Understanding how species adapt to different temperatures is crucial to predict their response to global warming, and thermal performance curves (TPCs) have been employed recurrently to study this topic. Nevertheless, fundamental questions regarding how thermodynamic constraints and evolution interact to shape TPCs in lineages inhabiting different environments remain unanswered. Here, we study Drosophila simulans along a latitudinal gradient spanning 3,000 km to test opposing hypotheses based on thermodynamic constraints (‘hotter-is-better’) versus biochemical adaptation (‘jack-of-all-temperatures’) as primary determinants of TPCs variation across populations. We compare thermal responses in metabolic rate and the egg-to-adult survival as descriptors of organismal performance and fitness, respectively, and show that different descriptors of TPCs vary in tandem with mean environmental temperatures, providing strong support to hotter-is-better. Thermodynamic constraints also resulted in a strong negative association between maximum performance and thermal breadth. Lastly, we show that descriptors of TPCs for metabolism and egg-to-adult survival are highly correlated, providing evidence of coadaptation and that curves for egg-to-adult survival are systematically narrower and displaced towards lower temperatures. Taken together, results support the pervasive role of thermodynamics constraining thermal responses in Drosophila populations along a latitudinal gradient, that are only partly compensated by evolutionary adaptation. 

Animal collection and maintenance

The study collected D. simulans adult flies from eight locations in Chile, spanning almost 3,000 km from North to South, during summer 2018 and 2019. The flies were caught using nets over banana or decomposing fruit traps and brought to the laboratory for taxonomic identification. Separate laboratory lines were established from each location and maintained in plexiglass cages under controlled conditions for at least three generations before experiments. Two replicate subcultures per line were maintained in separate cases, and populations were pooled every generation to minimize genetic drift. Maternal and grand maternal effects were removed by maintaining individuals under common garden conditions for at least three generations prior to experiments.

Metabolic rate

The study measured the standard metabolic rate (SMR or VCO2) in adult D. simulans flies using CO₂ production through a flow-through respirometry design. Individual female and male flies were measured in custom-made glass metabolic chambers at temperatures ranging from 12 to 37 °C. A total of 1,321 flies were measured, and data were extracted using LabAnalyst software. SMR was estimated for each 5 min period as the 30s average when VCO₂ was stable.

Overall egg-to-adult survival

New food plates were placed in maintenance cages to allow D. simulans flies to oviposit for 24 hours. Eggs were then collected and 20 eggs from each plate were transferred to vials with fresh culture medium at temperatures ranging from 12 to 32°C. Hatching was monitored every 24 hours until the last egg hatches or for up to 60 days. Egg-to-adult survival rate was assessed as the ratio of viability to development time, serving as a measure of fitness. A total of 1,119 vials were used for the experiment, with 20 replicates per temperature and population.

Statistical analysis

Descriptor parameters of the TPCs, including optimal temperature (T_opt), critical thermal maxima (CT_max), thermal breadth (T_breadth), and maximum performance (P_max), were extracted from nonlinear regressions. A principal component analysis was used to collapse TPC parameters into a single response variable, which was then tested with multiple regressions to determine the effects of diurnal mean temperatures and thermal range. The degree of correlation between P_max, T_opt, T_breadth, and CT_max was assessed for each trait, as well as the correlation between P_max and T_opt and T_breadth separately for metabolic rate and egg-to-adult survival. The assumptions of normality and homoscedasticity of residuals were validated visually, and all statistical analyses were performed using R software.

R-software is required to open and analyse the data files.