Aging is not equal across memory systems

Abstract

The present experiments compared the effects of aging on learning several hippocampus- and striatum-sensitive tasks in young (3–4 month) and old (24–28 month) male Fischer-344 rats. Across three sets of tasks, aging was accompanied not only by deficits on hippocampal tasks but also by maintained or even enhanced abilities on striatal tasks. On two novel object recognition tasks, rats showed impaired performance on a hippocampal object location task but enhanced performance on a striatal object replacement task. On a dual solution task, young rats predominately used hippocampal solutions and old rats used striatal solutions. In addition, on two maze tasks optimally solved using either hippocampus-sensitive place or striatum-sensitive response strategies, relative to young rats, old rats had impaired learning on the place version but equivalent learning on the response version. Because glucose treatments can reverse deficits in learning and memory across many tasks and contexts, levels of available glucose in the brain may have particular importance in cognitive aging observed across tasks and memory systems. During place learning, training-related rises in extracellular glucose levels were attenuated in the hippocampus of old rats compared to young rats. In contrast, glucose levels in the striatum increased comparably in young and old rats trained on either the place or response task. These extracellular brain glucose responses to training paralleled the impairment in hippocampus-sensitive learning and the sparing of striatum-sensitive learning seen as rats age, suggesting a link between age-related changes in learning and metabolic substrate availability in these brain regions.

1. Introduction

Many studies of age-related changes in learning and memory use tasks that rely on hippocampal processing. For example, old rats learn spatial versions of land mazes, including the 8-arm radial maze and the circular platform task, more slowly than do young adult rats (Barnes, 1979; Barnes, Nadel, & Honig, 1980; Barrett, Bennie, Trieu, Ping, & Tsafoulis, 2009; McLay, Freeman, Harlan, Kastin, & Zadina, 1999) and also show more rapid forgetting of previously learned spatial information (Barnes, 1979). The cognitive deficits of aged rats in the circular “Barnes” maze are accompanied by parallel impairments in synaptic plasticity in the dentate gyrus (Barnes, 1979). Acquisition of hippocampus-sensitive trace eyeblink conditioning (e.g., Disterhoft et al., 1996, 1999; Knuttinen, Gamelli, Weiss, Power, & Disterhoft, 2001) and trace fear conditioning (Blank, Nijholt, Kye, Radulovic, & Spiess, 2003; McEchron, Cheng, & Gilmartin, 2004; Villarreal, Dykes, & Barea-Rodriguez, 2004) are also sensitive to aging, with impairments paralleled by reduced intrinsic membrane excitability related to changes in calcium and potassium currents in hippocampal pyramidal neurons; notably, a longer post-burst afterhyperpolarization phase and increased spike-frequency adaptation accompanies aging (cf.: Oh & Disterhoft, in press), and are thought to contribute to cognitive impairment. Additional experiments report impaired spatial learning in aged rats trained on the swim task (Chen, Masliah, Mallory, & Gage, 1995; Gallagher & Nicolle, 1993; Gallagher, Burwell, & Burchinal, 1993; Quirion et al., 1995; Rose & Rowe, 2012). Moreover, on this task, several biological measures distinguish memory-impaired from memory-unimpaired aged rats, including enlarged after-hyperpolarizations (Oh & Disterhoft, in press) and, for example, markers of neurotransmitter-related functions (Abdulla, Abu-Bakra, Calaminici, Stephenson, & Sinden, 1995; Jiang, Owyang, Hong, & Gallagher, 1989; Le Jeune, Cécyre, Rowe, Meaney, & Quirion, 1996) implicated in learning and memory (e.g., Gold, 2003; Izquierdo et al., 1993; McDaniel, Mundy, & Tilson, 1990).

While age-related learning and memory deficits across hippocampal tasks are seen consistently, there is also evidence that cognitive abilities sensitive to manipulations of other brain regions are preserved with age. For example, when using the swim task to assess impaired spatial learning in aged rodents, the hidden platform (spatial) version of the task is typically compared with results obtained in a cued version, the latter of which aged rats learn as quickly as do young rats. The cued version is generally included as a control task to show that impairments on the spatial task result from learning and memory deficits in old animals rather than contributions of non-mnemonic variables such as sensory/motor function or motivation to escape (cf. Gallagher et al., 1993). Of particular interest here, the retained ability of aged rats to learn the cued version of the swim task also suggests retained cognitive functions across late age that may rely on intact functioning of non-hippocampal memory systems. However, in both young and aged rats alike, the cued task is learned more readily than is the spatial task, confounding interpretations of whether spared learning of the cued task depends on cognitive attributes of the task or cognitive load. Additional evidence for differential changes across memory systems with aging comes from studies showing that strategies to solve tasks switch away from hippocampal-sensitive spatial learning to non-hippocampal stimulus-response solutions in both rats (Barnes et al., 1980; Nicolle, Prescott, & Bizon, 2003; Pereira, Gallagher, & Rapp, 2015) and humans (Bohbot et al., 2012; Harris, Wiener, & Wolbers, 2012; Wiener, de Condappa, Harris, & Wolbers, 2013).

Given the considerable evidence that distinct brain systems process different types of learning and memory (Gold, 2004; Gold, Newman, Scavuzzo, & Korol, 2013; Graybiel & Grafton, 2015; Korol, 2004; Korol & Pisani, 2015; Korol & Wang, 2018; Poldrack & Packard, 2003; White & McDonald, 2002; White, Packard, & McDonald, 2013), reports of decrements and savings in learning and memory with advanced age may be explained by dissociations in the effects of age by memory system. That is, applying a multiple memory system perspective to cognitive aging highlights the exciting possibility of shifts in learning and memory with age as some memory systems are spared or augmented while others decline (e.g., Brown, Robertson, & Press, 2009; Mitchell, 1989; Nilsson, 2003). If aging modulates the relative use of specific memory systems, then demonstrations of age-related impairments, preservations, and enhancements of different learning and memory abilities would be expected, reflecting which memory system mediates the processing for individual tasks. Although observing changes within the hippocampus across age may be important for identifying neurobiological processes related to age-related impairments in learning and memory, comparisons across memory systems impaired or preserved in old age may provide insights into the mechanisms mediating these bidirectional functional trajectories of aging.

Brain glucose availability is one such process that may contribute to age-related shifts in learning and memory. For example, young male rats exhibit significant but relatively modest decreases in hippocampal extracellular fluid (ECF) glucose levels while tested on a hippocampus-sensitive spontaneous alternation task (McNay & Gold, 2001; McNay, Fries, & Gold, 2000; Newman, Korol, & Gold, 2011). Aged rats, however, have a precipitous drop in ECF glucose levels under similar memory testing conditions that also reveal age-related memory impairments (McNay & Gold, 2001). Systemic injections of glucose blunt the depletion of hippocampal glucose levels during memory testing and restore memory scores in aged male rats. More broadly, either systemic (Gold & Korol, 2014; Gold, 2014; Korol, 2002; Meikle, Riby, & Stollery, 2005; Messier, 2004; Smith, Riby, van Eekelen, & Foster, 2011) or intrahippocampal (Morris & Gold, 2013) administration of glucose enhances memory and fully reverses age-related memory impairments in laboratory rodents across several hippocampus-sensitive tasks. Glucose administration also reverses age-related deficits in cognitive functions in healthy elderly individuals (e.g.: Manning et al., 1990, 1992; Wilkniss, Jones, Korol, Gold, & Manning, 1997) and enhances cognitive functions in individuals with Alzheimer’s disease and other brain disorders (Craft, Zallen, & Baker, 1992; Manning et al., 1993, 1998; cf.: Gagnon, Greenwood, & Bherer, 2010; Gold, 2001, 2014; Gold and Korol, 2012, 2014; Korol, 2002; Korol & Gold, 1998; Macpherson et al., 2015; Messier, Tsiakas, Gagnon, & Desrochers, 2010; Seetharaman et al., 2015; Watson & Craft, 2004). These findings fit well with evidence that decreases in glucose utilization (Gage, Kelly, & Björklund, 1984) and oxidative metabolism (Villarreal, Gonzalez-Lima, Berndt, & Barea-Rodriguez, 2002) in the hippocampus are related to cognitive impairments in aged rats. Thus, deficient circulating and brain glucose availability may contribute importantly to age-related deficits in learning and memory. If so, during tasks that show age-related cognitive sparing, glucose availability in brain regions supporting these tasks should be similarly spared in old rats.

Experiments presented here used three different sets of tasks developed to assess the contributions of the hippocampus and striatum to learning and memory. Young adult and aged male rats were tested on a pair of object recognition tasks, a dual-solution navigational task, and a pair of single-solution navigational tasks, all of which dissociate the hippocampal or striatal memory system required for optimal task performance on the bases of lesions, pharmacological and hormonal manipulations, and neurochemical responses to training (e.g.: Colombo, 2004; Gold et al., 2013; Hawley, Witty, Daniel, & Dohanich, 2015; Kathirvelu & Colombo, 2013; Korol, 2004; Korol, Gardner, Tunur, & Gold, 2019; Korol & Pisani, 2015; Kundu et al., 2018; McDonald & White, 1993; Packard & Goodman, 2012; White et al., 2013; Wingard, Goodman, Leong, & Packard, 2015). Given the tight relationship between glucose and age-related changes in learning and memory, we also measured extracellular glucose levels in the hippocampus and striatum of young and aged rats while they learned hippocampus-sensitive place and striatum-sensitive response tasks.

2. Materials and methods

2.1. Animals

Young (3-month-old) or old (24–28-month-old) male Fischer-344 rats were obtained from Charles River Laboratories via the National Institute on Aging. Rats were singly housed on a 12:12 h light:dark cycle with ad libitum access to food and water except when food-restricted prior to maze tasks described below. Old rats underwent daily checks to monitor health across the entirety of the experiments. Young rats were similarly handled daily during the week prior to training. All methods were carried out in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals and were approved by the University of Illinois and the Syracuse University Institutional Animal Care and Use Committees. The animal facilities at both institutions are accredited by AAALAC.

The study included three experiments using tasks to dissociate hippocampal and striatal functions in learning and memory and a fourth experiment to measure extracellular glucose responses to training on single-solution place and response mazes. Training was conducted 4–8 h after the start of the light cycle.

2.2. Object recognition tasks

Object recognition tasks rely on a rat's propensity to explore novelty. Based on past work (Goodrich-Hunsaker, Hunsaker, & Kesner, 2008; Kesner, Bolland, & Dakis, 1993; Oliveira, Hawk, Abel, & Havekes, 2010), we recently developed two versions of object recognition tasks that are selectively sensitive to hippocampal- and striatal-dysfunction (Korol et al., 2019). Compared to more common paradigms that alter one object during the recognition test, our tasks involve changes to both objects during the test phase. Thus, in place of measuring exploration differences between an old and new object or location, both versions of tasks used in the current studies measure combined exploration of the two altered objects. One version is a double object location (dOL) task in which detection of changes in the relative positions of the two objects is sensitive to intrahippocampal but not intrastriatal infusions of lidocaine or α-cyano-4-hydroxycinnamate (4-CIN), drugs that block neural activity and neuronal uptake of lactate (Erlichman et al., 2008), respectively. The second version is a double object replacement (dOR) task based on detection of changes in the two objects themselves, i.e. introduction of two novel objects on the test trial; this task is sensitive to intrastriatal but not intrahippocampal infusions of lidocaine or 4-CIN (Korol et al., 2019). The dOL task requires memory of relative object locations and fits well with reports showing a role for the hippocampus in object-location associations (Barker & Warburton, 2011). Although Korol et al. (2019) showed the dOR task is sensitive to manipulations of the striatum and not hippocampus, the particular attribute of the task that involves striatal control is unknown. The role of the dorsal striatum in object recognition suggests that the striatum may link object novelty with approach behaviors important for subsequent learning of stimulus-response associations. Consistent with this view, fMRI data in humans reveals striatal activation with single object change related to landmark navigation as compared to hippocampal engagement during changes in spatial context (Doeller, King, & Burgess, 2008). Regardless of the specific cognitive attributes of novel object recognition related to the dorsal striatum, the structure by task double-dissociation of object and location recognition allows tests of memory system aging without possible confounds of altered energy substrates introduced by food deprivation required by many land maze tasks.

The training protocol and the tasks used for object recognition training and testing are illustrated in Fig. 1A, modified from Korol et al. (2019) to allow detection of both improvements and impairments in recognition. Young (N = 5 dOL, N = 5 dOR) and old (N = 5 dOL, N = 5 dOR) rats were placed in a square black, Plexiglas^® arena (70 × 70 cm floor, 50 cm walls) for exploration of two objects. The objects were small toys that were secured to the floor of the arena with a magnet. Rats received a series of three exploration trials in which they could examine the same two objects for 5 min on each trial, with 5 min between trials. These trials served as study sessions (S1, S2, S3) during which rats showed habituation to the objects, i.e. decreased time spent exploring the two objects across trials. Rats were tested for recognition memory 5 min after the last habituation trial. On the test trial (T) for the dOL task, rats were placed in the arena for 5 min and allowed to explore the original objects that had both been moved to new locations relative to each other and to the walls of the arena. On T for the dOR task, rats were placed in the arena for 5 min and allowed to explore two new objects placed in the same locations as the original objects. Rat behavior on all trials was recorded with a camera above the arena. Recognition memory was measured as the difference in exploration time on T minus S3, with increases in exploration times indicating novelty recognition.

Fig. 1.

Illustrations of the three sets of behavioral tasks. A) Object recognition tasks. Rats received three, 5-min study sessions (S1, S2, S3), during which the rats were allowed to explore the arena, with intertrial intervals of 5 min. On the test trial (T) given 5 min after S3, either the location of both objects was changed (dOL) or both objects were replaced (dOR). The recognition index was the time spent exploring both objects on the test trial minus the time spent exploring both objects on S3. B) Dual-solution task. Rats were trained to find food in the right (or left) arm of the maze, starting in the south arm on all training trials, until attaining 8/9 correct choices on consecutive trials. The rats were then given three probe trials from the north start arm. If the rat returned to the position in the room on the probe trial, the choice was classified as a place strategy to solve the maze. If the rat maintained a turn in the same direction as that of training, i.e. turning away from the room position, the choice was classified as a response strategy to solve the maze. C) Place and response training. These single-solution tasks are optimally solved by using either place cues to find the food (upper panel) or a body response (right/left turn) to find the food (lower panel).

Exploratory behavior was scored by hand using the videos together with custom software (ClickCounter; compliments of G. Dohanich, Tulane University). Object exploration was defined as whisking or sniffing at or any contact by the mouth, nose, or paw with the object. Because there were two objects changed or moved to new locations on the test trials, object exploration time refers to the combined time spent exploring both objects. Arena exploration, defined as general exploration that did not include exploration of the objects, grooming, or sitting still, was also measured. Total exploration was the sum of the arena and object exploration times. Each experimental group had N = 5 at the outset. However, two young rats in the dOL task stopped exploring the objects on S2, S3, and T (< 1 sec) and had extremely low arena exploration on S3 and T (< 20 sec). These rats were therefore eliminated from further data analyses.

The dependent measure used to calculate object recognition was the recognition index, calculated for each rat as the difference in time spent exploring the objects during the test, T, and the objects on the last study trial, S3. In pilot experiments, we saw continued habituation to familiar objects or object locations on a fourth study trial without introduction of novelty (i.e. an S4 trial; D.L. Korol, unpublished data); thus, our recognition index (object exploration on T – S3) likely underestimates the actual magnitude of recognition. Nonetheless, comparisons across age groups and interactions across age and task were robust with the T-S3 dependent variable. Planned comparisons of age effects within each task were performed using two-tailed t-tests. To confirm dissociable effects across tasks by age, a 2 (young, old) × 2 (dOL, dOR) factorial ANOVA examined the interaction of age by task. All analyses were run with alpha = 0.05.

2.3. Dual-solution task

The dual-solution task can be solved using either place or response strategies (Restle, 1957; Tolman, Ritchie, & Kalish, 1946). Rats were trained within a single session to find food in one of the two goal arms (i.e. east/right or west/left) of a T-maze (Fig. 1B). During training, correct choices were made by learning to turn in the proper direction, using an egocentric strategy, or by learning to turn to the arm in a particular room location, using an allocentric strategy. Which strategy individual rats used to solve the task was identified on probe trials administered after rats had learned to find the reward to a preset criterion. Converging data from pharmacological manipulations as well as neurochemical and molecular measures show that the hippocampus and striatum participate, respectively, in place and response solutions to this task (Canal, Stutz, & Gold, 2005; Chang & Gold, 2003a, 2003b; Gardner et al., 2016; McIntyre, Marriott, & Gold, 2003; Packard & McGaugh, 1996; Packard and Goodman, 2012, 2013; Packard, 1999). The present experiment determined whether the use of place and response solutions to the dual solution task shifts with age.

Rats were placed on a food-restriction regimen to lower their body weights to and then maintain their weights at 80–82% of individual baseline weights typical of regimens used by us and others (e.g., Newman, Scavuzzo, Gold, & Korol, 2017; Peters & Smith, 2020; Pisani, Neese, Katzenellenbogen, Schantz, & Korol, 2016). To familiarize rats with the reward used during training, they received Frosted Cheerios^® in their home cages for two days prior to training.

The maze apparatus, constructed of black Plexiglas^®, had 4 arms and was configured into a T-shape by blocking one arm. Each arm of the maze was 45 cm long, 14 cm wide, and 7.5 cm tall, with a food cup at the end; the center area was 14 × 14 cm. The training room contained an array of 2- and 3-dimensional extra-maze cues.

Prior to training, rats were placed in a clean holding cage and acclimated to the testing room for 15 min. On each trial, rats were placed in the start arm (the stem of the T) and were trained to find food in one goal arm (east/right or west/left) of the maze, with goal arms counterbalanced across rats within groups. Upon leaving the start arm, if the rat entered the unrewarded arm, it was allowed to remain there for up to 10 sec before being removed from the maze. If the rat made a correct choice, it was allowed to consume the reward fully. If the rat did not make a choice within two minutes, the experimenter removed the rat from the maze until the next trial. Training trials were conducted with an intertrial interval of 40 sec during which rats were placed in a holding cage in the training room. Arm entries were defined as four-paw entries into an arm.

To minimize the use of intra-maze and olfactory cues to find the reward, the food cups on the non-rewarded arms also had pieces of Frosted Cheerios^® placed beneath a screen, rendering the reward inaccessible, and the maze was rotated 90° during the intertrial interval.

After reaching a criterion of 8/9 correct on consecutive trials, each rat received three probe trials to identify the dominant learning strategy (place or response). On the probe trials, the start arm was rotated 180° from the position used during training and a food reward was available on both goal arms. With this procedure, when rats leave the start arm to approach the goal from the new location, a turn in the same direction as that used during training reflects the use of a response strategy. A turn toward the place in the room in which the goal arm was found during training reflects the use of a spatial strategy. The strategy assigned for each rat was the strategy, place or response, used by each rat on the majority of its probe trials; 21% of the rats used the same strategy on all three probes.

Rats were placed into four groups based on age and strategy selection: young place (N = 14), young response (N = 5), old place (N = 7), and old response (N = 13) groups. Age-related differences (young, old) in strategy selection (place, response) were analyzed with a Chi-Square test. Within-age tests of strategy bias were also conducted with a Chi-Square test. A 2 × 2 factorial ANOVA was used to compare number of trials to reach criterion across probe trial strategy (place, response) and across age (young, old). All analyses were run with alpha = 0.05.

2.4. Single-solution place and response tasks

Place and response versions of the 4-arm plus-shaped maze were used to compare learning in young and old rats (Fig. 1C). In contrast to the dual-solution task, these versions assess place and response learning in two tasks for which the alternate strategy does not provide the rat with an effective solution for finding food (Korol & Kolo, 2002). Learning on the place version of the maze is modulated by manipulations of the hippocampus, while learning on the response version is affected by manipulations of the striatum (Gold et al., 2013; Korol & Pisani, 2015; Zurkovsky, Brown, Boyd, Fell, & Korol, 2007).

Separate groups of rats were trained to find a food reward (~1/2 Frosted Cheerios^® piece) on either the place or response version of the 4-arm maze, as shown in Fig. 1C. Food restriction was as described for the dual-solution task. General training procedures were also similar except that during training, rats were placed in either the north or south arm on different trials and were explicitly trained to go to the correct arm in a specific room location, i.e. place learning, or to go to the correct arm by turning in a specific direction, i.e. response learning. The start arm was assigned quasi-randomly across north and south arms of the maze and counterbalanced across blocks of 15 training trials. The goal arm (east vs. west; place training) or the turning direction (left vs. right; response training) was counterbalanced across rats within age groups. Place training was conducted in a room with a rich array of 2-and 3-dimensional extra-maze cues; use of extramaze visual cues was minimized during response training by surrounding the maze with a beige wall curtain. Rats were trained in a single session for 75 trials with a 30-sec intertrial interval.

Young and old rats were randomly assigned to either place or response training to create the following groups: young place-trained (N = 6), young response-trained (N = 5), old place-trained (N = 7), and old response-trained (N = 4). Percent correct in blocks of 15 trials was calculated for each rat and group means for each block were used to generate learning curves across training. Two-way ANOVAs with repeated measures across blocks were used to compare age differences in learning rates for each task.

2.5. Glucose responses to training on single solution place and response tasks

Other rats were prepared with unilateral guide cannulae for later biosensor placements to measure training-related changes in extracellular glucose in the dorsal hippocampus and dorsolateral striatum. These regions were selected based on prior reports of their contributions to place and response learning, respectively (Packard & McGaugh, 1996; Packard, 1999), and of modulation of strategy selection following direct dorsal hippocampal and striatal glucose infusions (Canal et al., 2005). Notably, the dorsolateral striatum targeted here is reported to mediate habit or response learning, while the dorsomedial extent is involved in a goal-directed network supporting flexible action-selection and reversal learning (Palencia and Ragozzino, 2004, 2005; Ragozzino, Ragozzino, Mizumori, & Kesner, 2002; Regier, Amemiya, & Redish, 2015; Yin & Knowlton, 2004; Yin, Knowlton, & Balleine, 2004).

Young and old rats with hippocampal and striatal implants were randomly assigned to training conditions to create the following groups: hippocampus probes, place, young (N = 8) and old (N = 6); hippocampus probes, response, young (N = 5) and old (N = 4); striatum probes, place, young (N = 6) and old (N = 5); striatum probes, response, young (N = 5) and old (N = 6). For rats receiving food only without maze training, the groups were: hippocampus probes, young (N = 6) and old (N = 5); striatum probes, young (N = 7) and old (N = 7).

At least one week prior to food restriction, each rat was anesthetized with isoflurane and placed in a stereotaxic apparatus. A guide cannula (Pinnacle Technology, Lawrence, KS) was positioned above either the dorsal hippocampus (coordinates: 3.8 mm posterior and 2.5 mm lateral to bregma, 0.5 mm ventral to dura) or the dorsolateral striatum (coordinates: 0.2 mm anterior and 2.0 mm lateral to bregma, 1.5 mm ventral to dura). Striatal cannula guides were angled 15° lateral from the median plane to allow correct placement and still position the headstage over the midline. The guide cannula and a headstage that served as the housing unit for a wireless potentiostat was then affixed to the skull using dental cement and skull screws. Rats were allowed at least one week to recover before food restriction was initiated. Separate groups of rats were trained on place and response tasks during bioprobe recordings for glucose. One day prior to training, the biosensor (Pinnacle Technology, Inc., Lawrence, KS) was placed into the respective brain region extending ~ 3.2 mm past the guide cannula; the biosensor was left in place until the end of training.

The biosensor was used to measure changes in brain glucose levels using methods similar to those we have used before to measure brain extracellular lactate concentrations (Newman et al., 2011, 2017). The glucose-sensing portion of the probe is 1 mm long and sits above a 200-μm epoxy tip. The size of the active zone of the probe, as well as contributions of diffusion from nearby areas, precludes specification of the particular subregions of the hippocampus or striatum from which measures were taken.

The sensor is a platinum iridium electrode coated with glucose oxidase that reacts with glucose to produce gluconic acid and hydrogen peroxide. The hydrogen peroxide is oxidized at the electrode, generating a current proportional to the concentration of glucose. A wireless potentiostat in the headstage detects changes in current and transmits the information to a computer for signal recording every second throughout training using Sirenia software (Pinnacle Technology); the 1-sec values were averaged to produce 10-sec blocks for data presentation.

To ensure the probe’s functionality and specificity to glucose, each glucose biosensor was calibrated in vitro before and after use, i.e. prior to initial insertion of the bioprobe and immediately after removal of the probe after training. During calibration, each probe was placed in 10 ml of phosphate-buffered saline. Known amounts of glucose were added to the saline and changes in current were recorded. Data obtained with probes that did not show a strong linear relationship (r²s > 0.95) between glucose concentration (range: 50 μM to 2744 μM) and current were excluded. Data were also excluded if probes showed enhanced current from the addition of the interferent ascorbate, which would indicate a break in the ascorbate oxidase coating that protects the probe. The line generated by the post-test calibration standard curve was used to convert current generated by the probe to glucose concentrations.

On the day of training, rats were placed in the testing room in a clean holding cage and allowed to acclimate to the testing environment before maze training. At this time, glucose bioprobe recordings began and continued uninterrupted throughout the entire session. Baseline glucose measures were collected for ~75 min, with at least 15 min of stable recordings prior to the initiation of place or response training. To distinguish increases in hippocampal and striatal ECF glucose levels in response to maze training from those in response to ingestion of food, additional groups of young and old rats that were not maze-trained received Frosted Cheerios^® rewards in a holding cage at rates equivalent to those of their maze-trained counterparts. These groups received the following number of rewards in total across the testing session (mean ± s.e.m.): young, hippocampus 58 ± 2.0, old, hippocampus, 56.5 ± 2.5, young, striatum, 54.3 ± 2.5, old, striatum, 55.9 ± 2.0.

Immediately following training or feeding, rats received an overdose of sodium pentobarbital. Trunk blood was collected and centrifuged (1057g for 15 min) to isolate the serum fraction, which was subsequently measured for glucose concentration using commercial glucometers (Accu-Chek, Roche, Indianapolis, IN). Brains were removed and the bioprobe hemisphere was fixed for at least 2 days in 10% formalin. Brains were subsequently transferred to a 20% glycerol solution in phosphate buffered saline, sectioned coronally (40 μm), and mounted on gelatin-coated slides. The sections were stained with cresyl violet and later visualized for bioprobe placements.

Learning curves in the maze tasks were calculated as described above. Rats with hippocampal biosensors and those with striatal biosensors showed equivalent learning rates across training on both tasks (p's > 0.10) and the behavioral data were therefore pooled across biosensor placements. Two-way ANOVAs with repeated measures across blocks were used to compare age differences in learning rates for each task.

For each rat, baseline glucose values were calculated by taking the average of the measurements during the 5 min just prior to the start of maze training. After the start of training, testing-induced changes in ECF glucose were calculated for each rat by taking the difference between each 10-sec glucose value and the mean of the 5-min baseline. These difference scores were averaged across rats within groups and plotted against time to create glucose responses across the complete testing session.

To evaluate the full glucose response to training, the area under the curve (AUC) for changes in glucose levels from baseline during the first 60 min of maze-testing, the minimum time to complete testing, was calculated for each rat and used as the dependent variable for group means. Evaluation of the interaction of age (young vs old), task (place vs response), and brain region (hippocampus vs striatum) in the glucose response for AUC was made with a three-way ANOVA. A priori planned comparisons within each brain region were made across age groups using two-tailed t-tests with Bonferroni correction. To evaluate further age-related differences in glucose responses across the training sessions, we parsed the glucose values into six 10-min blocks. The values were analyzed using a mixed-measures ANOVA for each task and brain region, with age as a between subject factor and session block (six 10 min intervals) as a within subject factor; Greenhouse-Geisser correction was applied to violations of sphericity.

Differences in glucose responses between maze-trained rats and feeding controls were analyzed with four separate one-way ANOVAs for each brain structure and age: (1) hippocampus, young; (2) hippocampus, old; (3) striatum, young; (4) striatum, old. Planned paired comparisons were conducted using the Dunnett’s Multiple Comparisons test with food only as the control group. For each brain region, two-tailed t-tests were used to determine significant age differences in glucose responses to feeding. Paired t-tests were used to determine significant training-related or feeding-related increases in glucose levels (AUC) for each experimental group. Serum glucose was evaluated by age and task with a two-way ANOVA. Alpha = 0.05 for all statistical analyses.

3. Results

3.1. Object recognition tasks

Object recognition memory scores revealed a double dissociation by age and task, with performance of old rats lower than that of young rats on the hippocampus-sensitive dOL task but above that of young rats on the striatum-sensitive dOR task (Fig. 2). The double dissociation was confirmed by a 2 × 2 factorial ANOVA on the recognition index, which produced a significant interaction of age and task (F_1,14 = 10.83; p < 0.01), with no main effects of either variable (p’s > 0.2).

Fig. 2.

Object recognition in young and old rats. Upper graphs show total object exploration across study sessions S1–3, followed by exploration on the test trial T in (A) dOL and (B) dOR tasks. Note that object exploration for both ages and tasks declines across S1-S3, suggesting habituation to the objects. Lower graphs show the recognition index (sec) for (C) dOL and (D) dOR. Note that young rats performed better than did old rats on the dOL task but old rats performed better than did young rats on the dOR task. * = p < 0.05.

Object exploration times during the three study trials were comparable across young and old rats, decreasing from means of 14–25 sec on the first study session to means < 5 sec on the third session (Fig. 2A, B). During T, object exploration times increased in young rats on the dOL task and in old rats on the dOR task, whereas they did not increase in the old rats on dOL or the young rats on dOR tasks. Specifically, when tested on the dOL task, young rats showed an increase whereas old rats showed a decrease in the recognition index, resulting in a significantly higher recognition score for young vs. old rats (t₆ = 2.53; p < 0.05; Fig. 2C).

In marked contrast, an opposite pattern of results was seen on the dOR task. When tested with two novel objects on the test trial, aged rats demonstrated a 5-fold increase in exploration of the objects in comparison to a modest decrease in young rats. Thus, on the dOR task, old rats had a recognition index significantly higher than that of young rats (t₈ = 2.51; p < 0.05; Fig. 2D). Total arena exploration times on T, collapsed across tasks, were 76.1 ± 23.3 and 84.5 ± 16.3 sec (mean ± s.e.m.) for young and old rats, respectively; these values did not differ by age (data not shown; t₁₆ = 0.32; p > 0.7).

3.2. Dual solution task

A comparison of the proportion of rats using place versus response solutions in the dual-solution task revealed a significant difference in the strategy displayed on probe trials by young and aged rats. As shown in Fig. 3A, most young rats (74%) exhibited place solutions while most aged rats (65%) exhibited response solutions with a significant shift in strategy across age group (young vs. old, X² = 4.41; p < 0.05). The strategy preference in young rats reflected a significant strategy bias towards place learning (X² = 4.263, p < 0.05). The strategy preference in aged rats towards response learning was not significant (X² = 1.800, p > 0.1). Trials to reach criterion were similar across all groups; rats readily learned to solve the task, regardless of age and strategy used, with mean ± s.e.m. trials to reach criterion ranging from 17.7 ± 1.4 to 19.4 ± 3.4 (Fig. 3B). There were no significant effects of age (F_1,35 = 0.71, p > 0.4), strategy expressed on the probe test (F_1,35 = 0.001, p > 0.9), or their interaction (F_1,35 = 0.02, p > 0.8) for trials to criterion. Thus, aged rats solved the dual-solution task as quickly as did young rats but most often used a different strategy to do so.

Fig. 3.

Dual solution task performance in young and old rats. (A) Percent of rats choosing place or response strategies on probe trials. Note the significant age-related shift from place to response strategies with age. (B) Trials to reach the learning criterion (8/9 correct); these did not differ by age or by strategy selection. * = p < 0.05 across ages; ^ = p < 0.05 within age using Chi-Square analyses.

3.3. Single-solution place and response tasks

The acquisition curves for the place task revealed that young rats learned the hippocampus-sensitive task more quickly than did old rats, while learning rates for the striatum-sensitive response task were comparable across ages (Fig. 4). For place learning, young rats made substantially more correct responses throughout training except for the first block when both age groups were at chance. Interestingly, old rats failed to achieve more than 70% correct on average on the place task though they reached 90% on the response task.

Fig. 4.

Learning curves for (A) place and (B) response tasks in young and old rats. (A) Young rats learned the place task significantly faster than did old rats. (B) Young and old rats learned the response task at comparable rates. * = p < 0.05, main effect of age and age X trial block interaction.

A mixed measures ANOVA comparing learning on the place task revealed a main effect of age (F_1,11 = 14.41, p < 0.05) and a significant interaction between trial block and age (F_4,44 = 2.97, p < 0.05). An ANOVA examining learning on the response task showed no significant effect of age (F_1,7 = 0.35, p > 0.5) or interaction between age and trial block (F_4,28 = 0.14p > 0.9). Thus, aged rats were impaired at learning the place task compared to young rats, but learned the response task as quickly as did young rats. This conclusion was supported by repeated measures one-way ANOVAs on learning curves for the four groups. These analyses showed significant learning for the young rats on both the place (F_4,20 = 18.4, p < 0.001) and response (F_4,14 = 4.6, p < 0.02) tasks. The aged rats showed significant learning on the response task (F_4,12 = 6.1, p < 0.01) but not on the place task (F_4,24 = 1 0.4, p > 0.2).

3.4. ECF glucose responses to training

The behavioral results obtained in rats with implants for brain glucose measurements (Fig. 5) were consistent with those above using unimplanted rats (Fig. 4), providing a replication of age-related shifts in the learning of place and response mazes. As in rats that had not undergone surgery, an ANOVA comparing the learning rates of young and old rats on the place task revealed a significant interaction of trial block and age (F_4,92 = 2.54, p < 0.05) and significant main effect of age (F_1,23 = 5.73, p < 0.05). An ANOVA comparing the learning rates of young and old rats on the response task did not show a significant interaction of trial block and age (F_4,72 = 0.58, p > 0.6) or main effect of age (F_1,18 = 0.04, p > 0.8). In cannulated rats, all groups showed significant learning (young-place: F_4,52 = 22.3, p < 0.001; old-place: F_4,40 = 8.5, p < 0.001; young-response: F_3,36 = 15.9, p < 0.001; old-response: F_4,36 = 9.1, p < 0.001).

Fig. 5.

Learning curves for (A) place and (B) response tasks in young and old rats after cannula surgeries. As in unimplanted rats (Fig. 4), young rats learned the place task faster than did old rats, but learning was comparable for both ages on the response task. * = p < 0.05, main effect of age and age X trial block interaction.

During training on the place and response single-solution tasks, ECF glucose levels increased substantially above baseline values in both the hippocampus (Fig. 6) and striatum (Fig. 7) for both ages and tasks (Ns = 4–8; all p's < 0.02). The training-related responses in hippocampus and striatum were qualitatively and quantitatively quite different for young and old rats. The three-way ANOVA (age × task × brain area) on the increase in ECF glucose levels produced a significant three-way interaction (F_1,37 = 4.51, p < 0.05; see Table 1 for full statistical results including all main effects and interactions), suggesting specificity in how ECF glucose responses change across these factors.

Fig. 6.

Changes in extracellular glucose concentration (μM) from baseline in the hippocampus during place and response learning. (A) and (B) show glucose changes in 10-sec bins across 60 min of training on (A) place and (B) response maze tasks. Note that hippocampal glucose levels increased dramatically in both ages within 30 min of training on both tasks. Hippocampal glucose responses to place training were significantly lower in old relative to young rats: * = p < 0.05. (C) Areas under the curve for increases in ECF glucose levels during training. Note that hippocampal glucose increases were greater in young compared to old rats during place training but not during response training or feeding. Compared to fed controls, young rats trained on the place task had signficantly higher glucose responses to maze testing, with a similar trend in old response-trained rats. ^ = p < 0.05 vs food. # < 0.06 vs food.

Fig. 7.

Changes in extracellular glucose concentration (μM) from baseline in the striatum during place and response learning. (A) and (B) show glucose changes in 10-sec bins across 60 min of training on (A) place and (B) response maze tasks. Note that striatal glucose levels increased dramatically in both ages within 30 min of training on both tasks. C) Areas under the curve for increases in ECF glucose levels during training. Note that there were no age- or task-related differences in glucose responses to maze training. However, the increase in glucose during feeding was significantly higher in old vs young rats. Moreover, in young, but not old rats, the increase in striatal glucose during response training was substantially higher than the rise during feeding. * = p = 0.05 vs young; ^ = p < 0.05 vs food.

Table 1.

Statistics summary for ECF glucose changes.

		Glucose (0–60 min)
	Degrees of Freedom	F Value	P Value
Age	1	1.242	0.272
Task	1	0.782	0.382
Region	1	0.566	0.457
Age * Task	1	0.799	0.377
Age * Region	1	0.000	0.995
Task * Region	1	0.199	0.658
Age * Task * Region	1	4.506	0.041
Error	37
Total	45

Increases in ECF glucose concentrations in the hippocampus during place learning were significantly lower in aged rats than were the increases seen in young rats (t₁₂ = 2.59, p < 0.05; Fig. 6A, C). In contrast, increases in glucose in the hippocampus during response training were comparable across ages (t₇ = 0.99, p > 0.7; Fig. 6B, C). There was also a significant interaction between age and task on hippocampal glucose responses to training (F_1,19 = 5.73, p < 0.05). As in maze-trained rats, hippocampal ECF glucose responses to feeding in young (N = 6) and old (N = 5) rats were significantly higher than baseline values (p’s < 0.005), but increases to feeding were not different between young and old rats (t₉ = 0.11, p > 0.9). In young rats, the training-related increases in hippocampal glucose during the place task were significantly higher than increases recorded as rats consumed food without maze training (p < 0.05; Fig. 6C). This was not the case in old rats, in which increases in hippocampal glucose levels after feeding alone were comparable to those in rats trained on the place task (p > 0.9; Fig. 6C). Moreover, the increases in hippocampal glucose during response training appeared larger than those in food controls in old rats, an effect that approached significance (p < 0.06 after correcting for multiple comparions), but not in young rats (p > 0.6; Fig. 6C).

The profiles of glucose responses in the striatum (Fig. 7A, B) were quite different from those described above for the hippocampus. When rats were trained on either the place or response task, the magnitude of increases in striatal glucose levels during training did not differ significantly by age (place: t₉ = 0.05, p > 0.9; response: t₉ = 1.13, p > 0.4; Fig. 7C), and there was not a significant interaction across age and task (F_{1, 18} = 0.62, p > 0.4).

As in trained rats, ECF glucose in striatum of young and old rats rose significantly above baseline during feeding alone (Ns = 7; p’s < 0.01). However, in fed untrained controls, old rats had a remarkably larger increase in striatal ECF glucose levels than did young rats (t₁₂ = 2.17, p = 0.05; Fig. 7C). It is important to note that when compared with the respective food controls, there was a significant training-induced increase in striatal glucose in young rats during response training (p < 0.05); the same was not true for old rats trained on the response task (p > 0.7) or for young and old rats trained on the place task (young, p > 0.2; old, p > 0.7); Fig. 7C). Thus, striatal glucose went substantially beyond those of fed controls only in young rats during response training.

To test further the possibility that age-related differences in glucose responses varied within testing sessions, we ran mixed-measures ANOVAs with six 10 min testing blocks set as a within-subject factor and age as a between subject factor. In line with prior analyses, we found a significant main effect of age (F_{1, 12} = 7.65, p < 0.05) and interaction between age and testing block (F_{2.3, 27.5} = 7.65, p < 0.01) on hippocampal glucose responses to place training, with no significant effect for age during response training (F_{1, 7} = 0.97, p > 0.3); although an interaction effect between age and testing block approached significance (F_2,14.1 = 3.42, p = 0.061) during response training, there was relatively large overlap in glucose responses among age groups across the session. In the striatum, no significant effects were found in either task for age (Place: F_1,9 = 0.002, p > 0.9; Reponse: F_1,9 = 1.28, p > 0.2) or interactions of age and testing block (Place: F_1.4,12.3 = 0.52, p > 0.5; Reponse: F_1.8,15.8 = 2.03, p > 0.1).

Note that the number of rewards consumed by food-only controls were comparable across age groups and, on average, largely matched those consumed in maze-trained counterparts. However, given their relatively low rate of choice accuracy, old place-trained rats also consumed fewer rewards overall (mean ± s.e.m.: 45.4 ± 2.7) than did old food control rats or young place-trained rats (56.2 ± 1.6 and 54.7 ± 1.8, respectively). However, the number of rewards consumed by food controls were not correlated with glucose responses across individual rats within each brain region (data not shown; Hippocampus: r² = 0.04, p > 0.5; Striatum: r² = 0.12, p > 0.2; Overall: r² = 0.05, p > 0.2). Moreover, hippocampal glucose responses to place training distinguished between young and old rats very early in training (see Fig. 6A), a time point when choice accuracy and therefore the number of food rewards consumed was comparable between ages (Fig. 5A, trial block 1). Thus, it is likely that any variability in food consumption across groups had negligible effects on glucose signals and does not readily explain our findings.

In blood samples taken at the end of training or feeding, there were no substantial differences in serum glucose levels across training conditions, feeding controls, or age (Fig. 8). ANOVAs revealed no main effects of age (F_1,52 = 2.05, p > 0.1), task (F_2,52 = 2.24, p > 0.1), or age × task interaction (F_2,52 = 0.61, p > 0.5).

Fig. 8.

Serum glucose concentrations across conditions in young and old rats. Note that these levels were similar in the fed controls compared to place- and response- trained rats. There were no differences in serum glucose levels by age.

All biosensor placements were accurately positioned in the hippocampus or striatum. Histological examples of the placements are shown in Fig. 9.

Fig. 9.

Representative histology showing placements of A) hippocampal and B) striatal biosensors. Thin arrows point to the tract left by the bioprobe; wide arrows indicate damage from the guide cannula. Biosensors were correctly placed with minimal damage in all rats in this experiment.

4. Discussion

The present results support the view that aging is accompanied by impaired hippocampal contributions and maintained or enhanced striatal contributions to learning and memory. More generally, the results presented here show that aging brings differences, and not merely impairments, in learning based on shifts in the predominant strategies used for learning and remembering. These age-related differences in optimal learning strategies add to the list of other variables or manipulations that shift the use of strategies from hippocampal- to striatal-based learning, a list that includes stress and anxiety (Gold & Korol, 2012; Gruber & McDonald, 2012; Packard and Goodman, 2012, 2013; Sadowski, Jackson, Wieczorek, & Gold, 2009; Schwabe, 2013) and depleted estrogen status (Korol & Pisani, 2015).

Notably, these findings suggest that aging can be characterized by cognitive deficits or improvements depending on the contexts in which learning and memory are tested and specific attributes of the tasks. These bidirectional shifts may indeed reflect independent effects of age-related changes in hippocampal and striatal functions, altered levels of competition between brain areas (cf.: Gold et al., 2013; Korol & Pisani, 2015; Kosaki, Poulter, Austen, & McGregor, 2015; Lee, Duman, & Pittenger, 2008; Packard & Goodman, 2012, 2013; Poldrack & Packard, 2003; White et al., 2013) or both. Our recent finding that inactivation of the striatum blocks age-related deficits in place learning in old rats without altering learning in young rats provides support for increased competition of the striatum on the hippocampus during aging (Gardner, Gold, & Korol, 2020).

We observed age-related differences on both location recognition (dOL) and object recognition (dOR) tasks, shown previously in young adult males to be sensitive to pharmacological manipulations of the hippocampus and striatum, respectively (Korol et al., 2019). In particular, relative to memory in young rats, old rats showed impaired memory for locations and enhanced memory for objects. The age-related impairment we found on the dOL task matches prior reports showing object-location learning and memory impairments in aged rodents and humans (e.g., see Bortolatto, Wilhelm, Chagas, & Nogueira, 2012; Hernandez et al., 2015; Muffato, Hilton, Meneghetti, Beni, & Wiener, 2019; Wimmer, Hernandez, Blackwell, & Abel, 2012). Here, we also found opposite and robust age-related effects on the dOR task. Surprisingly, during the dOR test session, young rats did not display increased exploration of the novel objects compared to exploration of the familiar objects in the third habituation session. The absence of increased exploration may reflect the ongoing habituation seen in the absence of the novelty manipulation (D.L. Korol, unpublished observation). However, other reports have noted prominent novel object detection in three-month old Sprague Dawley rats (Korol et al., 2019) and in Lister-Hooded rats using similar object recognition procedures (McTighe, Cowell, Winters, Bussey, & Saksida, 2010). Further work will help clarify the experimental details (e.g., rat strain, duration of habituation, inter-session and retention intervals, etc.) that modify object detection abilities and underlying neural correlates. Nevertheless, the findings here show clear dissociable effects of age on dOL and dOR novelty recognition tasks that align with prior work and that are internally consistent across tasks reported here, showing decreased hippocampus-sensitive learning and increased striatum-sensitive learning in senescent rodents.

On the dual-solution T- maze task, learning rates were similar in young and old rats, but the predominant strategy expressed on the probe trial was different, shifting across age groups from a majority of place to a majority of response strategies. This finding is similar to those seen in other dual solution paradigms in rats (Nicolle et al., 2003) and humans (Bohbot et al., 2012) and also to results in a triple solution maze that allowed for use of intramaze cues, egocentric responses, or extramaze cues (Barnes et al., 1980) showing increased probability of response strategies and decreased probability of place strategies in old compared to young rats..

When rats were tested on mazes with only a single strategy optimal for solving the task, aging was accompanied by impairments in learning the place version of the maze but by maintained learning ability on the response version. It is important to note that the place and response mazes are the same in terms of motivation, reward, and locomotor needs; in essence, what differs is the cognitive rule and, from a multiple memory standpoint, the brain region necessary for optimal problem solving. That rats of both ages showed accuracy scores at chance at the start of training on place and response tasks suggests also that age differences reported here relate to cognitive style and not necessarily to nonmnemonic processes. Viewed across these single- and dual-solution tasks, the findings suggest that aging alters how a rat learns not necessarily how much a rat learns.

The variability within groups trained on the single-solution tasks was low, even for aged rats trained on the place version of the maze. In one study with relatively greater variability in spatial learning by aged rats (Pereira et al., 2015), some old rats performed at a level similar to that of young rats, while performance of others was impaired. Interestingly, the aged rats exhibiting spatial impairments showed enhanced response learning, fitting well with findings here. One interpretation of their findings is that the hippocampus is dysfunctional in the spatially impaired subset of old rats, yet equally likely is that increased engagement of the striatum during aging competes with the hippocampus, leading to the same outcome.

Our behavioral findings in the place and response tasks were paralleled by changes in ECF glucose levels in the hippocampus and striatum during training. The hippocampus in aged rats exhibited diminished training-related responses in ECF glucose levels during place training but not response training, while the striatal glucose response did not differ by age in either maze task, unless the contribution of food reward is considered. The striatal glucose responses to maze training in young rats, particularly during striatum-sensitive response learning, were larger than the responses seen in fed-controls, whereas the glucose responses to maze training in old rats failed to increase beyond the already potentiated change in glucose due to food ingestion. In interpreting these results, it is important to note that measures of extracellular neurochemical levels represent both recruitment and utilization of glucose. In tasks without food reward, depletion in hippocampal ECF glucose, particularly in old rats, corresponds to the cognitive load of the task perhaps because recuitment lags behind utilization (Korol, 2002; McNay et al., 2000). Thus, the absence of increased glucose responses to training over that in food controls in old but not young rats might reflect an age-related dysfunction in striatal activation, reduced glucose recruitment from blood flow, and/or enhanced transport into and utilization by brain cells during training. In addition, it is possible that in the aged striatum, a ceiling effect of glucose recruitment was established such that training-related needs for utilization were already met. From this perspective, preserved response learning in old rats supports the hypothesis that the relatively large striatal glucose increase (from food or otherwise) preserves glucose availability and is sufficient to support striatum-sensitive learning abilities. Likewise, relative to food controls, a rise in hippocampal ECF glucose in old rats during response training but not during place training (Fig. 6C) opens the possibility that the hippocampus in aged rats is preferentially activated or engaged during response learning. However, this effect could also result from age- and task-specific alterations in activity-dependent glucose transport into hippocampal cells; for example, in aged rats, place training may induce a relatively fast and substantial influx of glucose into cells resulting in reduced glucose availability compared to that in other conditions, such as during response learning.

It may seem surprising that ECF glucose increases were similarly evident in the hippocampus and striatum during both place and response training, suggesting activation of both systems no matter which was optimal for learning. However, similar increases in acetylcholine release (Pych, Chang, Colon-Rivera, Haag, & Gold, 2005) and in ECF levels of lactate (Newman et al., 2017) have been found in the hippocampus and striatum during both place and response learning. These data fit the notion that although a single system may be more relevant to the learning rule, both systems may incidentally process or acquire information about the task or environment. As one example, when young adult male rats are trained on the dual solution task, they initially express place strategies on test trials. However, with further training, the rats predominately express response strategies. If the striatum is inactivated at that later time, the rats do not lose the learned response overall but instead switch to a place strategy (Packard & McGaugh, 1996), showing that the memory system no longer controlling the expressed response still has the information to solve the task successfully.

The responses of glucose to training parallel well the results obtained in learning ability for place and response tasks in young and old rats, with age-related declines in hippocampus-sensitive place learning but preserved striatum-sensitive response learning. At one level, increases in glucose concentrations can be viewed as a marker of activation of the two brain areas, thereby suggesting that activation of the hippocampus when rats are trained on a hippocampus-sensitive task is reduced in aged rats compared to activation in young rats. Beyond serving as a marker of activation, however, considerable evidence suggests that glucose is itself a potent regulator of information processing important for learning and memory, with many demonstrations that glucose treatments reverse age-related cognitive impairments and enhance learning and memory for multiple tasks in senescent laboratory rodents and humans (cf.: Gold, 2001; Gold & Korol, 2014; Korol, 2002; Macpherson et al., 2015; Messier, 2004; Owen, Finnegan, Hu, Scholey, & Sünram-Lea, 2010; Seetharaman et al., 2015; Smith et al., 2011; van der Zwaluw, van de Rest, Kessels, & de Groot, 2015), including individuals with Alzheimer’s disease (Manning, Ragozzino, & Gold, 1993; Watson & Craft, 2004) or Down syndrome (Manning, Honn, Stone, Jane, & Gold, 1998).

The robust and pervasive influences of glucose on memory fit well with the present findings showing that impaired and preserved cognitive functions in the hippocampus and striatum are associated with impaired and preserved responses of extracellular glucose levels during learning. Training-related increases in blood glucose might contribute to age-related changes in memory in some circumstances, for example increases following release of adrenal epinephrine initiated by stressful or arousing events (Gold & Korol, 2014; Mabry, Gold, & McCarty, 1995; Morris, Chang, Mohler, & Gold, 2010); blood glucose in turn may regulate brain glucose availability, overriding depletion of brain glucose during behavioral testing and rescuing age-related memory impairments (McNay & Gold, 2001; McNay et al., 2000). However, such differences do not readily explain task-related specificity of the impairments and preservation of learning and memory for hippocampus-and striatum-sensitive tasks observed here. In the present experiments with appetitive maze learning, blood glucose levels increased comparably after ingestion of food with or without place or response training. Therefore, the maintained responses of ECF glucose in the striatum vs. attenuated responses in the hippocampus of old male rats during place training suggest that circulating glucose levels per se do not account for the age-related differences in learning and memory. Instead, glucose levels in each brain area appear to respond selectively to processes intrinsic to those brain areas, perhaps reflecting differential increases in blood flow, transport of glucose from blood to brain, or consumption of glucose in each brain area. These interpetations are complicated by the interplay of food reward effects on both blood and brain glucose levels. In this report, we chose to assess brain glucose responses to training using the place and response tasks because of the robust literature dissociating hippocampal and striatal contributions to the tasks. Future work with tasks that do not require a food reward, e.g., dOL and dOR novelty detection, will benefit tests of the role of glucose in learning and memory across the life span and memory systems.

The results presented here lead to questions of how glucose regulates cognitive activity. One possibility is that glucose provides energy directly to cells in brain areas engaged in learning and memory to promote mechanisms of neural plasticity. This explanation is consistent with evidence that stimulation of brain areas results in increased blood flow, glucose uptake into neurons and astrocytes, and glycolysis (Bak et al., 2009; Díaz-García & Yellen, 2019; Díaz-García et al., 2017; DiNuzzo, Mangia, Maraviglia, & Giove, 2010; Hung, Albeck, Tantama, & Yellen, 2011). This view is also consistent with findings of depressed responses of energy metabolism to cognitive activity in aged rodents and humans (e.g.: Gage et al., 1984; Ivanisevic et al., 2016; Mosconi et al., 2008; Villarreal et al., 2002). Prior work shows a link between an age-related decline in regional blood flow and the metabolic rate of glucose across the hippocampus and striatum (Bentourkia et al., 2000; Noda et al., 2002). Reductions in hippocampal blood flow and glucose metabolism are connected to age-related decline in spatial learning and memory (Gage et al., 1984; Heo et al., 2010), consistent with our findings. However, to our knowledge, these measures have not been evaluated in the striatum during striatum-sensitive tasks.

Another mechanism by which glucose may regulate learning and memory is by providing a substrate for lactate production. Like glucose (Morris & Gold, 2013; Ragozzino, Pal, Unick, Stefani, & Gold, 1998), infusions of lactate into the hippocampus also enhance memory (Newman et al., 2011; Suzuki et al., 2011). Lactate might mediate the effects of glucose on learning and memory in several ways. One view is that upon stimulation of a brain area, glucose enters astrocytes, where it can be stored as glycogen and later broken down to lactate (Chuquet, Quilichini, Nimchinsky, & Buzsáki, 2010; Fryer & Brown, 2015; Magistretti, Pellerin, Rothman, & Shulman, 1999; Pellerin, 2003). Lactate in turn may be released from astrocytes to support the metabolic needs of neurons (Gold & Korol, 2014; Gold, 2014; Magistretti et al., 1999; Pellerin & Magistretti, 2012; Pellerin, 2003; Steinman, Gao, & Alberini, 2016); lactate may also support the metabolic needs of astrocytes (Dienel & McKenna, 2014; Dienel, 2019; Sonnewald, 2014). Other possible mechanisms include lactate acting as a neuronal or vascular signal through its G-protein coupled receptor (Barros, 2013; Bergersen & Gjedde, 2012; Bozzo, Puyal, & Chatton, 2013; Scavuzzo, Rakotovao, & Dickson, 2020; Tang et al., 2014), perhaps coordinating region-specific regulation of neurovascular coupling to increase glucose availability in a brain area- by task-specific manner. In this way, lactate may serve as a mediator of a feed-forward process to promote increases in glucose availability upon demand (Gordon, Choi, Rungta, Ellis-Davies, & MacVicar, 2008; Lauritzen et al., 2014).

In summary, our behavioral results suggest that the functions of the hippocampus and striatum follow different trajectories during aging: the ability to use the hippocampus to solve tasks is impaired with age while the ability to use the striatum to solve tasks is spared or even enhanced with age. The behavioral results are paralleled by changes in extracellular glucose levels during training. Since glucose fully restores learning and memory in aged rats when injected peripherally or centrally, the development of treatments to augment availability of glucose, or to augment its downstream actions, may offer targets for development of treatments to enhance memory in aged humans. The results also reveal intact cognitive mechanisms that could be tapped by behavioral therapies, redirecting problem-solving strategies to those supported by spared functions to ameliorate age-related dysfunctions.