Characterizing neutral and adaptive genomic differentiation in a changing climate: The most northerly freshwater fish as a model

Abstract Arctic freshwater ecosystems have been profoundly affected by climate change. Given that the Arctic charr (Salvelinus alpinus) is often the only fish species inhabiting these ecosystems, it represents a valuable model for studying the impacts of climate change on species life‐history diversity and adaptability. Using a genotyping‐by‐sequencing approach, we identified 5,976 neutral single nucleotide polymorphisms and found evidence for reduced gene flow between allopatric morphs from two high Arctic lakes, Linne'vatn (Anadromous, Normal, and Dwarf) and Ellasjøen (Littoral and Pelagic). Within each lake, the degree of genetic differentiation ranged from low (Pelagic vs. Littoral) to moderate (Anadromous and Normal vs. Dwarf). We identified 17 highly diagnostic, putatively adaptive SNPs that differentiated the allopatric morphs. Although we found no evidence for adaptive differences between morphs within Ellasjøen, we found evidence for moderate (Anadromous vs. Normal) to high genetic differentiation (Anadromous and Normal vs. Dwarf) among morphs within Linne'vatn based on two adaptive loci. As these freshwater ecosystems become more productive, the frequency of sympatric morphs in Ellasjøen will likely shift based on foraging opportunities, whereas the propensity to migrate may decrease in Linne'vatn, increasing the frequency of the Normal morph. The Dwarf charr was the most genetically distinct group. Identifying the biological basis for small body size should elucidate the potential for increased growth and subsequent interbreeding with sympatric morphs. Overall, neutral and adaptive genomic differentiation between allopatric and some sympatric morphs suggests that the response of Arctic charr to climate change will be variable across freshwater ecosystems.


| INTRODUC TI ON
The Arctic has been profoundly affected by climate change within a brief period of time, with sea surface temperatures increasing significantly faster than the global average, and sea ice extent and thickness declining rapidly (Intergovernmental Panel on Climate Change, 2013). Further temperature increases will advance the break-up of ice (Prowse et al., 2011), expanding the open period for some of the largest rivers in the world . Accelerated glacial melting will also contribute to rising sea level and a higher frequency of landscape-altering, glacial lake outburst floods (Harrison et al., 2018). Other stressors such as freshwater acidification, eutrophication, and ozone depletion will be exacerbated by climate change as well (Adrian et al., 2009;Heino, Virkkala, & Toivonen, 2009;Schindler & Smol, 2008;Woodward, Perkins, & Brown, 2010).
Prolonged climate change is expected to have a significant impact on Arctic freshwater ecosystems (Woodward et al., 2010), and have wide-ranging effects on freshwater and anadromous Arctic fish species (Reist, Wrona, Prowse, Dempson et al., 2006;Reist et al., 2006aReist et al., , 2006bRouse et al., 1997;Wrona et al., 2006). Elevated temperatures, increased light penetration, and a higher concentration of atmospheric carbon dioxide are likely to increase the productivity of lotic and lentic systems and cause shifts in habitat availability and quality across the region (Adrian et al., 2009;Brander, 2010;Heino et al., 2009;Michelutti, Wolfe, Vinebrooke, Rivard, & Briner, 2005;Prowse et al., 2006;Woodward et al., 2010). Novel competitors, prey, parasites, and diseases from lower latitudes are likely to invade with increased temperatures (Davidson et al., 2011;Fossheim et al., 2015;Parkinson & Butler, 2005). In turn, such a changeover in diversity is likely to modify foraging preferences and the trophic structure of Arctic fish (Reist et al., 2006a), which will affect commercial and subsistence fishing yields, as well as food safety (Davidson et al., 2011;Reist et al., 2006a;White, Gerlach, Loring, Tidwell, & Chambers, 2007).
The Arctic charr Salvelinus alpinus (Linnaeus, 1758) has a broad, circumpolar distribution and is the most northerly distributed freshwater fish species (primarily above 60°N; Klemetsen, 2010;Wilson et al., 2004). Consequently, Arctic charr may provide a useful model to study the impact of climate change on cold water adapted fish species and Arctic freshwater ecosystems in general. Although some models predict that small temperature increases (<2.5°C), may increase Arctic charr growth rates (Budy & Luecke, 2014;Elliott & Elliott, 2010;Reist et al., 2006a), Lehnherr et al.. (2018) have attributed a rapid decline in physiological condition of Arctic charr in Lake Hazen on Ellesmere Island to climate change. Furthermore, studies indicate that the eggs of the Arctic charr are not expected to survive a freshwater temperature increase of 5°C, and temperatures exceeding 22-23°C will likely result in mass adult mortality unless cold water refugia are present (Elliott & Elliott, 2010). Range expansion of competitive salmonid species into the Arctic is likely to cause a substantial decrease in the geographic range of the species (Ficke, Myrick, & Hansen, 2007;Reist et al., 2006a), with a 73% reduction predicted in Sweden by 2100 (Hein, Öhlund, & Englund, 2012).
At high Arctic latitudes, lakes are typically of recent postglacial origin with low biodiversity, and the Arctic charr is often the only freshwater fish species present (Christiansen & Reist, 2013;Power, Reist, & Dempson, 2008). Arctic charr exhibit complex migratory behavior that has likely aided their colonization of such remote regions, including the Svalbard Archipelago (Svenning, Klemetsen, & Olsen, 2007). For instance, some individuals complete their entire life cycle in freshwater (residency), whereas others hatch and grow as juveniles in freshwater, migrate to sea (anadromy) to forage for weeks or months, and then return to freshwater to overwinter or reproduce (Craig, 1989;Gilbert, Donad, Swanson, & Tierney, 2016;Jensen & Rikardsen, 2008;Moore, Harris, Tallman, & Taylor, 2013;Radtke et al., 1996;Tallman, Roux, & Martin, 2018). This life history is different from anadromy in Pacific salmon (Oncorhynchus) because all migrants, including juvenile fish, return annually to freshwater to overwinter. Notably, Arctic charr populations in the southern range of the species distribution do not exhibit anadromy (Klemetsen et al., 2003). This phenomenon has been attributed to increased productivity in lower latitude lakes, which negates the energetic benefit of migrating to feed at sea where the risk of predation is higher (Finstad & Hein, 2012;Jensen, Finstad, & Fiske, 2018).
The degree of genetic differentiation among Arctic charr morphs is an ongoing topic of research, and their distinction has mostly been investigated using neutral microsatellites and mitochondrial genes.
It is currently uncertain how Arctic charr morphs will be affected by climate change. As freshwater productivity increases with climate change, resident morphs are predicted to increase in frequency, with fewer anadromous migrants occurring (Finstad & Hein, 2012;Reist, Wrona, Prowse, Dempson et al., 2006). This change may result in a reduced rate of gene flow and an increased degree of genetic differentiation among allopatric populations.
Elevated freshwater temperatures and an increase in ice-free days are also predicted to lengthen the potential growing season for Arctic charr (Elliott & Elliott, 2010;Reist et al., 2006a), which may cause small-sized morphs to grow larger (Budy & Luecke, 2014).
This size shift could increase the intensity of competition among small and larger sized sympatric morphs (Budy & Luecke, 2014) and result in more matings between morphs, eroding their genetic differentiation.
To investigate how Arctic charr may respond to climate change, we focus on morphs from two high latitude lakes, Linne´vatn and Ellasjøen, in the Svalbard archipelago, a Norwegian territory in the Arctic Ocean. An improved understanding of the genetic diversity within and among these morphs should indicate the adaptability of Arctic charr in Svalbard and other high latitude locations to factors such as increased freshwater temperatures and ice-free days. Using a genotyping-by-sequencing (GBS) approach (Elshire et al., 2011), we identify both neutral and putatively adaptive loci and test for genetic differentiation among the sympatric and allopatric morphs to determine the degree of reproductive isolation that can be attributed to genetic drift or local adaptation.  Svenning et al., 2007).
Since climate change is predicted to increase freshwater productivity, potentially leading to a reduction or total loss of anadromy (Finstad & Hein, 2012;Jensen et al., 2018), we test for evidence of gene flow between the Anadromous and two resident morphs. If Anadromous individuals represent a distinct gene pool from resident fish, the propensity to migrate could be lost entirely among Linne´vatn Arctic charr. Increased temperatures may also lengthen the growing season of small-sized Arctic charr (Budy & Luecke, 2014;Elliott & Elliott, 2010;Reist et al., 2006a), such that Dwarf individuals in Linne´vatn reach larger body sizes. This size change could increase the intensity of competition and rate of interbreeding between the Dwarf and typical-sized Anadromous and Normal morphs, eroding any potential genetic differentiation and possibly leading to the extinction of the Dwarf morph.
Our sample size was limited from Ellasjøen, but as a preliminary investigation, we tested for gene flow between the Littoral and Pelagic morphs. Climate change is likely to increase freshwater temperatures and productivity, facilitating the invasion of novel prey and parasite species (Davidson et al., 2011;Fossheim et al., 2015). This change in diversity is likely to modify the trophic structure of Lake Ellasjøen, altering foraging preferences and intensity of competition between the Littoral and Pelagic morphs. In turn, such change may affect the relative frequency of the two morphs. If gene flow is limited between Littoral and Pelagic fish, then each morph will need to adapt independently to the effects of climate change, or face extinction.

| Field sites and sample collections
Linne´vatn (78°03′N, 13°50′E) is located on western Spitsbergen, the largest island in the Svalbard archipelago ( Figure 1). It is the second largest lake on Svalbard with an area of 4.6 km 2 and a maximum depth of 37 m, occurring ~10 m above sea level (Bøyum & Kjensmo, 1978). There is a brief ice-free period from late July to mid-August, and the high water flow in the 2-km long river between the lake and the coast allows access to the sea (Bøyum & Kjensmo, 1978). Lake Ellasjøen (74°30′N, 19°00′E) is located on Bear Island, a remote smaller island, equidistant between Spitsbergen and mainland Norway. The lake is 21 m above sea level with an area of 0.73 km 2 and a maximum depth of 43 m (Bertram & Lack, 1933;Klemetsen, Grotnes, Holthe, & Kirstoffersen, 1985). Lake Ellasjøen flows into the short River Fossåa, but the lake outlet is too steep to permit the migration of Arctic charr (Klemetsen et al., 1985). The ice-free period occurs from late June to mid-September or October (Bertram & Lack, 1933).
Anadromous and resident fish were sampled from Linne´vatn in August 2013. Anadromous fish were captured in the main inlet river using a box trap net while resident fish were captured in the lake using a gill net. Two gill nets were used, the first of which had dimensions 1.5 m high and 30 m long with a mesh size of 26-39 mm, whereas the second net was 1.5 m high and 40 m long with a mesh size 10-45 mm. Both nets were used at a depth of 0-5 m in the littoral zone. Fork length (FL) measurements and sex (M/F) were recorded for each individual. A small piece of the adipose fin was removed from each fish and stored in a 1.5-ml vial filled with 95% ethanol. In total, 24 Anadromous fish (FL range: 315-587 mm; M/F = 12/12) were collected from Linne´vatn River as they migrated from the sea to the lake. Resident fish were classified as either a Dwarf or Normal morph based upon FL, morphology, and sexual maturity. Fish caught in the lake were dissected to determine sexual maturity. Fish that were sexually mature and <205 mm were classified as Dwarf charr. According to these criteria, twenty Normal (FL range: 207-540 mm; M/F = 5/15) and 28 Dwarf (N = 28; FL range: 109-204 mm; M/F = 9/19) samples were collected from Lake Linne´vatn (Table 1).
Arctic charr were sampled from Lake Ellasjøen in 2009 as part of a previous study (Hawley et al., 2016). In total, 19 resident individuals were obtained for genetic analysis. From these samples,  Table 1). Sex was not recorded for seven Lake Ellasjøen samples.

| Library preparation and sequencing
DNA extraction, library construction, and sequencing were conducted at Oregon State University's Center for Genome Research and Biocomputing, largely following the methodology of Elshire et al. (2011); modifications included the incorporation of a second restriction enzyme and a size selection step (see below). Briefly, F I G U R E 1 Location of the two lakes, Linne´vatn and Ellasjøen in the Svalbard archipelago. Anadromous Arctic charr were sampled in the Linne´vatn River, and both resident Normal and resident Dwarf charr were sampled from Lake Linne´vatn. Littoral and Pelagic Arctic charr samples were collected from Lake Ellasjøen. The inset map for the Svalbard archipelago was generated using the R package maps (Becker & Wilks, 2008)

| SNP calling and quality filtering
Raw Illumina reads were first assessed for quality using fastqc (v.  programs is known to vary according to sample size and the observed pattern of genetic variation (Ahrens et al., 2018;Narum & Hess, 2011), outlier was required to be identified by both lositan and bayescan in order to be cl assified as putativel y adaptive. These outlier loci are considred putaitvely adaptive until future validation studies are conducted (Shafer et al., 2015). A conservative approach was used for categorizing SNPs as neutral, using a p-value range estimated by lositan of 0.10-0.90, and a q-value threshold >0.50 as calculated by bayescan. A list containing putative adaptive and neutral SNPs were then used to extract each set of loci from the filtered dataset using plink v. 1.07.

| Differentiation among morphs based on neutral loci
Pairwise linkage disequilibrium (LD) of neutral SNPs was esti-

| Signatures of selection
In order to disentangle whether one or multiple morphs were driving

| SNP calling and quality filtering
Following library preparation and sequencing, 342,141,354 raw Illumina reads were evaluated using stacks process_radtags script.
Of these reads, 321,417,116 (93.9% of total reads) were retained, with demultiplexed samples averaging 3.84 million reads (±0.91 SD). A preliminary stacks run identified two individuals with low genotyping rate (>80% missing data; 2 Anadromous samples), which were excluded from the analysis. Assembling loci across the remaining individuals (N = 89) resulted in retention of 12,271 loci, containing a total of 18,095 SNPs (mean = 1.5 SNPs/locus).
After restricting the analysis to one SNP per locus, and excluding 49 SNPs that were homozygous, 1,308 SNPs were identified as putative paralogs (Supporting Information Figure S1). stacks popu-

| Identification of outlier loci
Using the full filtered dataset (8,075 single SNP loci), bayescan identified 17 candidate outliers (0.21% of SNPs), with q-values of candidate outliers ranging from 0.0002 to 0.049 (Table 2; Supporting   Information Table S2). All of these 17 candidate outliers were also identified by lositan, which classified 418 SNPs as candidates for positive selection, with probabilities (simulated F ST < sample F ST ) ranging from 0.996 to 1.0. Overall, examination of bayescan and lositan output resul ted in agreement of 17 putativel y adaptive and 5,980 neutral SNPs (Table 2; Supporting Information Table S2), which were subsequently extracted from the filtered dataset using plink.

| Differentiation among morphs based on neutral loci
Out F I G U R E 2 Cluster analyses of 5,076 neutral loci among five different Arctic charr morphs from lakes Ellasjøen and Linne´vatn: (a) first two axes from a discriminant analysis of principal components with points corresponding to individuals. Shape of points is based upon actual phenotypic group, with color corresponding to inferred group; (b) barplot of inferred group membership probability (k = 3 clusters) from k-means clustering; and (c) k = 3 also identified by admixture as best describing the structure in the data pairwise genetic differentiation was observed among morphs (F ST range = 0.002-0.178 Evaluating the BIC scores following successive k-means clustering identified k = 3 as the optimal number of genetic groups in the dataset (Supporting Information Figure S2a). This inferred group size was used for the DAPC after retaining seven principal components (

| Signatures of selection
Using individuals from both lakes, we identified 17 putatively adaptive SNPs supported by both bayescan and lositan. Simil ar to the set of neutral SNPs, high differentiation was observed among allopatric morphs (F ST = 0.808-0.853), yet a substantial range in genetic differentiation occurred among sympatric morphs (F ST = −0.04-0.511;  Figure S4).
Examining the distribution of allele frequencies for these 17 loci indicated that a majority (15/17) showed allele frequency disparities among allopatric morphs, largely driven by the presence of multiple monomorphic loci for Ellasjøen samples. Repeating bayescan using only the Ellasjøen samples failed to identify any outliers. Two loci (56596|Sequence 13 and 85782|Sequence 16) showed substantially different allele frequencies within the Linne´vatn samples (Supporting Information Figure S3), and repeating bayescan analysis using only the Linne´vatn samples identified these loci as repeat outliers at a local scale, with q-values ranging from 0.0002 to 0.0003.
Using all of the consensus outlier sequences as queries (n = 17), blastn had sol id hits against the Salvelinus alpinus reference genome, with E-scores ranging from 9.00e-21 to 6.00e-67 (Supporting Information Table S3). Four outlier sequences aligned to an unplaced scaffold, two sequences aligned to a linkage group with no coding regions within a 50 kb window, and the remaining 11 sequences mapped to a genomic location containing one or more coding regions within a 50 kb window (Supporting Information Table S3).
These targets for selection included genes associated with the regulation of transcription, regulation of telomeres, ion transport, iron binding, protein folding, embryonic development, bone morphogenesis, and memory (Supporting Information Table S3). The two local outliers (56596|Sequence 13 and 85782|Sequence16) identified within the Linne´vatn samples aligned to an unplaced scaffold and Linkage Group 6.1, respectively. The outlier sequence on Linkage Group 6.1 (85782|Sequence 16) mapped to within 50 kb of the interleukin 1 receptor accessory protein like 2 (IL1RAPL2), which is located on the X chromosome in humans and is linked to cognitive disability (Supporting Information Table S3). An important paralog of this gene is IL1RAPL1, which is expressed at a high level in the hippocampal memory system, suggesting a specialized role in the physiological processes underlying memory and learning abilities (www. genecards.org/cgi-bin/carddisp.pl?gene=IL1RAPL1).

| Neutral and adaptive genomic differentiation
We found that landlocked Arctic charr in Ellasjøen had lower genetic diversity (i.e., mean allelic richness and expected heterozygosity) than those in Linne´vatn. Varying levels of genetic differentiation were detected among allopatric and sympatric morphs using 5,976 neutral and 17 adaptive loci. Fifteen of the identified adaptive loci were shared between Linne´vatn and Ellasjøen, and consensus sequences had solid hits against the Salvelinus alpinus reference genome (E-scores range: 9.00e-21-6.00e-67; Supporting Information Table   S3). Two adaptive loci were restricted to samples from Linne´vatn.
Only one (85782|Sequence 16) of the two adaptive loci mapped to an annotated genomic location. We found one coding region within a 50 kb window of this locus, interleukin 1 receptor accessory protein like 2 (IL1RAPL2), which is located on the X chromosome in humans.
While mutations in this gene have been shown to cause cognitive disabilities in humans, its role in fishes has yet to be determined. In contrast, we failed to detect F ST outliers using only samples from Ellasjøen, suggesting that there were no sequenced adaptive genetic differences between the Littoral and Pelagic morphs. However, these findings should be considered preliminary given our limited sampling from Ellasjøen. A more comprehensive study focusing on all morphs within this system is definitely warranted.

| Comparison of allopatric morphs
Pairwise F ST estimates based on neutral and adaptive loci indicated high differentiation among allopatric morphs (Table 3), and cluster analyses of neutral loci separated samples from Linne´vatn and Ellasjøen (Figure 2). Considering that fish within Ellasjøen are landlocked, the neutral and adaptive genetic differences observed between morphs from the two lakes were expected. High genetic differentiation based on neutral markers has also been previously reported for six landlocked populations of Arctic charr across Fennoscandia (Shikano et al., 2015), and 12 landlocked populations in Maine (Bernatchez et al., 2002). Like Linne´vatn and Ellasjøen (Bøyum & Kjensmo, 1978;Klemetsen et al., 1985), these lakes are of postglacial origin (<12,000 years old) but it is uncertain if any have been geographically isolated for a similar length of evolutionary time as Ellasjøen (Bernatchez et al., 2002;Shikano et al., 2015). Our findings suggest that the allopatric morphs in Linne´vatn and Ellasjøen are locally adapted, and thus, responses to climate change will be variable between the lakes.

| Comparison of sympatric morphs
We could not genetically distinguish the sympatric Littoral and Pelagic morphs in Ellasjøen, indicating that they represent a single breeding population. The F ST values based on both neutral and adaptive loci were statistically insignificant (Table 3), and cluster analysis of neutral loci failed to distinguish any subgroups of genetic variation within Ellasjøen (Figure 2). Interestingly, previous research differentiated between the Littoral and Pelagic morphs based on foraging differences as well as different sensitivities to photoperiod variation (Hawley et al., 2016(Hawley et al., , 2017. Our genetic results from Ellasjøen, however, are consistent with other genetic studies which found that certain sympatric Arctic charr morphs are not genetically or reproductively distinct (see review Jonsson & Jonsson, 2001;e.g., Nordeng, 1983). However, the failure to detect genetic differentiation (i.e., neutral or adaptive) between the Littoral and Pelagic morphs in Ellasjøen may reflect inadequate sampling of these sympatric morphs. Future genetic investigations of Ellasjøen charr should also include the previously described dwarf morph (Hawley et al., 2016) to determine whether those individuals represent a genetically distinct group. A more comprehensive investigation, sampling additional individuals and all morphs in Ellasjøen, should indicate whether adaptive differences are present among these morphs. If our preliminary genetic results are accurate though, Arctic charr in Lake Ellasjøen may be more resilient to climate change than expected. For instance, a single population exhibiting substantial variation in morphology, foraging preferences, and photosensitivity (Hawley et al., 2016(Hawley et al., , 2017) may adapt more readily to increased light penetration and elevated freshwater productivity (Heino et al., 2009;Michelutti et al., 2005;Prowse et al., 2006;Woodward et al., 2010) than a set of smaller, fragmented populations with distinct phenotypes.
Within Linne´vatn, cluster analysis of neutral loci failed to distinguish the Anadromous and resident Normal morphs (Figure 2). The F ST values estimated for neutral and adaptive loci were both moderate and statistically significant between these sympatric morphs, although differences were somewhat higher for adaptive loci (Table 3).

This result is not necessarily a surprise, as anadromous and resident
Arctic charr both overwinter in freshwater, and the relationship between migratory behavior and gene flow is known to vary (Moore et al., 2013). To assess the opportunity for reproduction between morphs, otolith chemistry, tagging, and radio telemetry (albeit logistically challenging in the Arctic environment) could be used to determine the spatial and temporal patterns in migration and spawning of the Anadromous and resident Normal morphs in Linne´vatn (e.g., Brenkman & Corbett, 2007).
Our genetic results for the Anadromous and resident Normal morphs in Linne´vatn are consistent with previous studies of Arctic charr in the Salangen River, Norway, which suggested that anadromous and resident individuals belonged to the same population (Nordeng, 1961(Nordeng, , 1983. In addition, Nordeng (1983) conducted extensive rearing experiments and found that a fraction of the resident Arctic charr always transformed into anadromous fish. That said, a higher proportion of the offspring from anadromous parents smoltified (i.e., underwent physiological changes to adapt to seawater) when released into the river compared to offspring from resident parents, indicating that there is some genetic influence on anadromy (Nordeng, 1983).
Climate change is predicted to increase light penetration, water temperature, and carbon dioxide concentrations for freshwater systems (Heino et al., 2009;Michelutti et al., 2005;Prowse et al., 2006;Woodward et al., 2010). Such changes are expected to enhance freshwater productivity, making the risk of predation while feeding at sea more costly. As a result, the frequency of resident Arctic charr may increase (Finstad & Hein, 2012;Jensen et al., 2018;Reist, Wrona, Prowse, Dempson et al., 2006 (Figure 2). Our findings are consistent with other studies of sympatric variation in Arctic charr that also found evidence for strong genetic differentiation between Dwarf individuals and larger sized morphs (e.g., Adams et al., 2008;Alekseyev et al., 2014;Gordeeva et al., 2015;Samusenok et al., 2006).
Previous studies have documented differences in breeding time between typical-sized and dwarf morphs from other lakes (Jonsson & Hindar, 1982;Pavlov & Osinov, 2008). A similar "adaptation by time" (Hendry & Day, 2005) could explain the genetic differentiation observed for the Dwarf morph in Linne´vatn. It is possible that assortative mating based on spawning site or depth could also be an important driver of reproductive isolation among sympatric morphs (Jonsson & Hindar, 1982). Alternatively, like many profundal morphs of Arctic charr from other lakes, the Dwarf charr in Linne´vatn may be an example of paedomorphosis where juvenile traits are retained during adulthood (Danzmann et al., 1991;Jonsson & Jonsson, 2001;Klemetsen et al., 2003;Knudsen et al., 2006;Parsons et al., 2011;Skoglund et al., 2015;Skùlason et al., 1989 (Adams et al., 2008;Samusenok et al., 2006) or focused upon benthos (Alekseyev et al., 2014;Fraser et al., 1998). Larger morphs often cannibalize dwarf or small-sized morphs (Knudsen et al., 2016;Samusenok et al., 2006). Svenning et al. (2007) found that the diet of small individuals in Linne´vatn (<15 cm, predominantly Dwarf) was strongly characterized by chironomid Oliveridia tricornis (Oliver, 1976) larvae, whereas larger Arctic charr (>15 cm; Anadromous and Normal) within the lake were mainly cannibals (Svenning et al., 2007). These feeding differences could affect habitat use and the potential for gene flow, given that cannibalism by Anadromous and Normal morphs is potentially mortal risk to the reproductive success of Dwarf individuals. If cannibalism during spawning is common, we would expect selection to be strong against Dwarf individuals mating outside of their morph, unless such matings provide increased fitness, which could lead to the evolution of sexual cannibalism (Elgar & Schneider, 2004). In future studies, it would be useful to examine the rates of breeding and cannibalism between larger sized and Dwarf Arctic charr to determine whether this is a plausible factor to have affected gene flow.
Overall, the Dwarf morph represents a genetically distinct component of Arctic charr diversity in the Linne´vatn. Considering similar genetic evidence from other studies of Dwarf morphs (Alekseyev et al., 2014;Gordeeva et al., 2015;Samusenok et al., 2006), the potential response of small or profundal resident morphs to climate change warrants further investigation. It should be noted, however, that elevated freshwater temperatures are predicted to result in a lengthened growing season for Arctic charr (Budy & Luecke, 2014). If Dwarf individuals are not restricted to small body sizes due paedomorphosis or other heritable genetic changes, it is possible that Dwarf individuals could reach large enough sizes to compete or reproduce more often with larger morphs (Budy & Luecke, 2014).
Such a change could lead to increased introgression among morphs and a reduction in overall genetic diversity in Linne´vatn. Further research is needed to determine whether Dwarf individuals are restricted to small body sizes, and whether increased freshwater temperature affects the growth rate of the morph.

| CON CLUS IONS
The results of this study demonstrate the value of utilizing putatively adaptive loci to investigate intraspecific variation. Geographically isolated, allopatric populations of Arctic charr are likely to be genetically distinct, but genetic divergence among sympatric morphs varies, and response of populations and morphs to climate change probably needs to be considered on a case-by-case basis. That said, results from Anadromous and Normal morphs in Linne´vatn indicate that migratory differences do not necessarily indicate substantial genetic differences within a single lake. In concordance with previous studies (Alekseyev et al., 2014;Gordeeva et al., 2015;Samusenok et al., 2006), Dwarf Arctic charr in Linne´vatn were the most genetically distinct group observed, and the response of Dwarf morphs to climate change therefore warrants further investigation. Given that we observed substantial adaptive and neutral genetic diversity between just two lakes in the geographically remote Svalbard Archipelago, out results suggest that Arctic charr exhibit significant genetic diversity across the Arctic. The impact of continued climate change upon the diversity of other Arctic anadromous and freshwater fish species, and in turn, the wider ecosystem, fisheries, and affected indigenous communities, requires further investigation.

ACK N OWLED G M ENTS
We would like to thank our CharrAdapt project collaborators C.