Age-related decrease in appetitive associative memory in fruit flies

In animals and humans alike, memory scores typically decrease from adolescence through adulthood and old age into senescence, reflecting a combination of a wide variety of processes (reviewed in Bishop et al., 2010; Grady, 2012). These include degrading mnemonic ability and sensory-motor faculties, motivational and metabolic changes, as well as changes in cognitive strategy. Regardless of the relative contribution of these processes, the age-related decrease in memory scores is observed not only across species but also across experimental paradigms, sensory modalities, valence and temporal domains, and in both sexes. It was therefore surprising to realize that for Drosophila melanogaster, a versatile study case for understanding ageing and age-related memory impairment (reviewed in He and Jasper, 2014; Piper and Partridge, 2018), most studies have focused on age-related decreases in aversive, odour–electric shock associative memory, following the discovery of such decreases by Tamura et al. (2003) (for earlier reports in other, likewise aversively motivated paradigms, see Fresquet and Médioni, 1993; Guo et al., 1996; Savvateeva et al., 1999). In contrast, Tonoki et al. (2020) concluded that there is no age-related decrease in odour–sucrose associative memory in Drosophila (for details, see Discussion). Here, we report such a decrease, using a study design comprising three experimental conditions with five age cohorts each.

Drosophila melanogaster of the Canton Special wild-type strain were kept in mass culture on standard cornmeal–molasses food at 60–70% relative humidity and 25°C temperature under a 12 h:12 h light:dark cycle. To establish five 6 day age cohorts, approximately 500–1000 freshly hatched adult flies at a sex ratio of approximately 1:1 were collected in food vials and kept for 1–6, 7–12, 13–18, 19–24 or 25–30 days. To prevent the flies from getting stuck too frequently in the food once the food was churned up by their larval offspring, and to remove them long before their own offspring hatched as adults, the flies were placed on fresh food vials every 2–3 days.

The behavioural experiments followed standard methods as described in detail in Yarali and Gerber (2010), unless mentioned otherwise. In brief, 16–20 h before training started, randomly selected, mixed-sex cohorts of flies were transferred from their food vials to a vial without food but with only a moist tissue paper soaked with tap water (‘wet starvation’). Training and testing of the flies then took place using a custom-made apparatus (CON-ELEKTRONIK, Greussenheim, Germany) at 23–25°C and 60–80% relative humidity. Training was performed in white room light, and testing in red light invisible to the flies. As odorants, 50 µl benzaldehyde (BA) or 250 µl 3-octanol (OCT) (CAS 100-52-7 and 589-98-0, respectively; Merck, Darmstadt, Germany) were applied undiluted to 1 cm deep Teflon containers of 5 or 14 mm diameter, respectively. From these, odour-loaded air could be shunted into the permanent airstream flowing through the apparatus.

Flies were trained and tested en masse, in cohorts of ∼100 individual animals each. For training, the flies were gently loaded into the setup. After 1 min, the flies were transferred to a training tube lined with filter paper that had been soaked the day before with 2 ml of 2 mol l⁻¹ sucrose (d-sucrose; CAS: 57-50-1, Roth, Karlsruhe, Germany) solution and left to dry overnight. At the same time the rewarded odour (in Pavlovian terminology, the conditioned stimulus or CS+) was shunted into the permanent airflow running through this tube. After 45 s, the odour presentation was terminated, and after an additional 15 s, the flies were removed from the sucrose-containing tube. At the end of a 1 min waiting period, the flies were transferred into another training tube, which was empty. At the same time and thus in the absence of the sucrose reward, the reference odour (CS−) was presented. After 45 s, the odour presentation was finished, and 15 s later the flies were removed from this training tube. As indicated in the Results, we used either one or two such training trials. In a between-subjects experimental design, the use of BA and OCT as CS+ and CS− was balanced across repetitions of the experiment, resulting in reciprocally trained groups of flies. In addition, the sequence of the CS+/CS− presentations during training trials (BA+/OCT or OCT/BA+ when BA served as CS+ and OCT as CS−, and OCT+/BA or BA/OCT+ for the reciprocal cases), the position within the setup at which BA and OCT were presented during training and during the test (see next paragraph), and the sequence of running the experiment for the reciprocal training groups were balanced (Fig. S1A) (see fig. S1B in König et al., 2019).

To measure 24-h memory, the flies were tested the next day. In order to ensure reasonable survival rates in these cases, the flies were removed from the training apparatus and first kept for 1 h on standard food; only then were they wet-starved until the test took place the following day.

Two-tailed non-parametric statistics were used throughout. To compare across multiple independent experimental groups, Kruskal–Wallis tests (KW tests) with subsequent pair-wise comparisons by Mann–Whitney U-tests (U-tests) were used (Statistica 13, RRID: SCR_014213, StatSoft Inc., Tulsa, OK, USA). For comparisons of data from an experimental group with chance levels (zero), one-sample sign tests (OSS tests) were applied. To keep the within-experiment error rate below 5%, a Bonferroni–Holm correction for multiple comparisons was employed. The experimenter was blind to experimental groups during the counting of flies after the experiments. The data are presented as box plots showing the median as the middle line, the 25% and 75% quantiles as box boundaries, and the 10% and 90% quantiles as whiskers. Sample sizes were chosen on the basis of previous, similar studies aimed at detecting partial decrements in relatively low memory scores. The results of the statistical tests and the source data of all experiments performed are documented in Table S1.

All procedures comply with applicable law and ethics regulations.

Flies in 6-day age cohorts from the day of adult hatching to 30 days after hatching were starved for 16–20 h and received a single trial of differential conditioning such that one odour was paired with a sucrose reward, whereas the other odour was not. Then, the flies were tested for their relative preference between the two odours. From these preferences, associative memory scores were calculated that reflect their learned preference for the previously reward-associated odour (Fig. 1A) (using a between-subjects reciprocal design, the chemical identity of the odours was balanced across repetitions of the experiment). As shown in Fig. 1B, memory scores decreased across age cohorts: relative to the youngest cohort, memory scores were reduced from the 13–18 day age group onwards (see Fig. S1Bi for the underlying preference scores). The results were similar when two conditioning trials were used (Fig. 1C; Fig. S1Bii) (in this case a decrease was significant from the 7–12 day age cohort onwards) and also when testing was carried out 24 h after such two-trial differential conditioning (Fig. 1D; Fig. S1Biii). Notably, memory scores remained significantly above chance levels, i.e. were larger than zero, for all cases with the exception of the 24-h memory scores in the very oldest age cohort.

Memory scores for the youngest age cohort differed across the three experimental conditions, with a moderate gain in memory scores immediately after two-trial versus one-trial training and with the expected decrease in two-trial memory assessed 24 h later (KW test: H₂=45.62, N=127, P<0.05; U-tests: one-trial versus two-trial memory tested immediately after training U=630, P<0.05; one-trial memory tested immediately versus two-trial memory assessed 24 h later U=431, P<0.05/2; two-trial memory tested immediately versus two-trial memory assessed 24 h later U=114, P<0.05/3) (Schwaerzel et al., 2003; Kim et al., 2007; Tonoki et al., 2020). We note that previous reports did not uncover gains in memory scores from a second reward training trial (e.g. Colomb et al., 2009). In order to compare the rate of the age-related decrease in memory scores across these three experimental conditions, we therefore normalized all the memory scores for each of the three experimental conditions to the median of its respective youngest age cohort. This showed that the age-related decrease in memory scores was strongest for two-trial 24-h memory, in particular across days 7–18 (Fig. 1E). Of note is that the more partial decrease in memory scores for one-trial and two-trial memory in this 7–18 day age window assessed immediately after training was more pronounced in females than in males (Fig. 2; Fig. S1Ci,ii) (for the even older age cohorts, a separation by sex was not carried out, because the number of surviving flies, in particular the number of males, appeared to be too low, leading to excessively variable memory scores).

Thus, across experimental conditions, the present study reveals an age-related decrease in appetitive associative memory scores during the first 2–3 weeks after adult hatching (Fig. 1E), which is more pronounced in females than in males (Fig. 2). While the present study does not attempt to disentangle what are arguably multiple chains of cause and effect leading to this decrease, it suggests that future research could focus on this relatively early time window when looking into the interplay of sensory, motor, motivational, metabolic, mnemonic and cognitive processes contributing to age-related decreases in appetitive memory scores. For the practitioner, such an early time window of the effect is good news because the experiments take less time than those focusing on later stages, and require less attention to fly husbandry. Also, the contribution of non-target age-related effects is conceivably more pronounced for older stages. However, decreases in memory scores at this relatively young age might be of limited translational validity for processes at an older age that are more relevant from a biomedical point of view.

In contrast to the present study, Tonoki et al. (2020) concluded that there is no age-related decrease in memory scores for the association of odours with sucrose in Drosophila. Although their experimental paradigm is in principle similar to the one used here, there are a number of procedural differences. These include aspects of the starvation procedure and odour presentation, the light conditions during training, the duration of the feeding period between training and testing for the measurement of 24-h memory, and the sequence of CS+/CS− presentations during training. These sequences were balanced in the present study (Fig. S1A) but always started with CS− in Tonoki et al. (2020). For the relatively young age cohorts of 2 versus 10 days of age, moreover, Tonoki et al. (2020; see their fig. 1E–G) used a more extended starvation period of 40 h as compared with the 16–20 h used in the present study, which did uncover a decrease in memory scores in flies over roughly that age range (Fig. 1E). Strikingly, using a shorter starvation period of 18 h, similar to the 16–20 h used here, Tonoki et al. (2020) found higher memory scores (sic) for 30-day-old than for 10-day-old flies (see fig. 1C in Tonoki et al., 2020) (compare with Fig. 1E). Which of the above-mentioned differences in procedure might account for the discrepancies in the results must remain unresolved for now.

Concerning the more pronounced age-related decrease in memory scores in females than in males that we report (Fig. 2), it seems significant that females are less susceptible to starvation than males (Robinson et al., 2000). Accordingly, for an older age at least (7–18 days), our data suggest that survival rates for the 24-h memory test are better in female than in male flies (Fig. S1Ciii). Thus, at that age, the nutritious sucrose reward might be less motivationally salient, and indeed less rewarding, to the females than to the males – despite the ‘costs’ of reproduction and mating in females (Chippindale et al., 1993; Chapman et al., 1995).

Given that the decrease in memory scores uncovered here both takes place relatively early in adult life (Fig. 1E) and is more pronounced in females than in males (Fig. 2), one wonders whether female life-history traits and in particular the powerful post-mating changes in food-related motivation play a role (see discussion in Kubli, 2010). Indeed, Scheunemann et al. (2019) have shown that memory scores after aversive associative learning between odours and electric shock are more negative for mated than for non-mated females. As the proportion of mated females increases across early adult life, it seems possible that the less positive memory scores observed here for 7–18 day old than for younger female flies may partially reflect the fact that already mated, older females establish a relatively more negative ‘take home message’ both from the appetitive training procedure employed here and from the aversive training procedure of Scheunemann et al. (2019).

We thank Bettina Kracht and Anna Ciuraszkiewicz-Wojciech for technical assistance, Bert Klagges, Thomas Niewalda and Naoko Toshima for comments on earlier versions of the manuscript, and R. D. V. Glasgow, Zaragoza, Spain, for language editing.