Computational protocol: Morphometric and Genetic Differentiation of Two Sibling Gossamer-Wing Damselflies, Euphaea formosa and E. yayeyamana, and Adaptive Trait Divergence in Subtropical East Asian Islands

Similar protocols

Protocol publication

[…] DNA sequences were aligned using the Clustal W method in MegAlign (DNASTAR, The McDonald Kreitman Test (MKT) implemented in DnaSP v. 4.0 () was used to detect the signature of natural selection in cox2 by comparing proportions of synonymous and nonsynonymous substitution within vs. between populations. The aligned cox2 sequences were translated into amino acid sequences in DnaSP using a genetic code of the Drosophila. The significance of deviations on the ratio of replacement to synonymous substitutions was determined using two—tailed Fisher's exact tests. For maximum parsimony (MP) analyses, the most parsimonious trees were searched using parsimony ratchet procedure () implemented in Pauprat () and PAUP* v. 4.0b10 (). The ratchet procedure was run 20 times using 200 replicates in each run and repeated with 15% of weighted characters using batch files implemented in Pauprat. Branch supports were calculated using non—parametric parsimony bootstrapping with 1000 iterations, each with 100 stepwise random sequence additions and tree—bisection and reconnection (TBR) branch swapping. For maximum likelihood (ML) and Bayesian inference (BI) analyses, the best—fitted nucleotide substitution model was selected in Modeltest v. 3.7 () using Bayesian Information Criterion (BIC). ML tree searches of 1000 iterations and parameter optimization were performed using a rapid approximation algorithm implemented in RAxML v. 7.03 () with starting parameter values derived from the best—fitted substitution model. ML bootstrap analyses of 1000 replicates were conducted with a rapid bootstrapping procedure (-f a) and GTRMIXI model in RAxML to accommodate the proportion of invariable sites (I) and rate heterogeneity using a gamma distribution (Γ). MrBayes v. 3.12 () was used to search BI trees and calculated Bayesian posterior probabilities (BPP) of the trees. Prior values of the model parameters in BI analyses were estimated in Modeltest. Two independent Bayesian analyses with random starting trees were run simultaneously with each run containing four Markov Chains. The Markov Chain Monte Carlo (MCMC) processes were run for 1 × 107 generations with a tree sampling frequency of every 1000 generations. MCMC searches were monitored for the convergence of separate runs after the average split frequencies of two runs fell below the value of 0.01, and the convergence diagnostic potential scale reduction factor reached one (). The first 25% of MCMC samples were discarded as burn—in. BPP of the BI trees was calculated using a 50% majority rule tree from the remaining 7500 trees in PAUP*. For statistical network analyses, TCS v. 1.21 () was used to construct a parsimony network with 95% probability of haplotype connection. [...] Recent studies analyzing wing shapes and DNA sequences successfully discriminated morphologically cryptic insect species and populations (; ; ; ; ). The geometric morphometric method based on landmarks can separate information concerning shape from size and scaling of morphological structures, therefore allowing these structural characters, which are often correlated, to be tested independently (; ). The right wing of each damselfly was carefully removed from the preserved specimen and mounted on a glass slide with the dorsal side of the wing facing upwards. A ruler with minimum scales of 1 mm was placed on the glass slide to calibrate of the measurement. A Nikon D80 digital camera with 105 mm Micro Nikkor lens f 2.8 ( mounted on a copy stand was used to photograph the wings at 7–8× magnification, with two white lights projected from 45-degree angles above the slide and one light directly below the slide. Before taking each image, the slide surfaces were manually adjusted with the aid of a gradienter so that they were perpendicular to the camera. Images were saved in JPEG format (300 dpi) and imported into tpsDig () for digitization of landmarks. A series of twelve landmarks for forewings and hind wings were chosen to quantify wing shape variation (). Two additional landmarks 1 mm apart on the reference ruler were digitized to calculate the centroid sizes of wings but not used for shape analyses. Centroid size was used as an estimate of wing size, which represented a surrogate for body size of the damselfly.The x and y coordinates of the landmarks were digitized on each wing image and converted into TPS format using tpsDig. The TPS files were imported into CoordGen6h of the IMP () for subsequent statistical analyses of wing shape. The Procrustes superimposition method was used to remove non—shape variations including scale, position, and orientation differences among specimens, and to extract shape variables among homologous landmarks using a Generalized Least Square (GLS) criterion (; ). For morphometric analyses, samples from all populations within each species were pooled because the main purpose of this study was to distinguish between species, and the sample size was insufficient to allow sensitive statistical tests (3–10 samples per site, ). The geometric shape variables obtained from the GLS were used to conduct the Principal Component Analysis (PCA) implemented in PCAGen6p of the IMP for characterizing wing shape differences between species. Anderson's test was used to determine the numbers of statistically significant PCs that discriminate between the two species (). The consensus wing shape (mean wing shape) of all specimens was compared with a consensus for each of four categories (the forewing and hind wing of two species) to characterize changes of wing shapes. Thin—plate spline deformation grids were generated between each of the four categories and the consensus in PCAGen6p to visualize the level of deformation in wing shapes. Multivariate analyses of covariance (MANCOVA) were performed in SPSS v. 12.0 () to statistically evaluate the wing shape differences between species and between forewings and hind wings. The shape variables (uniform components and partial warps) of fore or hind wings were used as dependent fixed variables and the centroid size and population as a covariate. A multivariate regression of wing shape variables against centroid sizes using pooled samples of both species was conducted in TpsRegr v. 1.38 () to test for a linear pattern of wing shape and body size between both species. […]

Pipeline specifications

Software tools TpsDIG, TpsRegr
Application Macroscope & basic digital camera imaging
Organisms Euphaea formosa, Euphaea yayeyamana