Mate choice in fruit flies is rational and adaptive

Abstract

According to rational choice theory, beneficial preferences should lead individuals to sort available options into linear, transitive hierarchies, although the extent to which non-human animals behave rationally is unclear. Here we demonstrate that mate choice in the fruit fly Drosophila melanogaster results in the linear sorting of a set of diverse isogenic female lines, unambiguously demonstrating the hallmark of rational behaviour, transitivity. These rational choices are associated with direct benefits, enabling males to maximize offspring production. Furthermore, we demonstrate that female behaviours and cues act redundantly in mate detection and assessment, as rational mate choice largely persists when visual or chemical sensory modalities are impaired, but not when both are impaired. Transitivity in mate choice demonstrates that the quality of potential mates varies significantly among genotypes, and that males and females behave in such a way as to facilitate adaptive mate choice.

Figure 1. Male D. melanogaster display transitive mate choice.

(a) Mate choice network for Canton-S males, combining two independent replicates. More attractive lines (a line that received more than half of the matings in any pairwise competition) point to less attractive lines. The width of a line is proportional to the mating skew for each pairwise interaction. Absent arrows (for example, R517 versus R313) denote ties between two lines. Line number (for example, R362) corresponds to DGRP ID. (b) Examples of transitive and non-transitive triads. The overall transitivity of a network, t_tri, is calculated as the proportion of transitive triads in the full network. (c) Transitivity scores from 100,000 permutations of mate choice data, used to calculate the significance of our mate choice network. On average, we expect that 75% of relationships would be transitive at random, as 6/8 possible relationships within a triad are transitive. Dashed line represents the observed transitivity score for Canton-S males, which is significantly greater than transitivity scores expected to be generated at random. The data used to generate panels (a) and (c) are the combined results of the separate mate choice assays using Canton-S males, while the statistics reported in the text were calculated for each assay separately.

Figure 2. Males with impaired signal detection make similar mate choices.

(a) Correlation of the proportion of the first female line mated for all 45 pairwise combinations of 10 DGRP female lines between Canton-S males and Oregon-R (wild type), Canton-S males in the dark (blind), ninaB^−/− mutant males (blind) and ppk23^−/− mutant males (deficient CHC detection). Regression lines represent geometric mean regression. The significance of Pearson's correlation coefficients were calculated via 10,000 permutations of the data (see Methods for details). (b) Mating rates for all males tested (±s.e.m.). Bar colour represents genetic background (purple=Oregon-R, blue=Canton-S, orange=ninaB^−/−, red=ppk23^−/−), whereas black hashmarks signify that mating trials were conducted in the dark.

Figure 3. Evidence for male choice among female genotypes.

(a) Relationship between time to courtship in single-male single-female pairings and the mate choice of males in two-choice trials. Significance of Pearson's correlation coefficients were calculated via 10,000 randomizations of the data (see methods for details). (b) The proportion of time males spend with the more attractive female (as determined by previous trials with active females) when males interact with two decapitated females. The specific female line choices are given along the x axis, with the attractive (+) and unattractive (−) female denoted for each comparison. P-values are from Wilcoxon signed-rank tests against the null hypothesis of 50% time with the more attractive female. Box boundaries represent s.e.m. and whiskers show 80th and 20th percentiles of data.

Figure 4. Male mating preferences coincide with female offspring production.

(a) Relationships between female offspring production and mass, relative abundance of CHC 9, relative abundance of CHC 10 and female receptivity. Black lines and text show correlations between female traits with data from an outlier isogenic line removed, whereas grey lines and text show correlations with all data. Reported correlation coefficients and P-values are from Pearson's correlation. (b) Correlations between the proportions of the first female line mated and pairwise differences in female offspring production when an outlier line was removed. Significance was tested via permutation tests to avoid pseudoreplication. Regression lines for all plots represent geometric mean regression.