Genetic architecture of dispersal behaviour in the post-harvest pest and model organism Tribolium castaneum

Abstract

Dispersal behaviour is an important aspect of the life-history of animals. However, the genetic architecture of dispersal-related traits is often obscure or unknown, even in well studied species. Tribolium castaneum is a globally significant post-harvest pest and established model organism, yet studies of its dispersal have shown ambiguous results and the genetic basis of this behaviour remains unresolved. We combine experimental evolution and agent-based modelling to investigate the number of loci underlying dispersal in T. castaneum, and whether the trait is sex-linked. Our findings demonstrate rapid evolution of dispersal behaviour under selection. We find no evidence of sex-biases in the dispersal behaviour of the offspring of crosses, supporting an autosomal genetic basis of the trait. Moreover, simulated data approximates experimental data under simulated scenarios where the dispersal trait is controlled by one or few loci, but not many loci. Levels of dispersal in experimentally inbred lines, compared with simulations, indicate that a single locus model is not well supported. Taken together, these lines of evidence support an oligogenic architecture underlying dispersal in Tribolium castaneum. These results have implications for applied pest management and for our understanding of the evolution of dispersal in the coleoptera, the world's most species-rich order.

Fig. 1. Artificial selection on dispersal.

A Experimental arena setup used to assay the dispersal behaviour of experimental Tribolium castaneum populations and provide a basis on which to artificially select individuals displaying high and low dispersal propensity. B Mean dispersals per individual (of a maximum of three) for each T. castaneum selection line across four generations of selection on high and low dispersal (generation 1 represents behaviour in the stock population). Orange and blue represent high and low selection regimes respectively. Solid lines connect repeated measurements from the same selection line, while dashed lines show predictions generated by a GLM fitting the interaction of selection regime and generation modelled as a second order polynomial.

Fig. 2. Number of dispersers from assayed mixed-sex populations of 200 T. castaneum offspring from crosses within and between dispersal selection regimes.

On the X-axis, ‘H’ and ‘L’ represent the selection regime and ‘m’ and ‘f’ the sex of parents, e.g. Hm-Lf indicates that the male parent came from a high selection line and the female parent from a low selection line. Data for males and females are shown in purple and green respectively. Small filled points show experimental measures from individual crosses, while large hollow points show predictions of a GLMM fitting the effect of the interaction of sex and cross type on number of dispersers, while controlling for line ID.

Fig. 3. Combined data from experimental (column 1, row A) and simulated (all other panels) evolution experiments selecting on dispersal behaviour.

Column 1: (A) Experimental results showing the number of dispersers from each of 64 lines of T. castaneum following 11 generations of extreme inbreeding and (B–E) data from agent-based simulations designed to model the same inbreeding design and dispersal assay as used experimentally. Numbers of dispersers are shown prior to (blue) and following (red) 11 generations of inbreeding. Sample sizes for simulations were 250 independent populations for each genetic architecture, each modelling the trait as controlled by either 1, 3, 5 or 10 additive biallelic loci (rows B-E respectively). For each simulated architecture, starting dispersal allele frequencies were 0.8, 0.4, 0.4 and 0.4 in these architectures respectively, this and other parameters were selected as those maximising agreement with experimental results during a parameter scan (Fig. 4). For simulated data, additional columns display: 2) Allele outcomes at each locus at generation 11, i.e. whether either allele had become fixed, or both alleles remained in the population. 3) The distribution of allele frequencies averaged across loci for each individual simulated line.

Fig. 4. Outputs from parameter scans using an agent-based simulation designed to model an emigration selection experiment using T. castaneum under different trait architectures.

Colours represent R² values assessing the fit of a model defined on experimental data, and applied to simulated data across scenarios comprising combinations of dominance (d), trait heritability (h) and starting dispersal allele frequency (A), across a sex-linked single-locus, and unsex-linked single-, 3-, 5- and 10 locus architectures.