Skip to main content
Dryad logo

Data from: Does local adaptation along a latitudinal cline shape plastic responses to combined thermal and nutritional stress?


Chakraborty, Avishikta; Sgrò, Carla M.; Mirth, Christen Kerry (2020), Data from: Does local adaptation along a latitudinal cline shape plastic responses to combined thermal and nutritional stress?, Dryad, Dataset,


Thermal and nutritional stress are commonly experienced by animals. This will become increasingly so with climate change. Whether populations can plastically respond to such changes will determine their survival. Plasticity can vary among populations depending on the extent of environmental heterogeneity. However, theory conflicts as to whether environmental heterogeneity should increase or decrease plasticity. Using three locally-adapted populations of Drosophila melanogaster sampled from a latitudinal gradient, we investigated whether plastic responses to combinations of nutrition and temperature increase or decrease with latitude for four traits: egg-adult viability, egg-adult development time, and two body size traits. Employing nutritional geometry, we reared larvae on 25 diets varying in protein and carbohydrate content at two temperatures: 18ºC and 25ºC. Plasticity varied among traits and across the three populations. Viability was highly canalized in all three populations. The tropical population showed the least plasticity for development time, the sub-tropical showed the highest plasticity for wing area, and the temperate population showed the highest plasticity for femur length. We found no evidence of latitudinal plasticity gradients in either direction. Our data highlight that differences in thermal variation and resource predictability experienced by populations along a latitudinal cline are not sufficient to predict their plasticity. 


Fly stocks

We used mass bred populations collected from tropical (Townsville, latitude:19.29S), sub-tropical (Ballina, latitude: 28. 75), and temperate (Melbourne, latitude: 37.73) regions along the east coast of Australia (Supplementary Figure 1). The flies to seed these populations were collected in April 2016, and were maintained as mass bred populations at a population size of approximately 1500 flies at constant temperature of 25ºC on a 12-hour light/dark cycle on yeast-dextrose-potato medium (potato flakes 20g/L; dextrose 30g/L; 95 Brewer’s yeast 40 g/L; agar 7g/L; nipagen 6mL/L; and propionic acid 2.5 mL/L), for 55 generations prior to the experiments described below.

Nutritional geometry

Twenty-five diets were used in this study, following Kutz et al., (2019). These diets were one of five protein and carbohydrate (P:C) ratios, 1:8, 1:4, 1:3, 2:3, or 3:2, made by changing the quantities of inactive yeast, dextrose, and potato flakes. Each ratio was prepared at two concentrations of 2547.4kcal and 1273.7kcal. The 1273.7kcal was diluted for each ratio sequentially by 50% to make the remaining three caloric concentrations: 636.84kcal, 318.42 kcal, 159.21kcal per P:C ratio.

To obtain focal flies for the experiments, parental flies from each of the three populations were acclimated to egg laying chambers containing standard food for 24 hours, changing the egg plates every 12 hours. Eggs were collected from an overnight egg lay, and 20 eggs were transferred from each population into vials containing 7ml of each of the experimental diets. We used two constant thermal conditions, 18oC and 25oC to impose shifts in thermal regime.  These two temperatures represent the average winter and summer thermal conditions experienced along the east coast of Australia, from temperate Melbourne up to tropical Townsville (

Five replicate vials for each population, Ballina, Melbourne, and Townsville, and for each of the 25 diets were placed at either 25ºC or 18ºC, where they were left to develop from egg to adult eclosion.  Temperature and humidity-controlled cabinets with 12-hr light and 12-hr dark photoperiod were used for the experiment. The vials were moved within the cabinets every 12 hours to remove any biases of temperature and light gradients.

Egg-to-adult development time and egg-adult-viability

Egg-to-adult development time was recorded as the time in hours from the mid-point of egg laying to adult eclosion. Once counted, flies were removed from their vial. Vials were checked every 12 hours until four consecutive time points yielded no flies, after which vials were discarded. Egg-to-adult viability was measured as the proportion of adults that emerged from the 20 eggs transferred into each replicate vial.

Adult body size

The sizes of D. melanogaster body parts differ in their response to nutrition and temperature (Kutz et al. 2019; Shingleton et al. 2017). To account for these differences, body size was estimated using wing centroid size and femur length. Thirty female flies from each treatment were preserved in 70% ethanol/30% glycerol solution (SH solution) and were subsequently used for dissections. We chose to focus on female flies, as the size of their body parts is known to be more sensitive to nutrition (Shingleton et al. 2017), and because assessing sex-specific plasticity was not the focus of the present study. Left first legs and left wings were mounted in SH solution on a microscope slide and photographed using a Leica M80 stereo microscope (Leica, Heerbrugg, Switzerland). All images were processed using ImageJ software. Wing area was estimated by calculating the centroid size from landmarks from 6 vein positions (Supplementary Figure 2a).  Femur length was estimated as the distance between two landmarks: the anterior-most junction between the femur and coxa, and the posterior-most junction between the femur and tibia (Supplementary Figure 2b)

Statistical analyses

The response of all traits to temperature and to both the protein and carbohydrate composition of the diet was analysed following Kutz, et al; 2019. For development time, wing size, and femur length, data were fit using linear mixed effects models. Survival (egg-to-adult viability) was fit using generalised linear models assuming a binomial distribution. Fit of the data was validated by visual inspection of the residuals. The full models include the following:

Y = P + P2 + C + C2 + Temp + Pop + P*C + Temp*P + Temp*P2 + Temp*C + Temp*C2 + Temp*Pop + Pop*P + Pop*P2 + Pop*C + Pop*C2 + Pop*P*C + Temp*P*C + Temp*P*Pop + Temp*P2*Pop + Temp*C*Pop + Temp*C2*Pop + Temp*Pop*P*C ,

where P is protein, C is carbohydrate, Temp is temperature, and Pop is population. Experimental block (for development time) and replicate vial (all traits) were included as random effects.

Analyses were first performed on the full dataset for each trait to determine if there were significant interactions between protein concentration, carbohydrate concentration, temperature, and population. To estimate plasticity, we next scaled the data to a mean of 0 with unit standard deviations. We first compared differences in plasticity by comparing the shapes of the response surfaces, using partial F tests. We then contrasted trait responses to the linear and quadratic components of protein and carbohydrate concentration across populations and temperature, similar to methods described in Morrissey and Liefting (2016) using emtrends from the emmeans package in R. Thin-plate spines (TPS) were used to visualize the response surfaces for each trait, following Kutz, et al (2019). All statistical analyses were performed in R Studio (version 3.4.1, R Development Core Team 157 2017,


Australian Research Council, Award: DP180103725

Australian Research Council, Award: FT170100259