Offspring wet weight and fat
Across all treatments, wet weight of pupae increased and then declined with maternal age until the end of the experiment (Fig. 3). Model selection results provided strong evidence for a quadratic effect of maternal age on offspring wet weight (control – quadratic effect including random intercept and slope ω = 0.719; mating delay – quadratic effect with random intercept and slope ω = 0.562, without random slope ω = 0.272; and nutritional stress – quadratic effect with random intercept and slope ω = 0.866) (Tables S6 7-9). There was no support for a log, or a linear effect of maternal age for any of the treatments (ω = 0.000). In addition, across treatments, the predicted effect of mother age at the population level, assuming a quadratic fit, was within the standard errors of predictions from GAM fits to the data (S8 File).
The fitted quadratic curves and coefficients were similar between treatments (from the model with lowest AIC, which included a random intercept and slope, coefficient for maternal age: control 0.520 (0.466 – 0.574); mating delay 0.564 (0.406 – 0.723); and nutritional stress 0.502 (0.420 – 0.584), coefficient for maternal age squared: control -0.0040 (-0.0045 – -0.0035); mating delay -0.0042 (-0.0054 – -0.0030); and nutritional stress -0.0040 (-0.0048 – -0.0032)) (Tables S6 10 - 12).
For all three treatments, there was strong evidence for individual variation among mothers in the wet weight of their pupae (for all treatments, all models without random effects ω = 0.000). This variation was highest in the mating delay treatment with a random intercept variance from the most parsimonious model of 11.820, compared with 7.307 for the control and 2.379 for the nutritional stress treatment (Tables S6 13 - 15).
There was evidence for variation among mothers in the effect of maternal age on wet weight for the control and nutritional stress treatments (quadratic fit with random intercept and slope: control ω = 0.719; and nutritional stress ω = 0.866). There was less support for this in the mating delay treatment (quadratic fit with random intercept and slope ω = 0.562, next lowest AIC model - quadratic fit with random intercept only ω = 0.272).
Offspring fat increased linearly with offspring wet weight (Pearson’s correlation coefficient 0.554) (File S9), thus for brevity we focus on wet weight as our trait of interest, but the patterns are qualitatively similar if offspring fat is considered instead.