Age-induced memory impairment (AMI) is a common condition that is characterized by symptoms of cognitive decline that occur as part of the aging process. To date, molecular interventions to counteract AMI remain largely elusive, with the long lifespan of many animal models being a major limitation for studying age-related memory impairment. Drosophila, with its comparatively short lifespan and advanced genetics, is an ideal model system for unraveling the molecular mechanisms associated with AMI and testing putative means for preventing AMI. The age-dependent decline of (aversive) olfactory memory in Drosophila serves as an established model for AMI1–4. In this learning procedure, groups of flies are trained by presenting them with an odor that temporally coincides with the application of electric shocks and presenting a second odor without this punishment. Later on, flies can choose between the two arms of a T-maze containing either the odor that was paired with the electric shocks or the neutral odor. Learning is assessed by calculating the relative proportion of flies avoiding the punished odor. This type of associative olfactory memory is usually maintained over several hours5. Studies in various model organisms have implicated autophagy as a crucial regulator of the aging process6. Autophagy is a process of cellular self-digestion in which portions of the cytoplasm are sequestered in double- or multi-membraned vesicles (autophagosomes) and then delivered to lysosomes for bulk degradation.

Age-induced memory impairment (AMI) is a common condition that is characterized by symptoms of cognitive decline that occur as part of the aging process. To date, molecular interventions to counteract AMI remain largely elusive, with the long lifespan of many animal models being a major limitation for studying age-related memory impairment. Drosophila, with its comparatively short lifespan and advanced genetics, is an ideal model system for unraveling the molecular mechanisms associated with AMI and testing putative means for preventing AMI. The age-dependent decline of (aversive) olfactory memory in Drosophila serves as an established model for AMI1–4. In this learning procedure, groups of flies are trained by presenting them with an odor that temporally coincides with the application of electric shocks and presenting a second odor without this punishment. Later on, flies can choose between the two arms of a T-maze containing either the odor that was paired with the electric shocks or the neutral odor. Learning is assessed by calculating the relative proportion of flies avoiding the punished odor. This type of associative olfactory memory is usually maintained over several hours5. Studies in various model organisms have implicated autophagy as a crucial regulator of the aging process6. Autophagy is a process of cellular self-digestion in which portions of the cytoplasm are sequestered in double- or multi-membraned vesicles (autophagosomes) and then delivered to lysosomes for bulk degradation.

 In C. elegans, increased autophagy is necessary for lifespan extension by reduced insulin-like signaling7 and dietary restriction8. Similarly, increasing autophagy specifically in neurons can extend the lifespan of flies9, whereas reducing autophagy shortens lifespan and gives rise to neurodegeneration9,10. Although autophagy is a key regulator of the aging process, its role in cognitive aging has not yet been addressed. We found that the levels of simple polyamines (spermidine and putrescine) decreased in the heads of aging Drosophila, consistent with a decline in olfactory aversive memory in aging flies. Notably, restoration of polyamine levels by dietary spermidine supplementation suppressed AMI. Spermidine administration prevented the age-associated decrease of autophagy, and genetically induced deficits in the autophagic machinery occluded spermidinemediated protection of AMI. In addition, we found that the effects of spermidine on memory were not a result of generically improved health, but instead reflected neuron-intrinsic regulations. These results suggest that autophagy is important for cognitive aging and that polyamines, endogenous metabolites, are candidate substances for treating AMI.

RESULTS

Polyamine levels decline in aging Drosophila brains

A decrease in polyamine levels has been reported in aged brains of rodents and humans11,12. To investigate the role of polyamines in AMI, we began by measuring polyamine levels in heads of aged Drosophila, and analyzed the chemically related (and inter-convertible) species spermidine, putrescine and spermine (for a review, see ref. 13; Fig. 1a). In fact, both spermidine and putrescine levels were markedly reduced in the heads of 15-d-old flies compared with young flies (Spd− 1–3 d old; Online Methods and Fig. 1b,c). To determine whether simply feeding spermidine to flies would be sufficient to restore its expression in aged heads, we administered either 1 or 5 mM spermidine to isogenized wild-type Drosophila (w1118) in standard fly food (Spd1mM+ or Spd5mM+, respectively). Spermidine levels (in fly heads) were largely protected by this treatment from age-dependent decline (Fig. 1b). Simultaneously, putrescine levels were greater than those in normal juvenile flies (Fig. 1c), indicating that the cellular uptake of spermidine is accompanied by either an increased conversion into putrescine or inhibition of the endogenous conversion of putrescine to spermidine (Fig. 1a). This observation is consistent with the previously reported homeostatic control of polyamine levels14. On the other hand, levels of spermine declined considerably with age in both Spd+ as well as Spd− flies (Supplementary Fig. 1).

Potentially, polyamine effects could be mediated through developmental changes, such as during critical periods in early adulthood. To address such putative developmental effects, we shifted flies between spermidine-containing and spermidine-free food. When tested on day 30, in flies fed with spermidine only between days 0 and 20 (with spermidine supplementation being withdrawn for the last 10 d before testing; Fig. 2f,g), we found that the levels of polyamines (both spermidine and putrescine) declined to levels comparable to those of controls (no spermidine supplementation for days 1–30; Fig. 2f,g), as was the memory when tested at day 30 (Fig. 2e). In contrast, in flies fed with spermidine for the last 10 d before testing, levels of both polyamines (spermidine and putrescine) rose (Fig. 2f,g), and memory was partially, but substantially, restored (Fig. 2e). The fact that restoring polyamine levels during the 10 d before testing prevented AMI rules out the idea that the effects of spermidine-feeding on suppression of AMI are a result of affected developmental processes. Given that spermidine feeding promotes longevity14, it might be argued that protection of memory is a byproduct of increased life expectancy and generally improved health. Thus, we asked whether spermidine feeding preserves the function of the fly nervous system in all respects or whether it rather has a more specific effect on memory. In our learning assay, we found that 30-d-old naive flies to exhibit decrease odor avoidance scores compared with 3-d-old flies, whereas the shock reactivity of these naive flies did not change (Table 1), consistent with previous AMI studies1,2. Notably, spermidine-feeding had no influence on this age-dependent decline in odor avoidance scores (Table 1). 

What might be the mechanistic underpinnings of spermidinemediated protection from AMI? Spermidine administration induces autophagy in several model organisms14 and autophagy is crucially important for spermidine-mediated promotion of lifespan in yeast, C. elegans and Drosophila14. Notably, reduced basal autophagy in the nervous system of mice and flies has been shown to cause neurodegeneration9,18–20. In addition, the expression of several key genes in the autophagic pathway has been reported to decline with aging in the brain of humans and flies, potentially increasing neuronal vulnerability to the toxic effects of protein aggregates6,9. Thus, we investigated whether autophagy might be critical for spermidine-mediated protection from AMI. To address this, we first assayed the levels of Atg8a (a widely used marker for autophagy) in western blots from fly head extracts21. Consistent with a previous report9, Atg8a protein levels were considerably reduced in heads from 30-d-old Spd− flies compared with 3-d-old Spd− flies (Fig. 4a). Spermidine administration, however, blocked this age-related decline in Atg8a protein levels effectively (as seen by protein levels in the heads from Spd5mM+ flies; Fig. 4a). Suppression of autophagy has been associated with the accumulation of ubiquitinated protein aggregates9,10,22. In fact, we observed that the ageassociated increase in the amount of poly-ubiquitinated proteins was largely blocked in heads of spermidine-treated Drosophila (Spd5mM+; Fig. 4b).We next asked whether these ubiquitinated proteins were really being degraded by autophagy.

 The p62 family of proteins is closely associated with protein inclusions containing ubiquitin, as well as with key components of the autophagy pathway, thereby mediating autophagic clearance of ubiquitinated proteins19,23. In addition, ref(2)P, a Drosophila homolog of p62, was recently reported to accumulate with ubiquitinated neural protein aggregates in aged wild-type flies and autophagy mutants22. For a nervous system– specific readout, we stained brains from flies of different ages with antibody to ref(2)P. As expected, ref(2)P levels increased with age, and spermidine administration suppressed this age-dependent increase (Fig. 4c–g). Thus, spermidine administration seems to prevent the accumulation of ubiquitinated proteins, most likely as a direct consequence of enhancing autophagy in aged Drosophila brains.We then tested whether autophagy is functionally required for spermidine-mediated protection from AMI. Both Atg7 and Atg8 are essential for autophagy in Drosophila9,10,24–26. Given that Atg7−/− and Atg8a−/− flies have a mean lifespan of only 30 d (ref. 10), we decided to test memory in both mutants at 20 d of age (Fig. 5). We found that Atg7−/− flies (Fig. 5a,c) showed reduced memory scores at a young age (3 d of age; Fig. 5a), which further declined relative to controls at later time points (20 d of age; Fig. 5c). 

 However, we reasoned that if transcriptional reprogramming is causally involved in the protective effects of polyamines, then it might precede, or at least be concomitant with, its effects on memory. In another experiment, we analyzed the kinetics of polyamine decline and memory. A substantial decline of both polyamines (spermidine and putrescine), together with memory, was observed by 10 d of age, which only slightly decreased further by 15 d (Supplementary Fig. 4). We then carried out RNA-seq on head extracts prepared from 3-d-old and 10-d-old Spd5mM+ flies and compared them with age-matched Spd− flies. Sequenced reads were then aligned to the Drosophila genome with high stringency, allowing only one mismatch per read, and an average of 96% of aligned reads mapped to exons (Supplementary Table 1). Thereafter, the number of reads mapping to each gene was quantified and normalized for library size using DESeq29. Hierarchical clustering of the normalized reads revealed a high degree of consistency between the biological replicates, clearly showing that spermidine-treated samples clustered away from untreated samples (Fig. 6a). Using an established statistical method based on the negative binomial distribution of sequenced reads29 (Supplementary Fig. 5), we then identified large global changes in gene expression induced by spermidine feeding, with 2,051 genes and 4,076 genes being modulated by a factor of 1.5-fold (P < 0.05) in the heads of 3-d-old and 10-d-old flies. respectively (Fig. 6b and Supplementary Table 2). Notably, 84% of the genes that were differentially regulated after 3 d were also altered at 10 d of age (Fig. 6c).

To provide insights into the function of the spermidineregulated brain transcriptome, we subjected all of the genes found to be either up- or downregulated by spermidine feeding to gene ontology enrichment analysis30 (Supplementary Table 3). The genes that were modulated by spermidine feeding in both 3-d-old and 10-d-old flies showed strong enrichment for GO terms, such as response to starvation, response to oxidative stress, phagocytosis and cellular homeostasis (Supplementary Table 3.3). This is consistent with the idea that general protective effects are mediated by spermidine administration. In addition, we also found over 2,000 transcriptional changes that were specific for 10-d-old flies (Supplementary Tables 2.1 and 2.2). On one hand, the GO term ‘aging’ was considerably enriched in 10-d-old Spd5mM+ flies, indicating that ageprotective gene functions are induced when polyamine levels are remain high during the onset of aging. On the other hand, transcriptional changes that are specific for neuronal genes appear to have a role in Spd5mM+ flies, with ‘neuron differentiation’ and ‘neuron development’ being the top two enriched GO terms (Supplementary Table 3.2). In addition, several genes (14 in total) with the GO annotation ‘learning’ were found to be specifically upregulated in 10-d-old Spd5mM+ flies when compared with untreated flies (Supplementary Tables 2.2 and 3.2). Notably, several of the learning-related genes (along with other genes) that were modulated by spermidine feeding were subsequently analyzed and validated by quantitative reversetranscription real-time PCR (Supplementary Fig. 6). 

Although the decline of polyamines might be regarded as an established biochemical correlate of aging, the causal relationship to age-related deficits in cognitive functions has not been established. Simple dietary supplementation of spermidine allowed us to restore polyamine levels in the heads of aged flies (both spermidine and putrescine) to those seen in juveniles. This simple procedure was sufficient to effectively protect both short- and intermediate-term olfactory memory from age-induced decline. The effects of spermidine to the protection of memories were specific in several regards. First, spermidine feeding had no effect on memory in young flies, either in terms of short or intermediateterm components, arguing against the possibilty that spermidine might function as a general memory enhancer (Fig. 2a,b). Second, ITM has two components, with aging strongly affecting ASM, but not ARM1. We found that spermidine feeding had only a negligible effect on ARM; instead, spermidine administration specifically prevented the age-related decline of ASM (Fig. 2b,d). Third, polyamine restoration appeared to specifically suppress AMIs, as olfactory avoidance scores of naive flies (Table 1) and locomotion activity (Supplementary Fig. 3) declined with age in both Spd+ flies and age-matched Spd− flies. Fourth, in flies fed with spermidine for the last 10 d before testing (Spd−, 1–20 + Spd5mM+, 21–30), polyamine levels (spermidine and putrescine) increased (Fig. 2f,g), and memory was considerably restored (Fig. 2e), indicating that AMI suppression by spermidine administration is not a result of altered development.

 Furthermore, expressing Odc-1 in just Kenyon cells was sufficient to ameliorate AMI (Fig. 7e,f). These findings indicate that spermidine-mediated suppression of AMI is not executed via its effects on systemic regulations or a generally improved health of the organism, but rather result from an intrinsic regulation of a small fraction of neurons. The formation of memory requires dynamic changes in the neurons, including synapse formation and synaptic plasticity, steered by regulated protein synthesis and equally important protein degradation. In fact, the execution of effective quality control over proteins appears to be important for neurons to maintain proper neuronal physiology and functioning35. The process of autophagy is an important route for removing misfolded proteins and damaged organelles from cells via lysosomal-mediated bulk degradation20. Spermidine has been shown to operate as a natural inducer of autophagy in various model systems, including yeast, C. elegans, Drosophila and mice14,36. In fact, we found that spermidine feeding alleviated the age-induced dysfunction of autophagic machinery in flies, thereby preventing the accumulation of poly-ubiquitinated proteins and ref(2)P (Fig. 4). What might be the mechanism by which spermidine administration prevents the decline of autophagy in aged Drosophila? Spermidine treatment induces autophagy in enucleated cells within a few hours as effectively as the well-known autophagy inducer rapamycin27, arguing for fast, post-transcriptional regulation. On the other hand, using RNA-seq, we found that positive regulators of autophagy (such as Atg1a)37,38 were upregulated by spermidine feeding (Supplementary Fig. 6 and Supplementary Table 2).