Materials and Methods
General methods
Experiments were conducted in a 25°C humidified room held under a 12 h light: 12 h dark cycle. Flies were maintained in 75 x 25 mm glass vials containing 7 ml sugar-yeast-agar (SYA) medium (100g brewer’s yeast, 50g sucrose, 15g agar, 30mL Nipagin (10% solution), 3mL propionic acid, and 0.97L water per litre of medium). Wildtype flies were sampled from the Dahomey population (Bretman et al. 2009). Females were allowed to oviposit on agar-grape juice plates (50g agar, 600mL red grape juice, 42mL Nipagin (10% solution), 1.1L water) and larvae reared under a controlled density of 100 per vial. Adults were collected and separated by sex within 8 h of eclosion to ensure virginity and stored 10 per vial.
Sensory cues removal
Each male was randomly assigned into one of three treatments: housed with a rival male with all sensory cues intact (+ all), housed with a rival with the auditory cue removed (+ no sound), or housed alone (- all). The experiment was repeated in two independent replicates, which were pooled for analysis. The auditory cue of rival presence was removed by using a physical manipulation in which the wings of the rival males were removed under CO2 anaesthesia, preventing them from producing the song that signals their presence to competitors. To control for handling, the rival males in the +all treatment were also subjected to CO2 anaesthesia and the tips of their wings were clipped, allowing identification of the focal male but not affecting the capacity of rival males to produce song (Ehrman 1966). The focal and rival males in the +no sound and +all treatments were housed together in a single SYA vial. The males in the -all treatment were housed alone in a vial. Focal males were maintained in their respective treatments for three days.
Effect of cue removal on responses to rivals and reproductive success and sperm competitive ability
Virgin wildtype females were transferred to individual vials of SYA one day prior to mating. Each treatment male was introduced to a female directly from their rival treatments by using aspiration. Latency to mate (the time from when the male was introduced to when mating began) and mating duration were recorded to the nearest minute. Pairs that did not mate within 3 h were discarded. Males were removed from the vials by aspiration shortly after mating finished to prevent any rematings. Females were allowed to oviposit in a first set of vials for 24 h, following which they were moved to second set. The first set of vials were then incubated, and offspring that emerged from them were counted.
Approximately 24 h after the first mating, females were given the opportunity to mate a second time, to males with a ‘stubble’ (Sb ) mutation. Sb mutant individuals are identifiable by the shorter, thicker bristles on the back of the thorax (Overton 1967), allowing for offspring paternity to be determined by eye. Sb males came from aSb1 stock which had been backcrossed into the Dahomey wild type background at least 4 times. The proportion of females that remated was recorded and as in the first mating assay, the latency and duration of the rematings were recorded to the nearest minute. Pairs that did not mate within 3 h were discarded. Males were removed shortly after mating. Females were allowed to oviposit in the vials for 24 h, after which they were discarded. The vials were retained and incubated. Offspring that developed from eggs laid following the second mating had mixed paternity, some being fathered by the first (treatment) male and some by the second Sb male. Paternity was thus determined by the presence of the Sb phenotype allowing us to calculate the proportion of the offspring fathered by the first (treatment) male (P1) and by the second (Sb competitor) male (P2).
Statistical analysis
Statistical analyses were carried out in R v 3.4.2 (R Core Team 2016). The data from the two replicates were pooled, then analysed and plotted as one dataset with replicate as a random factor. Shapiro-Wilk and Levene’s tests were used to assess whether the data were normally distributed and whether variances were equal across treatments, respectively. Where the data were normally distributed or could be transformed to fit a normal distribution, mating duration and latency data were analysed using linear mixed models. When latency and duration data significantly differed from normal distribution and could not be successfully transformed, they were analysed using generalised linear mixed models with a gamma distribution and a log link.
Offspring counts from the first mating in both blocks were zero-inflated, so were analysed using hurdle models. The number of zero offspring counts in each treatment and the non-zero counts were manually separated. The number of zeroes was analysed with a binomial generalised linear mixed model. Where the non-zero offspring counts were normally distributed or could be transformed to fit a normal distribution, they were analysed with a linear mixed model. Otherwise, non-zero counts were analysed using a generalised linear mixed model with a Poisson distribution and a log link. In order to infer the effect of treatment on overall offspring counts, including zeroes and non-zeroes, Kruskal-Wallis tests were used.
The proportion of offspring produced following the second mating that were fathered by the treatment male (P1) was analysed as a dual response variable using a binomial generalised linear mixed model with a logit link.
Where significant treatment effects were found, pairwise differences between groups were determined using post-hoc Tukey tests with the ‘multcomp’ package (Hothorn et al. 2008).