Environmental factors such as dietary nutrients can shape the expression of developmentally plastic sexual traits in many species. However, while there has been extensive research into the developmental plasticity of sexual traits at the individual level, the broader consequences of this variation at the population scale remain poorly understood. Here, we asked whether plastic responses to the developmental environment can shape sexual competition and initiate reproductive isolation between populations. We reared neriid flies, Telostylinus angusticollis, on nutrient-rich and nutrient-poor larval diets, generating adult flies that differed in body size and secondary sexual trait expression. We then investigated sexual competition in experimental populations from each developmental environment, and tested for reproductive isolation between flies from mismatched environments. We found that, compared with poor-diet populations, rich-diet populations exhibited more frequent and escalated male-male combat and more frequent mating and mate-guarding. However, we found no evidence that sexual selection was affected by the developmental environment. Mismatched female-male pairs tended to take longer to mate and rich-diet females often rejected poor-diet males, but mismatched pairs were not less likely to mate within 1 hour or produce viable offspring. Our findings suggest that developmental plasticity could generate dramatic differences in sexual competition between populations, and could contribute to reproductive isolation.

Rearing and Culturing of Flies

The individuals used in the experiments described below were second-generation individuals reared from T. angusticollis collected from Fred Hollows Reserve in Randwick, New South Wales, Australia (33˚54’44.04”S, 151˚14’52.14”E). The flies were housed in population cages with moist cocopeat and were given three separate petri dishes containing brown sugar, yeast, and oviposition medium. The oviposition medium consisted of a nutrient-rich diet (as described below) which had been left to mould for approximately 4 days and then mixed to encourage oviposition. The cages were sprayed with reverse osmosis (RO) water every second day. To generate focal flies for experiments, eggs were collected from oviposition medium  in stock cages and transferred to 500 mL jars containing 200 mL of either a nutrient-rich diet (“rich diet”) or a nutrient-poor diet (“poor diet”) to represent developmental environments differing in availability of macronutrients. Approximately 20 eggs were transferred to each container, with 19 containers for each diet treatment (N = 760 eggs in total). The rich diet consisted of 32.9 g soy protein (Nature’s Way Instant Natural Protein) and 41.3 g of brown sugar (Coles Brown Sugar), whereas the poor diet consisted of 5.5 g soy protein and 6.9 g brown sugar, mixed with 1 L dry cocopeat (coconut husk shavings), and approximately 800 mL RO water (Sentinella et al., 2013). Virgin adults were separated within 24 hours after emergence into 12 L plastic tanks by sex and diet, with approximately 20 males or 25 females per tank (N = 18 tanks in total). Tanks contained adult flies of similar age, with flies from multiple larval containers combined randomly in adult tanks. Each tank contained a layer of moist cocopeat for humidity, and petri dishes containing an excess of brown sugar, yeast, and oviposition medium that were replaced approximately every 7 days. The tanks were sprayed with RO water every second day. Flies, larval medium containers, and experiments were all kept and conducted in a controlled lab environment with a temperature of 25 °C (± 2 °C). The focal flies were used in two separate experiments, as described below.   

Sexual Competition Experiment

We observed a total of 13 experimental replicate populations (i.e., unique groups of 5 males and 5 females combined and observed inside an experimental arena) from each larval diet (N = 26 replicate populations in total). Each replicate population was created by combining five virgin females and five virgin males (all of similar age and reared on the same larval diet, and each male marked with a different colour as described below) together in a transparent plexiglass arena (26 x 36 x 20 cm) with a mesh sleeve (Fig. 2). Replicate populations were comprised of focal individuals drawn randomly from the same adult tank (when possible), and focal individuals were not re-used in this experiment. Focal flies were between 14 and 45 days old when used in experiments. Flies reared on the poor larval diet emerged ~ 3 days later on average than flies reared on the rich larval diet, as typically observed in this species (Bonduriansky, 2007; Hooper et al., 2017). However, since T. angusticollis can live for >130 days in the laboratory (Hooper et al., 2017), all focal flies were relatively young when used in experiments and the small average difference in adult age between rich- and poor-diet focal flies is unlikely to have substantially influenced our results. The arena was designed to simulate natural mating aggregations on rotting tree bark, where females oviposit and feed and males compete for matings (Bonduriansky, 2006, 2007). The bottom of the arena contained a layer of moist cocopeat, and a petri dish (5cm diameter) filled with oviposition medium was placed in the centre. The arena was illuminated with a broad-spectrum light source placed above the oviposition medium to encourage interaction.

In order to identify individual males in replicate populations, the focal males were anaesthetised with CO₂ using a Flystuff Benchtop Flow Buddy System (59-122BCU, Genesee Scientific, USA) with Ultimate Fly Pad and Gun, and a spot of enamel paint (Tamiya Color Enamel Paint, Japan) was applied to the thorax, with each of the five males in each replicate population marked with a different colour. The males were placed in individual vials to allow the paint to dry and for recovery from the anaesthesia, and then kept in tanks with other males (as described above) for at least 24 hours before being allocated to replicate populations for the experiment.

During each day of the experiment, we collected data on one rich-diet replicate population and one poor-diet replicate population using the same dish of oviposition medium. Each replicate population was left in the arena for 30 minutes to acclimatise, and was then observed for one hour. For each individual male, we recorded the duration and number of matings and mating attempts, number and duration of guarding bouts after mating, number of male wing flicking bouts, number and duration of combat interactions, and number of times the male was rejected by a female. Mating was recorded when a male positioned himself above or behind a female and mounted the female for at least 20 s (Bath et al., 2012; Wylde et al., 2019b). Mating duration was quantified as the time in seconds between a male mounting a female and removing his genitalia from the female’s oviscape (Wylde et al., 2019b) (Fig. 1a). A mounting interaction that lasted less than 20 s was counted as a mating attempt. Female rejection was recorded when a female kicked a male or flew and/or ran away when a male attempted to mate with her (Bath et al., 2012). Guarding was identified as a male standing above a female after mating and using the span of his legs to enclose the female (Bonduriansky, 2006) (Fig. 1b). Male-male combat interactions were recorded when two males used their forelegs and/or body to strike each other (Bath et al., 2012) (Fig. 1c). Wing flicking involved males flicking their wings at another individual (Wylde et al., 2019b).

Reproductive Isolation Experiment

To determine whether individuals reared on different larval diets can mate and produce offspring, males and females were paired in all possible combinations (poor male with rich female, poor male with poor female, rich male with poor female, and rich male with rich female), with 23 replicates of each combination (Fig. 3). To ensure that the males were sexually experienced, males from the sexual competition experiment were re-used in the reproductive isolation experiment. However, all females were virgins. Each pair was placed in a 500 mL container with moist cocopeat on the bottom and a small petri dish of oviposition medium. After the female and male flies were combined, they were left to adjust for 10 s, and then observed for one hour. We recorded latency to mate, duration and number of matings, number of mating attempts, and female rejection behaviours. Here, female rejection was evidenced by a female wing-flicking or kicking a male as he tried to mate, or when a female ran/flew away from a male during a mating attempt. On each day of the experiment, we set up one replicate of each of the four larval diet combinations. A broad-spectrum light source was positioned above the containers throughout the experiment.

After one hour, the male was removed from the container and the female was left to lay eggs. Each oviposition dish was checked for eggs every day for a maximum of six days after the pairing was conducted. After six days, the female was removed and frozen for body size measurement. For each pair, 20 eggs (where possible) were transferred to a container with 200 mL of standard larval diet consisting of 11 g soy protein and 13.8 g brown sugar per 1 L of dry cocopeat and 800mL of water (Sentinella et al., 2013). After the first adult emergence in each container, the container was left for 10 days and then the emerged flies were counted. The larval containers with eggs but with no emerging flies were left for 40 days before they were discarded.

Body Size Measurement

The focal males used in the sexual competition experiment and reproductive isolation experiment, and females used in the reproductive isolation experiment, were frozen at -20°C and then used to quantify body size. One wing from each individual was removed, mounted on a microscope slide using double-sided tape, and then covered with cling-wrap. The wings were then photographed at a magnification of 6.3 x using a Leica MC170 HD camera mounted on a Leica MS5 stereo-microscope (Wetzlar, Germany). The linear distance from the intersection of the R₂₊₃ wing vein with the wing margin to its intersection with the R_s wing vein was measured using ImageJ software (Schneider et al., 2012). 

Statistical Analysis

All statistical tests were carried out using R 4.0.3 (R_Core_Team, 2020). To verify that our larval diet manipulation resulted in the expected effects on adult phenotype, we modelled wing length using a linear model with larval diet, sex and their interaction as fixed effects. Although we did not keep track of family identity or larval container, the flies were derived from eggs laid in stock tanks containing numerous flies and eggs were distributed among 38 larval containers. Effects of genotype and shared (container-specific) environments are therefore unlikely to have biased the results of this analysis.

In the sexual competition experiment, some males did not fight or interact with females, such that the individual data set was zero-inflated. Thus, analysis was carried out on means (for duration of mating, combat interactions, guarding) or total counts (number of matings, guarding bouts, female rejections, combat interactions, and wing flicks) for replicate populations. General and generalised linear mixed models (GLMMs) were fitted to the replicate population means or sums using the R package lme4 (Bates et al., 2015), with each response variable modelled separately. Models included larval diet as a fixed effect, and day as a random block effect (but the random effect of day yielded extremely small variance components and was removed from two models to facilitate model fit: see Table 1). Combat duration was log-transformed to improve model diagnostics. For over-dispersed count data, an observation-level random effect (unique code for each replicate population) was included. For Gaussian models, effects were tested using t-tests with Satterthwaite’s degrees of freedom using the package LmerTest (Kuznetsova et al., 2017). For Poisson models, effects were tested using z-tests. For rejection behaviours by females, the model failed to converge and a Wilcoxon matched-pairs test was used instead to compare data from rich- vs. poor-diet replicate populations within days.

Mating success skew represents the opportunity for sexual selection (Cattelan et al., 2020; Jones, 2009). To estimate mating success skew, we calculated coefficients of variation of the number of matings by individual males within replicate populations. The coefficients of variation were then compared using a Gaussian mixed-effects model with larval diet as the fixed effect and day as a random block effect. To test for and compare sexual selection on male body size in rich-diet vs. poor-diet replicate populations, we used a Poisson model of individual male mating success (number of matings) as a function of male body size, with larval diet and body size as fixed effects and day as a random block effect. For this analysis, individual male mating success and body size (wing length) were both standardized (z-transformed) within replicate populations. In this model, a main effect of male body size would indicate overall sexual selection on male body size, whereas a larval diet × body size interaction would indicate a difference between larval diet treatment groups in sexual selection on male body size. Note that the main effect of larval diet is not meaningful in this model because standardization within replicate populations brings the effect estimate to ~ 0.

To test for reproductive isolation, models were fitted with male larval diet, female larval diet, and their interaction as fixed effects and day as a random block effect. An environmentally induced reproductive barrier would be indicated by reduced propensity to mate or produce viable offspring by flies reared on mismatched diets, detectable statistically as a male larval diet x female larval diet cross-over interaction. Mating outcome and female rejection behaviour were modelled as binomial (0 or 1) response variables. A binomial model was also used to investigate treatment effects on egg-to-adult viability, represented for each replicate pair as a matrix of successes (number of eggs that resulted in adult flies) and failures (number of eggs that did not result in adult flies). For over-dispersed binomial data, an observation-level random effect (unique code for each pair) was included in the model. Latency to mate was modelled with Gaussian error.