Patterns of phenotypic and genetic variability show hidden diversity in Scottish Arctic charr


C. E. Adams, University Field Station, Institute of Biomedical and Life Sciences, University of Glasgow, Glasgow G63 0AW, UK; e-mail:


Abstract –  This study examined the degree and pattern of variability in trophic morphology in Arctic charr (Salvelinus alpinus L.) at three spatial scales: across 22 populations from Scotland and between and within two adjacent catchments (Laxford and Shin) in northern Scotland. In addition, the variability at six microsatellite loci between and within the Laxford and Shin systems was determined. Habitat use by charr differed significantly between populations. The pattern of variability in trophic morphology, known to influence foraging ability in charr, showed a very high degree of between-population variation with at least 52% of population pairs showing significant differences in head shape. Trophic morphology and genetic variation was also high over small geographical scales; variation being as high between charr from lakes within the same catchment, as between adjacent catchments. The pattern of both phenotypic and genotypic variation suggests a mosaic of variation across populations with geographically close populations often as distinct from each other as populations with much greater separation. Very low levels of effective migrants between populations, even within the same catchment, suggest that this variation is being maintained by very low straying rates between phenotypically and genetically distinct populations, even when there is no apparent barrier to movement. We conclude that the genetic and phenotypic integrity of charr populations across Scotland is high and that this adaptive radiation constitutes a ‘hidden’ element of diversity in northern freshwater systems. Two consequences of this are that the population (rather than the species) makes a more rational unit for the consideration of conservation strategies and that the habitat requirements and therefore management needs may differ significantly between populations.


Broad-scale phenotypic differences between populations of freshwater fish species are well documented, particularly within the salmonids. Across the Arctic charr (Salvelinus alpinus), the species exhibits a significant degree of variability in a broad range of phenotypic characters (Johnson 1980). The functional significance of much of this variation is unclear, however, variation in trophic morphology (i.e., in the morphological features associated with the detection, capture and handling of prey (sensu Skúlason & Smith 1995) has been linked with variation in feeding ecology (Snorrason et al. 1994; Adams et al. 1998; Fraser et al. 1999) and in foraging ability (Adams & Huntingford 2002a).

However, until relatively recently, the very wide range of observed phenotypic variability did not seem to be matched by similar variability at the genetic level between populations (Hartley et al. 1995). Recent use of microsatellite DNA techniques have, however, now enabled between-population discrimination at a genetic level (Gíslason et al. 1999; Wilson et al. 2004).

One of the consequences of phenotypic variability across Arctic charr is that it has led to speculation over the integrity of the species, with some individual populations being proposed as putative species (see, e.g., Adams & Maitland 2006). Although there is a strong case for a re-evaluation of the taxonomic status of alpinus (see Kottelat 1997) here for simplicity we refer to Arctic charr, S. alpinus, as a single species.

A second consequence of phenotypic and genetic variability across the species that has received little attention, is the effect of this extreme variability on the strategies for effective conservation and management of Arctic charr.

Two critical questions stem from the extent of the phenotypic variability found in the Arctic charr which are less acute in other species.

Firstly, is it possible to generalise about the conservation and management requirements across this species? If the biological or physical requirements differ substantially across the species, as the result of variability, then it may be difficult to develop generic conservation criteria.

Secondly, what is a reasonable unit of conservation? If there is significant structuring in the phenotypic or genetic variability within the species, a focus on conservation and management at the species level may not result in the protection of important elements within the species.

Here we address these questions using data on the phenotypic, genetic and behavioural variation in charr from across Scotland. Specifically, we examine:

  • 1Variation in habitat use by Arctic charr across a discrete, geographic area (across Scotland).
  • 2The degree and pattern of phenotypic variation across three spatial scales: across Scotland, between adjoining catchments and between lakes within catchments.
  • 3The degree and pattern of genotypic variation between adjacent catchments and between lakes, within catchments.

Materials and methods

Arctic charr were collected from a total of 22 lakes across Scotland. Collection sites spanned the most southerly, to the most northerly Scottish charr populations and populations from the east to the west of the country (Fig. 1). For 21 of these sites, a structured randomised sampling procedure was used (see below); for one further site (Loch Grilsta) charr samples were obtained but not in a structured sampling protocol. This site was omitted from analysis of comparative catch rates and habitat use.

Figure 1.

 The 22 sampling sites for charr from Scotland and Ireland. aG, a'Ghriama; Ae, Awe; Cr, Clair; Cn, Coulin; De, Doine; Dn, Doon; En, Earn; Ek, Eck; Ga, Grilsta; Lt, Langavat; Le, Lee; Lg, Lubnaig; Mr, Maree; Mt, Mealt; Md, Merkland; Me, More; Rh, Rannoch; Sn, Shin; Sk, Stack; Ta, Talla; Ue, Uaine; Vl, Voil.

To look for variation at smaller spatial scales, two adjoining catchments and all the lakes containing charr within them were sampled. The Laxford system (Sutherland, Scotland) drains west and contains two lakes which support charr population (lochs Stack and More). The Shin catchment adjoins the south-east edge of the Laxford system but drains to the east of Scotland. It contains three lakes which support charr populations (lochs Merkland, a'Ghriama and Shin) (Fig. 2).

Figure 2.

 The westerly draining Laxford and easterly draining Shin catchments showing lochs Stack and More (Laxford catchment), Merkland, a'Ghriama and Shin (Shin catchment).

Structured random sampling for fish

Fish were sampled from each loch by gill-net. Standard mono-filament survey gill-nets of 60-m long and 1.5-m deep (Lundgrens, Sweden) comprising 10 panels of 8–45 mm half-mesh size, were set on the bottom of the loch overnight. To sample in the pelagic zone, nets 60-m long and 3-m deep of mesh sizes 12.5–33 mm half-mesh were suspended 1 m below the surface. Catches were then adjusted to a catch per unit effort (CPUE) (100 m2 net 24 h−1) to compare differences in habitat use by charr, nets were set in each of four habitat types in each lake:

Littoral zone, on the bottom in 2–10 m water depth.

Sublittoral zone, on the bottom in 10–20 m water depth.

Profundal zone, on the bottom in 30+ m of water.

Pelagic zone, 1 m below the surface over the deepest area of the loch.

If catches of charr at any lake were less than 50 individuals, sampling was repeated over a second night. Charr were removed from gill-nets at the bankside and given an individual identification code. Where genetic samples were required, the adipose fin was removed and stored in 100% alcohol. Fish were placed in ice for return to the laboratory where they were immediately frozen at −16 °C.

Characterisation of trophic morphology

To look for variation across populations, 14 head landmarks were located from lateral view photographs of each fish. These were used to generate 11 head morphometric variables and a single measure of pectoral fin length (see Fig. 3). As these linear head morphometric variables were strongly correlated with fish size, the residuals derived from the regression of each head variable on fork-length (all populations pooled) were used as a size-independent measure of for each morphometric variable. Principal component scores from PC1 and PC2 were derived using principal component analysis, to enable a relatively simple comparison of shape differences between populations.

Figure 3.

 The 11 measures of head morphology from charr from each site. 1, eye diameter (ED); 2, anterior head length (AHL); 3, posterior head length (PHL); 4, head length (HL); 5, snout length (SL); 6, maxillary length (ML); 7, maxillary width (MW); 8, lower jaw length (LJL); 9, head depth at eye (HDE); 10, Head depth at operculum (HDO). Snout plus maxillary length (SML). Not shown pectoral fin length (PFL).

PC1 coefficients all show similar loadings and all in the same direction (Table 1), thus charr with a relatively large robust overall head shape for a given size have a small PC1 score. PC2 shows large negative factor loadings for pectoral fin length against positive loadings for posterior head length and depth and maxillary width. Thus, fish with relatively short pectoral fins, long, deep heads and a wide maxillary bone will score highly in PC2 scores.

Table 1.   The factor loadings for principal components 1 and 2 for the 12 morphological variables.
  1. ED, eye diameter; AHL, anterior head length; PHL, posterior head length; HL, head length; SL, snout length; ML, maxillary length; MW, maxillary width; LJL, lower jaw length; HDE, head depth at eye; HDO, head depth at operculum; SML, snout plus maxillary length; PFL, pectoral fin length; PC, principal component.



The genetic structure of Arctic charr from lakes within the two drainages was determined from microsatellite genotypic data collected as part of a larger study of European Arctic charr. Details of laboratory procedures, primer references, and summary statistics of allelic diversity are fully presented in detail elsewhere (Wilson et al. 2004). In brief, genomic DNA was isolated from fin tissue samples preserved in ethanol, and individual fish were then genotyped at six microsatellite loci: Sco19, Ssa85DU, SSOSL456, Omy301UoG, One11ASC and OtsG253b. Loci were individually amplified in the polymerase chain reaction (PCR) and products separated using gel electrophoresis. Alleles were visualised using the Hitachi FMBIOII fluorescence imaging system (Hitachi, Mississauga, Canada) and sized by comparison with 350-TAMRA lane standards loaded on each gel. Across the six populations, sample sizes ranged from 50 to 22 with a mean of 40, while the number of alleles per locus ranged from 1 to 20 with a mean of 9.03 over all loci and all populations.

To test for genetic differentiation between lakes under both an assumed infinite alleles model (IAM), and an assumed stepwise mutation model (SMM) of microsatellite mutation, pairwise values of FST and RST, respectively, were estimated and their significance assessed using the permutation procedure implemented in arlequin version 2.000 (Schneider et al. 2000). For all tests a permutation number of 1000 was used. A multilocus hierarchical analysis of genetic variation (amova) was performed using arlequin. Genetic variation was partitioned into among drainage, among lake within region and within lake components. Phi-statistics associated with these components of variance (ΦCT, ΦSC and ΦST, respectively) were estimated and their significance tested using a nonparametric permutation approach (Excoffier et al. 1992). This analysis was also performed under both IAM and SMM assumptions.

Gene flow

Gene flow between populations within each drainage was determined using the coalescent-based Monte Carlo Markov chain (MCMC) method implemented in migrate (Beerli & Felsenstein 1999). For each pair of populations, we estimated gene flow as Nexmyx, the effective number of migrants per generation from lake y to lake x. This method does not require assumptions of equal effective population size (Ne) or migration rate (m) across populations. Data were analysed under an assumed IAM of microsatellite mutation (with mutation rates assumed to be equal across loci). Multiple runs were performed to assess solution convergence with parameter estimates for each drainage obtained using MCMC parameters as follows: 20 short chains with 1000 genealogies followed by five long chains with 10,000 genealogies and a burn-in of 10,000.


Multiple comparisons were subject to Bonferoni corrections, for clarity the uncorrected P equivalent of the Bonferoni corrected P is presented.


Patterns in habitat use

There was significant variation in the catch rate of charr between lochs from all habitat types (F1,22 = 3.77; P < 0.001) (Fig. 4). CPUE (charr per 100 m2 per net per 24 h) ranged from 25 at Loch Doine to 1.5 at Loch Coulin. In addition, there were differences in habitat use between different lochs. For most lakes, charr catches were most abundant in the littoral and sublittoral zones (Fig. 5). However, in lochs Awe and More the profundal zone, and for lochs Langavat and Stack the pelagic zone catches dominated those of other habitat types.

Figure 4.

 Total catch per unit effort (CPUE) for charr from all habitat types at each loch. aG, a'Ghriama; Ae, Awe; Cr, Clair; Cn, Coulin; De, Doine; Dn, Doon; En, Earn; Ek, Eck; Ga, Grilsta; Lt, Langavat; Le, Lee; Lg, Lubnaig; Mr, Maree; Mt, Mealt; Md, Merkland; Me, More; Rh, Rannoch; Sn, Shin; Sk, Stack; Ta, Talla; Ue, Uaine; Vl, Voil.

Figure 5.

 Habitat use by charr in each of the study sites, the charr catch per unit effort (CPUE) expressed as a percentage of the total CPUE, from the littoral, sublittoral, profundal and pelagic zones separately. aG, a'Ghriama; Ae, Awe; Cr, Clair; Cn, Coulin; De, Doine; Dn, Doon; En, Earn; Ek, Eck; Ga, Grilsta; Lt, Langavat; Le, Lee; Lg, Lubnaig; Mr, Maree; Mt, Mealt; Md, Merkland; Me, More; Rh, Rannoch; Sn, Shin; Sk, Stack; Ta, Talla; Ue, Uaine; Vl, Voil.

Broad-scale country-wide phenotypic variation

There was significant variation in head morphology across the populations examined in this study. Amongst the 22 populations examined here there was significant, between-population variation in both PC1 scores (F1,22 = 29.1; P < 0.001) and in PC2 scores (F1,22 = 20.1; P < 0.001) (Fig. 6).

Figure 6.

 Loch mean and standard error of principal component 2 (PC2) versus PC1 scores of head morphology of charr. aG, a'Ghriama; Ae, Awe; Cr, Clair; Cn, Coulin; De, Doine; Dn, Doon; En, Earn; Ek, Eck; Ga, Grilsta; Lt, Langavat; Le, Lee; Lg, Lubnaig; Mr, Maree; Mt, Mealt; Md, Merkland; Me, More; Rh, Rannoch; Sn, Shin; Sk, Stack; Ta, Talla; Ue, Uaine; Vl, Voil.

Pairwise analysis of component scores for PC1 showed differences between 109 of 210 loch pairs (52%); for PC2 score 100 of the 210 loch pairs (48%) were significantly different (Table 2).

Table 2.   Pairwise loch comparisons (Bonferroni corrected) of PC1 (below the diagonal) and PC2 (above the diagonal) scores
  1. Blank cell = not significant; PC, principal component; *P < 0.05; **P < 0.0001.

  2. aG, a'Ghriama; Ae, Awe; Cr, Clair; Cn, Coulin; De, Doine; Dn, Doon; En, Earn; Ek, Eck; Ga, Grilsta; Lt, Langavat; Le, Lee; Lg, Lubnaig; Mr, Maree; Mt, Mealt; Md, Merkland; Me, More; Rh, Rannoch; Sn, Shin; Sk, Stack; Ta, Talla; Ue, Uaine; Vl, Voil.

aG  ****  ** **  **  **  ** ****
Ae** ** *  * **  **  **  ** ****
Cr *    **   **  ******     
Cn **    *   **  * ***     
De**  *  *   **  ** ****   * 
Dn *       *     *** * ** 
En********** ** **  **  **  *******
Ek** **** ****   **  *** **   * 
Ga **  *  **  *    **     
Lt **  ** ****  **** **** *** *  
Le**     *****  **  ** *********
Lg **  * ****    *  *** ** ***
Mr*     **      *** **     
Mt **    ****       **  *******
Md **** **** **  ** *** *** ** ***
Me** ********* ******* **** **** **  
Rh **  * ****      *** **********
Sn *    ****      ***    ** 
Sk **  * ****       **      
Ta **** **** **  ******* ******  ** 
Ue** ********* ******* **** ********  
Vl **  ** ***  *    **    ** 

Catchment and subcatchment phenotypic variability

To examine phenotypic variation between and within catchments (between lochs) for the Shin and the Laxford systems, a one-way anova with lochs nested within catchments was used to look at variation in PC1 and PC2 scores.

PC1 scores of trophic morphology did not differ significantly between Shin and Laxford catchments (F1,186 = 1.55) but did differ significantly between lochs (F3,186 = 22.3; P < 0.001) (Fig. 7). Between-catchment variation comprised only 9.2% of the total variation in PC1 scores but between-loch variation comprised 32.7% of variation.

Figure 7.

 Loch mean and standard error of principal component (PC2) versus PC1 scores of head morphology of charr from the Shin and Laxford systems. aG, a'Ghriama; Md, Merkland; Me, More; Sn, Shin; Sk, Stack

For PC2, catchments explained a significant proportion of the variation in trophic morphology (50.0%) (F1,186 = 15.1; P < 0.05) with lochs explaining a smaller proportion of the variation in trophic morphology (6.6%; F3,186 = 6.8; P < 0.01) (Fig. 7).

Catchment and subcatchment genetic variability

Analysis of the microsatellite data confirmed the presence of significant genetic structuring within and between drainages. In addition, significant within-population structuring was evident in Loch Stack, with evidence of two genetic subgroups (nominally Stack A and B) which are subsequently dealt with separately here. All pairwise estimates of FST were significant at P < 0.001, while higher levels of differentiation were found between lakes of different drainages (mean FST = 0.181) than between lakes within drainages (means of 0.067 and 0.145 between populations in the Shin and Laxford drainages, respectively) (Table 3). Pairwise estimates of RST were slightly higher but showed the same qualitative pattern with means of 0.336 (between populations from different drainages), 0.113 (Shin drainage) and 0.198 (Laxford drainage). All RST estimates were significant at P < 0.001 except the estimate between lochs Shin and a'Ghriama (RST = 0.015, P = 0.09). The hierarchical amova analysis based on allele frequencies indicated that, over all loci, 82.4% of the total genetic variance was found within populations (Table 4). The among-population-within-drainage component and the among-drainage hierarchical level both accounted for similar proportions of the variance (9.52% and 8.11%, respectively). The corresponding Phi-statistics were highly significant, including the estimate of ΦCT = 0.081 (P < 0.001) indicating the presence of significant differentiation between the two drainages. With an assumed SMM, a greater proportion of the variance was assigned at the among-drainage level (21.5%) and corresponding ΦCT was higher although marginally nonsignificant (ΦCT = 0.215, P = 0.074).

Table 3.   Pairwise estimates of FST (below diagonal) and RST (above the diagonal).
 MoreStack AStack BMerklanda'GhriamaShin
Stack A0.0600.3520.5270.4340.409
Stack B0.1950.1800.2390.1950.225
Table 4.   Hierarchical analysis of microsatellite molecular variance.
 Among regionsAmong populations within regionsWithin populations
  1. Associated with each hierarchical level are; components of molecular variance (V ), percentage of total variance explained (%), estimates of statistical significance (P), F-statistics [for Infinite Alleles Model (IAM)-based analyses using allele frequencies only] and R-statistics [for Stepwise Mutation Model (SMM)-based analyses with allelic size information incorporated].


Table 5 shows the estimated effective number of migrants per generation between each of the pairs of lochs in the Laxford and Shin catchments and between the two genetic groups within Loch Stack. The effective migration rate between all populations is extremely low, ranging from a maximum of 2.7 (between Stack A and More) to 0.067 (between Stack A and Stack B).

Table 5.   Estimates of the effective number of migrants Nexmyx between charr populations within drainages (95% confidence limits are shown in parentheses).
Receiving population (x )Migrant source (y )
Laxford drainageShin drainage
MoreStack AStack BMerklanda'GhriamaShin
More2.755 (2.455–3.091)0.586 (0.444–0.745)   
Stack A1.853 (1.648–2.075)0.581 (0.47–0.707)   
Stack B0.262 (0.204–0.33)0.067 (0.04–0.105)   
Merkland   1.169 (1.026–1337)0.328 (0.256–0.411)
a'Ghriama   1.058 (0.906–1.162)2.077 (1.878–2.288)
Shin   1.166 (1.013–1.334)1.724 (1.514–1.943)


Recent microsatellite analysis of the genetic structure of Arctic charr populations across north-west Europe, has revealed clear between-population genetic structuring, high levels of genetic diversity and significant genetic differentiation among populations (Wilson et al. 2004).

Here we show that this genetic variation is also matched by significant phenotypic variation across the Arctic charr from a broad geographical area (Scotland). The evidence from this study is that variation in head and mouth morphology is discrete, differing significantly among the majority of the populations examined here.

Microsatellite analysis has proven to be a useful tool for the identification of distinct gene pools and is generally thought to be examining a part of the genome that is subject to very low levels of direct selection. In contrast, the trophic morphology of fish in general and of charr in particular has been shown to have a direct effect on foraging ecology. For example, in sympatric polymorphic systems, variation in head and mouth shape correlates with variation in foraging ecology (Snorrason et al. 1994; Adams et al. 1998; Fraser et al. 1999). In laboratory experiments, variation in trophic morphology has also been shown to affect prey handling ability in charr (Skúlason et al. 1993; Adams & Huntingford 2002a) and this variation has been shown to be inherited, at least in some populations (Skúlason et al. 1996; Adams & Huntingford 2002b, 2004; Klemetsen et al. 2002). Thus, the variation in trophic morphology shown across Scotland in this study would strongly suggest that this phenotypic radiation is adaptive.

Also evident in this study, is that trophic morphology can differ significantly over very short geographical distances. Trophic morphology variation was as great within catchments (between lakes) as it was between catchments for the Laxford and Shin systems examined here. This finding was also supported by the pattern of microsatellite genetic variation within and between catchments, which showed clear and distinct discrimination between lakes from different catchments as well as between lochs within catchments. Thus charr from lakes within the same catchment, where there are no evident barriers to fish movement (for example, between lochs Stack and More in the Laxford catchment and between lochs Shin, a'Ghriama and Merkland in the Shin catchment) differ significantly in trophic morphology and represent different gene pools. The ultimate mechanisms through which this phenotypic and genetic structuring could have occurred are uncertain, however, there is strong evidence that Scottish and other northern European populations resulted from a postglacial invasion from a single refugium (Brunner et al. 2001; Wilson et al. 2004) and that diversification has occurred during the postglacial period (Gíslason et al. 1999). This study does, however, suggest one mechanism through which between-population variation may be maintained. Estimates of effective, interbreeding migrants between populations even within the same catchment were very low (one to two per generation) thus each loch population is operating as a separate gene pool even within systems with interconnected lochs in the same catchment.

The pattern of both phenotypic variation shown here and genetic variation on large and small spatial scales (also see Wilson et al. 2004) strongly supports the conclusion that extensive differentiation between populations has occurred across Scotland, such that even neighbouring populations may differ significantly. This mosaic of phenotypic and genetic diversity has a number of specific consequences and challenges for conservation.

Firstly, it is clear that the genetic and phenotypic integrity of populations across Scotland is very high and that the mixing of populations that has occurred as a result of ‘management’ practice in other species has not occurred or has not been successful to any significant extent within Scotland. This is probably because until now, Arctic charr has been largely ignored within Scotland as an angling and culture species (Maitland 1995).

Secondly, the extensive adaptive radiation in phenotype and genetic diversity of Arctic charr within Scotland, contributes significantly to the effective biodiversity of our northern freshwater systems. Notwithstanding the current single species status (although see argument in Kottelat 1997), the Arctic charr is clearly a genetic and phenotypically different fish in different places that also differs in its functional role in ecosystems.

A third conclusion that follows from the pattern of genetic and phenotypic variation is that a focus on the species as the unit for conservation for Arctic charr would be highly inappropriate. Any conservation strategy for Arctic charr to be successful in conserving the essential elements of the nature of the Arctic charr would need to recognise the importance of individual populations.

Lastly, it is clear that the management needs of charr in different populations may differ significantly. In this study, we show very high degrees of variation in traits with a functional significance for foraging and significant habitat use differences by charr from different populations. We also know that spawning requirements may differ between populations (see Walker 2006). Thus the specific requirements for good habitat quality may differ substantially between populations, making definition of good management practice for the whole single species impossible.


We thank Alan Grant, Nicola Bissett, Deborah Hamilton and Alastair Duguid for technical support. Field work was funded by EC grant FAIR CT96-1981. This study was facilitated by the EU Charrnet Programme, EC QLRI CT-2001-00007.