Computational protocol: Living in Heterogeneous Woodlands – Are Habitat Continuity or Quality Drivers of Genetic Variability in a Flightless Ground Beetle?

Similar protocols

Protocol publication

[…] We collected beetles by using ten live pitfall traps per plot baited with red wine on cellulose during the spring and summer of 2011 (Schwäbische Alb) and of 2012 (Schorfheide-Chorin). In all plots, the traps were placed in a straight line, 10 m apart along the plot border to ensure equal sampling area. We gathered all of the Abax parallelepipedus individuals we found in the traps approximately once a week and rebaited the traps until we had caught 33 individuals in the plot. We pooled the beetles trapped in all the traps of a plot each collection round and froze them at-80°C. Field work permits were issued by the responsible state environmental offices of Baden-Württemberg and of Brandenburg (according to §72 BbgNatSchG).We extracted DNA using the CTAB extraction protocol [] from three legs from each of 24 randomly selected beetles for each plot. We genotyped 14 polymorphic microsatellite loci using an ABI 3730 Genetic Analyzer (Applied Biosystems, Foster City, CA, USA). For PCR, sequencing protocols, and information about the loci see [].Deviation from Hardy-Weinberg Equilibrium (HWE) was tested using GENEPOP 4.2 [] and no significant deviation was found (percentage of local populations not in HWE: Schwäbische Alb = 5.288%, Schorfheide-Chorin = 4.240%). Suspected presence of null alleles was tested using MICRO-CHECKER 2.2.3 [] and no null alleles were found. Linkage disequilibrium (LD) was checked using FSTAT [,]. No significant LD was found. Allelic richness, the rarefied numbers of alleles as a measure of genetic diversity, was calculated for each plot using FSTAT The rarefaction was done per local population, with a minimum sample size of 20 individuals, to account for isolated instances of ineffective PCR reactions.Overall FST values among local populations, a measure of genetic differentiation, were calculated for each region using Arlequin []. Private alleles, meaning those found in only one local population, and unique alleles, meaning those found in a specific group of plots either by region, by ancient or recent woodlands, or by population density, were counted and tallied. The grouping by local population density was done by grouping plots into percentiles based on the number of individuals caught in the 2008 killing traps (see ). Genetic clustering was tested using the algorithm developed by Pritchard [] as implemented in STRUCTURE 2.3.4 for each region separately, to ensure that no underlying clustering is affecting the results. In this analysis we used the admixture model with no use of previous information about sampling location. Burnin length was 20,000 and there were 100,000 MCMC repeats after burnin. Number of clusters was run from K = 1 to K = the number of plots+1 for the Schorfheide-Chorin (K = 43), and for the Schwäbische Alb (K = 47). For the Schwäbische Alb for higher values of K, the runtime was insufficient to find proper solutions, and therefore a second run of K = 1 to K = 30 was analyzed (). We used CLUMPAK [] and HARVESTER [] to find the most likely K using the Evanno method [] and to visualize the results. [...] We modeled the relationship between plot characteristics, including measures of habitat continuity as well as environmental parameters, and allelic richness. We started with 27 predictor variables (). We first tested for collinearity between the predictor variables and removed the smallest number of variables possible while eliminating all instances where Spearman's rho > |0.7| [], leaving us with 18 predictors (). When we could not choose which variable to eliminate based on maximizing the number of remaining variables, we chose to retain the one more correlated with allelic richness. We created a general linear model for each region using allelic richness as a response variable. We modeled the regions separately as the means and variances of allelic richness are different due to the environmental conditions and history of the regions. The models were reduced using a backwards step reduction process based on AICc scores. The models with the lowest AICc scores, and the smallest numbers of predictors in the case of ΔAICc<2 between two models, were selected (see []). The residuals were checked to ensure that they are normally distributed and the residuals were plotted against the fitted values to investigate homogeneity of variance. As stand age and the FORMI index are significantly correlated (Spearman Rank Correlation: rho = -0.700, p<0.001, ) and we were interested in testing both of these parameters, we ran these models twice, once using stand age and once using FORMI as a possible explanatory variable.We tested for spatial autocorrelation using Moran's I both of the allelic richness values themselves for each region using the APE package [] and of the residuals of the model using the ncf package [] and corrected using Bonferroni's correction for multiple testing. We examined the effects of long-term habitat continuity on genetic differentiation using two methods. We first ran an AMOVA in Arlequin [] for each region separately, grouping the plots by whether they are located in an ancient or in a recent woodland. We then tested the effects of stand age, of location in ancient or in recent woodlands, and of the interaction between them on genetic differentiation using GESTE 2.0 []. This program uses hierarchical Bayesian methods to find population-specific FST values, which are then modeled with the historical variables we provided in a generalized linear model. GESTE was run using default parameters. If not otherwise stated, all statistical analyses were done using R 3.0.0 []. […]

Pipeline specifications

Software tools Genepop, Arlequin, Clumpak, APE
Applications Phylogenetics, Population genetic analysis
Organisms Abax parallelepipedus