Field tests reveal genetic variation for performance at low temperatures in Drosophila melanogaster


  • Johannes Overgaard,

    1. Zoophysiology, Department of Biological Sciences, Aarhus University, Aarhus, Denmark
    Search for more papers by this author
  • Jesper Givskov Sørensen,

    1. Ecology and Genetics, Department of Biological Sciences, Aarhus University, Aarhus, Denmark
    2. National Environmental Research Institute, Department of Terrestrial Ecology, Aarhus University, Silkeborg, Denmark
    Search for more papers by this author
  • Louise Toft Jensen,

    1. Ecology and Genetics, Department of Biological Sciences, Aarhus University, Aarhus, Denmark
    Search for more papers by this author
  • Volker Loeschcke,

    1. Ecology and Genetics, Department of Biological Sciences, Aarhus University, Aarhus, Denmark
    Search for more papers by this author
  • Torsten Nygaard Kristensen

    Corresponding author
    1. Ecology and Genetics, Department of Biological Sciences, Aarhus University, Aarhus, Denmark
    2. Department of Genetics and Biotechnology, Aarhus University, Tjele, Denmark
    Search for more papers by this author

*Corresponding author. E-mail:


  • 1. Physiological and evolutionary responses underlying thermal adaptation and acclimation are often investigated under controlled laboratory conditions. Such studies may fail to assess ecologically relevant parameters as they do not account for the complexity of the natural environment.

  • 2. We investigated a population of Drosophila melanogaster for performance at low temperature conditions in the field using release recapture assays and in the laboratory using standard cold resistance assays. The aim of the study was to get a better understanding of the nature and underlying mechanisms of the trait measured in field recapture studies and the association between field performance and fitness measures estimated in the laboratory.

  • 3. We performed one generation of selection on the ability to reach a resource at low temperature under field conditions. Flies that reached a resource (‘mobile’) and those that never left the release site (‘stationary’) were reared to the F1 and F2 generation in the laboratory. Subsequent field releases with these flies demonstrated a clear genetic differentiation between mobile and stationary flies in their ability to reach resources at low temperatures in the field. This indicates that mobility at low temperature is under additive genetic influence. In contrast mobile and stationary flies were generally indistinguishable when tested in standard laboratory tests of cold performance. The genetic differentiation between the two sub-populations was not linked to allelic variation in known candidate genes for cold adaptation. However, using transcriptomics we identified new candidate genes (transcripts) and pathways that differed between the mobile and stationary flies.

  • 4. The current study reveals an irregular relationship between cold performance in the field and in the laboratory. Based on these results, we suggest that the ecological relevance of laboratory assays should be evaluated more critically in studies of thermal adaptation and hardening/acclimation.


Ectothermic animals must cope with a changing thermal environment on both daily and seasonal time-scales and numerous studies have investigated the importance of both acclimatory and evolutionary responses for fitness under different thermal conditions (Chown & Terblanche 2007). In the case of terrestrial insects, laboratory reared fruitflies (Drosophila melanogaster) have often been the model animal of choice in studies of thermal performance under laboratory conditions (Hoffmann, Sørensen, & Loeschcke 2003). Results obtained from such studies are often interpreted in the context of the environmental conditions that the organism is expected to meet in the field. However, laboratory studies may fail to consider the true complexity of environmental stimuli that ectotherms face under natural conditions and the ecological relevance of results obtained in the laboratory has therefore been questioned (Gibbs 1999; Harshman & Hoffmann 2000). An alternative approach to studying thermal adaptation and acclimation is to investigate wild caught insects under controlled laboratory conditions (Deere & Chown 2006; Terblanche et al. 2007). Although this approach provides insight into the ‘real’ thermal performance of insects, it offers limited ability to separate the effects of developmental acclimation, adult acclimation and genetic variation. Clearly common garden approaches and/or tests of model animals under field conditions enable a better separation of these phenomena (e.g. Hoffmann, Shirriffs, & Scott 2005; Kristensen et al. 2008a; Ragland & Kingsolver 2008).

A number of studies have investigated laboratory reared flies under natural or semi-natural conditions. Release–recapture experiments have been used to investigate components of fitness in nature (Jaenike, Benway, & Stevens 1995; Kingsolver 1999; Loeschcke & Hoffmann 2007; Kristensen et al. 2008a) and field cage studies have revealed the importance of acclimation in the field (Mitrovski & Hoffmann 2001; Hoffmann et al. 2003; Overgaard & Sørensen 2008). In the release–recapture assays the success in finding resources is often used as a proxy of fitness for D. melanogaster because all active life stages of this species feed and breed on ephemeral rotting fruits. One conclusion drawn from these studies is that laboratory results can be misleading as a predictor of field performance (Kingsolver 1999; Kristensen et al. 2008a). For example, laboratory studies may underestimate the costs and benefits associated with heat and cold hardening (Thomson, Robinson, & Hoffmann 2001; Loeschcke & Hoffmann 2007; Kristensen et al. 2008a). Therefore results from field experiments on plastic responses are an important supplement to traditional laboratory investigations.

Until now release–recapture studies have mainly been concerned with how laboratory reared flies perform in the field and no studies have yet investigated how recaptured flies perform in laboratory assays of temperature tolerance. Yet, in order to use the results from field–recapture studies to answer important questions in ecology, physiology and evolutionary biology we need a better understanding of the nature and underlying mechanisms of the trait measured in field recapture studies and the association between field performance and fitness measures estimated in the laboratory.

To investigate these questions we released a large number of flies from a mass population in the field under cold conditions and recaptured them at capture points (buckets with banana). After recapture, we tested flies that reached the resource (‘mobile flies’) against those that never left the release point (‘stationary flies’) in a number of laboratory assays. In addition, we established F1 and F2 generations founded by flies from these two groups and subsequently tested the F1 and F2 flies under field and laboratory conditions. Finally, we investigated whole genome gene expression patterns and single nucleotide polymorphism (SNP) variation in mobile and stationary flies to search for candidate genes and/or transcripts potentially involved in the observed differentiation in the ability to disperse and locate resources under cold conditions in the field.

Materials and methods

Origin of experimental flies

The D. melanogaster population used in this experiment originated from a non-selected mass bred population that had been kept in high numbers (N > 1000) for several years in the laboratory (see Bubliy & Loeschcke 2005 for details). Flies were kept at 20 °C in a climate room (20 ± 1 °C) under non-crowded conditions on agar-yeast- sugar-oatmeal medium at 12h:12h light:dark cycles and 50% relative humidity. To ensure that only virgin flies were used in the experiment (see later for the reason for using virgin flies only), flies were anaesthetised using short exposure to CO2 < 15 h after emergence and separated into females and males kept at low density. These flies were left for 2 days at 20 °C so that they were sexually mature at the onset of the experiment.

Field release and separation of stationary and mobile flies in the F0 generation

The aim of the release–recapture study was to compare flies that were able to locate and reach resources in the field (mobile flies) with those that did not leave the position at which they were released and where resources were not available (stationary flies). In order to separate these groups, approximately 10·000 2–3-day old virgin female flies were released on the morning of September 24, 2007 in the park surrounding Aarhus University, Denmark. Wild living D. melanogaster are not present in this area at this time of the year. Virgin flies were used to increase our selection differential. The habitat consisted of grass and bushes. Temperatures in the field were recorded with data loggers (Tinytalk II, Chichester, UK) positioned in the shade approximately 1 m above ground. At the time of release (08:30), air temperature was approximately 15 °C ensuring minimal activity of the flies. Five meters from the release point three buckets, 2 m apart, were placed on the east and west side of the release point. The temperature increased gradually during the day with an average of 17·7 °C and a maximum of 22·5 °C. Flies were caught at the resource stations every 30 min and stored in empty glass vials near the release site so that both experimental groups experienced similar environmental conditions. In the afternoon (15:45) when approximately 460 female flies were caught at the resource stations (mobile flies) the collection was ended by catching approximately the same number of female flies at the release site (stationary flies).

The experiment was repeated for male flies the following day using the exact same experimental approach. On this day the temperature was approximately 13 °C at the onset of release and the average and maximum temperatures were 14·8 °C and 17·5 °C, respectively. We caught a low number of male flies at the resource stations (probably due to low temperatures). The experiment was terminated at 14:30 after 35 male flies were caught at the resource stations and a similar number were then collected from the release point. These flies were only used to establish the F1 generation and no further experiments were conducted with F0 male flies.

Laboratory tests of cold temperature performance in mobile and stationary flies

The mobile and stationary groups of females caught were randomly split into three sub-groups. One sub-group (210 flies) was tested in various cold performance assays immediately after collection (day 0) (see details later). A second sub-group (210 flies) was placed in food vials at 20 °C and tested in the same cold performance assays 2 days after collection (day 2), and finally a third sub-group (∼ 40 flies) was used to establish a F1 generation of mobile and stationary flies by mating them with the corresponding group (mobile or stationary) of field caught males. The F1 generation was raised under controlled density (50 eggs per food vial) to avoid confounding factors of variable levels of crowding. Egg-to-adult viability and developmental time were recorded during development of the F1 generation. The emerging F1 flies were tested for cold performance in the laboratory similarly to the F0 flies (day 0 and 2) and the F1 generation was also tested in field release–recapture experiments (see below). In addition, 50 randomly chosen pairs of F1 flies from each group (mobile/stationary) were used to establish a F2 generation that was also tested in field release–recapture experiments.

Cold tolerance of mobile and stationary female flies was tested on F0 flies at day 0 and day 2 after field capture and F1 flies were tested when they were 2–3 days old. Cold tolerance was tested using assays of (i) chill coma recovery (ii) egg production at 16 °C (iii) spontaneous locomotor activity and (iv) stimulated activity in a maze.

(i) Chill coma recovery

Forty flies from each group were placed individually in Eppendorf tubes and exposed to −3 °C for 2 h. After this, flies were quickly returned to room temperature where they were placed in a position that made it easy to score recovery time for each individual as the time it takes for flies to be able to stand after the cold exposure without disturbing flies.

(ii) Egg production at 16 °C

Forty virgin female flies from each group were mated with 40 ‘neutral’ male flies (males from the original mass bred population). These pairs were placed in vials at 16 °C and allowed to mate and deposit eggs for 48 h, with vials replaced every 24 h. After replacement, the vials were returned to 20 °C for development. Productivity was scored as the number of emerging adults.

(iii) Spontaneous locomotor activity at 16 °C

Forty flies from each group were placed in small glass tubes (5 mm diameter) containing standard fly media. These tubes were placed horizontally in a TriKinetics Drosophila Activity Monitoring System (TriKinetics Inc, Waltham, MA, USA) that continuously monitors activity. The monitors were placed at 16 °C and flies were allowed to settle at least 1 h before the recording started. The system counts the number of times each individual fly passes an infrared light beam. To avoid erroneous locomotor activity measurements from flies standing still directly under the beam, we scored the presence/absence of activity every minute. Activity was always measured in the time frame from 18:00 on the first day until 15:00 on the following day.

(iv) Stimulated activity at low temperature

Stimulated activity was tested by releasing the flies in a horizontally placed maze, similar to a Hirsch T-maze (Hirsch 1959). Flies were released in one end of the maze and vials with food were attached to the opposite end. In order for flies to reach the food resource, they must walk approximately 60 cm in the maze, which consists of a large number of crossings pointing towards or away from the food. Forty flies from each group (mobile/stationary) were tested simultaneously in the maze and these were distinguished by different colours of fluorescent micronized dust applied immediately prior to the test. Flies were collected from the food vials every 30 min during day time and every third hour during night time and scored for dust colour under UV light. The maze-test was started at 17:00 where the maze initially was placed at 16 °C. Since no activity was recorded during the first 8 h of the first test, the maze was moved to a more moderate temperature of 20 °C at 01:00 the following morning. The experiment continued at this temperature until the test was terminated after 39·5 hours at 08:30 the following morning. This procedure was subsequently used for all three treatment groups (F0 day 0, F0 day 2, F1). Less than 5% of the flies died during the tests and these were discarded from the dataset.

Field test of mobile and stationary flies from the F1 and F2 generation

Two releases were undertaken using F1 female flies and one release with each sex was undertaken using F2 flies. Capture points in all releases were 5 m apart starting 5 m from the release site and extending up to 20 m in opposite directions from the release point. The releases were undertaken on a shady day at cool temperatures. Temperatures were continuously recorded with data loggers (Tinytalk II, Chichester, UK) (Fig. 2). Prior to the release, groups of 200 flies were transferred into vials with 1·5 ± 0·5 mg of fluorescent micronized dust. Dust colors were randomly assigned to the two groups (mobile/stationary), and changed between releases. Flies were released in the morning between 09·00 and 10·00 and they were captured from resources (buckets with mashed banana) by netting and/or aspirating, starting 1 h after the release and repeatedly every hour until 17·00. Captured flies were immediately placed on ice to minimize transfer of dust color between flies. They were then transported to the laboratory where the colors were scored under UV light. For further detail see Fig. 2 and Kristensen, Loeschcke, & Hoffmann (2007).

Figure 2.

 Number of captured flies from the mobile and stationary Drosophila melanogaster groups in field releases. Two releases with female flies were performed in the F1 generation (a and b), and one release was performed for each sex with F2 flies (c (females) and d (males)). Details on the number of released flies and overall recapture success are given in the individual headings and the inserted figures show the field temperature at the individual release sites. The proportion of mobile vs. stationary flies caught is given in each figure together with the P values from logit models and χ2-tests, respectively.

Morphological measurements, metabolic rate and longevity

Several morphological measurements were made on female flies from mobile/stationary groups of flies from the F0, F1 and F2 generation. For each experimental group (F0 and F1) we measured approximately 30 individual female flies with respect to wing size, thorax length, dry mass and lipid content. All traits were measured on the same flies, and these flies had previously been used for measurement of spontaneous activity. For measures on fly wings we fixed the right wing to a microscope slide and photographed it with a digital camera attached to a dissecting microscope (IM1000 version 1·2.). We measured length of longitudinal vein 2 and width (distance between the intersection of the wing edge and longitudinal vein 2 and 5; see Kjærsgaard et al. 2007). Distances were calculated from landmarks by the software package ImageJ version 1·33u (Rasband 2001). Thorax length was measured from the anterior margin to the posterior tip of the scutellum using a dissecting microscope fitted with a digital filar eyepiece (Los Angeles Scientific Instrument Company, Inc., CA, USA). Finally, the wingless flies were dried overnight at 60 °C for measurements of dry body mass. Each individual fly was subsequently washed in 1 ml chloroform overnight to remove all lipids. The chloroform was removed and the flies dried for 24 h at 60 °C before being reweighed. The difference in dry weight before and after the chloroform wash was used as an estimate of total lipid content.

Longevity and standard metabolic rate was measured on F1 flies only. Oxygen consumption rates were measured using closed-system respirometry. Six replicates of 20 female flies from each group were placed in a closed 20 ml syringe containing room air and after 4 h at 16 °C a 15 ml gas sample was injected into a O2 (S-3A/I) analyser (Applied Electrochemistry, Sunnyvale, CA, USA) connected to an AcqKnowledge (version 3·7·1) MP100 data-acquisition system running at 100 Hz. The injected sample was then sucked through the oxygen analyser by an air-pump ensuring constant flow so that the amount of oxygen consumed could be calculated by integrating the resulting decrease in O2. Metabolic rate is reported relative to wet mass and reported as STPD (Standard Pressure, Standard Temperature, Dry air).

Longevity estimates were based on 10 replicates of 10 female flies for each group. These flies were placed at 16 °C and survival was checked every fourth day until all flies were dead.

Genotyping of candidate genes for cold resistance

DNA from F0 female flies was extracted by the CTAB method (Doyle & Doyle 1987) and alleles were scored for the following genes: hsr omega, adh, pgi, frost, dca and hsp26. These genes have all been associated with cold tolerance of D. melanogaster and other insects with the exception of pgi which has been associated with variable flight ability at low temperature in butterflies (see Table S1 in Supporting information). hsr omegaS/L alleles were scored as in Anderson et al. (2003), adhF/S was scored as in Umina et al. (2005) and hsp26 alleles as in Frydenberg, Hoffmann, & Loeschcke (2003). The pgi SNP polymorphism described in Sezgin et al. (2004) was scored by sequencing using the primers F: GGCCATTGGACTGTCAATCT and R: GCAATGGGATTGTGTGTCTG. Variation in frost and dca promoters were scored as outlined in Rako et al. (2007), while another polymorphism in the frost gene was investigated using primers F: TGGTCATCATGGCAACAATC and R: ATCCTCGGTGGTCAACTCAG, amplifying a variable proline rich region in the coding region of the gene (Mark Blacket, personal communication). All candidate genes were scored in each of 48 flies per mobile/stationary group.

Transcriptional analysis of mobile and stationary flies

RNA was purified from 2 × 20 F0 virgin females (mobile/stationary, 20 flies for each probe array) that were maintained at 20 °C for 2 days following the field release–recapture experiment. Isolation and labelling of RNA and microarray processing were performed as described in Kristensen et al. (2005). In short, total RNA from whole flies was isolated using Trizol Reagent (Invitrogen). A total of 5 μg was labelled using the superscript Choice system (Life Technologies). Biotin labelled cRNA was prepared using the BioArray High Yield RNA Transcript Labelling Kit (ENZO). A total of 15 μg of cRNA was loaded onto the Affymetrix probe array cartridge (Drosophila Genome Array Version 2). Hybridization mixture from each of the two groups of flies was loaded to two arrays yielding a total of four probe arrays for the experiment.


Statistical analyses were performed using Sigmastat version 3·5 and SPSS version 15·0. Values from all laboratory tests (except transcriptional data) are reported as average ± SEM and an effect was considered significant at the 0·05 level. A two-way anova was used to test for effects of sex and group on developmental time (square root transformed). Due to occasional lack of normality we tested the effect of group (mobile/stationary) on morphological traits (bodymass, lipid content, thorax length and wing size), chill coma recovery time and spontaneous activity using individual Mann–Whitney rank sum tests. A similar test was used to separate groups with respect to longevity, productivity, egg–to-adult viability and metabolic rate at 16 °C.

Differences in stimulated activity in the maze between the two groups were tested using a survival LogRank test. Here flies that passed the maze were scored as ‘dead’ and the time at which they were scored was noted. Flies that failed to pass the maze was scored as ‘alive’ at the end of the test (time 39·5 h after onset). This test separates the groups in terms of how quickly they pass the maze as well as the proportion of flies that pass through the maze.

In the field releases of F1 and F2 flies, we tested whether the total number of flies caught differed between mobile and stationary flies. Capture data was treated as categorical and analyzed separately for each release. The difference between the two groups in capture success was assessed using logit models, assuming that each fly represents an independent data point for treatment and sex. We also undertook a simpler analysis where capture rates were compared directly between the groups and tested using the chi-square statistic against an expectation of equal numbers being captured from each group. To indicate the relative capture success of flies from the mobile group relative to flies from the stationary group, we computed relative capture rates comparing the likelihood of capture of flies from the mobile group vs. flies from the stationary group in each of the four releases. Here, a relative capture success above 1 indicates that more flies from the mobile group were caught than flies from the stationary group.

For the examined candidate genes, each gene locus was tested using Fstat (Goudet 2002) where a FST-value significantly different from 0 between the two groups implies a difference in allele frequency. Transcriptomics data were pre-processed using the Robust Multi-array Analysis algorithm (Irizarry et al. 2003; Wu et al. 2004). To exclude genes that could not be confidently detected in the data analysis, probe sets with less than one present call across the four chips were excluded. The filtered gene set contained 8884 transcripts. Differential expression of individual transcripts was assessed using significance analysis of microarrays (SAM) proposed by Tusher, Tibshirani, & Chu (2001). An overall test of significance for single genes was performed using moderated F-statistics in a multiclass analysis. For each gene the two groups (mobile/stationary) were tested for differentially expressed gene transcripts. The moderated F-statistic tests whether any of the contrasts are non-zero for that gene among the two groups. Multiple testing was accounted for by controlling the false discovery rate at 10% in the single gene analysis (Benjamini & Hochberg 1995). To investigate pathways differentially expressed in the two groups, we detected single genes using the same analysis as above but now using a less conservative false discovery rate at 20%. Differential expression of functional groups of genes between the mobile and stationary flies was tested on the basis of a modified t-statistics using the two-class unpaired analysis. The SAM analysis was performed as implemented in the R package siggenes (Gentleman et al. 2004). Groups of genes being differentially expressed were assigned to functional categories using the KEGG (Kyoto Encyclopedia of Genes and Genomes) pathway information (Kanehisa & Goto 2000) downloaded from the David database (URL:


Laboratory tests of performance at low temperature in mobile and stationary flies

There was no significant difference between the mobile/stationary groups in neither the F0 (0 and 2 days) nor the F1 generations with respect to chill coma recovery and productivity. However, there was a significantly higher spontaneous activity at 16 °C of the mobile flies when tested immediately after recapture but this was not apparent in either the 2 day test of F0 flies or in the F1 generation (Table 1). There was a clear effect of group on stimulated activity in the F0 generation with mobile flies being considerably more active than the stationary flies at both day 0 and day 2 (Fig. 1a,b; P < 0·001). Although we observed a trend in the same direction, the differences between groups were no longer significant when tested in the F1 generation (Fig. 1C; P = 0·27).

Table 1.   Cold tolerance of mobile and stationary Drosophila melanogaster tested acutely (F0 day 0), after 2 days (F0 day 2) and after one generation (F1)
  F0 day 0F0 day 2F1
  1. All values are average ± SEM and N-values are listed in parentheses. An asterisk and numbers in bold indicate groups that differ significantly using Mann–Whitney rank sum tests.

Chill coma recoverySeconds402·4 ± 7·7 (39)393·7 ± 8·0 (39)361·2 ± 7·5 (39)345·6 ± 7·5 (40)321·6 ± 8·0 (39)325·1 ± 7·2 (40)
(U = 842, P = 0·42)(U = 632·5, P = 0·15)(U = 829·5, P = 0·63)
Productivity 16 °CProgeny (pair/48 h)13·6 ± 1·4 (37)12·2 ± 1·1 (40)17·4 ± 1·8 (39)16·4 ± 1·7 (39)7·2 ± 1·6 (38)4·8 ± 1·1 (38)
(U = 644·5, P = 0·33)(U = 701, P = 0·55)(U = 629, P = 0·30)
Spontaneous activity, 16 °Cno of active periods/21h81·1 ± 3·7* (40)50·9 ± 5·5 (40)88·1 ± 7·7 (40)75·7 ± 1·8 (40)77·5 ± 7·5 (40)85·2 ± 8·9 (39)
(U = 523, P = 0·008)(U = 711, P = 0·39)(U = 710·5, P = 0·50)
Figure 1.

 Stimulated activity of Drosophila melanogaster at low temperature. The curves show the cumulative success of completing the walk through a maze for mobile and stationary flies. The test was performed immediately after separation of the two groups (F0 day 0) (a); after the two groups had been kept for 2 days at 20 °C (F0 day 2) (b); and on the offspring (F1) from mobile and stationary flies (c). See text for further explanation.

Longevity, metabolic rate, developmental time and egg-to-adult viability

There was no significant effect of group on longevity or metabolic rate (Table 2). Egg-to-adult viability was also similar in the two groups of emerging F1 flies. However, developmental time was faster for both sexes in mobile compared to stationary flies and as expected, female flies developed slightly faster than male flies (Table 2).

Table 2.   Developmental time, egg-to-adult viability, oxygen consumption rate and longevity of the F1 generation of mobile and stationary Drosophila melanogaster kept under standard laboratory conditions at 20 °C
  1. All values are average ± SEM and N-values are listed in parentheses (N-values for oxygen consumption are number of samples with 20 individuals). An asterisk and numbers in bold indicate groups that differ significantly.

Egg-to-adult viability%93 ± 1 (34) 89 ± 2 (34)
(U = 707, P = 0·11)
(Mann–Whitney test for effect of group)
LongevityDays155·7 ± 3·1 (95) 149·0 ± 3·6 (88)
(U = 4765·5, P = 0·10)
(Mann–Whitney test for effect of group)
Oxygen consumption at 16 ºCμL O2/h/mg ww (STPD)1·46 ± 0·05 (6)1·41 ± 0·03 (6)
(U = 14, P = 0·59)
(Mann–Whitney test for effect of group)
Developmental timeHours324·7 ± 0·7 (34)332·6 ± 0·7 (34)326·7 ± 0·9 (34)334·6 ± 0·8 (34)Two-way anova
Sex: (F1,132 = 6·72; P = 0·011)
Group: (F1,132 = 108·2; P < 0·001)
Interaction: (F1,132 = 0·049; P = 0·83)

Field releases of F1 and F2 generation flies

The percentages of flies captured in the four releases varied between 0·8% and 4·9% (Fig. 2). In all releases, flies caught at the resources in the two directions from the release point, were combined. The relative capture success was more than 2·5 times higher for flies from the mobile group relative to flies from the stationary group in the two F1 releases (Fig. 2). The difference between the numbers of flies caught from the two groups was highly significant in the first but marginally non-significant in the second F1 release. In the release of F2 flies the relative capture success was 83% and 100% higher for female and male flies from the mobile group relative to female and male flies from the stationary group (Fig. 2). Also in the F2 releases the number of flies caught only differed significantly in one of the two releases.

Morphological measurements

There were no consistent morphological differences across generations between mobile and stationary flies. Mobile flies had larger wings in the F0 generation while wings and thorax length of flies from this group were smaller in the F2 generation. There was a significantly higher lipid content of mobile flies in the F1 generation, but this was not seen in F0 or F2 generations (see Table S2 in Supporting information).

Genotyping of candidate genes for cold resistance

We failed to detect any significant differences in allele frequencies in the studied candidate genes between the groups (see Table S1 in Supporting information). frost alleles in both the promoter and coding region showed no polymorphism at all, and hsr omegaL and adhS alleles were almost completely fixed in both groups (data not shown).

Transcriptional differences between mobile and stationary flies

Using a false discovery rate of 10% 13 genes were differentially expressed between the mobile and stationary flies from the F0 generation. These were all down-regulated in the mobile flies (Table 3). In order to investigate patterns of transcriptional differences at higher levels of organisation, we employed a less conservative false discovery rate of 20% and tested for functional groups of genes being differently expressed between the two groups of flies using the KEGG pathway information. Using this less conservative false discovery rate criterion 779 and 221 genes were significantly up- and down-regulated when contrasting mobile and stationary flies. Subsequent analysis identified four pathways to be characterised by disproportionate high numbers of up-regulated genes in the mobile flies. Two of these pathways are associated with inositol phosphate metabolism, one with DNA polymerase and one with ubiquitin mediated proteolysis (see Table S3 in Supporting information).

Table 3.   Differentially expressed genes identified using affymetrix gene chips
Gene symbolAffymetrix IDInferred molecular function and/or biological process (FlyBase)Chromosome location
  1. Gene transcription data are from F0 flies kept under standard laboratory conditions (20 °C) for 2 days after recapture. All genes have unknown function in D. melanogaster but in some cases function can be inferred from homology to other genes. All genes are down-regulated in mobile relative to stationary flies.

CG312171633893_atSerine-type endopeptidase activity (proteolysis)3R
CG130951635453_atAspartic-type endopeptidase activity; pepsin A activity (proteolysis)2L
CG21911633425_atSodium-dependent multivitamin transmembrane transporter activity (membrane transport)3R
CG59211633783_a_atProtein bindingX
CG155311638528_atStearoyl-CoA 9-desaturase activity; iron ion binding (fatty acid biosynthetic process)3R
CG66601633308_atUnknown (membrane associated)3R


One primary aim in this study was to explore the relationship between performance of D. melanogaster at low temperatures in field and laboratory tests. A wide range of laboratory tests are used when investigating the genetic or physiological background for thermal adaptation in insects, and the results are often assumed to correlate with field fitness. Thus, cold shock survival, or chill coma recovery are used as proxies for cold tolerance, and conversely, heat shock survival and heat knock down temperature are used to assay heat tolerance (Hoffmann, Sørensen, & Loeschcke 2003; Chown & Terblanche 2007). However, such simple assays of thermal performance cannot fully predict the fitness consequences of thermal stress or the selective pressure on adaptive mechanisms enabling organisms to cope with thermal stress. In recent years, a number of studies have used release-recapture or field cage experiments to assay components of fitness in the field (Jaenike, Benway, & Stevens 1995; Kingsolver 1999;Mitrovski & Hoffmann 2001; Hoffmann et al. 2003; Thomson, Robinson, & Hoffmann 2001; Loeschcke & Hoffmann 2007; Kristensen et al. 2008a, b). In release–recapture experiments D. melanogaster is often used to test the ability to locate resources under different thermal conditions. The underlying logic is that larvae and adults of D. melanogaster feed and breed in necrotic fruits, implying that the fitness of the individual is closely linked to the ability to find resources. In previous studies on D. melanogaster it has been demonstrated that laboratory based acclimation improves re–capture rates in the field indicating that acclimation is important under field conditions (Loeschcke & Hoffmann 2007; Kristensen et al. 2008a). However, some of these studies also indicate that laboratory and field results may not be equivalent. Kristensen et al. (2008a) demonstrated that laboratory cold acclimation during development improved both laboratory and field performance at low temperatures. However, costs of cold acclimation at high temperature were not seen in laboratory tests while there was a clear cost of cold acclimation at high temperatures in the field. This indicates that the complex nature of the field release test detects subtle differences in thermal performance not evident in simple laboratory tests. Our release–recapture experiments show that both F1 and F2 generations of mobile flies are better able to locate a food source at low temperature in the field. In contrast, nearly none of the laboratory tests revealed consistent differences between stationary and mobile flies. Thus, there seems to be no straightforward relationship between field and lab cold performance.

Field performance in the release-recapture assay relies on both the perception and integration of external stimuli and on the ability to perform the basic physiological functions that underlie movement and orientation. Simple laboratory tests are unlikely to uncover such complexity or may score irrelevant traits. In this context, it is interesting to note that the laboratory assay that showed the greatest difference in performance between mobile and stationary flies was the stimulated activity measurements in the maze (Fig. 1). In this relatively complex laboratory test we saw that mobile flies were significantly more active in the two tests performed in the F0 generation. The difference was not significant in the F1 generation, although mobile flies still tended to be more active (Fig. 1). Two additional tests were performed in the F1 generation, and both of these also showed a non-significant tendency towards higher activity of the mobile flies (see Fig. 1S in Supporting information). Developmental time also differed significantly between stationary and mobile groups of flies in the F1 generation, with mobile flies developing slightly faster (Table 2). Although the biological significance of this result is hard to judge, individuals with faster development are generally considered to have higher fitness (Lewontin 1965).

An alternative explanation for the observed differences between results obtained in the laboratory and field tests may be that the ‘mobile’ flies identified in the field test may have had better locomotor performance, rather than improved cold performance per se. However, there are a number of reasons why we do not interpret the results in that way. Firstly, the thermal conditions in our release experiments are close to those limiting flight performance in drosophila (Frazier et al. 2008). Secondly, a previous study selected lines for heat and cold stress resistance using classical laboratory traits during the selection procedure (Bubliy & Loeschcke 2005). When these flies were released under field conditions similar to those used in the present study we did observe a positive effect of laboratory selection for heat and cold resistance, respectively (Kristensen, Loeschcke, & Hoffmann 2007). This suggests that our results should not be attributed alone to a temperature independent difference in locomotor activity of the mobile flies, but that differences are linked to performance at low temperature. Thirdly, when F1 mobile and stationary flies were tested at high temperature in the maze, we found no difference in locomotor activity between the two treatment groups (Fig. S1 in Supporting information). Finally, we saw no difference in spontaneous activity in F1 and F2 generations while the difference in cold performance in the field persisted. Although the present study does not determine a direct causal relationship between cold tolerance and locomotor performance our observations indicate that the traits selected for in the present study is related more to cold performance than to locomotor activity per se. Consequently we argue that the mobile flies in the present study have higher field fitness under low temperature field conditions.

Most traits associated with thermal resistance investigated in the laboratory are highly heritable in D. melanogaster (Hoffmann, Sørensen, & Loeschcke 2003; but see Blows & Hoffmann 2005) and the present study indicates that this is also the case for the ability to locate resources at low temperatures in the field as also suggested by Hoffmann & O’Donnell (1922). The likelihood of catching flies from the mobile group relative to flies from the stationary group at the resources was higher in both the F1 and F2 generation. Although we cannot quantify the heritability or the additive genetic variance based on data from this study, the observed differentiation of the two groups after just one generation of selection indicates that there is significant additive genetic variation for this trait. Moreover, it seems unlikely that maternal effects can explain the apparent differentiation between mobile and stationary groups of flies since the mobile flies were just as superior relative to the stationary flies in the F2 generation as they were in the F1 generation when tested in the field.

We found no significant allelic differentiation between mobile and stationary flies when screening loci that have previously been associated with cold tolerance or flight ability under cold conditions. Although this indicates that these genes are unlikely to play a role in the genetic differentiation observed in this study it does not exclude an importance of these candidate genes with respect to other aspects of cold tolerance in D. melanogaster. Moreover, it is possible that one generation of selection is insufficient to reveal allelic differentiation. A potentially interesting polymorphism to investigate in relation to activity at low temperature in the future could also be the ‘rover/sitter’ polymorphism described to be important for foraging behaviour in D. melanogaster (Pereira & Sokolowski 1993).

The significant transcriptional differentiation found between the mobile and stationary flies does not directly imply a genetic differentiation but it does point out putative genes and pathways of interest. Moreover, it is important to stress that transcriptomic differences from a single release experiment does not necessarily give a full and comprehensive assessment of all relevant transcriptomic responses in mobile and stationary flies. Unfortunately, the function of all 13 differentially expressed genes detected when using a false discovery rate of 10% is so far unknown making it difficult to functionally interpret the result (Table 3). However, 6 of the 13 differentially expressed genes identified in this study (CG7916, CG13095, CG6660, CG10513, CG10514, CG6164) were also differentially expressed (and down-regulated) in response to heat stress exposure in D. melanogaster (Sørensen et al. 2005). These genes are all involved in catalytic and/or peptidase activity indicating that some common metabolic pathways may confer thermal tolerance at high and low temperature. Using a higher false discovery rate we identified four up-regulated KEGG pathways (see Table S3 in Supporting information) which could all serve as interesting starting points for future investigations on mechanisms of importance for performance at low temperature in insects. Two of the four pathways that were differentially expressed are involved in inositol phosphate metabolism. Interestingly, genes involved in inositol phosphate metabolism in D. melanogaster and several plant species have been shown to be important for cold and heat resistance partly through their involvement in phototransduction processes suggesting that perception of light may be a temperature stress sensing mechanism (Nielsen et al. 2006; Abreu & Aragao 2007). With respect to the transcriptomic results presented here it should be stressed that we only assessed expression levels of flies exposed to benign conditions (20 °C). It is possible that different transcripts would be differentially expressed between stationary and mobile flies had we examined the flies after exposure to cold temperatures.

In conclusion, the current study reiterates the asymmetrical relationship between field and laboratory performance. Since field conditions are variable in time and with regard to local geographical conditions one should not dismiss the use of standard laboratory tests that are reproducible and controllable. Instead, we propose that ecologically relevant assays should be used more often in studies of thermal adaptation and hardening/acclimation as also suggested by Terblanche et al. (2007). For example we found that stimulated activity using a maze reflected the field test better than other laboratory assays of performance at low temperatures, and it is possible that such tests could be used alternatively. Ideally, hypotheses based on results from laboratory experiments should be tested in field experiments to investigate the ecological relevance of the results obtained in the laboratory. The present study also demonstrates that the ability to locate resources in the field is a heritable trait for which ample genetic variance exists in a laboratory population of D. melanogaster. We did not find allelic variation associated with field performance but identified gene transcripts and metabolic pathways which may be linked to the observed genetic differentiation between mobile and stationary flies. Clearly, targeted studies of mechanisms downstream of these genes/pathways are of great interest for the understanding of thermal adaptation and performance at low temperature.


We are grateful to Doth Andersen for excellent laboratory assistance, Mogens Kruhøffer for running the microarrays, Tobias Wang for help with measurements of metabolic rate, Ary A. Hoffmann for helpful discussions, two anonymous reviewers and Raymond Huey for constructive input to earlier versions of this paper. This work was funded by the Danish Natural Research Council with Steno stipends (JO, TNK and JGS) and centre and frame grants (to VL) and the Danish Veterinary- and Agricultural Research Council (TNK).