Socially plastic responses in females are robust to evolutionary manipulations of adult sex ratio and adult nutrition

Socially plastic behaviours are widespread among animals and can have a significant impact on fitness. Here we investigated whether the socially plastic responses of female Drosophila melanogaster can evolve in predictable ways following long term manipulation of adult sex ratio and adult nutrient availability. Previous reports show that female D. melanogaster respond plastically to their same-sex social environment, and lay significantly fewer eggs after mating when previously exposed to other females. In this study, we tested two hypotheses, using females drawn from lines with an evolutionary history of exposure to variation in adult sex ratio (male biased, female biased, or equal sex ratio) and adult nutritional environment (high or low quality). The first was that a history of elevated competition in female-biased regimes would select for increased plastic fecundity responses in comparison to females from other lines. The second was that these responses would also be magnified under poor nutritional resource regimes. Neither hypothesis was supported. Instead, we found that plastic fecundity responses were retained in females from all lines, and did not differ significantly across any of them. The lack of differences does not appear to be due to insufficient selection, as we did observe significant evolutionary responses in virgin egg laying patterns according to sex ratio and nutritional regime. The lack of variation in the magnitude of predicted plasticity is consistent with the idea that the costs of maintaining plasticity are low, benefits high, and that plasticity itself can be relatively hard-wired.

Methods
Dataset collected as described in the methods. Processing according to the statistical pipelines described. All statistical analysis was performed using R Core team V-4.0.2 (2020 2020). All three replicates were analysed simultaneously, with the replicates (‘Population’) designated as a random factor. The Shapiro-Wilk test, Q-Q plots, and histograms were used to check data were normally distributed and the Levene’s test to check the homogeneity of variances across treatments. Analysis of egg number and progeny were analysed using linear mixed effects models from the lme4 package (Bates et al. 2015) and Chi-squared test were used to drop non-significant terms (supplementary material). Mating latency and duration were also analysed using the same method after being log10 transformed. To analyse differences between group treatments, a Tukey post hoc analysis was conducted using the ‘emmeans’ package (Lenth 2022). Additionally, a Generalized Linear Mixed Effects Model (GLMER) with a Poisson error distribution, and a negative binomial GLMER were used to check model fit versus the LMER. The GLMER with poisson structure did not fit the data well, whilst the negative binomial reported similar results to the LMER. Models were compared using Log-likelihood, Akaike’s information criterion (AIC) and residual plots. The data were initially analysed using the whole dataset. In subsequent analyses, zero egg counts were removed to allow the data for the egg laying and non egg laying females to be analysed separately, using a binomial generalised linear mixed model. For virgin egg data, it was not distinguish between the eggs laid by focal and non-focal females in the grouped treatments. Therefore, we analysed the virgin egg count data separately for alone and grouped treatments.