Models of local adaptation to spatially varying selection predict that maximum rates of evolution are determined by the interaction between increased adaptive potential owing to increased genetic variation, and the cost genetic variation brings by reducing population fitness. We discuss existing and new results from our laboratory assays and field transplants of rainforest Drosophila and UK butterflies along environmental gradients, which try to test these predictions in natural populations. Our data suggest that: (i) local adaptation along ecological gradients is not consistently observed in time and space, especially where biotic and abiotic interactions affect both gradient steepness and genetic variation in fitness; (ii) genetic variation in fitness observed in the laboratory is only sometimes visible to selection in the field, suggesting that demographic costs can remain high without increasing adaptive potential; and (iii) antagonistic interactions between species reduce local productivity, especially at ecological margins. Such antagonistic interactions steepen gradients and may increase the cost of adaptation by increasing its dimensionality. However, where biotic interactions do evolve, rapid range expansion can follow. Future research should test how the environmental sensitivity of genotypes determines their ecological exposure, and its effects on genetic variation in fitness, to predict the probability of evolutionary rescue at ecological margins. This article is part of the theme issue ‘Species’ ranges in the face of changing environments (Part II)’.

Crossing design

In 2011, we collected female D. birchii from sites along three elevation gradients in north-east Queensland (Mt Lewis, Mt Edith and Paluma) and established isofemale lines, as described previously (Bridle et al. 2009; O'Brien et al. 2017). We established 7 – 10 lines per site, with 10, 7 and 6 sites respectively at Mt Lewis, Mt Edith and Paluma.

We used a series of fully reciprocal diallel crosses to estimate additive genetic (co)variances of quantitative traits of D. birchii, and assess how this changes along elevation gradients. Crosses were between isofemale lines from within each site. For each site, crosses were set up between all possible line combinations, including reciprocals and within-line crosses (full diallel design). The total number of unique crossing combinations at each site was therefore n², where n = the number of isofemale lines from that site. Each crossing combination was replicated 2 - 4 times. While the number of lines in each diallel varied slightly between gradients, where possible it was kept constant between sites within a gradient, to avoid variation in the number of crosses affecting genetic variance estimates along gradients. 

In the two generations prior to establishing crosses, isofemale lines were reared at low density by limiting the number and laying time of flies laying in vials. For each of five vials per isofemale line, five mated females were placed on 10ml standard potato media, supplemented with live yeast, and left to oviposit for 72 hrs. This typically resulted in 30 - 50 offspring per vial (Paluma mean = 34.8, Danbulla mean = 53.0, Mt Lewis mean = 45.3), which is well below the maximum larval density that can be supported by this volume of media. Eclosing offspring were sexed over three days under CO₂ anaesthesia within 24 hours of emergence to ensure they were unmated. Sexes were then held separately in food vials at densities of ≤ 10 flies per vial for 5 – 7 days to allow them to recover from effects of CO₂, and to ensure they were reproductively mature. Crosses were established by placing a single virgin female in a vial with a single virgin male. Mating pairs were left in food vials to mate and lay for 72 hrs. Offspring were sexed upon eclosion (± 12 hrs) over a four day period to obtain flies for use in trait assays. 

Measurement of quantitative traits

Flies emerging from each cross were screened for cold tolerance and wing size. Wing size was measured on male offspring, and cold tolerance was measured on female offspring. All offspring were virgin prior to screening. Approximately 3 - 4 offspring from each cross were screened for each trait, or as many flies as were available if the cross produced fewer than this. Each fly was only screened for a single trait. Flies were held at 19 °C on a 12 hr:12 hr light:dark cycle until they were assayed.

Cold tolerance

Cold tolerance was assayed on 13-day-old females as the productivity of a female following a cold shock, since reproduction is often more sensitive than viability to environmental stress (Hoffmann 2010). Flies were placed individually in 40 ml vials containing 10 ml of standard potato food media and cold-shocked for 2 hours at 0 °C, which put all flies into a cold-induced coma. After cold shock, flies were left to recover at 25 °C for one hour (long enough for the vast majority of flies to emerge from chill coma). A single virgin male from a mass bred stock population was then added to each vial. The flies were left to mate and lay for 48 hrs, and then removed. The total number of flies emerging from each vial was counted as a measure of productivity. Any females that did not produce any offspring were excluded from analysis of productivity because reproductive failure could occur for reasons other than cold stress.

Wing size

To measure wing size, the right wing of male flies was removed and mounted on a microscope slide. Wings were photographed at 40 X magnification using a digital camera attached to a Nikon microscope. Wing images were landmarked using tpsDig v 2.16 (F.J. Rohlf, http://life.bio.sunysb.edu/morph/) at the same 10 landmarks used by Griffiths et al. (2005). Wing size was measured as the wing centroid size (square root of the squared distance between each landmark and the centroid), obtained using MorphoJ (Klingenberg 2011). This is given in mm.

A detailed description of the contents of each data file is included in the README.txt file.