Aging is the major risk factor driving age-dependent memory impairment (AMI) even in the absence of neurodegenerative diseases, representing one of the top health burden of the 21st century1 . Rather than gross neuronal loss characteristic of neurodegenerative diseases2 , AMI is associated with subtle but selective and region-specifc changes such as dysregulation in synapse number and structure, dendrite morphology, cellular connectivity and/ or Ca2+ homeostasis3–5 . Aging has also been postulated to be associated with a decline in autophagy/lysosomal proteolysis, a cytoprotective degradation pathway responsible for the turnover of long-lived aggregate-prone proteins and damaged organelles (importantly mitochondria). Autophagy is considered to exert protective functions against aging and age-associated diseases6 . Irrespective of the exact and potentially complex relation of aging and protective programs such as autophagy, stimulation of autophagy has been proposed to protect from aging efects in various models7–9 . We have previously demonstrated that feeding fruit fies with the naturally occurring polyamine spermidine, whose levels decline with age in a broad spectrum of organisms likely including humans7,9–12, protects from AMI. Spermidine supplementation counteracts ultrastructural and functional changes at aging synapses via a mechanism that requires a functional autophagy machinery9,11.

Aging is the major risk factor driving age-dependent memory impairment (AMI) even in the absence of neurodegenerative diseases, representing one of the top health burden of the 21st century1 . Rather than gross neuronal loss characteristic of neurodegenerative diseases2 , AMI is associated with subtle but selective and region-specifc changes such as dysregulation in synapse number and structure, dendrite morphology, cellular connectivity and/ or Ca2+ homeostasis3–5 . Aging has also been postulated to be associated with a decline in autophagy/lysosomal proteolysis, a cytoprotective degradation pathway responsible for the turnover of long-lived aggregate-prone proteins and damaged organelles (importantly mitochondria). Autophagy is considered to exert protective functions against aging and age-associated diseases6 . Irrespective of the exact and potentially complex relation of aging and protective programs such as autophagy, stimulation of autophagy has been proposed to protect from aging efects in various models7–9 . We have previously demonstrated that feeding fruit fies with the naturally occurring polyamine spermidine, whose levels decline with age in a broad spectrum of organisms likely including humans7,9–12, protects from AMI. Spermidine supplementation counteracts ultrastructural and functional changes at aging synapses via a mechanism that requires a functional autophagy machinery9,11.

 First analysis in aging mice has shown that spermidine feeding prolongs life span and exerts cardioprotective efects7,13 by 1 Department of Biology, Chemistry, Pharmacy, Freie Universität Berlin, 14195, Berlin, Germany. 2 Department of Molecular Pharmacology and Cell Biology, Leibniz Forschungsinstitut für Molekulare Pharmakologie (FMP), 13125, Berlin, Germany. 3 NeuroCure Cluster of Excellence, Charité Universitätsmedizin, 10117, Berlin, Germany. 4 Institute of Molecular Biosciences, NAWI Graz, University of Graz, 8010, Graz, Austria. 5 BioTechMed-Graz, 8010, Graz, Austria. 6Department of Anatomy and Histology, University of Veterinary Medicine Budapest, 1078, Budapest, Hungary. 7 Charité – Universitätsmedizin Berlin, corporate member of Freie Universität Berlin, Humboldt-Universität zu Berlin, and Berlin Institute of Health, 10117, Berlin, Germany. 8 Cellular Neurophysiology, Hannover Medical School, 30625, Hannover, Germany. *email: Haucke@fmp-berlin.de; stephan.sigrist@fu-berlin.de autophagic reprogramming13,14. Interestingly, a recent pilot trial study indicates that spermidine might exert benefcial efects on cognitive performance in elderly humans at risk of dementia15. Tus, taken together, spermidine is a promising candidate for anti-AMI interventions. Synaptic plasticity in the hippocampus, a pivotal cortical area for learning and memory, is particularly sensitive to aging16,17 and age-induced impairment of hippocampus-dependent memories in mice has been suggested to be linked to the downregulation of select autophagy factors8,18. Two well-studied types of hippocampal synapses are the (i) mossy fber (MF)-CA3 synapses formed by dentate granule cell (DG) axons onto CA3 pyramidal neurons and (ii) the CA3-CA1 synapses of the Schafer collateral pathway.

 MF-CA3 and CA3-CA1 Schafer collateral synapses display distinct pre- and postsynaptic forms of long-term potentiation (LTP), respectively19, and, hence, might be diferentially afected by dietary spermidine in aging animals. Here we show that both types of hippocampal synapses sufer from an aging-induced decrease in synaptic vesicle (SV) density and mitochondria abundance. Dietary spermidine supplementation protected from age-related synaptic alterations and partially restored defective LTP selectively at hippocampal mossy fber (MF)-CA3 but not CA3-CA1 synapses while preventing the aging-induced loss of neuronal mitochondria at both types of synapses. Our results unravel synapse-specifc benefcial efects of spermidine at hippocampal synapses characterized by diferential synaptic plasticity, thereby providing information for the future development of potential therapies against AMI.

Results

Dietary spermidine supplementation age-protects autophagy in the mouse hippocampal CA3 area. Based on previous studies demonstrating that spermidine supplementation stimulates autophagy7,20, we set up several independent cohorts of aged mice undergoing a six-months long spermidine treatment in their drinking water, starting at 18 months of age. To determine if spermidine (Spd) treatment was able to protect from aging-induced decline of autophagy, we analyzed the expression levels of the key autophagy protein LC3 in the hippocampus of young, aged Spd− (24 months old) and aged Spd+ (18+6 months) mice. We focused on hippocampal granule cells that form MF synapses onto CA3 pyramidal neurons. In the dentate gyrus granule cell body layer, aging tended to decrease LC3 levels. Importantly, these levels were signifcantly increased by spermidine treatment (Fig. 1a,b). 

Similar observations were made in the CA3 stratum lucidum area, where granule cells form MF-CA3 synapses. Akin to the dentate gyrus LC3 levels also tended to decline with age in the synaptic CA3 area, while spermidine treatment led to a signifcant increase in LC3 expression (Fig. 1c,d). Tese data suggest that dietary spermidine indeed boosts the steady-state levels of the key autophagy protein LC3 in hippocampal granule cells. Next, we analyzed the expression levels of WIPI2, an essential autophagy protein whose overexpression restores autophagosome biogenesis in aged dorsal root ganglion neurons21. Notably, WIPI2 immunoreactivity appeared particularly pronounced in the stratum lucidum (SL), e.g. a region where MFs are located (Fig. 1e). In the CA3 SL spermidine treatment tended to increase WIPI2 levels compared to identically aged controls (Fig. 1e,f) and WIPI2 levels were signifcantly upregulated in the spermidine treated cohort in comparison to young control mice (Fig. 1e,f). Reduced autophagic protein turnover is typically associated with the accumulation of p62, an autophagy receptor for aggregated proteins that itself is a substrate for autophagy8 . Although the number and size of p62 puncta was highly variable in aged mice, we found that spermidine treatment signifcantly reduced the p62 spot area as compared to aged matched non-spermidine treated controls (Fig. 1g–i). Collectively, these data suggest that dietary spermidine may indeed promote autophagy in the mouse hippocampal CA3 area in vivo.Dietary spermidine stabilizes and protects mitochondria at aging hippocampal synapses. Mitophagy, a specialized form of selective autophagy, is an essential mechanism for the long-term maintenance and protection of mitochondrial abundance and functional integrity.

 Spermidine previously was found to activate the degradation of dysfunctional mitochondria by mitophagy in the heart7 . We therefore ultrastructurally analyzed mitochondria in presynaptic terminals of MF-CA3 synapses (Fig. 2a–d). Although aging did not alter the percentage of mossy fber boutons containing at least one mitochondrion (Fig. S1a), the density of mitochondria in aged mossy fber boutons was signifcantly decreased compared to MFBs of young adult mice (Fig. 2b), confrming previous observations at mossy fber boutons of aged rats22. Dietary spermidine rescued the age-dependent decrease in mitochondria density at mossy fber boutons (Fig. 2b), without afecting the fraction of boutons containing mitochondria (Fig. S1a). Interestingly, adult MF terminals ofen displayed several mitochondria arranged in local networks (Fig. S1b), while in aged mossy fber boutons such networks tended to decrease (Fig. S1b). Instead, individual mitochondria appeared enlarged, a feature typically associated with a dysfunctional state (Fig. 2a–d). Aging previously was reported to lead to increases in mitochondrial size (“swelling”)23, a feature that we also observed in our aged cohort (Fig. 2a,d). Spermidine treatment appeared to protect from this age-induced swelling phenotype and tended to rescue network formation (Figs. 2a,d and S1b). When analyzing the relative bouton area covered by mitochondria in presynaptic MF terminals, we observed a trend towards rejuvenated values in the spermidine supplemented animals (Fig. 2c), though not to statistical signifcance. To see whether a similar protective efect of spermidine can be observed at other types of hippocampal synapses, we analyzed mitochondrial abundance at CA3-CA1 presynaptic terminals (Fig. 2e–g).

 At this synapse type, age tended to decrease the number of boutons containing at least one mitochondrion (Fig. 2f). Spermidine treatment signifcantly increased the fraction of mitochondria-containing boutons (Fig. 2f) and also tended to increase the bouton area covered by mitochondria (Fig. 2g). Finally, we asked whether the observed increase in presynaptic mitochondrial abundance by spermidine treatment was associated with an overall increase in mitochondrial mass. To probe this, we performed quantitative real time qPCR to measure mitochondrial DNA content of young, aged mock, and aged spermidine-treated brain tissue (Fig. 2h). Analysis of the abundance of mitochondrial DNA (normalized to nuclear β actin DNA) revealed a high variability in aged mice in comparison to young controls. Spermidine treatment signifcantly increased the relative mitochondrial DNA abundance in comparison to aged mice and young controls (Fig. 2h), indicating that spermidine supplementation can indeed increase net mitochondrial mass. Taken together, our analyses demonstrate that chronic dietary spermidine supplementation results in age-protection of mitochondria at sites of neurotransmitter release of both MF-CA3 and CA3-CA1 synapses via an overall increase in mitochondrial mass.Aging efects on MF-CA3 synapse ultrastructure are responsive to spermidine supplementa- tion. Next we asked whether aging promotes changes in ultrastructure at mossy fber synapses onto CA3 pyramidal neurons (MF-CA3; Fig. 3) and CA3-CA1 synapses (Fig. 4).

 MF-CA3 synapses display multiple release sites, large paired-pulse and frequency facilitation19,24,25 and marked synaptic plasticity of presynaptic origin19,26. MF-CA3 synapses thus markedly difer from CA3-CA1 synapses, the vast majority of which form a single synaptic contact27. Structural changes at MF-CA3 release sites during aging might therefore conceivably underlie age-dependent alterations in synaptic plasticity at this synapse. EM analysis of the ultrastructure of aged hippocampal MF-CA3 synapses revealed a signifcant reduction in synaptic vesicle (SV) density by approximately 30% when compared to synapses from young control mice, while the area occupied by mossy fber boutons remained unchanged (Fig. 3a–c), similar to previous reports for mossy fber boutons in aged rats22. While aging did not signifcantly alter the density of active zones (Fig. 3d), the length of individual active zones also signifcantly decreased with age (Fig. S2). Importantly, spermidine treatment for 6 months rescued the age-dependent reduction in SV density of aged MF-CA3 synapses to the density observed in young animals (Fig. 3b). Spermidine did not afect active zone number or active zone length when compared to age-matched controls (Figs. 3d, S2). To see whether a similar protective efect of spermidine is observed at other types of hippocampal synapses, we analyzed the ultrastructure of CA3-CA1 presynaptic terminals (Fig. 4a–d). Tis widely investigated synapse executes long-term synaptic plasticity postsynaptically28. Consistent with our fndings at MF-CA3 synapses, we found that aging decreased SV density at CA3-CA1 synapses (Fig. 4b). In contrast to MF-CA3 synapses treatment with spermidine for 6 months did not rescue these changes (Fig. 4a,b).

 Te visible presynaptic bouton profle tended to increase in aged control and spermidine treatment signifcantly increased it (Fig. 4c), suggesting that spermidine may not be able to counteract aging-induced alterations in presynaptic bouton size of CA3-CA1 synapses. Similarly, aging tended to be associated with a slightly increased active zone (AZ) length, which could not be reversed by spermidine treatment (Fig. 4d). Collectively, these data show that dietary spermidine supplementation selectively prevents aging-induced loss of SVs at MF-CA3 but not CA3-CA1 synapses in the mouse hippocampus.Dietary spermidine age-protects long-term plasticity specifcally at MF-CA3 synapses. Given these encouraging results, we investigated whether the observed protective efects of spermidine with respect to MF-CA3 synapse ultrastructure would translate into the preservation of synaptic function and plasticity. We therefore analyzed long-term potentiation (LTP), a cellular paradigm for information storage, learning and memory28, at MF-CA3 synapses. LTP has been reported to decline with age at hippocampal synapses29,30, including MF-CA3 synapses31. Indeed, when we analyzed acute hippocampal slices from aged mice (24 months) by feld recordings (Fig. 5a), we observed a severe reduction of MF-LTP in comparison to young adult mice (5 months; Fig. 5b,c). Frequency facilitation, a form of short term plasticity pronounced at MF-CA3 synapses in which a switch from low to modest stimulation frequencies lead to an increase in synaptic strength19, was unaltered (Fig. S3). 

Importantly, we found that spermidine supplementation signifcantly rescued the age-dependent decrease of MF-LTP in 24 months-old animals (Fig. 5b,c). When we analyzed EPSP/fber volley curves, we observed an age-dependent increase in basal neurotransmission at MF-CA3 synapses that, similar to LTP, was rescued to juvenile levels by spermidine treatment (Fig. 5d,e). As a control we also studied CA3-CA1 synapses that exhibit largely postsynaptic forms of LTP. Consistent with previous studies29,30, we observed a trend towards lower LTP of fEPSP slopes at CA3-CA1 synapses of aged animals (Fig. 6a,b). Chronic spermidine administration did not rescue this trend towards an age-dependent decrease in CA3-CA1 LTP (Fig. 6b), consistent with the lack of efect of spermidine with respect to SV density (Fig. 4b). When we analyzed CA3-CA1 EPSP/fber volley curves, we did not observe any signifcant changes with either aging or spermidine treatment (Fig. 6c,d). Taken together, our results reveal a synapse-specifc restoration of presynaptic LTP at MF-CA3 synapses by dietary spermidine in aged mice.Discussion Elucidating the cellular and molecular mechanisms that underlie the aging of brain function are a major challenge of neuroscience. Finding dietary regimes to fundamentally reduce the rate at which these processes occur is equally relevant. Recent reports suggest that with aging neuronal autophagy might decline6,8 and, importantly, that restoring hippocampal autophagy in aged mice allows for a rescue of age-induced memory impairment8,32. In the present study, we show that dietary supplementation of spermidine, a substance usually declining with age, can restore levels of core autophagic proteins.

 We found that spermidine treatment for 6 months starting late in life (at 18 months), a feeding scheme previously shown to extend life-span of aged mice7 , enhanced clearance of the autophagy substrate p62 and increased expression of key autophagy proteins (LC3, WIPI2). Tese data are consistent with previous studies in other organ systems demonstrating that spermidine administration can stimulate autophagic fux20,33–35 and a prominent autophagic role for WIPI2 in aged neurons21. Tese data thus indicate that chronic spermidine supplementation might promote autophagy in the aged mouse brain. Age-dependent cognitive decline seems to be tightly associated with subtle synaptic changes but is not mandatorily linked to neuronal cell death3 , a prominent feature of neurodegenerative disorders rather than normal aging. Changes in the density of the pool of synaptic vesicles distant from active zones (e.g. the recycling and reserve pools) have been previously described at hippocampal Schafer collateral synapses of aged rats36 and at MF-CA3 synapses22. Our data confrm that aging indeed leads to decreased SV density at both MF-CA3 and CA3-CA1 synapses (Figs. 3a,b and 4a,b). Most importantly, we found that spermidine treatment late in life could rescue the age-dependent reduction in SV density and in LTP at MF-CA3 synapses. Strikingly, this phenotype was synapse-specifc as spermidine administration failed to elicit postsynaptic LTP at and to restore SV density at CA3-CA1 synapses (Figs. 5b,c and 6b). 

Spermidine supplementation also rescued the age-dependent increase in EPSP/fber volley amplitudes observed at MF-CA3 synapses (Fig. 5d,e), which might indicate higher basal neurotransmission at these synapses with age, as previously reported for aged synapses in the Drosophila olfactory circuitry11. Our results are in contrast with a recent study at aged MF-CA3 synapses in rats using 4-6 week old animals as young controls31, which might refect developmental changes still occurring at this early age. Te discrepancy might be due to the diferent species and age windows used. Diferent from CA3-CA1 synapses, MF-CA3 synapses are characterized by presynaptic plasticity, a process controlled by presynaptic cAMP and cAMP-dependent protein kinase (PKA) via Rab3a, a SV protein that regulates SV fusion19,37,38. Interestingly, we observed similar benefcial efects of spermidine at Drosophila mushroom body synapses that are also characterized by cAMP-driven presynaptic plasticity to form new memories9,11,39. Our results might indicate that spermidine acts on a mechanism that specifcally regulates the SV pool at MF terminals. Tis is not unlikely considering that MF-CA3 synapses are remarkably diferent in term of structure, release probability and plasticity from CA3-CA1 synapses19,24–27,40. MF-CA3 synapses display multiple release sites, low release probability41, strong facilitation and execute long term plasticity via presynaptic mechanisms. In contrast, CA3-CA1 synapses form a single synaptic contact and are characterized by higher release probability42,43. Past studies have demonstrated a diferential regulation or role of proteins involved in SV cycling at MF-CA3 versus CA3-CA1 synapses, supporting our fndings. 

Te Rab3-interacting protein Rabphilin is a PKA efector, that controls the recovery of the ready releasable pool (RRP) of SVs following extensive synaptic activity44. Active PKA has been shown to diferentially phosphorylate Rabphilin at MF vs. CA3-CA1 synapses, suggesting a MF-specifc mechanism regulating SVs exocytosis upon RRP depletion45. We thus speculate that spermidine might exert benefcial efects specifcally at synapses executing synaptic plasticity via presynaptic mechanisms, possibly including autophagic turnover of presynaptic components. While these mechanisms may be complex, our data showing that spermidine supplementation prevents aging-induced defects in presynaptic mitochondria (Fig. 2), suggests a possible role for mitochondrial maintenance in presynaptic Ca2+ homeostasis46–48. Interestingly, spermidine has been found to increase the rate and afnity of Ca2+ uptake in brain mitochondria49.  Tis may be of particular importance for pre- but not postsynaptic forms of LTP, i.e. at MF-CA3 synapses. Future studies will be needed to address this possibility. We would like to note that efects of spermidine on CA1 mitochondrial morphological parameters could not be observed in another cohort with very diferent housing conditions (smaller group size, diferent environmental enrichment) in a diferent animal facility which also showed less clear aging efects (data not shown). Tis suggests that environmental factors can impinge on neuronal ultrastructure of the aging brain per se or on the efects of spermidine in particular. Defning these factors should also be subject of future research.

 All experiments were performed in accordance with the relevant guidelines and regulations.Spermidine supplementation. C57BL6 WT mice were purchased from Janvier Labs (C57BL/6 J:Rj males). Spermidine supplementation at a fnal concentration of 3mM in drinking water started late in life (18 months of age) for 6 months7 . A more detail description of housing conditions is found in the Supplemental Information. Notably, the efects of spermidine on CA1 mitochondrial morphological parameters were not observed in another cohort with very diferent housing conditions (smaller group size, diferent environmental enrichment) in a different animal facility (data not shown), All animal experiments were approved by the animal welfare committee of Charité Universitätsmedizin Berlin, Leibniz Institut für Molekulare Pharmakologie (FMP) and the Landesamt für Gesundheit und Soziales Berlin and by the Bundesministerium für Wissenschaf, Forschung und Wirtschaf, BMWFW, Austria: BMWF66.007/0011-II/3b/2013, BMWFW-66.007/0002-WF/V/3b/2015. All experiments were performed in accordance with the relevant guidelines and regulations.Electron microscopy. Following transcardial perfusion with a mixture of 2% formaldehyde (FA) and 2% glutaraldehyde in 0.1 M phosphate bufer (PB, pH 7.4) under anesthesia, brains were post-fxed overnight in 4% FA at 4 °C. Ultrathin sections were examined with a JEM-1011 transmission electron microscope (JEOL, Tokyo, Japan). Data collection was performed blindly. A more detail description is found in the Supplemental Information. Ultrastructural analysis. We analyzed six-seven mice per group.

 Visible bouton area, synaptic vesicles, active zones, mitochondria and their visible area were manually annotated in ImageJ (1.48 v). A more detail description is found in the Supplemental Information. Electrophysiology. For recordings of MF-fEPSPs and CA3-CA1 fEPSP, mice were anesthetized with isofurane overdose and transcardially perfused with ice cold dissection artifcial cerebrospinal fuid (ACSF)50 or directly decapitated, respectively. 350 μm-thick (MF recordings) or 300 µm-thick (CA3-CA1 recordings) hippocampal sections were incubated at 35 °C for 30 minutes immediately afer preparation and kept in a resting chamber at 22–24 °C (MF recordings) or at RT (CA3-CA1 recordings) for at least an hour before use. A detailed description of all the electrophysiological experiments is found in the Supplemental Information. Statistic. Statistic was performed with IBM SPSS Statistics version 25 (IBM) or with GraphPad Prism (GraphPad). Data distribution was assessed by Kolmogorov-Smirnov normality test and by inspecting histograms and normality Q-Q plots. Data were analyzed with one-way ANOVA followed by Tukey post-hoc test or by Kruskal-Wallis test followed by Mann Whitney U test, Bonferroni corrected with α set to 0.016667 to account for multiple comparisons, unless otherwise stated. Only two tail p-values were considered. Values are expressed as mean±SEM, n indicates the number of boutons analyzed for EM analysis or the number of brain slices for electrophysiological experiments, unless otherwise stated. A detailed description of the statistical analysis is found in the Supplemental Information.