Correlated responses to selection for stress resistance and longevity in a laboratory population of Drosophila melanogaster


Volker Loeschcke, Department of Ecology and Genetics, University of Aarhus, Ny Munkegade, Building 540, DK-8000 Aarhus C, Denmark.
Tel.: +45-89423268; fax: +45-89422722; e-mail:


Laboratory studies on Drosophila have revealed that resistance to one environmental stress often correlates with resistance to other stresses. There is also evidence on genetic correlations between stress resistance, longevity and other fitness-related traits. The present work investigates these associations using artificial selection in Drosophila melanogaster. Adult flies were selected for increased survival after severe cold, heat, desiccation and starvation stresses as well as increased heat-knockdown time and lifespan (CS, HS, DS, SS, KS and LS line sets, respectively). The number of selection generations was 11 for LS, 27 for SS and 21 for other lines, with selection intensity being around 0.80. For each set of lines, the five stress-resistance parameters mentioned above as well as longevity (in a nonstressful environment) were estimated. In addition, preadult developmental time, early age productivity and thorax length were examined in all lines reared under nonstressful conditions. Comparing the selection lines with unselected control revealed clear-cut direct selection responses for the stress-resistance traits. Starvation resistance increased as correlated response in all sets of selection lines, with the exception of HS. Positive correlated responses were also found for survival after cold shock (HS and DS) and heat shock (KS and DS). With regard to values of resistance across different stress assays, the HS and KS lines were most similar. The resistance values of the SS lines were close to those of the LS lines and tended to be the lowest among all selection lines. Developmental time was extended in the SS and KS lines, whereas the LS lines showed a reduction in thorax length. The results indicate a possibility of different multiple-stress-resistance mechanisms for the examined traits and fitness costs associated with stress resistance and longevity.


Environmental abiotic stresses play an important role as factors determining the distribution and abundance of species. Although a number of adaptive features can improve survival under specific stressful conditions, evolutionary physiologists and geneticists always sought for universal mechanisms that promote resistance to multiple forms of stress (e.g. Hoffmann & Parsons, 1991). Laboratory studies on Drosophila have been widely used to accumulate experimental evidence, formulate and test for different hypotheses on the physiological basis of such mechanisms in insects.

One of these hypotheses suggested that lowered metabolic rate may be a general explanation for stress resistance because reduction in metabolism conserves resources and minimizes exposure to stress. It was advanced by Hoffmann & Parsons (1989a,b), who found a decrease in mass-specific metabolic rate and activity levels in lines of Drosophila melanogaster selected for desiccation resistance. The selected lines also showed enhanced resistance to knockdown by heat, starvation, gamma radiation, toxic concentrations of ethanol and acetic acid. However, the association between reduced metabolic rate and stress resistance did not receive strong experimental support in studies of laboratory Drosophila populations (e.g. Djawdan et al., 1997; Loeschcke et al., 1997; Harshman & Schmid, 1998). On the other hand, an increased level of lipid and carbohydrate was frequently observed in strains selected for desiccation and starvation resistance (for a review, see Hoffmann & Harshman, 1999) causing some authors (Service, 1987; Graves et al., 1992; Djawdan et al., 1997) to suggest that the accumulation of metabolic energy reserves underlies multiple-stress resistance. Another widely recognized candidate mechanism for such resistance is heat-shock proteins (HSPS), which are induced by a variety of stresses (for a review, see e.g. Feder et al., 1995; Sørensen et al., 2003). Many HSPS are known to act as molecular chaperones involved with folding, trafficking, protection and renaturation of cellular proteins (e.g. Hartl & Hayer-Hartl, 2002; Walter & Buchner, 2002). The focus of Drosophila studies was on the role of HSPS in resistance to thermal stresses. There is evidence that expression of the major inducible Drosophila heat-shock protein, Hsp70, is associated with increased resistance to high temperatures (Dahlgaard et al., 1998; Lansing et al., 2000; Norry & Loeschcke, 2003) and its extra copies improve protection against heat stress in transgenic individuals (Krebs et al., 1998). In addition, it was shown that cold-stressed larvae and adults may express Hsp70 during the recovery period (Burton et al., 1988; Goto & Kimura, 1998; Sejerkilde et al., 2003).

It has been also hypothesized that a generalized stress resistance is a prerequisite for the expression of the extended longevity phenotype and changes in lifespan can be regarded as mainly secondary to primary selection for stress resistance (Parsons, 1995). This suggestion, known as the stress theory of ageing, seems not to be supported by current experimental evidence in Drosophila. Strains selected for stress resistance are not necessary long-lived (Harshman et al., 1999b; Norry & Loeschcke, 2002a) and long-term laboratory selection for increasing stress resistance can lead to a breakdown in positive correlation between resistance and longevity (Archer et al., 2003; Phelan et al., 2003). Moreover, there is a large amount of data demonstrating that longevity may be specifically associated with resistance to oxidative stress due to a common physiological basis involving antioxidant defense system (ADS) enzymes (e.g. Arking et al., 2000a,b; Le Bourg, 2001). Nevertheless, results of many Drosophila studies indicate that resistance and longevity are genetically and physiologically linked. Selection experiments revealed a positive association between various stress-resistance traits and lifespan (Service et al., 1985; Rose et al., 1992; Force et al., 1995; Zwaan et al., 1995; Chippindale et al., 1998; Norry & Loeschcke, 2003). Long-lived strains had an increased level of metabolic reserves, such as lipid (Rose, 1984; Service, 1987; Djawdan et al., 1998) and glycogen (Graves et al., 1992; Force et al., 1995) and the beneficial effects of small HSPS on aging were shown in mild stress and overexpression experiments (reviewed by Morrow & Tanguay, 2003). A decrease in metabolic rate has also been reported in some Drosophila strains selected for a longer lifespan (Service, 1987; Simmons & Bradley, 1997) and in desiccation-selected lines with extended longevity (Hoffmann & Parsons, 1993b). However, this was not found in other comparisons between long- and normal-lived strains (Arking et al., 1988; Djawdan et al., 1996) and in recombinant inbred lines with different longevity values (Van Voorhies et al., 2003).

The main aim of the present study was to test for genetic associations between stress-resistance traits and between stress resistance and longevity by comparing correlated responses to artificial selection in D. melanogaster. Such associations may reflect shared physiological mechanisms and indicate a possibility of joint evolution for the above-mentioned traits. Analysis of correlated responses in laboratory selection experiments on Drosophila has been a traditional tool for testing various hypotheses on stress resistance and longevity (Harshman & Hoffmann, 2000). However, in the majority of studies selection was conducted for only one trait with subsequent estimation of indirect responses in selection lines. To make our analysis more powerful, we simultaneously selected for five resistance traits and longevity (under nonstressful conditions) in the same laboratory population. The resistance traits were increased survival rates in adults after severe cold, heat, desiccation and starvation stresses as well as increased heat-knockdown time. They have been previously studied in selection experiments using Drosophila but multiple-stress resistance was usually analysed in desiccation/starvation-resistant and long-lived strains (for references, see e.g. Harshman et al., 1999b; Hoffmann & Harshman, 1999). Only a few studies (Huey et al., 1992; Hoffmann et al., 1997; Bubli et al., 1998; Norry & Loeschcke, 2002a,2003; Sejerkilde et al., 2003) considered correlated responses for stress resistance and/or lifespan in thermal-selected lines. However, the genetic relationship between cold and heat resistance and between thermal resistance on the one hand and desiccation/starvation resistance and longevity on the other is still poorly investigated.

To test for possible fitness costs associated with stress resistance and longevity we estimated preadult developmental time, early age productivity and thorax length under nonstressful conditions. A trade-off between extended longevity and early age reproduction has been reported in many studies that indicated its possible involvement in senescence evolution by pleiotropy (for reviews, see Zwaan, 1999; Tatar, 2001; Zera & Harshman, 2001). Preadult developmental period has also been widely studied but no consistent evidence on its contribution to lifespan was obtained (for reviews, see e.g. Zwaan, 1999; Partridge et al., 1999). Results of some selection experiments showed that trade-offs may exist between fecundity and stress resistance (Hoffmann & Parsons, 1989a; Watson & Hoffmann, 1996) as well as between stress resistance and developmental rate (Chippindale et al., 1996,1998; Harshman et al., 1999a; but see Hoffmann & Parsons, 1993b). A decrease in body size has been found in some longevity-selected strains (Buck et al., 2000); however, selection for stress resistance was reported to produce larger flies (Chippindale et al., 1994,1996,1998). Though Drosophila body size is not a fitness component itself, it is genetically correlated with many life-history parameters (for references, see e.g. Partridge & Fowler, 1993; Roff, 2000) and can thus be used as a morphological index of fitness.

Materials and methods

Base population

The replicate selection and control lines were derived from a mass laboratory population of D. melanogaster established in September 2002. To increase the genetic diversity of the starting material, this population was created by mixing 600–700 flies from each of four pre-existing laboratory stocks. The latter had been kept in the laboratory as discrete generations using large numbers of breeding individuals. All flies were reared on standard Drosophila mediums under low to moderately high density conditions at 25 °C unless otherwise stated. The stocks were as follows:

Hov–Copenhagen basic strain

The flies were collected in October 1997 from two Danish localities, Hov (Jutland peninsula, east coast) and Hvidovre (Zealand island, near Copenhagen). They were kept as 27 and 30 isofemale lines, respectively, until February 1998. Then the lines were mixed and maintained as one large interbreeding population.

Supermass Hov–Copenhagen population

It was founded in September 2001 by mixing a few sets of heat and longevity selection lines. The heat selection lines were started using flies from the 17th generation of the Hov–Copenhagen basic strain. There were four sets of lines: lines selected for increased survival after heat shock (38.6 °C) with and without prior heat-hardening (37.0 °C), lines which were heat-hardened but not selected and lines kept at cycling temperatures (25 and 35 °C for 18 and 6 h, respectively). Since May 2001, the first three sets were maintained without selection and/or hardening treatment. The longevity selection lines were established in April 2000 also by sampling flies from the Hov–Copenhagen basic strain. They were selected for extended longevity at two temperature regimes, 25 and 29 °C.

Heat-knockdown selection lines

These were a few lines from two sets of highly inbred laboratory lines described by Norry et al. (2004). The first of them (SH) originated from Australian flies collected near Melbourne in February 1994 and selected for increased heat-knockdown resistance. The flies for the second set of lines (D) were sampled from the 10th generation of the Hov–Copenhagen basic strain and selected for reduced heat-knockdown resistance.

Leiden strain

This strain was represented by 30 isofemale lines founded by females collected near Leiden (the Netherlands) in October 1999. For the first five generations it was maintained at 25 °C and then at 20 °C.

The mass population was kept at 25 °C on a standard oatmeal-sugar-yeast-agar Drosophila medium. Every subsequent generation was founded by a mix of parents collected from different bottles. There were 25 bottles in total with ca. 50 pairs of parents per bottle. The experimental lines were established by flies from the fourth generation of the mass population.

Selection regimes

Seven regimes were used with five independent replicate lines per regime. They were as follows:

Unselected control (UC)

These lines were maintained during the course of the whole experiment without selection treatment. After emergence, flies were collected from the culture bottles and aged for 5 days in vials with the standard oatmeal-sugar-yeast-agar medium. Then they were placed in a new set of bottles to propagate the next generation.

Cold-shock resistance selection (CS)

After emergence, flies were aged for 2 days and transferred to a new set of vials containing medium seeded with live yeast. These vials were put in an incubator at 11 °C for another 5 days to acclimate flies. The acclimated individuals were placed in empty vials and chilled at 0.5 °C using an ice-water-filled container placed in a 5 °C cold temperature room. The relative humidity (RH) in the container during the test was close to 100%. The initial chilling period of 27 h was gradually increased to 50 h following changes in cold resistance of the selected lines. The cold-stressed flies were allowed to recover for 24 h at 25 °C in vials with the standard medium; then the survivors were placed in culture bottles for reproduction.

Heat-shock resistance selection (HS)

After emergence, flies were aged for 5 days and transferred to fresh vials with medium to be hardened for 30 min at 36.0 °C in a preheated water bath. The hardened flies were kept for 20 h at 25 °C and then placed in empty vials for the heat-shock test. The test was first carried out by exposure of the flies to 38.0 °C for 1 h in a preheated water bath. Then the temperature was gradually increased to 38.5 °C to maintain the mortality constant as the lines improved their resistance due to selection. The surviving flies were placed in culture bottles after the 24 h recovery period at 25 °C.

Heat-knockdown resistance selection (KS)

After emergence, flies were collected from culture bottles and aged for 5 days before the test. The knockdown time was assessed at 40.0 °C using a preheated knockdown tube (Huey et al., 1992; Sørensen et al., 2001) in a 25 °C temperature room under controlled humidity conditions (RH = 50%). For each run, 300 pairs of flies were placed in the tube. The flies succumbed to heat and rolled down a series of baffles reaching a collecting vial, which was replaced every 30 s. Each individual fly was assigned a knockdown time according to when it fell into the collecting vial. Flies with longer knockdown time were mixed and transferred to culture bottles to start a new generation.

Desiccation resistance selection (DS)

After emergence, flies were aged for 5 days and transferred to empty vials, which were put in a standard exsiccator containing a silica gel (SORBSIL® C, Oker-Chemie GmbH). The humidity level measured in the exsiccator at a room temperature (22–24 °C) without flies showed a rapid decrease and was close to zero after 2 h. The test was done under the same temperature conditions and time of exposure to desiccation stress was increased from 12 to 20 h during the course of the experiment. The surviving flies were placed in culture bottles for reproduction. The time of exposure to the desiccation stress was not enough to result in statistically significant mortality in the starvation resistance test. Thus, selection for increased desiccation resistance did not lead to increased starvation resistance.

Starvation resistance selection (SS)

After emergence, flies were immediately put in standard vials containing 7 mL of pure agar medium (2% of agar) to provide moisture and avoid desiccation while starving. The time needed to reach the desired mortality rate due to starvation stress was gradually increased from 35 to 60 h during selection. The survivors were put in culture bottles and allowed to found the next generation.

Longevity selection (LS)

After emergence, flies were immediately placed in vials with the standard medium. Every second day they were transferred to fresh food vials until the desired mortality rate was reached. This period was 4 weeks at the beginning of the experiment and was later increased to 6 weeks. The live individuals were used to start a new generation.

In all lines except KS emerging adults were collected and sexed under CO2 anesthesia. The acclimation/hardening treatment in the CS and HS lines was done to focus selection on resistance to the highest stress, which flies can survive and still reproduce. For the mortality assays (CS, HS, DS, SS) and longevity test (LS), flies from each replicate line were distributed among 10 shell vials (100 × 24 mm, 7 mL of medium) with 30 pairs per vial. The stress doses and the ageing period in the longevity test were such as to give a mortality rate of ca. 50%. In the heat-knockdown test, 50% of flies with highest knockdown scores were selected. Thus, the intensity of selection in all lines was around 0.80. Selection was implemented every second generation to allow the population to recover and avoid possible cross-generation effects (see e.g. Watson & Hoffmann, 1996). Relaxed generations were maintained using the same procedures as the control lines (UC). In all generations the flies were reared at 25 ± 1 °C and a 12 : 12 h light : dark cycle in 100-mL bottles containing 35 mL of the standard oatmeal-sugar-yeast-agar medium. Each replicate line was kept in five culture bottles using ca. 30 pairs of parents per bottle. The level of larval density was moderately high and did not lead to a marked delay in preadult development (most of the flies emerged on the ninth day). For nonstressed parents in the selection lines as well as for parents in the control lines the period of egg laying was 1 day, whereas for stressed parents it was either 2 (KS, DS) or 3 days (CS, HS, SS). The selected parents in the longevity test were kept in the bottles for 3 days. In every generation, progeny collected from different bottles of the same replicate line were mixed.

Estimation of selection responses

Number of generations

After 19–21 months of the experiment, the selection and control lines were tested for the above-mentioned resistance traits and longevity as well as preadult developmental time, early age productivity and thorax length. All traits, except heat-knockdown resistance, were assayed after the selection lines were reared for 2–3 generations without selection under standard laboratory conditions. Heat-knockdown resistance was estimated when selection had been relaxed for one generation. Due to differences between selection regimes in the amount of time needed for the selection process, the lines passed different numbers of generations. The CS, HS, KS and DS lines had passed 21 selection (not relaxed) generations before they were tested for longevity and for survival after cold, heat, desiccation and starvation stresses. The SS and LS lines were analysed for these traits after being selected for 26 and 11 generations, respectively. Heat-knockdown resistance, developmental time, productivity and thorax length were estimated after all above-mentioned lines had passed one more selection generation. The total number of generations including both selected and relaxed ones for the CS, HS, KS and DS lines was 45–48, whereas for the SS and LS lines it was 55–58 and 25–26, respectively. The control lines passed from 45 to 48 generations.

Stress resistance and longevity

To minimize any genotype-by-environment effects, the stress-resistance traits and longevity were assayed under the same conditions as those used in the selection procedures. The differences in estimation techniques are described below.

All selection traits were tested only in females reared under relatively low uncontrolled larval density with ca. 300–500 larvae per bottle. In the mortality assays, the number of individuals per experimental vial was always equal to 10. Survival was checked after 24 h at 25 °C and flies that were able to walk on vertical walls of a glass vial were scored as living. The cold-shock and desiccation resistance were tested using small plastic tubes (75 × 12 mm) instead of standard glass vials. This allowed us to increase the capacity of the ice container and exsiccator and to avoid experimental blocks in the two above-mentioned assays. For the heat-shock and knockdown tests, five and four blocks, respectively, were formed. The heat-shock resistance was tested at 37.5 °C simultaneously in five water baths and in each water bath one vial represented one replicate line. Knockdown resistance was assayed during four consecutive days, with each replicate line being tested once a day (but using a different sample of flies). The number of females per run was 50 and the temperature in the knockdown tube was 39.0 °C. The time of exposure to desiccation stress was 30 h. In the starvation and longevity assays, mortality was recorded every 8 and 72 h, respectively, until all flies had died. The vials in the longevity test were replaced every 3 days.

Other traits

To estimate developmental time, early age productivity and thorax length all experimental lines were simultaneously reared under controlled larval density conditions. Upon emergence, flies were collected from the culture bottles and aged for 6–7 days. Then groups of 40–50 pairs were placed in bottles with standard Drosophila medium seeded with live yeast to stimulate mating and oviposition. After 24 h, each group was transferred from its bottle to a separate empty vial containing a plastic spoon filled with 2 mL of the yeast-seeded standard medium. The flies were allowed to oviposit on spoons for 1.5 h. Thirty first-instar larvae (1–3-h-old) were transferred from each spoon to a vial containing 7 mL of medium without live yeast. All vials were synchronously put in the 25 °C temperature room. Eclosed adults were collected and counted every 6 h at 6:00, 12:00, 18:00 and 24:00.

In all experimental lines, the peak of female eclosion was on the eighth day in the morning and females collected on that day at 12:00 (i.e. emerged between 6:00 and 12:00) were used for the productivity assay. To obtain enough males of the same age for this test, additional bottle cultures with low larval densities were set up for each set of lines in parallel with the thirty-larva vials and kept in the temperature room at 25 °C. On the eighth day at 6:00 all emerged flies were removed from the bottles and at 11:00 new young males were collected for the productivity test. The test was started 5 h later by placing five females from each thirty-larva vial and five males from the corresponding replicate line bottle in a new vial with yeast-seeded medium. After 24 h, the flies were transferred to bottles with live yeast for another 24 h and this was repeated two more times. Then the males were discarded and females placed in plastic microtubes filled with a fixative liquid (1/2 glycerine + 1/2 70% ethanol) together with other females collected from the same culture vials. The flies emerged in the productivity test were scored using a Drosophila counter (model BM 8310, Institute of Genetics, Copenhagen).

The females stored in the microtubes were estimated for thorax length. It was measured with an ocular micrometer as the distance from the anterior margin of the thorax to the posterior tip of the scutellum from the dorsal view. Five individuals were sampled for measurements from each culture vial.

Data treatment

Measurement units and the number of observations

In the cold-shock, heat-shock and desiccation resistance assays, the proportion of surviving flies (survival rate) was computed for each experimental vial and these values were used as individual observations in the further data analysis. For starvation resistance, proportions of dead flies were recorded every 8 h and the time taken for 50% of flies to die (LT50) was estimated by linear interpolation. In the knockdown test, each individual fly was assigned a knockdown time according to when it fell into the collecting vial (replaced every 30 s) and the mean knockdown time was calculated for each run to be used as individual observation. Longevity was estimated as the time interval in days from the starting day of the test to the recorded time of a fly's death. Developmental time was measured as the time interval in hours from the mid-point of the oviposition period to the recorded time of eclosion. For both traits, mean-vial values were computed to be considered further in the statistical analysis, with males and females being pooled in the case of developmental time. Productivity was computed as progeny number per female per vial/bottle over 4 days. The number of individual observations used for the data analysis corresponded to the number of vials/bottles. Thorax length was measured in ocular micrometer units and converted to millimetres (1 unit = 0.014 mm). Each measurement represented an individual observation.

The number of test vials per replicate line was equal to five for the longevity assay and for the resistance tests, which estimated survival rates. In the heat-knockdown assay, each replicate line was represented by four measurements because it was tested once a day during four consecutive days. For developmental time and thorax length, six vials per replicate line were used, whereas productivity was assayed using five vials/bottles per replicate line. Taking into account the number of replicate lines within selection regimes and the measurement units (see previous paragraph), cold-shock, heat-shock, desiccation and starvation resistance and productivity were estimated using 25 observations per selection regime. For developmental time this value was equal to 30; for heat-knockdown resistance and thorax length it was 20 and 150, respectively.

Statistical analysis

The statistical analysis was based on replicate line means, i.e. five observations per selection regime for each trait. Before analysis, the data were subjected to standard transformations to improve their normality and make the variances less dependent of the means. The survival proportions for cold-shock, heat-shock and desiccation resistance were arcsine-square-root-transformed. Other traits were log-transformed except productivity, which was square-root-transformed. In accordance with the goal of the study, each selection regime was compared to the unselected control for every trait using nonorthogonal planned comparisons in one-way anova with selection regime as factor. The sequential Bonferroni procedure was used to test for statistical significance of these comparisons (see e.g. Sokal & Rohlf, 1995). To reveal possible relationships among selection regimes across five stress-resistance tests, the UPGMA method was applied. A rank 1–7 (lowest and highest resistance, respectively) was assigned to each of seven regimes within each resistance assay and a matrix of rank-based Euclidean distances between the regimes was analysed using the UPGMA technique. All computations were done with STATISTICA package (StatSoft, Inc, 2001).


Figure 1 shows the values of all analysed traits in each of five replicate lines for seven selection regimes (line sets). Table 1 presents regime means based on the replicate line values and F-ratios for one-way anovas of these values with selection regime as factor. The statistically significant effect of selection treatment was found for all traits except productivity, with stress-resistance traits and developmental time showing highly significant F-ratios (Table 1). The planned selection vs. control comparisons (every set of selected lines vs. UC lines) following the anovas revealed statistically significant differences for the five resistance traits, developmental time and thorax length. These differences ranged from the largest negative to the largest positive selection-minus-control value are presented in Fig. 2. As seen from the figure, all stress-resistance traits had highly significant direct responses to selection. Correlated responses were most pronounced in the starvation resistance test, where only one set of lines, HS, did not show a difference from the unselected control. In contrast, heat-knockdown and desiccation resistance did not reveal a statistically significant increase in the lines selected for other traits. The HS and DS lines had higher cold resistance than the UC lines, whereas KS and DS showed improved survival in the heat-shock test. Developmental time was significantly extended in the SS and KS lines as compared to UC. Thorax length revealed a negative correlated response to selection for longevity.

Figure 1.

Stress-resistance and fitness-related traits in replicate selection and control lines (UC, unselected control; CS, cold-shock selection; HS, heat-shock selection; KS, heat-knockdown selection; DS, desiccation selection; SS, starvation selection; LS, longevity selection). All traits are measured in females except developmental time, which is calculated for pooled samples of two sexes.

Table 1.  Trait means (± SE) in all sets of lines and F-ratios for the selection regime effect based on one-way anovas of replicate line values.
TraitSelection regimeRegime effect, F6,28
  1. *P < 0.05; **P < 0.01; ***P < 0.001; ns, nonsignificant.

  2. UC, unselected control; CS, cold-shock selection; HS, heat-shock selection; KS, heat-knockdown selection; DS, desiccation selection; SS, starvation selection; LS, longevity selection.

Cold-shock resistance (survival rate)0.140 ± 0.0440.668 ± 0.0370.341 ± 0.0390.268 ± 0.0450.368 ± 0.0860.260 ± 0.0390.236 ± 0.0529.26***
Heat-shock resistance (survival rate)0.160 ± 0.0470.318 ± 0.0530.660 ± 0.0600.456 ± 0.1020.424 ± 0.0470.200 ± 0.0260.300 ± 0.0797.11***
Heat-knockdown resistance (min)1.74 ± 0.101.32 ± 0.101.85 ± 0.314.42 ± 0.621.45 ± 0.051.29 ± 0.111.56 ± 0.0916.79***
Desiccation resistance (survival rate)0.190 ± 0.0370.304 ± 0.0360.345 ± 0.0370.312 ± 0.0300.748 ± 0.0480.176 ± 0.0280.277 ± 0.05518.00***
Starvation resistance (LT50, h)56.00 ± 0.8061.95 ± 1.0557.42 ± 0.7360.32 ± 0.8061.79 ± 0.9572.80 ± 1.3062.69 ± 1.0829.53***
Longevity (days)37.33 ± 1.1033.70 ± 1.5935.75 ± 1.9235.93 ± 2.2538.30 ± 1.9435.30 ± 1.1144.29± 2.213.52*
Developmental time (h)193.33 ± 0.15193.40 ± 0.41192.37 ± 0.31194.54 ± 0.29194.26 ± 0.38194.69 ± 0.32194.41 ± 0.137.95***
Productivity (progeny number)271.92 ± 13.13279.78 ± 9.31286.59 ± 10.57283.94 ± 9.98276.69 ± 6.80268.79 ± 9.24285.28 ± 7.930.48ns
Thorax length (mm × 100)111.19 ± 0.49110.25 ± 0.26110.53 ± 0.45110.15 ± 0.54111.74 ± 0.32109.99 ± 0.21109.25 ± 0.294.02**
Figure 2.

Responses to selection for seven traits presented as mean differences between selection and control lines (selection minus control) ranged from the largest negative to the largest positive value. The dark and light bars denote statistically significant and nonsignificant differences in accordance with planned selection vs. control comparisons (one-way anova, selection regime as factor). The stars indicate significance levels: *P < 0.05, **P < 0.01 and ***P < 0.001.

Associations between selection regimes with regard to their stress-resistance profiles were investigated using ranks based on the trait values from Table 1. Each regime within each of five resistance tests was coded 1–7 depending on its relative (to other regimes) resistance, with 1 being the lowest and 7 the highest value. A matrix of Euclidean distances between selection regimes was formed using the rank data (not shown) and analysed with the UPGMA method. The resulting tree diagram with average resistance ranks for each set of lines is shown in Fig. 3. As seen from the diagram, there were two major clusters in accordance with differences in resistance ranks. The cluster with higher ranks included only stress-resistance selection lines representing all thermal selection regimes and the desiccation selection regime. The other cluster consisted of the starvation and longevity selection lines as well as unselected control lines. The closest association was observed between the regimes that were involved in high-temperature selection, KS and HS. Two other linked pairs, DS–CS and SS–LS, showed relatively larger linkage distances.

Figure 3.

Cluster analysis of resistance ranks for all selection regimes across five resistance tests. In each test, the line sets were ranked 1–7 depending on their resistance values (as presented in Table 1) to give a Euclidean distance matrix, which was further analysed using the UPGMA method.

When compared using the Mann–Whitney U-test, the mean resistance ranks of two major clusters were significantly different (4.90 vs. 2.80, U20,15 = 60.00, P < 0.01). There was also a significant difference between the KS-HS-DS-CS cluster and the SS-LS subclaster (4.90 vs. 3.20, U20,10 = 49.00, P < 0.05). The average resistance rank of the selection lines both with and without LS lines was significantly higher than that of the control UC lines (4.33 vs. 2.00, U30,5 = 25.00, P < 0.05 and 4.48 vs. 2.00, U25,5 = 20.50, P < 0.05, respectively).


Stress resistance

All stress-resistance traits showed clear-cut direct responses to selection, which suggests a high level of genetic variation for stress resistance in the base population. In three of five stress assays (cold shock, heat shock, starvation) increased resistance was observed as correlated responses. Also, the resistance ranks averaged over all tests were higher in the selected lines compared to the control lines. These results are in agreement with numerous physiological-genetic studies reported that selection for resistance to one stress can confer resistance to other stresses. However, variation of correlated responses across stress assays and line sets indicates no single general resistance mechanism. Three sets of thermal-selected lines and the DS lines were markedly different with regard to their resistance profiles from the SS lines. Selection for starvation resistance did not result in enhanced resistance to other stresses. At the same time, each set of selection lines except HS showed a well-pronounced increase in survival rate for the starvation test. This suggests that starvation resistance easily evolves as correlated response to selection for resistance to desiccation, cold shock and knockdown by heat but the physiological mechanism of starvation resistance alone cannot explain for resistance to the above-mentioned stresses.

Survival rates under starvation and desiccation conditions in Drosophila are two most extensively studied resistance traits, for which a large number of correlations with other traits have been documented (for a review, see Hoffmann & Harshman, 1999). Particularly, a positive correlation of starvation resistance with lipid level in adult flies was consistently reported (e.g. Chippindale et al., 1996; Djawdan et al., 1998; Harshman et al., 1999a) suggesting that the accumulation of metabolic energy reserves may underlie this kind of stress resistance. There is also evidence on association of starvation resistance with increased level of glycogen (Djawdan et al., 1998). The latter was shown to contribute to enhanced desiccation resistance (e.g. Graves et al., 1992; Chippindale et al., 1998; Bradley et al., 1999) that can provide a common physiological basis for two resistance traits. The key components underlying desiccation resistance are thought to be reducing rates of water loss and increasing water storage (Hoffmann & Harshman, 1999; Gibbs & Matzkin, 2001; Folk et al., 2001). Chippindale et al. (1998) hypothesized that glycogen may function as a sponge, binding some of the extra water stored in desiccation-resistant flies. Increased starvation resistance has been repeatedly shown in Drosophila strains selected for resistance to desiccation (e.g. Hoffmann & Parsons, 1989a, 1993a; Rose et al., 1992).

In contrast to desiccation resistance, survival after cold shock has not been reported to correlate with starvation resistance in laboratory selection experiments on Drosophila. However, these traits may not be independent because cold resistance in adults is associated with the amount of lipid and carbohydrate (Ohtsu et al., 1992; Chen & Walker, 1994) and qualitative changes in the lipid composition (Ohtsu et al., 1993,1998; J. Overgaard, J.G. Sørensen, S.O. Petersen, V. Loeschcke & M. Holmstrup, unpublished). Particularly, Overgaard and coauthors revealed such changes analyzing three replicate lines from our CS regime after 10 selection generations. Accumulation of glycogen might also promote the increased cold resistance of the DS lines. Resistance to cold and desiccation should work together to prevent cellular injury and death caused by ice crystal formation. A connection between cold hardiness and desiccation has been found in a few species of arthropods (Ring & Danks, 1994). In most examples, the insect dehydrated to avoid body freezing and thus the role of glycogen as metabolic water sponge might be important.

There are also no physiological data, which could explain the increased survival of the KS lines in our starvation test. It could be suggested that the accumulation of metabolic energy reserves such as lipid and/or carbohydrate was a correlated response in lines selected for heat-knockdown resistance. However, this factor is unlikely to contribute to increased knockdown resistance: neither SS nor DS lines differ in their knockdown time from the unselected control. Some indirect evidence indicates that individuals resistant to knockdown by heat might have an increased lipid level. Bubli et al. (1998) and Sørensen & Loeschcke (2001) found enhanced heat-knockdown resistance in D. melanogaster reared at high larval densities, which has been shown to result in increased fat content (Zwaan et al., 1991; Borash & Ho, 2001).

The HS lines were the only set of lines, in which starvation resistance was not affected by selection. This suggests that selection for higher survival after heat shock was associated with a resistance mechanism different from the lipid/carbohydrate accumulation. Since individuals were heat-hardened before being selected, this mechanism could involve inducible heat-shock proteins. In some studies, expression of Hsp70 was reported to increase in hardened heat-resistant flies, although this usually did not explain all variation in heat resistance (e.g. Dahlgaard et al., 1998; Lansing et al., 2000; Norry & Loeschcke, 2003). However, the enhanced survival of the KS lines in the heat-shock test and similar responses of the KS and HS lines to other stresses could not be explained by genes controlling the Hsp70 expression. We selected for knockdown resistance without prior heat hardening and thus such genes were not a primary target of selection. More likely, other Hsp genes with constitutive expression selected under both regimes were responsible for the observed parallelism in the KS and HS resistance values. The possibility of association of constitutively expressed heat-shock genes with knockdown resistance to heat has been demonstrated by studies of hsr-omega in D. melanogaster (McKechnie et al., 1998; McColl & McKechnie, 1999). However, hsp-omega does not seem to be a part of any general resistance mechanism. The selection lines used by McKechnie et al. (1998) revealed no correlation between heat-knockdown resistance and other resistance estimates, including survival after heat shock, cold shock and desiccation (Hoffmann et al., 1997).

Induced HSPS might be involved in the increased cold resistance of the HS lines in the cold-shock test. It has been reported that Drosophila larvae and adults exposed to cold express Hsp70 during recovery (Burton et al., 1988; Yiangou et al., 1997; Goto & Kimura, 1998; Sejerkilde et al., 2003) and survival after heat shock and cold shock can be improved by cold and heat acclimation/hardening, respectively (Burton et al., 1988; Sejerkilde et al., 2003). The fact that the CS lines did not show an increase in survival following heat shock indicates that the expression of inducible HSPS might be not affected by selection for cold resistance. This is in agreement with results by J. Overgaard, J.G. Sørensen, S.O. Petersen, V. Loeschcke & M. Holmstrup (unpublished), who found no changes in cold-induced Hsp70 expression for the CS lines after 10 selection generations.

The increased resistance of the DS lines in the heat-shock test suggests that HSPS could also be involved in resistance to desiccation. There is no data on Hsp expression under desiccation stress in Drosophila but some evidence has been obtained for the flesh fly Sarcophaga crassipalpis by Tammariello et al. (1999). These authors revealed that two heat-shock protein transcripts, hsp23 and hsp70, were upregulated in pupae of S. crassipalpis following a nonlethal desiccation stress. Hoffmann (1990,1991) provided indirect evidence on the possible association between induced HSPS and desiccation resistance in Drosophila. He found that flies become more resistant to desiccation after prior exposure to a nonlethal desiccation stress and after heat hardening, although the latter had a relatively small effect. In the context of these data and data on the Hsp expression in cold-stressed flies (see previous paragraph), the increased cold resistance in the DS lines might also be explained by selection on Hsp-related genes.

In the above discussion we considered two possible explanations for the observed multiple-stress resistance, one of which is accumulated metabolic energy reserves and the other the nonspecific cellular defense system based on HSPS. Apparently these physiological responses are not alternative and supplement each other in providing resistance to some stresses. For instance, it may be suggested that the increased cold resistance of our DS lines is due to a correlated response involving both carbohydrate biosynthesis genes and Hsp genes. However, HSPS and/or other factors not related to the lipid/carbohydrate accumulation appear to be more important for resistance to most of the stresses considered here. This is indicated by the fact that the lines selected for increased thermal and desiccation resistance had generally higher survival rates over stress assays than the starvation-selected lines.

Reduced metabolic rate, which has also been hypothesized as a general stress-resistance mechanism (Hoffmann & Parsons, 1989a,b,1991), may explain resistance to a variety of, but not to all stresses. One exception seems to be cold resistance: at low temperature a reduced metabolism would decrease rather than increase survival, leading to an early stop of vitally important physiological processes (see e.g. Wieser, 1973). Hoffmann & Parsons (1989b) reported no change in cold resistance for the lines of D. melanogaster, which were selected for enhanced desiccation resistance and had reduced mass-specific metabolic rate. However, in our experiment a similar selection resulted in increased resistance to cold shock. Also, the HS lines had improved survival in the cold-shock assay. These results are not consistent with the above-mentioned hypothesis indicating that reduced metabolic rate should not be a key factor of the observed cross-resistance.


Unlike stress-resistance traits, the direct response to selection on longevity was not statistically significant. This fitness parameter was measured after a lower number of selection generations, which together with the fact of a generally low lifespan heritability may explain the present results. Whereas estimates of heritability for desiccation and starvation resistance reached 60 and 100%, respectively (Hoffmann & Harshman, 1999), for longevity they ranged from 3 to 20% being typically around 3–4% (Baret et al., 1995). Selection on age of reproduction often resulted in a significant longevity increase only under stressful or suboptimal conditions such as high larval density (e.g. Clare & Luckinbill, 1985; Luckinbill & Clare, 1986; Service et al., 1988). We also found no correlated response for longevity in any set of the resistance-selected lines. Selection for stress resistance has been reported to extend lifespan in some Drosophila studies (Rose et al., 1992; Hoffmann & Parsons, 1993b; Norry & Loeschcke, 2003) but not in others (Harshman et al., 1999b; Norry & Loeschcke, 2002a). The lack of evidence on longevity changes in experiments with stress-resistant lines including present work does not agree with the hypothesis known as the stress theory of ageing (Parsons, 1995). The latter suggests that a generalized stress resistance is the major mechanism underlying extended longevity and that a more direct procedure for increasing longevity is to select for stress resistance as the primary trait.

However, our starvation test showed that the lines selected for longevity were more resistant than the unselected control. In addition, the values of resistance ranks across all tests for the LS lines were similar to that of the SS lines. An increase in starvation resistance has been reported for different long-lived Drosophila strains (e.g. Service et al., 1985; Zwaan et al., 1995; Rose & Archer, 1996) and a recent QTL analysis in D. melanogaster discovered a linkage group on the second chromosome that conferred both starvation resistance and longer lifespan (Wang et al., 2004). There is also evidence on increased lipid content in strains selected for extended longevity (Rose, 1984; Service, 1987; Djawdan et al., 1998). The fact that the LS lines showed a clear-cut correlated response for starvation resistance, which is known to be consistently associated with the lipid accumulation (Hoffmann & Harshman, 1999), suggests that the latter might be a factor contributing to prolonged lifespan. On the other hand, there was no indication that the LS lines were more resistant in other stress assays and this was again not in agreement with the stress theory of ageing.

It should be noted that data on association between lifespan and resistance traits are frequently inconsistent, suggesting that different long-lived strains can utilize different combinations of physiological mechanisms to alter their longevity and stress resistance. For instance, one set of long-lived lines sometimes called ‘the UC Irvine lines’ (Rose, 1984) were reported to have increased starvation, desiccation and oxidative stress resistance (Service et al., 1985; Rose et al., 1992; Harshman & Haberer, 2000) as well as decreased metabolic rate and increased lipid and glycogen content (Service, 1987; Graves et al., 1992). In contrast, another set of longevity-selected lines known as ‘the Wayne State University lines’ (Luckinbill et al., 1984) showed increased resistance to oxidative stress but no change in metabolic rate, starvation resistance and lipid content, and any correlation with desiccation resistance and glycogen content was marginal (Arking et al., 1988; Dudas & Arking, 1995;Force et al., 1995). Apparently, such a difference in correlated responses could be due to initial genetic differences between the base populations, which is typical for laboratory selection experiments on Drosophila (Harshman & Hoffmann, 2000). On the other hand, the evident increase in starvation resistance observed in the LS lines as well as in most of the stress-resistance selection lines seems to reflect another feature of laboratory selection studies. As has been pointed out by Harshman & Hoffmann (2000), the selection response may tend to be based on the accumulation of energy storage compounds when food resources are abundant in every generation during the course of the experiment. In contrast, adaptation by energy conservation via reduced metabolic rate could be a common response to stress in natural populations where food supplies are limited (Hoffmann & Parsons, 1991).

Other traits

Correlated responses to selection for stress resistance and longevity were observed under optimum laboratory conditions in two of three additionally considered fitness-related traits. Productivity estimated over the first 4 days after eclosion showed no statistically significant heterogeneity among the selection regimes. The trade-off between longer lifespan and decreased early age reproduction in Drosophila has been observed in numerous age-at-reproduction selection experiments (e.g. Rose, 1984; Luckinbill et al., 1984; Partridge et al., 1999) as well as in an experiment on direct selection for extended longevity (Zwaan et al., 1995). There is also some evidence on a possible trade-off between fecundity and adult stress resistance in lines selected for resistance to cold (Watson & Hoffmann, 1996) and desiccation (Hoffmann & Parsons, 1989a). It should be mentioned, that in our study differences in early productivity between lines could be expected not only due to the above-mentioned trade-offs but also because of variation in age of parents at reproduction. To start every second generation, 1.5–2.5-day-old and 4–6.5-week-old flies were used in the SS and LS lines, respectively. For other lines, this age ranged from 5–6 (control) to 10–13 (CS) days. In fact, selection for extended lifespan in the LS lines was associated with delayed reproduction as compared to other lines, whereas the SS lines were selected for the earliest reproduction. However, as it is seen from the homogeneity of productivity values, any selection on this trait was inefficient. Apparently, the low heritability of fecundity (for references, see e.g. Roff & Mousseau, 1987) and relatively short duration of selection did not allow obtaining differentiation in early productivity.

In contrast, developmental time and thorax length, which are known to have relatively high heritability (Roff & Mousseau, 1987), were affected by selection. Developmental time was higher in two sets of lines, SS and KS, as compared to the control lines. A positive correlation with developmental time in Drosophila has been previously reported for starvation resistance by Chippindale et al. (1996) and Harshman et al. (1999a). Chippindale et al. (1996) suggested a trade-off between larval developmental time and adult storage compound acquisition arguing that slower development allows the lipid accumulation for the adult stage. Our data do not contradict this hypothesis: both SS and KS line sets had an increased survival in the starvation test. Selection for slower development in the SS lines was probably quite intensive because it overbalanced selection in the opposite direction due to earlier reproduction in the SS lines as compared to the control lines. At the same time, the absence of statistically significant correlated responses for developmental time in other sets of lines with enhanced starvation resistance (LS, CS, DS) indicates that delayed development is not a necessary cost of this resistance.

Thorax length in our experiment was reduced only in the longevity-selected lines. In most Drosophila studies, body size did not reveal a correlated response to selection on lifespan (Rose et al., 1984; Luckinbill et al., 1988; Zwaan et al., 1995; Norry & Loeschcke, 2002b). However, Buck et al. (2000) reported a decrease in dry body weight for two independent sets of long-lived strains, in which developmental time was also reduced. The parallel changes in body size and developmental time in the above-mentioned study were in the direction predicted by a well-known positive genetic correlation between them (for references, see Roff, 2000). In our LS lines, developmental time was not different from that of the control lines but body size was decreased, indicating a response to selection for lifespan independent of developmental time. Such independence is in agreement with other studies that compared the effects of developmental time and adult body size on longevity (Zwaan et al., 1991; Hillesheim & Steams, 1992; Partridge et al., 1999). Small body size is known to result not only from shortened preadult period but also from decreased rate of growth. A decrease in growth rate appears to be a plausible explanation for the reduced body size in our study that may be considered as a cost of selection for extended longevity.


Comparing indirect responses showed that selection for resistance to one stress can result in increased resistance to other stresses. However, the pattern of responses suggests no single general resistance mechanism. The most common and well-pronounced response was enhanced starvation resistance found in all selection lines except those selected for increased survival after heat shock. Taking into account data of previous physiological studies on Drosophila, this indicates that the accumulation of metabolic energy reserves such as lipid and/or carbohydrate could be associated with the observed increase in stress resistance. At the same time, it should not be a key factor for resistance to stresses other than starvation because the starvation-selected lines revealed no multiple-stress resistance.

The lines selected to survive a severe heat shock after heat hardening probably evolved a physiologically different resistance mechanism involving induced HSPS. The latter might also contribute to enhanced cold resistance of the heat-shock selection lines and improve the survival of the desiccation-selected lines in the heat-shock test. The lines selected for resistance to knockdown by heat and those selected for survival after heat shock had similar resistance values across stress assays. This might be due to selecting some Hsp-related genes not involved in the induced heat-shock response. Reduced metabolic rate as response to selection for stress resistance seems not to be an appropriate explanation of the observed cross-resistance when the results of the cold-shock test are taken into account.

The longevity-selected lines had increased starvation resistance but like the starvation-selected lines revealed no other correlated responses. These data show that longevity and multiple-stress resistance are not closely associated in the laboratory population considered. The similarity of the starvation and longevity selection lines with regard to resistance values across different stress assays indicates that selection for extended lifespan could promote the accumulation of metabolic energy compounds.

Two of three additionally considered fitness-related traits, developmental time and thorax length, were affected by selection. Developmental time was longer in the lines selected for increased resistance to starvation and knockdown by heat. Such a delay might reflect a fitness cost associated with more intensive accumulation of metabolic reserves in the larval stage. The reduced thorax length in the longevity selection lines could represent another fitness cost. This result also shows that a well-known trade-off between body size and speed of development may not necessarily evolve in strains selected for increased longevity.


We are grateful to Doth Andersen, Jesper Givskov Sørensen and Torsten Nygaard Kristensen for continuous support in running the Drosophila lab as well as Mia Frank Nielsen, Iben Skov Jensen and Janne Pleidrup Andersen for maintaining the selection lines. Funding for this research was supported by the Danish Natural Sciences Research Council through a Centre Grant.