Re: Neural plasticity in the ageing brain - nature 리뷰

[문형철]

beyond reason

Neural plasticity in the ageing brain

Abstract 

The mechanisms involved in plasticity in the nervous system are thought to support cognition, and some of these processes are affected during normal ageing. Notably, cognitive functions that rely on the medial temporal lobe and prefrontal cortex, such as learning, memory and executive function, show considerable age-related decline.  It is therefore not surprising that several neural mechanisms in these brain areas also seem to be particularly vulnerable during the ageing process.

In this review, we discuss major advances in our understanding of age-related changes in the medial temporal lobe and prefrontal cortex and how these changes in functional plasticity contribute to behavioural impairments in the absence of significant pathology.

Ageing is associated with a decline in cognitive function that can, in part, be explained by changes in neural plasticity or cellular alterations that directly affect mechanisms of plasticity. Although several age-related neurological changes have been identified during normal ageing, these tend to be subtle compared with the alterations that are observed in age-associated disorders, such as Alzheimer’s disease and Parkinson’s disease. Moreover, understanding age-related changes in cognition sets a background against which it is possible to assess the effects of pathological disease states. In this review, we discuss functional alterations that occur during normal ageing in the medial temporal lobe and the prefrontal cortex (PFC) and how these ageassociated changes might contribute to the selective cognitive impairments that occur in advanced age. We first discuss data that suggest that profound loss of neurons does not significantly contribute to age-related cognitive impairments. We then review the subtle changes in neuronal morphology, cell–cell interactions and gene expression‎ that might contribute to alterations in plasticity in aged animals and how these changes disrupt the network dynamics of aged neuronal ensembles that ultimately contribute to selective behavioural impairments.

Morphology of the ageing brain

Age-related changes in the morphology of neurons are selective and it seems that there is no universal pattern across the entire brain. However, one finding that does seem to be consistent is that in most brain areas neuronal loss does not have a significant role in age-related cognitive decline. Rather, small, region-specific changes in dendritic branching and spine density are more characteristic of the effects of ageing on neuronal morphology (FIG. 1).

This is contrary to early investigations of aged nervous tissue in which profound neuron loss was reported to occur in advanced age. In 1955, Brody was the first to suggest that age-related reductions in brain weight were due, in part, to a decline in neuron number in all cortical layers1 . Subsequent investigations corroborated this work, reporting a 10–60% decline in cortical neuron density between late childhood and old age2 . In addition, profound cell loss was found in the hippocampus of ageing humans3 and the hippocampus and PFC of non-human primates4 . The data obtained from these early reports, however, were confounded by various technical and methodological issues, such as tissue processing and sampling design, that later called into question their accuracy5 . In the 1980s, when new stereological principles were developed, it became possible to identify and eliminate many of the confounding factors of the previous studies that had indicated a profound decline in neuron number occurring in advanced age6 . The resulting conclusion was that in humans7,8, non-human primates9–12 and rodents13–15, significant cell death in the hippocampus and neocortex is not characteristic of normal ageing. A notable exception to this idea, however, has recently been reported. In aged non-human primates, there is a ~30% reduction in neuron number in all layers in area 8A of the dorsolateral PFC, which significantly correlates with impaired performance on a working memory task. By contrast, area 46 of the PFC shows conservation of neuron number16. Similar to early reports of a decline in neuronal density with ageing, early investigations of dendritic branching suggested massive deterioration in the human entorhinal cortex and hippocampus17,18 (FIG. 1a).

These experiments, however, included both healthy individuals and people with dementia. Subsequent investigations, which were more precisely controlled for the participants’ mental status and applied stereological controls, found that normal aged individuals had extensive dendritic branching in layer II of the parahippocampal gyrus, the origin of the perforant pathway to the dentate gyrus19,20. Moreover, dendritic branching and length appeared to be greater in aged individuals than in younger adults or patients with senile dementia. Other investigations have reported increased dendritic extent in the dentate gyrus of old compared with middle-aged humans21,22. In other subregions of the human hippocampus, however, including areas CA1 (REF. 23) and CA3 (REF. 24), and the subiculum25, there is no change in dendritic branching with age. Studies of dendritic extent in other animals have, in general, confirmed that there is no regression of dendrites with age. In rats, there is no significant change in dendritic length of hippocampal granule cells between young (3 months), middle-aged (12–20 months) and aged (27–30 months) rats, with a trend towards an increase between middle-age (20 months) and old age (27 months)26. There is also no decrease in dendritic extent between young (3 months) and old rats (26 months) in area CA1 (REFS 27,28), although there is some evidence that a small subset of CA1 neurons from 24-month-old rats have increased basilar dendritic length and branching compared with 2-month-old rats28 (but see also REF. 29). The morphology of PFC neurons seems to be more vulnerable to the effects of ageing than that of hippocampal neurons. In rats, dendritic branching of pyramidal neurons decreases with age for both apical and basal dendrites in superficial cortical layers30. A reduction in dendritic branching with age has also been observed in anterior cingulate layer V of the rat31 and the human medial PFC32,33.

Similar to the investigations on dendritic branching during ageing, the data on spine density suggest that ageassociated alterations are also region-specific. Even in the hippocampus, changes in spine density are not consistent across subregions. In the dentate gyrus, there is no significant reduction in spine density in aged humans34 or rats35. There is also no reduction in spine density in area CA1 in aged compared with young rats29. In the subiculum of non-human primates, however, significant reductions in spine density with age have been observed in monkeys between the ages of 7 and 28 years36.

Biophysical properties of aged neurons

In all subregions of the hippocampus, most electrical properties remain constant over the lifespan37. These include resting membrane potential27,38–45; membrane time constant27,46,47; input resistance39–48 (but see also REF. 49); threshold to reach an action potential42,47; and the width and amplitude of Na+ action potentials27,40,42–45,47,50. Numerous studies, however, have shown an increase in Ca2+ conductance in aged neurons. CA1 pyramidal cells in the aged hippocampus have an increased density of L-type Ca2+ channels51 that might lead to disruptions in Ca2+ homeostasis52, contributing to the plasticity deficits that occur during ageing53,54. Moreover, Ca2+ activates outward K+ currents that are responsible for the afterhyperpolarizing potential (AHP) that follows a burst of action potentials41,43. Aged neurons in areas CA1 and CA3 have an increase in the amplitude of the AHP that results, at least in part, from age-related increases in Ca2+ conductance41,50. Other factors that might contribute to the larger AHP in aged animals include reduced basal cyclic AMP (cAMP) levels55. The larger AHP observed in aged hippocampal neurons suggests that aged CA1 pyramidal cells are less excitable, as they are further from action potential threshold than are young neurons during the AHP. The only evidence that supports this idea is the finding that, in an in vitro hippocampal slice preparation, aged CA1 neurons fire fewer action potentials than do young neurons in response to a prolonged depolarization50. This is not the case, however, when pyramidal neurons are recorded in vivo under normal physiological conditions. In awake, behaving rats, there is no difference in the firing rates of CA1 pyramidal neurons according to age56–64, and, in fact, the firing rates of CA3 pyramidal neurons are actually slightly higher in aged than young rats64. Similar to neurons in the hippocampus, many electrophysiological properties of neurons in the PFC remain the same during normal ageing, including resting membrane potential; membrane time constant; threshold to elicit an action potential; and rise time and duration of an action potential65. There is some evidence of a small increase in the input resistance in PFC neurons of aged monkeys as well as a decrease in the amplitude and fall time of action potentials65. However, cognitive performance is not related to action potential amplitude, action potential fall time or input resistance65. Neurons in the PFC of aged monkeys also have a significantly larger AHP compared with young neurons65, which suggests that Ca2+ homeostasis might also be disrupted in PFC neurons in advanced age. Changes in cell–cell interactions Aged animals have alterations in the mechanisms of plasticity that contribute to cognitive functions. One functional alteration that could directly affect plasticity is reduced synapse number, which could make it more difficult to attain the sufficient amount of cooperatively active synapses that is necessary to lead to network modifi cation. An early electron microscopic investigation at the perforant path–granule cell synapse showed that aged rats have a 27% decrease in axodendritic synapse number in the middle molecular layer of the dentate gyrus compared with young rats66,67.

Moreover, spatial memory deficits have been shown to correlate with a reduction in perforated synapses at the medial perforant path–granule cell synapse68. When these results were replicated with stereologically controlled measures of synapse number, the total number of synaptic contacts per neuron was found to be diminished significantly in the dentate gyrus middle molecular layer and inner molecular layer of aged rats relative to young adults. Both perforated and non-perforated axospinous synapses showed age-dependent decreases in numbers69. The primary difference between the new stereological synapse count data and the old synapse count data is the observation that age-related synaptic loss involves axospinous, but not axodendritic, junctions. Electrophysiological data support the anatomical observation that there is a reduction in synapse number in the dentate gyrus of older animals. In the aged rat, the field excitatory postsynaptic potential (EPSP) recorded in the dentate gyrus is reduced38,47. This reduction is accompanied by a decrease in the presynaptic fibre potential amplitude at the perforant path–dentate gyrus granule cell synapse47,70. Because there is no loss of entorhinal cells13, this decrease is probably due to a reduction in axon collaterals from layer II of the entorhinal cortex to the granule cells. Interestingly, whereas the field EPSP in the aged dentate gyrus decreases, for a given magnitude of afferent fibre response, old animals show a larger syn aptic field potential47, which indicates that fewer fibres are able to elicit larger postsynaptic currents in aged animals. Consistent with this, the unitary EPSP is increased in old granule cells in response to stimulation of single afferent perforant path fibres71. The increase in unitary EPSP size in the dentate gyrus is probably mediated by an increase in AMPA (α-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid) receptor currents and is suggestive of a compensatory mechanism that increases postsynaptic sensitivity in response to the reduced medial perforant path input39,47,71. Another possible mediator of the larger currents is the increase in electrotonic coupling that is observed between aged hippocampal neurons. In fact, there is a 15% increase in the number of gap junctions between granule cells in aged compared with young rats (FIG. 1b). Old rats also show an increase in electrotonic coupling between neurons in areas CA1 (15% more gap junctions; FIG. 1d) and CA3 (18% more gap junctions) compared with young rats72. A reduction in axospinous synapses in the dentate gyrus is correlated with spatial memory deficits in aged rats68. This is not the case for Schaffer collateral–CA1 synapses, as the total synapse number remains the same across different age groups73. When the post synaptic density area of axospinous synapses in area CA1 is compared between aged learning-impaired and learningunimpaired rats, however, the impaired animals show a profound reduction in the postsynaptic density area of perforated synapses74. These findings support the idea that many hippocampal perforated synapses become non-functional or silent in aged learning-impaired rats, and this loss of functional synapses might contribute to cognitive decline during normal ageing. Electrophysiological data also support the hypothesis that there is a loss of functional synapses in area CA1. The amplitude of the field EPSP recorded in area CA1 of aged rats is reduced compared with young rats39,75,76. The unitary EPSP size in the CA1 region of the hippocampus in aged rats is preserved39, however, and there is no change in the amplitude of the Schaffer collateral presynaptic fibre potential77–79, which indicates that there is no axonal pruning of aged Schaffer collateral fibres. Interestingly, the postsynaptic density area of perforated synapses declines most significantly in aged learningimpaired Long Evans rats74. In an electrophysiological study, however, the field EPSP amplitude was found to be reduced in both impaired and unimpaired aged Fischer 344 rats80, whereas plasticity mechanisms were defective only in the impaired group80. Combined, these data implicate the perforated synapse postsynaptic density area in plasticity mechanisms independent of fast synaptic transmission processes. The effects of altered morphology, biophysical properties and synaptic connections of aged neurons on plasticity can be assessed by measuring age-associated alterations in long-term potentiation (LTP) and long-term depression (LTD). LTP can be divided into an induction phase (earlyphase LTP) and a maintenance phase (late-phase LTP). The induction phase involves the temporal association of presynaptic glutamate release with postsynaptic depolarization (necessary to eject Mg2+ from the pores of NMDA (N-methyl-d-aspartate) receptors), which results in an increase in intracellular Ca2+ (REF. 81). LTP maintenance is the continued expression‎ of increased synaptic efficacy that persists after induction. It probably involves changes in gene expression‎ and insertion of AMPA receptors into the postsynaptic membrane82. Aged rats have deficits in both LTP induction and maintenance. These deficits, however, are complex, protocol-dependent and region-specific. Although there is a reduction in the field EPSPs recorded both in the dentate gyrus47,71 and in area CA1 (REFS 39,75,76), aged animals can show intact LTP induction at the perforant path–granule cell synapse38,83,84, the CA3–CA1 Schaffer collateral synapse85,86 and the perforant path–CA3 pyramidal cell synapse87 when robust high-frequency, high current amplitude stimulation protocols are used (FIG. 2a). Even when supra-threshold stimulation parameters are used, however, aged rats have a deficit in the maintenance of LTP in both the dentate gyrus38,47 and CA3 compared with young rats87. When peri-threshold stimulation parameters are used, LTP induction deficits can be observed in both the dentate gyrus and CA1. In the dentate gyrus, when weak presynaptic stimulation is combined with direct depolariza tion of the granule cell, larger amplitude current injection is required to elicit LTP at the perforant path–granule cell synapse of aged rats compared with young rats88. This indicates that aged granule cells in the dentate gyrus have an increased threshold for LTP induction. The pattern of age-related LTP deficits in CA1 pyramidal cells is different from that observed in the dentate gyrus. Aged neurons in area CA1 do not have an increased threshold for LTP89, but when peri-threshold stimulation parameters are used, the level of LTP induction in aged rats is less than in young rats78,80,90,91 (FIG. 2b). For example, when LTP induction is measured in young and old rat hippocampal slices using four-pulse stimulation at 100 Hz90 or the primed-burst protocol, in which a single priming pulse is followed 170 ms later by four stimulus pulses at 200 Hz91, the increase in field EPSP slope is less in the aged rats than in the young rats. These induction deficits occur even if the stimulus intensity of the Schaffer collaterals is increased to match field EPSP amplitudes between young and aged rats78. Although the aged rats’ Schaffer collateral axons can follow high-frequency stimulation as well as those of young rats, aged CA1 neurons show weaker temporal summation of the multiple EPSPs induced by high-frequency stimulation. Therefore, during high-frequency bursts, CA1 pyramidal cells are less depolarized, which explains the age-related LTP induction impairment in CA1 (REF. 78). It is possible that age-related changes in Ca2+ regulation cause some portion of the observed age-related plasticity deficits. In particular, it has been proposed that post synaptic intracellular Ca2+ concentrations are involved in setting the synaptic modification threshold. This threshold might then affect the probability that a synapse will be depressed or potentiated at a given time53,92,93. Ca2+ dyshomeostasis in aged animals51,53,54 could, therefore, alter the probability that synaptic activity will induce either LTP or LTD. This idea is supported by Ca2+ imaging studies, which have shown that the resting Ca2+ concentration does not differ substantially with age in area CA1. Greater elevation of somatic Ca2+ and greater depression of EPSP frequency facilitation, however, develop in aged CA1 neurons in response to stimulation94. In line with the Ca2+ hypothesis of age-related plasticity impairments is the finding that aged rats are more susceptible than are young rats to LTD95 (FIG. 2c) and to the reversal of LTP53 (FIG. 2d). Moreover, it was recently shown that inhibition of Ca2+ release from intracellular Ca2+ stores attenuated LTD induction in aged CA1 neurons96 (FIG. 2c).

Age-related changes in gene expression‎

It is known that the maintenance of LTP requires gene expression‎ and de novo protein synthesis; therefore, it is not surprising that aged animals also show alterations in these processes. The investigation of the role of immediate-early genes (IEGs) in neural plasticity began in 1987 with the observation that the IEG c-fos is rapidly induced in neurons following seizures97. Subsequent investigations showed that IEGs are expressed following LTP induction98–100, which led to the hypothesis that IEGs are dynamically regulated by specific forms of patterned synaptic activity believed to underlie information storage98. It was later shown that IEGs are expressed by cells that are activated during behaviour such as spatial exploration101. A crucial event for the induction of expression‎ of IEGs is the phosphorylation of CREB (cAMP-responsive element-binding protein). Once phosphorylated, CREB promotes the transcription of IEG mRNA that may then be translated into protein. On the basis of the functional role of the protein, IEGs can be grouped into two classes: inducible transcription factors and effector proteins. Among the inducible transcription factors are c-jun, c-fos and zif268. After c-jun and c-fos mRNA are translated into protein, their protein products can form a heterodimer called the activator protein 1 (AP1) complex. AP1 is a transcription factor that promotes the expression‎ of late-response genes, some of which are important for the growth of new synapses or the modification of synaptic structure102,103. The expression‎ of zif268 is necessary, although probably not sufficient, for the maintenance of LTP and long-term memory104.

region of the dendrite that receives the synaptic input that initiated transcription107, and is proposed to be involved in the structural rearrangement of activated dendrites108. This probably involves AMPA receptor trafficking, as Arc protein has also been shown to reduce AMPA receptor currents (P. F. Worley, personal communication). Finally, Arc expression‎ is necessary for the maintenance, but not the induction, of LTP and long-term memory109. Age-associated changes in gene expression‎ have been investigated using several techniques (BOX 1), each of which has specific advantages and disadvantages. Gene microarray technology allows researchers to monitor the expression‎ level of thousands of genes in a given brain region and make comparisons between young and old animals. The initial use of this method, in mice, showed age-related alterations in the expression‎ of hundreds of genes110,111, but did not involve the use of formal statistical tools to eval‎uate age effects and their behavioural relevance. In a later study that used behaviourally characterized rats, gene expression‎ alterations in area CA1 were found to correlate with age-related cognitive decline. The behaviourally relevant upregulated genes included several that are associated with inflammation and intracellular Ca2+ release pathways, whereas genes associated with energy metabolism, biosynthesis and activity-regulated synaptogenesis were downregulated. Arc and Narp were two of the genes that were shown to be downregulated112. These results should be interpreted with caution, however, as the data reflect resting levels of gene expression‎. As many of the genes that are necessary for learning and memory are only robustly expressed after synaptic activity, resting levels of expression‎ might not capture an age difference that may occur in gene expression‎ during behaviour. Northern blots can be used to measure the amount of RNA transcribed from a particular gene. When this technique was used to compare resting levels of c-fos, c-jun and AP1 activity between young and aged rats there was no age-associated difference113. The expression‎ levels of Arc, c-fos, c-jun, zif268 and Narp mRNA have been measured following LTP-inducing stimuli using a reverse northern strategy. In adult and aged memory-impaired rats, the induced levels of ARC, c-jun, junB, Zif268 and NARP mRNA are similar but the amount of c-fos mRNA is significantly higher in aged animals114. Both the microarray and northern blot techniques are limited, however, by their lack of cell specificity. Changes in the proportion of cells that express a gene can be assessed using fluorescence in situ hybridization. This allows exact determination of which individual cells are expressing which genes. For example, in aged rats, granule cells of the dentate gyrus, but not the pyramidal cells of areas CA1 and CA3, have a significantly smaller proportion of neurons that express Arc following spatial exploration115. Interestingly, in studies using MRI methods in humans and monkeys, the granule cells also seem to be particularly vulnerable to the effects of normal ageing115,116. Fluorescence in situ hybridization alone, however, does not allow determination of the magnitude of expression‎ of a particular gene in a cell.

By using real-time quantitative reverse transcriptase polymerase chain reaction (RT-PCR) concurrently with fluorescence in situ hybridization, changes in gene expression‎ levels within a single cell can be determined. For example, in area CA1, the proportion of cells that express c-fos mRNA is similar between young and aged rats but when RT-PCR is carried out, young rats are found to have higher levels of c-fos mRNA compared with the old animals (M. K. Chawla, unpublished observations). This indicates that although a similar number of pyramidal neurons express c-fos across different age groups, the individual cells from old animals transcribe less c-fos mRNA, which may lead to dysregulation of other genes that depend on the AP1 transcription factor.

Dynamics of aged neural ensembles

It is widely agreed that modifiable neuronal ensembles support cognition. Therefore, alterations in these networks could be responsible for the behavioural impairments observed with ageing. Advances in multiple single unit recording methods (BOX 2) have allowed the dynamics of hippocampal cell populations to be investigated in behaving rats, and studies using these methods have shown that certain properties of these networks are compromised during ageing. Interestingly, many of the agerelated changes that have been discovered can be linked to plasticity deficits, as blockade of NMDA receptors in young rats results in ensemble dynamics that resemble those of aged rats117,118. Neuronal recordings from the hippocampus of adult rats reveal that when a rat explores an environment, pyramidal119 and granule120 cells show patterned neural activity that is highly correlated with a rat’s position in space (that is, the ‘place field’ of the cell; BOX 2). Between 30% and 50% of CA1 pyramidal cells show place-specific firing in a given environment101,121, which has earned these neurons the name ‘place cells’. When the firing patterns of many hippocampal neurons are recorded simultaneously, it is possible to reconstruct the position of a rat in an environment from the place cell firing patterns alone121. The composite cell activity is ‘map-like’ and, in different environments, hippocampal place maps change markedly. Although these maps can be driven by external environmental features, internal events are also important and a new map might be generated in the same environment if the demands of the task change122–124. In young rats, CA1 place fields expand asymmetrically during repeated route following (for example, traversing a circular track), which results in a shift in the centre of mass of place fields in the direction opposite to the rat’s trajectory125. This observation is consistent with neural network models dating back to Hebb’s 1949 concept of the ‘phase sequence’ of cell assemblies, which suggested that an associative, temporally asymmetric synaptic plasticity mechanism could serve to encode sequences or episodes of experience126. The magnitude of this place field expansion, however, significantly decreases in aged rats60. It is likely that this age-associated reduction in experience-dependent plasticity is due to LTP deficits, as it does not occur when the NMDA receptor antagonist CPP (3-(2-carboxypiperazin-4-yl)propyl1-phosphonic acid) is administered to young rats118. In addition to age-related alterations in experiencedependent place field expansion, the maintenance of place maps also differs between young and old animals. In normal young rats, a place map for a given environment can remain stable for months127. Therefore, when a rat is returned to the same environment, the same place map is retrieved. A similar stability of CA1 place maps in aged rats is observed within and between episodes of behaviour in the same environment. Occasionally, however, if the old rat is removed from the environment and returned later, the original place map is not retrieved and an independent population of place cells may be activated even in a familiar room59. This ‘remapping’ predicts that rats should show bimodal performance on tasks that require the functional integrity of the hippocampus. For spatial tasks, good performance should correspond to retrieval‎ of the original map, and poor performance should correspond to retrieval‎ of an incorrect map. This prediction seems to be correct. When trained on the spatial version of the Morris swim task, the performance of both young and aged rats is bimodal in early trials. This means that for some trials rats find the hidden escape platform with a short path but for other trials the rats do not recall the location of the platform and take a longer path. By the final training trials, however, the young rats’ performance is unimodal, with most rats taking a direct path to the platform. By contrast, the aged rats’ performance remains bimodal. The trials on which the old rats fail to correctly remember the location of the hidden escape platform could correspond to map retrieval‎ failures59. A probable mechanism for map retrieval‎ failures is defective LTP in aged rats. Although place-map stability within an episode does not require plasticity, the maintenance of place maps between episodes depends on an LTP-like mechanism. In young rats, blockade of NMDA receptors117 or protein synthesis inhibition128 has been shown to result in map retrieval‎ errors when the rat is returned to the same environment. When CA1 and CA3 place cell recordings are pooled it seems that spatial representations in old rats do not change when they should (for example, in response to major changes in the environment)61,129. Combining results across areas is problematic, however, as it has recently been shown that ensemble activity in these different subregions is dissociable130–132. This dissociation could reflect two competing functions of the hippo campal network: pattern completion versus pattern separation133,134. Moreover, there is evidence of a dissociation of the effects of ageing on CA1 and CA3 ensembles. In area CA1, spatial representations are less stable in aged compared with young rats59. By contrast, spatial representations in CA3 seem to be more rigid in aged rats. When an aged rat explores a familiar environment for 7 min and is then placed into a novel environment, spatial representations in area CA3 remain the same even though the environment has changed64. In young rats, however, CA3 place maps are independent between familiar and new environments64,131. A disruption in the ensemble characteristics of dentate gyrus granule cells, a structure known to be particularly vulnerable to the ageing process115,116, could contribute to the failure of aged CA3 networks to form new spatial representations. It is believed that the dentate gyrus makes information stored in hippocampal networks more dissimilar (that is, it is involved in pattern separation), thereby increasing storage capacity133. Because the transfer of information between granule cells and CA3 pyramidal cells declines, this might contribute to the inability of the aged CA3 network to form new spatial representations when required. 

Age-related changes in behaviour

Because the hippocampus and the PFC are particularly vulnerable to the ageing process, it is not surprising that performance on tasks that require information processing in these brain regions declines with age. Below is a brief discussion of selective examples of these behavioural changes in humans, non-human primates and rats. One such example is an age-related decline in spatial memory, which is a key element of most episodic experience. Compared with younger adults, episodic memory declines in aged humans, who show deficits in retrieving the contextual details of these memories135,136. In addition, aged humans137,138, monkeys139,140, dogs141, rats38,142,143 and mice144 all show deficits on tasks designed to test spatial navigation. This is consistent with the neurobiological data that suggest that hippocampal function is compromised with age145. An additional hippocampal-dependent impairment that is consistently observed during ageing across species is a deficit in trace eyeblink conditioning. The trace eyeblink conditioning task tests associative learning using a classical conditioning paradigm. In this task, a neutral or conditioned stimulus, which is usually a tone, is predictive of an aversive unconditioned stimulus (air puff or stimulation of eye). The time taken to acquire the eyeblink reflex to the neutral stimulus alone and the retention of this association is measured. Acquisition and retention of the learned eyeblink response both require the involvement of the hippocampus, as well as other brain structures such as the cerebellum146. Aged mice147, rats148, rabbits149,150 and humans151 are impaired in the acquisition of trace eyeblink conditioning. Whereas the hippocampus is crucially involved in spatial and episodic memory, the PFC is necessary for working memory152,153 and executive function154. In animals and humans, working memory function can be measured using the delayed non-matching-to-sample (DNMS) task. Aged rats155 and non-human primates156–158 show time-dependent deficits on the DNMS task, with the magnitude of the deficit increasing as the delay increases. Humans also show working memory impairments in advanced age159. The DNMS task also tests recognition memory, which depends on the perirhinal cortex153,160,161. It is therefore possible that age-related changes in this task could involve additional plasticity deficits in the perirhinal cortex, although little is known about the impact of ageing on the functional integrity of perirhinal circuits. In addition to its involvement in working memory, the PFC is considered to be the neural substrate of executive function, which also declines during normal ageing. One way to measure executive function is with the Wisconsin card sorting task (WCST). Aged humans are impaired on the WCST and make more perseverative errors162. Animal analogues of the WCST have been design ed, which also show that normal ageing leads to a decline in executive function. For example, relative to young adult monkeys, aged monkeys show a significant difficulty in the acquisition of a conceptual set shifting task and demonstrate a high degree of perseverative responding163.

Conclusions

In summary, during the normal ageing process, animals experience age-related cognitive decline. Historically, it was thought that primary contributions to the aetiology of this decline were massive cell loss1 and deterioration of dendritic branching17,18. However, we now know that the changes occurring during normal ageing are more subtle and selective than was once believed. In fact, the general pattern seems to be that most age-associated behavioural impairments result from region-specific changes in dendritic morphology, cellular connectivity, Ca2+ dysregulation, gene expression‎ or other factors that affect plasticity and ultimately alter the network dynamics of neural ensembles that support cognition. Of the brain regions affected by ageing, the hippocampus and the PFC seem to be particularly vulnerable, but even within and between these regions the impact of ageing on neuronal function can differ. The morpho logy of neurons in the PFC is more susceptible to age-related change, as these cells show a decrease in dendritic branching in rats30,31 and humans32,33. There is also evidence of a small but significant decline in cell number in area 8A of monkeys that is correlated with working memory impairments16. Although there is evidence of Ca2+ dysregulation in aged PFC neurons65, the functional consequences of this are not yet known. Moreover, so far, there are no reports of multiple single unit recordings in the PFC of awake behaving animals. More is known about the impact of ageing on hippo campal function. Ca2+ dysregulation51,53,54 and changes in synaptic connectivity69,74 might affect plasticity and gene expression‎, resulting in altered dynamics of hippo campal neuronal ensembles. Because more is known about the neurobiology of ageing in this brain region, there are therapeutic approaches on the horizon that might modify hippo campal neurobiology and slow age-related cognitive decline or partially restore mechanisms of plasticity. For example, agents that reduce intracellular Ca2+ concentra tion following neural activity could modulate the ratio of LTD and LTP induction, thereby partially restoring normal network dynamics. Considering that the average lifespan is increasing worldwide, understanding the brain mechanisms that are responsible for age-related cognitive impairment, and finding therapeutic agents that might curb this decline, becomes increasingly important.