Trophic variation within a piscivorous lake trout morph from Great Bear Lake, Canada: The initial step toward ecological specialization?

Ecological opportunities present during colonization of novel environments can drive divergent selection on traits, resulting in specialization of morphs to enhance efficient use of resources. Thus, in an ecologically polymorphic species, differences in resource specialization should be found among morphs, and homogeneity in resource use expected within a morph. Using one of four morphs in Great Bear Lake, we investigate whether specialization of trophic resources among individuals occurs within this single morph, which could indicate a potential for continued divergence. Four distinct dietary patterns of resource use within the lake trout morph were detected from fatty acid composition. Feeding habits of different groups within the morph were not associated with detectable morphological or genetic differentiation, suggesting that behavioral plasticity may have caused the trophic variation within this morph. A low level of genetic differentiation was detected between exceptionally large-sized individuals and other individuals. Investigating a geologically young system that displays high levels of intraspecific diversity and focusing on dietary patterns of resource use variation of individuals suggested that individual specialization can occur within a morph.

investigate whether individual specialization may be contributing to trophic breadth and 136 variation observed among individuals in this morph. Specifically, our aims were to 1) compare 137 resource use among piscivorous lake trout individuals (Morph 2) by characterizing their fatty 138 acids profiles, 2) determine whether resource-use differences were influenced by life-history 139 traits (e.g., size and age), 3) characterize the extent of morphological variation individuals 140 present among groups expressing different feeding strategies, and 4) determine if genetic 141 differences existed among groups. In addition, we examined a sub-set of large lake trout from 142 our collections (> 900 mm in fork length) referred to locally as "Giants" (Fig. 1), to determine if 143 they showed any ecological and genetic differences. These exceptionally large individuals 144 comprise < 1% of the lake trout population in Great Bear Lake, and are among the largest lake 145 trout in the world (Chavarie et al. 2016 ). Except for their large body-size, these individuals 146 exhibit no major morphological or spatial and temporal distribution differences relative to other 147 co-occurring piscivorous lake trout. By focusing on trophic variation within a specific morph, we 148 aimed to advance our understanding of ecological and evolutionary processes operating within a 149 geologically young ecosystem that provides resource potential sufficient for promoting 150 intraspecific divergence (Bhat et  Bear Lake lacks a commercial fishery but plays an important role in the local economy, 162 supporting a fly-in sport fishery for tourists and a subsistence fishery for the small Sahtu 163 community of Déline. Great Bear Lake has considerable intraspecific diversity within lake trout, 164 lake whitefish (Coregonus clupeaformis), and cisco (C. artedi) ( All fatty acids values were converted to a mass percentage of the total array, and were 208 named according the IUPAC nomenclature as X:Y n-z, where X is the number of carbon atoms 209 in the fatty acids, Y is the number of methylene-interrupted double bonds in the chain, and n-z 210 denotes the position of the last double bond relative to the methyl terminus (Ronconi et al. 2010).

Life-history 229
To determine if fatty acid groups differed in size-at-age, length vs. age was modeled 230 using the Von Bertalanffy length-age model fit to length at age-of-capture of individual fish 231 (Quinn and Deriso 1999): 232 The length-age model describes length L t at age-of-capture t as a function of theoretical 234 maximum length (L ∞ = mm), instantaneous rate at which L t approaches L ∞ (K = 1/year), 235 theoretical age-at-zero length (t 0 = years), and multiplicative error (ε). Model parameters, L ∞ , K, 236 and t 0 , and associated standard errors were estimated using nonlinear regression. Residual sums-237 of-squares were compared between a full model (separate models for each group) to a reduced 238 model (a single model for all groups) in a likelihood-ratio test (Hosmer Jr et al. 2000). If the 239 likelihood-ratio test was significant (P < 0.05), we concluded that growth differed among groups 240 identified by fatty acids (79 lake trout). If the likelihood-ratio test was not significant (P > 0.05), 241 we concluded that growth did not differ among groups. The same test was repeated for each pair 242 of groups, with and without the Giant form (fork length ≥ 900 mm) included in each group, to 243 isolate the influence of this sub-set in our size-at-age comparison due to the prevalence of Giants 244 in Group 3. 245

Genetic analyses 246
To determine if genetic differences existed among individuals expressing different 247 feeding strategies, 79 lake trout classified by fatty acid composition into four groups were 248 genotyped to determine genetic variation and structure within and among groups. To allow a 249 sample size sufficient for making a genetic comparison of the Giant to the other dietary groups, 250 22 additional individuals determined non-randomly by their size (≥ 900 mm ; Giant sub-set) 251 from the 2002-2015 collections were added to the Giant processed for fatty acids, for a total of 252 39 Giants for genetic analysis. Lake trout DNA was extracted from pectoral fin tissue preserved 253 in ethanol using DNEasy extraction kits (Qiagen Inc., Valencia, CA) following manufacturer 254 protocols. Piscivorous groups were assayed using a suite of 23 putatively neutral microsatellite 255 markers amplified in four multiplexes previously described in Harris et al. (2015). Amplified 256 microsatellite fragments were analyzed using an automated sequencer (ABI 3130xl Genetic 257 Analyzer; Applied Biosystems, Foster City, CA). The LIZ 600 size standard was incorporated 258 for allele base-size determination. All genotypes were scored using GeneMapper software ver. Genetic structuring was tested among lake trout groups using several different methods. 279 First, genotypic differentiation among lake trout groups was calculated using log-likelihood (G) 280 based exact tests (Goudet et al. 1996)  determine the optimal number of principal components to retain in the analysis. 306 Morphology 307 Morphological variation was quantified for the 79 lake trout to compare fatty acid 308 groupings (different feeding strategies) to morphological variation within the piscivorous morph.  Table 1). The first two axes of the fatty acids PCA 346 explained 65.2 % of the variation in diet and the four groups were supported by PERMANOVA 347 (F 3,76 = 39.4, P < 0.01) and pairwise comparisons between all pairs (all P < 0.01). Finally, depth 348 of capture did not differ among groups identified by fatty acids profiles (p ≥ 0.05). For all 349 groups, the majority of lake trout were caught between 0-20 m (Fig. A3). 350

Life-history 351
Overall, life history parameters did not differ among lake trout groups identified by fatty 352 acid composition. Length-age models did not differ among fatty acid groups, based on overall 353 likelihood-ratio tests ( Fig. 3; F 9 Table 2). Allelic richness ranged from 9.57 (Group 2 372 and 4) to 9.87 (Group 1), while expected private allelic richness ranged from 0.87 (Group 3) to 373 1.08 (Group 2; Table 2 Table 3) whereas comparisons that included Giants always differed the most from the other fatty 389 acid groups, and were also the only significant pairwise comparisons (P < 0.05, Table 3). The F ST 390 value for the Giant vs. Group 1 and 4 were similar to genetic differentiation previously observed 391 among four lake trout morphs in Great Bear Lake (Table 3) (Table A2). The admixture plot based on 394 K=2 showed no clear genetic structure among groups defined by fatty acid analysis, however, 395 some differentiation of the Giant sub-set from the fatty acid groups was observed (Fig. 4). 396 Finally, the Bayesian information criterion in the DAPC analysis (BIC = 185.42, Table  397 A3, Fig. A5 A) suggested that two clusters best explained genetic structure in our study (30 PCs 398 retained as suggested by the cross-validation procedure; Fig. A5

B). A compoplot (barplot 399
showing the probabilities of assignment of individuals to the different clusters) for K=2 revealed 400 no clear genetic structure between two groups identified by the DAPC analysis with the 401 exception of the Giant group which appeared to have more individuals assigned to cluster two 402 (Fig 4). Density plots of the discriminant function, however, do show that the two clusters 403 identified through the DAPC analysis are mostly non-overlapping (Fig. A5 C). 404

Morphology 405
Morphological variation was low among four dietary groups within the piscivorous 406 morph. The first canonical axis for body shape CVA was significant (P>0.05), but head shape 407 CVA revealed no significant canonical axes (P>0.05) in groupings (Fig. 5 a, b, c). MANOVAs 408 for body and head shape were not significant (P>0.05). Linear measurements CVA revealed one 409 significant canonical axis (P>0.05). MANOVA permutation tests confirmed differences in linear 410 measurements among groups for linear measurements (P = 0.047). Most distinctions were related 411 to linear measurements of heads, whereas upper and lower jaws, head depth, and snout-eye 412 lengths differed between Group 3 and Group 4 (P ≤ 0.05), and head length differed between 413 Group 1 and 4 (P = 0.03; Fig. 6). Caudal peduncle length and anal fin length differed marginally 414 between Group 2 vs 3 (P = 0.068) and Group 1 vs 3 (P = 0.075), respectively. The first two PCA 415 axes explained 44.3% and 12.3 % of variation for body shape, 35.1% and 30.7 % of variation for 416 head shape, and 39.6 % and 20.9 % for linear measurements (Fig. 5 d, e, f). Eberhard 2003). Theoretical models suggest that exploiting a wide range of resources is either 529 costly or limited by constraints, but plasticity is favored when 1) spatial and temporal variation 530 of resources are important (i.e., highly present in Great Bear Lake; Fig. A4), 2) dispersal is high, 531 3) environmental cues are reliable, 4) genetic variation for plasticity is high and 5) cost/limits of 532 plasticity are low (Ackermann et al. 2004; Hendry 2016). 533 The expression of plasticity in response to particular ecological conditions (e.g., habitat 534 structure, prey diversity) can be evolutionarily beneficial (i.e., result in increases in fitness). 535 While most studies of diet variation focus on morphological differences among morphs in a 536 population, diet variation can also arise from behavioral, biochemical, cognitive, and social-rank 537 differences that cause functional ecology to be expressed at a finer scale rather than at the morph

Competing interests 637
The authors declare that they have no competing interests. 638

Availability of data and materials 639
The datasets supporting the conclusions of this article are included within the article. Raw data 640 are available from the corresponding author upon reasonable request. 641 Group 3, and Group 4 of piscivorous Lake Trout morph identified from Great Bear Lake.  clusters in the DAPC performed on piscivorous Lake Trout from Great Bear Lake. The function 1311

Consent to publish
find.clusters was run with a maximum number of clusters of 10 to identify the optimal number of 1312 clusters based on the BIC values. A K value of 2 (the lowest BIC value) represents the best 1313 summary of the data (most probable number of (K)). (C) The results of the discriminant function 1314 that shows that the two clusters are mostly non-overlapping. 1315 55 of n