Responsiveness of dentate neurons generated throughout adult life is associated with resilience to cognitive aging

Abstract During aging, some individuals are resilient to the decline of cognitive functions whereas others are vulnerable. These inter‐individual differences in memory abilities have been associated with differences in the rate of hippocampal neurogenesis measured in elderlies. Whether the maintenance of the functionality of neurons generated throughout adult life is linked to resilience to cognitive aging remains completely unexplored. Using the immediate early gene Zif268, we analyzed the activation of dentate granule neurons born in adult (3‐month‐old), middle‐aged (12‐month‐old), or senescent (18‐month‐old) rats (n = 96) in response to learning when animals reached 21 months of age. The activation of neurons born during the developmental period was also examined. We show that adult‐born neurons can survive up to 19 months and that neurons generated 4, 10, or 19 months before learning, but not developmentally born neurons, are activated in senescent rats with good learning abilities. In contrast, aged rats with bad learning abilities do not exhibit activity‐dependent regulation of newborn cells, whatever their birthdate. In conclusion, we propose that resilience to cognitive aging is associated with responsiveness of neurons born during adult life. These data add to our current knowledge by showing that the aging of memory abilities stems not only from the number but also from the responsiveness of adult‐born neurons.

formation (Gonzalez-Escamilla, Muthuraman, Chirumamilla, Vogt, & Groppa, 2018). Several models and theories (maintenance, reserve, compensation) emerged in an effort to account for variability in cognitive outcome across old subjects, and high level of neural plasticity has been proposed for brain reserve and resilience to cognitive aging (Nyberg et al., 2012).
The ability of the adult brain, and in particular the dentate gyrus (DG) of the hippocampus, to create new neurons is a peculiar form of plasticity to protect the aging brain. Briefly, new dentate granule neurons (DGNs) generated throughout the entire life of an individual (Aimone et al., 2014), humans included (Moreno-Jiménez et al., 2019), are integrated into functional circuits and play a crucial role in complex forms of learning and memory (Clelland et al., 2009;Dupret et al., 2008;Tronel et al., 2012).
In particular, both the addition and the elimination of new neurons before, during, or after learning were found to be important for learning and forgetting in young adult rodents (Akers et al., 2014;Dupret et al., 2007).
During aging, the rate of cell proliferation (and thus neurogenesis) decreases (Drapeau & Abrous, 2008), a process associated with the progressive loss or phenotypic and functional changes of  (Drapeau et al., 2003). Moreover, we have found that corticosterone dampening from middle age on has a beneficial effect on both the rate of neurogenesis and spatial memory measured once animals have reached senescence (Montaron et al., 2006). Altogether this dataset raises the fascinating hypothesis that neurons generated throughout adult life could constitute a mechanism that promotes resilience to cognitive aging.
To tackle this question, we examined whether the activation of DGNs generated throughout adult life was related to the maintenance of memory function by imaging them when animals reached senescence. DGNs born in adult (3-month-old), middle-aged (12-month-old), or senescent (18-month-old) rats were labeled with analogs of thymidine, and their activation in response to spatial learning was measured using Zif268, an immediate early gene (IEG) (Tronel, Lemaire, Charrier, Montaron, & Abrous, 2015), when animals have reached senescence. The activation of DGNs born during development was also examined.

| Fate of dentate granule neurons born in old rats
In a first step, we sought to determine whether new neurons born during senescence are recruited by spatial learning. To do so, 18-month-old rats were injected with BrdU according to a previously described protocol (Table 1) and were trained 4 months later in the water maze using a reference memory protocol (Drapeau et al., 2003). Animals were trained for 11 days ( Figure S1a,b) until the aged-unimpaired rats (AU) learned the task (day effect on Latency: F 11,66 = 2.35, p = .016; day effect on Distance: F 11,66 = 2.76, p = .005) and reached asymptotic levels of performances (with no statistical significant differences over the last 3 days). In contrast, the aged-impaired (AI) rats did not learn the task although they were searching and finding the platform most of the time ( immunohistochemistry. At the time of sacrifice, BrdU-IR cells were 4 months old and their majority was located within the granule cell layer (GCL) (Figure 1a).
These cells were more numerous in the GCL of animals with good learning abilities (AU) compared to animals with memory deficits (AI) ( Figure 2a, F 2,16 = 7.64, p = .05 with C = AI<AU at p < .01). This finding is consistent with our previous study showing that the number of neurons generated 1 month after learning is higher in AU compared to AI (Drapeau et al., 2003)  To determine whether newborn neurons are recruited by learning, we used Zif268 since this IEG is still expressed in the old DG (Gheidi, Azzopardi, Adams, & Marrone, 2013). Given that a substantial fraction of cells generated during senescence did not express NeuN, we verified in trained animals that Zif268-expressing cells were expressing NeuN ( Figure 1c). We found that the vast majority of activated cells (Zif268) were neurons (NeuN) and that this ratio was similar in good and bad learners (AI: 96.4 ± 0.5; AU: 96 ± 1.3, p > .05). Then, we examined the activation of adult-born cells, meant to be neurons, in response to learning (Figure 1d). We found that the percentage of BrdU-IR cells expressing Zif268-IR in aged animals with good learning abilities was greater than that of aged animals with memory deficits and that of the untrained control group (Figure 2c, F 2,16 = 3.70, p = .05 with C = AI<AU at p < .05). In contrast, the total number of Zif268-IR nuclei did not differ between groups

| Fate of dentate granule neurons born in middle-aged and young rats
Then, we asked whether neurons born earlier, that is, in middle age or young adulthood, are also recruited by learning during aging. For this purpose, animals were injected with CldU when 3 months old and with IdU when middle aged (12 months old;

| Fate of dentate granule neurons born during development
Finally, we explored the role of dentate granule born during development of the DG by tagging neurons born in adolescent rats (PN28, Figure 1k) and neurons born in embryos (E18.5, Figure 1l) with CldU.  (Table S1).

| D ISCUSS I ON
To determine whether neurons generated during adult life participate in learning abilities in old age, the expression of the IEG Zif268 was assessed in new neurons. We found that cells generated during young adulthood, middle age, and senescence survive for a long period of time and are functionally integrated into the dentate network.
When taking into account individual differences in memory abilities, we highlight that although the number of new cells generated in 12-month-old animals (IdU-IR cells) is decreased 10-fold compared to 3-month-old rats (CldU-IR cells), the total number of CldU-IR or IdU-IR cells measured when animals reached senescence is similar in AU and AI and not different from untrained control animals.
These conclusions have been obtained using the hippocampal-dependent version of the Morris water maze (measuring reference memory with variable starting points), one of the most widely used behavioral tests to study normal aging in rats; this task does not involve food restriction or the administration of shock to encourage participation of old animals and is very sensitive to the aging process (Kennard & Woodruff-Pak, 2011;Lubec et al., 2019;Mota et al., 2019). Although inter-individual differences in memory in aged rats have been demonstrated in other type of mazes (Barrett, Bennie, Trieu, Ping, & Tsafoulis, 2009;Temido-Ferreira et al., 2018), in object location memory tasks (Lux, Masseck, Herlitze, & Sauvage, 2017), and in the hole-board task (Lubec et al., 2019), we chose to use the water maze because it is the most widely used task to study the implication of adult-born neurons in aged rats (our own work, (Bizon & Gallagher, 2003;Marrone, Ramirez-Amaya, & Barnes, 2012)) whereas their role remains largely unexplored using the other tasks.
To map neuronal activity, Zif268 was preferred over many other markers such as c-Fos or Arc that have been frequently used in the memory field (Gallo, Katche, Morici, Medina, & Weisstaub, 2018). Zif 268 presents the advantage to have a high "basal" expression which increases the probability to map new neurons, the number of which decreases with aging. In addition, we have previously shown using Zif 268 and c-Fos that new neurons are recruited during both spatial memory acquisition and retrieval , indicating that either one or the other IEG can be used to analyze adult-born neurons' activation in response to learning.
To study adult-born neuronal activity, we used immunohistochemistry, the only method available to study their role without impairing their activity. This technical approach is dependent on F I G U R E 3 Neurons produced during middle age are activated by spatial learning in aged good learners. Top: Experimental design. (a) The number of IdU-IR cells generated at mid-age is independent of the memory abilities measured when rats reached senescence. (b) The percentage of cells differentiating into neurons (IdU-IR cells expressing calbindin) is slightly increased in AI (compared to C and AU). (c) The expression of Zif268 in IdU-IR cells is increased in AU rats compared to AI rats and C rats. *p < .05, ** < .01 compared to AU. °p < .05 compared to c were processed together from perfusion to labeling, we tend to exclude a technical bias. However, visualizing the activity of new neurons during learning using calcium imaging is now the best available technic to confirm our conclusions in behaving rats in the course of aging.
Our conclusion that the responsiveness of dentate neurons generated throughout adult life is associated with resilience to cognitive aging has been obtained using two different cohorts of rats. The first one was utilized to study Adu-DGNs generated in senescent DG and the second one to study Adu-DGNs generated in young adult and middle-aged DG. When comparing the behavior of the two batches of rats, it appears that deficits in the aged-impaired rats were much more pronounced in the first batch of animals. This cohort effect, a well-known phenomenon in aging research (Schaie & Willis, 2015), could be related to housing conditions. Indeed, the first batch was raised in the vendor facilities until 16 months of age whereas the second one was raised in-house. Supporting this, in our previous experiments performed in rats not aged in-house, the difference between AU and AI was more pronounced than that observed in the second experiment (Drapeau, Montaron, Aguerre, & Abrous, 2007).
However, independently of the cohort of rats the same profile of activation was observed (C = AI <AU).
While the process of neurogenesis has been well characterized in young adult rodents (Aimone et al., 2014), information about their aging and their function is less abundant (Drapeau & Abrous, 2008;Encinas & Fitzsimons, 2017;McAvoy & Sahay, 2017). The number of stem cells, their rate of proliferation, and neuronal differentiation dramatically decrease with age, and the number of immature neurons (<4 weeks old) is thus significantly decreased. Recently, the development and functional integration of these cells have been described to be delayed by age. Indeed, 3-week-old neurons generated in middle-aged mice (10-14 months) displayed shorter and simpler dendrites and a dramatic reduction in spine number compared to those generated in 2-month-old mice and exhibited immature neuronal electrophysiological properties as revealed by the lack of functional glutamatergic synaptic inputs (Trinchero et al., 2017). In fact, their overall mature excitability and maximal glutamatergic connectivity are delayed compared to neurons born in younger animals (Trinchero, Herrero, Monzon-Salinas, & Schinder, 2019). The longterm destiny of adult-born neurons generated in young adult animals has not been explored in depth. We and others have shown that F I G U R E 4 Neurons produced during young adulthood are activated by spatial learning in good leaners. Top: Experimental design. (a) The numbers of CldU-IR cells generated when animals are young adult are independent of the memory abilities measured when rats reached senescence. (b) The percentage of CldU-IR cells expressing calbindin is increased by training. (c) The expression of Zif268 in CldU-IR cells generated in young adult DG is increased in AU compared to AI rats and C rats. °p < .05 compared to AU, +p < .05 compared to AI. *p < .05, **p < .01 compared to AU   and even years in the DG (present results) and do not show signs of decline in excitability when they age: 5-month-old neurons are as excitable as 1-monthold-cells; they can even exhibit high levels of excitability following either enriched environment exposure or induction of LTP (Ohline et al., 2018). This latest very exciting result supports our hypothesis that even when several months old, Adu-DGNs are still plastic, they do not retire and participate in memory functions (Abrous & Wojtowicz, 2015), and even more so their persistence is not passive, but a result of their activity.
Here, we found that between middle age and senescence, the number of cells is further decreased, but then a difference between the AU and AI groups appears. Based on previous data, it is likely that the emergence of such a difference results from a difference in cell proliferation (Drapeau et al., 2003), neuronal differentiation (Drapeau et al., 2003;Qiao et al., 2019), cellular senescence we have shown that the senescent neurogenic niche is capable to rejuvenate upon removal of corticosterone (Montaron et al., 1999) or addition of pregnenolone sulfate (Mayo et al., 2005), indicating that neural stem cells are not depleted and keep their abilities to divide.
The main finding of our study is that there is a strong link between the ability for newborn cells to be recruited by learning and memory abilities in aged rats. Indeed, the percentage of adult-born cells expressing Zif268 was higher in animals that learned the task compared to animals that did it to a lesser extent. This finding is in accordance with our previous data showing that (a) when compared to control rats (naïve rats or rats trained to find a visible platform), adults required to use an hippocampal-dependent strategy in the water maze (or the dry maze) exhibit an increased percentage of mature adult-born neurons expressing Zif268 , and (b) ablating mature adult-born neurons generated 4 months before training (when animals were 3 months old) delays the ability of rats to learn such a task (Lemaire et al., 2012). In the present experiment, the percentage of adult-born cells expressing It could be argued that neurons born during development, which represent a major part of the DG, are also involved in differences in spatial memory abilities in old age. However, three arguments seem to rule out this hypothesis. First, the total number of granule cells is similar in AU and AI groups (Drapeau et al., 2003(Drapeau et al., , 2007Rapp & Gallagher, 1996). Second, if neurons generated during development (prenatal and postnatal periods) were activated by spatial learning, given their high numbers, differences in the total number of Zif268 cells should have emerged as a function of the cognitive status. Third, we have shown that neurons born in neonates are not activated by spatial learning when they are mature compared to neurons of the same age born in adults. Indeed, the former are not recruited by spatial learning in the water maze when animals are tested at 7 months . Here, we extended this observation showing that neurons born during the juvenile (PN28) or the embryonic (E18.5/19.5) period are not differentially recruited in good and bad learners. Recently, we began to explore the reasons for which developmentally and adult-generated neurons do not respond in the same way. By comparing the dendritic arbor of neurons born at different ontogenetic stages (embryonic, neonatal, adolescence, adulthood), we found that they display distinct morphological features and also different location (Kerloch, Clavreul, Goron, Abrous, & Pacary, 2019;Mathews et al., 2010) that may underlie different inputs and functions.
One question that we did not address is whether the three neuronal populations studied participate to the same extent to learning.
To address this point, sophisticated models that allow to selectively tag new neurons generated within a defined period of time (adulthood, middle age, or senescence) and to ablate them during training performed at senescence are required. One possibility would be to take advantage of the recently developed pharmacogenetic or optogenetic approaches or in order to tag specifically neurons born in young adult rats and manipulate them when animals have reached senescence.
A previous study has shown that 4-month-old neurons generated in old rats exhibiting spatial memory deficits are recruited in response to spatial exploration behavior with the same probability than 4-month-old neurons generated in aged good learners or in young adult rats (Marrone et al., 2012). From this dataset, it was concluded that disrupted information processing at old age may be linked to a reduced number of adult-generated granule neurons and not to a deficit in their functionality. However, in this study the activation of adult-generated neurons was evaluated in response to a simple form a learning (spatial exploration). Taking the present data into consideration, we rather suggest that adultborn neurons in AU are sufficiently connected to integrate simple stimulations generated during simple forms of learning but insufficiently integrated to process the complex stimulations generated during spatial navigation.
Zif268 is known to be regulated in an activity-dependent manner by learning (for review, see Gallo et al., 2018;Veyrac, Besnard, Caboche, Davis, & Laroche, 2014). It is overexpressed in response to different types of learning in distinct structures and circuits that are processing the ongoing information, and several arguments indicate that it is required for the stabilization (and not acquisition) of long-lasting memories. Although the mechanisms are not fully understood, the activation of Zif268 may strengthen/stabilize the memory trace. It can be hypothesized that during learning the activation of Zif268 in adult-born neurons may be involved in the formation, stabilization, and reactivation of place cells in the hippocampal network, events known to support spatial learning.
Here, we hypothesize that adult-born neurons that do not exhibited activity-dependent regulation of Zif268 become functionally silent in the course of aging, leading to memory deficits. Although the firing patterns that are sufficient to induce Zif268 in adult-born neurons in "behaving" animals are so far unknown, adult-born neurons silencing may have several origins. It may result from a loss of synaptic inputs (Geinisman, Toledo-Morrell, & Morrell, 1986) altering the ability to fire properly (Ahlenius, Visan, Kokaia, Lindvall, & Kokaia, 2009); these synaptic alterations of Adu-DGNs could be linked to the acceleration of senescence through epigenetic changes (Penner, Roth, Barnes, & Sweatt, 2010), decreased autophagy activity (Glatigny et al., 2019), or changes of the local and systemic milieu (Villeda et al., 2011).
The HPA axis (and corticosterone) deserves a special attention in relation to its role in age-related memory disorders (Sapolsky, 1992).
In his pioneer work, Issa and colleagues showed that AU rats were characterized by high level of corticosterone in basal condition and in response to a restraint stress (Issa, Rowe, Gauthier, & Meaney, 1990). We replicated and extended this finding by showing that animals with the heaviest adrenal glands, indicative of chronic HPA axis hyperactivity, exhibited the worst memory performance and the lowest number of proliferating cells or 3-week-old surviving cells in comparison with AU rats (Drapeau et al., 2003). Although some reports failed to support a relationship between cell birth and corticosterone levels during aging (using different rat strain (Heine, Maslam, Joels, & Lucassen, 2004), we can hypothesize that in the present experiment AU and AI rats exhibit different HPA axis activity both in basal condition and in response to training. This raises the possibility that differences in the reactivity of adult-born neurons in these animals are related to differences in the activity of the HPA axis. The influence of learning in the water maze on HPA axis activity is largely unknown at the exception of one study showing that corticosterone secretion is increased (in adult rats) immediately after the first training session and that corticosterone has pro-mnesic effect on learning (Sandi, Loscertales, & Guaza, 1997). Whether or not learning still increases corticosterone at the end of training when animals get habituated to the task is not known. But independently of this, AI exhibit a hyperactive HPA axis and it cannot be excluded that excessive levels of corticosterone may impair adult-born neurons' functioning and memory processing. As a consequence, blocking/ removing glucocorticoid receptors expressed by adult-born neurons generated early in life at the time of training may increase their responsiveness to learning and promote memory. In this context, we have previously shown that blocking age-related increased in HPA axis activity form middle-age onward rejuvenates memory and the neurogenic niche in senescent subjects (Montaron et al., 2006).
In conclusion, our results highlight the importance of neurons born throughout adult life in providing resilience to age-related memory disorders. They reveal a novel perspective for developing therapies to promote resilience to age-related memory disorders or to rejuvenate the DG by acting throughout adult life on adult-born dentate neurons.

| Animal
For these experiments, a total of 96 male Sprague Dawley rats (OFA, Janvier, France) were used. Animals were housed collectively until behavioral testing under a 12-hr:12-hr light/dark cycle with ad libitum access to food and water. Temperature (22°C) and humidity (60%) were kept constant.
In the first experiment, male rats (n = 19) were 16 months old on delivery. In the second experiment, rats (n = 32) were 2 months old on delivery. In the third and fourth experiments, rats (n = 25) were 21 days old on delivery. In the fifth experiment, pregnant Sprague

| Thymidine analog injections
Newly born cells were labeled by the incorporation of synthetic thymidine analogs (XdU, Sigma-Aldrich, Table 1). In the first experiment, rats were injected with 5-bromo-2′-deoxyuridine (BrdU) according to a previously described protocol (Drapeau et al., 2003(Drapeau et al., , 2007. These animals received one daily BrdU injection (50 mg kg -1 day -1 ; ip) for 5 days when 18 months old, that is, 4 months before training.

| Water-maze training
Rats were tested in the water maze when 22 months old (experiments 1, 2, and 4) or 15 months old (experiments 3 and 5). The apparatus consisted of a circular plastic swimming pool (180 cm diameter, 60 cm height) that was filled with water (20 ± 1°C) rendered opaque by the addition of a white cosmetic adjuvant. Before the start of training, animals were habituated to the pool for 2 days for 1 min per day. During training, the Learning group (L) was composed of animals that were required to locate the submerged platform, which was hidden 1.5 cm under the surface of the water in a fixed location, using spatial cues available within the room. Rats were all trained for four trials per day (90 s with an inter-trial interval of 30 s and released from 3 different starting points that varied randomly each day). If an animal failed to locate the platform, it was placed on it at the end of the trial. The time necessary to reach the platform was recorded using a video camera that was secured to the ceiling of the room and connected to a computerized tracking system (Videotrack, Viewpoint). Daily results were analyzed in order to rank animals according to their behavioral score calculated over the last 3 days of training (when performances reached an asymptotic level).
The behavioral scores calculated over the whole training duration of aged-unimpaired (AU) rats were below the median of the group, whereas those of aged-impaired (AI) animals were above the median of the group. Control groups consisted of animals that were transferred to the testing room at the same time and with the same procedures as trained animals but that were not exposed to the water maze.

| Immunohistochemistry
Animals were sacrificed 90 min after the last trial (Table 1)

| Activation of new cells
The activation of adult-born cells was examined using immunohistofluorescence. To visualize cells that incorporated thymidine analogs, one-in-ten sections were incubated with different anti-BrdU antibodies (BrdU and CldU, rat primary antibodies at 1/200 Accurate Chemical; IdU, mouse primary antibodies at 1/200, BD Biosciences).

| Statistical analysis
All data are expressed as mean ± SEM. Data were analyzed using an ANOVA or Student's t test (2 tails) when necessary.

ACK N OWLED G M ENTS
The authors thank Dr M Koehl (Inserm, UMR 1215) and Dr A Schinder (Leloir Institute, BA, Argentina) for useful discussion. We greatly acknowledge C Dupuy for animal care and A Mangin for his technical help. This work was supported by Institut National de la

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interest.

AUTH O R CO NTR I B UTI O N S
MFM performed experiments. VC analyzed the recruitment of DGNs of rats from batches 4 and 5 and revised the paper. NB analyzed the recruitment of DGNs of animals from batch 3. PG performed experiment from batch 1. DNA conceived experiments, analyzed data and wrote the paper.

DATA AVA I L A B I L I T Y S TAT E M E N T
The authors elect to not share data.