Environmental enrichment alters dentate granule cell morphology in oldest‐old rat

Abstract

The hippocampus of aged rats shows marked age‐related morphological changes that could cause memory deficits. Experimental evidence has established that environmental enrichment attenuates memory deficits in aged rats. We therefore studied whether environmental enrichment produces morphological changes on the dentate granule cells of aged rats. Fifteen male Sprague‐Dawley rats, 24 months of age, were randomly distributed in two groups that were housed under standard (n= 7) or enriched (n= 8) environmental conditions for 26 days. Quantitative data of dendritic morphology from dentate gyrus granule cells were obtained on Golgi–Cox stained sections. Environmental enrichment significantly increased the complexity and size of dendritic tree (total number of segments increased by 61% and length by 116%), and spine density (88% increase). There were large interindividual differences within the enriched group, indicating differential individual responses to environmental stimulation. Previous studies in young animals have shown changes produced by environmental enrichment in the morphology of dentate gyrus granule cells. The results of the present study show that environmental enrichment can also produce changes in dentate granule cell morphology in the senescent brain. In conclusion, the hippocampus retains its neuroplastic capacity during aging, and enriched environmental housing conditions can attenuate age‐related dendritic regression and synaptic loss, thus preserving memory functions.

Introduction

The process of aging is usually followed by cognitive decline that is characterized by impairments in a variety of tests for spatial memory [1, 2, 3, 4, 5, 6]. It is suggested that the anatomical substrate for these impairments is to be found in age‐related changes that appear mostly in hippocampal formation [1, 2, 3, 7, 8, 9, 10, 11, 12, 13]. In particular, the dendritic trees of dentate granular cells are the main receptive field for afferent projections in the hippocampus. These cells receive sensory information from neocortical areas through the perforant pathway synapses from neurons in the superficial layers of the entorhinal cortex [14, 15]. Numerous anatomical studies indicate that the dentate gyrus of aged rats shows dendritic atrophy, partial deafferentation, concomitant loss of synapses, astrocyte hypertrophy [16, 17, 18, 19, 20, 21, 22, 23, 24], as well as decrease in neurogenesis in comparison with young animals [25]. Electrophysiological data have shown a decrease in the NMDAN (N‐methyl‐D‐aspartic acid)‐receptor‐mediated response at perforant path synapses onto dentate gyrus granule cells, and impairments of synaptic plasticity, which include deficits in the induction and maintenance of long‐term potentiation, lower thresholds for depotentiation and long‐term depression [26]. Neurochemical studies showed decrease in neurotrophin levels [27, 28].

Since the early 1960s it has been known that housing higher number of animals in large cages, which contain various stimulating objects that were regularly changed, produces changes on differential behavioural, neuroanatomical, neurophysiological and neurochemical parameters in the rat brain (reviewed in Rosenzweig [29]). These changes could be produced even in aged brain although their pattern in not necessarily the same as in young animals [30, 31, 32].

Previous studies have reported that environmental enrichment produces functional and morphological changes in the dentate gyrus of young and adult animals. For example, electrophysiological studies demonstrated that exposure to an enriched environment regulates excitability, synaptic transmission and long‐term potentiation in the rat dentate gyrus [33]. Changes in dentate granule cells morphology can be induced by environmental enrichment in the weaning and post‐weaning period [34, 35, 36, 37, 38, 39, 40, 41], although data in adult animals are contradictory [42, 43, 44, 45, 46, 47].

In old age, environmental enrichment can improve performance on tests of learning and memory [48, 49, 50, 51, 52], reduce age‐related gliosis [49], increase levels of synaptophisin [53] and neurotrophic factors [54, 55] and stimulates dentate gyrus neurogenesis [56, 57]. To our knowledge the effect of enriched environment on dentate granule cells dendritic morphology has not been investigated in the aged animals.

Because environmental stimulation in old age can prevent or ameliorate impairments in a variety of spatial learning and memory tasks [31, 49, 50, 51, 58], the working hypothesis of the present study is that positive behavioural effects of housing in stimulating environment are associated with morphological changes in the hippocampus of aged animals. In this study, we have examined the effects of differential housing in oldest‐old rats (24 months of age) on dendritic morphology on dentate gyrus granule neurons as assessed by Golgi–Cox method.

Materials and methods

Animals and environmental housing

The animals used were 15 male Sprague‐Dawley rats, 24 months of age, obtained from a commercial breeder (Alab AB, Sollentuna, Sweden). They were housed on a 12‐hr on/12‐hr off illumination schedule (lights on 6.00 a.m.) with unlimited access to food and water. The animals were randomly divided into two experimental groups and exposed to differential environmental housing, enriched conditions (EC; n= 8) and social conditions (SC; n= 7, one rat died at the end of housing), respectively, for 26 days. During the experimental period the animals were kept in the same room at the animal department so that additional stimulation from environment (handlers, noise) were the same for both groups.

The EC consisted of large wire mesh cages measuring 100 × 60 × 35 cm (eight rats per cage) containing ladders, shelves, tunnels and additional diverse toy objects. EC rats were exposed to three different objects (toys) that were changed daily. The control rats were housed in standard laboratory Plexiglas cages measuring 45 × 30 × 20 cm (four rats per cage) conditions we refer as SC. All the experimental procedures and housing conditions used in this study followed the guidelines of Swedish animal protection legislation and were approved by the Animal Ethics Committee at the Karolinska Institute.

Tissue preparations and identification of analysed region

After the 26 days of the experimental period all rats were deeply anaesthetized intraperitoneally with pentobarbital overdose (60 mg/kg) and then decapitated. The brains were removed and divided in the midsagital direction. Both hemispheres were alternatively cut in a coronal plane into three blocks of tissue that were placed in Golgi–Cox solution for 3 weeks, with one change of solution after 3 days [59]. After impregnation, the tissue was dehydrated, embedded in celloidin and sectioned coronaly at 180 μm. For developing the staining, the sections were immersed in 20% ammonium hydroxide for 5 min. and then transferred to a 15% solution for 25 min. After rinsing they were further processed with 1% thiosulphate for 7 min., dehydrated in alcohol, cleared in Histoclear and covered with Histomount (National Diagnostics, Atlanta, Georgia, USA) mounting media [59].

The subjects were coded so that the investigators were not aware of the housing conditions. Anatomical level of analysis was between –3.5 and –4.5 from the position of Bregma according to the rat atlas [60]. We marked the rectangle in the middle part of the dentate gyrus upper blade (x–y‐dimension), which represents the area of analysis. All neurons included in this area with their cell body positioned in the middle third of sections thickness (z‐dimension starting from –60 μm from the surface) were reconstructed (Fig. 1). In Golgi–Cox slices neurons that lie in the middle third of the section thickness have highest level of impregnation [59], and this criterion could reduce the number of segments cut at the section surface. 12 to 15 consecutive stained granule neurons for each animal were sampled for quantitative analysis in the specified area regardless of their position in the granular layer. Because all brain tissue was processed under the same laboratory conditions, the obtained values were not corrected for the shrinkage factor [61, 62].

Figure 1.

Low magnification micrograph of rat hippocampal formation impregnated by Golgi–Cox method shows the basic anatomical relations, CA1 region, CA3 region and dentate hilus (hil). Box indicates the position of selected portion of dentate upper blade granular cell layer (gl) where neurons were analysed. Micrograph was taken from the animal housed in the enriched environmental conditions. Molecular layer (ml). Scale bar indicates 500 μm.

Dendritic tree reconstruction

Quantitative morphometric analysis was performed with Neurolucida 3.18 software (Microbrightfield Inc., Colchester, VT, USA) and automatic dendrite measuring system that provides three‐dimensional data of the dendritic tree [15, 63]. Measurements were made using a 60× air objective with actual enlargement on a screen equal to 4200×. Neurons were drawn over the live picture on the PC screen, bringing the signed point into the sharp focus when drawing. Changes in depth (z‐dimension) were identified for each drawn point, and automatically corrected according to Snell's law for diffraction air correction factor (1.515). X–Y coordinates were also given to the each point in relation to the reference point.

Morphometric measurements

The following parameters of the granule cell morphology were analysed: (i) somatic cell surface, i.e. the area of the cell soma projected onto a two‐dimensional (x–y) plane of sectioning to indicate its size, (ii) average dendritic diameter, (iii) base dendritic diameter, (iv) number of primary dendrites, (v) total number of segments, (vi) total dendritic length per neuron, (vii) average length of intermediate and terminal segments, (viii) spine density on the entire tree, (ix) spine density on intermedial and terminal segments and (x) spine number per individual dendrite.

For the analysis of segment length and spine density, each dendritic tree was divided into subgroups of intermediate and terminal segments depending on their further branching and not their distance from cell body. These subgroups were shown to have different length distributions [61, 62]. Intermediate segments are the segments between the dendritic origin and the first bifurcation point, or between two consecutive bifurcation points [61, 62]. Terminal segments are segments between the terminal tip of dendrites and the last bifurcation point before the terminal tip [61, 62]. Incomplete segments refer to segments that were impossible to trace completely, because they were cut at the surface of the section or ran into a precipitation, and these were excluded from the analysis of individual segment length. Spine counts were done on whole dendritic tree of selected neurons.

Measurements of thickness of molecular and granular layers were made on a Golgi–Cox stained sections using Leica Q Win Pro v. 2.3 (Leica Imaging Systems, Wetzlar, Germany). Eight measurements per subject were done, on two neighbouring sections. All measurements were taken on the different positions through dentate gyrus upper blade.

Our sample of neurons differs from samples in other studies [36, 39, 42, 44, 64, 65, 66, 67] in which neurons were selected individually, to avoid analysing incompletely impregnated neurons. In that way neurons with low dendritic length and low spine density were not selected for analysis, assumed to be incompletely impregnated. Given that we were not able to consistently differentiate neurons affected by regressive changes and those affected by incomplete impregnation, we decide to count all neurons in the middle third of section thickness of the defined area. At this depth of the section neurons tend to be completely impregnated, and we did not find clear signs of incomplete impregnation on any neuron reconstructed.

However, this might raise question if our result are comparable with result of other Golgi studies performed on dentate gyrus granule cells. Therefore, we have performed additional statistical analysis on the selected population of reconstructed neurons (20% with the largest dendritic field from each animal). Also, in parallel with the old rats, two young adult rats (4 months of age) were killed to serve as a control of impregnation quality. The obtained values for the total length of dentate granule cells (1177 μm) were around 1000–1400 μm, as described in the other Golgi–Cox studies [47, 67]. We also performed study on layer III pyramidal neurons in the occipital cortex (data not shown) of the same animals analysed in this study, where basal dendrite values (SC 885 μm, EC 1188 μm) corresponds to the values described in the literature for adult rats (SC 600 μm, EC 870 μm) [38].

Data analysis

All the parameters were calculated as mean values per neuron for each animal, and these data were used for the statistical analysis. A non‐parametrical Mann‐Whitney U‐test was applied. Differences between groups were considered significant at P < 0.05.

Results

Qualitative observations

Neurons positioned in the two distinct parts of dentate gyrus, i.e. the upper blade portion of the granule cell layer adjacent to the CA1 region of the hippocampus, and the opposite portion of granule cell layer (the lower blade) showed differences in morphology. Neurons located in the upper blade tended to have larger dendritic trees, and higher spine density than cells located in the lower blade, as previously observed [30]. To maintain a uniform population of neurons for analysis in both groups and to avoid the technical/staining problems on the edge of the structure, neurons from the middle part of dentate gyrus upper blade were drawn for all the animals (Fig. 1).

In the Golgi–Cox stained sections dentate granule cells have elliptical cell bodies and a characteristic cone‐shaped tree of spiny dendrites with the branches directed towards the superficial portion of molecular layer. The granular layer consists of several rows of cells. Those that are positioned in the deeper half of the layer have more elliptically shaped elongated soma with only one or two dendrites. By contrast, those cells positioned in the superficial half of the granular layer tend to have more rounded soma with two to six primary dendrites rising from it and a wider dendritic tree relative to cells located in the deeper portions [39, 68]. Granule neurons located in the centre of the granule layer are likely to exhibit intermediate characteristics. Considering previous findings that both types of neurons are affected by environmental manipulation [36, 39, 67], and that the morphology of both cell types changed in a similar manner [36] we analysed all neurons in the medial part of the dentate upper blade regardless of their position in the molecular layer.

In both groups the granule cell dendritic tree never reached the most superficial parts of the molecular layer (Fig. 2), whereas the dendrites of granule neurons in EC animals tended to extend closer to the hippocampal fissure (Fig. 2B).

Figure 2.

Microphotograph of Golgi–Cox impregnated dentate gyrus granule cells in rats under standard social conditions (A, C) and enriched environment (B, D). The granule cell dendritic trees (gl – granular layer) protrude almost throughout entire thickness of molecular layer (mz) in enriched (B) that was not visible in control group (A). In control group (C) dendritic complexity was much lower than in enriched group (D). Higher magnification (box indicated in C, D and magnified in upper right corner of the Fig. C, D) of fine details of dendritic spines in enriched group (D) showed higher density then in control group (C). Most of spines on magnified part are mushroom and stubby spines are indicated by arrows. Scale bar indicates 200 μm (A, B) and 20 μm (C, D).

The dendritic trees of granule neurons display differences in spine density and types of dendritic spines depending on their distance from the cell body. Spine density was much higher for distal segments. The main spine type on proximal segments was stubby spines, while distal segments contained a high proportion of mushroom spines. Comparison between groups revealed increase in spine density on terminal segments in EC group (Fig. 2D). Environmental influence on terminal parts of dendritic tree was also evident in change of morphology of mushroom spines that tend to have longer necks than in the SC group (Fig. 2C).

Variability in the appearance of neighbouring neurons was evident in each case (Fig. 3). Neurons ranged from those with rich dendritic arborization extending over large distances with densely positioned spines, to those with regressed dendritic tree of only few segments with only a few spines that were more numerous in the SC group. A certain amount of variability in the appearance of granule neurons dendritic tree was evident between the individuals in the same group, as well (Fig. 3).

Figure 3.

Neurolucida reconstructions of best‐impregnated dentate granule neurons from five different specimens of each, control (A) and enriched (B) group. Every neuron was reconstructed from different specimen. (A) In control group the neurons with most complex dendritic tree (arrows) exceed in size and complexity the best‐developed neurons from some of enriched specimens. (B) In some of the specimens from enriched groups part of neurons have very complex and expanded dendritic tree (arrows) and in control group not a single cells have dendritic tree developed at similar manner. Point on dendrites indicate dendritic spines. Scale bar indicates 100 μm.

Quantitative results

The results showing an effect of enriched environment on the main morphometric parameters of granule cells are represented in the Table 1. The number of primary dendrites arising from the cell body and the total number of segments per neuron were significantly affected by environmental enrichment. Increase in the total segment number by 61% and the number of primary dendrite by 35%, suggests that outgrowth of new segments is more intensive on the distal part of dendritic tree. A significant environment effect was also observed for total dendritic length, where two‐fold increase was observed in enriched group. The mean values were 179.96 ± 65.68 μm for SC rats and 389.57 ± 120.57 μm for EC rats. The percentage of cut segments was not significantly different between SC and EC animals (mean values were 3% and 6%, respectively). The average length of intermediate and terminal segments was significantly greater in the EC group, but the differences were more marked for intermediate segments (P < 0.01 for the intermediate segments versus P < 0.05 for the terminal segments; Mann‐Whitney U, Table 1). This can be also explained as more intensive segment outgrowth on the distal part of the dendritic tree. We found no difference between the groups studied, in soma surface and dendritic diameter.

Table 1.

Effects of enriched environment on granule cell dendritic tree morphology

	SC Mean ± S.D.	EC Mean ± S.D.	Inc. %	P
1. Soma surface (μm2)	128.0 ± 13.4	146.3 ± 23.6	14.3	n.s.
2. Average dendritic diameter (μm)	1.8 ± 0.2	1.8 ± 0.2	0	n.s.
3. Base dendritic diameter (μm)	2.4 ± 0.3	2.3 ± 0.3	0	n.s.
4. Primary dendrites per neuron (n)	2.0 ± 0.5	2.7 ± 0.4	35	*
5. Segments per neuron (n)	6.7 ± 1.2	10.8 ± 2.7	61.2	**
6. Total dendritic length (μm)	180.0 ± 65.7	389.6 ±120.6	116.4	**
7. Average length of int. seg. (μm)	18.1 ± 3.2	30.9 ± 5.5	70.7	**
8. Average length of ter. seg. (μm)	30.0 ± 7.6	38.7 ± 4.4	29	*

Values shown in the table are means and standard deviations of means (S.D).

SC, control animals; EC, animals housed in enriched environment; Inc., increase in percentage.

Non‐parametrical Mann‐Whitney U‐test: *P < 0.05; **P < 0.01; n.s., not significant.

All the values obtained on the selected sample of 20% of the neurons with the largest dendritic tree from a single animal showed between 60% and 100% increase as compared to the values of the whole sample. The ratio between the SC and the EC group in all analysed parameters was slightly lower than on the whole sample, as was the level of statistical significance. Increase in average length of terminal segments did not reach statistical significance (Table 2). Considering that the sample including 20% of the neurons with the largest dendritic tree in the SC group includes the population of neurons less affected by regressive changes, we can assume that there will be fewer outgrowths of new segments and that elongation will mostly affect intermediate segments. These data also strongly suggest that the neurons sampled in our study represent a ‘real’ aging regression and not a methodological artefact of Golgi method.

Table 2.

Effects of enriched environment on morphology of selected granule cells, corresponding to 20% of cells with longest dendritic tree from each animal

	SC Mean ± S.D.	EC Mean ± S.D.	Inc. %	P
1. Segments per neuron (n)	12.7 ± 3.0	17.6 ± 4.4	38.6	*
2. Total dendritic length (μm)	363.0 ± 186.5	740.0 ± 207.5	103.9	**
3. Average length of int. seg. (μm)	33.2 ± 9.5	48.6 ± 9.5	46.4	*
4. Average length of ter. seg. (μm)	55.7 ± 13.1	60.2 ± 7.4	8.1	n.s.

Values shown in the table are means and standard deviations of means (S.D).

SC, control animals; EC, animals housed in enriched environment; inc., increase in percentage; int. seg., intermediate segments; ter. seg., terminal segments.

Non‐parametrical Mann‐Whitney U‐test: *P < 0.05; **P < 0.01; n.s., not significant.

Spine density was also influenced by the different environmental conditions, and that effect was evident on terminal segments (P < 0.05; Mann‐Whitney U; Table 3). Considering that spine density in EC was 89% higher than in SC, and at the same time that the total dendritic length was also 116% longer, it can be calculated that total number of spine is about four times larger in EC than in SC group. We calculate also the spine number per dendrite (Table 3), showing marked increase (238%) on whole sample and smaller (113%) on selected group of dendrites with highest spine density. This suggests increase in spine density on present dendrites and spine formation on new branches.

Table 3.

Effects of enriched environment on granule cell spine density

	SC Mean ± S.D.	EC Mean ± S.D.	Inc. %	P
1. Spine density (n/μm)	0.104 ± 0.057	0.196 ± 0.071	88.5	*
2. Spine density (n/μm) – 20%	0.226 ± 0.123	0.355 ± 0.105	57.1	*
3. Spine number per dendrite	9.4 ± 4.1	31.8 ± 9.9	238.3	**
4. Spine number per dendrite – 20%	31.5 ± 16.5	67.4 ± 17.5	113.3	**
5. Spine density on int. seg. (n/μm)	0.117 ± 0.168	0.152 ± 0.073	29.9	n.s.
6. Spine density on ter. seg. (n/μm)	0.113 ± 0.07	0.222 ± 0.08	96.5	*

Values shown in the table are means and standard deviations of means (S.D).

SC, control animals; EC, animals housed in enriched environment; inc., increase in percentage; int. seg., intermediate segments; ter. seg., terminal segments; 20%, analysis made on 20% of dendrites with highest spine density.

Non‐parametrical Mann‐Whitney U‐test: *P < 0.05; **P < 0.01; n.s., not significant.

Differences for total dendritic length and number of segment between animals with the lowest and the highest values within each group were around twofold. These differences were slightly lower for terminal and intermediate segment length (Fig. 4). It is interesting that one individual subject in the SC group does not show high regression in dendritic length and bifurcation when compared to the EC group, as well as that one subject in the EC group shows very low values, indicating that its response to EC was low. This effect was more expressed on selected population of the 20% of the neurons with the highest dendritic tree, indicating different response to SC and EC between individual animals, as selective influence on different neuronal classes.

Figure 4.

Mean values per analysed neuron expressed for individual animal. (A) Total dendritic length per neuron (μm); (B) Number of segments per neuron; (C) average length of intermediate segments (μm) and (D) average length of terminal segments (μm). Note clear significant shift between social condition (SC) and enriched condition (EC) group, but also overlap of values for half of animals from both groups. Values are shown for the whole sample of neurons reconstructed (100%) and for selected 20% of neurons with most complex dendritic tree in each of specimen (20%) (see ‘Material and methods’). However, the same difference between enriched and social condition was present in whole and selected sample, with respect to average length of individual terminal segment (D). Standard deviation of mean was mostly between 30% and 50% of mean.

Changes in neuronal morphometric parameters induced by environmental stimulation were not accompanied by changes in the thickness of the molecular and the granular layer (Table 4).

Table 4.

Thickness of molecular and granular layer

	SC Mean ± S.D.	EC Mean ± S.D.	P
1. Thickness of molecular layer (μm)	239.58 ± 31.14	236.47 ± 26.97	n.s.
2. Thickness of granular layer (μm)	96.55 ± 13.33	109.83 ± 20.65	n.s.

Values shown in the table are means and standard deviations of means (S.D).

SC, control animals; EC, animals housed in enriched environment.

Mann‐Whitney U‐test: n.s., not significant.

Discussion

Our study shows that exposure of old rats to an enriched environment for 26 days induces marked morphological changes in the dendritic tree of dentate gyrus granule neurons, and increase in the complexity and size of the dendritic tree as well as in spine density. Significant changes were observed in the number of primary dendrites, dendritic tree branching, total dendritic length, average length of both terminal and intermediate segments, spine density in general and spine density on terminal segments, whereas an increase in intermediate segment length of 30% was not found to be statistically significant. These results are in line with previous studies in younger animals [33, 34, 35, 36, 37, 38, 39, 40, 42, 43, 44, 47]. Dendrite diameter and cell soma size, as well as molecular and granular layer thickness were unaffected by environmental stimulation [69, 70].

The increase in total dendritic length and total number of segments found in the EC group, where the effect of elongation is more obvious on intermediate than on terminal segment, might be a consequence of two distinct events: (i) an outgrowth of new dendritic segments, that according to smaller increase in number of primary dendrites appears mostly on distal part of dendritic tree and (ii) elongation of previously present segments. Such findings have been obtained on pyramidal neurons in the cortex although some other studies show a different pattern [41, 71]. In addition, although results obtained on a selected population of reconstructed neurons (20% of the neurons with the largest dendritic trees) showed less segment outgrowth and an absence of elongation of terminal segments in the EC group, the total dendritic length has doubled. This is in line with available data [63, 72] and supports the view that the sample in our SC represents a real regression and not a methodological artefact as a result o incomplete impregnation.

The values for total dendritic length and spine density of granule cells obtained in our study were much lower in both groups (especially for the SC group) than the values obtained in other rat Golgi studies [36, 39, 42, 43, 44, 64, 65, 66, 67, 68]. It is possible that the dendrites had started to express regressive changes, a phenomenon that has been observed in the ‘oldest old’ stage in rodents, monkeys and human beings [65, 69, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84]. This might imply that the effect of enriched environment is actually in slowing down the speed of regression. Another explanation could be due to a limitation of the Golgi–Cox method, that according to Desmond and Levy [68] provides less detailed pattern of branching and does not regularly impregnate the distal dendritic tips. Although visualization of neurons in quantitative studies using various Golgi methods and intracellular staining studies consequently influences metric results giving large differences in absolute dendritic extent [82] it is important to note that until now, the use of different staining methods in other research fields has produced results with similar differences in dendritic changes (for review, see de Brabander et al.[82]).

Due to differences from other studies in neuron selection, we also performed an analysis on those 20% of neurons which had the largest dendritic tree (see ‘Materials and methods’). However, the data obtained on that sample for the EC group showed values close to the adult levels obtained in previous studies of adult rats housed in SC [67], and showed the same effect between the SC and the EC group as for the whole reconstructed population. We believe that the lower values obtained on the whole sample in our study simply reflect a different sampling design, because in the other studies only the best impregnated neurons were analysed. Taken together, and with full respect to methodological factors affecting the results, we consider it more likely that oldest‐old Sprague‐Dawley rats housed in SC have a large dendritic regression of granular cells in the dentate gyrus. This is in line with a large regression within neuronal circuitries obtained during normal aging in certain brain regions [8, 9, 81, 85], that can be modulated by environmental stimulation [80].

An increase in spine density in addition to an increase in dendritic complexity further increases the surface area of granular cells by means of which they interact with other neurons. Because dendritic spines are the primary site of excitatory synapses [86] the experience of living in enriched environment probably increases the overall number of excitatory synapses in the dentate gyrus of oldest‐old rats. The increase in spine density after being housed in an enriched environment observed in our study is in line with previously obtained data on the cerebral cortex and hippocampal pyramidal neurons [69, 83, 87, 88, 89, 90]. However, in the dentate gyrus of young deer mice housed in an enriched environment such an effect was not observed [44].

The present data also imply that environmental stimulation has a higher influence on neurons that were less affected by regressive changes. This can also explain the high interindividual differences within the analysed group, as subjects with the lowest values in the EC group are close to subjects with the highest values in SC group. This suggests that if the regression has taken a higher level, an enriched environment has less impact. The data presented here, as our data in dentate gyrus of 4‐month‐old rats and on the occipital cortex in this material (see ‘Material and methods’), also suggest the hypothesis that there is a large ‘real’ dendritic regression in dentate gyrus of 24‐month‐old Sprague‐Dawley rats compared to young adult. It is possible that this rat strain may be susceptible to faster aging [74] as compared to other rat's strains, as for example Norway rats (preliminary data, not shown).

The variability of granular cells morphology is dependent on interindividual and intragroup differences [65, 74, 84, 91]. The slightly increased diversity of neurons we found in EC animals could imply the existence of a neuron specific susceptibility to morphological change in response to environmental manipulation. Individual differences observed in our SC subjects ought to be a consequence of the normal aging process, because such differences have also been reported in human morphological studies [65, 82, 92, 93, 94]. This finding could also be supported with findings from behavioural studies of marked variability in the learning ability of rats [95, 96] that increases during senescence [1, 2, 3, 4, 5, 97, 98, 99]. Interindividual differences in level of regression [74] could possibly be explained not only by the social position of the rat within the group, but by genetic background as well.

The possible mechanisms by which an enriched environment can affect changes in granule neurons dendritic complexity and spine density include: (i) by increasing the expression of neurotrophic factors, such as brain‐derived neurotrophic factor [54, 55, 100, 101] or (ii) by lowering levels of glucocorticiods [30] that are increased in aging and can be responsible for age‐related behavioural impairments and decrease in spine density [102]. Furthermore, the hippocampal formation is the most ‘plastic’ region of the cortex in which granular neurons in the dentate gyrus continue to be generated throughout the lifespan [103, 104, 105, 106, 107], and this process can be modified environmentally [58, 108, 109, 110, 111], even in old age [25, 58, 79, 110, 112, 113, 114, 115].

In conclusion, our results showed that in the rat Sprague‐Dawley species some individuals show marked regression during normal aging. Furthermore our data also showed that branching and spine density of the dentate granule neurons in the aged brain could be modulated by means of a relatively short period (26 days) of cognitive stimulation, except individuals with very high regression. Our finding is compatible with the view that stimulating environment occurring at an appropriate time can help individuals to maintain cognitive function during aging and, moreover, may diminish the deleterious effects of neurodegenerative disorders (such as Alzheimer's disease) on the hippocampus. This demonstrates the importance of cognitive stimulation for arresting regression in the hippocampus during aging, as has been shown for earlier periods during normal development [63, 72].