Alleles favoured by sexual selection in males can reduce fitness when expressed in females, resulting in intralocus sexual conflict. It remains unclear whether such conflict is fully resolved by the evolution of sexual dimorphism. If conflict persists, then female reproductive performance might covary negatively with the expression of secondary sexual traits in male relatives, and potentially with the expression of homologous traits in females themselves. However, because secondary sexual traits often exhibit strong condition dependence, a resource-poor developmental environment could weaken the covariance between female fitness and the expression of such traits. We tested these predictions using a split-brood experiment in which neriid flies (Telostylinus angusticollis) were reared on nutrient-rich or nutrient-poor larval diets, resulting in high (rich diet) and low (poor diet) adult sexual dimorphism. Consistent with predictions, in families reared on the rich larval diet, we found that females whose brothers exhibited relatively large secondary sexual traits produced fewer viable offspring. Moreover, rich-diet females with relatively large homologues of male sexual traits exhibited increased latency to oviposition. By contrast, in families reared on the poor larval diet, we found no evidence of negative covariation between male secondary sexual trait expression and female performance, and females with relatively large male trait homologues exhibited reduced latency to oviposition and increased fecundity. Our results confirm that sexually dimorphic morphology can reflect sexually antagonistic fitness variation, and suggest that intralocus sexual conflict remains unresolved in this species. Our results also suggest that the nutritional environment can modulate the signal of sexual antagonism.

Experiment set-up

Approximately 30 neriid fly (Telostylinus angusticollis) individuals were collected from a natural population in Fred Hollows Reserve (33°54 44.04S 151°14 52.14E), Sydney, Australia, and mixed with stock flies (around 20 individuals) that had originated from the same location and had been reared in the lab for several generations. The flies were maintained in an 8L container with a mesh top and a layer of cocopeat on the bottom in a controlled-temperature room (~25 °C, 12 h light/dark cycle). The flies were supplied with food (sugar and yeast) ad libitum and watered periodically. To collect eggs, we provided oviposition medium in a large petri dish. The oviposition medium was prepared using rich larval food (described below) that was kept in a room at 25 °C for approximately seven days to allow mould growth, and watered and mixed periodically. From this lab population, 430 eggs were collected and raised on a standard larval diet in 500mL containers, with ~40 eggs per 200 g of food. The standard larval medium consisted of 10.9 g of soy protein (Nature’s Way brand, Pharm-a-care Pty. Ltd., Warriewood, NSW, Australia), 29.7 g of brown sugar (Coles brand, Bundaberg, Australia) and 500 mL of water per L of dry cocopeat. The larval medium was homogenized using a hand-held beater and frozen at –20°C until the day of use. The larval containers were incubated in a controlled-temperature chamber set to 25 °C ( 1.5 °C) and a 12 h light/dark cycle.

After the F0 adult flies emerged, 100 male-female pairs were formed and placed in separate 200 mL vials with a substrate of moist cocopeat (Figure 4). After two weeks, when they reached maturity (Wylde et al. 2019), a small petri dish with oviposition medium was provided to each of the pairs, and the flies were given 5 days to oviposit. We collected eggs from pairs that laid at least 40 eggs (n = 62 pairs, 2,480 eggs in total). From each of these pairs, 20 eggs were transferred to a container with poor larval diet (20 eggs per 100 g of food) and 20 eggs were transferred to a container with a rich larval diet (20 eggs per 100 g of food). The poor diet consisted of 5.5 g of protein and 14.8 g of brown sugar, whereas the rich diet consisted of 32 g of protein and 89 g of brown sugar, both mixed with 500 mL of water and one L of dry cocopeat (Bonduriansky 2007a; Sentinella et al. 2013). The vials with the larval medium were placed inside one-L containers with a layer of cocopeat to facilitate pupation and collection of F1 adult flies (N = 1,148). The full-sibs descended from a male-female pair and split between rich and poor larval diets constitute a “family” in the analysis. A further 520 eggs were collected from flies that were not used to produce the experimental families and reared on a standard larval diet to generate standard males for crosses used to assess female performance (see below).

A maximum of 24 hours after emergence of the experimental flies (i.e., F1 flies from the 62 families) and standard flies from their puparia, males and females from each family were separated by sex to avoid mating. Upon emergence, two males randomly selected from each family were allowed 24 h for their exoskeletons to sclerotize fully and then frozen at −20°C for morphological measurements. We chose to measure two males from each family as there is little variation in body size or shape within families. Families that did not produce enough males or females were excluded from the analyses. Males were imaged using a Leica MC170HD camera mounted on a Leica MZ16A stereoscope (Wetzlar, Germany). For each male, we measured the lengths of the thorax as a proxy of body size, and the combined lengths of the head capsule and the right antenna as an index of secondary sexual trait expression, using ImageJ (Schneider et al. 2012). We used the residuals of the linear regression between family mean male secondary sexual trait length and family mean male thorax length to identify the F1 families with the lowest and highest expression of male secondary sexual traits relative to body size within each larval diet treatment group. Families with the largest negative residuals (12 families from rich larval diet and 10 families from poor larval diet) or largest positive residuals (12 families from rich larval diet and 10 families from poor larval diet) were selected as focal families for experimental assays.

From this subset of F1 families, two females (replicates) per family were randomly selected for the reproductive performance and longevity assays. Females were paired with standard males in separate 1 L containers and provided with oviposition medium. Females that developed in rich diet were paired with males at 10 +/- 2 days old, and females that developed in poor diet were paired with males at 20 +/- 2 days old as poor diet females take longer to reach reproductive maturity (Wylde et al. 2019). Eggs were counted every second day for eight days. From each female, twenty eggs (where possible) were transferred to containers with a standard diet (20 eggs per 100 g of food) in order to quantify F2 offspring viability, a measure that incorporates egg hatching success and larval and pupal survival until the adult stage. We recorded the number of F2 adults that eclosed from the eggs over 14 days after the first adult eclosion in each replicate brood. To estimate F1 focal female longevity, the focal females were kept with their male partner in the same 1 L containers and provided with food (sugar and yeast) and water but not oviposition medium. Mortality was checked and recorded three times per week. Following death, the focal females were imaged and their head capsule length, antenna length and thorax length were measured from the images as described above.

Statistical analyses

We tested whether larval diet manipulation affected male and female body size using a Gaussian linear mixed model with thorax length as the dependent variable, larval diet as the predictor, and family identity as a random effect using the package glmmTMB (McGillycuddy et al. 2025). For this analysis, thorax length was standardized (z-transformed; mean = 0, standard deviation = 1) within sexes but across both larval diet treatments. We used a similar model to test larval diet effects on male and female body shape, quantified as head length/thorax length (using unstandardised morphological measurements). To test for covariation of relative head lengths of male and female siblings, we first carried out separate linear regressions of head length on thorax length (both variables standardised within sex × larval diet combinations) within each larval diet × sex combination and obtained the standardised residuals for each individual (representing individual relative head length). We then calculated the standardised family-mean residual for each sex within each larval diet. Finally, we fitted a Gaussian linear mixed model with family-mean female residual head length as the dependent variable and family-mean male residual head length, larval diet, and their interaction as fixed effects, as well as family identity as a random effect. We also fitted separate mixed models for each larval diet treatment group.

We then tested for effects of brothers’ relative head length and focal female relative head length on focal female performance. We built linear mixed models that included larval diet, brothers’ relative head length (a categorical variable representing the sign of the family mean residual male head length), female relative head length (a continuous variable representing residual head length of each individual focal female), female body size (thorax length, standardized within larval diets to avoid redundancy with the categorical effect of larval diet), and the larval diet × brothers’ residual head length and larval diet × female residual head length interactions, as fixed effects using glmmTMB. We used a Gaussian linear mixed model to test effects on the latency to lay eggs and generalized linear mixed models to test effects on the number of eggs laid (Poisson error distribution) and offspring viability (binomial error distribution, eggs that produced viable F2 adults versus eggs that did not produce viable F2 adults). We included an observation-level random effect to account for overdispersion when modelling the number of eggs laid. To test for effects on female survival, we fitted a Cox proportional hazards model using the package coxme (Therneau 2010). Family identity was included as a random effect in all models. Effects in glmmTMB and coxme models were tested using Wald z-tests. For dependent variables that yielded significant or near-significant interactions, we investigated the data further by carrying out post-hoc Tukey tests (for the larval diet × brothers’ residual head length interaction) or separate analyses within each larval diet treatment group (for larval diet × female residual head length interactions). All statistical analyses were conducted using R version 4.4.2 (R_Core_Team 2021).