Selection on heritable social network positions is context-dependent in Drosophila melanogaster

Abstract

Social group structure is highly variable and can be important for nearly every aspect of behavior and its fitness consequences. Group structure can be modeled using social network analysis, but we know little about the evolutionary factors shaping and maintaining variation in how individuals are embedded within their networks (i.e., network position). While network position is a pervasive target of selection, it remains unclear whether network position is heritable and can respond to selection. Furthermore, it is unclear how environmental factors interact with genotypic effects on network positions, or how environmental factors shape selection on heritable network structure. Here we show multiple measures of social network position are heritable, using replicate genotypes and replicate social groups of Drosophila melanogaster flies. Our results indicate genotypic differences in network position are largely robust to changes in the environment flies experience, though some measures of network position do vary across environments. We also show selection on multiple network position metrics depends on the environmental context they are expressed in, laying the groundwork for better understanding how spatio-temporal variation in selection contributes to the evolution of variable social group structure.

Fig. 1. Genotypic effects on network position.

Twenty heterozygous Drosophila melanogaster genotypes significantly differed in five commonly studied measures of social network position (all LRT, P_R < 0.05). Broad-sense heritability estimates (H²) for each network position are presented in each panel. Boxplots present the medians (horizontal lines), interquartile ranges (boxes), and values within ± 1.5x IQR (whiskers) for each genotype (n for each genotype = 56). Genotypes are ordered in each panel by their median value for each network position. Genotype labels are presented as the maternal parent genotype number, crossed with (“x”) the paternal genotype number. Genotype numbers are arbitrary and unindicative of any similarities/differences between genotypes. The sex of each genotype is indicated by the color-fill of the boxplots (yellow = females, navy = males). Source data are provided as a Source Data file.

Fig. 2. Sex-by-nutritional environment effects on network position.

Sex differences in outstrength (a, b) and clustering coefficient (c, d) significantly differed across nutritional environments of varying protein-to-carbohydrate ratio (a, c) and caloric concentration (b, d). Boxplots present the medians (horizontal lines), interquartile ranges (boxes), and values within ± 1.5x IQR (whiskers) for each sex in each nutritional environment (1:1–1:4, 1x–4x; n = 160, 120, 130, 50, 100, 130, respectively for each sex). Sexes are delineated by the color fill of the boxplots (yellow = females, navy = males). Source data are provided as a Source Data file.

Fig. 3. Selection on network position is context-dependent.

Selection gradients, quantified as the correlations between male mating success and three network position metrics measured the first day after social groups were established, varied across nutritional environments that varied in protein:carbohydrate ratio (a–c) and caloric concentration (d–f). Lines show best fit linear relationships with standard error (shaded areas) (1:4–1:1, 4x–1x; n = 40, 60, 60, 40, 70, 50, respectively). The colors of each line correspond to the given protein:carbohydrate ratio or caloric concentration. Source data are provided as a Source Data file.