The common hill-flowers wither, but they blossom again. The laburnum will be as yellow next June as it is now. In a month there will be purple stars on the clematis, and year after year the green night of its leaves will hold its purple stars. But we never get back our youth. The pulse of joy that beats in us at twenty becomes sluggish. Our limbs fail, our senses rot. We degenerate into hideous puppets, haunted by the memory of the passions of which we were too much afraid, and the exquisite temptations that we had not the courage to yield to. Youth! Youth! There is absolutely nothing in the world but youth!
– Oscar Wilde, The Picture of Dorian Gray (1891)
Dietary restriction (DR) extends lifespan in an impressively wide array of species spanning three eukaryotic kingdoms. In sharp contrast, relatively little is known about the effects of DR on functional senescence, with most of the work having been done on mice and rats. Here we used Drosophila melanogaster to test the assumption that lifespan extension through DR slows down age-related functional deterioration. Adult virgin females were kept on one of three diets, with sucrose and yeast concentrations ranging from 7% to 11% to 16% (w/v). Besides age-specific survival and fecundity, we measured starvation resistance, oxidative stress resistance, immunity, and cold-stress resilience at ages 1, 3, 5, and 7 weeks. We confirmed that DR extends lifespan: median lifespans ranged from 38 days (16% diet) to 46 days (11% diet) to 54 days (7% diet). We also confirmed that DR reduces fecundity, although the shortest-lived flies only had the highest fecundity when males were infrequently available. The most striking result was that DR initially increased starvation resistance, but strongly decreased starvation resistance later in life. Generally, the effects of DR varied across traits and were age dependent. We conclude that DR does not universally slow down functional deterioration in Drosophila. The effects of DR on physiological function might not be as evolutionarily conserved as its effect on lifespan. Given the age-specific effects of DR on functional state, imposing DR late in life might not provide the same functional benefits as when applied at early ages.
One may question the self-absorbance and hedonistic lifestyle of Dorian Gray, the main character in Oscar Wilde's novel. But do we not all envy his lack of physical degeneration? While aging is inevitable, not all traits degenerate at the same rate as an organism ages (Herndon et al., 2002; Simon et al., 2006). Biologists often define senescence as the age-related decline in fitness traits that arises due to internal physiological deterioration (Rose, 1991). Surprisingly, there have been few direct tests of the assumption that demographic and functional senescence are coupled. Studies that have looked at this give equivocal results (Barger et al., 2003; see below).
One way to test the relationship between demography and function would be to manipulate lifespan, and then to examine correlated effects on physiological traits. For decades, researchers have used dietary restriction (DR) to successfully extend lifespan in an impressive array of species spanning three eukaryotic kingdoms, including ciliates, yeast, rotifers, nematodes, water fleas, spiders, fruit flies, guppies, mice, rats, hamsters, dogs, and possibly monkeys (Verdonesmith & Enesco, 1982; Weindruch & Walford, 1988; Austad, 1989; Lane et al., 2001; Masoro, 2005). In sharp contrast, relatively little is known about the effects of DR on functional senescence, with most of the work having been done on mice and rats (reviewed by Weindruch & Walford, 1988; Masoro, 2002). Studies of rodents have shown that DR can retard age-related deterioration of the nervous system, muscular system, cardiovascular system, endocrine system, reproductive system, and immune system, and that DR retards the onset of age-associated diseases. In yeast, DR retards the age-related increase in generation time (an aging phenotype in yeast) (Jiang et al., 2000). In nematodes, DR increases resistance against oxidative and heat stress, although there are methodological problems in applying DR to worms (Houthoofd et al., 2005).
Recent studies suggest that extension of lifespan may not necessarily go hand in hand with slower functional senescence. In rodents, DR appears to have no effect on age-related deterioration in the nervous system of males or on male reproductive senescence, and can delay puberty and retard skeletal growth (Weindruch & Walford, 1988; Masoro, 2002). Other manipulations that extend lifespan, such as mutation or selection, generally increase stress resistance (Grotewiel et al., 2005), but the effects on physiology can be absent (Cook-Wiens & Grotewiel, 2002), trait and age specific (Kang et al., 2002; Goddeeris et al., 2003, 2005; Wessells et al., 2004; Broughton et al., 2005), and condition dependent (Marden et al., 2003). In a study of the long-lived daf-2 mutant in Caenorhabditis elegans, Jenkins et al. (2004) showed that mutants may appear healthy in physiological terms, but can quickly go extinct in a competitive environment because of reduced fertility relative to wild-type worms. Reznick et al. (2004) found that guppies from a high-predation environment lived longer under laboratory conditions than those from a low-predation environment. However, the long-lived guppies had higher rates of physiological senescence. Taken together, these studies suggest that lifespan extension does not necessarily slow down physiological deterioration of all functional traits.
The fruit fly provides an excellent system in which to explore the consequences of DR on physiological aging. Previous studies have examined the effect of different levels of DR on age-specific mortality and fecundity (Partridge et al., 2005b). Building on this previous work, here we study the effects of DR not only on mortality and fecundity, but also on four physiological traits in young and old flies, including resistance to starvation, oxidative challenge, bacterial infection, and cold shock. While each of these traits has been associated directly or indirectly with aging, they are likely to have distinct genetic pathways, and so provide a broad test of whether the ability of DR to increase lifespan extends to functional traits. Our results show that DR in Drosophila does not universally slow down functional deterioration, but that its effects are trait and age dependent.
Figure 1 illustrates the effect of diet on survival (Fig. 1A) and mortality (Fig. 1B) as a function of age for data from all trials pooled. In Fig. 1C, we plotted for each trial the probability of dying per time step for flies on the 11% and 16% diets, relative to those on the 7% diet. These parameter estimates were determined from the mixed-effects Cox model (Equation 1). Median lifespan ranged from 38 ± 1.2 days on the 16% diet, to 46 ± 2.1 days on the 11% diet, to 54 ± 12 days on the 7% diet (mean ± SE, n = 18 cages per diet) (Fig. 1A). Thus, our results confirm previous findings that DR extends lifespan in flies [H0(β2 = 0): z≥ 15.0, P < 0.0001 in all six trials]. As in an experiment like this there is no ad libitum control, one could also state that the concentration of sucrose and yeast decreases lifespan and increases the probability per time step of dying (Fig. 1C). Generally, diet had a linear effect on demography, except in fecundity Experiment 1, where the 11% diet was similar to the 7% diet, and fecundity Experiment 2, where the 11% diet was similar to the 16% diet (Fig. 1C). There is some debate about whether DR extends lifespan through a decrease in baseline mortality or in the rate at which mortality increases with age. In flies, Partridge and colleagues consistently find support for a decrease in baseline mortality (e.g. Pletcher et al. 2002; Main et al. 2003; Partridge et al. 2005c). However, DR has been shown to change the rate parameter in rats (Pletcher et al. 2000) and even in flies (Bross et al. 2005). An attempt to run a decomposition analysis (Pletcher et al., 2000) was unsuccessful, presumably because of cage frailty (see below). Visually, diet lowered the initial mortality rate more than it decreased the slope of the mortality curve (Fig. 1B), which suggests that its effect was of similar size across all ages.
Figure 2 shows the effect of diet on reproductive output of individual females when held either continuously with three males (Fig. 2A) or intermittently with two males for 1 day per week (Fig. 2B). For both experiments, we plotted the effect of diet on the total number of eggs laid in 33 days (Fig. 2C) and the rate of reproductive senescence (Fig. 2D). Although flies on the 7% diet lived longer (Fig. 1), they laid significantly fewer eggs than flies on either the 11% or 16% diet (Fig. 2C), confirming previous studies. Diet had no significant effect on the rate of reproductive senescence (Fig. 2D), which suggests that the negative effect of DR on fecundity was similar across all ages. The mated flies (Fig. 2) clearly lived shorter than the virgin females that were used to measure demography (Fig. 1). Interestingly, the effect of diet on reproduction depended on the mating conditions. When males were present continuously, reproduction peaked at intermediate concentrations of sucrose and yeast (Fig. 2C). In contrast, when males were present intermittently, fecundity was much higher (Fig. 2C), the rate of reproductive senescence was lower (Fig. 2D), and the 16% diet gave similar fecundity as the 11% diet (Fig. 2C). Note that for Experiment 2, there was also no difference in survival of females maintained on the 11% and 16% diets, but kept as virgins (Fig. 1C). The important point here is that although flies on the 16% diet were short lived, the short lifespan was not simply caused by malnutrition as the 16% diet was a good food source for reproduction. Continuous presence of males, on the other hand, did reduce fecundity, especially under rich food conditions.
In 4-day-old flies, DR flies were much more resistant against starvation than control flies [Fig. 3A and Supplementary Fig. S1A; H0(β2 = 0): z = 4.21, P < 0.0001]. However, 2 weeks later we found the opposite effect, with DR flies appearing much less resistant to starvation than the controls (Fig. 3A; Supplementary Fig. S1B; z = −4.62, P < 0.0001). The same negative effect of DR was seen at later ages (Fig. 3A; Supplementary Fig. S1C; z = −8.71, P < 0.0001; and Supplementary Fig. S1D; z = −7.47, P < 0.0001). Under starvation conditions, median lifespan of flies kept previously on the 7% diet decreased with age from 66 h (4 days, Fig. S1A) to 40 h (20 days, Fig. S1B) to 36 h (33 days, Fig. S1C) to 24 h (47 days, Fig. S1D) when food deprived, whereas median lifespan of flies kept previously on the 16% diet actually increased with age from 46 h (4 days, Fig. S1A) to 56 h (20 days, Fig. S1B) to 60 h (33 days, Fig. S1C; 47 days, Fig. S1D) when food deprived. During the first two age classes, virtually no flies died in the control treatment (Fig. S1A,B). During the last two age classes, control flies also died in the presence of a sucrose solution. However, the strong negative effect of DR on starvation resistance was already obvious before control flies started dying off [i.e. within 2 days in 33-day-old flies (Fig. S1C) and within 20 h in 47-day-old flies (Fig. S1D)]. Although flies on the 11% diet had an intermediate lifespan (Fig. 1C), diet did not affect starvation resistance in a linear fashion.
Oxidative stress resistance
In 6-day-old flies, there was no significant effect of diet on oxidative stress resistance [Fig. 3A and Supplementary Fig. S2A; H0(β2 = 0): z = 0.081, P = 0.94]. At later age classes, however, DR significantly decreased oxidative stress resistance (Fig. 3 and Supplementary Fig. S2B: z = −2.885, P = 0.0039; Supplementary Fig. S2C: z = −4.22, P < 0.0001; Supplementary Fig. S2D: z = −4.07, P < 0.0001). Survival of control flies kept on a sucrose solution remained fairly high at all age classes. In 34-day-old flies (Fig. S2C), DR had a positive effect on survival of control flies. This reinforces our argument that DR had a negative effect on oxidative stress resistance. This effect was approximately linear.
In 7-day-old flies, DR had no significant effect on survival post-infection with the Gram-negative Pseudomonas aeruginosa [Fig. 3A and Supplementary Fig. S3A1; H0(β2 = 0): z = −0.16, P = 0.87]. Two weeks later, there was still no effect of diet (Fig 3A and Supplementary Fig. S3B1; z = −0.18, P = 0.86), although the same bacterial concentration resulted in a much steeper decline in survival. Thus, a potential effect of diet could have been masked by the severity of the treatment. This is unlikely the case in the first age class, because a considerable proportion of flies survived the infection. To make the treatment more benign, we lowered the dose to about half the original OD600. In older flies, DR had a positive effect on immunity against P. aeruginosa (Fig. 3A and Supplementary Fig. S3C1; z = 2.03, P = 0.042; Supplementary Fig. S3D1; z = 2.69, P = 0.0072). This effect appeared earlier and to a larger extent than the effect of diet on sham-infected flies.
Immunity against the Gram-positive Lactococcus lactis gave less clear results. Young, 7-day-old flies on 7% food had significantly increased resistance compared with flies on 11% food [Fig. S3A2; H0(β1 = 0): z = 2.95, P = 0.0032]. The effect was in the same direction but not significant compared with flies on 16% food [Fig. 3B and Fig. S3A2; H0(β2 = 0): z = 1.67, P = 0.095]. Two weeks later, flies on both 7% and 11% food had significantly increased resistance compared with flies on 16% food, although the effect was strongest in flies on 11% food [Fig. 3B and Fig. S3B2; H0(β1 = 0): z = −2.28, P = 0.023; H0(β2 = 0): z = 2.34, P = 0.019]. In 34-day-old flies, there was no effect of DR on survival post-infection with L. lactis [Fig. 3B and Fig. S3C2; H0(β2 = 0): z = −0.49, P = 0.62]. In 48-day-old flies, DR significantly increased immunity [Fig. 3B and Fig. S3D2; H0(β2 = 0): z = 2.77, P = 0.0057]. In general, DR can have some weak positive effects on immunity against L. lactis throughout life.
In the first two age classes, there was no significant effect of diet on cold-stress resilience [Fig. 3B and Supplementary Fig. S4A; H0(β2 = 0): z = −0.23, P = 0.82; Supplementary Fig. S4B: z = −1.35, P = 0.18]. In the last two age classes, flies kept on 7% diet recovered slightly more slowly from chill coma than flies kept on 16% diet (Fig. 3B and Supplementary Fig. S4C: z = 2.09, P = 0.036; Supplementary Fig. S4D: z = 2.29, P = 0.022).
The effect of diet on demographic senescence was large and always in the same direction, but in each trial we also found significant variation for cage frailty (≥ 7.74, P≤ 0.0054; trial 1: = 0.076; trial 2: = 0.143; trial 3: = 0.080; trial 4: = 0.010; trial 5: = 0.057; trial 6: = 0.019). This means that the cage in which a fly was housed contributed significantly to the variation in the probability per time step of dying. Variation in cage frailty appeared as a temporary leveling off in survival (Fig. 1A) and mortality (Fig. 1B), as some cages went extinct. More importantly, the standard error of the average median lifespan across cages increased dramatically with decreasing sucrose and yeast concentration. This suggests that cage frailty was diet dependent. Existing statistical packages do not provide a means to test for a diet by frailty interaction in proportional hazards modeling, although this interaction appears qualitatively to be an important component of overall variation.
In this study, we set out to test the assumption that life-extending manipulations will also delay age-related functional deterioration. We used DR to slow demographic senescence and measured its age-specific effects on several functional traits. We confirmed previous studies that DR extends lifespan and reduces fecundity. Negative effects of DR on fecundity have been shown in rodents and Drosophila, although long-term effects can be positive in female rats (Chippindale et al., 1993; Chapman & Partridge, 1996; Masoro, 2002). Our results also confirm earlier studies that showed a negative effect of mating or male harassment on female fecundity (Linder & Rice, 2005), and a diet-dependent cost of mating (Chapman & Partridge, 1996). Mair et al. (2004) argued that since DR also extends lifespan in sterile Drosophila females, lifespan extension through DR is not simply a result of a re-allocation of resources from reproduction to survival.
In contrast to the effects of DR on survival and reproduction, the consequence of DR on other traits varied not only from trait to trait, but also among ages. The most striking result was that DR initially increased starvation resistance, but dramatically reduced it later in life. We have since qualitatively confirmed this finding in an additional set of experiments using a population of wild-type flies collected from a peach orchard in Watkinsville, Georgia (D. S. Hwangbo, unpublished data). The positive effect of DR on starvation resistance in Drosophila has been known for over a decade (Chippindale et al., 1993), and increased stress resistance is often used to argue that lifespan extension is not compromised by physiological side effects (Kenyon, 2001). Our data clearly show that the positive effects early in life are offset by negative effects later in life. The initial positive effect of DR on starvation resistance and lipid content is consistent with the hormesis hypothesis (Masoro, 1998), which proposes that DR is a low-intensity stressor that induces investment of resources into stress resistance and survival. DR increases lipid content in young flies (K. J. Min, personal communication). This enables organisms to overcome adverse periods of food scarcity. As food conditions remain poor, however, this increased stress resistance cannot be sustained.
We also found that DR had no effect on oxidative stress resistance early in life, and negative effects at later ages, in contrast to the free radical theory of aging (Harman, 1956; Balaban et al., 2005). Although the attenuation of oxidative damage is a popular hypothesis to explain lifespan extension through DR, the evidence for this hypothesis is mixed (Masoro, 2000, 2005; Merry, 2004; Sinclair, 2005). Our data do not support the hypothesis, at least in terms of the ability to resist exogenous sources of oxidative damage. We do not know about rates or levels of endogenous damage. Furthermore, it remains possible that DR flies demonstrate compensatory feeding and therefore ingest more paraquat. There is currently no consensus regarding the effect of DR on feeding behavior. Mair et al. (2005) have found that DR does not affect the proportion of time that flies spent on feeding. Others found that DR has no effect on food intake (Bross et al., 2005), that DR increases food intake (Carvalho et al., 2005), and even that DR decreases food intake (Min & Tatar, 2006). The effect of DR on feeding behavior in the paraquat assay remains to be tested. A useful alternative to the paraquat assay would be to increase oxidative stress by maintaining flies at high oxygen levels. In this case, oxygen exposure would presumably be unaffected by feeding rate.
DR initially had no effect on immunity against P. aeruginosa, but positive effects later in life. DR had some weak positive effects on immunity against L. lactis throughout life. In Drosophila, there are at least two signaling pathways involved in innate immunity. Generally, the Toll pathway is involved in resistance against fungi and Gram-positive bacteria, and the Imd pathway is involved in resistance against Gram-negative bacteria (Lemaitre et al., 1997; Tanji & Ip, 2005). If DR affects gene expression differently in these two pathways, this could potentially explain why the effects of DR on immunity against P. aeruginosa were different from the effects of DR on immunity against L. lactis. Our results support recent studies that suggest a connection between longevity and immunity (Brummel et al., 2004; Kurz & Tan, 2004). The positive effects of DR on immunity imply that there may be a causal relationship between lifespan and immunity, or that these two traits share a common regulatory mechanism. On a side note, it remains unclear what the relationship is between survival post-infection, and, for example, expression of antimicrobial peptides (Zerofsky et al., 2005).
DR initially had no effect on cold-stress resilience, but weak negative effects later in life. The relationship between lifespan and cold-stress resistance remains unclear. Reduced production of insulin-like peptides extends lifespan but decreases cold-stress resistance (Broughton et al., 2005). Overexpression of hsp70 extends lifespan but does not affect cold-stress resistance at any age class (Minois, 2001). Selection for increased lifespan increases cold-stress resistance (Luckinbill, 1998; Bubliy & Loeschcke, 2005), but selection for increased cold-stress resistance either decreases (Anderson et al., 2005; Bubliy & Loeschcke, 2005) or does not change lifespan, depending on temperature (Norry & Loeschcke, 2002). Thus, in contrast to the stress theory of aging (Parsons, 1995), it is unlikely that a universal stress-resistance mechanism underlies lifespan extension (Bubliy & Loeschcke, 2005).
Our results provide a challenge for molecular geneticists and physiologists to test hypotheses about the mechanisms that underlie the trait- and age-specific functional responses to DR. This could be achieved by studying the effects of DR in mutants that have reduced or increased resistance against starvation, oxidative stress, pathogens, cold, heat, etc. (as Mair et al., 2004, did with fecundity). Several such mutants are already available (De Gregorio et al., 2002; Harbison et al., 2004; Mourikis et al., 2006). Another approach would be to switch animals from one diet to another and measure correlated responses in functional traits. Switch experiments measuring mortality (Mair et al., 2003) and fecundity (Chippindale et al., 1993; Good & Tatar, 2001) suggest that DR does not slow down age-related accumulation of damage. Instead, the effects of DR on mortality and fecundity are acute and independent of feeding history. The question of why mortality varies across cages also needs to be addressed, especially because frailty appeared to be diet dependent.
We conclude that DR does not universally slow down functional deterioration in Drosophila. Compared with the beneficial effects of DR in rodents, our results suggest that the effects of DR on function might not be as evolutionarily conserved as its effect on lifespan. It has been suggested in the literature that application of DR late in life may provide the same functional benefits as application of DR early in life, since DR has acute effects on mortality (Mair et al., 2003). Our results suggest a more cautious interpretation may be called for. Even if feeding history would have no effect on an organism's functional state, functional benefits of DR early in life can be offset by functional costs at later ages.
Drosophila melanogaster (Canton-S) were obtained from Stephen L. Helfand and kept in a large demography cage at room temperature. For each of the six trials (see below), flies were expanded in 177-mL plastic bottles. From these bottles, we collected virgin females that were either placed individually in 25 × 95-mm glass vials (reproduction) or equally distributed among 0.95-L demography cages at 200 flies per cage (other traits). Food was changed every other day. Cages were maintained at 24 °C, about 30% r.h. on a 12:12 h light–dark cycle, and rotated throughout the incubator to prevent confounding effects of microclimate.
Adult flies had access to one of three diets. The composition of each diet was modified from a recipe provided by Scott D. Pletcher and is given in Table 1. The yeast and sucrose concentration ranged from about 7%, to about 11%, to about 16% (w/v). Water and agar concentrations were adjusted to control the consistency of the food. Active dry yeast was obtained from SciMart (St. Louis, MO, USA).
|Weight (g)||% (w/v)||Weight (g)||% (w/v)||Weight (g)||% (w/v)|
For each experiment with a functional trait, we simultaneously collected 1800 virgin females for a demography experiment. These flies were divided into nine cages of 200 flies each. Survival of these flies was scored every other day. Thus, we measured age-specific survival for 6 trials × 3 diets × 3 cages × 200 flies for a total of 10 800 flies.
Fecundity was measured as daily number of eggs laid by individual females. To measure the effects of diet on fecundity, we set up two experiments. For each experiment we collected 30 females per diet. Females were held individually in vials from the day of eclosion onwards. In the first experiment, females were kept continuously with three males. The second experiment was set up to reduce the effects of harassment of females by males, which can reduce female life expectancy considerably (Partridge et al., 2005a). In the second experiment, females had access to two young males of less than 1 week old for the first 2 days and for 1 day per week thereafter. In both trials, flies were transferred daily to new food vials and the number of eggs laid was counted up to 33 days. Thus, we measured age-specific reproduction in 2 trials × 3 diets × 30 flies for a total of 180 flies.
Starvation resistance was measured as survival of flies without access to food but with access to water (wet starvation). For each age class, we prepared six vials per diet. Each vial contained one Kimwipe and two circular filter papers. One vial per diet received 2 mL of a 1% sucrose solution (control), and the other five vials received 2 mL of water (starvation treatment). We aspirated 15 flies per diet from the demography cages into each vial without anesthesia. Survival was scored roughly every other hour until virtually all flies were dead. Starvation resistance was measured at 4, 20, 33 and 47 days of age. Thus, we measured starvation resistance for 4 age classes × 3 diets × (1 control vial + 5 treatment vials) × 15 flies for a total of 1080 flies.
Oxidative stress resistance
Oxidative stress resistance was measured as survival of flies with access to paraquat. Paraquat is a compound that generates the free radical superoxide (), a reactive oxygen species that is also produced during mitochondrial respiration. Paraquat therefore increases oxidative stress. For each age class, we prepared six vials per diet. Each vial contained one Kimwipe and two circular filter papers. One vial per diet received 2 mL of a 1% sucrose solution (control), and the other five vials received 2 mL of a 1% sucrose solution and 10 mm paraquat (Aldrich, St. Louis, MO, USA) (oxidative stress treatment). We aspirated 15 flies per diet from the demography cages into each vial without anesthesia. Survival was scored roughly every other hour until virtually all flies were dead. Oxidative stress resistance was measured at 6, 19, 34, and 47 days of age. Thus, we measured oxidative stress resistance for 4 age classes × 3 diets × (1 control vial + 5 treatment vials) × 15 flies for a total of 1080 flies.
Immunity was measured as survival after septic infection with either P. aeruginosa or L. lactis bacteria. Pseudodomonas aeruginosa is a Gram-negative bacterium and an opportunistic pathogen that occurs in a variety of habitats including soil, water, and plant surfaces. Lactococcus lactis is a Gram-positive bacterium that occurs naturally in plant material and insects. Both bacteria species are found in flies derived from natural populations (V. Corby-Harris, unpublished data). Bacteria were obtained from Brian P. Lazzaro, who obtained P. aeruginosa from the American Type Culture Collection (ATCC) stock center and isolated L. lactis from hemolymph and thorax muscle of wild-type flies collected in a fruit orchard (University Park, PA, USA). Both bacteria were cultured in standard nutrient broth. The day before flies were infected, a working stock was incubated at 37 °C, and we aspirated 140 flies per diet from the demography cages into food vials at 10 flies per vial. The day at which flies were infected, the incubated working stock had reached the log phase of growth and was diluted with nutrient broth until an optical density of about 0.2 cm−1 at 600 nm was reached using a Helios Delta spectrophotometer (Thermo Electron, Waltham, MA, USA). After the second age class, the bacterial concentration was reduced to an OD600 of about 0.1 cm−1, because flies immunosenesce and the initial concentration was considered too virulent for 5- and 7-week-old flies. For each vial, flies were carbon dioxide anesthetized and poked in the lateral thorax using a minute, stainless steel pin with a 0.1-mm diameter (Fine Science Tools, North Vancouver, BC, Canada) that was dipped in either sterile broth (control) or bacterial culture. Between infections, the pin was sterilized by dipping it in 70% ethanol, wiping it with a Kimwipe, dipping it in another 70% ethanol tube and wiping it again. Out of the 14 vials per diet, we used two vials as control and six vials per bacterial species as treatment. All 42 vials were handled in random order to prevent confounding circadian effects (Lazzaro et al., 2004). All flies were infected within about 3 h, after which survival was scored for 48 h. Flies were infected at 7, 20, 34 and 48 days of age. Thus, we measured immunity for 4 age classes × 3 diets × 2 bacterial species × (1 control vial + 6 treatment vials) × 10 flies for a total of 1680 flies.
Cold-stress resilience was measured as recovery time after induction of chill coma by cold stress (David et al., 1998). To prevent confounding effects of handling order, we divided the experiment into three trials per age class and used a Latin square design to randomize diets within trials. For each trial, we aspirated 20 flies per diet from the demography cages into an empty vial. For each trial, one vial per diet was put on melting ice in a cooler, which knocks flies down into a chill coma. After 3 h, vials were returned to room temperature and flies were spread out on filter paper in an open Petri dish. A fly was aspirated into a time- and diet-specific empty vial as soon as it was able to stand on its legs. We changed time-specific vials every minute. Trials were separated by about 30 min. Cold-stress resilience was measured at 5, 18, 33, and 43 days of age. Thus, we measured cold-stress resilience for 4 age classes × 3 diets × 3 trials × 20 flies for a total of 720 flies.
All data except reproduction were times to an event. For cold-stress resilience, the event was recovery from chill coma. For the other traits, the event was death. These times to an event were analyzed using a mixed-effects Cox model with cage or vial as random effect (R 2.2.1, macro coxme) (Therneau & Grambsch, 2000; Therneau, 2003):
where λij(t) is the probability per time step that the event occurs on diet i in cage or vial j at time t; λ0j(t) is the baseline probability per time step; X are indicator variables: X1 = 1 for the 11% diet and 0 otherwise, X2 = 1 for the 16% diet and 0 otherwise; β are parameters for the direction and magnitude of the relative effect of the 11% diet (β1) and the 16% diet (β2); and Zj is the frailty for cage or vial j, which has a normal distribution with mean 0 and variance . Thus, the expected hazard rate in an average cage or vial is λ0(t) for the 7% diet, for the 11% diet, and for the 16% diet. The random effect was tested by a likelihood ratio test, where the test statistic equals twice the difference between the maximum likelihood of a model with random cage or vial effect (R macro coxme) and the maximum likelihood of a model without random cage or vial effect (R macro coxph). This test statistic is χ2 distributed with one degree of freedom. For Fig. 1, mortality was estimated by µx ≈ –ln(px) (Lee, 1992), where px is the probability of surviving from x– 1 to x.
Analysis of fecundity data was complicated by the fact that repeated-measures data are not independent across ages and count data have non-normal errors. The simplest way to avoid pseudoreplication is to analyze the total number of eggs laid per individual. To test for the age-specific effect of diet, we first fit the following four-parameter model for each individual (R macro nls) (McMillan et al., 1970):
where Y is the number of eggs laid at age X, a is the intercept parameter that represents the potential maximum number of eggs laid if there was no maturation period [Y(0) if e−c(X–d) = 0], b is the slope parameter that represents the rate of reproductive senescence, c is a parameter that represents the maturation rate, and d is the age at first reproduction. If this model did not converge, we fit a generalized linear model with Poisson errors and log link [e−c(X–d) = 0; R macro glm], ignoring egg counts during the maturation period. We then tested the effect of diet on the total number of eggs laid and on the estimate for parameter b (the rate of reproductive senescence), using the Wilcoxon rank sum test for pairwise comparisons (R macro wilcox.test). We applied Bonferroni's correction for multiple comparisons by using an adjusted significance level of α′ = 0.05/3.
We thank Isabell Scherer and Chris Rodgers for preparing the food, Stephen Helfand for flies, Scott Pletcher for diet recipes, Brian Lazzaro for bacterial stocks, and Jake Moorad, Jodie Linder, Chih-Horng Kuo and two anonymous reviewers for comments on an earlier draft of this manuscript. This research was supported by an Ellison Medical Foundation Senior Scholar Award to D.E.L.P. and the Alton Graduate Fellowship to V.C.H.