SEARCH

SEARCH BY CITATION

Keywords:

  • cold;
  • Drosophila;
  • starvation;
  • stress resistance;
  • trade-off

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

In insects changes in lipid metabolism may underlie a trade-off between cold resistance and starvation resistance. To test this we examined correlated responses in independent sets of Drosophila melanogaster lines selected for increased cold resistance and increased starvation resistance. The starvation lines showed correlated patterns found in other D. melanogaster populations selected for this trait, including higher lipid levels and increased resistance to desiccation, although the selected lines did not show a longer development time as found in some other studies. Consistent with the trade-off hypothesis, selected lines with increased starvation resistance showed decreased resistance to a cold stress as measured by mortality, whereas selected lines with increased cold resistance showed a decrease in starvation resistance. To counter the possibility of inadvertent selection accounting for these patterns, selected and control lines from both selection regimes were crossed to form mass bred populations, which were left for four generations prior to establishing isofemale lines. By scoring starvation and cold resistance in these lines derived from both sets of selection regimes, we confirmed the negative association between resistance to these stresses in females but not in males. Potential implications of this trade-off for surviving cold conditions when food resources are limiting are discussed.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

The evolution of resistance to starvation stress has received a lot of attention in the last few years because of the apparent association between this trait and longevity. Drosophila lines with increased starvation resistance can have increased longevity (Hoffmann & Harshman, 1999) although this is not always the case (Harshman et al., 1999a). In addition, mutants of Caenorhabditis elegans and Drosophila that have increased levels of starvation resistance often also show increased longevity (Lin et al., 1998, 2001; Clancy et al., 2001).

Less attention has been given to interactions between starvation resistance and other traits. Lines with increased starvation resistance may show an increase in larval development times (Harshman et al., 1999b). There may also be a decrease in early fecundity (Hoffmann & Parsons, 1989a; Leroi et al., 1994). However correlated responses are not always consistent, perhaps because multiple mechanisms underlie starvation resistance, and specific environmental conditions can make selection responses via some mechanisms more likely (reviewed in Hoffmann & Harshman, 1999). When making generalizations about evolutionary interactions among traits, it is important to understand the specific mechanisms involved as well as the genes underlying the mechanisms (Zera & Harshman, 2001).

A survey of P element insertion lines of D. melanogaster for starvation resistance (Harbison et al., 2004) has suggested that a large number of genes can influence this trait, including genes involved in reproduction, metabolism and larval feeding rate. Mutant studies have linked starvation resistance to the insulin signalling pathway in both Drosophila (Clancy et al., 2001) and C. elegans (Munoz & Riddle, 2003) as well as to the storage of lipids (Hader et al., 2003). Physiological studies indicate that lipid levels often underlie starvation resistance. This includes evidence that plastic changes in starvation resistance are closely correlated to levels of lipid carried by the flies (Service, 1987), and that lipid levels increase following selection (Chippendale et al., 1996; Harshman et al., 1999b). Associations between lipid and starvation resistance may extend to the interspecific level (Van Herrewege & David, 1997) although not in all comparisons (Bharathi et al., 2003). Changes in lipid levels following selection for starvation resistance may account for correlated changes in reproduction in flies, because lipid metabolism can form the basis of life-history trade-offs involving reproductive output (Zera & Zhao, 2003).

Although selection experiments suggest that resistance levels to different environmental stresses are often positively correlated (Hoffmann & Parsons, 1989b; Harshman et al., 1999b), this may not apply to starvation resistance and cold resistance. In Drosophila, changes in phospholipid composition, triacylglycerol accumulation and proline accumulation can play a role in resistance to cold temperatures (Chen & Walker, 1994; Misener et al., 2001). There might be a trade-off between cold resistance and starvation resistance if lipids are shunted towards these mechanisms rather than being accumulated for countering starvation. Such a trade-off would be particularly important from an adaptive point of view because insects facing winter may need to counter cold stress and a reduction in food at the same time.

We are unaware of direct tests for such an association by examining correlated selection responses. We therefore tested for a trade-off between starvation resistance and cold resistance by selecting for increased starvation resistance in replicated lines of Australian D. melanogaster and scoring correlated responses in cold resistance. The correlated responses of other traits to selection for starvation resistance in these lines are also described. To further test the association between the resistance traits, we scored starvation resistance in lines from a different Australian base population selected for decreased recovery times following exposure to cold temperatures leading to a coma (Anderson et al., 2004). These selected lines also showed decreased mortality following a more severe cold shock. We undertook crosses between selected and control lines in these experiments and allowed several generations of assortment to occur before deriving isofemale lines to further test for a negative association between the two resistance traits in both sexes.

Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

Starvation selection

Isofemale lines collected from 10 sites along the east coast of Australia in February 2000 were combined to generate the mass bred stock used in this selection experiment. The stocks were cultured at 25 °C, 24 h light on a medium comprised of sugar (5.4% w/v), dried yeast (3.6% w/v) and agar (1.8% w/v). The medium was treated with the preservatives nipagin (1.4% v/v), tartaric acid (0.8% w/v) and propionic acid (1.1% v/v). A solution of antibiotics (2% streptomycin and 0.6% penicillin G) was also added to the surface of the medium to counter bacterial growth. Stocks were reared in 250 mL glass bottles with 40 mL of the medium.

Selection began after the mass bred population had been cultured for 40 generations. Three replicate-selected lines and three control lines were maintained. The selection process involved separating 3–6-day-old individuals into groups of 50 (sexes mixed) using CO2 anaesthesia. These were then allowed to recover for 24 h before being placed in an empty vial sealed with gauze and inverted over a second vial containing cotton wool soaked with distilled water. The coupled vials were sealed together with Parafilm®(Pechiney Plastics, Neenah, WI, USA) and left until 90% mortality was achieved. The remaining individuals were then divided evenly between three culture bottles for each selected line and left for 3–4 days to propagate the next generation. We ensured that at least 100 individuals were transferred per line for each generation. Control lines were continuously maintained on media at a similar population size. This process was repeated for 19 generations of selection.

Correlated responses in starvation lines

To characterize the starvation-selected lines for their selection response and other traits, selection was relaxed for a generation and eggs were obtained from parents for rearing at a standard density of 100 eggs per 250 mL culture bottle. The emerging adults were scored for a number of traits.

To determine the selection response, we set up 15 replicate vials of 10 nonvirgin females for each line and assayed these females for starvation resistance following the selection protocol. Mortality was scored every 8 h at 25–26 °C and continued until all the females had died. We then computed the time taken for half the flies to die (LT50) and 90% of the flies to die (LT90) estimated by linear interpolation.

Desiccation resistance was assessed on 3–4-day-old mated females at 25–26 °C. Fifteen replicates of 10 females per line were stressed in empty vials covered with gauze. These were placed in a sealed tank containing silica gel generating a relative humidity of less than 10%. Mortality was scored every hour until all the flies had died, and LT50s were determined by linear interpolation.

To score lipid content and dry weight, groups of 20 mated females were isolated after being aged for 8 days, and were then left to recover for 24 h. We set up 10–12 replicates per line and transferred these to empty vials. The vials were placed in an oven at 100 °C for 48 h for drying. After this period females were weighed on a microbalance to obtain dry weight. We then added 15 mL ether to each vial to extract the lipids. The flies were removed from the ether after 24 h and returned to the oven for a further 48 h. Females were then reweighed to estimate lipid content.

The development time of the lines was scored by setting up 30 replicate vials each with 10 eggs collected from a 2 h laying period at 25 °C. Emergence was scored every 6 h beginning 9 days from the time of egg lay.

Finally, we scored cold resistance as a correlated response in both females and males. Offspring were aged for 5 days before being separated into groups of 10 individuals of each sex. Each group was then transferred to an empty vial and submerged in a cooled 10% glycol solution. Females were exposed to −2 °C for 3 h and males were exposed to a more severe stress of −4 °C for 4 h to produce similar mortality levels in the sexes. All replicates were then transferred to new media and scored for mortality after 24 and 48 h, although only the 24 h data are presented because the results were almost identical.

Starvation resistance of cold-selected lines

Selection for chill coma resistance was described in Anderson et al. (2004). Briefly, lines were established from a mass bred population derived by mixing five isofemale lines from each of five populations from northern Queensland, Australia. Lines responded rapidly to an intermittent selection regime for decreased time to recovery once flies had been placed in a comatose state. Females and males from selected lines showed decreased recovery time after exposure to a broad range of low temperatures and also had a lower mortality following a cold shock.

To characterize starvation resistance of females from each of the three control and selected lines, three replicates of 10 females were tested for starvation resistance as described above. Starvation was tested two generations after 10 cycles of cold selection had been completed.

Starvation and cold resistance in matings from line cross populations

For the starvation-selected lines, one of the control lines (CSt-1) and one of the selected lines (SSt-1) were crossed after they had been selected for 20 generations. Reciprocal crosses between these two lines were established from 150 virgin females and males, spread over three culture bottles. Offspring from crosses were combined. Crosses were kept as mass bred populations of several thousand individuals for an additional four generations to allow assortment of genes controlling starvation and cold resistance. Virgin flies were then collected and 50 pairs of males and females mated. Offspring of these crosses were scored for starvation and cold resistance. Flies were aged for 3 days before being sexed and separated into groups of 10 individuals for testing. For starvation resistance, we tested one to three replicate offspring groups of each sex per pair mating. These were left for a day before being assayed for resistance as described above. Mortality was scored every 8 h until all flies had died. To measure cold mortality, one to three replicates of 10 individuals of each sex from each pair mating were placed at −2 °C (females) or −4 °C (males) for 4 h as described above. Survival was scored after 24 h recovery on food medium at 25 °C.

For the chill coma-selected lines, a control line (CCo-2: defined as C2 in Anderson et al., 2004) and a selected line (SCo-2: S2) were crossed reciprocally by placing 50 virgin females and 50 virgin males in each of five bottles. This population was maintained for four generations before single pair matings were set up as described above for the starvation line crosses (50 single pair matings per cross). F1 males and females from the pair matings were scored for chill coma recovery, cold mortality and starvation resistance. Because the lines had been selected for chill coma recovery, this trait was scored as well as mortality following cold shock. For the chill coma recovery assay, 15 mL glass vials containing 10 males or 10 females (4 days old, nonvirgin) were placed at 0 °C for 4 h and returned to 25 °C for recovery. At 2-min intervals the number of flies recovered in each vial was recorded. We set up two replicate vials per sex for each pair mating. Cold mortality and starvation resistance was scored as described above for the starvation cross lines (one to three replicates for each pair mating).

Analysis

Prior to analysis, proportions (for lipid levels, mortality following cold shock) were arcsin transformed. Selected and control lines were compared with nested anovas, replicate lines being nested within selection regime. Correlation coefficients were computed to test for associations among trait means for families established from populations derived following line crosses. Correlation coefficients computed between starvation resistance and cold resistance (measured as either the proportion that died or chill coma recovery) were multiplied by −1 so that negative coefficients reflected increased resistance to a stress being correlated with decreased resistance to the other stress.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

Correlated responses in starvation lines

Artificial selection for starvation resistance increased the resistance of females in all lines as indicated by the nested anovas on LT50s and LT90s (Table 1). All selected lines had higher resistance than the controls (Fig. 1a), particularly for the LT90s which differed by almost 30 h.

Table 1.  Nested anovas comparing effects of selection and lines nested within selection or control regimes on stress resistance and other traits.
TraitSelection (d.f. = 1)Line (d.f. = 4)Error MS (d.f.)
MSPMSP
Starvation-selected lines
 Starvation (LT50)3359.7880.00286.9490.00722.916 (65)
 Starvation (LT90)8754.903<0.001127.9290.09762.351 (65)
 Dry weight (×109)15.9300.2549.179<0.0010.175 (57)
 Lipid proportion (×104)550.013<0.0011.9990.7464.116 (57)
 Desiccation (LT50)429.7640.0028.1320.0983.998 (76)
 Development time47.6310.576128.4030.00228.568 (168)
 Mortality (females)1.4340.0280.1260.0210.036 (24)
 Mortality (males)0.1870.0480.0240.6320.038 (30)
Cold-selected lines
 Starvation resistance769.5330.01240.8410.36735.683 (18)
 Desiccation resistance15.1700.57340.3100.46242.793 (18)
image

Figure 1. Selection responses and correlated responses in females following selection for starvation resistance. Starvation resistance is expressed as the time taken for 50% (LT50) of the flies to die (left bars in pairs) or 90% (LT90) of the flies to die (right bars in pairs). Desiccation resistance was expressed as LT50, and lipid content as the proportion of fly weight consisting of ether-extractable lipid. Hatched bars are selected lines, open bars are controls, and error bars are standard errors.

Download figure to PowerPoint

The desiccation resistance of females showed a correlated response to starvation selection as there was a significant difference in the nested anova (Table 1). Females from all three starvation-selected lines were more resistant to desiccation than the control lines, with a difference in LT50 of around 4 h (Fig. 1b).

The dry weight of females did not differ between the selected and control lines although there were significant differences among the replicate lines (Table 1). Lipid levels differed significantly between the two sets of lines (Table 1). The lipid content of the selected line females was relatively higher than that of the controls, reflecting a difference of about 4% (Fig. 1c). For development time, there was no significant difference between the two sets of lines although there was a difference between the replicate lines (Table 1).

Cold resistance as scored by mortality showed a significant correlated response to selection for both sexes (Table 1). Mortality levels in males and females from the selected lines were higher than in flies from the control lines (Fig. 2), suggesting that selection for starvation resistance had decreased cold resistance.

image

Figure 2. Response of starvation-selected lines to cold stress as measured by mortality following a cold shock. Error bars are standard errors.

Download figure to PowerPoint

Starvation resistance of cold-selected lines

A nested anova indicated that females from selected lines differed from control line females for starvation resistance (Table 1). Selected line females had a significantly reduced LT50 for starvation resistance when compared with females from the control lines (Fig. 3). In contrast, desiccation resistance of the females had not been altered by selection (Table 1).

image

Figure 3. Resistance of females from chill coma-selected lines to starvation as scored by the time taken for 50% of the flies to die (LT50). Error bars are standard errors.

Download figure to PowerPoint

Starvation and cold resistance in matings from line cross populations

For crosses derived from the starvation lines, there was a negative correlation between starvation resistance and cold resistance in the females but not in the males (Table 2). Resistance to starvation was correlated between the sexes, but the cold resistance of the two sexes was uncorrelated. Male starvation resistance also correlated negatively with female cold resistance. These results suggest that the sexes shared a common genetic basis for starvation resistance that is negatively associated with female cold resistance.

Table 2.  Correlations among means of strains derived from crosses between selected and control lines. Means above diagonal are for lines selected for chill coma resistance, and those below diagonal are for starvation-selected lines.
 Female cold mortality (d.f.)Female starvation (d.f.)Male chill coma (d.f.)Male cold mortality (d.f.)Male starvation (d.f.)
  1. *P < 0.01; **P < 0.01; ***P < 0.001.

Female chill coma0.503** (76)−0.272* (81)0.496** (83)0.007 (56)−0.286* (71)
Female cold mortality1−0.226* (81)0.394** (79)−0.083 (57)−0.107 (75)
Female starvation−0.347* (82)10.186 (85)0.122 (62)0.424** (81)
Male chill coma  1−0.108 (70)−0.131 (87)
Male cold mortality0.088 (66)0.036 (66) 10.139 (64)
Male starvation−0.246* (66)0.444*** (66) 0.025 (76)1

For the pair matings derived from crosses between the chill-coma lines, the female data indicated a significant negative correlation between starvation resistance and both the chill coma recovery and cold mortality measures of cold resistance (Table 2). However, for the males there was no correlation between starvation resistance and either of the cold resistance traits. For the females, the two cold resistance traits were significantly correlated with each other, but this was not the case for the males. The starvation resistance of the two sexes was correlated, and male chill coma resistance (but not cold mortality resistance) was correlated with both measures of female cold resistance. There was also a negative correlation between female cold resistance and male starvation resistance. These results are consistent with correlations obtained from crosses of the starvation-selected lines, and they indicate a trade-off between cold resistance and starvation resistance in one of the sexes.

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

We have found evidence for a trade-off between starvation resistance and cold resistance in females from two sets of lines selected independently for different traits. This type of comparison where independent selection lines produce convergent results is fairly uncommon. In the context of stress resistance, selection for desiccation resistance produced a correlated increased starvation resistance (Hoffmann & Parsons, 1989a) that matched the correlated response in desiccation resistance when starvation was selected on a different base stock (Harshman et al., 1999b). Selection for starvation resistance also produced changes in reproductive patterns that matched correlated changes in starvation resistance when late reproduction was selected (Leroi et al., 1994; Rose & Archer, 1996). However, patterns of correlated responses are often specific to stocks (Hoffmann & Cohan, 1987), and even change in the same set of lines during the course of long-term selection (Archer et al., 2003). For instance, the lack of a correlated response in development time in the starvation lines described here contrasts with data from other starvation-selected lines (Chippendale et al., 1996).

The trade-off between starvation and cold resistance was confirmed by crossing the selected and control lines and allowing four generations of assortment before establishing families to test for trait associations. This approach was previously used to establish associations between different stress resistance traits in D. melanogaster (Hoffmann & Parsons, 1989b). It counters the possibility that correlated selection responses have arisen from inadvertent selection rather than genetic correlated responses due to close linkage or pleiotropy. Inadvertent selection has particularly been a problem in establishing patterns of correlations in life-history experiments (Partridge et al., 1999). Spurious correlated responses can also arise when inbreeding in lines causes the accumulation of deleterious genes that influence many traits. However, these are likely to lead to positive genetic correlations among stress resistance traits rather than a trade-off, as alleles that are generally deleterious are unlikely to increase resistance to one stress but decrease resistance to another stress.

The results suggest that the trade-off between starvation and cold resistance is sex-specific. The correlations among family means derived from both sets of selection line crosses indicated that starvation resistance was negatively correlated with cold resistance in females but not in males. This may be related to the sex-specific expression of family differences for cold resistance rather than starvation resistance; starvation resistance of the sexes was positively correlated, whereas for one measure of cold resistance there was no correlation between the sexes. Moreover, the negative correlation between male starvation resistance and female cold resistance suggests that genes controlling starvation resistance in males as well as females influenced female cold resistance. However one of the results – that selection for starvation resistance led to a correlated response in both male and female cold resistance as measured by mortality – is inconsistent with this simple sex-specific expression hypothesis.

We are a long way from understanding the exact mechanistic basis of this trade-off. Quantitative trait loci (QTL) mapping points to several genes that might influence starvation resistance by being involved in lipid allocation (Harbison et al., 2004). Moreover, physiological studies on mutants indicate that increased starvation resistance can be associated with metabolic changes leading to the accumulation of lipids (Hader et al., 2003) and hormonal induction of carbohydrate and lipid mobilization as well as feeding behaviour (Lee & Park, 2004). It is not known if these mutants or QTLs also influence susceptibility to cold stress. Starvation resistance was associated with an increased proportion of a fly's body weight being associated with lipids, but we have no information on the proportion of a fly's body that consists of phospholipids as opposed to triglycerides; it is the relative proportion of these different types of lipids that is likely to underlie a trade-off between membrane function to improve cold resistance and storage to increase starvation resistance.

The results may have implications for adaptive shifts in these traits as well as geographical patterns. Although negative genetic correlations among traits are not essential for trait interactions to act as selection limits (Pease & Bull, 1988), they can facilitate such limits. In some species, starvation resistance is relatively high in tropical populations but lower in temperate areas (Karan et al., 1998). The lower starvation resistance in temperate areas could reflect selection for increased cold resistance, which can also vary clinally (Hoffmann et al., 2002). The results also raise the issue of whether acclimation procedures that increase cold resistance decrease starvation resistance, and whether such plastic changes as well as a genetic trade-off between these traits might limit range expansion of species. Given that many organisms are likely to have problems in finding food in winter as well as countering stressful cold conditions, the trade-off might contribute to range limits in invertebrate species. However there is only a moderate negative correlation between these traits; ideally selection experiments for both increased starvation and cold resistance simultaneously would be needed to test if the interaction between these traits can influences selection responses.

Acknowledgments

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

This research was supported by the Australian Research Council via their Special Research Grant Program.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References