Computational protocol: Phenotypic and genetic divergence within a single whitefish form – detecting the potential for future divergence

Similar protocols

Protocol publication

[…] All analyses were carried out on freshly dead fish, within four hours after catch. To specifically address the feeding morphology of the gangfisch, we counted the number of gill rakers on the first right gill arch. For geometric morphometrics, each individual was photographed with its fins spread and fixed to the surface of a polystyrene bed. We chose 16 landmarks on the left side of each specimen following general guidelines for placement of landmarks (Zelditch et al. ) (Fig. S1). Because large filled gonads can influence the body shape in the ventral body regions (Helland et al. ), we omitted the commonplace landmark at the pelvic fin from all analyses. To quantify morphological variation in body shape among individuals, we performed multivariate geometric shape analysis. After digitizing the landmarks using TpsDig, we analyzed each landmark's relative position and hence overall variation in body shape using TpsRW (Thin-Plate Spline Relative Warp), (Rohlf ) (all Tps-software and information available for download at http://life.bio.sunysb.edu/morph/index.html). TpsRW allowed calculation of the partial warp and uniform scores that denote the differences in body shape among the individuals. Both partial warps and uniform scores were scaled to centroid size as part of a generalized Procrustes analysis (GPA; please refer to Rohlf and Slice () for details of the method). We then analyzed the partial warps and uniform scores using a multivariate discriminant function analysis (DFA) based on the classification of individuals into the different spawning depths. A subsequent canonical variance analysis (CVA) combined all partial warp and uniform scores for each individual into two CVA scores that maximally discriminate between the three depths. The CVA scores were used solely for visualization of the differences in morphology because they represent single values for an individual that are easy to use in software designed to visualize shape differences. For visualization of the body shape differences among depths, we connected the landmarks of two extreme individuals that lie on opposite ends of the morphology spectrum ranging from what the CVA discriminated as the most littoral and the most pelagic individual. Body shape depictions were created using the software TPSregr that regresses the variation in body shape with independent variables such as CVA scores or measures of genetic differentiation such as FCA scores. To test for specific differences in morphology between the different depths, we performed a mancova with all partial warps and uniform scores as response variables and depth as a categorical predictor variable. To account for any level of allometric size variation caused by differences in size at age in gangfisch, we included fish size (log centroid size) as a covariate in our model. To control for a possible artifact of body arching (cf. Valentin et al. ), we used the function ‘unbent’ as implemented in the software tpsUtil. We therefore first connected the landmarks 1, 6, 9, and 10 with a straight line using the software's graphical interface (Fig. S1). The software then fits a quadratic curve through this line of landmarks and ‘unbends’ the configuration so that the estimated quadratic fit becomes a perfectly horizontal line. The tps-file created by applying this unbent configuration contains the shape information without any bending artifacts were used for all further analyses. All statistics were performed using Statsitica vers. 11 (Statsoft® (Europe), Hamburg, Germany) except for the tps-software (see above), niche metrics, and genetic analyses (see below). [...] All loci were checked for the presence of null alleles with the software MicroChecker (Van Osterhout et al. 2004). Observed (HO) and expected (HE) heterozygosity of samples from the three different depths were calculated using arlequin version 3.5 (Excoffier and Lischer ). Deviations from Hardy–Weinberg equilibrium (HWE) were tested using exact tests (Guo and Thompson ), for each locus and sample using genepop vers. 4.0. (Raymond and Rousset ). We ran 20 batches and used 10 000 dememorizations and 5000 iterations per batch. For multiple comparisons, significance levels were adjusted by sequential Bonferroni corrections (Holm ). Deviations from linkage disequilibrium (LD) between all pairs of loci for each sample were tested using arlequin. The global genetic differentiation (global FST) of all gangfisch and the one locus pairwise estimates (FST′s) between fish caught at different depths were calculated using genepop. We calculated the pairwise genetic differentiation (FST) between fish caught at different depths with 10 000 permutations using arlequin. For comparison of genetic with morphological data, the difference between individuals based on allele frequencies was calculated with factorial correspondence analysis (FCA) using the software genetix version 4.03 (Belkhir et al. ). […]

Pipeline specifications

Software tools Arlequin, Genepop
Application Population genetic analysis
Organisms Homo sapiens, Hemisus marmoratus