Phenotypic plasticity of body size in a temperate population of Drosophila melanogaster: when the temperature-size rule does not apply.

A natural population of Drosophila melanogaster in southern France was sampled in three different years and 10 isofemale lines were investigated from each sample. Two size-related traits, wing and thorax length, were measured and the wing/thorax ratio was also calculated. Phenotypic plasticity was analysed after development at seven different constant temperatures, ranging from 12 degrees C to 31 degrees C. The three year samples exhibited similar reaction norms, suggesting a stable genetic architecture in the natural population. The whole sample (30 lines) was used to determine precisely the shape of each reaction norm, using a derivative analysis. The practical conclusion was that polynomial adjustments could be used in all cases, but with different degrees: linear for the wing/thorax ratio, quadratic for thorax length, and cubic for wing length. Both wing and thorax length exhibited concave reaction norms, with a maximum within the viable thermal range. The temperatures of the maxima were, however, quite different, around 15 degrees C for the wing and 19.5 degrees C for the thorax. Assuming that thorax length is a better estimate of body size, it is not possible to state that increasing the temperature results in monotonically decreasing size (the temperature-size rule), although this is often seen to be the case for genetic variations in latitudinal clines. The variability of the traits was investigated at two levels-within and between lines-and expressed as a coefficient of variation. The within-line (environmental) variability revealed a regular, quadratic convex reaction norm for the three traits, with a minimum around 21 degrees C. This temperature of minimum variability may be considered as a physiological optimum, while extreme temperatures are stressful. The between-line (genetic) variability could also be adjusted to quadratic polynomials, but the curvature parameters were not significant. Our results show that the mean values of the traits and their variance are both plastic, but react in different ways along a temperature gradient. Extreme low or high temperatures decrease the size but increase the variability. These effects may be considered as a functional response to environmental stress.