Delayed formation of neural representations of space in aged mice

Abstract Aging is associated with cognitive deficits, with spatial memory being very susceptible to decline. The hippocampal dentate gyrus (DG) is important for processing spatial information in the brain and is particularly vulnerable to aging, yet its sparse activity has led to difficulties in assessing changes in this area. Using in vivo two‐photon calcium imaging, we compared DG neuronal activity and representations of space in young and aged mice walking on an unfamiliar treadmill. We found that calcium activity was significantly higher and less tuned to location in aged mice, resulting in decreased spatial information encoded in the DG. However, with repeated exposure to the same treadmill, both spatial tuning and information levels in aged mice became similar to young mice, while activity remained elevated. Our results show that spatial representations of novel environments are impaired in the aged hippocampus and gradually improve with increased familiarity. Moreover, while the aged DG is hyperexcitable, this does not disrupt neural representations of familiar environments.


| INTRODUC TI ON
As the world's population ages, it becomes more important to understand the cognitive effects of aging and to develop therapeutic strategies for the aging brain.While healthy aging has milder cognitive effects than neurodegenerative states, several aspects of cognitive function exhibit decline (Jagust, 2013).Memory deteriorates with age, with spatial memory being particularly susceptible to deficits in older humans and animals (Konishi et al., 2017;Techentin et al., 2014).Older adults show orientation deficits (Diersch et al., 2021;Moffat, 2009), but they tend to perform better in familiar locations or tasks than in novel ones (Merriman et al., 2016;Muffato et al., 2015).
Novelty is a powerful modulator of brain activity, and several brain regions respond more strongly to novel stimuli than they do to familiar ones (Kafkas & Montaldi, 2018).In memory systems, the hippocampal dentate gyrus (DG) is known to play an important role in novelty detection.Novelty, such as the exploration of a novel environment or new objects, has been found to increase DG excitability and long-term potentiation (LTP) (Kitchigina et al., 1997;Sierra-Mercado et al., 2008).In vivo whole-cell recordings found that DG granule neurons undergo a transient depolarization in novel environments, which is driven by metabotropic acetylcholine receptors (Gómez-Ocádiz et al., 2022).Another study found that hilar mossy cells in the DG have increased c-Fos expression, a proxy for increased activity, when animals explore novel objects (Bernstein et al., 2019).Both aged mice and mice with DG lesions show impaired novelty detection in an object placement task, which tests spatial memory, whereas lesions of CA3 resulted in no such deficits (Hunsaker et al., 2008).On the contrary, aged mice show an increase in novel-object exploration behavior when spatial memory is not involved (Denny et al., 2012).
Aging-related deficits in spatial memory are thought to be caused by vulnerabilities in the hippocampal circuits that mediate spatial learning (Morris et al., 1982;Scoville & Milner, 1957).
The hippocampus undergoes diverse alterations during the course of aging (reviewed in Leal & Yassa, 2015).These include cell loss, synaptic degradation of inputs from the entorhinal cortex (EC), and decreases in neuro-modulatory inputs.There are also decreases in synaptic plasticity and neurotrophic support, as well as epigenetic changes.The DG is one of the hippocampal subregions that are most vulnerable to aging (Small et al., 2004;Yassa et al., 2011).The inputs from the EC to DG are a main area of dysfunction in both Alzheimer's disease (AD) patients and AD mouse models (Moreno et al., 2007).During aging, there are reduced synaptic contacts from both the medial and lateral EC (Burke & Barnes, 2010) onto DG neurons.This is accompanied by reduced synaptic plasticity of these inputs, including a higher threshold for LTP induction (Barnes, Rao, & Houston, 2000;Froc et al., 2003).In addition, the DG undergoes a sharp age-related decrease in adult neurogenesis (Kuhn et al., 1996;Moreno-Jiménez et al., 2019;Spalding et al., 2013).Activity levels in different hippocampal areas are known to change during aging.
In CA3, which is downstream of the DG, aging is associated with an increase in activity (Wilson et al., 2005;Yassa et al., 2011).Aging is also correlated with a reduction in the number of somatostatinexpressing DG interneurons, particularly in cognitively impaired rodents, suggesting that loss of inhibition and hyperactivity in this area could be at the origin of aging-related cognitive deficits (Koh et al., 2014).However, there are very limited data about how DG activity changes with aging, primarily because neuronal activity in this region tends to be too sparse for in vivo electrophysiological recordings (GoodSmith et al., 2017;Leutgeb et al., 2007).Most studies have therefore resorted to immediate early gene (IEG) expression as a proxy of neuronal activity.Aging was associated with reductions in DG expression of Egr1/Zif268 (Desjardins et al., 1997;Penner et al., 2016), Arc (Marrone et al., 2012) and Npas4 (Qiu et al., 2016), while the number of cells expressing c-Fos was unchanged (Chawla et al., 2013;Desjardins et al., 1997).However, IEG expression is an indirect measure of activity, so aging-related changes could be due to changes in DNA methylation, rather than a decrease in neuronal activity (Penner et al., 2011).
Since the DG is upstream of other hippocampal areas, changes in this region could help elucidate the causes of aging-related changes in neuronal activity (El-Hayek et al., 2013;Lee et al., 2021;Simkin et al., 2015;Wilson et al., 2005) and neural representations of space (Barnes et al., 1983;Tanila et al., 1997a;Wilson et al., 2005) found in some areas of the hippocampus but not in others.Because different areas have different susceptibilities to aging, either hyper-or hypoactivity or disrupted spatial representations in the DG could be relayed along the hippocampus tri-synaptic circuit to other downstream areas such as CA3, where strong recurrent excitatory connections could further contribute to hyperexcitability, leading to "runaway" excitation and cognitive impairment (Tamminga et al., 2010).The DG exerts a strong gating effect on the excitatory inputs to CA3, and aging-related disfunction in the DG is thought to contribute to CA3 hyperexcitability and epileptiform activity (Patrylo & Williamson, 2007).Activity patterns in both the human DG/CA3 and in the EC inputs to those areas appear to change differently during aging, with the EC being hypoactive while CA3 is hyperactive.This imbalance in activity between regions was associated with deficits in memory discrimination both in humans and rodents and has been suggested as a cause of age-related cognitive impairment (Lee et al., 2021;Reagh et al., 2018).
In this study, we investigated whether aging is associated with impaired DG activity and spatial representations, which could account for the spatial memory deficits seen in aged individuals.Using in vivo two-photon microscopy, we imaged calcium activity in a large population of DG neurons of young and aged mice as the animals walked head-fixed on a treadmill to which they had never been exposed.We found that aged mice have increased DG single-cell calcium activity and disrupted neural representations of space upon their first exposure to the treadmill setting.However, further imaging sessions on subsequent days showed that spatial representations become similar to those of young mice as the animals familiarize themselves with the treadmill, whereas single-cell activity remains elevated.Our results highlight the importance of novelty and familiarity to spatial encoding in aged animals.

| RE SULTS
To record neuronal activity in DG excitatory neurons, we injected young (3-4 month old) and aged (21-26 month old) C57Bl6 mice with an AAV encoding the calcium indicator jRGECO1a (Dana et al., 2016) under the CaMKII promoter (Figure 1a).We then implanted the mice with a hippocampal imaging window above the DG (Danielson et al., 2016;Gonçalves et al., 2016) to enable in vivo two-photon calcium imaging in this region (Figure 1a,b).After waiting 3-4 weeks to allow for viral expression and recovery from surgery, we imaged the granule cell layer of the DG over four consecutive days.During each imaging session, the mice ran head-fixed along a treadmill containing tactile cues with four distinct segments (Figure 1a,c) that was manually moved (Figure S1).We recorded 10 min videos of calcium activity and identified the regions corresponding to active cells using Suite2p software (Figure 1d).Previous studies found that hippocampal activity in area CA3 becomes elevated during aging (Wilson et al., 2005;Yassa et al., 2010), and we asked if this was also the case in the DG.We determined single-cell calcium activity of dentate granule neurons by measuring the area under the curve of transients and normalizing to the distance run by each mouse (Frechou et al., 2022) (Figure 1e-g).We found that the aged group had higher activity levels than the young group (Figure 1e, Young N = 8, n = 910; Aged N = 8, n = 699; p = 0.0046, nested bootstrap), which aligns with previous reports of hyperactivity.We also estimated the fraction of active neurons during each imaging session and found no changes between groups (Figure S2).We next investigated whether there were differences between the spatial representations of young and aged DG neurons.Both young and aged DG neurons have a variety of spatial responses, including "place cells" (Jung & McNaughton, 1993;Leutgeb et al., 2007;Neunuebel & Knierim, 2012;O'Keefe & Dostrovsky, 1971) that only respond to a specific location and neurons whose activity shows low spatial preference (Figure 2a,b), although since the animals are head-fixed the activity patterns of these cells are likely different from what they would be in freely moving animals (Ravassard et al., 2013).We determined the spatial selectivity of neuronal activity using a previously described spatial tuning index (see methods).
DG neurons in aged mice had significantly lower tuning indices than those in young mice (Figure 2c, Young N = 8, n = 910; Aged N = 8, n = 699; p < 0.0001, nested bootstrap).This difference indicates that the activity of DG neurons in aged mice was not as place-specific as that of young DG neurons, suggesting an aging-related degradation of the spatial code.We also computed single-cell Fisher information (FI), as a measure of how much spatial information was encoded by individual DG neurons.By plotting whole-recording calcium activity raster plots for neurons in the 10th percentile of FI, we verified that neurons with high spatial information behave like place cells in both young and aged animals, as their activity is concentrated at a single location on the treadmill (Figure 2d,e).This place-specific response was present across laps as neurons in both cohorts of mice largely To confirm that the changes in spatial representations in aged mice corresponded to a change in spatial memory, mice underwent behavioral testing using an object placement paradigm.This test takes advantage of a mouse's natural preference for novelty to assess how well animals can remember the location of objects in space, a behavior that is thought to be hippocampus dependent (Haettig et al., 2013).Mice explored an arena with visual cues and two novel and identical objects during a training trial.We then displaced one object to another location in the arena and again allowed the mice to explore during a test trial (Figure S3a).Young mice had a preference for the novel object that was significantly above chance (p = 0.0017, One sample Wilcoxon test) whereas aged mice did not score above chance (Figure S3b,c, p = 0.64, One sample Wilcoxon test).This change in preference can also be seen in the exploration times per objects, in which young mice spend more time engaged with the displaced object, while aged mice spend about an even amount of time between objects (Figure S3d, p = 0.021, two-way ANOVA with Sidak multiple comparisons test).Overall, these data confirm earlier findings of spatial memory deficits in aged mice, which we now show is accompanied by impaired neural representations of space in the DG.
Older individuals are better able to distinguish places in an environment that they are familiar with than in a novel environment (Merriman et al., 2016;Muffato et al., 2015), so we asked how the representations of space in the DG changed across four consecutive days of imaging, as the treadmill belt became more familiar to the mice.Single-cell calcium activity remained elevated in aged mice through all recording days, even though this difference was not significant in Days 2 and 3 (Figure 3a, nested bootstrap with Bonferroni correction).While in young mice there was a slight but significant decrease in activity (p = 0.00005) across days, in aged mice activity increased over time (p < 0.0001) (Figure 3b,c, nested bootstrap).
In contrast to activity, the differences in tuning index and FI were erased over the 4 days of imaging, as young and aged groups converged.The spatial tuning index of young mice remained approximately constant over the course of all imaging sessions (p = 0.20), whereas the tuning of aged mice underwent an increase so that the differences between both groups were no longer significant in days 3 and 4 (p = 0.00005).(Figure 3d-f, nested bootstrap, Bonferroni corrected in 3d).FI rose in both young and aged groups over the four imaging days, but the increase was more pronounced in aged animals so that the differences between young and aged animals were no longer significant in Days 3 and 4 (Figure 3g-i, nested bootstrap, Bonferroni corrected in 3g).These data suggest that representations of space in aged animals can be rescued by repeated exposure to the same environment.
Previous studies found the stability of neural representations of space was altered in CA1 and CA3 in aged animals, but the results differed between aged place cells being more "rigid" or "multistable" depending on the experimental paradigm (Barnes et al., 1997;Tanila et al., 1997b;Wilson et al., 2004).Given these diverging findings, we asked whether spatial representations in the DG would show changes in stability in our imaging paradigm.
We tracked the activity of individual neurons across days to investigate the stability of the active cell ensemble (Figure 4a-c).We first asked whether cells that were active in the initial imaging session were reactivated on subsequent days.Surprisingly, we found no difference between groups, suggesting that both DG neurons in young and aged mice have similar reactivation rates (Figure 4d, mixed effects model with Sidak multiple comparisons test).We went on to analyze the cells that were active both on Day 1, when the treadmill was novel, and on Day 4, when it had become familiar.The single-cell activity of matched cells increased across days in the young mice (p = 0.0023, nested bootstrap) but not in the old mice (p = 0.16, nested bootstrap) (Figure 4e,f).This may simply reflect a ceiling for activity since the aged mice already had higher activity on Day 1.While the tuning index did not differ in either group across days (Figure 4g,h, p = 0.076, p = 0.21), both groups saw higher spatial information, as FI was significantly increased in reactivated cells on Day 4 (Figure 4i,j, p = 0.00015, p = 0.00024), which again may simply reflect the fact that FI increases sharply for the entire population across the 4 days.We then asked whether the reactivated cells had a distinct profile compared to those neurons that were not active across days.Reactivated cells did not differ from other cells in single-cell activity levels.Whether cells were reactivated or not, their activity levels were higher in the aged mice on Day 1, which recapitulates the pattern seen in the general population (Figure 4k).This hyperexcitability was gone by Day 4, despite maintained hyperactivity overall (Figure 4l).In the aged group, cells that were reactivated showed higher tuning levels than other cells on Day 1 (Figure 4m), but this effect was absent on Day 4 (Figure 4n).Reactivated cells had significantly higher FI than other neurons in both the young and aged groups on both Day 1 and Day 4 (Figure 4o,p).This suggests that higher spatial information is a predictor of whether neurons will be reactivated over several days.Overall, we did not see evidence of any changes in the stability of coding ensembles in aged animals when compared with the young cohort.However, since reactivated neurons initially have higher tuning and spatial information than nonreactivated cells in aged animals, and increase their FI over time, it is possible that they contribute to the improvement of spatial representations in aged mice over the 4 days of imaging.
In summary, we have found that DG neurons in aged animals are hyperexcitable and impaired in their encoding of spatial features in novel environments.Our results highlight how the defects in spatial representations in the DG are specific to the first introduction to a novel context and can be rescued by increasing familiarity with a given environment.

| DISCUSS ION
Aging is associated with cognitive decline in spatial memory in healthy aged humans and rodents (Konishi et al., 2017;Techentin et al., 2014).In this study, we asked whether aged mice have impaired representations of space in the DG circuits that mediate spatial memory, which could account for the memory deficits seen in aged individuals.
Several studies have found that neuronal activity is elevated in the hippocampus of aged animals specifically in hippocampal area CA3 (El-Hayek et al., 2013;Lee et al., 2021;Simkin et al., 2015;Wilson et al., 2005).Human studies using fMRI in older and younger adults have generally confirmed the hyperexcitability findings in the DG-CA3 axis (Diersch, Valdes-Herrera, Tempelmann, & Wolbers, 2021;Yassa et al., 2011), although fMRI does not have the resolution to distinguish between areas DG and CA3.This aging-related hyperexcitability phenotype is thought to contribute to the memory deficits seen in aged individuals (Haberman et al., 2017;Jiménez-Balado & Eich, 2021).In order to understand the causes of this hyperexcitability, it would be important to record neuronal activity from the DG, as this region provides the main inputs to CA3.
We used in vivo two-photon calcium imaging to record activity from young and aged DG neurons.This allowed us to image a large population of cells and therefore have a better chance of capturing active cells in a sparsely active population.Additionally, using an imaging-based approach allowed us to study the same group of cells across time.We have found that single-cell calcium activity is indeed increased in aged DG neurons when compared to young controls (Figure 1e).
We also found that during the initial exposure to the treadmill, there is a significant reduction in spatial information and tuning in aged DG neurons, as compared to their young counterparts (Figure 2b,c).By using untrained mice for our experiments, we were able to capture the very first neuronal representations of a novel environment.We found that during the course of four consecutive imaging days, spatial information and tuning in the aged mice converged with those of young mice, which leads us to conclude that aged mice require additional exposure to a novel environment in order to form adequate neural representations of space.It can therefore be said that the spatial encoding defects seen in aged mice are specific to novel environments.This is in agreement with previous reports that found that spatial representations in the aged CA1 were disrupted when novel environmental cues were introduced during a navigation task (Tanila, Sipilä et al., 1997;Wilson et al., 2004).Interestingly, both young and aged groups saw an increase in spatial information through the four days of imaging, whereas spatial tuning increased in aged mice but was not significantly different in young mice.We did not find a convergence in single-cell calcium activity levels between aged and young groups, meaning that the hyperexcitability is no impediment to aged mice eventually forming accurate spatial representations.There are several interesting parallels between our findings and human data showing that aged individuals are better able to navigate familiar environments than novel ones (Muffato et al., 2015) leading them to avoid taking novel routes in favor of navigating familiar routes.
To verify that aged mice have impairments in spatial memory, we used an object placement test that is not physically tasking and does not involve an aversive component.Our behavior data confirmed that aged mice show a deficit in this task compared to their younger counterparts.This goes along with studies analyzing similar spatial memory tasks that have found deficits in aged animals (Wimmer et al., 2012;Yanai et al., 2022).Our results also highlight the welldescribed heterogeneity present during normal aging, as some of the aged mice were still able to perform well in this test (Hamieh et al., 2021).
In order to better compare our data to previous studies of DG activity using IEG labelling, we determined the percentage of active cells in our imaging fields but found no difference in the number of active cells between young and aged animals.However, there are also some caveats to our experimental data that must be kept in mind: We calculate the total number of cells from mean projections of calcium imaging movies, which can introduce several biases, for example active cells will tend to be brighter and therefore more visible.We found a correlation between the percentage of active cells and the total number of cells in the imaging field (Figure S2b), as fields of view with fewer detected cells had much higher variance of the fraction of active cells.Fields with more total cells generally had more cells that were dimmer and less active but were still counted in the total.In contrast, in fields with few cells due to areas of lower viral expression or obscured by blood vessels, there are likely more cells present than can be quantified in the mean projection image.
These differences in imaging fields across animals make it difficult to calculate exact percentages, limiting the usefulness of this approach.
Previous studies have found that increased hippocampal excitability contributes to aging-related cognitive impairments.Our experimental design did not allow us to determine whether hyperexcitability contributes to the deficits in spatial information seen in the novel environment during the first day of imaging.However, our results suggest that instead dysfunction in the plasticity mechanisms that underlie spatial selectivity may be to blame.Spatial selectivity can develop very fast in CA1 hippocampal neurons through a mechanism termed behavioral timescale synaptic plasticity (Bittner et al., 2017) that is mostly active in novel contexts, leading to the development and stabilization of spatial tuning within the first few minutes of exposure to a new environment (Priestley et al., 2022).In addition, other experience-dependent changes occur in CA1 place cell activity, including an increase in the skewness of tuning curves as the animals become familiar with a new environment (Mehta et al., 1997).Less is known about the dynamics of spatial selectivity in the DG, which plays a different role in spatial encoding than CA1 and contains cells with much different firing patterns (Alkadhi, 2019;Jung & McNaughton, 1993;Leutgeb et al., 2007).However, spatially tuned cells in the DG also seem to emerge and stabilize within the first few minutes of exposure to a novel environment, with further refinement as the animals are re-exposed to the same environment over the following days (Kim et al., 2020).This is in line with our finding that spatial information increases over several days both in young and aged animals.Previous literature has shown that aging is associated with plasticity deficits in the hippocampus (Burke & Barnes, 2006;Froc et al., 2003), namely in long-term potentiation and depression.Given that these plasticity mechanisms are likely most active as mice map out new environments (Priestley et al., 2022), we speculate that this may explain the protracted development of spatial selectivity that we see in aged mice.Expression of some IEGs, such as Arc and Egr1/Zif268, which are thought to be regulators of neuronal plasticity, is reduced in the DG of aged animals (Desjardins et al., 1997;Marrone et al., 2012;Penner et al., 2016).Deficits in the expression of these genes following neuronal activity could be at the origin of plasticity defects in aged animals.Another factor potentially contributing to the delayed formation of spatial representations in the aged DG is a reduction of adult neurogenesis.
Adult-born neurons are known to enhance DG neuronal plasticity (Snyder et al., 2001) and reduce responses to novelty in non-spatial tasks (Denny et al., 2012;Lemaire et al., 1999), but their numbers fall to almost zero in aged animals (Kuhn et al., 1996).Overall, our findings support the idea that despite the loss of inputs from EC (Burke & Barnes, 2006;Geinisman et al., 1992;Smith et al., 2000), and the lower numbers of IEG-expressing neurons, activity is increased in the aged DG, which likely contributes to downstream hyperexcitability.
The increased excitability of aged DG neurons could therefore also be seen as a compensation mechanism for both the reduced synaptic input from EC and the reduced expression of IEGs.Several potential mechanisms could underlie this.One possibility is that EC inputs, though fewer in number, could be strengthened and more efficient at depolarizing DG and CA3 neurons (Barnes & McNaughton, 1980).
Additionally, the loss of hilar interneurons could further contribute to DG hyperactivity (Koh et al., 2014).
It is likely that the increase in DG excitability will also spread to CA3 and contribute to the previously described aging-related hyperexcitability phenotypes (El-Hayek et al., 2013;Lee et al., 2021;Simkin et al., 2015;Wilson et al., 2005), as dentate granule cells form strong synapses with CA3 pyramidal neurons and constitute the major source of inputs into this region.Furthermore, small changes in the activity of DG inputs could be further amplified by recurrent excitatory activity in CA3 (Tamminga et al., 2010).Ultimately, it is likely that this would result in cognitive deficits, as hyperactivity in the DG and CA3 is linked with defects in memory discrimination between similar experiences (Yassa et al., 2010), one of the cognitive functions that have been associated with these areas.Treatment with an anti-epileptic drug aimed at reducing neuronal excitability was found to reduce CA3/DG hyperexcitability and improve memory performance in patients with mild cognitive impairment (Bakker et al., 2015), suggesting that interventions aimed at correcting the hyperexcitability phenotype could potentially be effective at restoring DG spatial information in novel environments and cognitive function.
Aging-associated memory deficits have also been associated with changes in the stability of spatial representations as some previous studies have found that the aged hippocampal spatial code undergoes larger changes over different sessions than in young mice (Barnes et al., 1997), whereas others have found that spatial representations are more rigid (Tanila et al., 1997a;Wilson et al., 2004).
We did not see changes in the reactivation of cells from one session to another between groups (Figure 4d).This is in contrast to data showing that more distinct dentate populations express the IEG Arc when aged mice are re-exposed to the same environment (Marrone et al., 2012), which could be due to age-related changes in DNA methylation rather than changes in neuronal excitability (Penner et al., 2011).In order to further investigate the properties of cells in the aged DG over time, we matched cells that were active in our imaging fields across days.We did not find major differences in the reactivated cells in any measures between young and aged groups, suggesting that ensembles of neurons follow the same basic mechanisms in the aged DG.When comparing reactivated neurons to the remainder of the population, we found that reactivated cells had significantly higher spatial information (Figure 4o,p), indicating that this may be a marker of whether a neuron will contribute to spatial representations over time.
To summarize, our data show that new DG spatial representations are initially impaired and take longer to form in aged mice.These data expand our knowledge of the network activity and spatial representations in the aged hippocampus and suggest that agingassociated hippocampal hyperactivity is not an impediment to the formation of rich spatial representations.The protracted refinement of spatial representations suggests that the underlying plasticity mechanisms responsible for the development of spatial selectivity are impaired in aged animals, and that these processes are likely to be a relevant therapeutic target for ameliorating the memory deficits associated with aging.

| Animals
Aged mice were 21-26 months old C57Bl6J females originally from Jackson Labs and kindly gifted by Dr. Carolyn Pytte of Queens College.Young mice were 3-4 months old C57Bl6J females purchased from Jackson Labs.A total of 8 aged mice and 9 young mice were used for imaging experiments, and 13 aged mice and 13 young mice were used for behavior.Not all implanted mice were imaged in every daily session, as in some cases the window implants were not sufficiently clear and stable for in vivo calcium imaging.All mice were housed in standard conditions with a 14/10 h light/dark cycle.
Mice were provided food and water ad libitum.All procedures were done during the light part of the cycle and in accordance with the Einstein Institutional Animal Care and Use Committee (Protocol #00001197).

| Viral injections
Mice were anesthetized (induction: 5%, maintenance 0.5% isoflurane in O 2 vol/vol).Following anesthetization, mice were attached to a stereotactic apparatus and the right hemisphere of the dentate gyrus (DG) was injected with 1 μL of a DJ-serotype AAV vector encoding the red-shifted calcium indicator jRGECO1a under the CamKII promoter at 10 12 GC/mL titer.Viral injections were done with a pulled glass pipette using a Nanoject III microinjector (Drummond).The virus was injected at 33 nL per cycle for 31 cycles at a rate of 10 nL/s with a 20 s delay between cycles, followed by a 15 min wait period before the pipette was removed.
The location of the viral injection was determined by first finding the midpoint between bregma and lambda.The mediolateral coordinate was 1.5-1.75mm from the the lambda-bregma midpoint and the dorsoventral coordinate was 1.8-2.0mm from the dura, depending on bregmalambda distance, which ranged from 3 to 4 mm (Table S1).

| Window implantation
Following viral injection, dexamethasone (1 mg/kg) was administered via subcutaneous injection to minimize brain swelling.Optibond (Kerr Dental) adhesive was applied to the skull and cured to help secure the implant once attached.A 3 mm diameter craniotomy was made over the right dorsal DG and the overlaying tissue was removed by aspiration down to the hippocampal fissure where a custom-built cylindrical titanium implant with a glass coverslip on the bottom was inserted over the dorsal surface of the DG (Gonçalves et al., 2016).
The implant and a titanium bar for head-fixation were attached to the skull with dental cement.All mice were given carprofen (5 mg/ kg, subcutaneous) as a post-surgery analgesic.

| Two-photon imaging
In vivo calcium imaging movies were acquired with a custom twophoton laser scanning microscope (based on Thorlabs Bergamo) using a femtosecond-pulsed laser (Coherent Fidelity 2, 1070 nm) and a 16x water immersion objective (0.8 NA, Nikon).Imaging sessions were started 3-4 weeks after surgery to allow for recovery and for optimal viral expression.During imaging sessions, mice were headfixed to the microscope with a titanium headbar and the microscope stage was adjusted so that the hippocampal window was perpendicular to the axis of the objective for optimal imaging conditions.
Mice were awake and walking on a manually rotated treadmill belt throughout imaging.The treadmill contained four zones with different textures as previously described (Jordan et al., 2021).Ten-minute videos were acquired at 15.253 fps with a 343.6 × 343.6 μm field of view.Treadmill data were acquired using National Instruments analog-to-digital converter and synchronized with the imaging data using Thorsync software.To track the same field of view, the initial location was noted on the first imaging day using a coordinate system and by taking an image of the field.During following sessions, the coordinates and initial field image were used to match as close as possible to the same field.

| Data analysis
The open-source Suite2p software (version 10.0) was used to register and motion correct videos, for cell detection and spike deconvolution (Pachitariu et al., 2017).Regions of interest corresponding to active neurons were identified by Suite2p and further manually curated in order to identify all active cells in the field of view.We used the ROIMatchPub function for Suite2p used to automatically identify and match cells in the same field of view across imaging sessions.
Matched ROIs were manually confirmed after automated detection.
Single-cell calcium activity was determined as previously described (Frechou et al., 2022).Briefly, ΔF/F data were filtered with a zero phase shift, third order Butterworth lowpass filter.This filtered ΔF/F was thresholded to 2 standard deviations (σ) above a rolling-mean baseline.Single-cell activity was determined as the cumulative sum of the thresholded trace for each cell.This was then normalized to the total distance traveled by each mouse.
Tuning indices were using deconvolved calcium traces, which were thresholded to 2σ above the baseline.Any point not significantly above that noise threshold was set to zero.Videos were cut to only include periods of movement >1 s in duration.The treadmill band was segmented into 100 position bins, and transients were mapped to these bins according to the location of the mice on the treadmill in order to generate a tuning vector for each cell (Danielson et al., 2016).The mean of the thresholded activity at each location was calculated for every neuron and normalized to the time the mouse spent at that position.The tuning index was defined as the modulus of this normalized tuning vector.
Single-cell Fisher information (FI) was calculated as previously described (Frechou et al., 2022;Kanitscheider et al., 2015).A biascorrected signal to noise ratio was computed using the unfiltered ΔF/F fluorescence data for each individual cell, where the signal is the square of the difference of the mean activity at two locations on the treadmill, and the noise is the average variance of the activity at each location.Position data were segmented into 20 bins.

| Statistics
Nested bootstrap analysis-When pooling cell data from different mice into a single experimental group, significance testing was done using a muti-level bootstrapped approach as previously described (Frechou et al., 2022).This approach allows for the fact that cells observed within the same animal are not truly independent observations and takes intra-animal variance into consideration (Field & Welsh, 2007;Saravanan et al., 2020).To assess whether differences between two experimental groups are significant, a null surrogate distribution was constructed for each mouse in each group by resampling with substitution from the pool of all imaged cells.
The difference between the null distributions generated for each group was calculated and the previous steps repeated to generate 100,000 bootstrap estimates of the difference between two groups' null distributions.The empirically observed value of the difference between conditions was then compared to the null distribution values with a statistical significance level (α) set at 0.05.If the empirical group differences fall outside of the 95th percentile of the 100,000 bootstrap estimates of the difference between the null distributions, then it is considered to be a statistically significant difference.The p-value is the proportion of bootstrap different estimates that are larger than the empirical difference between groups (or smaller, if the difference is negative).Whenever comparing more than two groups, Bonferroni's correction for multiple comparisons was applied.

F
I G U R E 1 DG is hyperactive in aged animals.(a) Experimental timeline, including AAV injection, window implantation, and two-photon imaging.(b) Diagram of chronic window implant over the right hemisphere of the DG.(c) Diagram of imaging setup, with side view (left) and top view (right) of mouse headfixed to treadmill with multiple tactile zones.(d) Example field of view with regions of interest of active cells shaded in color.Scale bar = 100 μm.(e) Mean single-cell calcium activity.(f) Example calcium traces (black) and corresponding treadmill positions (red) from young mice.(g) Example calcium traces (black) and corresponding treadmill positions (red) from aged mice.a.u., arbitrary units.Statistics done with nested bootstrap analysis.Bars represent mean ± SEM. a.u., arbitrary units.Young N = 8 mice, n = 910 cells; Aged N = 8 mice, n = 699 cells.**p < 0.01.See also Figures S1 and S2.kept the same place response in both even and odd laps.Aged mice encoded less spatial information than young mice as denoted by a significant decrease in FI (Figure 2f, Young N = 8, n = 910; Aged N = 8, n = 699; p = 0.00005, nested bootstrap).

F
I G U R E 2 DG representations of space are impaired in aged mice.(a) Tuning vector polar plots representing a highly spatially tuned cell and (b) a cell with low spatial tuning.(c) Spatial tuning index.(d) Raster plots of tuning vectors of young mouse neurons in 10th percentile Fisher Information, sorted by the position of activity maximum for whole recording (All Laps).Raster plots of partial recording (Even Laps and Odd Laps) keep same sorting used for whole recording.(e) Same raster plots for aged mice.(c) Singlecell Fisher information.Statistics done with nested bootstrap analysis.Bars represent mean ± SEM.Young N = 8 mice, n = 910 cells; Aged N = 8 mice, n = 699 cells.*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

F
Reactivated cells do not have a distinct profile in aged animals.(a) Example field of view used to match cells across days.Neurons that were active during recording session are shaded in color.(b) Matched regions of interest from the field of view in panel A, red and white contours represent cells active on Days 1 and 4. (c) Examples of matched cells on Days 1 and 4. (d) Reactivation rate of neurons that were active on Day 1. Percentages are based on imaging day versus Day 1, regardless of whether the cell was active on other days.(e, f) Matched single-cell calcium activity in reactivated cells in young (e) and aged (f) mice.(g, h) Matched tuning index in reactivated cells in young (g) and aged (h) mice.(i, j) Matched single-cell Fisher information in reactivated cells in young (i) and aged (j) mice.(k, l) Single-cell activity of reactivated and nonreactivated cells on Day 1 (k) and Day 4 (l).(m, n) Tuning of reactivated and nonreactivated cells on Day 1 (m) and Day 4 (n).(o, p) Single-cell Fisher information of reactivated and nonreactivated cells on Day 1 (o) and Day 4 (p).Statistics done with nested bootstrap analysis.Bars represent mean ± SEM. a.u., arbitrary units.Young: N = 6; Day 1 n = 617; Day 1 react.n = 140; Day 4 n = 396; Day 4 react.n = 140; Aged: N = 5; Day 1 n = 375; Day 1 react.n = 80; Day 4 n = 196; Day 4 react.n = 80.ns, not significant, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.
Object Placement-A square-shaped arena 42cm × 42cm × 30cm (L × W × H) was set up with visual cues along the walls and with two novel and identical objects on the floor.Mice were individually placed in the arena for a 4 min training trial.Following a 45 min retention period, one object was displaced to another location in the arena and the mice were placed again for a 4 min test trial.Exploration of the objects was quantified by an experimenter and confirmed with computer tracking (Viewer, Biobserve).
Mixed-effects models with Geisser-Greenhouse correction for matched values were used to compare mean values by mouse across imaging days.Sidak's multiple comparisons test was used to compare groups on each imaging day.Nonparametric one sample Wilcoxon tests were used to determine if group medians were significantly above chance levels in behavior experiments.A two-way ANOVA with Sidak's multiple comparison test was used to compare exploration times between objects in behavior experiments.