Asymmetrical habitat coupling of an aquatic predator—The importance of individual specialization

Abstract Predators should stabilize food webs because they can move between spatially separate habitats. However, predators adapted to forage on local resources may have a reduced ability to couple habitats. Here, we show clear asymmetry in the ability to couple habitats by Eurasian perch—a common polymorphic predator in European lakes. We sampled perch from two spatially separate habitats—pelagic and littoral zones—in Lake Erken, Sweden. Littoral perch showed stronger individual specialization, but they also used resources from the pelagic zone, indicating their ability to couple habitats. In contrast, pelagic perch showed weaker individual specialization but near complete reliance on pelagic resources, indicating their preference to one habitat. This asymmetry in the habitat coupling ability of perch challenges the expectation that, in general, predators should stabilize spatially separated food webs. Our results suggest that habitat coupling might be constrained by morphological adaptations, which in this case were not related to genetic differentiation but were more likely related to differences in individual specialization.

In addition, the ability of predators to couple distinct aquatic habitats can be limited by factors like temperature (Tunney, McCann, Lester, & Shuter, 2014), habitat size and morphometry (Dolson et al., 2009;Eloranta et al., 2015), and interspecific competition (Eloranta, Knudsen, & Amundsen, 2013;Vander Zanden, Casselman, & Rasmussen, 1999). Thus, the dynamics of habitat coupling are complex and may depend on a combination of external as well as population and individual level factors, such as individual specialization.
Individual specialization, where individuals use a small subset of a population's resource base, is common in many animal taxa (Bolnick et al., 2003). The causes of individual specialization can be an individual's preference for resources, food resource use efficiency related to morphology, or their behavioral or physiological ability to handle resources (Smith & Skúlason, 1996;Svanbäck & Eklöv, 2003). In addition, intraspecific competition can affect prey densities and drive individual specialization, leading to variation in net increases in energy, individual fitness, and even fine-scale genetic population differentiation (Adams et al., 2006;Bolnick et al., 2003;Gerlach, Schardt, Eckmann, & Meyer, 2001;Smith & Skúlason, 1996;Svanbäck & Bolnick, 2005). For example, site fidelity, behavioral traits, spawning time, and kin preference have all been associated with population differentiation over much smaller scales than the dispersal ability of fish (Behrmann-Godel, Gerlach, & Eckmann, 2006;Gerlach et al., 2001;Smith & Skúlason, 1996). Therefore, the individual specialization of predators has the potential to affect their habitat coupling ability to such an extent that local adaptation and/or genetic differentiation occur.
A predator that is known to exhibit strong dietary and morphological trade-offs between different habitats is the Eurasian perch (Perca fluviatilis; hereafter perch) (Svanbäck & Eklöv, 2003).
Perch have a more elongated body in pelagic habitats, where open water favors stronger/faster swimming ability to catch prey, whereas perch in littoral habitats have a deeper body, which is more suited to moving and foraging in complex environments Olsson & Eklöv, 2005;Svanbäck & Eklöv, 2003. The morphological variation and dietary trade-offs in perch have been well studied and are largely plastic responses to their environments (Faulks, Svanbäck, Eklöv, & Östman, 2015a;Olsson & Eklöv, 2005;Svanbäck & Eklöv, 2003). If morphological and genetic variations of perch affect their habitat use, then we would also expect these factors to be important for their ability to couple habitats.
Here, we aimed to investigate the associations between individual specialization, morphological adaptations, and/or genetic variation in the predator perch in relation to its ability to couple spatially separate habitats. We define habitat coupling as the ability of perch to feed on more than one spatially separated food web, where the dominance of consumed food resources from only one habitat would indicate low habitat coupling. We consider fluxes among spatially distinct habitats in predator and prey populations, and detritus and nutrients, as habitat linkages (sensu Schindler & Scheuerell, 2002). Due to potential local adaptations of perch, we assume that genetic and morphological adaptations to specific habitats and resources may limit movements between habitats and thereby constrain habitat coupling. Conversely, intraspecific competition can lead to increased individual specialization (IS) on resources that spans across habitats, thereby facilitating habitat coupling. This suggests that habitat coupling is determined by the relative strength of population differentiation constraining habitat coupling and individual specialization promoting habitat coupling. We spe- a substantial amount of pelagic zooplankton whereas the opposite pattern was not found (Scharnweber, Strandberg, Marklund, & Eklöv, 2016). Thus, it is possible that variation in resource preference can lead to asymmetries in habitat coupling.
The nets were set in the evening, left for approximately 12 hr, and collected the following morning. To maximize our chances of capturing fish from the littoral zone, nets were set at the lake bottom just outside the reed belt at a water depth of approximately 1.5 m where both the vegetation and the nets extended from the bottom to the water surface. Pelagic nets extended from the water surface and 6 m down in the water column. All fish were frozen immediately after retrieval from the nets and stored frozen until further processing. At each site, 30 perch of 100-140 mm in length were sampled to minimize age cohort variation.
This size range of perch is also appropriate to capture niche variation because perch typically undergo two ontogenetic niche shifts and all types of resources used by perch are covered within this size range; zooplankton, macroinvertebrates and fish (Persson, 1988) (see Supporting Information Table S1 for total perch abundance, catch per unit effort, and littoral north (five sites); pelagic habitat sites were divided into pelagic west (three sites) and pelagic east (three sites). In the laboratory, each fish underwent the following procedure: A digital photograph was taken; stomach contents were dissected; muscle tissue was sampled and dried for stable isotope analysis; and a fin clip was taken for genetic analysis (stored in 95% ethanol at −20°C until extraction).

| Habitat coupling assessed by habitat-specific diet and niche use
Stomach contents were analyzed using a dissecting microscope following the approach described in Scharnweber et al. (2016) and Svanbäck and Eklöv (2003). Prey were counted and identi- (h) crayfish, and (i) terrestrial fauna. The length of the first 10 prey items of each category was measured to the nearest 0.1 mm. The mean length of prey items was converted to biomass (mg dry mass) using our own length-mass regressions, and proportional diet data were calculated for each category and individual. These data were then used to calculate diet indices.
In order to test whether individual variation in niche was related to habitat coupling, we assessed individual specialization (IS), total niche width (TNW), and the between-individual component (BIC). IS ranges from 1, when all individuals use the full range of the population's niche, to 0, when individuals only consume a single prey type.
This means that when the index IS is at 1, individual specialization is at minimum, and when IS is at 0, the specialization is at its maximum and  Svanbäck, 2002;Roughgarden, 1972Roughgarden, , 1974.
Diet indices for each site were estimated using RInSp (Zaccarelli, Bolnick, & Mancinelli, 2013) 2016). Data were plotted and, if necessary, arcsine-square root transformed to achieve normal distribution. An ANOVA with sites nested within region and habitat was run to investigate the effects of habitat and region on diet specialization. Primer v 7.0.9 (Clarke & Gorley, 2015) with the PERMANOVA add-on package was used for all multivariate diet data analyses. Sites were nested within habitat and region, and all of these factors were treated as fixed effects. All diet data were arcsine-square root transformed (% dry mass), and ordinations were based on Bray-Curtis similarity. Unrestricted permutations (9,999 permutations) with type III sums of squares were used to test for the significance of models.

| Habitat coupling assessed by stable isotope analysis
Stable isotopes indicate the long-term resource use of pelagic and littoral perch and were used to indicate habitat coupling occurring over a longer period of time (Scharnweber et al., 2016). Analyses of δ 13 C and δ 15 N on perch muscle tissue and baseline samples were done at the Stable Isotope Facility, University of California, Davis (CA), using PDZ Europa ANCA-GSL elemental analyzer interfaced to a PDZ Europa 20-20 isotope ratio mass spectrometer (Sercon Ltd., Cheshire, UK). Samples were dried at 60°C for 48 hr, ground to a powder and weighed, and then placed into tin capsules (ca. 1 mg of tissue). Perch muscle samples were not corrected for lipid content due to their low C:N ratios (Pinnegar & Polunin, 1999;Post et al., 2007). A third of all perch were processed in duplicate, and the analytical error was 0.16‰ and 0.17‰ for δ 13 C and δ 15 N, respectively.
We defined the isotopic baseline at the time of sampling for each site by isotope values of Bivalvia, Gastropoda, and zooplankton.
For more details of sampling design and methods, see Supporting Information Appendix S1.
To assess the proportion of pelagic and littoral components in the diet of perch at each site, a Bayesian mixing model MixSIAR v 3.0.2 (Stock & Semmens, 2013) in R v 3.3.0 (R Development Core Team, 2016) was used, following the approach of Scharnweber et al. (2016). Fractionation factors of 0.4 ± 1.3 for δ 13 C and 3.4 ± 1.0 for δ 15 N were used to correct for trophic fractionation (Post, 2002). Prior to model runs, isotope biplots were inspected to validate the models (see Supporting Information Figure S1 for biplot of raw stable isotope values). We used Gelman-Rubin and Geweke diagnostics to check for chain convergence. Informative prior distributions were included based on a previous study of perch diet in the same study system (Bartels, Hirsch, Svanbäck, & Eklöv, 2012

| Habitat coupling assessed by morphological and genetic analysis
Perch were photographed on their left side with a bright background and under constant light conditions. Landmark-based thinplate spline analysis was used to characterize perch morphology.
These relative warp scores were used to calculate the Euclidean distance (ED) of each perch. This distance is the total vector distance between the centroid of the perch and the landmarks for the individual compared to the average perch within its site. Euclidean distances were analyzed by ANOVA in R with sites nested within region and habitat.
Body form changes were visualized using MorphoJ (Klingenberg, 2011). The digitized data from TPS-dig2 were imported into MorphoJ and checked for outliers. The data were size corrected for body size by regressing shape scores (Procrustes coordinates) against the centroid size following the approach outlined in Klingenberg (2016). The residuals of this regression were then used for all further analysis.
Canonical variate analysis (CVA) and discriminant function analysis (DFA) were used to assess the significance of shape differences (Mahalanobis distance) of perch between regions.
Peaks were binned automatically and manually checked. Genotyping errors due to allelic dropout and null alleles were screened using curtin.edu.au) (Raymond & Rousset, 1995;Rousset, 2008), using sequential Bonferroni correction for multiple testing where appropriate. None of the nine loci showed significant allelic dropout, null alleles, linkage disequilibrium, or deviations from Hardy-Weinberg equilibrium. However, one locus (SviL7) was excluded from further analysis due to missing data (genotypes missing at 66% of perch).
Of the total of 494 perch, 461 were successfully genotyped on the remaining eight loci.
As long-term isolation between sites was not expected, pairwise relatedness (r) Lynch and Ritland (1999) was also calculated using GenALEx v 6.5 (Peakall & Smouse, 2006. Mean relatedness within region was estimated in GenALEx. To further explore whether region was a barrier to gene flow, the distributions of relatedness between pairs of individuals within (e.g., pelagic east vs. pelagic east; littoral north vs. littoral north) and between (e.g., pelagic east vs. littoral north) regions were visually compared. If regions were a barrier to gene flow, a higher proportion of related individuals would be expected within rather than between regions.

| Diet
Of Reliance on pelagic resources was high in perch caught in all regions, but the pelagic reliance was highest in the pelagic regions (Figure 2b).
The total niche width in isotope space was highest in the littoral north region, followed by littoral south, pelagic west, and pelagic east.
Stable isotope standard ellipse areas differed among the regions, being higher in the littoral habitats indicating a wider range of resource use in these two regions (Littoral North: 3.15; Littoral South: 2.06) compared to the two pelagic regions (Pelagic West: 0.48; Pelagic East: 0.29). All ellipses showed substantial overlap, indicating common resources were used to some extent across the regions (Figure 2c).

| Diet specialization and niche width
Individual specialization was significantly higher (i.e., low IS values) in the littoral sites compared to the pelagic sites (ANOVA: F 1,12 = 46.07, p < 0.001, Figure 3a). There were also significant differences between regions (ANOVA: F 2,12 = 4.87, p < 0.05). Post hoc Tukey's HSD tests showed that the littoral north and littoral south regions were significantly different (p < 0.05), with higher individual specialization (i.e., low IS values) found in littoral south than littoral north (Figure 3a), whereas the two pelagic regions did not differ (p = 0.99). IS did not correlate with intraspecific competition (measured as perch abundance, see Supporting Information Table S1 for

(b) (c)
p < 0.05). Post hoc tests showed that differences between the littoral north and south regions were marginally not significant (p = 0.065).
The total niche width was largest in the littoral south followed by the littoral north. Total niche width was low in both pelagic regions.
There were significant differences in between-individual component (BIC) between littoral and pelagic habitats (ANOVA: F 1,12 = 44.13, p < 0.001, Figure 3c) and between regions (ANOVA: F 2,12 = 3.99, p < 0.05). Post hoc tests revealed that the differences between the littoral north and south regions were marginally not significant (p = 0.066). The between-individual component was highest in the littoral habitats and lowest in the pelagic habitats.  Table S2). Compared to pelagic perch, perch from the littoral regions had a more downward positioned snout and an operculum bone that was angled in a way to indicate a downward-pointing snout (negative CV1 values; Figure 4b). The position of the operculum bone, snout length, caudal and pectoral fin positions, and body height were the main morphological differences ( Figure 4b). CV1 explained 57% of the variance in morphology, which was associated with the position of the operculum bone, snout length, pectoral fin position, and body height. CV2 explained 27% of the variance and was associated with the caudal fin shape and body height (Figure 4b).

| Microsatellites
Genetic diversity was similar in each region (Supporting Information   Table S3). Within-region relatedness ranged from −0.004 in pelagic east to 0 in littoral south (Supporting Information Figure S5).
Pairwise relatedness showed no difference within or between regions ( Figure 5, Supporting Information Table S3). Genetic differ-

| D ISCUSS I ON
We show that perch display asymmetrical habitat coupling, where littoral perch clearly feed from both littoral and pelagic habitats to a higher degree than pelagic perch. Both diet and stable isotope results indicated that littoral perch include littoral and pelagic resources in their diet, while the pelagic perch were more reliant on pelagic resources only. This niche expansion of littoral predators by including pelagic prey items in their diet is consistent with general theory showing that predators normally considered as generalists, such as perch, may have a greater ability to couple habitats than more specialist predators (McCann et al., 2005;Rooney et al., 2008). However, littoral perch also showed higher individual specialization than the pelagic perch, which is contrary to the expectation that littoral perch display higher generalist feeding behavior and stronger habitat coupling. Based on our results, we propose that habitat coupling, at least in this aquatic system, can be controlled by a balance between factors promoting (e.g., competitive release, increased foraging opportunity due to increased habitat diversity) and inhibiting habitat shifts (e.g., local adaptation leading to population differentiation). Below, we elaborate on how these factors, alone or together, can regulate habitat coupling.
Predator mobility can be driven by distinct usage of prey resources in different habitats, which in this case were the littoral and pelagic zones of a freshwater lake. Higher mobility of predators has been suggested as a mechanism behind habitat coupling ( has been shown that the contribution of benthic resources is important for many pelagic feeding predatory fish species (Vander Zanden & Vadeboncoeur, 2002), our results emphasize that differences in the accessibility of habitat and resources can lead to differential habitat coupling. Potentially, this asymmetry in habitat coupling could be explained by the pelagic food web compartment providing essential resources for growth and reproduction that are not found in the littoral habitat (Scharnweber et al., 2016). Alternatively, population dynamics could influence the coupling of habitats. For example, high population densities can drive habitat shifts in perch where a part of the population moves from the preferred littoral habitat out to the pelagic habitat (Svanbäck & Persson, 2004). This asymmetry in migration might also skew habitat coupling since habitat and diet profitability will change as competition changes in the population (Svanbäck & Bolnick, 2005). While some studies have provided evidence for asymmetrical resource use (Bartels et al., 2012;Bartels, Hirsch, Svanbäck, & Eklöv, 2016;Knudsen et al., 2010;Scharnweber et al., 2016), these studies did not discuss the mechanistic drivers of asymmetrical habitat coupling.
We observed the highest individual specialization in littoral perch, which also had the broadest population niche. Our stable isotope findings corroborated with our diet results, indicating that littoral perch also had a much broader niche space than pelagic perch over a longer period of time. While total niche width is based on diet and only gives a snapshot estimate, SEAc is based on stable isotopes, which give longer time-frame and integrated measurement of resource usage (Bolnick et al., 2002;Scharnweber et al., 2016).  (Svanbäck, Quevedo, Olsson, & Eklöv, 2015). Total niche width of a population can relate to niche width within individuals (within-individual component, WIC), as well as between individuals (between-individual component; BIC) (Bolnick et al., 2010). We show that perch in our study system had a low BIC in the pelagic habitat and a high BIC in the littoral habitat. Therefore, while littoral perch had a broader overall niche width, the littoral population was composed of relatively specialized individuals, and while we found no significant relationship between IS and intraspecific competition (perch abundance), it is possible that intraspecific competition was decreased by individuals dividing the resources among themselves within the population (Bolnick et al., 2010;Faulks, Svanbäck, Ragnarsson-Stabo, Eklöv, & Östman, 2015b;Lister, 1976). Having more specialized individuals suggests that habitat coupling is not driven by individuals expanding their resource niche, but rather the population is expanding its niche. Niche expan-

sion often implies that generalization occurs once individuals have
shifted to the other habitat (i.e., the pelagic zone in our case). We anticipate a similar process here, but that littoral individuals switched between the pelagic and littoral habitats, which is supported by a substantial amount of pelagic resources found in the diet of littoral perch (i.e., pelagic Cladocera, Copepoda and pelagic macroinvertebrates). The conclusion that higher individual specialization leads to higher habitat coupling might seem counterintuitive if the littoral individual specialization would only consider littoral prey items.
However, the specialization also includes pelagic prey, which implies that some littoral predators shifted to the pelagic zone and included pelagic prey in their diet. A similar dietary pattern among individuals in the pelagic sites was not expected, since the distance from the pelagic populations to the littoral zone was relatively large and we therefore do not expect individuals captured in the pelagic zone to easily move to the littoral zone. Conversely, at some of our littoral sites, the individuals consumed almost exclusively littoral prey items.
The morphological variation (ED) and SEAc results support expectations in most cases, where morphological variation was often positively correlated with diet variation (Roughgarden, 1972;Snowberg, Hendrix, & Bolnick, 2015;Van Valen, 1965) and individual specialization . This suggests that if individuals trade-off habitat-specific foraging adaptations with differences in competition or predation between habitats, such adaptations might lead to constraints in habitat coupling Quevedo et al., 2009). However, one littoral region was seen to have a large niche space without a correspondingly high morphological variation and individual specialization (littoral south). A potential explanation for this result might be the composition and the accessibility of resources, where the range in habitat heterogeneity and presumably resource diversity in this region was lower than in the other littoral region. Lower habitat heterogeneity can also be related to higher individual specialization and lower morphological variation, as shown in an experiment that manipulated habitat accessibility (Marklund, Svanbäck, Zha, Scharnweber, & Eklöv, 2018).
This raises the question of how habitat heterogeneity correlates to individual specialization and morphological variation. One possibility is that habitat heterogeneity increases resource diversity, which may counteract intraspecific competition promoting individual niche expansion. Intraspecific density did not affect individual specialization between habitats, instead it is possible that the relationship between intraspecific competition and individual specialization is modified by habitat heterogeneity in this littoral region.
Limitations in habitat coupling might also be a result of localized genetic differentiation, if local adaptation leads to constrained individual movements across habitats. While a previous study suggested that assortative mating of perch in this study lake might be taking place over small spatial scales (Bergek & Björklund, 2007), we found no evidence of assortative mating based on region, habitat, or morphology. Indeed, even though genetic differentiation has been found between perch populations from pelagic and littoral habitats, the main component of morphological adaptation is believed to be due to phenotypic plasticity (Faulks, Svanbäck, Eklöv et al., 2015a;Olsson & Eklöv, 2005;. Phenotypic plasticity might play a big role for predators to couple different habitats, especially when considering changing environmental conditions over longer timescales (Siepielski, DiBattista, & Carlson, 2009). Lower plasticity could potentially result in slower responses to changing foraging opportunities, reducing their habitat coupling ability, and also potentially affecting ecosystem stability (McCann et al., 2005;McMeans et al., 2016;Rooney et al., 2006).
The simple and straightforward hypothesis that mobile consumers have strong stabilizing effects in food webs is a very attractive one (McCann et al., 2005). However, this leans on the assumption of rapid behavioral response to fluctuating resources. Our study demonstrates that this ability may also be influenced by intraspecific niche partitioning and the degree of individual specialization of the predators (see also Knudsen et al., 2010, Quevedo et al., 2009. We show that niche expansion, where the littoral perch included more pelagic prey types, might have led to asymmetrical habitat coupling; that is, the littoral perch population coupled the littoral and pelagic habitats more than the pelagic perch population. Our study emphasizes the role of individual variability in niche and trophic traits for spatial interactions, and highlights the need for further studies that explore the relationship between evolutionary related traits and spatially related food web interactions.

ACK N OWLED G M ENTS
This study was funded by Malméns stiftelse to MHKM, the German Academic Exchange Service (DAAD), Olsson Borghs stiftelse to KS, and the Swedish Research Council (VR) to PE and RS. We thank