Does selection on increased cold tolerance in the adult stage confer resistance throughout development?

Authors


Anneke Dierks, Zoological Institute and Museum, University of Greifswald, Johann-Sebastian-Bach-Str. 11/12, D-17489 Greifswald, Germany. Tel.: +49 3834 864266; fax: +49 3834 864252; e-mail: anneke.dierks@uni-greifswald.de

Abstract

Artificial selection is a powerful approach to unravel constraints on genetic adaptation. Although it has been frequently used to reveal genetic trade-offs among different fitness-related traits, only a few studies have targeted genetic correlations across developmental stages. Here, we test whether selection on increased cold tolerance in the adult stage increases cold resistance throughout ontogeny in the butterfly Bicyclus anynana. We used lines selected for decreased chill-coma recovery time and corresponding controls, which had originally been set up from three levels of inbreeding (outbred control, one or two full-sib matings). Four generations after having terminated selection, a response to selection was found in 1-day-old butterflies (the age at which selection took place). Older adults showed a very similar although weaker response. Nevertheless, cold resistance did not increase in either egg, larval or pupal stage in the selection lines but was even lower compared to control lines for eggs and young larvae. These findings suggest a cost of increased adult cold tolerance, presumably reducing resource availability for offspring provisioning and thereby stress tolerance during development, which may substantially affect evolutionary trajectories.

Introduction

At least at evolutionary timescales, all organisms are faced with environmental change such as climatic alterations, new diseases or new predators (Frankham & Kingsolver, 2004; Frankham, 2005). Therefore, the ability to respond to such changes is of crucial importance to ensure a species’ longer-term persistence. This can be achieved either through phenotypic plasticity, that is, quick environmentally induced adjustments within a set genotype, or through genetic adaptation involving changes in allele frequencies (Bijlsma & Loeschcke, 2005; David et al., 2005; Sørensen et al., 2005). Although phenotypic plasticity allows for maximal flexibility, it does not seem to be generally favoured by selection, suggesting that plastic responses involve nontrivial costs (DeWitt et al., 1998; Relyea, 2002; Pigliucci, 2005). In contrast, it has been repeatedly suggested that genetic adaptation may come at reduced costs, thus being favoured in stable environments (DeWitt et al., 1998; Relyea, 2002; Aubret & Shine, 2010). This notion, however, has been challenged by some recent findings (Meyer & Di Giuliuo, 2003; Bourguet et al., 2004).

Perhaps the most obvious limit to genetic adaptation is negative genetic correlations (caused by antagonistic pleiotropy), in which a beneficial genetic change in one trait is linked to a detrimental change in another (Stearns, 1989; Roff, 2002). Over recent decades, our knowledge on genetic trade-offs has much increased, especially due to the use of artificial selection experiments, representing the most straightforward approach (e.g. Bell & Koufopanou, 1986; Brakefield, 2003; Czesak & Fox, 2003; Bubliy & Loeschcke, 2005; Fischer et al., 2006Bauerfeind & Fischer, 2007; Bertoli et al., 2010). For instance, artificial selection in Drosophila melanogaster revealed a strong trade-off between larval development time and adult weight (Nunney, 1996) and between cold tolerance and starvation resistance (Hoffmann et al., 2005). In general, there is agreement that such negative genetic correlations between traits related to fitness have the potential to maintain genetic variation within populations and more importantly to bias or constrain evolutionary change (Cheverud, 1984; Roff, 2002).

Although such genetic trade-offs have been quite frequently measured among different traits, hardly any data are available on potential trade-offs across developmental stages (but, for example, Tucic, 1979; Cheverud et al., 1983; Loeschcke & Krebs, 1996). Throughout ontogeny, though, selection at any age is expected to result in correlated responses in all other life history stages. Such genetic influences may on principle yield (i) consistent effects throughout development, (ii) divergent patterns across developmental stages or (iii) phenotypic effects in one or two consecutive ages only (Cheverud et al., 1983).

Against this background, we here investigate genetic links between the expression of a specific phenotype at different time points in the ontogeny of the butterfly Bicyclus anynana. To this end, we used lines artificially selected for an ecologically important trait (Kristensen et al., 2011), increased cold tolerance in the adult stage and corresponding control lines (Dierks et al., 2012a,b). We predict that increased cold tolerance should prevail throughout development (i.e. also in the egg, larval and pupal stages; cf. Tucic, 1979). This prediction rests on the straightforward assumption of genetically positively correlated gene expression, which should be beneficial throughout ontogeny if cold tolerance is selected for in the adult stage. Note in this context that at least some studies indicate that the costs involved in thermal adaptation may be minor (Anderson et al., 2005; Bertoli et al., 2010; Dierks et al., 2012b). On the other hand, thermal sensitivity and tolerance may change during ontogeny, as morphology and physiology will undergo fundamental changes (Chen et al., 1987; Block et al., 1990; Tsutsayeva & Sevryukova, 2001; Jensen et al., 2007). For instance, Krebs & Loeschcke (1997) found that pupae of Drosophila buzzatii were most heat resistant, followed by eggs, first- and finally third-instar larvae. Furthermore, using selection for increased heat-shock resistance in adult and larval D. buzzatii, Loeschcke & Krebs (1996) found no evidence for correlated responses across developmental stages.

The selection lines used here were initially set up from different levels of inbreeding (F = 0.00/0.25/0.375), to investigate effects of selection regime and inbreeding on the response to selection (Dierks et al., 2012b). Although inbreeding effects are beyond the scope of this study and although lines went through several generations of outbreeding prior to our current experiments, we nevertheless also report the respective results here for completeness.

Material and methods

Study organism

We used the butterfly B. anynana Butler, 1879 (Lepidoptera, Nymphalidae), a species that is distributed in the tropical areas of Southern Africa up to Ethiopia (Larsen, 1991), for this study. Adults feed on a diversity of fallen and decaying fruits (Larsen, 1991; Brakefield, 1997). This species exhibits two seasonal morphs as an adaptation to the wet and dry season in its natural environment, and the associated changes in resting background and predation (Brakefield, 1997; Lyytinen et al., 2004; Bauerfeind & Fischer, 2007). Due to its amenability to experimental manipulations, B. anynana has been used frequently in studies on life history evolution and developmental plasticity before (e.g. Brakefield & Mazzotta, 1995; Van Oosterhout et al., 2000; Bauerfeind et al., 2009; Fischer et al., 2010). The origins of our laboratory stock population, which has been founded at Greifswald University, Germany, in 2007, lead back to a stock population at Leiden University, The Netherlands. The latter population has been established in 1988 from over 80 gravid females, collected at Nkatha Bay, Malawi. Since then, several hundred individuals are reared in each generation to avoid inbreeding and to maintain high levels of heterozygosity (Van’t Hof et al., 2005; Bauerfeind & Fischer, 2007).

Inbreeding treatments and selection regime

To investigate effects of selection in the adult stage throughout development, we used in total 12 selection lines, which had been selected for increased cold stress tolerance starting from different levels of inbreeding (Dierks et al., 2012a,b; also for further details). First, three levels of inbreeding had been established using a full-sib breeding design: inbreeding 1 (I1) with individuals resulting from matings between full-sibs, inbreeding 2 (I2) resulting from matings between full-sibs in two consecutive generations and outbred controls (C) resulting from random mating. To initiate artificial selection, butterflies within inbreeding levels were pooled across families (120 each). Per inbreeding level, four groups were set up, two for selection on increased cold tolerance and two as unselected controls, resulting in a total of 12 lines (three inbreeding levels × 4). Per generation and line, 40 males and 40 females were selected to found the next generation, being either the most cold-tolerant ones or selected at random (control lines). Selection was applied to chill-coma recovery time, that is, the time needed to regain mobility following cold exposure, on day 1 following adult eclosion. This trait is considered a reliable proxy of climatic cold adaptation and has been used successfully in B. anynana previously (e.g. Geister & Fischer, 2007; Fischer et al., 2010). Selection was continued for 10 generations, yielding highly divergent phenotypes with the lines selected for increased cold tolerance showing an approximately 28% shorter chill-coma recovery time compared with unselected controls (Dierks et al., 2012b). Lines had been kept without selection under standard rearing conditions for four generations prior to this experiment. Several hundred butterflies were reared per line in each generation.

General rearing

Unless otherwise stated, butterflies were reared in a climate room at 27 °C, high relative humidity (70%) and a photoperiod of L12:D12 (24-h light cycle) throughout. Larvae were reared in population cages (50 × 50 × 50 cm) and fed on young maize plants. Adults were kept in cylindrical hanging cages (30 × 38 cm) and provided with banana and moist cotton wool. Cages were arranged in a randomized block design within the climate cabinet to balance potential slight temperature and humidity variation (Bauerfeind & Fischer, 2007; Fischer et al., 2010; Dierks et al., 2012a).

Experimental design

The effects of (inbreeding and) selection on increased cold tolerance in the adult stage were tested in various developmental stages, specifically in (i) eggs (tested 1 day after oviposition), (ii) second- and fourth-instar larvae, (iii) 1- and 5-day-old pupae and (iv) one- (control) and 7-day-old adults. As proxies of cold tolerance, we used chill-coma recovery time for adults and survival rates for the other developmental stages. Adequate cold stress conditions were selected based on pilot experiments on each developmental stage (data not shown), aiming at survival rates of approximately 50%. Throughout, each individual was tested only once. Eggs for the experiments outlined below were collected from 100 males and 100 females per line, which were allowed to mate randomly.

Experiment 1: Egg stage

Two hundred eggs per line were collected, being randomly distributed over 10 replicate petri dishes (20 eggs each) containing moistened filter paper. The petri dishes were exposed to −1 °C for 90 min on the day following oviposition and were afterwards returned to 27 °C. Unsuccessful (dead) eggs were counted 6 days after the cold shocks had been performed (note that egg development is approximately 4 days at 27 °C). Throughout, petri dishes containing eggs from the different lines were arranged in a randomized block design.

Experiment 2: Larval stage

For measuring survival rates of larvae, second- and fourth-instar larvae were placed individually in translucent plastic cups (125 mL) and were exposed for 9 h to −5 °C in a climate cabinet (Sanyo MIR-553, Sanyo Biomedical, Mariguchi, Japan). Twenty-four hours after their return to 27 °C, during which larvae had access to a fresh leaf of maize for feeding, surviving larvae were counted. Sixty seconds- and 50 fourth-instar larvae were tested per line.

Experiment 3: Pupal stage

One or 5 days after pupation, pupae were individually placed in translucent plastic cups (125 mL) and were exposed for 2 h to −3.5 °C in a climate cabinet (Sanyo MIR-553). After their return to 27 °C, cups were checked daily for eclosed butterflies. We scored the number of successfully eclosed individuals. We tested 28–39 male and 31–37 female 1-day-old pupae and 30–48 male and 30–48 female 5-day-old pupae per line.

Experiment 4: Adult stage

We used chill-coma recovery time for measuring cold tolerance of butterflies. This method produces highly repeatable results, has been used successfully in B. anynana previously and is considered a reliable proxy for climatic cold adaptation (e.g. Geister & Fischer, 2007; Fischer et al., 2010). Unmated 1- and 7-day-old butterflies were placed individually in translucent plastic cups (125 mL) and arranged on a tray in a randomized block design (maximum of 72 butterflies per block). We exposed them for 19 h to 1 °C in a climate cabinet (Sanyo MIR-553) to induce a chill coma. Afterwards, they were transferred to a room with a constant temperature of 20 °C to determine recovery times. Recovery time was defined as the time elapsed between taking the trays out of the 1 °C climate cabinet until a butterfly was able to stand on its legs (Geister & Fischer, 2007; Fischer et al., 2010). Observation time was restricted to a maximum of 60 min. Butterflies that did not recover within this time span were given the maximum recovery time of 60 min (if still alive) or were excluded from further analyses (if dead; typically < 1%). We tested 30–45 males and 30–40 females per line and treatment.

Statistical analyses

Survival rates of eggs, larvae and pupae were analysed using nominal logistic regressions on binary data (dead or alive). Selection regime, inbreeding level and sex (only for pupae) were used as fixed factors and replicate (nested within selection regime and inbreeding level) as random factor. Similar general linear models were applied to test for differences in chill-coma recovery time in the adult stage. Significant differences between groups were determined with the Tukey’s HSD post hoc test. Throughout, means are given ± 1 SE. All statistical tests were performed using jmp (4.0.0; SAS Institute, Cary, NC, USA) or statistica (6.1; StatSoft Inc., Tulsa, OK, USA).

Results

Survival rates

Selection regime significantly affected survival rates of eggs and tended to affect survival rates of young larvae, but did not influence survival rates of pupae or older larvae (Table 1; Fig. 1). Contrary to expectations, though, survival rates after cold exposure of eggs (control lines: 57.3 ± 2.1%; cold-tolerant lines: 43.7 ± 2.3%) and young larvae (control lines: 58.6 ± 2.0%; cold-tolerant lines: 53.9 ± 4.4%) were higher in the control lines than in the cold-tolerant lines. Inbreeding level significantly affected survival rates of eggs, young larvae and young pupae, but not of old larvae and old pupae. Egg survival rate was highest in the I1 followed by the I2 and finally the C treatment (I1: 56.9 ± 2.4%; I2: 53.9 ± 2.8%; C: 40.8 ± 2.9%). Like in eggs, young larvae survival was highest in the I1 followed by the I2 and finally the C treatment (I1: 59.6 ± 2.8%; I2: 57.5 ± 2.6%; C: 51.7 ± 6.3%). Pupal survival was also highest in the I1, followed by the C and finally the I2 treatment (I1: 76.0 ± 4.4%; C: 71.0 ± 3.0%; I2: 59.4 ± 4.5%). Variation between replicate lines was significant in eggs, 2nd-instar larvae, 4th-instar larvae and 5-day-old pupae. The significant interaction between selection regime and inbreeding level for 4th-instar larvae reflects that negative effects of selection were confined to the I1 and I2 treatments, whereas an opposite pattern was found in the outbred controls (Fig. 1c). Additionally, a significant interaction between inbreeding level and sex for 1-day-old pupae indicates that survival rates of males compared to females were higher in the I1 and I2, but lower in C treatment (I1: 82.6% > 68.9%; I2: 63.4% > 55.3%; C: 67.9% < 74.2%).

Table 1.   Nominal logistic regressions for the effects of selection regime and inbreeding level on survival rates after cold exposure in Bicyclus anynana. Throughout, replicate line (random factor) was nested within selection regime and inbreeding level. Sex was added as factor in the analyses of pupal survival rates only (note that sex determination of eggs and larvae was not possible).
Developmental stageFactord.f.Chi-squaredP
  1. Significant P-values are given in bold.

EggsSelection regime126.35< 0.0001
Inbreeding level241.36< 0.0001
Replicate [Sel., Inbr.]612.840.0457
Sel. × Inbr.25.470.0649
Larvae (2nd instar)Selection regime13.650.0561
Inbreeding level29.910.0070
Replicate [Sel., Inbr.]612.700.0481
Sel. × Inbr.23.890.1429
Larvae (4th instar)Selection regime11.880.1705
Inbreeding level20.580.7502
Replicate [Sel., Inbr.]615.910.0142
Sel. × Inbr.210.230.0060
Pupae (day 1)Selection regime10.010.9452
Inbreeding level27.680.0215
Sex13.210.0734
Replicate [Sel., Inbr.]611.840.0656
Sel. × Inbr.24.470.1077
Sel. × Sex10.000.9820
Inbr. × Sex28.610.0135
Sel. × Inbr. × Sex22.080.3534
Pupae (day 5)Selection regime10.140.7122
Inbreeding level24.680.0963
Sex10.370.5419
Replicate [Sel., Inbr.]624.590.0004
Sel. × Inbr.22.460.2930
Sel. × Sex10.220.6414
Inbr. × Sex23.730.1553
Sel. × Inbr. × Sex22.710.2580
Figure 1.

 Survival rates of eggs (a), 2nd-instar larvae (b), 4th-instar larvae (c), 1-day-old pupae (d) and 5-day-old pupae (e) in Bicyclus anynana in relation to inbreeding level (C: outbred control; I1: one full-sib mating; I2: two full-sib matings) and selection regime.

Chill-coma recovery time

After four generations without selection, chill-coma recovery time was significantly shorter in 1-day-old cold-tolerant than in 1-day-old control butterflies (by 15.7%; 1152 ± 24 s < 1366 ± 25 s; Table 2). Despite similar effect size (recovery time by 15.0% shorter in cold-tolerant lines than in control lines), there was no significant effect of selection regime in 7-day-old butterflies (1864 ± 42 s = 2191 ± 40 s, note the increase in SEs; Fig. 2). The latter pattern is at least partly caused by males, in which cold-tolerant and control individuals had similar recovery times (cold-tolerant lines 1857 ± 65.1 s = control lines 2042 ± 51.3 s), whilst the difference was still significant in females (cold-tolerant lines 1870 ± 55.0 s; control lines 2336 ± 58.6 s; Tukey’s HSD after anova; significant selection-regime-by-sex interaction). For 1-day-old butterflies, the significant inbreeding-treatment-by-sex interaction indicates that recovery times tended to decrease with increasing inbreeding level in males, but to increase in females (males: C: 1291 ± 44.5 s, I1: 1251 ± 48.2 s, I2: 1181 ± 46.6 s; females: C: 1161 ± 46.3 s, I1: 1209 ± 49.5 s, I2: 1308 ± 48.9 s).

Table 2.   Nested analyses of variance for the effects of selection regime (fixed), inbreeding level (fixed), sex (fixed) and replicate line (random) on cold stress resistance in 1- and 7-day-old Bicyclus anynana butterflies. Throughout, replicate line was nested within selection regime and inbreeding level.
AgeFactorMSd.f.FP
  1. Significant P-values are given in bold.

1 day oldSelection regime49354.41,620.550.0037
Inbreeding level319.52,60.130.8780
Replicate [Sel., Inbr.]2418.96,8891.520.1688
Sex323.91,8890.200.6522
Sel. × Inbr.1694.62,60.710.5301
Sel. × Sex108.61,8890.070.7941
Inbr. × Sex9231.82,8895.800.0032
Sel. × Inbr. × Sex683.22,8890.430.6513
Error1593.0889  
7 days oldSelection regime5401.31,62.500.1648
Inbreeding level1043.52,60.480.6391
Replicate [Sel., Inbr.]2161.66,7531.590.1459
Sex1339.51,7530.990.3206
Sel. × Inbr.89.72,60.040.9596
Sel. × Sex9066.41,7536.690.0099
Inbr. × Sex2783.62,7532.050.1291
Sel. × Inbr. × Sex541.22,7530.400.6710
Error1356.0753  
Figure 2.

 Means (+ 1 SE) for chill-coma recovery time (CCRT) in 1- (a) and 7-day-old (b) Bicyclus anynana butterflies in relation to inbreeding level (C: outbred control; I1: one full-sib mating; I2: two full-sib matings) and selection regime.

Discussion

Response to selection on increased cold tolerance in the adult stage

In line with studies using Drosophila (e.g. Anderson et al., 2005; Mori & Kimura, 2008; Bertoli et al., 2010; Udaka et al., 2010), artificial selection on increased cold tolerance yielded a significant response in B. anynana (Dierks et al., 2012b). When measured directly after the course of selection, however, chill-coma recovery time was approximately 28.9% shorter in the selection as compared to the control lines (Dierks et al., 2012b); however, the difference was only 15.7% in the current study undertaken four generations later (as measured in 1-day-old butterflies, the age at which selection was carried out). This rather rapid convergence towards control line levels indicates that the selected genes and alleles had not yet become fixed in the selection lines and furthermore (for obvious reasons) a lack of a selective pressure maintaining increased cold tolerance under laboratory conditions. Perhaps, there was even selection against increased cold tolerance, which would be expected if significant costs were involved (Marshall & Sinclair, 2010; Duncan et al., 2011; Stoks & De Block, 2011). However, a previous study could not detect any such costs in an array of traits (Dierks et al., 2012b).

In seven- as compared to 1-day-old butterflies, a similar pattern was observed (by 15.0% faster recovery in the selection lines), although the difference between selection and control lines was here significant in females only. The lack of a significant main effect of selection regime thus seems to be caused by a differential response across the sexes in combination with an increase in variance and the low statistical power of anovas involving random factors (Charmantier et al., 2006; Descamps et al., 2008). Note in this context the striking effect of age, with chill-coma recovery time being 61% longer in 7-day-old than in 1-day-old butterflies. As individual performance is expected to decrease with age because of senescence, that is, a decline in physiological functions (Descamps et al., 2008), this effect was expected. Under laboratory conditions, B. anynana typically lives and reproduces for 3–4 weeks (Ferkau & Fischer, 2006; Fischer, 2007), but survival in the field can be expected to be much shorter, averaging at around 10 days in many butterfly species (Brakefield & Reitsma, 1991; Fischer & Fiedler, 2001). Therefore, a difference in 6 days is considered to be ecologically highly relevant. Previous studies on effects of age on chill-coma recovery times though yielded contradictory results, thus challenging the notion of a linear correlation between age and cold tolerance (Fischer et al., 2010; Lalouette et al., 2010). In summary, artificial selection applied in the adult stage clearly increased cold tolerance in this stage. However, does this also confer increased cold tolerance in other developmental stages?

Effects on cold tolerance during development

In our study, we found no evidence for increased cold tolerance in either the egg, larval or pupal stage as a consequence of selection on shorter chill-coma recovery time in the adult stage. Note that we have necessarily measured different traits, namely cold stress survival during egg, larval and pupal development and chill-coma recovery time in the adult stage. However, both traits are clearly related, for which there is direct evidence in B. anynana (Fischer et al., 2010). Although positive effects were thus lacking throughout, increased cold tolerance in the adult stage was even associated with reduced cold stress survival for eggs, with corresponding tendencies being found in young larvae (except in the I2 treatment) and furthermore in all other developmental stages investigated (though not in all groups). Similarly, Watson & Hoffmann (1996) found a negative effect of selection for increased cold tolerance in Drosophila, with selected lines having a lower fecundity compared to control lines. In our study, increased cold tolerance in the adult stage seems to come at a cost in terms of a reduced cold tolerance during development.

In another study using Drosophila, in contrast, positive correlated responses to selection on increased cold tolerance were found, with an increase in cold tolerance across developmental stages after 52 generations of selection (Tucic, 1979). Note though that in this study, selection was applied to all stages (eggs, larvae, pupae and adults) using predefined survival rates. Effects were strongest in the adult stage and moreover more pronounced in the stages closer to the ones in which selection took place (Tucic, 1979). Against this background, our results seem counter-intuitive, which might be caused by (i) the involvement of different mechanisms in cold stress survival vs. chill-coma recovery time, although both traits are related (Hoffmann et al., 2002; Anderson et al., 2005; Fischer et al., 2010), (ii) the involvement of different mechanisms facilitating cold tolerance in the adult stage and during development (note that Gilchrist et al., 1997 revealed clear effects of thermal experimental evolution in the adult but not in the egg stage), and (iii) the presence of resource allocation trade-offs with an enhanced investment into adult cold tolerance leaving less resources available for offspring provisioning.

Distinguishing between these possibilities will be an important task for future research. However, based on the (at least partly) negative effects of increased adult cold tolerance on offspring performance during development, being most pronounced in early developmental stages, we favour for the time being a resource-based explanation. A wealth of studies indicates that allocation trade-offs, with energy-demanding functions competing for limited energy resources, play an important role in life history evolution (Stearns, 1989, 1992; Roff, 2002, 2007; Roff & Fairbairn, 2007). Consequently, resources allocated to increased cold tolerance in the adult stage may not be available anymore for offspring provisioning, which may in turn diminish subsequent offspring performance. Such effects should be most pronounced in early developmental stages, which is indeed the case here.

Effects of inbreeding

As mentioned above, results on inbreeding are included here because of the initial set-up of the selection experiment, which had included effects of inbreeding on the response to selection (Dierks et al., 2012b). After 14 generations of random mating, we found significant differences between inbreeding groups in the survival rates of eggs, 2nd-instar larvae and 1-day-old pupae. Whereas in pupae survival rates were significantly lower in the I2 compared to the other treatments, egg and 2nd-instar larvae survival was lower in the control compared to the inbred groups. Whether these differences reflect effects of initial inbreeding though is questionable, given the rather long period of random mating, which should on principle set off any inbreeding effects (see also Phillips et al., 2001). Therefore, it seems more likely that the above differences result from random genetic drift within lines.

Note in this context that inbreeding yielded pronounced negative effects on, for example, egg hatching success immediately after full-sib matings (Dierks et al., 2012a), but that such effects were not detectable anymore after the course of selection (Dierks et al., 2012b). Additionally, nonadditive effects may play a major role in inbreeding depression of B. anynana (Dierks et al., 2012b).

Conclusions

Our study shows that, despite a significant response to selection on increased cold tolerance in the adult stage, there was no correlated positive response in other developmental stages. In contrast, increased adult cold tolerance tended to decrease cold tolerance during development, especially so during early development. This finding suggests a cost of increased adult cold tolerance, presumably reducing resource availability for offspring provisioning and stress tolerance in other developmental stages. Our results thus highlight that increased performance in a specific trait may be traded off not only against other energy-demanding functions but also against performance throughout ontogeny, which may substantially affect optimal responses to selection pressures (Marshall & Sinclair 2010; Stoks & De Block, 2011).

Acknowledgments

We thank Christin Michalowsky for help with the experiment and Wolf Blanckenhorn and Jarrod Hadfield for valuable comments on an earlier version of this manuscript.

Data deposited at Dryad: doi: 10.5061/dryad.65r9c

Ancillary