Optimal temperature range of a plastic species, Drosophila simulans

Authors


Correspondence author. E-mail: ajmoehring@gmail.com

Summary

  1. When a species experiences a new climate, it can adapt in two main ways: become genetically adapted to the new temperature, or adopt a plastic approach that allows it to survive at a range of temperatures.
  2. The constraint on fitness for genetically adapted populations that are exposed to a new temperature has been well studied, but the range of optimal temperatures and their effect on fitness has never been examined across the worldwide distribution of a plastic species.
  3. Here, we determined the optimum temperature range of 11 populations of the phenotypically plastic species Drosophila simulans. We measured the influence of temperature on eggs, larvae and adults at six temperatures that span the natural range the flies experience during their primary breeding season.
  4. We found no correlation between optimum temperature and native temperature, an effect that is not likely due to laboratory maintenance, suggesting that the species has not locally adapted to temperature. We also found that this species had equal survival and reproductive success at most of the temperatures and life stages that we tested, regardless of the native temperature where the flies originated.
  5. Thus, this genetically plastic species has an optimum fitness at a surprisingly wide range of temperatures, and is the first example of a cosmopolitan species exhibiting this large amount of plasticity across its sampling distribution.

Introduction

A population becomes better suited to its habitat through heritable changes over successive generations. Organisms adapt to their environment in two main ways: local adaptation and phenotypic plasticity. Local adaptation occurs when a population of organisms becomes genetically specialized to the unique environment the population resides in (Angilletta 2009). This allows for highly fit individuals to be uniquely adapted to the local environment, with the consequence that they are potentially less fit in others. In contrast, organisms may adapt through phenotypic plasticity, which allows for individuals to temporarily adjust their phenotype to the local conditions they experience, and maintain their ability to also temporarily adjust to other environments (Angilletta 2009). Although historically plasticity has been defined in a variety of ways, here we restrict the definition of plasticity to the ability of organisms to adjust their phenotype to survive and reproduce when exposed to one environment vs. another. Through plasticity, organisms can also adjust their phenotype during development to be best suited to the current environment, and thus are ‘environmental generalists’ (David et al. 2004). Just as with generalist vs. specialist feeding, there are costs associated with exhibiting phenotypic plasticity. If an individual adjusts incorrectly to a given thermal environment in a way that is detrimental to their life cycle, it may result in no direct offspring in the next generation. In addition, although the generalist approach may allow individuals to survive, they may not be quite as fit in that particular environment as other species that are locally adapted, and therefore may be outcompeted by the specialist species (Angilletta 2009). While there have been many studies examining the costs associated with local adaptation when those populations are later exposed to a new environment (Visser 2008), the limitations of phenotypic plasticity have not been well characterized.

The ability to survive shifts in temperature is even more important in the face of ongoing and rapid climate change. Populations of organisms that are locally adapted are going to be forced to change rapidly to adjust to new environmental conditions, or change their habitat range to reflect their temperature tolerance. In contrast, those individuals that are able to exhibit plasticity can adjust their development to cope with changes in temperature. Thermal adaptation has been relatively well studied across a variety of groups, including algae (Eggert et al. 2003), reptiles (Licht et al. 1966; van Berkum 1988), crustaceans (Gaston 1998), amphibians (Miller & Packard 1977), and insects (Addo-Bediako et al. 2000; David et al. 2003; Matute et al. 2009). Most of these studies focus on the tolerance of populations to extreme temperatures rather than examining their fitness at intermediate temperatures (but see Dell et al. 2011). However, with climate change, organisms will first be exposed to these intermediate shifts in temperature, and their ability to adapt to these initial shifts will determine their fitness and survival.

Thermal adaptation has been well investigated among many groups of organisms, but is of particular interest in ectotherms (Huey & Stevenson 1979; Angilletta 2009), whose body temperatures closely follow the ambient temperature. To thoroughly examine maximum viability at a variety of temperatures, while controlling for other variables, including response to being housed in an environmental chamber, it is necessary to employ a model system that survives well in a laboratory. The quick development and generation time of Drosophila allows for its efficient use in studying life-history traits within the laboratory (Demerec 1950). The life cycle of Drosophila involves several discrete stages, beginning with the fertilized egg, then three larval stages (1st, 2nd and 3rd instar) before pupating, undergoing metamorphosis, and eclosing as adult flies (Demerec 1950). The adult fly becomes reproductively mature in hours to days following eclosion, depending on the species and sex (Pitnick et al. 1995; Promislow & Bugbee 2000). Although many species of the genus Drosophila are well studied at the genetic level, far less attention is given to their ecology (Powell 1997; Markow & O'Grady 2007; Matute et al. 2009). Most studies that use Drosophila to examine the survival and physiological response to changing temperatures, use samples from only one or two geographical regions (Hoffmann & Weeks 2007) or compare single strains of two Drosophila species (Tantawy & Mallah 1961; Giesel et al. 1982; Montchamp-Moreau 1983; David et al. 2004; Overgaard & Sørensen 2008); few have looked at the divergence of a species across its entire range (Capy et al. 1993). For example, the effect of temperature on developmental time and reproductive success was examined in detail for several strains of D. santomea and D. yakuba, but different strains of each species are all from the same geographic area (Matute et al. 2009). The lack of plasticity in these traits for D. santomea was thought to restrict the species to a defined mountain habitat, thus creating extrinsic reproductive isolation from D. yakuba. The effect of temperature on traits other than survival and reproduction has also been examined, including its effect on body size (Bakker 1959; Capy et al. 1993; Reeve et al. 2001; Noory & Loescheke 2002; Hoffmann & Weeks 2007) and offspring sex ratios (Tantawy & Mallah 1961; Burke & Little 1995; Pétavy et al. 2001; Marshall & Sinclair 2010).

The worldwide cosmopolitan species of D. simulans is a close relative of the more widely used D. melanogaster (divergence c. 3 MYA), with both species originating in the Afrotropical region (Lachaise & Silvain 2004). Many comparative studies have been done between the two species to measure their genetic and physiological differences (Throckmorton 1975; David et al. 1983; David et al. 2003; David et al. 2004; Tamura et al. 2004; Cutter 2008). Overall, D. melanogaster has greater genetic differentiation and more variation in morphological traits among different populations than D. simulans populations, although some variation does exist in life-history traits in D. simulans based on the sampling location, including adult longevity (Parsons 1977) and birth rate and fecundity (Murphy & Giesel 1983), but the level of variation among populations is generally less than that of D. melanogaster (David et al. 2004). It is therefore thought that D. melanogaster is more genetically adapted to its local environment, whereas D. simulans populations utilize phenotypic plasticity (Capy et al. 1993; Capy & Gibert 2004; Gibert et al. 2004; Trotta et al. 2006).

Here, we measure the level of temperature adaptation of a single species using populations that have been sampled across the entire species range, subjecting them to temperatures that span the median annual temperature of their environment, and measuring survival and reproductive traits at multiple life stages as an indicator of fitness. Furthermore, we have chosen a species that is considered to be phenotypically plastic, D. simulans, providing the first broad measures of optimal temperature range limits in a plastic species. We subject eleven populations of D. simulans to both native and non-native (and potentially suboptimal) temperatures at egg, larval and adult life stages and assess their survival, reproductive success, body mass and the sex ratio of their offspring. Due to the infeasibility of collecting flies simultaneously from around the world, we compared the optimum temperature of long-standing laboratory strains to a wild-caught strain, allowing us to define the limitations of long-standing maintenance in a constant laboratory environment. By comparing the ‘optimum temperature range’ in the laboratory to the average temperature of the region where the population was sampled from, we also determine the extent that individual populations of D. simulans have adapted to their native environment vs. maintained an optimum surrounding their ancestral Afrotropical climate.

Materials and methods

Drosophila stocks and rearing

Ten populations of D. simulans were sampled from a variety of geographic locations (Table 1) and maintained for many generations at the Drosophila Species Stock Center (maintained at 23 °C), before being acquired by the Moehring Lab. One wild-caught population of D. simulans was created from 30 isofemale strains that were collected in 2009 from Niagara Falls (NF), Ontario, by Mark Fitzpatrick and maintained in large population cages to reduce any potential inbreeding. Approximately 1 month before the experiment, a sampling of the pooled wild caught, NF population was provided to the Moehring Lab and subsequently stored in 30 mL (8 dram) vials. All Drosophila stocks were reared on standard corn syrup-cornmeal-agar recipe medium (Bloomington Drosophila Stock Center) in 30 mL (8 dram) vials (‘food vials’) and maintained at 21 °C on a 14 h : 10 h light-dark cycle and 75 ± 10% relative humidity. Assays were performed at a range of temperatures chosen to surround the reported optimum temperature for D. simulans of c. 21·3 °C (Pétavy et al. 2001): 14, 18, 21, 24, 27 and 30 °C.

Table 1. Drosophila simulans strain origin used for temperature assays
Strain numbera,bOrigin locationaYear population sampledaLatitude, longitudecMean of three warmest monthly temperatures (°C)dMean of three coldest monthly temperatures (°C)d
  1. a

    Data provided by the Drosophila Species Stock Center.

  2. b

    Names of strains are referred to by the last three digits (following 14021-0251).

  3. c

    Data from maps.google.com.

  4. d

    Data from the National Oceanic & Atmospheric Administration (NOAA).

14021-0251.176Rosevears, Australia199727°00′S, 133°00′E17.111.3
14021-0251.199Nanyuki, Kenya19980°02′N, 37°04′E19.49.3
Genetic strain c167.4Nanyuki, Kenya19980°02′N, 37°04′E19.49.3
14021-0251.196Antsirabe, Madagascar199819°86′S, 47°03′E20.714.9
NFNiagara Falls, Canada200943°06′N, 79°02′W21.6−3.5
14021-0251.005Lima, Peru195612°00′S, 77°09′W23.316.2
14021-0251.194Winters, CA, USA199538°53′N, 121°96′W25.58.5
14021-0251.216Winters, CA, USA199538°53′N, 121°96′W25.58.5
14021-0251.198Noumea, New Caledonia199122°30′S, 166°50′E26.120.2
14021-0251.197Joffreville, Madagascar199812°49′S, 49°20′E26.325.2
14021-0251.165/FCFlorida City, FL, USA198525°30′N, 81°20′W27.620.1

Egg hatchability

Eggs were temperature treated as in Matute et al. (2009). In brief, flies were transferred to population cages containing cornmeal medium in a Petri dish to allow for egg laying for 16 h at 21 °C; 50 eggs were transferred to food vials for the experimental temperature treatment (14, 18, 21, 24, 27 and 30 °C; n = 4 for each strain). The number of eggs that hatched after 24 and 48 h was counted, and the larvae were reared at 21 °C to adulthood. Our preliminary experiments suggested that no further eggs hatched after 48 h of temperature incubation at any of the experimental temperatures. The sex ratio and mass of males vs. females was measured for each strain and temperature c. 21 days following initial incubation, following eclosion as adults. Dry mass of individual flies (n = 3) was determined using an MX5 microbalance (± 0·5 μg; Mettler Toledo, Columbus, OH, USA).

Larval survival and development time

Five adult flies of each sex were placed together in a 30 mL food vial at 21 °C to allow for egg laying. After 48 h, the adults were removed from the vials and the 1st instar larvae were incubated at each experimental temperature (14, 18, 21, 24, 27 and 30 °C; n = 4 for each strain). The number of larvae was not standardized among strains, as we focused on the differences within a strain among temperatures. Flies that eclosed were counted, sexed and removed daily to prevent any additional eggs being laid on the food medium to determine the number of eclosing time and mean development time of each strain, at each experimental temperature. The number of males and females that eclosed from each vial was used to determine the development time and sex ratio. The assays were discontinued when 5 days had passed when no new flies eclosed from the vial. The dry mass of males vs. females was measured as outlined above.

Adult fitness: mating behaviour

Eggs and larvae were kept at a constant temperature of 21 °C until eclosion. Virgin flies were incubated for 5 days at each experimental temperature (14, 18, 21, 24, 27 and 30 °C; n = 20) and then paired with a virgin of the opposite sex in a no-choice mating assay in a 30 mL water-misted vial. Each pair was observed within 1 h of lights-on for 60 min for the incidence of courtship and copulation behaviours at 21 °C. A constant temperature was chosen for assays to control other variables that affect behaviour such as humidity, light, odour, etc. (Spieth 1974). The dry mass of each sex and strain at each temperature was determined as outlined above.

Statistical analysis

The program R 2·14·1 (R Development Core Team 2012) was used for statistical calculations and all hypotheses were tested at α = 0·05. For egg hatchability, the arcsine square-root transformed per cent survival of each replicate, strain and experimental temperature was compared using a multifactor ANOVA with Tukey's post-hoc test to discern any differences among strains and temperatures. For the larval assay, the mean number of eclosing flies of each replicate, strain and experimental temperatures, was compared using a multifactor ANOVA with Tukey's post-hoc test to discern any differences among strains and temperatures. Pearson's correlation compared the experimental optimum temperature of each strain with the mean temperature of the three warmest months, as well as the three coldest months, at the closest weather station to the collection site. The optimum temperature for each strain will be the one which had the highest per cent survival of eggs or larvae. Optimum temperature range was determined using overlapping 95% confidence intervals (CI) of individual data points. Pearson's chi-squared test compared the observed ratio of male and female eggs and larvae to a 1 : 1 male to female ratio. A full-factorial two-way within subjects ANOVA was used to compare the mean mass of males and females for each experimental temperature, where strain was kept as a random factor in the model.

For adult behaviour, the proportion of courtship and copulation behaviours of each replicate, strain and experimental temperatures was compared using a logistic regression. Pearson's correlation was used to compare the experimental optimum temperature of each strain with the mean temperature of the three warmest months, as well as the three coldest months, at the closest weather station to the collection site. The optimum temperature for each strain was determined as the one with the highest per cent courtship and copulation at each temperature. Breadth of optimum temperature was determined using overlapping 95% CI of individual data points. A full-factorial two-way within subjects ANOVA was used to compare the mean mass of males and females for each experimental temperature, where strain was kept as a random factor in the model. Page's trend test was used to compare the ranked order of populations among all four measures contributing to fitness for their overall breadth in tolerance, (°C where the population performs equally as well). This test determines if there is a significant correlation at the P = 0·05, 0·01 or 0·001 level, or if the correlation is not significant.

Results

The egg hatchability assay involved transferring a total of 13 200 eggs onto filter paper for incubation. The egg hatchability of D. simulans was significantly different among strains after 24 h (two-factor ANOVA, F10,198 = 1215·211, < 0·001; Table 2), and significantly different among experimental temperatures (F5,198 = 36·254, < 0·001). The egg hatchability strain by temperature interaction, however, was not significant (F50,198 = 1·118, = 0·292). A post-hoc test revealed significant differences among many individual strains and experimental temperatures (Tukey's HSD). The egg hatchability of D. simulans was significantly different among strains after 48 h (two-factor ANOVA, F10,198 = 43·054, < 0·001; Fig. 1a; Table 2), but was not significantly different among experimental temperatures (F5,198 = 0·375, = 0·866). The median temperature producing the greatest number of hatched eggs across all strains was 22·0 °C. The eggs had equally optimum hatchability over a wide temperature span, with the optimal range spanning an average of 16·0 °C. We would therefore not consider any of the tested strains to be ‘constrained’ in terms of their response to temperature as it relates to optimum egg hatchability (<5 °C spread) and all eleven strains we would consider ‘highly plastic’ in their optimum response (>10 °C spread), with all strains performing equally well across the full range of temperatures tested (Fig. 1a). The egg hatchability strain by temperature interaction was also not significant after 48 h (F50,198 = 0·781, = 0·848). A post-hoc test revealed significant differences among many individual strains (Tukey's HSD). There was no significant correlation between optimum temperature for egg hatchability and each strain's native temperature of the three warmest months (Pearson's correlation, = −0·395, = 0·229; Appendix S2a), or the three coldest months after 48 h (Pearson's correlation, = −.081, n = 11, = 0·8136; Table 2; Appendix S2b). There was no significant deviation from a 1 : 1 sex ratio at any of the experimental temperatures (Pearson's chi-squared test; Table 3). After temperature treatment at the egg stage, there was a no significant change in the mean mass of adult flies with temperature, sex, or their interaction (two-way within subjects ANOVA; Table 2).

Table 2. Summary statistics for 11 strains of Drosophila simulans, incubated at six experimental temperatures, during the egg, larval and adult stage. Two-factor ANOVA used for differences in egg and larval survival and developmental time among effects. Logistic regression analysis for predictors of temperature and strain on the proportion of adults exhibiting mating behaviours with Wald's chi-squared test. Two-way repeated measures ANOVA used for differences in body mass among effects. Pearson's correlation statistics for temperature of highest survival, N = 11 for all groups
MeasureEffect
StrainTempSexStrain × TempTemp × SexWarm months correlationCold months correlation
d.f.F or χ2P or Pr> χ2d.f.F or χ2P or Pr> χ2d.f. F P d.f.F or χ2P or Pr> χ2d.f. F P r P r P
Egg Stage
Hatchability (24 h)10, 1981215.21<0.0015, 19836.25<0.001N/AN/AN/A50, 1981.120.292N/AN/AN/A−0.3950.229−0.0810.8136
Hatchability (48 h)10, 19843.05<0.0015, 1980.3750.866N/AN/AN/A50, 1980.780.848N/AN/AN/A
Adult body massN/AN/AN/A1, 7420.5070.4991, 7424.6700.068N/AN/AN/A1, 7420.4590.520
Larvae Stage
# Eclosing10, 24295.6650.0075, 2425.300<0.001N/AN/AN/A50, 2420.9840.204N/AN/AN/A−0.4670.147−0.3090.786
Developmental time10.2421.4340.1725.242144.195<0.001N/AN/AN/A47.24242.284<0.001N/AN/AN/A
Adult body massN/AN/AN/A1, 2082.0370.1971, 20830.215<0.001N/AN/AN/A5, 1501.6800.143
Adult Stage
Courtship10.27410.5364108.95510.6006N/AN/AN/A104.07530.9439N/AN/AN/A−0.4310.185−0.3950.230
Copulation19.90250.44911041.2633<0.001N/AN/AN/A104.96000.8938N/AN/AN/A0.1350.3470.5620.072
Adult body massN/AN/AN/A1, 4540.2770.5991, 45435.158<0.001N/AN/AN/A1, 4541.1340.287
Table 3. Pearson's chi-squared analysis for comparison of sex ratios to 1 : 1 at six experimental temperatures, during the egg and larval life stages in Drosophila simulans (d.f. = 1)
Life stageTemperature (°C)Sex ratio (♀/♂) χ 2 P
Egg140.48640.22200.6376
180.36833.38800.0657
210.54352.39100.1220
240.57191.88000.1703
270.52450.27100.6028
300.52300.03200.8586
Larvae140.48150.11100.7388
180.54640.79300.3731
210.53711.93400.1643
240.51030.04100.8391
270.51520.10600.7444
300.65713.5450.0597
Figure 1.

Eleven strains of D. simulans incubated at six temperatures, measuring (a) mean ( ± SE) egg hatchability of eggs after 48 h of incubation (n = 4), (b) mean ( ± SE) number of eclosing flies after larval incubation (n = 4), (c) proportion ( ± 95% CI) of courting pairs after five days of incubation (n ≥ 13*), (d) proportion ( ± 95% CI) of copulating pairs after five days of incubation (n ≥ 13*), (e) range of mean monthly temperatures of native environment for each strain, where the small circles denote the mean temperature of the three warmest months for each strain and the shaded box on the x-axis denotes the temperatures assayed in this study. Note: some data are not shown, as some points are hidden behind identical points. Lines are colour coded from those with the lowest native temperature (purple) to the highest native temperature (dark red). Lines below figures denote ranges of temperatures with the highest success (e.g. egg hatchability) for that strain, based on a 95% confidence interval. Dotted lines denote significant decreases within the overall peak. *See Appendix S1.

The larval survival assay involved transferring a total of 2640 flies to vials for egg laying, and removing 1540 adults after eclosion. Although these data violated the assumptions of homoscedasticity and normality, ANOVA is considered robust to departures from normality (Dempster & Lerner 1950). The larval survival of D. simulans was significantly different among strains (two-factor ANOVA, F10,242 = 95·665, P = 0·007; Fig. 1b; Table 2) and experimental temperatures (F5,242 = 5·300, P < 0·001). The median temperature producing the greatest rate of larval survival across all strains was 22·0 °C. The range of experimental temperatures producing equal larval viability had an average span of 14·2 °C. None of the strains were considered ‘constrained’ in terms of their response to temperature as it relates to optimum larval survival (< 5 °C spread), while 10 strains would be considered ‘highly plastic’ in their response (> 10 °C spread), with eight strains having an optimum spanning the full range of temperatures tested (Fig. 1b). A post-hoc test also revealed significant differences among experimental temperatures (Tukey's HSD). The larval survival strain by experimental temperature interaction was not significant (F50,242 = 0·984, = 0·204). There was no correlation between optimum temperature for larval survival and each strain's native temperature of the three warmest months (Pearson's correlation, r = −0·467, = 11, = 0·147; Appendix S2a), or the three coldest months (Pearson's correlation, = −0·093, = 11, = 0·786; Table 2; Appendix S2b). There was no significant deviation from a 1 : 1 sex ratio at any of the experimental temperatures (Pearson's chi-squared test; Table 3). However, there were significant differences in the mean mass of eclosing flies based on the sex of the fly (two-way within subjects ANOVA), but not the temperature or the interaction between sex and temperature (Table 2).

The developmental time assay involved transferring a total of 2640 flies to vials for egg laying, and removing 1540 adults after eclosion. The developmental time of D. simulans was not significantly different among strains (two-factor ANOVA, F10,242 = 1·434, P = 0·172), but was significantly different among experimental temperatures (F5,242 = 144·195, P < 0·001; Appendix S3; Table 2). A post-hoc test also revealed significant differences among experimental temperatures (Tukey's HSD). The strain by temperature interaction was also significant (F47,242 = 42·284, P < 0·001).

The adult mating behaviour assay involved incubation and mating assays between 1168 pairs of flies (Appendix S1). The proportion of pairs of D. simulans found courting during mating assays was scored. Neither temperature nor strain were significant predictors for courtship, nor was the temperature by strain interaction significant (Table 2; Fig. 1c). The median temperature producing the greatest rate of courtship across all strains was 22·0 °C. The range of experimental temperatures producing equal courtship occurrence after treatment averaged a span of 12·6 °C. One strain we would consider ‘constrained’ in terms of their response to temperature as it relates to optimum amount of courtship (<5 °C spread, or statistically identical levels of courtship across a range of temperatures) and 10 strains we would consider ‘highly plastic’ in their response (>10 °C spread), with five strains having an optimum spanning the full range of temperatures tested (Fig. 1c). For the proportion of pairs of D. simulans found copulating during mating assays, temperature was a significant predictor for copulation (logistic regression, Table 2; Fig. 1d), but strain was not a significant predictor for copulation, nor was the temperature by strain interaction significant. The median temperature producing the greatest number of copulations across all strains was 24·0 °C. There was an equal likelihood of copulations occurring over an average temperature span of 12·4 °C. We would consider none of the strains ‘constrained’ in terms of their response to temperature as it relates to courtship (<7 °C spread). All 11 strains we would consider ‘highly plastic’ in their response (>10 °C spread), with one of the strains spanning the full range of temperatures tested as there was a consistently very low rate of copulation when flies were assayed at 14 °C (Fig. 1d). There was no significant correlation between optimum temperature for courtship and each strain's native temperature of the three warmest months (Pearson's correlation analysis, r = −0·431, = 11, = 0·185; Appendix S2a), or the three coldest months (Pearson's correlation, = −0·395, = 11, = 0·230; Table 2; Appendix S2b). There was also no significant correlation between optimum temperature for copulation and each strain's native temperature of the three warmest months (Pearson's correlation analysis, r = −0·135, = 11, = 0·347; Table 2; Appendix S2a), or the three coldest months (Pearson's correlation, = 0·562, = 11, = 0·072; Table 2; Appendix S2b). There were, however, significant differences in the mean mass of adult flies based on the sex of the fly (two-way within subjects ANOVA), but not the temperature or interaction between sex and temperature (Table 2). There was no pattern in the ranked order of populations among all four measures contributing to fitness (Page's trend test, = 289·5, > 0·05), indicating that individual strains did not consistently perform ‘well’ or ‘poorly’ for the four traits measured.

Discussion

Across all life stages, the maximum performance of D. simulans' survival and reproduction is consistent across a surprisingly wide range of temperatures that is centred around the average range that D. simulans experiences in the wild during the summer months (Table 1; Appendix S2a). In addition, there is no direct relationship between the optimum temperature in laboratory and mean temperature of the sampling location, primarily due to the very broad span of the optimum temperature range. Likewise, the optimum temperature in the laboratory was found to be fairly consistent in breadth across all life stages, with generally little impact on our proxy fitness measures within the range of temperatures we tested.

Other studies on other species of Drosophila have found that temperature has an effect on the survival of eggs (Matute 2009). Although we also found an effect on egg hatchability at 14 °C after the first 24 h of incubation, this difference disappeared after 48 h, making it likely that this effect was simply due to a delay in development time rather than a temperature effect on viability (Etges 1989). Our study found genetic differences in egg viability among populations, consistent with the results of past studies (Trotta et al. 2006). The eggs' response to the experimental temperatures was found overall to be extremely plastic, with each strain displaying consistent hatchability across the entire range of temperatures tested (Fig. 1a).

Similarly, at the larval stage, the number of eclosing adults was consistent across most of the temperatures tested, with only the extreme temperatures having a strong effect on larval survival (Fig. 1b). This result is consistent with past research that examined the larval stage, which has been reported to have the highest survival around 20–28 °C (Murphy and Giesel 1983). As expected, there were differences in the development time of larvae based on the experimental temperature (Appendix S3). As the temperature increased, the development time decreased, until it statistically plateaued starting at 21 °C. Graphically, there appears to be a decrease in developmental time with increasing temperatures, but the decrease among temperatures is not statistically significant (Appendix S3). The general trend towards decreasing developmental time with increasing temperature is consistent with past studies (Pétavy et al. 2001), and may result from physiological pathways that proceed faster at warmer temperatures, or adaptive plasticity that delay maturity at cooler temperatures (Angilletta et al. 2004). However, within the developmental pathway, there are likely physiological costs associated with faster development that do not allow for the eclosion of healthy adults (Chippindale et al. 1997). The strain by temperature interaction results from a single point at 24 °C, which the authors feel is an artefact of the experiment and is not an actual increase in developmental time.

Lastly, at the adult stage, both courtship and copulation behaviours were also consistent across the range of temperatures, but slightly more sensitive than in the developmental stages (Fig. 1c and 1d). Temperature did not have a statistically significant effect on the quantity of courtship behaviour, when the effect was examined across all sampling locations. However, individual differences were found when separating the effects of strain from temperature (Fig. 1c), and this trait had by far the greatest amount of variance in the size of the optimum temperature range of the fitness-correlated traits we measured. Interestingly, the three strains with the smallest range of temperatures in their native habitats (Fig. 1e) also had the most restricted optimum range for proportion courting (Fig. 1c), providing a modest indicator that courtship behaviour may be linked to the native temperature for these populations, and may be an early sign of local adaptation. Temperature also had an effect on the success of attempted copulations (Fig. 1d), primarily driven by very low copulation values at 14 °C, which indicates either that the female's ability to accept potential mates was affected by this temperature treatment, or that the temperature has an effect on the quality, but not quantity, of courtship behaviours performed by males. The possibility that the temperature incubation interfered with sexual maturity in males and females is unlikely since there was a delay of 24 h before temperature incubation, and sexual maturity is reached by more than 80% of individuals within the first 24 h of eclosion (Carracedo and Casares 1987). The wide tolerance of both courtship and copulation to different temperatures spans the native temperatures experienced in the wild (Fig. 1e), as well as the temperature typical of many African countries, which is akin to the ancestral temperature (Table 1).

In contrast with the effect of temperature on survival and reproductive life-history traits, there is no effect of temperature on the resulting sex ratio of flies from each strain, at any of the life stages. This suggests that temperature is not harming one sex more than the other or inducing meiotic drive, as both sexes are being produced in equal numbers. This result is in contrast with past studies, which found female biases at extreme temperatures, which they postulated was due to females being hardier than males at extreme temperatures (Tantawy & Mallah 1961). Our results, however, suggest that D. simulans does not have differential success of one sex over the other, or that these differences can only be seen at more extreme values for temperature incubation than at the ones we tested.

In contrast, there were differences in the effect of sex on the mass of flies across the larval stages, but no effect of temperature was found in our study. Increasing temperatures tend to have a strong negative effect on the size and mass of developing Drosophila, which is consistent with the temperature-size or Bergmann's rule (Bergmann 1847; Lehmann 1999). The mass of an adult fly has also been shown to vary between the sexes, with females being larger (Bakker 1959; Nunney & Cheung 1997; Pétavy et al. 1997). In a similar study, differences in body size between sexes and changes in temperature were found in developing D. melanogaster, which suggests that body size may be less dependent on temperature in D. simulans (David et al. 2011). In addition, there was a non-significant increase in mass at the highest temperature tested (30 °C; Appendix S4), which may have affected our results. Similarly, there was no effect of temperature on the mass of flies after the adult temperature incubation, indicating that the flies reached their final developmental mass by the time they were incubated (24 h after eclosion), and that temperature had no effect on further acquisition of adult mass through feeding behaviour.

Although there were differences in our fitness measurements among strains, such as the different egg hatchability rates, these results may be the result of overall hardiness of each stock. The most recently wild-caught strain, the NF strain, had the highest overall egg hatchability (dark green line, Fig. 1a), which suggests that there may be an effect of inbreeding and long-term maintenance in a laboratory environment on the egg hatchability of the other stocks. However, at the larval and adult stages, the wild-caught strain's survival, developmental time and behaviour were not different from the other laboratory strains, indicating that this effect disappears after the egg stage. In general, laboratory maintenance and inbreeding does not seem to have caused a large effect on the ability of flies to adjust to the differences in temperature tested in our study. Past studies have shown some effects of laboratory maintenance and adaptation; however, these effects were not seen in the current study, as not all peaks in fitness measures occurred within the laboratory temperature range at which the stocks have been maintained (21–23 °C). In addition, the optimum temperature range was fairly similar between strains originally sampled from the same location (see Table 1), but maintained separately at a consistent laboratory temperature, again suggesting that laboratory adaptation has not occurred to a great extent in these stocks. Lastly, given that there is no pattern in the ranked order of populations and their effect on measures related to fitness, it appears no strain is overall more (or less) successful than another, indicating that particular strains do not suffer a general overall reduction in vitality due to laboratory maintenance.

Across the life cycle of all populations of D. simulans, it appears that populations have not differentiated from one another, and have instead maintained a fairly wide tolerance for changes in temperature. The route of adaptation, therefore, seems to not be by local adaptation, but by displaying insensitivity to temperature. This result is surprising, as individuals are not becoming specialists, but retaining the ability to survive in many different environments. One possible explanation for this generalist strategy is an example from France, where temperate populations of D. simulans are founded each spring by migrants from warmer areas (David et al. 2004). This migration from warmer areas to cooler areas may serve as a barrier for local adaptation to occur. One route for individuals to display insensitivity is by exhibiting phenotypic plasticity, which is consistent with the results of this experiment. Across all life stages, the majority of the strains we tested fit the criterion of phenotypic plasticity, which involves optimal output at a wide range of temperatures. Phenotypic plasticity has been observed in a variety of other traits in D. simulans (Capy et al. 1993; Pétavy et al. 1997), and has been proposed as a mechanism by others as a strategy for thermal adaptation (Capy et al. 1993; David et al. 2004). Therefore, our results support the hypothesis that D. simulans uses adaptive plasticity, in contrast with local specialization to a thermal environment.

Although the optimal temperature range is quite large, there is no correlation between optimum temperature and native temperature save for the rate of eclosion after exposure at the larval stage, further supporting that, overall, D. simulans has not become locally adapted to their native environment. This is the first example of a cosmopolitan species exhibiting phenotypic plasticity across the entire sampling distribution. This lack of correlation could potentially be due to D. simulans becoming adapted to a range of microclimates within their habitat, which could have a different mean temperature than the mean ambient temperature (Stanley et al. 1980). For example, a shaded environment that receives minimal sunlight will have a more constant and lower temperature than a habitat in direct sunlight. Therefore, those individuals that are from cooler climates could find warmer microclimates, and individuals from warmer climates could find cooler microclimates to survive at an optimal level (Martin 1998; Nevo et al. 1998; Rashkovetsky et al. 2006). This may be what has occurred in strain 176, sampled from Australia, whose native mean of the three warmest monthly temperature is 17·1 °C, but has a wide range of optimal temperatures, with the point of highest fitness occurring for most traits at c. 24 °C (Table 1; purple line in Fig. 1a-d), indicating that this population may seek out warmer microclimates to perform optimally in its native environment. In a previous study using D. melanogaster, populations from temperate environments, which must deal with large ranges in temperature, have been found to have a high level of plasticity, whereas tropical strains face relatively constant temperatures across the year and had a lower level of plasticity (Ayrinhac et al. 2004; Trotta et al. 2006). Recently, it has been shown that autocorrelation of strain's thermal tolerance for adjacently located populations may affect a typical correlation of experimental fitness temperature maxima and the native temperature (Kellerman et al. 2012). However, no such correlation was found in our study, which must be taken into consideration when comparing the results of our study with results from other species. In our study using D. simulans, the level of plasticity seemed to be fairly constant, as well as surprisingly large across the entire sampling distribution. In general, the cosmopolitan D. simulans appears to use phenotypic plasticity to survive across its entire sampling distribution, while in contrast, D. melanogaster has been shown to exhibit plasticity only in temperate environments for the majority of traits studied to date (Trotta et al. 2006).

In addition to finding a general trend for D. simulans adapting through phenotypic plasticity, the study serves as the first comprehensive examination of contributors to fitness across the worldwide distribution of a plastic species. D. simulans populations sampled from four continents were relatively insensitive to differences in temperature spanning 18–30 °C, and the species as a whole has remained largely phenotypically plastic in response to this environmental condition. This finding is in agreement with a study that examined plasticity in D. simulans across its European distribution (Pétavy et al. 2001). Therefore, across the life cycle of D. simulans, the response to changes in temperature appears to be that of phenotypic plasticity, but is perhaps genetically constrained to centre around the ancestral optimum temperature.

Acknowledgements

An NSERC Discovery Grant to A.J.M. funded this research. Our gratitude goes to Rachelle Kanippayoor, Vanda McNiven and Daniel Matute for their advice on experimental design, to Brent Sinclair and Marc-André Lachance for their advice on this project, to Mark Fitzpatrick for supplying the wild-caught Niagara strain and to Angeli Humilde for her assistance with experimental work. Lastly, thanks to Dianne Kowalski, Natalie Leblanc, Victoria Ransberry and Sonia Wing for their critical input of this manuscript.

Ancillary