Cold adaptation in geographical populations of Drosophila melanogaster: phenotypic plasticity is more important than genetic variability



    1. Laboratoire Populations, Génétique, et Evolution, Centre National de la Recherche Scientifique, 91198 Gif-sur-Yvette Cedex, France
    2. University of Groningen, Evolutionary Genetics Group, 9750 AA Haren, The Netherlands
    Search for more papers by this author
  • V. DEBAT,

    1. Laboratoire Populations, Génétique, et Evolution, Centre National de la Recherche Scientifique, 91198 Gif-sur-Yvette Cedex, France
    Search for more papers by this author
  • P. GIBERT,

    1. Laboratoire Populations, Génétique, et Evolution, Centre National de la Recherche Scientifique, 91198 Gif-sur-Yvette Cedex, France
    2. Laboratoire Biométrie, Biologie Evolutive, Université Lyon I, 69622 Villeurbanne Cedex, France
    Search for more papers by this author
  • A.-G. KISTER,

    1. Laboratoire Populations, Génétique, et Evolution, Centre National de la Recherche Scientifique, 91198 Gif-sur-Yvette Cedex, France
    Search for more papers by this author
  • H. LEGOUT,

    1. Laboratoire Populations, Génétique, et Evolution, Centre National de la Recherche Scientifique, 91198 Gif-sur-Yvette Cedex, France
    Search for more papers by this author

    1. Laboratoire Populations, Génétique, et Evolution, Centre National de la Recherche Scientifique, 91198 Gif-sur-Yvette Cedex, France
    Search for more papers by this author

    1. Laboratoire Populations, Génétique, et Evolution, Centre National de la Recherche Scientifique, 91198 Gif-sur-Yvette Cedex, France
    Search for more papers by this author
  • J. R. DAVID

    Corresponding author
    1. Laboratoire Populations, Génétique, et Evolution, Centre National de la Recherche Scientifique, 91198 Gif-sur-Yvette Cedex, France
      †Author to whom correspondence should be addressed. E-mail:
    Search for more papers by this author

†Author to whom correspondence should be addressed. E-mail:


  • 1According to their geographical distribution, most Drosophila species may be classified as either temperate or tropical, and this pattern is assumed to reflect differences in their thermal adaptation, especially in their cold tolerance. We investigated cold tolerance in a global collection of D. melanogaster by monitoring the time adults take to recover from chill coma after a treatment at 0 °C.
  • 2Flies grown at an intermediate temperature (21 °C) showed a significant linear latitudinal cline: recovery was faster in populations living in colder climates.
  • 3The role of growth temperature was analysed in a subset of tropical and temperate populations. In all cases, recovery time decreased when growth temperature was lowered, and linear reaction norms were observed. This adaptive phenotypic plasticity explained more than 80% of the total variation, while genetic latitudinal differences accounted for less than 4%.
  • 4The beneficial effect observed in adults grown at a low temperature contrasts with other phenotypic effects which, like male sterility, appear as harmful and pathological. Our results point to the difficulty of finding a general interpretation to the diversity of plastic responses that are induced by growth temperature variations.


For ectotherm species, such as insects, temperature has long been recognized as a major environmental factor responsible for species abundance and geographic distribution (Andrewartha & Birch 1954; Precht, Christophersen & Hensel 1955; Cossins & Bowler 1987; Leather, Walters & Bale 1993). Ambient temperature varies according to daytime and season, so that natural populations are often exposed to heat or cold stress (Gibbs, Perkins & Markow 2003). The capacity to adapt to and tolerate such stresses is crucial for the persistence of populations (Hoffmann & Parsons 1991, 1997; Addo-Bediako, Chown & Gaston 2000; Chown, Addo-Bediako & Gaston 2002; Hoffmann, Sorensen & Loeschcke 2003; Klok & Chown 2003). In temperate countries, species must tolerate cold conditions during winter and have developed a diversity of adaptive mechanisms to do so, including the occurrence of diapause and the production of antifreeze compounds (Leather et al. 1993; Graham, Walker & Davies 2000).

Drosophilid species form convenient models with which to investigate ectotherm responses to stress because they are found across a full range of habitats from the Arctic to the equator. Drosophila melanogaster is a cosmopolitan species, and its broad geographical range is accompanied by genetic variation in a diversity of traits, which, in most cases, vary progressively with latitude (David & Capy 1988): this pattern suggests that local climate and especially temperature is the selective factor. Latitudinal clines have been demonstrated for several morphological traits (e.g. Capy, Pla & David 1993) and also for traits directly related to fitness, such as viability, rate of development, larval competitive ability, egg production and tolerance to stresses (heat, cold, desiccation, starvation) (Stanley & Parsons 1981; Boulétreau-Merle et al. 1982; Davidson 1990; James & Partridge 1995, 1998; Guerra et al. 1997; Karan et al. 1998; Hoffmann et al. 2003).

A specific problem in thermal adaptation investigations arises from the fact that most morphological and physiological traits exhibit a large amount of phenotypic plasticity related to growth temperature. For several morphological traits, the response curves are now well described and adaptive variations in the shape of the reaction norms are known across species or geographical populations (see David, Gibert & Moreteau 2004). However, plasticity in fitness traits is less commonly investigated because measurements at the adult stage include variation, which is a consequence of both growth and experimental temperature (Huey et al. 1999).

In the present work, we analyse cold tolerance in D. melanogaster with a recently developed assay (David et al. 1998), that measures the time needed for recovering from chill coma after cold treatment. Recovery from chill coma induced at 0 °C provided a clear-cut discrimination between temperate and tropical species (Gibert et al. 2001). Comparison of geographical populations in nine cosmopolitan species also showed that temperate strains recovered faster than tropical ones, suggesting a genetic adaptation (Gibert et al. 2001). Such adaptive genetic variations were confirmed and latitudinal clines were found in Australian populations of two species, D. melanogaster (Hoffmann, Anderson & Hallas 2002) and D. serrata (Hallas, Schiffer & Hoffmann 2002). Here we investigate a global set of populations of D. melanogaster and demonstrate a latitudinal cline which might exist on all continents. We also analyse the role of growth temperature and show that most adaptive variability arises from phenotypic plasticity. By changing the developmental temperature the adult phenotype varies from cold tolerance (as found in temperate species) to cold sensitivity (as found in purely tropical ones).

Materials and methods

strains investigated

We investigated 24 different strains from around the world, ranging from the equator to 60° of latitude (see Table 1). With a few exceptions (Table 1) all strains had been kept in the laboratory for less than 3 years before the start of experiments. All strains were founded from at least 10 pairs of wild-collected flies, and sometimes much more. They were kept in culture bottles at 20 ± 1 °C and the population size was at least 50 pairs in each generation.

Table 1.  List of the 24 populations of D. melanogaster investigated with the year of collection, their (North or South) latitude of origin, mean recovery time ± standard error. Standard deviations are calculated using the number of independent measures (n = 4·65 ± 0·41 per population)
PopulationsYear of collectionLatitudeMean ± SE (minutes)n
Sao Tomé2000 0°20 N 44·4 ± 4·78
Nairobi (Kenya)2000 1°16 S 49·6 ± 2·94
Mombasa (Kenya)2000 4°02 S 40·8 ± 2·15
Brazzaville (Congo)1999 4°15 N 43·9 ± 8·53
Pointe Noire (Congo)1999 4°46 N 39·4 ± 3·33
French Guyana1996 5°09 N 45·6 ± 4·95
Cotonou (Benin)1998 6°21 N 58·4 ± 1·24
Weita (Australia)199712°4 S 33·9 ± 2·893
Bahia (Brazil)199912°59 S 38·6 ± 3·43
Mananara (Madagascar)200016°10 S 44·1 ± 1·94
Campinas (Brazil)200022°53 S 39·6 ± 3·74
Rio (Brazil)200122°54 S 44·7 ± 2·64
Rohtak (India)199828°53 N 33·3 ± 7·83
Marrakech (Morocco)200131°38 N 46·0 ± 0·12
Montevideo (Uruguay)200034°51 S 40·2 ± 1·75
Chincoteague (USA)200137°55 N 40·6 ± 2·54
Capri (Italy)200040°33 N 31·0 ± 2·45
Foissac (France)200144°02 N 25·3 ± 1·35
Bordeaux (France)200144°51 N 32·2 ± 2·58
Besançon (France)200047°15 N 27·1 ± 2·94
Seattle (USA)200047°56 N 34·0 ± 1·85
Prunay (France)200148°41 N 29·0 ± 6·04
Draveil (France)200148°44 N 32·6 ± 1·911
Helsinki (Finland)>5 years60°10 N 33·9 ± 1·24
Total38·64 ± 7·81110

experimental procedures

Groups of adults (about 20 pairs) were taken from the mass strains and used as parents for the production of experimental flies. Each group oviposited at 21 °C for a short duration (6–8 h) in culture bottles (110 ml) containing a cornmeal–agar medium seeded with live yeast. A piece (about 5 g) of a killed yeast medium (David & Clavel 1965) was added to each bottle to improve the feeding quality and reduce possible larval crowding effects. With this procedure, the number of adults in each bottle was fewer than 500. After emergence, adults were transferred to fresh food and kept for 2–4 days before the experiment. All 24 strains were studied following development at 21 °C.

The influence of developmental temperature was analysed in a subset of nine strains, four from temperate and five from tropical origin. Oviposition took place at 21 °C as above, and culture bottles were transferred, after removing the parents, to constant temperature incubators regulated at ±0·1 °C. Seven experimental temperatures were utilised: 12, 14, 17, 21, 25, 28 and 31 °C. After emergence, adults were transferred to fresh food and kept at their growth temperature before being subjected to cold stress.

recovery time (rt) from cold treatment

Adults aged 2–4 days were transferred without anaesthesia in empty glass vials which were immediately set in a box containing melting ice, that is at a temperature of 0 °C (see David et al. 1998). The number of flies was not precisely controlled, but was about 50. Duration of cold treatment was always 16 h. For measuring recovery time, adults are placed into a Petri dish at room temperature (22–24 °C). At the beginning, all flies are in chill coma and unable to move. The transfer to room temperature permits a progressive recovery, starting by the capacity to move the tarsi, then the legs and finally to stand up. We considered a fly recovered from chill coma when it could stand on its legs: the fly was then removed from the Petri dish, and the time elapsed from the beginning noted as an estimate of RT. For each group, the mean RT was calculated and used as a basic observation.

In D. melanogaster, RT is slightly longer in males than in females (David et al. 1998), but the difference is small and the values in both sexes from the same batch are highly correlated. For simplicity we decided to pool the data of males and females, as in other investigations (Gibert et al. 2001; David et al. 2003). We know (David et al. 1998) that RT is prone to uncontrolled variations when different batches of flies from the same strain and the same age, grown at the same temperature but in different vials, are investigated. For that reason, as in previous works (Gibert et al. 2001; David et al. 2003) we always repeated the analysis several times on the same strain but on different batches of adults and in different generations. The accuracy with which the mean RT of a strain is known depends not on the total number of flies observed, but on the number of independent repeats (David et al. 1998).


genetic variation in relation to latitude of origin

The list of the 24 investigated populations is given in Table 1. For each population, several independent repeated measurements were made (range 2–11, average 4·65 ± 0·41). As in previous work (David et al. 1998), we found significant variations between independent repeats on the same strain, so that the mean of each population is defined by the number of repeats and not by the number of flies. anova (not shown) revealed a highly significant heterogeneity among investigated populations. We further investigated the origin of these variations by analysing the relationship between recovery time and latitude. A highly significant negative linear regression was found, shown in Fig. 1 (slope =−0·284 ± 0·056, P < 0·001, intercept = 45·81 ± 1·79, P < 0·001). Figure 1 reveals a large variability of the populations around the linear regression. We tested the possible influence of a quadratic component but this effect was not significant (ancova, P = 0·25). The general latitudinal cline reflects an overall adaptive genetic variation: populations living in tropical places recover more slowly and are more sensitive to cold than temperate populations.

Figure 1.

Relationship between recovery time and latitude for 24 populations of D. melanogaster. The linear regression is shown as well as its confidence intervals.

Population level analyses of this nature assume the populations are genetically independent and not connected by migration, although this is not always the case (Felsenstein 2002). We think that population non-independence is unlikely to be important for two reasons. First, geographically distant populations were selected, which should minimize connection due to human activities. Second, we investigate an adaptive trait that is maintained by local climate selection, sufficiently strong to suppress the effects of incoming migrants.

phenotypic plasticity according to growth temperature

The previous experiments were all undertaken on flies reared at 21 °C. Morphological traits are, however, highly plastic in relation to growth temperature (David et al. 2004) and we checked whether chill coma recovery time would be influenced by thermal conditions during development. We chose to analyse populations from the two ends of the cline, i.e. in four temperate populations from France, and four Afrotropical populations. We also investigated an Indian subtropical population from Rohtak, near New Delhi. Our results revealed a major effect of growth temperature: RT was always very short following development at 12 °C, and increased steadily when higher temperatures were used. This phenomenon is illustrated (Fig. 2) by the distributions of recovery time for three growth temperatures (12, 21 and 31 °C) for two populations from Kenya and France. The figure shows that there is no overlap between 12 and 21 °C although an overlap may exist between the distributions at 21 and 31 °C in the Kenyan population. It is also clear that an increase in mean value is accompanied by an increase in the variance of the distribution. To ensure homogeneity of variances, data were log-transformed before being submitted to anova (Table 2). Calculations revealed highly significant effects due to population, temperature and their interaction. More significantly, the major portion of the variance (80%) was due to temperature, i.e. to phenotypic plasticity, while genetic differences between populations accounted for only 3·7%. We subsequently analysed the shape of the response curves (recovery time as a function of growth temperature), that is the reaction norms, using a polynomial model (see David et al. 1997). It turned out that the best fit was provided by a linear regression, and the results of these regressions are given in Table 3 and shown in Fig. 3. For all populations, the slopes are significantly positive. The average value for the temperate populations is 3·11 ± 0·17, significantly less than the average (3·78 ± 0·11) obtained for the tropical populations (see legend of Table 3). A lesser slope in the temperate populations suggests that their reactivity to temperature might be less than that of the tropical populations. But the mean values of the two sets of populations are also different (Table 3) so that a scaling effect is possible. To check this possibility, we estimated, for each population, a relative magnitude of plasticity by calculating the coefficient of variation among the seven experimental temperatures. As seen in Table 3, coefficients of variation (CVs) were significantly greater for the temperate than for the tropical populations, suggesting a higher reactivity to growth temperature in temperate flies.

Figure 2.

Distribution of individual recovery times for three growth temperatures (12, 21 and 31 °C) in two populations: a tropical population from Mombasa (Kenya) and a temperate one from Prunay (near Paris, France).

Table 2.  Results of a mixed model anova with population as random and temperature as fixed effect on adult recovery time. Repeats (number of vials) are nested into population and temperature. Data were log-transformed to homogenize variances
 dfMSFVariation explained (%)
  • ***

    P < 0·001.

  • df, degree of freedom; MS, mean square; F, variance ratio.

Population (1)     8  71·53 32·66*** 3·7
Temperature (2)     62101·3278·31***80·7
Repeat (1 * 2)   203   2·19 24·33*** 2·8
Interaction 1 * 2    48   7·55  3·45*** 2·3
Error18 238   0·09 10·5
Table 3.  Results of linear regression analysis applied to the nine investigated populations. Comparisons of the two geographical groups are made with anova (F1,7). Slopes are significantly steeper in tropical populations (P = 0·018), and this difference does not allow a comparison of intercepts. Mean values of recovery time (RT) are calculated for the average development temperature of 21·1 °C, and difference is highly significant (P = 0·0001). Coefficients of variation (CV) are calculated between the seven experimental temperatures, and the plasticity is more pronounced in the temperate populations (P = 0·045)
ClimateStrainInterceptSlopeMean RTCV
TemperatePrunay −31·942·79***27·0475·9
Draveil −37·283·15***29·3177·9
Foissac −33·602·85***26·6577·2
Mean ± SE −37·68 ± 3·103·11 ± 0·1728·02 ± 0·6880·9 ± 3·4
TropicalSao Tomé−41·643·83***39·3370·8
Mombasa −39·573·75***39·7168·3
Mananara −36·723·48***36·8569·1
Pointe Noire−46·994·21***42·0172·8
Mean ± SE−41·29 ± 1·503·78 ± 0·1138·66 ± 1·1571·5 ± 1·3
Figure 3.

Linear relationship between recovery time and developmental temperature for four temperate and five tropical populations of D. melanogaster. Only regression lines are given for the sake of simplicity.


Recovery time after a cold shock at 0 °C is a laboratory assay aimed at measuring the cold sensitivity of a given Drosophila strain. Like most laboratory assays, it does not correspond to a situation occurring in nature. So the question is: are our results relevant to natural selection and adaptation? We believe that our assay provides a positive answer for four main reasons.

First it has been shown (Gibert et al. 2001) that recovery from chill coma induced at 0 °C permits an unambiguous classification of species as either tropical or temperate. Second, a brief survey (Gibert et al. 2001) revealed that this distinction between tropical and temperate species is also observed in other insect orders, suggesting a need for more extensive comparisons. Third, we have shown here that genetic variations among world populations of D. melanogaster vary according to a latitudinal cline. Populations living at high latitudes under temperate climates recover more quickly (i.e. are more cold tolerant) than tropical populations. Fourth, and still more interestingly, we show here that recovery time is a plastic trait, depending on developmental temperature. The results are in the expected direction for an adaptive plasticity: a development at low (12 °C) temperature considerably increases the cold tolerance of adults so that D. melanogaster (an Afrotropical species) approaches the values typical of purely temperate species (Gibert et al. 2001).

To our knowledge, this is the first time that a linear reaction norm based on development according to growth temperature has been observed in Drosophila. Other fitness traits, such as viability, offspring production, male fertility or adult longevity, exhibit a generally sharp decrease at both ends of the thermal range (Cohet & David 1978; David et al. 1983; Pétavy et al. 2001; Chakir et al. 2002). In such cases, phenotypes may be interpreted as pathological, reflecting the deleterious effects of extreme developmental temperatures. However, in the case of cold tolerance plasticity clearly has a beneficial effect. In this respect, chill coma recovery appears to support the beneficial acclimation hypothesis (Leroi, Bennett & Lenski 1994) but it could also be interpreted as ‘colder is better’ (Huey et al. 1999). Analogous trade-offs between cold tolerance and fertility were also observed after acclimation of D. melanogaster adults at 11 °C (Bubliy et al. 2002). The fact that a given environment may result in pathological effects on some fitness traits but in beneficial effects on others illustrates the complexity and diversity of plastic responses, and the difficulty of generalized interpretations.

In a temperate country such as France, autumn generations do develop at temperatures close to 12 °C. This is shown by examining morphometric traits and knowing their reaction norms. The two most informative traits are female abdomen pigmentation and wing/thorax ratio (David et al. 1994; Pétavy et al. 2002) which both exhibit monotonically decreasing reaction norms. A low developmental temperature in adult flies collected in October and November is inferred by the observation of a very dark pigmentation and high wing/thorax ratio (J. R. David, unpublished observations). During winter, adults developed at a low temperature probably overwinter in sheltered places, although this is not well understood (Izquierdo 1991). They are probably submitted to repeated cold stresses and the capacity to recover rapidly is an advantage, which is also correlated to a better survival (J. R. David, unpublished observations).

From an evolutionary point of view, a consensus exists that D. melanogaster is native to tropical Africa (David & Capy 1988) and that it could have reached Europe several millennia ago, independently of modern human transportation (Lachaise et al. 1988). The cold adaptation seen in temperate populations may be the consequence of a long-lasting directional selection. The observation of a similar adaptive plasticity in the ancestral tropical populations is, however, unexpected since temperatures approaching 0 °C and able to induce a chill coma are not encountered in nature. Two opposite kinds of explanation may be proposed. First, we may assume that plasticity, which presumably implies some specific changes in the nervous system (see David et al. 1998), is a by-product of some more general functional changes related to growth temperature. For example, a lowering of temperature in ectotherms increases the level of unsaturated fatty acids, so that the fluidity of cell membranes is kept stable (Cossins & Bowler 1987; Hazel 1995). In Drosophila, it has been shown in several species that acclimation to low temperature resulted in a qualitative change in the proportion of monoenes and diens, and also in a decrease in the length of fatty acids (Ohtsu, Kimura & Katagiri 1998). In this respect, chill coma plasticity might be considered as an exaptation, according to Gould & Vrba (1982). Second, it is also possible that D. melanogaster, in spite of its Afrotropical origin, was already adapted to cold. Cold temperatures are encountered in the mountains of tropical countries and the occurrence of mountain populations in that species is known. Indeed, the collection of a wild living male in a natural habitat at an altitude of 3000 m was already mentioned (David & Tsacas 1981). In other words, D. melanogaster could have been somehow preadapted to the colonization of temperate countries by first colonizing afrotropical mountains. To examine this hypothesis, it would be interesting to investigate colonizing species which, like D. ananassae (Morin et al. 1997) or Zaprionus indianus (Karan, Moreteau & David 1999), are readily transported by humans, but are restricted to tropical places and apparently unable to tolerate cold winters.