rear edge of species current ranges should be considered of high conservation priority, even if the species is not at higher risk across the whole range (Channell, 2004). This is even more important for cold-adapted species such as Scotch argus, for which populations at the rear edge are suffering higher exposure to climate warming but may also have become more adapted to warmed conditions, harbouring the adaptative potential for the species to survive future climate change (Abeli et al., 2018;Hill et al., 2011;Nadeau & Urban, 2019).
Moreover, if isolated rear-end populations are large, as is the case at least for one of our study sites, they may represent important sources for future conservation actions. Reintroductions and translocations of individuals to bolster or create new populations have been used extensively as conservation tools for recovering species with small and fragmented populations (Seddon et al., 2007;Taylor et al., 2017). However, only recently have those ideas been discussed in the context of climate change (e.g., assisting colonisation to help species tracking suitable climate in a highly human modified landscape; Ferrarini et al., 2016;Hoegh-Guldberg et al., 2008;Seddon et al., 2014;Thomas, 2011).
Conservation of cold-adapted species may also require these interventions to help these species surviving in colder spots within continuously warming landscapes (Massimino et al., 2020;Suggitt et al., 2018). Selecting the isolated rear-edge populations as the source of individuals maybe more sensible than using populations further north. For example, the reintroduced population examined here provide some support for this idea as the population has persisted for at least 10 years and is doing well (Dave Wainwright, personal communication).
In conclusion, our results highlight that isolated populations at the rear-edge of cold-adapted species should be considered of high conservation priority, as they hold high levels of genetic diversity and differentiation which may prove to be key for the survival of these species under global warming.

SUPPORTING INFORMATION
Additional supporting information can be found online in the Supporting Information section at the end of this article.