Glycogen phosphorylase inhibition improves cognitive function of aged mice

Abstract Inhibition of glycogen breakdown blocks memory formation in young animals, but it stimulates the maintenance of the long‐term potentiation, a cellular mechanism of memory formation, in hippocampal slices of old animals. Here, we report that a 2‐week treatment with glycogen phosphorylase inhibitor BAY U6751 alleviated memory deficits and stimulated neuroplasticity in old mice. Using the 2‐Novel Object Recognition and Novel Object Location tests, we discovered that the prolonged intraperitoneal administration of BAY U6751 improved memory formation in old mice. This was accompanied by changes in morphology of dendritic spines in hippocampal neurons, and by “rejuvenation” of hippocampal proteome. In contrast, in young animals, inhibition of glycogen degradation impaired memory formation; however, as in old mice, it did not alter significantly the morphology and density of cortical dendritic spines. Our findings provide evidence that prolonged inhibition of glycogen phosphorolysis improves memory formation of old animals. This could lead to the development of new strategies for treatment of age‐related memory deficits.


| INTRODUC TI ON
Glycogen phosphorylase (Pyg) catalyzes the first and rate-limiting step in the process of glycogen degradation (glycogenolysis).
Inhibition of Pyg was shown to block memory formation in young chickens (Gibbs et al., 2006) and induction of the Long Term Potentiation (LTP, a cellular/molecular mechanism of memory formation) in the hippocampi and hippocampal slices isolated from young rodents, respectively (Drulis-Fajdasz et al., 2015;Suzuki et al., 2011).
It was also shown that impairment of synaptic plasticity after Pyg inhibition was associated with decreased transport of glycogen-derived lactate from astrocytes to neurons in a process called the astrocyteneuronal lactate shuttle (ANLS; Magistretti & Allaman, 2015). The impact of the astrocytic glycogen-derived lactate on neuronal metabolism is the subject of ongoing debate (Dienel & Cruz, 2015) and the mechanism by which this pool of lactate stimulates the LTP is not fully understood. However, it is commonly accepted that disruption of the ANLS affects memory formation (Hertz & Chen, 2018).
In contrast to the young animals, inhibition of glycogen breakdown in hippocampal sections isolated from adult and aged rodents was shown to improve the LTP formation, elevating significantly its magnitude (Drulis-Fajdasz et al., 2015). Moreover, in hippocampal slices isolated from old animals, significant alterations in morphology of dendritic spines were observed after inhibition of Pyg, indicating changes in dendritic spines maturation (Drulis-Fajdasz et al., 2015).
Mechanisms underlying this different response to Pyg inhibition remain to be discovered but they might be associated with a different organization of hippocampal formation in young and aged animals (Drulis-Fajdasz et al., 2018) and global changes in the expression of hippocampal proteins , and in the NAD + / NADH metabolism (Zhu et al., 2015) during aging.
In this report, we deliver several lines of evidence that inhibition of Pyg alleviated memory deficits and restored neuroplasticity in aged, 20-to 22-month-old mice. We show that the 2-week intraperitoneal administration of the Pyg inhibitor BAY U6751 improved memory in old animals, in terms of behavioral skills tested using the 2-Novel Object Recognition (NOR) and Novel Object Location (NOL) tests, both partially dependent on the hippocampal mechanisms of memory formation. This reported memory improvement was correlated with significant alterations in the hippocampal formation, at both the cellular and molecular levels, that is, morphological changes in dendritic spines and restoration of young-like proteome in old animals. At the same time, BAY U6751 did not change performance in-also hippocampus-contingent-the Y-maze spontaneous alternations test. This points to the task-specificity of the BAY U6751-induced improvement in cognitive function, which is most likely attributable to changes in both hippocampus and in different neural circuits beyond hippocampus involved in the applied cognitive assays.
In contrast to the aged mice, the inhibition of Pyg in 8-week-old animals had no effect on overall memory formation as measured by the NOR test; however, it disturbed the hippocampal mechanism of memory formation, changed protein expression, and decreased density of dendritic spines in the hippocampal formation.
Both in young and in aged animals, BAY U6751 did not significantly alter the density and morphology of cortical dendritic spines.
The results presented here demonstrate that inhibition of glycogen breakdown may be a promising target of therapies improving aging-associated memory decline.

| RE SULTS
Previously, we showed that the blockade of glycogen degradation significantly improved the LTP induction in the CA1 region of hippocampal slices dissected from old animals (Drulis-Fajdasz et al., 2015). Based on these results, we hypothesized that inhibition of Pyg activity would lead to an improvement of ageassociated deficits in memory formation. To verify this hypothesis, we treated young (6-week-old) and old (20-22 months old) mice for 2 weeks with daily intraperitoneal injections of the Pyg inhibitor BAY U6751 (BAY; Figure 1a) and tested their behavior, neuronal morphology, and proteome in comparison with control groups of animals injected with saline. The titer of the inhibitor was chosen on the basis of studies using the hippocampal slices of young rats.
The lowest titer that completely blocked the LTP formation was selected ( Figure S1).

| Behavioral studies
The 2-Novel Object Recognition test (NOR) and Novel Object Location test (NOL) are routinely used to study memory and learning, preference for novelty, the influence of different brain regions on the process of recognition, and to determine effects of drugs on memory formation (Antunes & Biala, 2012;Denninger et al., 2018).
The tests are technically simple and do not require aversive or appetitive stimulation, thereby minimizing the animal's stress which could be a confounding factor in cognitive skills measurements. In our experiment, we used the NOR variant without the habituation step and with 6-h interval between the familiarization and test session. According to Leger et al., this experimental design allows the most sensitive discrimination between the time spent by an animal on exploring the novel object compared to the familiar one (Leger et al., 2013). The NOR test is known to measure multi-mode brain system interactions in cognition, but it is mainly associated with the cortex and hippocampus functions (Warburton & Brown, 2010).
Moreover, omission of the habituation phase enables detection of aspects which are related mainly to hippocampus-dependent contextual memory formation such as the time to the first exploration of the object (Oliveira et al., 2010).
Data obtained from the NOR test performed at the Day 0 (i.e., before the BAY administration) revealed that, as would be expected, old animals were characterized by impaired novel object recognition compared to young animals. Young mice explored the novel object significantly longer than the familiar one (Y-CTR p = 0.004;  (Figure 1b). We did not find any signs of side preferences in the object exploration ( Figure S2).

| Pyg inhibition disturbs hippocampal-dependent object recognition in young mice
After the 2-week intraperitoneal administration of the phosphorylase glycogen inhibitor (in BAY groups) or saline (in control groups), young animals from both groups continued to spend a significantly higher amount of time on the exploration of the novel object (Y-CTR p = 0.039; Y-BAY p = 1.4 × 10 −5 ) (Figure 1c), which suggested that the BAY treatment did not impair the formation of memory about the familiar/constantly present object.
However, in the NOR test with intersession interval (ISI = 6 h), the overall time of exploration of the new object reflects the sum of the cortex-and hippocampus-dependent mechanisms of the long-term memory formation (Antunes & Biala, 2012). Moreover, Oliveira et al. showed that if familiarization with object occurred when the contextual environment was relatively novel (here: without the habituation phase), the hippocampus played an inhibitory role in the consolidation of object recognition memory. Thus, hippocampal inactivation F I G U R E 1 Inhibition of glycogen phosphorylase stimulates memory formation of 20-to 22-month-old mice. (a) Schematic representation of the experiment; for more details, please refer to the Figure S2a. (b, c) Objects exploration time during the familiarization session of the NOR performed at Day 0 and at Day 14. (d) Latency to the first exploration of any object in the familiarization sessions, and (e) the ratio of the latency to the first exploration between Day 14 and Day 0 as an index of memory formation related to hippocampal plasticity. We observed increased spatial orientation for Y-CTR and O-BAY mice (time reduction at Day 14), and decreased spatial orientation for Y-BAY and O-CTR mice (time extension at Day 14). (f) Sucrose preference test at Day 14. (g) Body mass index as percentage of body mass at Day 0. (h) Time spent on the rod in the Rotarod test at Day 14. The number of subjects in each experimental group n = 9. The statistically significant changes between groups are indicated (*p < 0.05; **p < 0.01; ***p < 0.001).
could even enhance the cortex-related long-term memory in young animals (Oliveira et al., 2010).
In the light of these findings, it may be concluded that at Day 14, the Y-BAY mice expressed abilities similar to the Y-CTR as a result of the preserved function of cortex and inhibition of the hippocampal plasticity ( Figure 1c).
In turn, changes in latency to the first exploration of any object in the NOR test are a measure of the hippocampus-dependent memory (formed during the familiarization session at Day 0). This parameter is also an indicator of anxiety, as animals displaying elevated anxiety tend to delay exploration of objects (Heinz et al., 2021). Here, we did not observe any differences in the latency between the animal groups in the familiarization session at Day 0 (Figure 1d), which gave us confidence about the homogeneity of the animals in this aspect of behavior.
However, after the second familiarization session at Day 14, we found that the BAY-treated young animals did not change the latency to the first exploration (p = 0.192), whereas in the salinetreated young mice, the latency was significantly shorter (p = 0.01) ( Figure 1d,e). Since shortening of time to the first exploration points to improvement in spatial orientation in a familiar environment (Heinz et al., 2021) thus, the lack of such a shortening is an indicator of decreased hippocampal function (disturbance of memory about already seen environment) upon BAY treatment in young animals.

| Pyg inhibition improves hippocampal-dependent cognition in old mice
In contrast to young mice, old animals responded to the BAY treatment with an improved recognition of objects. After 14 days of the treatment, we observed that the novel object recognition index

| The effect of BAY on blood glucose level, sucrose preference, and motor skills
Since the BAY treatment may have significant effects on the metabolism of glucose, we decided to perform additional tests to control for possible alterations of the baseline physiology and behavior (not associated directly with learning and memory in tested animals) that could introduce bias to our NOR test results. We observed that 2-week treatment of animals with BAY had no effect on the body mass ( Figure 1g) and blood glucose level (with average concentration 8.4 mM) of the animals. Moreover, the treatment did not affect motor skills, ataxia and cerebellum-dependent coordination, as measured by the Rotarod test. However, as expected, the time which old animals spent on the Rotarod was about 2/3 of that of young animals (Y-CTR vs. O-CTR p = 0.005; Y-BAY vs. O-BAY p = 8 × 10 −4 ) ( Figure 1h). The lack of effects of BAY on locomotor skills in young and old mice may be unexpected; however, we did not use caloric restrictions and did not perform an endurance exercise test which could reveal changes related to the lack of glycogen availability in the tissues.
Since the hippocampus is implicated in the pathophysiology of depression, we decided to check whether the BAY treatment might result in changes of sucrose preference, a parameter used to measure hedonic deficit, a hallmark of depression (Campbell & MacQueen, 2004). The SPT test performed before the BAY treatment revealed no differences between the studied groups (data not shown). We found a slight decrease in sucrose preference in young but not old BAY-treated animals, compared to the respective saline-

| The effect of Pyg inhibition and aging on dendritic spines density and morphology
In order to check how the Pyg blockade affected neuronal plasticity in young and aged mice, we analyzed morphology and density of dendritic spines in secondary/tertiary dendrites of the CA1 hippocampal region and cortex. Here, we present the scale-free parameter, the length/width ratio, as cumulative curves to visualize spine distribution according to the spine shape classification-from stubby and mushroom-shaped, through thin and long thin, to filopodia-like (Michaluk et al., 2011). The analysis revealed significant differences between young and old groups of untreated mice both in the hippocampus and in the cortex (Figures 3c and 4c).
In agreement with previously published results (Toni et al., 2007), hippocampal and cortical dendritic spines of old mice were enriched in stubby and mushroom spines, as compared to spines of young mice whose dendritic spines were thinner and more filopodia-like Two weeks of BAY administration resulted in changes of the hippocampal dendritic spine morphology both in young and in aged animals, compared to their respective control groups.
In young animals, Pyg inhibition was associated with the shift of the CA1 hippocampal spines morphology toward shorter and thicker spines (Y-CTR vs. Y-BAY: D = 0.11, ks = 2.46, p = 1 × 10 −6 ) ( Figure 3a), that are associated with more stable synapses formed upon LTP induction (Szepesi et al., 2014). These morphological changes in response to the BAY treatment were, however, accompanied by a significant decrease in the spine density, compared to young control group (Y-CTR = 1.02 spines/μm; Y-BAY = 0.54 spines/ μm; p = 6.6 × 10 −6 ) ( Figure 3e) which may suggest that the Pyg activity blockade caused elimination of immature (long and thin) spines rather than spine maturation.
In contrast, in old animals, we did not find significant changes in the spine density of the CA1 hippocampal dendrites upon the  (Figure 3b). However, when we looked into the immature spine compartment (spines with the length-to-width ratio above 2), we found that the BAY treatment resulted elongation and thinning of the spines, that is, the spines became more filopodia-like (O-CTR vs. O-BAY: D = 0.166, ks = 2.5, p = 7 × 10 −6 ) ( Figure 3b). Since the abundance of the filopodia-like spines is associated with increased potential for new synapse formation thus, the improvement of memory formation of BAY-treated aged mice may be at least partially attributed to the improved synaptic plasticity.
The O-BAY group's changes in hippocampal dendritic spines were analogical (but more prominent) to that observed by us previously for rat acute hippocampal slices treated with the glycogen phosphorylase inhibitor ex vivo (Drulis-Fajdasz et al., 2015).
Remodeling of the dendritic spines shape is a marker of structural plasticity within hippocampus excitatory synapses, and this plasticity is probably responsible for the enhanced memory formation observed in old rats and mice. age (with the final NOR test performed on 8-week-old mice) thus, the high physiological lactate concentration in the brain (Chen et al., 2016), along with intensive neurodevelopmental processes that occur in such young animals , may be sufficient to compensate for the impairments caused by the BAY inhibitor in hippocampi.

| Inhibition of glycogen phosphorylase alters global hippocampal proteome of young and old mice
The overview of the proteomic experiment is presented in Figure 5a.
A total of 7959 proteins with at least two unique peptides were identified across all study samples. The mass spectrometry proteomics data have been deposited to the Proteome Xchange Consortium via the PRIDE partner repository with the dataset identifier PXD025978, file UU6-UU15.
For quantitative analyses, only proteins identified in at least 60% of samples belonging to a given study group were used (6147 proteins in total). The principal component analysis (PCA) of protein concentrations exhibited changes in each analyzed proteome related to age and BAY treatment ( Figure 5b). We observed that inhibition of glycogen phosphorylase significantly affected the abundance of several proteins in old mice (340 proteins in total, p < 0.05), 178 and 118 of which were up-and down-regulated, respectively (please see the Table S1). In the Figure  Similarly to previous studies, we found that many fundamental pathways were down-regulated with aging, including translation, nervous system development, protein transport, cell adhesion and migration, and neurotransmission ( Figure 6a). In turn, enrichment analysis of BAY-related proteins with increased abundance revealed a significant up-regulation of many of the above processes suggesting an overall shift of the aged hippocampal proteome toward the young-like proteome (Figure 6b). Notably, in BAY-treated old mice, we observed changes in the abundance of several proteins involved in neuronal transmission and memory formation toward young-like proteome. Among these proteins were key excitatory glutamate receptors (both the ionotropic, such as Gria, Grin, Grik, Grid, and the metabotropic, such as Grm) (Figure 6c). The BAY treatment F I G U R E 3 Glycogen phosphorylase inhibitor BAY U6751 affects dendritic spines morphology in hippocampi of young and old animals. (a-d) Distribution of dendritic spines shape parameter (length-to-width ratio). (e) Spine linear density in hippocampal neurons. (f) Examples of DiI-stained neurons in the CA1 area of mouse hippocampus (images show secondary apical dendrites; blue outline-exemplary regions of interest marked for individual dendritic spines; yellow outline-defined dendrite core). The data were analyzed using the Kolmogorov-Smirnov test (a-d) and Student's t-test (e). The number of subjects in each experimental group: animals (n = 4-5), cells per animal (n = 9-14), and spines per group (n = 1100-1250). *** p < 0.001.

| LC-MS quantification of the BAY in mouse hippocampi and cortex
We used the LC-MS method to identify and quantify BAY U6751 in brain structures, to confirm that intraperitoneal administration of the inhibitor allowed to achieve effective working concentrations in the hippocampus and cortex. Measurements were made based on three technical repetitions, and obtained concentration values are above the limit of quantification (LOQ) in the developed method (≥23.29 ng/mL).
We found that the average concentration for BAY in the hippocampus and cortex tissue extracts at Day 14 was 24.42 ± 3.64 ng/mL and 57.13 ± 3.13 ng/mL, respectively. The measurement precision was expressed as the coefficient of variation (CV) and did not exceed ±15% (for hippocampal = 14.92%; for cortex = 5.5%; see Methods 5.14). Representative extracted ion chromatograms for BAY in hippocampus and in cortex are presented in Figure S5.
The LC-MS measurements confirmed that BAY effectively penetrated the tissues and crossed the blood-brain barrier. The measured concentrations of the compound were similar to concentrations determined in our previous study: 23.66 ± 0.96 ng/mL (CV = 4.07%) in the hippocampus and 58.92 ± 1.21 ng/mL (CV = 2.05%) in the cortex (Pudelko- Malik et al., 2022). F I G U R E 4 Glycogen phosphorylase inhibitor BAY U6751 do not affect spine morphology in the cortex of young and old animals. (a-d) Distribution of dendritic spines shape parameter (length-to-width ratio). (e) Spine linear density column chart. (f) Examples of DiI-stained neurons in cortex (blue outline-exemplary regions of interest marked for individual dendritic spines; yellow outline-defined dendrite core). The data were analyzed using the Kolmogorov-Smirnov test (a-d) and Student's t-test (e). The number of subjects in each experimental group: animals (n = 4-5), cells per animal (n = 9-14), and spines per group (n = 1100-1250).

| DISCUSS ION
Memory loss and cognitive impairments are observed in numerous neurodegenerative diseases. However, a gradual decrease in cognitive functions such as memory formation and its retention also occurs naturally as a part of the aging process in individuals with no evident symptoms of pathological neurodegeneration (Harada et al., 2013). Endurance exercises and healthy lifestyle are Black circles represent all other significantly changed proteins. Gray circles display proteins not regulated by BAY/age. The number of upregulated (red) and down-regulated (green) proteins is presented. Proteins were graphed by fold change (difference) and significance (−log p-value) using a false discovery rate (FDR) of 0.05, with the Perseus software. Ogrodnik et al., 2021) and inhibitors of the integrated stress response may contribute to the improvement of age-related cognitive (Chou et al., 2017;Krukowski et al., 2020;Oliveira et al., 2021). Our previous studies on a molecular/cellular mechanism of memory formation-LTP, and aging-dependent alterations in hippocampal proteomes suggested that glycogen metabolism may be a promising target of an anti-aging brain therapy.
It is widely accepted that the astrocytic glycogen-derived lactate is indispensable for the hippocampus-associated memory formation in young individuals (Gibbs et al., 2006;Newman et al., 2011;Suzuki et al., 2011). The mechanism of the lactate-dependent stimulation of the LTP formation is not fully understood but it is hypothesized that the crucial role in this process plays an increasing NADH/NAD + ratio in neurons, which is a result of astrocytic glycogen-derived lactate oxidation (Magistretti & Allaman, 2015). The elevation of the NADH/NAD + ratio may lead to potentiation of NMDA signaling (Yang et al., 2014) and promotion of Camk2 autoactivation by dimeric fructose-1,6-bisphosphatase 2 (Fbp2) (Duda et al., 2020), an enzyme whose oligomerization (tetramerization) is NAD + -and AMPdependent (Gizak et al., 2019).
In contrast to young animals, in hippocampal slices isolated from old rodents, the inhibition of glycogen degradation significantly elevated the amplitude of the LTP (Drulis-Fajdasz et al., 2015). Thus, the main aim of the present study was to investigate whether a prolonged inhibition of glycogen phosphorolysis could modify old mice behavior.
The induction and maintenance of the LTP trigger persistent changes in gene expression. It was shown that the LTP induction correlated with an increase in the expression of more than 350 genes, while the level of transcripts for about 240 genes was lowered (Chen et al., 2017). Therefore, we also hypothesized that a pro-  Gibbs et al., 2006;Kilic et al., 2018), in the present work, we applied the 14-day protocol of daily injections of mice with the inhibitor (Figure 1a). We found that this treatment did not affect basic physiological parameters of young and old animals such as body mass, serum glucose level, and motor coordination. On the F I G U R E 7 Calcium/calmodulin-dependent protein kinase type IV (Camk4) expression in the hippocampus is altered by aging and BAY treatment. (a) Exemplary confocal images of Camk4 immunofluorescence (magenta) distribution within hippocampal Cornu Ammonis (CA1) stratum pyramidale (SP) region in young (Y-CTR), young BAY treated (Y-BAY), old (O-CTR) and old BAY treated (O-BAY) animals, respectively. (b) Localization of neuronal somata, dendrites, and nuclei was revealed with antibodies against β-Tubulin Isotype III (green) and DAPI (blue), respectively (merge channels). (c) Quantification of Camk4-related immunofluorescence in hippocampal slices calculated for total hippocampi CA1 and CA2 region (SP, stratum pyramidale; SR, stratum radiatum; stratum lacunosum-moleculare, SLM). Immunofluorescence was normalized to values obtained for Y-CTR samples (n = 6). Asterisks indicate a statistically significant difference (*p < 0.05, **p < 0.01, ***p < 0.001). (d) Relative immunofluorescence ratio between analyzed animal groups (IF ratio) in comparison with protein concentration ration obtained in proteomic analysis (TPA ratio).

F I G U R E 8 Cellular localization of Glutamate receptor 2 (Gria2 = GluR2) in the hippocampus is altered with aging and is altered with BAY inhibitor treatment. (a) Exemplary confocal images of Gria2 immunofluorescence (magenta) distribution within hippocampal Cornu
Ammonis (CA1) stratum pyramidale (SP) region in young (Y-CTR), young BAY treated (Y-BAY) old (O-CTR) and old BAY treated (O-BAY) animals respectively. (b) Localization of neuronal somata, dendrites, and nuclei as revealed with antibodies against β-Tubulin Isotype III (green) and DAPI (blue), respectively (merge channels). (c) Quantification of Gria2 normalized immunofluorescence in hippocampal slices calculated for total hippocampi CA1 and CA2 region (SP, stratum pyramidale; SR, stratum radiatum; stratum lacunosum-moleculare, SLM). Immunofluorescence was normalized to values obtained for Y-CTR samples (n = 6). Asterisks indicate a statistically significant difference (*p < 0.05, **p < 0.01, ***p < 0.001). (d) Relative immunofluorescence ratio between analyzed animal groups (IF ratio) in comparison with protein concentration ration obtained in proteomic analysis (TPA ratio). contrary, the consumption of sucrose by young animals treated with BAY was slightly but statistically significantly reduced, which indicated a tendency to the anhedonic-like behavior of that group of animals and might suggest alterations in the mesocortical limbic system activity (Campbell & MacQueen, 2004). Interestingly, we did not observe significant alterations in glycogen and lactate levels in hippocampi of the BAY-treated old mice compared to the untreated animals ( Figure S6a,b). However, these findings are consistent with results of studies on another glycogen phosphorylase inhibitor, 1,4-dideoxy-1,4-imino-D-arabinitol (DAB). They demonstrated that DAB had an inhibitory effect also on glycogen synthase and that prolonged exposure to the inhibitor reduced glycogen levels (Walls et al., 2008). BAY is supposed to have a similar secondary effect (Latsis et al., 2002).
Thus, after the long-term treatment with BAY only moderate changes in glycogen and glycogen-derived lactate could be expected.
The results of the NOR and NOL tests revealed that the 2-week treatment with BAY significantly improved spatial orientation and memory formation in old mice. The parameters of the NOR test that are related to both the cortical and hippocampal components of memory formation were moderately changed only in the BAYtreated old mice (Figure 1c). However, after the BAY-treatment, the parameters reflecting mainly the hippocampus-based mechanisms of memory formation were significantly improved in old (Figures 1d,e and 2c), but worsened in young animals (Figure 1d,e).
Intriguingly, it turned out that in contrast to the BAY-induced cognitive improvement observed in the NOR and NOL tests, the Y-maze test ( Figure S3a) showed no differences between the BAY-treated and untreated old animals in working memory ( Figure S3b-e).
We argue that such effects might be attributable to the at least partially different neural circuits underlying the diverse cognitive functions measured in NOR, NOL, and Y-maze assays. Indeed, the said circuits are partially located in the structures, which were not the scope of this study.
It is noteworthy, the BAY delivery in our experiments was systemic rather than targeted to a specific brain region. Hence, we are unable to unequivocally attribute the results of the testing for specific cognitive functions to the improvement of plasticity in hippocampus alone. For example, it is feasible that the NOL and NOR-related hippocampal processing of the input from the parahippocampal, and perirhinal/entorhinal cortex might be differently influenced by the drug than the heavily hippocampus-dependent spatial working and Localization of neuronal somata, dendrites, and nuclei as revealed with antibodies against β-Tubulin Isotype III (green) and DAPI (blue), respectively (merge channels). (c) Quantification of Gad1 normalized immunofluorescence in hippocampal slices calculated for total hippocampi CA1 and CA2 region (SP, stratum pyramidale; SR, stratum radiatum; stratum lacunosum-moleculare, SLM). Immunofluorescence was normalized to values obtained for Y-CTR samples (n = 6). Asterisks indicate a statistically significant difference (*p < 0.05, **p < 0.01, ***p < 0.001). (d) Relative immunofluorescence ratio between analyzed animal groups (IF ratio) in comparison with protein concentration ration obtained in proteomic analysis (TPA ratio). due to the process called LTP-occlusion (Li et al., 2005;Whitlock et al., 2006). We argue that due to the BAY-enhanced receptiveness of the neural circuits of the aged mice to change, plasticity saturation happened more readily than in the untreated controls, which led to the observed effects.
However, to deepen our understanding of the correlation between the BAY-induced enhancement of cognitive functions and synaptic plasticity, similar electrophysiological experiments should be performed before and after the NOR and NOL tests.
Taken together, our results showed (a) the task-specificity of the BAY-induced improvement in cognitive functions, which is most likely attributable to changes in both hippocampus and in different neural circuits beyond hippocampus involved in the specific behaviors tested, and (b) a complex and most probably non-linear relationship between the BAY treatment and synaptic plasticity induced by training.
The worsening of the hippocampus-based memory was observed by Magistretti's and Gibbs's groups which showed that inhibition of glycogen degradation and lactate release from astrocytes in young animals, impaired their hippocampus-based memory formation (Gibbs et al., 2006;Suzuki et al., 2011;Vezzoli et al., 2020).
The results presented here showed that the 2-week inhibition of Pyg in young mice weakened their memory to the level observed in old animals, whereas it improved the memory of old mice to the level typical for young animals (Figure 1c). The molecular mechanism of such opposite effects of Pyg inhibition in young and old animals is enigmatic and requires further studies. Pyg activity in the brain is attributed almost exclusively to astrocytes (Newman et al., 2011), and the role of the enzyme in neurons is controversial. Some literature data suggested that Pyg activity in neurons, although low, had a physiological significance (Saez et al., 2014) but other studies showed that the activity of neuronal Pyg was permanently blocked by phosphorylation (Vilchez et al., 2007). The brain aging was shown to correlate with an increasing acidification, elevation of lactate titer (Ross et al., 2010), mild upregulation of the overall glycolytic enzymes concentration (Drulis-Fajdasz et al., 2018;Gostomska-Pampuch et al., 2021), and an increase of the NADH/NAD + ratio (Zhu et al., 2015). It was also shown that supraphysiological concentrations of lactate in young mouse brains could mimic disturbances observed in the old brain (Das et al., 2022).
Thus, an attractive hypothesis is that the inhibition of glycogen degradation in old brains may reduce the permanently elevated lactate level and neutralize acidification thereby improving the ability to form the LTP.
Our observation that Pyg inhibition affected mainly hippocampal mechanisms of memory formation prompted us further to check the density and morphology of dendritic spines in both brain formations.
We found that BAY reduced the density of dendritic spines (Figure 3e) and significantly shortened their length in hippocampi of young mice ( Figure 3a). A reduction of these parameters is considered as a measure of lower plasticity and is related to a loss of spatial orientation and increase of an anhedonic behavior (Krzystyniak et al., 2019). On the contrary, in old animals, the BAY treatment led to a significant shift of dendritic spines population from the stubby and mushroomlike shapes (that are considered to form stable synaptic connections) (Toni et al., 2007), toward the long and thin, and filopodia-like structures (Figure 3b). Such filopodia-like dendritic spines are the main form of spines in young animals and are considered as a hallmark of high neuronal plasticity (Basu et al., 2016). In contrast to hippocampal formation, we did not observe any significant alterations in the density and morphology of cortical dendritic spines either in young or in old BAY-treated animals (Figure 4), which was consistent with our observation that inhibition of Pyg affected predominantly the hippocampus-based memory. Because the effect of BAY on memory formation is related to a disturbance in the synthesis of lactate from astrocytic glycogen thus, the lack of influence of Pyg inhibition on cortex-based mechanisms of memory formation and cortical dendritic spines suggests that the ANLS is essential for memory formation in the hippocampal formation, but in the cortex, the ANLS,  5 and 6). While in the control animals, aging was correlated with downregulation of protein groups associated with translation, nervous system development, protein transport, cell adhesion and migration, and neurotransmission, the BAY treatment of aged animals increased back the abundance of proteins engaged in many of these processes. Although we cannot definitely exclude that BAY directly affected gene expression by interactions with DNA regulatory sequences, the differences in BAY-induced changes between young and old hippocampal proteomes point to Pyg-inhibition-associated changes in metabolites as the trigger of proteomic remodeling.

| CON CLUS IONS
Our study presents a new insight into the role of glycogen in the aging brain. In contrast to young animals, inhibition of glycogen degradation improves memory formation in old animals, affecting mainly functions of the hippocampal formation. Therefore, our findings provide direct evidence that inhibition of glycogen phosphorolysis improves memory formation in old animals, and thus, they can contribute to development of better treatment strategies for aging-related memory deficits. However, detailed studies on the long-term effects of Pyg inhibitor(s)-also adminis-

| Behavioral tests
To determine the impact of glycogen phosphorylase inhibition on cognitive skills of the mice, all animals were subjected to two rounds of the 2-Novel Object Recognition test (NOR; please see Methods

| 2-Novel object recognition test (NOR)
The NOR procedure was performed without the habituation step according to M. Leger et al. (Leger et al., 2013) (see Figure S2). A 33 cm × 33 cm × 20 cm (length, width, height, respectively) box made from white opaque material and with top lightening (5 lux intensity) was used. The test consisted of the familiarization session-in this session each mouse was placed in the box, and allowed to explore two identical objects for 10 min (T1); the intersession interval (ISI) = 6 h-the mouse was returned to its home cage; and the test session-the mouse was placed in the box again, with one familiar and one novel object (T2). The novel object was different in shape, texture and color. To avoid stress resulting from removing only one animal from the cage, each pair of animals from one cage was tested simultaneously in identical boxes, with identical set of objects. After each session, the box and the objects were thoroughly cleaned with ethanol and distilled water to remove any residual scent. The behavior of mice during T1 and T2 was recorded by digital camera located approximately 1.5 m above the box. After 14 days of the pharmacological treatment, the NOR test was repeated with novel objects.
For the detailed representation of the NOR test design, please refer to Figure S2.

| Novel object location test (NOL)
For the NOL, we used the same behavioral setup and additional animal care procedures between test steps, as described in Methods 5.3. The NOL were performed according to Denninger et al. (2018).

| Behavioral analysis of the NOR test and the NOL test
Video recordings of mice behavior during the NOR and NOL tests were analyzed using a self-developed analysis system. Exploration time was counted whenever the nose of an animal was at least 2 cm from the object, facing toward the object. When an animal climbed the object with all four pawns, the exploration time was

| Sucrose preference test (SPT)
The SPT test was performed in the dark phase of the 24-h cycle. To decrease the likelihood of bias resulting from neophobia to sweet taste, mice were given 2.5% sucrose solution for 2 h instead of water 1 day before the actual test. The SPT test was performed before and after 14 days of BAY administration. The mice were given freechoice access to 1% sucrose solution and water that were provided in identical bottles for 12 h. To eliminate possible bias from side preference, the positions of the bottles were changed after 6 h of the test. The consumption of water and sucrose solution was estimated by weighing the bottles. The percentage of sucrose preference was calculated using the following equation: Other conditions of the test were as previously described by Strekalova et al. (2011).

| Rotarod test
To measure the motor skills of mice, the widely used Rotarod test was performed. After 1 hour of habituation to the room where the test took place, mice were gently placed on the rubber shaft, which was set to initial rotational motion (10 rpm). The shaft rolling speed was gradually increasing (0.125 rpm/s). The test lasted for maximum 240 s. The amount of time the mouse spent on the spinning shaft was used to determine the animal's motor abilities. The test was performed 4 times for each animal, and the average time spent on the shaft was used for further analysis.

| Tissue preparation
For the dendritic spine analysis (Methods 5.9), mice were anesthetized (ketamine/xylazine) and transcardially perfused with 1.5% paraformaldehyde in PBS. The brains were dissected, the hippocampi and frontal cortex separated in an ice-cold phosphate buffer, and sliced using a vibratome. For the proteomic and LC-MS analyses (Methods 5.11 and 5.14), animals were anesthetized with isoflurane and decapitated. The brains were dissected, the hippocampi and frontal cortex separated, frozen immediately in liquid nitrogen, and kept at −80°C for further analysis. For immunofluorescent staining

| DiI staining of brain slices
To visualize changes in the shape of dendritic spines, 1,1′-dioctade cyl-3,3,3,3′-tetramethylindocarbocyanine perchlorate (DiI) staining was performed in slices (140 μm thick) of different brain structures (Bączyńska et al., 2021). The slices were allowed to recover for at least 1.5 h at room temperature. Random dendrite labeling was performed using 1.6 μm tungsten particles (Bio-Rad) coated with propelled lipophilic fluorescent dye (DiI; Invitrogen) and delivered to the cells by gene gun (Bio-Rad) bombardment. Images of dendrites in different brain regions were acquired under 561 nm fluorescent illumination using a confocal microscope (63× objective, 1.4 NA) at a pixel resolution of 1024 × 1024 with a 3.43 zoom, resulting in a 0.07 μm pixel size.

| Morphometric analysis of dendritic spines
The analysis of the dendritic spine morphology and calculation of changes in spine parameters were performed as described previously (Krzystyniak et al., 2019). The images acquired from the brain slices were processed using ImageJ software (National Institutes of Health, Bethesda, MD, USA) and analyzed semi-automatically using the custom-written SpineMagick software (patent no. WO/2013/021001) and 3dSpAn software for three-dimensional dendritic segment reconstruction (Basu et al., 2016). The spine length was determined by measuring the curvilinear length along a fitted virtual skeleton of the spine. The fitting procedure was performed by looking for a curve along which integrated fluorescence was at a maximum. The head width was defined as the diameter of the largest spine section while excluding the bottom part of the spine (1/3 of the spine length adjacent to the dendrite). Based on that we calculated a scale-free parameter length-to-width ratio, which effectively describes the spine shape (Michaluk et al., 2011). Dendritic segments of at least 4 animals per condition were morphologically analyzed resulting in 1100-1250 spines per group. To determine the spine density, approximately 1000-1500 μm of dendritic length was analyzed per experimental group (9-14 cells per group).

| Proteomic analysis
The sample preparation and proteomic measurements were performed as described before  using two groups of female C57BL/10 J mice: young (8-week-old; n = 10) and old (20-to 22-month-old; n = 10), which were treated with BAY (n = 5/group) or saline (n = 5/group), as described in Methods 5.2.
The overview of the proteomic experiment is presented in Figure 5a.
In brief, hippocampi were dissected from the mice and the samples were prepared by tissue lysis in SDS containing buffer, followed by a two-step consecutive protein digestion by multi-enzyme digestion filter aided sample preparation (MED-FASP) protocol using Lys-C and trypsin (Wiśniewski et al., 2020). The obtained peptides were separately analyzed by LC-MS/MS, and MaxQuant (MQ) software (Max Planck Institute, Martinsried, Germany) was used for spectra searching. Specific protein concentrations were calculated by the "Total Protein Approach" (TPA) using raw intensity MQ output (Wiśniewski & Rakus, 2014). Statistical analysis was conducted using Perseus software v.1.6.14.0 (Max Planck Institute for Biochemistry). The

| LC-MS analysis
The quantification of BAY in the hippocampus and cortex was performed with LC-MS analysis. The sample preparation and experimental instrument parameters were described previously Si Q-TOF MS with an electrospray ion source (ESI) was combined with the Acquity UPLC I-class chromatographic system (Waters).
Chromatographic separation was performed on the Acquity UPLC CSH C18 (2 × 100 mm, 1.7 μm, Waters) analytical column. The mobile phase comprised of water (A) and methanol (B) with the addition of 0.1% formic acid. The MassLynx software (version 1.60.1774, Waters) with the QuanLynx application were used for data acquisition and quantitative analysis, respectively. The limit of detection (LOD) and limit of quantification (LOQ) were calculated using the calibration curve and were 7.76 and 23.29 ng/mL, respectively. In the presented analytical method, precision was expressed as the coefficient of variation (CV). In order to meet the acceptance criteria of the European Medicines Agency (EMA) Guidelines, the CV did not exceed ±20% for lower limit of quantification (LLOQ) and ±15% for higher value of concentration.

| Statistical analysis
All tests for 4 groups of animals were analyzed: Y-CTR, Y-BAY, O-CTR, and O-BAY. In behavioral tests; n = 6-9 animals for each group were analyzed. In the immunolabeling and Dil staining, n = 4 animals for each group, and the proteomic data we determined for n = 6 animals for each group. For statistical analysis, we used the parametric unpaired Student's t-test and/or the Kolmogorov-Smirnov test (D -max deviation; ks -Kolmogorov-Smirnov test statistic; pprobability). The statistical significance of differences was indicated by asterisks as follows: *p < 0.05, **p < 0.01, ***p < 0.001. For the LC-MS measurements, precision was calculated by coefficient of variation (CV).