Life history consequences of temperature transients in Drosophila melanogaster

Body temperature profoundly affects the performance and fitness of ectotherms (Huey and Stevenson,1979; Cossins and Bowler,1987). Temperature effects are typically studied by maintaining organisms in the laboratory either at different constant temperatures or on different fixed temperature cycles. Nevertheless, ectotherms in nature experience transient weather fronts that may alter their body temperatures for short periods (e.g. Kingsolver,2000). Weather fronts may not only alter an organism's performance during the front itself, but also induce carry-over effects that persist even after weather returns to normal. Surprisingly little is known about the impact of such thermal transients on the physiology and life history of ectotherms,except in regards to short-term exposure to extreme temperatures(Bubliy and Loeschcke, 2001; David et al., 2003; Gibert et al., 2001; Hercus et al., 2003; Krebs and Loeschcke, 1994a; Krebs and Loeschcke, 1994b; Lee et al., 1987; Maynard Smith, 1958; Rohmer et al., 2004; Sisodia and Singh, 2006; Zani et al., 2005a; Zani et al., 2005b).

Drosophila melanogaster is suitable for an examination of the physiological and fitness impacts of non-stressful thermal transients. These flies are widely distributed in temperate zones(Mueller, 1985) where weather fronts are common: thus studies of thermal transients are ecologically relevant to this species. Moreover, a wealth of information is known about the effects of chronic temperature exposure on performance and life history traits in D. melanogaster (David et al.,1983; Hoffmann et al.,2003). For example, fecundity increases gradually with temperature to ∼25°C and then decreases rapidly at temperatures above 28° to 30°C (David and Clavel,1969; Huey et al.,1995; McKenzie,1975; Schnebel and Grossfield,1986; Siddiqui and Barlow,1972).

We studied the impact of thermal transients on fecundity (eggs) and progeny production of Drosophila melanogaster. Initially we measured daily fecundity and progeny production of flies living at 25°C for 3 days and then transferred flies to one of three transient treatments (1 day at 18°C, 3 days at 18°C, or 1 day at 29°C) before returning them to 25°C for 3–5 days. The 1- or 3-day transient exposures are equivalent to typical weather front durations(Allen et al., 1996; Bosart et al., 1973; Robb and Forbes, 2006). We selected transient temperatures of 18°C and 29°C because these are displaced from the optimal temperature (∼25°C) of these flies(Cohet and David, 1978; Huey et al., 1995; Siddiqui and Barlow, 1972),but are not extreme (David et al.,2005). We also maintained some flies at constant temperatures(18°C, 25°C or 29°C) so we could compare the effects of chronic versus transient temperature exposures.

To determine whether transient exposures affected progeny of exposed females, we estimated development time of offspring produced during the experiment. From these data we computed generation time of offspring of experimental flies, net reproductive rate (R₀, that is,total female progeny produced during the experiment) and intrinsic rate of population growth (r). Thus our experimental design enabled us to determine how key life history traits changed during, as well as after, a thermal transient. Although we expected that temperature transients would depress egg production during the exposure, we were uncertain as to whether and how transient exposure would induce carry-over effects once flies were returned to 25°C. Consequently, we considered four competing hypotheses(Fig. 1):

1. No carry-over hypothesis. This hypothesis predicts no carry-over effects of a transient temperature change. If so, then life history parameters will quickly return to baseline levels after the transient.

2. Induced damage hypothesis. If flies are damaged by the thermal transient (Krebs and Loeschcke,1994a), their fecundity may not return to normal levels for some time, if ever. Because the transient temperatures we selected were intentionally non-extreme (David et al.,1983), this outcome is unlikely in our experiments, but would be likely following more extreme treatments.

3. Resource reallocation hypothesis. If flies interpret a sudden temperature change as a cue (sensu Levins,1968) that unfavorable weather will persist for some time, they might reallocate resources from reproduction to survival. If so, their fecundity might remain low for some time even following a return to normal temperatures (Mitrovski and Hoffmann,2001; Schmidt and Conde,2006). Note that the resulting pattern of fecundity parallels that for the `induced damage hypothesis' but has a different causal basis.

4. Compensation hypothesis. When transferred back to normal temperatures, flies might actually increase fecundity above baseline levels for a brief period, `catching up' for fecundity lost during the thermal transient. In effect, compensation could represent a `make hay while the sun shines' response. Compensation could also occur if flies are able to produce,but not lay, eggs during the thermal transient and then can dump accumulated eggs once conditions improve.

We find little effect of thermal transients on flies once they return to 25°C. This result strongly supports the `no carry-over hypothesis'. Consequently, D. melanogaster seems well buffered against short-term exposure to non-optimal (but non-extreme) temperatures. Parallel studies with exposure to more extreme temperatures (or perhaps with longer durations) will be necessary to set boundaries on the range of permissive temperatures.

Because natural responses to thermal transients might decay in laboratory stocks that have been evolving at constant temperatures for many years(Quintana and Prevosti, 1990),we chose to work with a relatively fresh stock. Accordingly, we collected 100 isofemale lines of D. melanogaster from an apple orchard near Wenatchee, WA, USA (47°30′N, 120°17′W) in June 2005, and established a base stock by pooling about 25 males and 25 females from each isofemale line. This stock was maintained at 25°C (12 h:12 h L:D) until July and early August 2006, when these experiments were run. Thus the flies had been in the laboratory for just over 1 year. To maintain the colony at 1000–3000 flies, eggs were collected approximately every 10 days and reared in controlled conditions (50–100 eggs/vial; 25°C and 12 h:12 h L:D; diet of cornmeal, molasses, yeast, agar, tegosept) before newly emerged adult flies were transferred into the colony. Colony food bottles (150 ml)were replaced every 3–5 days to ensure that all colony flies developed in controlled conditions.

Eggs collected from the base stock on July 12, 2006 were reared at low densities of 50–70 eggs/vial (25°C, 12:12 L:D; 2.5×9.5 cm glass vials containing ∼10 ml food). Within 12 h after flies emerged (7/21 and 7/22), adults were briefly anesthetized (CO₂) and triads of flies (one female, two males) were placed into individual vials seeded with live yeast. Egg production increases until females are about 3 days of age(Huey et al., 1995), so fly triads were maintained at 25°C until females were 3 to 4 days of age before assigning them to treatment groups. Flies that did not lay eggs during this initial period were eliminated from the experiment. Fly triads were then transferred (without anesthesia) into fresh vials without supplemental yeast(to facilitate egg counting) and then randomly assigned to one of six treatment groups, with about 18 triads per group. One transient group was kept for 3 days at 25°C, 1 day at 18°C and then 5 days back at 25°C. A second was kept for 3 days at 25°C, 3 days at 18°C, and 3 days back at 25°C. The third was kept for 3 days at 25°C, 1 day at 29°C, and 5 days back at 25°C. Three additional sets of fly triads were maintained at constant temperatures of 18°C, 25°C or 29°C for 9 days. Flies were housed in environmental chambers (Percival Scientific, Perry, IA, USA; 12:12 L:D cycle starting at 07:00 h). Chamber temperatures were monitored daily and vials within a chamber were rotated daily to minimize effects of within-chamber temperature variation. Each triad was transferred without anesthesia into a new vial every day at 09:30–10:00 h so as not to interfere with peak oviposition, which occurs in late afternoon in these flies. Flies were briefly at room temperature (∼23°C) during transfers.

After returning triads to environmental chambers, we counted `transient'eggs laid during the previous 24 h period and then immediately transferred them to 25°C for development. We maintained `constant' eggs at their respective temperatures. Subsequently, we counted newly emerged flies in late morning every day until no more emerged. Because eggs collected during transient exposure completed most of their development at 25°C, any between-groups difference in progeny viability should be due to effects of the transient on the female or on the eggs themselves.

We estimated egg-to-adult viability as the proportion of eggs that produced adults. Egg counts in vials are often in error, because some eggs either might have hatched by scoring time or might simply have been overlooked. Indeed, 156 of 834 vials had a few more flies emerged than eggs counted (treatment had no effect on the number or magnitude of miscounts). For these vials, viability was taken as the number of emerged flies instead of the number of eggs (thus these females were assumed to have 100% viability). Two vials were excluded from the analysis because they had many fewer eggs than emerged adults. However, the statistical results presented below are robust, even if these vials are included.

To determine fitness consequences of the various temperature treatments, we used a life-table analysis (Birch,1948) and estimated generation time (T_g), net reproductive rate (R₀, the total number of female eggs laid during the experiment), and intrinsic rate of increase (r,computed using an iterative technique; see Appendix A) for each female. Estimates of r require development time, which was not measured for experimental females themselves (because all were reared at 25°C, see above). Instead, the average development times of the females' progeny were used as estimates of their development times. (Thus, our estimate of r is based on fecundity patterns of female parents and development times of their progeny, rather than on data collected within a single generation.) For these calculations we assumed a 50:50 sex ratio of progeny.

During the experiment 6 females died (2 at 18°C, 3 at 25°C and 1 at 29°C), 6 escaped during transfers (2 at 18°C and 4 at 25°C), and 12 did not lay any eggs (3 at constant 18°C, 2 at constant 25°C, 1 at constant 29°C, 2 at short 18°C, and 1 at short 29°C). These were excluded from further analysis. Females for which one male escaped or died were included in the analysis. To correct for non-independence of daily measurements within females, we used a repeated-measures MANOVA with daily fecundity, viability (arcsine square root-transformed), or development time as the dependent variable, temperature treatment as the between-subjects effect,and day and treatment×day within-subjects effects. This technique is robust to violations of normality and sphericity(Obrien and Kaiser, 1985). For these analyses, we report several statistics (Pillai's Trace, Hotelling's Trace, Wilks's lambda, and Roy's statistic) because these did not always give similar results. For data not requiring repeated measures, we used multivariate ANOVA analyses with Tukey-HSD post-hoc tests when the data were normally distributed. For non-normal data, we used Kruskal–Wallis rank sum tests with non-parametric Behrens–Fisher post-hoc multiple comparisons tests. All statistics were done in R(R Development Core Team,2006), with packages multcomp(Bretz et al., 2004) and npmc(Helms and Munzel, 2006) for parametric and non-parametric post-hoc tests, respectively, nlme(Pinheiro et al., 2006) for linear mixed effects modeling, and gregmisc(Warnes, 2006), Hmisc(Harrell, 2006), IDPmisc(Ruckstuhl et al., 2006), and gridBase (Murrell, 2006) for advanced plotting features.

We analyzed daily fecundity of 94 females over 9 days (total eggs=14,870). Constant temperatures strongly affected female fecundity during the 9 day experiment (`total eggs per female', Fig. 2A, inset; ANOVA, F_2,43=8.62, P<0.001). Females at 18°C laid significantly fewer total eggs than did females at 25° or 29°C (Tukey HSD, P=0.016, and P<0.001, respectively), but females at 25° and 29°C did not differ significantly in total eggs laid (Tukey HSD, P=0.169).

Females in constant temperature treatments laid fewer eggs as they aged(Fig. 2A; Table 1, MANOVA,within-subjects Day effect, P<0.0001); and the decline in fecundity with age grew steeper with increasing temperature(Fig. 2A). The rapid decline in fecundity with age (Fig. 2A)confounded any simple analysis of effects of transient temperature on egg production, as an observed decline in egg production from day 3 to 4 could be due either to aging or to the transient temperature treatment. Consequently,we standardized mean daily fecundity of transient temperature groups by computing the percent difference between the fecundity of each transient treatment group and that of the constant 25°C group on each day(Fig. 2B).

Prior to exposure to the temperature transient (on days 1–3), females in the constant 25°C treatment and those destined for transient treatments all had similar fecundities (Kruskal–Wallisχ ₃²=2.18, 1.42, 1.42 on days 1, 2 and 3,respectively, all P>0.536; see Fig. 2B). Differences in fecundity among treatment groups on days 4–6 could therefore be attributed to the transient temperature exposure.

A 1-day `hot' temperature transient had no immediate or carry-over effects on female fecundity. Females at 29°C for 1 day had fecundities similar to those of females kept at 25°C on the transient day(Fig. 2B; day 4 post-hoc Behrens–Fisher P=0.700) and on each subsequent day at 25°C (days 5 to 9, Behrens–Fisher, all P>0.52).

Both 1- and 3-day `cold' temperature transients strongly reduced female fecundity, but these effects disappeared immediately when flies were returned to 25°C. While at 18°C for 1 or 3 days, females laid significantly fewer eggs on those days than did the constant 25°C females(Fig. 2B; Behrens–Fisher,all P<0.001). When returned to 25°C transient cold females had similar fecundities to constant 25°C females(Fig. 2B; Behrens–Fisher P=0.960 for post-hoc comparison of 1-day 18°C with constant 25°C on day 5; Kruskal–Wallisχ <sub>3</sub><sup>2</sup>=2.44, P=0.486 for day 7). These results strongly support the `no carry-over hypothesis'(Fig. 1).

Constant low temperatures strongly reduced egg-to-adult viability. Progeny developing at 25°C and 29°C had similar viabilities that were much higher than viabilities of those developing at 18°C (MANOVA,between-subject Temperature effect, P=0.005; Table 1; Fig. 3A).

Egg-to-adult viability did not differ among the three transient treatment groups and the 25°C control on days 1 to 3, indicating that these groups were homogeneous prior to temperature transients(Fig. 3B; Kruskal–Wallisχ ₃²=6.01, 1.88, 2.55; and all P>0.46). A 1-day exposure of flies to 29°C did not affect egg-to-adult viability of their progeny (Fig. 3B;Behrens–Fisher P=0.904 for comparison of 1-day 29°C group with 25°C group on day 4). A 1-day exposure of flies to 18°C reduced egg-to-adult viability of their progeny(Fig. 3B, broken line), but not significantly so (Behrens–Fisher P=0.110). However, a 3-day exposure of flies to 18°C significantly reduced egg-to-adult viability of their progeny on the first 2 days (Fig. 3B, broken line; Behrens–Fisher P=0.002, 0.036 for days 4 and 5, respectively), and marginally so on day 3 (Behrens–Fisher P=0.087 for comparison of 3-day 18°C group and 25°C group on day 6). In no case did temperature transients have carry-over effects on egg-to-adult viability (Fig. 3B; Behrens–Fisher, all P>0.800).

Development time decreased significantly with increasing constant temperature (Fig. 4A; Table 1, between-subjects Temperature effect, P<0.0001). Flies eclosed in 21.8±1.3 days at 18°C, in 9.8±0.8 days at 25°C, and in 8.3±0.7 days at 29°C. Progeny development time did not change significantly with female's age (Fig. 4A; Table 1, within-subjects Day effect, P=0.480). Therefore we compared development time of transient treatment groups directly to the 25°C control.

Transient temperature exposure did not affect development time. Progeny of females moved to 18°C or 29°C for 1 day developed in the same time as progeny of females held at constant 25°C(Fig. 4B, day 4, Kruskal-Wallisχ ₃²=4.95, P=0.175). Development times of progeny from eggs laid by females during their third day at 18°C were slightly but not significantly longer than eggs laid at 25°C on the same day (Fig. 4B, day 6;Behrens–Fisher P=0.085).

Low constant temperatures significantly reduced fitness. Mean generation time decreased significantly, and both net reproductive rate(R₀) and intrinsic rate of increase (r) increased significantly with temperature (Table 2; ANOVA, all P<0.002). Both R₀and r were significantly lower for 18°C flies than for 25°C and 29°C flies (Tukey HSD P<0.05), but these traits were not significantly different for 25°C versus 29°C flies.

Surprisingly, exposure to transient temperatures did not decrease female fitness. Despite major depression effects of transient exposure to 18°C on fecundity and on egg-to-adult viability(Fig. 2B, Fig. 3B), transient exposure to 18°C or to 29°C temperatures did not affect T_g, R₀, or r relative to the constant 25°C treatment (Table 3; ANOVA all P>0.257).

Temperature has strong effects on the performance and fitness of flies and other ectotherms (Huey et al.,1995; McCabe and Partridge,1997; Gilchrist and Huey,2001). Many ectotherms use behaviors to avoid – or at least reduce – the impact of fluctuations in operative temperatures(Cowles and Bogert, 1944; Stevenson, 1985; Huey et al., 2003). Nevertheless, ectotherms will likely be forced to experience non-optimal temperatures during weather fronts. Two issues are germane. First, how much are performance and fitness depressed during a weather front? Second, are there carry-over effects that persist once temperatures return to normal? Our results suggest that female flies experiencing even minor cold fronts (though not minor warming) will have reduced fecundity and egg-to-adult viability during those fronts. Nevertheless, they recover quickly when temperatures return to normal, supporting the `no carry-over hypothesis'.

The constant temperature treatments provide a necessary baseline for comparison of our fly population with other stocks. Overall, the flies used in this study responded to chronic temperatures as expected from previous work. Chronic exposure to 18°C reduced fecundity and viability in our flies, and increasing chronic temperatures reduced development time and changed the pattern of egg production over time. All of these findings qualitatively match previous work in Drosophila(Bochdanovits and de Jong,2003; Bouletreau-Merle et al.,2003; David and Clavel,1969; David and Clavel,1967; Fernandez and Lopez-Fanjul, 1997; Gilchrist and Huey, 2001; Huey et al.,1995; Schnebel and Grossfield,1986; Siddiqui and Barlow,1972; Trotta et al.,2006). However, chronic exposure to 29°C did not reduce fecundity, viability or fitness of our flies even though this temperature is stressful for many fly stocks (David and Clavel, 1967; Fernandez and Lopez-Fanjul, 1997; Gilchrist and Huey, 2001; Huey et al.,1995; Schnebel and Grossfield,1986; Siddiqui and Barlow,1972; Trotta et al.,2006). Thus our stocks appear heat tolerant relative to previously studied ones. This might reflect geographic variation(Schmidt and Conde, 2006; Hoffmann et al., 2003), or the fact that our flies had been in the laboratory for only a year, whereas prior studies often used lab-adapted stocks; heat tolerance in D. subobscura decays quickly in the laboratory(Quintana and Prevosti,1990).

We first review changes in life history parameters during transient exposure. Transient exposure to 18°C negatively affected both fecundity and viability. The reduction in fecundity is not surprising, as chronic exposure to 18°C is known to reduce fecundity in many Drosophilastocks (Bochdanovits and de Jong,2003; Bouletreau-Merle et al.,2003; David and Clavel,1969; David and Clavel,1967; Fernandez and Lopez-Fanjul, 1997; Gilchrist and Huey, 2001; Huey et al.,1995; Schnebel and Grossfield,1986; Siddiqui and Barlow,1972; Trotta et al.,2006), and in our flies (Fig. 2A). However, our finding that viability of progeny is reduced by a 1-day (or less) exposure to 18°C is novel and exacerbates the ecological impact of a weather transient. Eggs were exposed to only 1 day (or less) at 18°C and were at 25°C for the remainder of development (∼10 days),suggesting that fly embryos are very sensitive to even brief exposure to moderately cool temperatures. Whether later stages of development are also sensitive will require additional experiments.

Flies experiencing a 1- or 3-day 18°C transient had a larger decline in fecundity during that transient (70–80% decline relative to constant 25°C females, see Fig. 2B)than did same-age females maintained at constant 18°C (32–65%decline on days 4–6; Fig. 2A). Egg viabilities were similar for the two groups (30–50%for 18°C transients, 20–50% for constant 18°C flies). Thus,transient exposure to 18°C depressed fecundity more than did chronic exposure to 18°C; perhaps chronic exposure to 18°C post-eclosion induces a `beneficial acclimation' effect(Ayrinhac et al., 2004; Nunney and Cheung, 1997).

Transient exposure to 29°C did not significantly affect fecundity or viability relative to flies maintained at 25°C. We were surprised by this result because chronic exposure to 29°C typically reduces the performance and fitness of D. melanogaster(David and Clavel, 1967; Fernandez and Lopez-Fanjul,1997; David et al.,2005; Schnebel and Grossfield,1986; Siddiqui and Barlow,1972). However, in our flies, chronic exposure to 29°C did not depress fitness; thus 29°C appears not to be particularly `hot' for this stock.

Although fecundity and progeny viability were markedly reduced during a 1-or 3-day exposure to 18°C, these traits returned to baseline levels immediately after return of females to 25°C(Fig. 2B, Fig. 3B, days 5–9). This pattern strongly supports the `no carry-over hypothesis' and contradicts alternative hypotheses (Fig. 1).

For our flies, neither fecundity nor progeny viability changed during exposure to 29°C (Figs 2, 3), either chronically or transiently, so it is not surprising that these traits were unchanged(relative to constant 25°C flies) after return to 25°C. Our data document that transient exposures to temperatures as low as 18°C or as high as 29°C have no sustained effects on fecundity or viability. Thus these flies are buffered against temperature fluctuations of more than 10°C. Brief exposures to more extreme temperatures do have sustained negative effects on fitness traits of flies(Krebs and Loeschcke, 1994a; Sisodia and Singh, 2006). In any case, additional experiments will be required to demarcate the range and duration of transient temperatures that inflict sustained damage on flies.

Because transient exposure to 29°C did not alter fecundity or progeny viability relative to control flies, we focus only on the fitness consequences of transient exposure to 18°C. Although these transients reduced fecundity and viability, they did not significantly reduce lifetime fitness relative to constant 25°C flies (R₀, r, Table 3). Initially, this seemed paradoxical to us, but on reflection we believe two factors are responsible:

1. Flies experienced transients for only a fraction of the period for which we obtained life history data (1 or 3 days out of 9), and thus were at 25°C during most of the experimental period. This necessarily dilutes the impact of a transient. Had we terminated the experiment immediately after transient exposure, fitness impacts would be evident.

2. To allow for measurement of baseline fecundities and viabilities of all groups, we waited until flies were 7–8 days old (as adults) before exposing them to thermal transients. By this age, all flies had fecundities that were 70% lower than on day 1 of the experiment (when flies were 3–4 days old; Fig. 2A), and egg viabilities that were already 10% lower than on days 1–3(Fig. 3A). Thus relatively heavy reproductive success early in life likely swamped the negative effects of the thermal transient later in life. This is especially likely for r, which is particularly sensitive to reproduction early in life(Birch, 1948). Very likely,exposure of younger flies to a temperature transient would have significantly reduced fitness.

In conclusion, our experiments demonstrate that non-extreme thermal transients can reduce fecundity and progeny viability, but that these transients have no sustained effects on flies. Thus these flies seem relatively well buffered against moderate weather fronts. Future studies should attempt to delimit the range of transients that are tolerable by flies,as well as whether sensitivity to transients varies with age. Our egg viability data suggest that early embryos are very sensitive to brief exposure to temperatures as mild as 18°C.

We thank M. Frazier and two anonymous reviewers for helpful comments on the manuscript and NSF IOB-0416843 to R.B.H. for funding.