Sexual selection in animals has been mostly studied in species in which males are signallers and females are choosers. However, in many species females are (also) signallers. In species with non-signalling females, virgin females are hypothesized to be less choosy than mated females, as virgins must mate to realize fitness and the number of available males is generally limited. Yet, when females signal to attract males, mate limitation can be overcome. We tested how virgin and mated females differ in their calling behaviour, mating latency, and in mate choice, using the tobacco budworm Chloridea (Heliothis) virescens as an example for a species in which females are not only choosers but also signallers. We found that virgin females signalled longer than mated females, but virgin and mated signalling females were equally ready to mate, in contrast to non-signalling females. However, we found that virgin signalling females showed weaker mate preference than mated females, which can be explained by the fact that females increase their fitness with multiple matings. Mated females may thus further increase their fitness by more stringent mate selection. We conclude that signalling is a crucial aspect to consider when studying female mate choice, because signalling may affect the number of available mates to choose from.

excerpt from Zweerus, N. L., van Wijk, M., Smallegange, I. M., & Groot, A. T. (2022). Mating status affects female choice when females are signalers. Ecology and Evolution, 12, e8864

Study organism

Chloridea (Heliothis) virescens populations originated from North Carolina State University (YDK strain) and the Max Planck Institute for Chemical Ecology, Jena, and have been reared at the Institute for Biodiversity and Ecosystem and Dynamics (IBED), University of Amsterdam since 2011. We conducted the experiment for this study between April 2018 and November 2019. The moths were kept in a climate chamber at 60% relative humidity and 25 ± 1°C, with a 14 h light (photophase): 10 h dark (scotophase) photoperiod (lights off at 11 a.m. CET). Larvae were reared on artificial pinto bean diet (Burton, 1970) in individual plastic cups (37 ml, Solo, Lake Forest, Illinois). Pupae were checked daily for eclosion (i.e., hatching of adult) and emerged adults were fed 10 % sucrose solution provided through 1 cm cotton dental wick. All experiments were conducted with 2–3-day old non-sibling individuals and were conducted under the same environmental conditions as those used for rearing.

Procedure to obtain mated females and maintaining virgin females

To obtain mated females, we placed an adult female (first day after eclosion) in a large clear plastic cup (473 ml, Solo) with an adult male that was randomly chosen from the adults of the standard rearing. At the same time, for the ‘virgin group’ females were also selected on the first day after eclosion and individually isolated. We next observed all pairs at 30 min intervals during the scotophase (dark period) until all pairs were mating, or until 9 hours of the scotophase had passed. Because C. virescens matings last for several hours, only females that mated for ≥ 60 min qualified for the ‘mated group’ of the experiment. In the photophase that followed the scotophase (i.e., light period on day 2), we individually isolated the mated females. We conducted the experiments on the third day after eclosion, under the same environmental conditions as the rearing and the preparatory matings.

Procedure to obtain an extended range of male body sizes

Previously we found that male body size affects female fitness: females mating with larger males had more offspring than when mating with smaller males (Zweerus et al., 2021). To assess female preference for males of different body sizes (a proxy for quality), we increased the range of male sizes by altering the larval diet, following Zweerus et al. (2021). Briefly, we obtained males of average to large size by rearing larvae on a standard pinto bean diet (Burton, 1970), and obtained males of smaller than average size by rearing larvae on diet which nutritional value was 25% of the standard diet. It is important to note that even on standard diet, male mass varies. The low-nutritional diet extended this ‘natural’ range of masses at the lower end (see Zweerus et al., 2021, Fig. S1). The diet treatment was thus a tool to increase the range of male masses without being a factor in the experimental design. For the experiments, we did not use males from one or the other diet, but rather males that differed in pupal mass.

Mate-attraction effort experiment

To compare mate attraction between virgin and mated females using their signalling behaviour as a proxy of mate attraction effort, we quantified the signalling activity of all females (virgin: n = 24 and mated: n = 21) in the third scotophase after eclosion as follows. In the hour prior to the scotophase, we placed single females into large clear plastic cups (473 ml, Solo). From the onset of scotophase, we observed and scored the number of females signalling every 15 minutes. We stopped observing an individual if the female did not signal for at least 1 hour. To test if the proportion of signalling females differed between the virgin and mated group, we used a Chi-square test for independence. To assess if virgin females (n = 21) started signalling later than mated females (n = 17), we analyzed the normally-distributed data with equal variances for the onset time of signalling with a two-tailed t-test. To test if the duration of signalling differed between virgin and mated females, we used a Mann-Whitney-U-test, because the data were non-normally distributed (assessed by Shapiro-Wilk test and visual exploration of histograms). We visualized the data by fitting the proportion of signalling females per time point and mating status over the scotophase using the package ggplot2 (Wickham, 2016) in the software R (version 4.0.5, R Core Team (2021)).

No-choice experiment to assess readiness to mate in virgin and mated females

To test the hypothesis that virgin females have a shorter mating latency (as a proxy of readiness to mate) than mated females, we measured their mating latency (the time from the pairing until copulation) in no-choice mating assays. Firstly, we placed one female with one male into a clear plastic cup (473 ml, Solo) and covered the cup with a mesh. We then mounted each plastic cup in a hanging grid with a camera (GoPro Hero silver) underneath. The assay started at the beginning of the scotophase, after which we recorded a time-lapse series of 1pic / min. We collected the data on minimally 4 and maximally 20 samples per time over 8 scotophases between 11^th and 20^th of May 2018. To determine mating latency, we identified the timestamp of the picture showing a newly-formed mating pair in the time-lapse series. For individuals that did not mate, we censored their data by assigning a maximum time span of 600 min, which corresponds to an entire scotophase. To test if virgin females (n = 61) mated significantly earlier than mated females (n = 29), and whether male pupal mass or female pupal mass affected mating latency, we fitted a Cox proportional-hazards model with the explanatory variables mating status, male pupal mass, and female pupal mass as main effects and mating latency as the response variable in R using the packages survival (Therneau, 2021), and survminer (Kassambara et al., 2019).

Two-choice experiment to compare virgin and mated female mate preferences

To test the hypothesis that virgin females show a weaker preference for larger (i.e., higher-quality) males than mated females, we conducted two-choice tests, in which we placed one female (virgin: n =  61 or mated: n = 63) together with a larger and a smaller male into a BugDorm cage (H30 cm x W30 cm x D30 cm), and scored which male a female mated. To distinguish between the two males, we marked each male by clipping the tip of one randomly chosen wing. The experiment started 10 min before the onset of the scotophase and we checked all cages at 30 min intervals. Once a mating pair had formed, we removed the unmated male from the cage. To check if the matings were successful, we isolated and froze all individuals in the next photophase, dissected the females, and quantified the number of spermatophores per individual. We collected the data over a total of 13 scotophases between 13^th Augustus and 8^thNovember 2019. Since not all females mated in the two-choice assay (virgin: n = 56, mated: n = 45), we first tested for an association between mating status (virgin, mated) and mating occurrence (mating, no mating) with a Fisher’s exact test. Additionally, we confirmed that the mass range of larger and smaller males did not differ significantly between males offered to virgin females and to mated females, using a Welch’s two-sample t-test. Finally, we determined if the size range between virgin females and mated females differed using a Welch’s two-sample t-test.

To assess whether the males that females chose to mate with were on average larger than the rejected males, we first tested if mean pupal mass differed significantly between the chosen and not chosen males by computing paired t-tests. To then test whether virgin and mated females differed in the strength of their mate preference, we first randomly selected the data of one male per cage (i.e., this male was either chosen or not chosen by the female). Since each female made one choice, this step ensured that the number of data points for the analysis corresponded to the actual number of choices made in the experiment. Subsequently, we modelled the response variable female choice as a function of the variable larger/smaller (male), indicating whether the male was larger or smaller compared to the other male in the same cage, the difference in male pupal mass between the two males per cage, female mating status, and their three-way interaction. Including a three-way interaction in the model allowed us to let the slopes of the function vary independently, and thus enabled us to identify the differences in female mate choice with respect to female mating status. We fitted the model (glm) with a binomial error distribution and produced the ANOVA table using the package car (Fox & Weisberg, 2019) in R. The results were visualized using the package ggplot2 (Wickham, 2016).