Transgenerational memory effect of ageing in Drosophila


Frederic Mery, Laboratoire Evolution, Génomes et Spéciation, UPR 9034, CNRS, Avenue de la Terrasse, 91198 Gif sur Yvette, France. Tel.: +33169823732; fax: +33169823736; e-mail:


Children born to older parents tend to have lower intelligence and are at higher risk for disorders such as schizophrenia and autism. Such observations of ageing damage being passed on from parents to offspring are not often considered within the evolutionary theory of ageing. Here, we show the 25% memory impairment in Drosophila melanogaster offspring solely dependent on the age of the parents and also passed on to the F2 generation. Furthermore, this parental age effect was not attributed to a generalized reduction in condition of the offspring but was specific to short-term memory. We also provide evidence implicating oxidative stress as a causal factor by showing that lines selected for resistance to oxidative stress did not display a memory impairment in offspring of old parents. The identification of the parental age-related memory impairment in a model system should stimulate integration between mechanistic studies of age-related mortality risk and functional studies of parental age effects on the fitness of future generations.


Damage because of ageing should not be passed onto offspring through the germ line, because these effects could accumulate in a lineage and result in reduced fitness compared to lineages without germ line ageing. Ageing should, thus, be limited to the soma (Kirkwood & Holliday, 1979), at least at typical reproductive ages. However, there is evidence that offspring quality can be affected by parental age (Kern et al., 2001; Barnes, 1984; Cadieu, 1983; Fox, 1993; Kennedy et al., 1994) or even grandparental age (Hercus & Hoffmann, 2000). How does this evidence fit within evolutionary theories of ageing? One possibility is that there are trade-offs between offspring quality at later parental ages against quality at early parental age, or against survival and fertility (Kern et al., 2001). The manifestations of such relationships have been little studied, but an important step is to identify offspring phenotypes that are affected by parental age.

The cognitive function of offspring is not often considered within studies of ageing but could have significance even within human populations. For instance, people in western nations are having children at later ages (National Center for Family Growth, 2008). The advent of sophisticated fertility technology and changing social norms has facilitated childbirth by women at later ages and for those who otherwise would not have had the opportunity (Cohen et al., 1999). Some effects of parental age on children are well known (e.g. Down’s syndrome, Morris et al., 2002), but there is also good support for links to cognitive issues that may be delayed and more subtle. A great deal of evidence links increased paternal age with a greater likelihood of schizophrenia (Malaspina et al., 2002; Perrin et al., 2007) and autism (Croen et al., 2007; Glasson et al., 2004). Importantly, paternal age is also correlated with offspring IQ, with a decline particularly noted at paternal ages over 35 (Malaspina et al., 2005; Auroux et al., 1989; Saha et al., 2009). A debate over social vs. biological causes for these patterns has lasted decades. Correlations with other explanatory factors such as social environment, socioeconomic status, home stability or parental IQ confound our ability to isolate the importance of parental age on their children’s cognitive ability (Croen et al., 2007). Such studies are able to statistically identify the correlation of various factors with the cognitive health of children but lack the ability to experimentally isolate causal effects.

Accumulation of oxidative damage throughout a lifetime may explain decreased cognitive function in children with increased parental age. Reactive oxygen species are small molecules, such as oxygen free radicals, that are produced by normal metabolism and degraded by antioxidant defences (Wallace & Melov, 1998). These reactive oxygen species can damage proteins and DNA; if the DNA in germs cells is damaged by oxidative stress during ageing, this damage could be passed on to future generations.

The importance of maternal vs. paternal age effects is also under investigation. Large genetic abnormalities are more common in female germ cells, because DNA repair mechanisms in eggs are worse at detecting and fixing large mutations than the repair mechanisms in sperm (Hunt & Hassold, 2002). However, these types of abnormalities are usually lethal or have other dramatic effects. They are, therefore, unlikely to explain the gradual decline observed in offspring cognitive ability or mental health as parents age (Malaspina et al., 2005; Crow, 2000). Most new mutations in human populations can be traced to advancing paternal age (Crow, 2000), usually via single base substitutions. Mutation number may increase with age in a male’s germ cells because of reduced antioxidant enzyme activity (Cummins et al., 1994) or declines in DNA proofreading and repair during the many replication cycles (Hurst & Ellegren, 1998). As such, it has recently been proposed that the monotonic increase in the incidence of schizophrenia with paternal age is attributed to de novo mutations (Malaspina et al., 2002) or epigenetic misregulation (Perrin et al., 2007).

In this study, we address experimentally: (i) how parental age affects offspring cognitive performance, (ii) whether there are specific paternal or maternal age effects and (iii) how these effects are changed by selection for oxidative stress resistance. For several reasons, we chose to approach these problems using Drosophila melanogaster as a model system. Similarities have been observed between genes affecting D. melanogaster memory and human cognitive disorders (e.g. Bolduc et al., 2008). The D. melanogaster system permits the study of genetic factors with little social interference. Also, because we can readily measure multiple other phenotypes, it is possible to delineate the specificity of parental age effects on memory. Importantly, previous studies on Drosophila have already revealed some evidence of parental age-related deficits in offspring (Faurby et al., 2005; Hercus & Hoffmann, 2000; Magiafoglou & Hoffmann, 2003; Rogilds et al., 2005). Most of these studies observed a decreased probability of egg hatching success and larval viability related to increased maternal age. Kern et al. (2001) compared two sets of lines selected for early vs. late reproduction and found that old parents had lower egg hatching success and larval viability, but these effects were diminished in the lines selected for late reproduction. There is also genetic variation for resistance to oxidative stress in D. melanogaster (Arking et al., 2000; Mockett et al., 2001). As oxidative stress is a good candidate for the underlying cause of ageing, experimental manipulation of parental age and oxidative stress resistance in D. melanogaster may help illuminate mechanisms that contribute to or modulate this parental age effect.

Materials and methods

Fly stock and selection for paraquat resistance

Our stock line of D. melanogaster was derived from 2000 flies caught in Chavroches (France). From this stock, seven outbred lines were kept independently for 15 generations (unselected lines 1–7). Seven other lines were submitted to artificial selection for resistance to paraquat (selected lines 1–7), a powerful oxidative agent. Over 15 generations, these lines were selected for survivors that have ingested 33 mm paraquat, diluted in 5% sucrose solution, over 30 h (when 3–5 days old). Ingestion of paraquat by flies causes oxidative damage that mimics ageing (Phillips et al., 1989). Our selection lines show improved resistance to oxidative stress but no increase in longevity (Fig. 1), a not uncommon result in studies of oxidative stress resistance (Seto et al., 1990; Vermeulen et al., 2005). For each replicate, 20 flies (sex mixed) were kept until death on fresh food. The food was renewed every 3 days, and dead flies were sexed and counted. Longevity was measured for the selected and unselected lines (n = 5 replicates per line) at generation 15. We analysed the median longevity per cage using a linear mixed model which included selection regime as a fixed effect and line as a random effect nested within selection regime.

Figure 1.

 Survivorship of unselected and selected lineages of flies. (a) Survival (mean ± SEM based on variation among lines) of the unselected and selected flies when fed for 30 h the oxidative agent paraquat dissolved in sucrose solution, or sucrose solution alone, after 15 generations of selection. ***< 0.001 (n = 4 for each line) (b) Longevity of unselected and selected flies (20 males and females were housed together. n = 5 for each line).

Before each experiment described later (summarized in Table 1), we relaxed the selection for two generations to prevent maternal effects from paraquat exposure. To obtain old parents, flies were kept for 25 days on a standard cornmeal-based food (sex mixed) that was renewed every 3 days. All matings were performed within lines. After egg laying, parents were removed from the food vials and did not interact further with their offspring. F1 offspring were kept in standard food tubes until they were 3–5 days old, at which point they became subjects in one of the tests described later. Throughout the experiment, larval densities were between 100–150 individuals for 10 ml of food; this range of densities does not lead to any observable crowding effects such as increased development time or smaller body size (unpublished data).

Table 1.   Summary of the mating crosses performed for each experiment.
  1. Note: * indicates that short-term memory (STM), anaesthesia-resistant memory (ARM), long-term memory (LTM), olfactory sensitivity, locomotion, egg-to-adult viability and body mass were measured.

  2. †STM, ARM and LTM were measured.

  3. ‡STM was measured.

Parental age effectsSelectedOld ♂ × Old ♀Young ♂ × Young ♀
Young ♂ × Young ♀Young ♂ × Young ♀
UnselectedOld ♂ × Old ♀Young ♂ × Young ♀
Young ♂ × Young ♀Young ♂ × Young ♀
Sex-specific age effectsUnselectedOld ♂ × Young ♀TEST‡ 
Young ♂ × Old ♀TEST‡ 
Old ♂ × Old ♀TEST‡ 
Young ♂ × Young ♀TEST‡ 
Sex-specific selection regime effectsOldUnselect ♂ × Select ♀TEST‡ 
Select ♂ × Unselect ♀TEST‡ 
Unselect ♂ × Unselect ♀TEST‡ 
Select ♂ × Select ♀TEST‡ 

To examine transgenerational effects, we also tested 3–5-day-old F2 offspring from old and young parents from the selected and unselected lines. F2 offspring were produced from the matings of young (3–5 day old) F1 males and females. Thus, half of the F2 offspring had old grandparents (25 days old) and young parents, whereas the other half had both young grandparents and young parents.

We attempted to isolate maternal and paternal effects on offspring learning ability using various crosses. First, within each unselected line, we crossed old males (25 days old) with young females and also young males with old females (25 days old) to isolate maternal and paternal age effects. These crosses were compared to old–old and young–young crosses. To ensure the appropriate paternity for these crosses, 15-day-old females were removed from mixed-sex environments and placed in female-only environments for 10 days before mating with young males. Virgin females were used in old male–young female crosses. Second, to isolate possible sex-specific effects of our selection regime, we crossed old unselected males with old selected females, and old selected males with old unselected females. The different crosses were performed by taking old flies of one sex from unselected line i and old flies from the other sex from selected line i (1 ≤ i ≤ 7). These crosses were compared to crosses made between old parents within selection regime.

Memory training and testing

We have developed a simple training assay (similar to Tully & Quinn, 1985), in which the flies are trained to associate an odour with an aversive mechanical shock (Mery & Kawecki, 2005). Training and memory tests were performed on samples of ∼ 100 adult flies (3–5 days post-emergence, sexes mixed), raised in standard conditions. The training procedure consisted of training cycles immediately following one another [massed protocol that induces short-term memory (STM) and anaesthesia-resistant memory (ARM) formation] or separated by 20-min intervals [spaced protocol that induces long-term memory (LTM) formation]. Unlike ARM, LTM formation requires protein synthesis (Tully et al., 1994). STM was induced with three training cycles, whereas ARM and LTM were induced with five training cycles.

In each training cycle, flies were first exposed for 60 s to one odourant simultaneously with a mechanical shock (21 g vibration pulses for 1-s duration, delivered every 5 s by a test tube shaker). This period was followed by a 60-s rest period (no odour and no shock). Then, for 60 s, another odourant was delivered, without shock. The training cycle ended with a second rest period of 60 s. 3-octanol (OCT) and 4-methylcyclohexanol (MCH; both 2.0 ml L−1 of paraffin) were used as odourants. To reduce the possibility of nonassociative effects such as sensitization, each line was also trained and tested with the reverse association between the presence and absence of mechanical shock and the odours. There were, thus, two vials of ∼ 100 flies tested for each replicate for each line in each experiment. Flies trained for ARM and LTM were kept overnight at 20 °C in standard food vials.

Fifteen min (for STM assay) or 24 h (for ARM or LTM assay) after being trained, ∼ 100 flies were simultaneously transported to the choice point of a T-maze, in which they were exposed to two converging currents of air, one carrying OCT and the other MCH, and allowed to choose between the two odours for 90 s. The memory score was calculated as the difference in the proportion of individuals choosing OCT between flies trained to avoid MCH and those trained to avoid OCT. Memory scores range from -1 (all flies went towards the punished odour) to 1 (all flies avoided the punished odour). A score of 0 indicates no response to training. For statistical comparison of the memory scores (but not for graphical representation of the data), all proportions were arcsine-square-root-transformed before the analysis (Sokal & Rohlf, 1995). Memory scores were calculated for each replicate within each line and were analysed using anova. In the F1 generation, 9–10 replicates per line and parental age were tested for STM, six replicates for LTM and three replicates for ARM. In the F2 generation, three replicates per line and grandparental age were tested for STM and two replicates each for LTM and ARM. To account for the data structure, we followed the univariate approach including line as a random effect (Z subject) nested within selection regime. For crosses analysing maternal and paternal effects, five replicates per line were tested in the sex-specific age effect experiment, and three replicates per line were tested in the sex-specific selection regime effect experiment. anova was used with post hoc Tukey’s tests to examine group differences in these experiments.

Other phenotypes

We tested several phenotypes that could affect performance in the memory test or be associated with parental age. First, differences in olfactory sensitivity to the apparatus odours could clearly affect how well each group learns the association between odour and mechanical shock. Thus, olfactory sensitivity of naïve animals was tested in a T-maze with a choice between one of the odours and air. Approximately 100 flies were used in each test; one test was conducted for each odour and concentration per line within each parental age and selection regime. Both OCT and MCH are naturally aversive to the flies, so if they can detect the odour, they usually avoid it by entering the opposite arm of the T-maze. Flies were tested for 60 s at odour concentrations of 2.0 ml L−1, the same as the concentration for the memory tests, and at 0.4 ml L−1. Olfactory sensitivity was analysed using anova. Second, lower locomotor activity could reduce memory scores in the T-maze test, thus it is important to insure that this phenotype did not differ between groups. Locomotor activity of the flies was measured using the Drosophila Activity Monitoring System (Trikinetics, Waltham, MA, USA). Individual male flies (initially 3–5 days old) were subjected to 12 h/12 h light/dark cycles for 72 h. Two or three flies per line, parental age and selection regime, were tested. Locomotor activity (rate of photobeam crossing) was measured over the last 48 h and averaged for each individual flies, then analysed with anova. Prolonged observation did not show any difference among treatments in circadian rhythms (Burns and Mery, unpublished data). Third, differences in body mass between groups could indicate basic differences in condition. Thus, dry body mass was compared, within each sex, between offspring of old and young flies from the same line by weighing 10 flies for each group. Lastly, as egg-to-adult viability has previously been reported to differ between old and young parents (Kern et al., 2001), we reared 100 eggs from young or old parents in a standard food tube and counted the number of adults that emerged. We used one replicate per line, parental age and selection regime.


Parental age is associated with specific memory impairment in offspring

We tested olfactory memory in D. melanogaster using an aversive associative learning task (Mery et al., 2007; Tully & Quinn, 1985) to study how parental age may affect the cognitive performance of subsequent generations. F1 offspring of 25-day-old unselected parents had inferior STM compared to F1s of young unselected parents (3–5 days old), but there was no significant differences in two other commonly measured types of memory (anova, STM: F1,6 = 83.9, < 0.001; Fig. 2a. Anaesthesia-resistant memory (ARM): f1,6 = 0.01, = 0.9; LTM: f1,6 = 0.24, = 0.64; data not shown). The STM relationship also held when flies were trained with five, instead of three, cycles of mechanical shock (anova, F1,6 = 21.4, = 0.004). There was no significant difference between male and female offspring in STM (anova, Sex: f1,6 = 0.34, = 0.57; Age × Sex: f1,6 = 0.02, = 0.91). Comparing differences in olfactory memory between old and young parents themselves was not possible in our paradigm, because old flies had impaired olfactory acuity compared to young flies (f1,6 = 11.3, = 0.01).

Figure 2.

 Memory scores of offspring of old and young parents. (a) Olfactory short-term memory is lower in offspring (F1) and grand-offspring (F2; parents were 3–5 days old) of old (25 days old) parents relative to offspring of young (3–5 day old) parents. (b) No short-term memory deficit is observed in oxidative stress-resistant flies. Scores represent the memory score (mean ± SEM based on variation among lines). n = 9 or 10 for each line. ***< 0.001.

Strikingly, the STM impairment could be transgenerationally inherited; F2s of old parents (F1s were young) exhibited the STM impairment (anova, Grandparental age: f1,6 = 26.5, = 0.002; Fig. 2a). Thus, the memory impairment imposed by the parents is inherited by the F2s, even through young F1 parents.

To separate the effects of maternal and paternal age on STM, we crossed young males with old females, and old males with young females. We reasoned that a maternal age effect would be detected in the former cross, and a paternal age effect in the latter cross. Offspring of these two crosses showed better STM compared to offspring with two old parents (anova, f3,18 = 11.1, < 0.001; a post hoc Tukey’s test indicated that only the old male–old female cross differed from the other crosses, Fig. 3) suggesting a combinatory effect rather than a parental sex-specific effect.

Figure 3.

 Memory scores of offspring of young and old unselected parents and crosses between young and old unselected parents. Scores represent the memory score (mean ± SEM based on variation among lines). n = 5 for each line. ***< 0.001.

Oxidative stress is implicated in the memory impairment

To examine whether oxidative stress was involved in the offspring memory defect, we measured memory in flies artificially selected for resistance to paraquat, an oxidizing agent (Fig. 1a). These lines did not differ in longevity from the unselected flies described previously (anova, selection regime: F1,12 = 0.27, = 0.61; Fig. 1b), neither for males (F1,12 = 0.73, = 0.41) nor for females (F1,12 = 0.06, = 0.94). Thus, any memory differences are not because of postponed senescence. Offspring of old parents in the paraquat-resistant lines did not exhibit a STM impairment (anova, F1: F1,6 = 3.06, = 0.13; F2: F1,6 = 0.01, = 0.99, Fig. 2b), suggesting that, in the unselected flies, oxidative stress led to some alteration in the germ line that was passed on to offspring. As well, ARM and LTM did not differ between offspring of old and young selected lines (anova, ARM: F1,6 = 0.02, = 0.8; LTM: F1,6 = 0.01, = 0.90, data not shown).

We again attempted to separate paternal and maternal effects via crosses of groups that displayed or did not display the age effect; in this case, we crossed old unselected males with old selected females, and old selected males with old unselected females. The offspring of these crosses also did not show any STM deficit (anova, crosses F3,18 = 6.4, = 0.004, a post hoc Tukey’s test indicated the only the unselected male–unselected female cross differed from the other crosses, Fig. 4), and we did not detect any effect of the direction of the crosses and the contribution of both parents to the memory deficit.

Figure 4.

 Memory score of offspring of crosses between old selected and unselected parents. Scores represent the memory score (mean ± SEM based on variation among lines). n = 3 for each line. ***< 0.001.

Memory impairment is not related to low offspring quality

To strengthen these findings and analyse whether the observed parental age effect was attributed to a generalized reduction in offspring quality or rather to some specific impairments, we measured and compared other phenotypic traits. First, we tested for potential variations in olfactory acuity that might have biased the memory scores but found no differences among offspring of young and old parents of the selected and unselected lines (anova, Unselected: F1,6 = 3.03, = 0.13; Selected: F1,6 = 0.02, = 0.9 Table 2). Differences in general body condition may be indicated by differences in body mass, general activity or locomotion. However, there were no differences in dry body mass among offspring of young and old parents (anova, Males: Unselected, F1,6 = 0.2, = 0.7; Selected, F1,6 = 1.05, = 0.3; Females: Unselected, F1,6 = 0.3, = 0.6; Selected, F1,6 = 0.9, = 0.4; Fig. 5). We also did not observe any difference in general locomotor activity level between offspring of old and young parents (anova, F3,78 = 0.65, = 0.59). However, in accordance with previous work (Kern et al., 2001; Rose, 1984), we observed that egg-to-adult survival was lower for eggs laid by old flies compared to young ones [Mean % (SEM): Young unselected = 70.0 (3.7), Old unselected = 66.0 (2.2), Young selected = 76.4 (3.0), Old selected = 65.2 (3.0); anova, age: F1,12 = 14.5, = 0.002]. This decrease was equally observed for selected and unselected lines (selection × age: F1,12 = 2.6, = 0.13).

Table 2.   Olfactory sensitivity of the two test odours at two concentrations (memory tests were conducted at 2.0 ml L−1) for offspring of young and old parents in the F1 generation.
TreatmentOctanol (OCT)Methylcyclohexanol (MCH)
0.4 ml L−12.0 ml L−10.4 ml L−12.0 ml L−1
  1. Values are the percentage of flies that avoided the odour (mean ± SEM).

Unselected70.1 (2.5)73.1 (3.4)74.7 (2.0)79.7 (3.8)57.8 (3.9)51.5 (2.5)70.6 (3.2)76.4 (4.6)
Selected71.0 (2.5)77.3 (1.8)77.1 (2.1)74.4 (2.7)60.8 (3.5)61.3 (5.2)79.6 (3.5)73.9 (4.2)
Figure 5.

 Average dry body mass of female and male flies from the F1 generation. One group of 10 flies was weighed for each line.


This study experimentally isolated the influence of parental age on offspring cognitive ability. We present the first evidence of transgenerational inheritance of a memory impairment that is attributed to ageing in D. melanogaster. This effect was specific to STM and appears to be dependent on genetic variation for oxidative stress resistance. The specificity of this effect to STM, and not to ARM or LTM, is important, because it suggests that particular pathways are more sensitive to parental age effects. It should be stressed that 95% of the flies were still alive at 25 days. This high level of survival limits the opportunity for differential mortality which would have selected for long-lived parents; thus, possible negative correlations between longevity and STM (Burger et al., 2008) are unlikely to have affected our results. Decline in offspring STM, thus, appears to be a general feature of parental senescence in D. melanogaster.

Interestingly, the STM impairment effect was transmitted over more than one generation; the F2s of old grandparents also presented a decreased STM even if the parents were young. To our knowledge, this is the first demonstration of a grand-parental effect on offspring cognitive functions. How such effect may impact the evolution of ageing is likely to depend on the average age of reproduction in natural populations and on the fitness benefits of learning. The cellular damage incurred by an ageing individual through oxidative stress (Wallace & Melov, 1998) should be confined exclusively to the soma; the individual should protect its germ line from damage. In fact, ageing should not evolve unless offspring show no detrimental effects of their parent’s ageing, because lineages that do not fully protect the germ line can accumulate damage over generations and eventually go extinct (Ackermann et al., 2007). An important caveat to this point is that the ageing effect must be present at ages in which parents normally breed. The 25-day-old flies used in this study are likely older than those that typically mate in the wild. However, our study detected a ∼ 25% decline in short-term memory score, which is a substantial decrease in performance. Whether flies aged between 3 and 25 days would produce offspring with a memory impairment deserves attention, but measuring subtle cognitive deficits is difficult and may be beyond the present sensitivity of our apparatus.

Previous studies on Drosophila have demonstrated maternal age effects on egg-to-offspring survival (Kern et al., 2001), morphology (Kjaersgaard et al., 2007) and lifespan (Yilmaz et al., 2008). Maternal effects are typically more common than paternal effects owing to the tendency for females to invest a greater proportion of their resources in production or care of offspring. However, paternal effects because of environmental experiences have also been demonstrated. In the present experiment, we observed that associative learning ability is dependent on both male and female age. In general, maternal age effects are believed to be caused by epigenetic misregulation or degradation of the embryo environment. It is unlikely that this memory deficit is simply attributed to old females laying poor quality eggs. Offspring from the old female-by-young male crosses would have exhibited memory impairment if their eggs simply contained poor quality nutrients. The high-specificity of the parental age effect on STM is also inconsistent with a poor egg quality explanation. One should also notice that, in our experimental design, old and young females strongly differed not only by their age but also by their reproductive experience. While we were primarily interested in examining age effects on parents that had normal reproductive experience (at least within a laboratory context), the possible effects of differences in egg investment based on age or mating history could be examined in more detail through control of the number of matings parents experience prior to the production of offspring that will be tested.

The lack of a specific parental sex effect may result from a maternal or paternal effect having been rescued by the other gender. Maternal processes in a fertilized egg can repair permutation lesions in D. melanogaster (Vogel et al., 1985), but this is more likely to happen in eggs from females in good condition as opposed to when the females are in poor condition (Agrawal & Wang, 2008). Thus, a paternal age-specific effect might have gone undetected if maternal processes in eggs from young females repaired any defects.

Our study supports the existence of genetic variation for parental age effects. The parental age effect was not observed on lines previously selected for oxidative stress resistance. These flies had a similar lifespan as unselected flies, indicating that the absence of parental age effect was not because of a delayed mortality. The existence of genetic variation in parental age effect shows that the offspring performance decline is not a universal by-product of ageing (Mousseau & Dingle, 1991). It may also explain the large variability observed in studies on humans despite control for environmental factors and open new perspectives on the study of parental effects. The presence of genetic factors influencing the expression of transgenerational effects suggests that populations may evolve to alter expression of such effects in progeny.

A strong relationship has long been proposed between oxidative stress accumulation and ageing. Oxidative stress damages sperm DNA as well as mitochondrial and nuclear membranes (Aitken et al., 2003). Older men have a greater proportion of sperm with DNA damage as a consequence of age-associated increased oxidative stress in their reproductive tracts (Barroso et al., 2000). Experimental evidence in mice indicates that the DNA repair mechanisms of fertilized eggs are capable of repairing damage in paternal DNA that had been induced by UV radiation of sperm before fertilization (Sakkas et al., 2000). Resistance to oxidative stress is, therefore, likely to decrease DNA damage and thus transmission of deleterious mutation to the offspring.

Our work has two main implications. First, it strengthens the case that offspring phenotypes should be integrated into evolutionary theories of ageing, especially because factors such as oxidative stress resistance can affect age-specific mortality as well as offspring cognition. Second, it provides a novel approach for investigating the underlying mechanisms of the decline of cognitive abilities in children born to older parents. We anticipate that our findings will stimulate integration between mechanistic studies of age-related mortality risk and functional studies of parental age effects on the fitness of future generations. The existence of genetic variation for this parental age-related memory decline effect opens new perspectives on the study of parental effects. Considering the wide spectrum of genetic tools available in D. melanogaster, this work paves the way for future studies that will investigate the causes of this transgenerational effect.


We thank C. Moreno, F. Lagasse and R. Antoine for laboratory assistance. The work was supported by an ATIP Grant from the Life Sciences Division of the Centre National de la Recherche Scientifique and from the European Research Council under the European Community’s Seventh Framework Program (FP7/2007-2013)/ERC Grant agreement no 209540 to FM. JGB was supported by an NSERC post-doctoral fellowship.