Direct observation of evolution in response to natural environmental change can resolve fundamental questions about adaptation including its pace, temporal dynamics, and underlying phenotypic and genomic architecture. We tracked evolution of fitness-associated phenotypes and allele frequencies genome-wide in ten replicate field populations of Drosophila melanogaster over ten generations from summer to late fall. Adaptation was evident over each sampling interval (1-4 generations) with exceptionally rapid phenotypic adaptation and large allele frequency shifts at many independent loci. The direction and basis of the adaptive response shifted repeatedly over time, consistent with the action of strong and rapidly fluctuating selection. Overall, we find clear phenotypic and genomic evidence of adaptive tracking occurring contemporaneously with environmental change, demonstrating the temporally dynamic nature of adaptation.

Taken from SI of Rudman et al. 2022, Science

Establishment of experimental populations 

To examine the pace, magnitude, parallelism, and genomic architecture of adaptation in response to a temporally variable environment we created a genetically diverse founder population that was seeded into each outdoor replicate. This outbred founder population was constructed from 80 fully sequenced Drosophila melanogaster inbred lines to facilitate the use of haplotype inference to attain high effective sequencing coverage. These inbred lines were derived from wild-caught individuals collected June 1, 2012 from Linvilla Orchards, Media PA USA (1). Each line was subsequently inbred for 20 generations by full-sib mating during which time they were maintained at 25 ˚C with a 16:8 photoperiod and fed ‘Spradling Cornmeal Recipe’ media. Then, 30-50 individuals from each line were pooled for whole genome sequencing. Sequencing and variant calling were performed as described in (2), with the addition that genomic DNA from certain lines was resequenced on an Illumina HiSeq X to increase coverage to a minimum of 10x for all lines. Mapped and de-duplicated bam files from all original and resequencing runs can be found on SRA under BioProject PRJNA722305 (Table S1). To initiate the baseline population in this experiment, we combined 10 males and 10 females from each of the 80 lines into large cages in May 2014. We allowed 4 generations of unmanipulated recombination and population expansion to facilitate recombination between lines before using 500 males and 500 female flies to found each of 10 field cages. Inbred lines have many deleterious alleles; purifying selection against deleterious alleles fixed during inbreeding was likely strong during lab outcrossing, and potentially, the early field phase of the experiment.

Each field cage is a 2m x 2m x 2m mesh enclosure around a dwarf peach tree located outdoors (Philadelphia, PA) and features a natural insect and microbial community. The ground was fresh soil with clover planted as ground cover in each cage. The only food source and egg-laying substrate was 400ml of Drosophila media (‘Spradling cornmeal recipe’) contained in 900cm3 aluminum loaf pans that were added every second day for the duration of the experiment (July 13th - November 7th, 2014). After two days, each loaf pan was covered with a screen mesh lid and eggs were allowed to develop until eclosion ceased. Loaf pans of media within experimental cages were protected from rain and direct sun on shelving units oriented away from direct sunlight.

Measurement of population size and evolution of fitness associated phenotypes 

Census size of adults was estimated in each replicate over the course of the experiment by photographing an equal amount of the surface area (approximately 2.5%) of the ceiling in each cage at dusk (12 total census estimates per cage). This protocol did not disturb flies perched on the ceiling. The number of adult D. melanogaster in each of 8 standardized photographs in each estimate for each cage was counted and multiplied by 40 to correct for total surface area and obtain census estimates. Egg production was estimated every second day by counting the eggs present on a 1/24th portion of the exposed surface of the media. When adult population sizes were large and the weather was warm the density of eggs on the media was considerably higher than optimal lab conditions.

To assess the rate and direction of phenotypic evolution over the course of the experiment we collected ~2500 eggs from each cage, brought them to the laboratory, and reared them for an additional 2 generations in a common garden (25˚C, 16L:8D) while maintaining population sizes at ~2500 individuals. Fitness-associated phenotypes were measured on density and age-controlled replicates in the F3 generation. Fecundity was measured as the total number of eggs laid by a group of five females, counted each day for a period of three days, with twenty replicate vials for each cage at each time. Egg length was measured using a microscope and image processing software (3) on at least 15 eggs (average of 27) from each cage at each time point. Larval development rate was tracked as the time from when eggs were laid until pupation in three replicate vials from each cage at each time point with 30 eggs in each vial. Starvation tolerance was measured as time to starvation in three replicate vials containing moist cotton (1.5 ml water) (following (4)) and 10 female flies with three replicates for each cage at each time point. Desiccation tolerance was measured as time to death in desiccation chambers containing 10 female flies with three replicates for each cage at each time point (4). Chill coma recovery was measured as the time it took for flies buried in ice and placed in a 4℃ incubator for 2h to resume an upright stance at 25℃ (1). This was measured using groups of 10 female flies for each cage at each time point that had been allowed at least 24hrs to recover from light CO2 anesthetic. We also attempted to measure evolution in heat knockdown. However, the assay temperature we used for the founder population, a stressor that caused 50% of flies to knockdown by 12 minutes, was not sufficiently hot to cause knockdown by the second sample period. Thus, although we cannot quantify it, heat tolerance evolved rapidly. We assayed each of the remaining phenotypes in the founding population (founder assays failed for fecundity and egg size) and at four times during the experiment: day 11 (7/25/14), day 38 (8/19/14), day 61 (9/11/14), and day 90 (10/10/14). Census and phenotypic evolution data have been uploaded to Dryad. 

Calculation of evolutionary rates and statistical analysis of phenotypic data to test for evolutionary parallelism 

We calculated evolutionary rates in Haldanes by dividing the trait change over each interval by the pooled standard deviation and then by the number of generations elapsed (5,6). We calculated the rate of adaptation as the parallel change across replicates. To do so we took the average trait change across all 10 replicates and calculated a single rate in Haldanes. Haldanes were calculated for all six phenotypes for each experimental interval and over the whole experiment. We compared the rates of evolution measured in our experiment to those from a meta-analysis of evolutionary rates from field populations that focused on contemporary evolution (less than 200 generations) (7). The meta-analysis was focused on phenotypic change, which includes both genetic and environmental (plastic) effects, as few prior studies used common garden experiments to measure the rate of evolution. 

To test for parallel phenotypic evolution in each of the six phenotypes we carried out separate linear mixed effect models (e.g. lme(phenotype measured ~ time, random=~1|cage/time)) and tested for significance using an anova (nlme and R respectively). For phenotypic responses that were rate calculations (i.e., development, starvation, chill coma recovery, and desiccation tolerance) a GLM was used to find the time at which 50% of individuals completed the assay (i.e., pupated, starved, recovered, desiccated) in each technical replicate. These values were used in the subsequent analysis.

Please see the README file. Additional information is in  SI of the paper.