Temporal stability of individual feeding specialization may promote speciation


Correspondence author. E-mail: rune.knudsen@uit.no


1. Inter-individual differences in trophic behaviour are considered important in the disruptive selection process for resource specialization and may represent an early phase in the evolution of polymorphism and adaptive radiation. Here, we provide evidence of high stability of individual trophic niches of a fish predator from a 15-year study.

2. Individual resource specialization was investigated by combining data from analyses of stomach contents (recent trophic niche), trophically transmitted parasites (long-term niche) and trophic morphology (niche adaptations) from single specimens of a postglacial fish (Arctic charr) population sampled from contrasting pelagic and littoral habitats.

3. Based on the relationships between morphology, parasites and diet, high inter-individual temporal consistency of narrow niches (zooplanktivorous vs. benthivorous) was evident through the ontogeny of the charr, indicating low degree of switching both in habitat utilization and feeding strategy of individual fish. Co-occurrence of differently specialized behavioural phenotypes was sustained over multiple generations.

4. The stable long-term habitat and feeding specializations may represent an important initial step in an adaptive radiation process, and our findings suggest a case of sympatric speciation into two incipient forms diverging along the littoral–pelagic resource axis.


There is presently a strong theoretical foundation for the importance of ecological mechanisms as drivers of sympatric speciation (Dieckmann & Doebeli 1999; Abrams, Rueffler & Kim 2008; Pennings et al. 2008). A key factor amongst the ecological processes promoting evolution in both theoretical and empirical studies of ecological driven evolution is connected to niche specialization (i.e. Schluter 2000, 2001). Foraging behaviour, combined with resource competition, is thought to be important in the disruptive selection process for resource specializations (Ackermann & Doebeli 2004; Bürger, Schneider & Willensdorfer 2006; Rueffler, Van Dooren & Metz 2007), and a correlated driver in the selection of morphological or other secondary traits among vertebrates (Skúlason, Snorrason & Jonsson 1999; Streelman & Danley 2003; West-Eberhardt 2005). Inter-individual trophic niche differentiation (e.g. dietary and habitat specialization) may thus represent an early phase in the evolution of polymorphism (Skúlason & Smith 1995; Bolnick et al. 2003; Ackermann & Doebeli 2004), and may be particularly important for existing models of evolution as it provides a possible mechanism for sympatric speciation (Dieckmann & Doebeli 1999; Abrams 2006; Rueffler et al. 2007). Temporal stability in individual foraging specialization is a central empirical issue that remains largely unaddressed (but see Tinker, Bentall & Estes 2008).

Resource polymorphism related to habitat and dietary specialization along the littoral–pelagic resource axis is commonly observed in northern postglacial freshwater fish populations, both within single populations and between sympatric morphs (Skúlason & Smith 1995; Robinson & Parsons 2002; Bolnick & Lau 2008). The Arctic charr Salvelinus alpinus (L.) is an excellent model organism for studying individual feeding specialization as this behaviour has been indicated both from field and experimental studies (Amundsen et al. 1995; Knudsen, Klemetsen & Staldvik 1996; Bolnick et al. 2003). In addition to longitudinal feeding data (i.e. repeated diet data from single specimens), Bolnick et al. (2003) suggested three tentative methods through which the temporal stability of individual foraging specialization might be explored from correlations with the individual resource use, including (i) stable isotope ratios of tissue from the foraging individual, (ii) food borne parasite fauna or (iii) variation of functional morphology. Trophic morphology and body form of fish, including charr, are generally found to be closely associated with individual resource acquisition and have both genetic and plastic components (e.g. Adams & Huntingford 2002; Knudsen et al. 2007; Wilson & McLaughlin 2007). The gain of food transmitted parasites is also closely related to the past food acquisition of individual fish hosts (Amundsen et al. 2003; Knudsen, Curtis & Kristoffersen 2004).

The Arctic charr of the subarctic lake Fjellfrøsvatn has been studied for a period of 15 years, including analyses of diet, parasites and morphology (Knudsen et al. 2004, 2007; Amundsen, Knudsen & Klemetsen 2008). The lake has two reproductively isolated and genetically distinct morphs of charr (Klemetsen et al. 1997; Westgaard, Klemetsen & Knudsen 2004; Knudsen et al. 2006), a small-sized, cryptic morph that spawns in deep water in the winter and a common morph that spawns in shallow water in the autumn. They are termed profundal charr and littoral charr after their spawning habitats. Recently, Knudsen et al. (2007) suggested that the littoral morph may be in the process of incipient sympatric segregation along the benthic–pelagic resource axis.

In the present approach, we expanded the analysis of data from the littoral charr population by relating recent (diet and habitat) and long-term (accumulation of food transmitted parasites) niche use to the functional morphology of individual fish. This novel combination of the long-term data allowed further testing of the hypothesis of incipient sympatric segregation. We predicted that (i) individual fish adopt temporally stable foraging specializations (zooplanktivorous or benthivorous) throughout their lifetime and (ii) divergent foraging behaviour remains stable over multiple generations.

Materials and methods

Fjellfrøsvatn (69°05′N, 19°20′E, 125 m a.s.l) is a dimictic, oligotrophic lake situated in a tributary to the Målselv river, subarctic Norway. It has a surface area of 6.5 km², a maximum depth of 88 m and well-developed pelagic and littoral habitats. The littoral zone (<10 m) constitutes slightly less then one-third of the surface area. The only fish species present are Arctic charr and brown trout (Salmo trutta L.) that had an opportunity to enter this post-glacial lake from anadromous ancestors only within a short-time period (see Klemetsen et al. 1997 for more details). The presence of the reproductively isolated profundal and littoral morphs (Klemetsen et al. 1997) is unique to Fjellfrøsvatn in the Målselv drainage. No other upstream or downstream lakes have similar polymorphisms. Our definition of these morphs is made according to their known reproductive isolation in both time and place of spawning (see Klemetsen et al. 1997). Ecotypes within the littoral morph are defined according to each individual trophic niche use (i.e. benthivorous or zooplankivorous) because of the uncertainty regarding these groups as partial reproductive isolated.

Arctic charr were sampled by multi-mesh gill-net (mesh size 10, 12.5, 15, 18.5, 22, 26, 35, 45 mm, knot to knot) from the littoral (benthic gill-nets, 0–10 m) and pelagic zones (offshore floating gill-nets, >30 m depth). Further details of fish sampling and general charr biology is given by Knudsen et al. (2004). Sampling was conducted from September to November in 1992 and then annually during late September or early October from 1996 to 2006. To standardize the size composition between years, adult Arctic charr from 180 to 260 mm (fork-length) size-range (mainly 4–6 year old) were selected for the analyses. Prey items in the stomach contents of individual fish were identified and their relative contributions to total stomach fullness estimated according to Amundsen, Gabler & Staldvik (1996). Food items were categorized as limnetic (zooplankton and surface insects, i.e. emerging insect pupae and adult insects) or benthic prey (benthic crustaceans, molluscs, insect larvae). Individual charr were categorized as either zooplanktivore or benthivore based on the dominance (>50% contribution) of limnetic or benthic prey in the stomach contents. Additionally, fish with a dominance of Gammarus lacustris (Sars) in their stomach (>50%) were defined as Gammarus feeders.

Almost all fish were examined for parasites. Two highly positive correlated parasites species (Cystidicola farionis Fisher and Cyathocephalus truncatus Pallas, 1781) transmitted from amphipods (here G. lacustris) only (Knudsen et al. 2004), were used as biological markers of the former benthivore (or gammaridivore) diet of their fish hosts. We used autumn samples of parasites as a picture of accumulated diet specialization that has occurred during the ice-free growth season prior to the capture of the fish. The swim bladder nematode C. farionis was sampled on every occasion (in 1992 and yearly from 1996 to 2006), enumerated and identified as L3-larvae (recruitment stage; ∼2 months development to L4), preadults (L4-larvae) or adults. These nematodes have a lifespan of several years in their fish hosts (Knudsen, Amundsen & Klemetsen 2002). The short-lived intestinal cestode C. truncatus, which has a longevity of only 2 months in their hosts (Amundsen et al. 2003; Knudsen et al. 2008), was studied only in 1992, 1997 and 2006. The infrapopulation size (the number of parasites in an individual fish) of C. truncatus is considered as a marker of the diet of the individual fish during the latest ice-free season, whereas the long-lived C. farionis is an ecological marker of the diet throughout the life time of the fish host. To analyse the relationships between diet and parasite infection in individual fish we performed constrained correspondence analyses (CCA) followed by Monte Carlo permutation tests (Legendre & Legendre 1998).

In a single year of the study (1997), we also compared the diet (i.e. the most recent foraging niche) with food-borne parasites and trophic morphology (i.e. past niche) of each individual specimen to examine temporal consistency of behavioural specializations. Here, we defined trophic morphology as head- and jaw structures, and different body measurements and fin sizes according to Adams & Huntingford (2002). The morphometric variables were chosen on the basis of functional considerations corroborated by comparative studies on charr (Bertrand, Marcogliese & Magnan 2008). Morphometric measurements were corrected for body size using residuals from regression of morphometric variables against fork length. Morphometric characteristics were first inspected by principal component analysis (PCA) and indirectly related to diet. Thereafter, we used a CCA to model the relationships between resource use, morphometry and food-borne parasites, followed by Monte Carlo permutation test (Legendre & Legendre 1998). This was performed to obtain an additional indication of the degree of individual separation into benthivore or zooplanktivore ecotypes. All statistical analyses were performed using the statistical software r (package Vegan 1.8-8, by Jari Oksanen).


Diet and habitat specialization

There was a consistent pattern of diet choice within fish caught in the littoral and pelagic habitats, respectively, both through the ontogeny (left panels) of the Arctic charr and throughout the whole 15 years (right panels) study period (Fig. 1). Nearly, all pelagic-caught charr (98.6 %; = 322) were limnetic feeders and had only low quantities of benthic prey in their diets (see Fig. S1, Supporting information). The littoral-caught charr segregated in contrast into a benthivore and a planktivore group feeding mainly on benthic and limnetic prey respectively.

Figure 1.

 Proportion of benthic prey items as a total of all stomach content through the (a) ontogeny (different age groups; in 1992) and (b) over the 15-year study period in Arctic charr collected from the littoral and pelagic zones of Fjellfrøsvatn (= number of charr).

Parasites as long-term trophic behavioural indicators

Among the pelagic-caught fish, the abundance of both the short-lived C. truncatus and long-lived C. farionis (Fig. 2a) was low compared to the littoral caught charr throughout the ontogeny of their hosts. Furthermore, in all sampling years during the period from 1992 to 2006 the littoral-caught specimens had a significantly higher infection of C. farionis than the pelagic caught charr (Fig. 2b). Parasites were also used as markers of the consistency of individual feeding specialization within the littoral zone (Fig. 3). There were generally significant differences in the infection of both C. truncatus and C. farionis between Gammarus feeders and benthivores and zooplanktivores charr from the littoral zones in the two time periods (1992 and 1996–2006) with available data (Fig. 3). Similarly, in each of the years (1992, 1997, 2006), a CCA relating infection by food-borne parasites to diet produced very similar results (see Fig. S2a–c, Supporting information). The first ordination axis separates fish feeding on littoral benthic prey (on the right) from those feeding on pelagic prey (on the left). The diet gradient was significantly correlated to all of the parasite species and stages (Monte Carlo tests, < 0.05). The short-lived species C. truncatus and time-span stages (L3-larvae) of C. farionis showed similar patterns, and were highly correlated with most benthic prey, especially with Gammarus. Furthermore, the long-lived C. farionis (preadult/adult worms) also showed comparable results (Fig. 3, S2a–c, Supporting information), reinforcing the high consistency between the most recent benthivore diet and the past benthivore diet niche over several years in single specimens.

Figure 2.

 The abundance (±SE) of cestodes (Cyathocephalus truncatus) and swim-bladder nematodes (Cystidicola farionis) transmitted by Gammarus lacustris in Arctic charr caught in the littoral (solid line) and pelagic (dotted line) habitats (a) by age class of the fish (in 1992) and (b) over the 15-year study (C. farionis only) period (= number of charr).

Figure 3.

 The abundance (±SE) of cestodes (Cyathocephalus truncatus) and swim-bladder nematodes (Cystidicola farionis) transmitted by Gammarus lacustris in Arctic charr caught in the littoral and pelagic habitats, and according to foraging groups (zooplanktivore; benthivore; Gammarus-feeders) within the littoral zone over two time periods (1992 and 1996–2006) in Fjellfrøsvatn [= number of charr (lower panel: = data 1997 and 2006/yearly data 1996–2006); statistical significance MW U tests: *< 0.01; **< 0.001].

Relationships between trophic niche, morphological traits and food-borne parasites

The first two axes in the PCA of charr morphology accounted for 61.5% of the variation (Fig. 4a). The morphological traits were positively associated with the degree of benthos in the diet, as shown by the generalized additive model surface fitted to the two ordination axes (superimposed isolines in Fig. 4a). Fish feeding on benthic prey (grey symbols) had more robust heads, and deeper bodies compared with non-benthivore feeders (open symbols) from the littoral (triangles) and the pelagic (circles) habitats. This morphological pattern relative to diet is also evident within the littoral zone. The trophic morphology (PC1 axis in Fig. 4a) was significantly correlated with the proportion of benthos in the diet (R² = 0.27; < 0.0001; Fig. 4b), the infestation of short-lived parasites/stages (C. truncatus and larval stage of C. farionis: R² = 0.35; < 0.0001; Fig. 4c), and infestation of long-lived parasites (adult stage of C. farionis: R² = 0.09; < 0.01; Fig. 4d) of individual fish. A CCA confirmed that the benthic–limnetic diet-gradient (different prey types) was significantly (Monte Carlo test, < 0.005) associated with variation in morphological traits and also corresponded nicely with the density of parasites (see Fig. S3a,b, Supporting information). This pattern was independent of which of the food-borne parasite species (C. truncatus or C. farionis) or stages (C. farionis) that were used in the analyses (Fig. S3a,b, Supporting information).

Figure 4.

 Biplot (a) of principal component analysis results for trophic morphological traits with superimposed isolines (based on regression surface) of the percentage of benthos in the diet of Arctic charr (= 88) caught in littoral (triangles) and pelagic (circles) habitats. Each individual charr are classified as either benthivorous (grey symbols) or zooplanktivorous (open symbols) from their diet preferences. Large symbols represent centroids (bars, ±95% CI) for each dietary group. Relationships between trophic morphology (PC1 axis from a) and (b) diet (percent benthos in stomach content), (c) density (log transformed) of short-lived parasites/stages (Cyathocephalus truncatus and larval stage of Cystidicola farionis), (d) density (log transformed) of long-lived parasites (adults of C. farionis) for each individual fish. Abbreviations of morphological measurements: head height behind gills (Hg), head length (Hpl), jaw length (Jl), pectoral fin length (Fp), body width at dorsal fin (Bdw), dorsal fin width (Fdw), caudal height (Bch). See Supporting information Figs. S2 and S3 for more details.


We found that individual Arctic charr within the population had a highly specialized foraging behaviour throughout their ontogeny. The consistent individual specialization that was repeated over the whole 15-year study period indicates a stable, long-term foraging specialization with close habitat-diet associations. This was supported by differences in trophic morphology and food-web transmitted parasites as ecological markers of former diet niches. The morphological differences was clearly associated to diet of individual fish and not only the habitat choice, similarly also found among sticklebacks (Araújo et al. 2008). The morphology can also be explained functionally and appear to be related to inter-individual niche preferences as the benthivorous ecotype have a deeper body form and larger and broader jaws and head shape than the zooplanktivorous ecotype, which in turn have a more slender body (see also Knudsen et al. 2007 for a more detailed discussion of these aspects). There was a clear segregation in diet along the benthic–pelagic resource axis as the charr caught in the pelagic zone were almost exclusively zooplanktivores whereas the fish caught in the littoral zone mainly diverged into benthivorous and zooplanktivorous trophic niches. Only few fishes had a mixed diet. Although specialization in individual behaviour is commonly observed (Bolnick et al. 2003), the long-term persistence in individual foraging specialization has until now largely been unaddressed (Tinker et al. 2008; Woo et al. 2008). Temporal consistency in individual feeding specialization may expectedly have both important ecological and evolutionary implications for the population (Bolnick et al. 2003). Thus, our findings support models of ecological and evolutionary dynamics focusing on processes at the level of individuals (Bolnick et al. 2003; Rueffler, Egas & Metz 2006).

To be able to promote sympatric speciation, it is reasonable to assume that inter-individual differentiation into two contrasting resource niches should be persistent throughout the ontogeny of individuals (Dieckmann & Doebeli 1999) and remain stable over multiple generations in contrast with Taylor et al. (2006). These conditions seemed to be met in Arctic charr, as indicated by the evidence on morphology, parasite acquisition and prey selectivity. The composition and abundance of food-borne parasites that track past niche use were consistent with the most recent trophic niche as indicated by the habitat and diet use at the time of catch. This niche-stability persisted over the last months prior to catch as indicated by the intensity of the short-lived C. truncatus. In addition the infrapopulation size of the long-lived C. farionis, which accumulates over the lifespan of the fish hosts (Knudsen et al. 2002; Amundsen et al. 2003), showed trophic niche stability over a period of years. A long-term habitat and feeding specialization was also evident from the fact that all age-groups of zooplanktivorous pelagic fish had low infection rates of both Gammarus-transmitted parasite species. This suggests that individuals with low feeding frequency on Gammarus enter the pelagic zone and reside there feeding on zooplankton throughout the ice-free season, and also proceed with this behaviour year after year. Furthermore, the observed more slender body form of the zooplanktivores and the more robust heads of the benthivores is a general pattern frequently observed among fish (e.g. Schluter 2000; Robinson & Parsons 2002) and in the present Arctic charr population (see Knudsen et al. 2007 for functional relevance), which likely promotes the diversification into benthivorous and zooplanktivorous resource specialists.

Mechanisms promoting inter-individual variation in feeding specialization are often related to variability and plasticity in either behavioural or morphological traits (Bolnick et al. 2003). In the absence of morphological diversification, behavioural traits seem important (De Kerckhove, McLaughlin & Noakes 2006; Rueffler et al. 2007). Behavioural traits are very flexible (West-Eberhardt 2005; Wilson & McLaughlin 2007), but evidently highly stable for each individual predator according to the inferred foraging behaviour (habitat and diet choice). Beside high plasticity, planktivorous and benthivorous foraging behaviours are also heritable traits (Adams & Huntingford 2002; Klemetsen et al. 2002, 2006). Furthermore, the present benthivorous and zooplanktivorous ecotypes were found to differ in trophic morphology and body shape related to their individual trophic niches (see also Knudsen et al. 2007, 2008); traits that appear to be strongly heritable in Arctic charr (Adams & Huntingford 2002; Klemetsen et al. 2002). Such morphological specializations to exploit contrasting benthic or limnetic diet resources within a single gene pool population may have an origin from both phenotypic plasticity and heritability (e.g. Adams & Huntingford 2002; Prolux & Magnan 2004; Svanbäck & Persson 2004), and probably promotes disruptive selection for stable divergent morphological traits (i.e. Skúlason et al. 1999; Rueffler et al. 2007). Thus, the origin of individual niche specializations seems to be an interplay in plasticity of both behavioural and morphological adaptations with the persistent behaviour component as a key factor.

The diversification along a benthic–pelagic resource axis is a recurrent pattern of resource polymorphism in fish from postglacial lakes (e.g. Schluter 2000; Robinson & Parsons 2002; Østbye et al. 2006). The stability of these distinct foraging niches suggests selection against generalist feeders. Trade-offs related to resource utilization of different prey types are evident among genetically differentiated sympatric zooplanktivorous and benthivorous morphs (Schluter 2000). Furthermore, Knudsen et al. (2007) suggested that niche-based assortative mating on specific diet-morphological mediated traits were seen among the present ecotypes, as also observed in an unimodal population and among several species pairs of postglacial fishes (Boughman, Rundle & Schluter 2005; Snowberg & Bolnick 2008). Model studies stress the importance of assortative mating for the completion of reproductive isolation through arbitrarily small steps (e.g. Doebeli et al. 2007; Pennings et al. 2008). Recently, Pennings et al. (2008) proposed partial isolation as a possible stable outcome of frequency-dependent disruptive selection instead of speciation. If so, the stability of the present Arctic charr ecotypes suggests that incipient speciation is presently taking place along the benthic–pelagic resource axis. This may, however, be a labile evolutionary situation as a collapse from diverged ecotypes to a hybrid swarm is found in both theoretical and observational studies (e.g. Taylor et al. 2006; Pennings et al. 2008). In an Arctic charr population in a lake close to Fjellfrøsvatn, either a lack of ecological opportunities or a loss of previous polymorphism has probably occurred due to resource competition from a fish assemblage with several benthivore competitors (Knudsen et al. 2007).

Postglacial lakes provide excellent cases for studying early evolutionary processes as they are young in evolutionary time, with low species packing and vacant niches. The profundal morph of Fjellfrøsvatn, not included in this study, is strictly specialized to exploit the deep-water soft bottom resource niche (Klemetsen et al. 2002, 2006; Knudsen et al. 2006). Knudsen et al. (2006) suggested that it originated sympatrically from the ancestral invader through niche expansion. The complete reproductive isolation, the very distinct phenotype and the genetic differences suggest that this was an early event in the postglacial history of the lake. Thus in Fjellfrøsvatn, there is a possible incipient adaptive radiation into a three morph situation: one old and well-segregated profundal morph and two incipient forms (zooplanktivorous and benthivorous ecotypes). A similar three-morphs system, utilizing each of the main habitats (i.e. littoral, pelagic, profundal habitats) in postglacial lakes, has recently been documented among European whitefish (Kahilainen & Østbye 2006). Trimorphism is also theoretically considered as a plausible scenario (Abrams 2006; Bolnick 2006; Bürger et al. 2006). Skúlason et al. (1999) proposed four stages in the diversification process of fishes, starting with behavioural specialization (i) subsequently manifested in trophic morphology (ii) in association with assortative mating (iii) that finally may lead to reproductive isolation and genetic fixation of traits (iv). Thus, according to the evolutionary stages of Skúlason et al. (1999), the profundal morph in Fjellfrøsvatn has evidently full-filled the incipient speciation process with reproductive isolation and genetically different morphs as a result (Klemetsen et al. 2002, 2006; Westgaard et al. 2004; Knudsen et al. 2006). For the two ecotypes addressed in this study, an incipient speciation process is apparently still in progress.


We thank staff and students of the Freshwater ecology group at the University of Tromsø for field assistance through many years, Laina Dalsbø and Cesilie Lien for laboratory assistance and Prof Colin Adams, University of Glasgow, and Grethe Robertsen for comments on the manuscript.