Stereological Investigation of the Effects of Treadmill Running Exercise on the Hippocampal Neurons in Middle-Aged APP/PS1 Transgenic Mice

Abstract

The risk of cognitive decline during Alzheimer’s disease (AD) can be reduced if physical activity is maintained; however, the specific neural events underlying this beneficial effect are still uncertain. To quantitatively investigate the neural events underlying the effect of running exercise on middle-aged AD subjects, 12-month-old male APP/PS1 mice were randomly assigned to a control group or running group, and age-matched non-transgenic littermates were used as a wild-type group. AD running group mice were subjected to a treadmill running protocol (regular and moderate intensity) for four months. Spatial learning and memory abilities were assessed using the Morris water maze. Hippocampal amyloid plaques were observed using Thioflavin S staining and immunohistochemistry. Hippocampal volume, number of neurons, and number of newborn cells (BrdU+ cells) in the hippocampus were estimated using stereological techniques, and newborn neurons were observed using double-labelling immunofluorescence. Marked neuronal loss in both the CA1 field and dentate gyrus (DG) and deficits in both the neurogenesis and survival of new neurons in the DG of middle-aged APP/PS1 mice were observed. Running exercise could improve the spatial learning and memory abilities, reduce amyloid plaques in the hippocampi, delay neuronal loss, induce neurogenesis, and promote the survival of newborn neurons in the DG of middle-aged APP/PS1 mice. Exercise-induced protection of neurons and adult neurogenesis within the DG might be part of the important structural basis of the improved spatial learning and memory abilities observed in AD mice.

Keywords

neuron newborn neuron running exercise transgenic AD mice

INTRODUCTION

Alzheimer’s disease (AD) is characterized by the progressive decline of memory and cognitive abilities. Accelerated hippocampal atrophy and neuronal loss within the hippocampus have been associated with a loss of memory and cognitive abilities during AD [1 –4]. However, neural stem cells (NSCs) have been identified in the subgranular zone of the adult hippocampus [5, 6]. The presence of stem cells in the adult hippocampus has led to an increased interest in the role of adult neurogenesis in hippocampal function. Previous studies have indicated that plasticity-related changes are observed in the hippocampus with ongoing neurogenesis in AD [7 –11]. Therefore, neurogenesis may be a new candidate target for studies of the neurobiology and treatment of AD [12, 13].

Approximately one-third of AD cases worldwide might be related to potentially modifiable risk factors [14, 15], and the largest proportion of AD cases in the USA, Europe, and UK might be attributable to physical inactivity [15]. Previous studies have reported that the risk of cognitive decline during aging can be reduced if physical activity is maintained [16 –18]. Many clinical and basic studies have suggested that running exercise could reduce cognitive decline in patients with AD and AD mice [19 –23]. However, other studies have shown that running exercise does not improve the spatial learning and memory of AD mice [24, 25]. After a review of these publications, we speculated that these different outcomes might be attributed to variations in the genetic models, age of the mice, intensity and duration of exposure, and type of exercise used in these studies. In a previous study, we have shown that long-term, mild and regular treadmill exercise can delay the decline of learning and memory abilities in the early disease stage in APP/PS1 mice [26]. We next wondered whether exercise exerted protective effects during the late disease stage in APP/PS1 mice. Um et al. and Kang et al. have reported that treadmill running exercise can ameliorate learning and memory ability dysfunction and repress hippocampal neuronal cell death in 24-month-old Tg-NSE/PS2 mice [27, 28]. Furthermore, Mirochnic et al. have reported that voluntary wheel running exercise increases the number of newborn granule cells in the dentate gyrus (DG) of APP23 mice at 18 months [29]. However, very few studies have used the three-dimensional stereological techniques to quantitatively investigate the effects of running exercise on neurons and neurogenesis in the hippocampus during the late disease stage in APP/PS1 mice.

In this study, we investigated the effect of treadmill running exercise on neurons and neurogenesis in the hippocampus during the late disease stage in an AD mouse model, APP/PS1 mice [heterozygous species B6C3-Tg (APPswe, PSEN1de9) 85Dbo/J], which express mutant human APP and PS1 in the brain [30]. The pathological changes in APP/PS1 mice show a number of further similarities with AD, for example the abundant age-dependent severe neuropathology is associated with global brain atrophy and decreased glucose metabolism in the hippocampus [30]. Therefore, the APP/PS1 mice could represent a better model for investigating therapeutic interventions. For this purpose, 12-month-old male APP/PS1 transgenic mice were forced to run for four months (regular and moderate intensity). Then, spatial learning and memory abilities, amyloid plaques in the hippocampus, the volume and neuronal number of the CA1 field, CA2-3 fields and DG and number and density of newborn cells in the DG were investigated.

MATERIALS AND METHODS

Mice and running procedures

All mice were provided by the Animal Model Institute of Nanjing University, and the mice reproduced in the Experimental Animal Center at Chongqing Medical University, P. R. China. All animals were housed and treated in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals (NIH Publication No. 85-23). This study was approved by the Animal Care and Research Committee of Chongqing Medical University, P. R. China.

Twelve-month-old APP/PS1 transgenic male mice were randomly assigned to a control group (AD control, n = 11) or running group (AD runner, n = 11), and age-matched non-transgenic male littermates were used as a wild-type group (WT control, n = 11). All mice were housed in groups of 5 in sedentary conditions; only AD runners were placed on a treadmill to run 20 min per day, 5 days per week, for 4 months [26 –28]. During the first two weeks, the running speed was gradually increased from 5 m/min to 10 m/min. Afterwards, the running speed was maintained at 10 m/min. For the first 7 treatment days, all mice were injected with 50 mg/kg 5-bromo-2^′-deoxyuridine (BrdU) once a day [31].

Morris water maze

The Morris water maze has been widely used to study hippocampal-dependent spatial learning and memory in transgenic mouse models of AD [32]. After four months of treatment, all mice were assessed using the Morris water maze with a hidden platform task and a probe trial task in succession.

For the hidden platform task, the mice were monitored with a camera and trained with four trials per day over 5 days. A circular platform was submerged 1.5 cm beneath the surface of the water, which was made opaque with milk, and was invisible to the mice throughout training. At the beginning of each test, the mice were placed on the platform for 10 s prior to each training trial to reduce stress. The starting points were designated as stationary positions in advance and changed in a pseudorandom order every day. During each trial, the mice were allowed to find and escape onto the submerged platform within 60 s. After each trial, the mice were placed on the platform to rest for 10 s and then placed into a holding cage under a warming lamp until the start of the next trial. In the hidden platform task, the escape latency to reach the platform location was recorded. Twenty-four hours after the last training session on day 5, the platform was removed. The two farthest points from the platform were chosen as the two starting points. The mice were allowed two probe trials and performed a free swim for 60 s per trial. In the probe trial task, swimming speed, length of swimming paths in the target quadrant, total length of swimming paths in the whole pool, the times of platform location crosses and the time spent in the target quadrant during training were recorded [26].

Tissue processing

Five mice were randomly selected from each group. All mice were anesthetized with 1% pentobarbital sodium and then transcardially perfused with normal saline plus heparin and 4% paraformaldehyde. Right or left hemispheres of the mouse brains were chosen at random after the skulls were opened. The hemispheres were dehydrated for 24 h in 10% sucrose, 20% sucrose and 30% sucrose in succession. After dehydration, the hemispheres were embedded in optimal cutting temperature compound (OCT Compound, SAKURA, 4583, USA) and then coronally sliced at 50-μm equidistant intervals with a cryo-ultramicrotome (Leica Microsystems, CM1950, Germany), starting randomly at the rostral pole. All slices, from the rostral to caudal end of the hemisphere, were collected in 96-well plates with phosphate-buffered saline (0.01 M PBS) in succession. For the quantification of the cells of interest in the hippocampus, all slices containing sections of the hippocampus were sampled in a systematic and random fashion according to a stereological sampling method [33, 34]. For the quantification of BrdU-positive cells in the hippocampus, one section was systematically sampled every 6 sections, with the first section randomly sampled from the first 6 sections [33, 34]. Therefore, the section sampling fraction (ssf) was 1/6. On average, 13 sections were sampled per hippocampus.

Analyzing amyloid plaques

A group of tissue sections was randomly selected from six groups of tissue sections per animal. Among them, half of the tissue sections were used for anti-beta amyloid antibody immunohistochemistry staining [35]. The immune reagents were washed and diluted with 0.01 M phosphate buffered solution (PBS) with 0.3% Triton X-100 (PBS-T). The sections were then pretreated with 70% formic acid at room temperature for 20 min for antigen retrieval, followed by two 30-min washes with PBS-T. Next, the sections were pretreated with 0.3% H₂O₂ at room temperature for 20 min to reduce endogenous peroxidase activity, followed by two 15-min washes with PBS-T. The sections were subsequently treated with mixed serum (1.5 ml 5% bovine serum albumin + 1.5 ml 10% goat serum albumin+ 0.5% fetal bovine serum albumin) at 37°C for 2 h to prevent non-specific binding. The samples were then incubated with primary antibodies (mouse monoclonal antibody for beta amyloid, Abcam plc., Product code ab11132) diluted with PBS-T (0.04% final concentration) overnight at 4°C, followed by four 15-min washes with PBS-T. The primary antibodies were then detected using a Streptavidin-Peroxidase Kit (ZSBO-Bio, SP9002), as follows. The sections were incubated at 37°C for 2 h with a goat anti-mouse IgG working solution, followed by four 15-min washes with PBS-T. Then, the sections were incubated at 37°C for 2 h with horseradish peroxidase working solution followed by four 15-min washes with PBS-T. Staining was developed with a diaminobenzidine (DAB) Horseradish Peroxidase Color Development Kit (Beyotime, P0202). After immunostaining, the sections were mounted on gelatin-coated slides, stained with hematoxylin, air-dried, dehydrated with an ethanol gradient, vitrified by dimethylbenzene and mounted with neutral balsam.

Fig.1

A) After Nissl’s staining (toluidine blue), the contours of the field CA1 (yellow box), CA2-3 (red box) and DG (blue box) are traced and the stereological probe (counting points) was superimposed onto the images using stereology analysis software. Bar = 1000μm. B) An illustration of the way to count the number of the neurons with the optical disector technique is shown. At left, the neurons that their nucleoli are clearly in focus in guard zone are not counted. At right, the neurons that their nucleoli are clearly in focus in counting zone but not in focus in guard zone are counted. Arrows show the neurons counted. Bar = 45μm. C) An illustration of the way to count the number of newborn cells with the optical disector technique is shown. At left, the cells that their nucleoli are clearly in focus in guard zone are not counted. At right, the cells that their nucleoli are clearly in focus in counting zone but not in focus in guard zone are counted. Arrow shows the cell counted. Bar = 45μm.

The other half of the tissue sections were used for Thioflavin S staining [23, 36]. The sections were mounted on gelatin-coated slides and air-dried. Next, the sections were washed with 0.01 M PBS at room temperature for 30 min, followed by treatment with 0.3% Thioflavin S (Sigma-Aldrich, T1892) (diluted with 50% ethanol) at room temperature for 8 min in the dark. Then, the sections were washed with 0.01 M PBS in a dark environment for 30 min, followed by treatment with anti-fade solution to reduce fluorescence quenching. After Thioflavin S staining, amyloid plaques per mm² in the hippocampi of the three groups were quantified.

Estimation of hippocampal volume

A group of tissue sections was randomly selected from five additional groups of tissue sections per animal. All sections were used for Nissl staining (toluidine blue). The sections were mounted on gelatin-coated slides and air-dried, followed by an incubation in PBS-T at room temperature for 120 min and treatment with 0.01% toluidine blue at 56°C for 45 min in the dark. Then, the sections were washed with ultra-pure water for 2 min, followed by color separation with 95% ethanol 3 times for 2 min each, vitrification by dimethylbenzene and mounting with neutral balsam.

After Nissl staining, the contours of the hippocampal subregions of each section could be clearly traced and delineated under a light microscope at 4×magnification [37, 38]. A stereological probe (counting points) was superimposed onto the images of the frozen sections taken with stereological analysis software (New CAST, Denmark) (Fig. 1A). The area associated with each counting point was 0.01 mm² at the tissue level. The volumes of the CA1 field, CA2-3 fields and DG were calculated using Cavalieri’ s principle, as follows [33]: $V (CA 1) = t \times a (p) \times \sum P (CA 1)$ (1) $V (CA 2 - 3) = t \times a (p) \times \sum P (CA 2 - 3)$ (2) $V (DG) = t \times a (p) \times \sum P (DG)$ (3) where V(CA1), V(CA2-3) and V(DG) are the volumes of the CA1 field, CA2-3 fields and DG, respectively; t is the distance between two adjacent sampled sections, which is 50μm multiplied by 5 here; a(p) is the area associated with each point in the grid; and Σ P(CA1), Σ P(CA2-3) and Σ P (DG) are the number of points that hit the CA1 field, CA2-3 fields and DG, respectively.

Counting neurons

After Nissl staining, all sections were observed under a modified Olympus BX51 microscope (Olympus, Tokyo, Japan). The Opitiphot microscope was equipped with a motorized stage for precise, automatic movements in the X and Y directions, a video camera to project images onto a computer screen, and a microcator (Heidenhain, USA) attached to the stage for precise measurement of the focal depth (in 0.1μm). First, all sections were observed at 1.25×magnification, and an image of each section was then captured integrally using the super navigator tool in the stereological analysis software (New CAST, Denmark). Second, the contours of the hippocampus subregions were traced and delineated using the super navigator tool at 4×magnification [37, 38]. Third, the counting frame was activated, and the size of the counting frame, height of the guard zone, height of the counting zone and the area sampling fraction (asf) were set at 100×magnification. Fourth, neurons in the CA1 field, CA2-3 fields and DG of each section were counted at 100×magnification (N.A.1.40) according to the parameters that were set in advance [37, 38] (Fig. 1B). During counting, the average thickness of each section after staining was measured. The ratio between the height of the counting zone and the average thickness of the sections represented the thickness sampling fraction (tsf). Finally, the total number of neurons in the CA1 field, CA2-3 fields and DG of each hemisphere were estimated using an optical fractionator [33 , 38] as follows:

N (CA 1) = \sum Q (CA 1) \times \frac{1}{ssf} \times \frac{1}{asf} \times \frac{1}{tsf}

(4)

N (CA 2 - 3) = \sum Q (CA 2 - 3) \times \frac{1}{ssf} \times \frac{1}{asf} \times \frac{1}{taf}

(5)

N (DG) = \sum Q (DG) \times \frac{1}{ssf} \times \frac{1}{asf} \times \frac{1}{tsf}

(6)

where N(CA1), N(CA2-3) and N(DG) are the total numbers of neurons in the CA1 field, CA2-3 fields and DG, respectively; ssf is the section sampling fraction, here ssf is 1/6; asf is the area sampling fraction, here asf is 1%; tsf is the thickness sampling fraction calculated as the height of the counting zone/average thickness of the sections, here the height of the counting zone is 15μm, and the average thickness of sections is 18.2μm, so tsf is 0.82; and Σ Q(CA1), Σ Q(CA2-3) and Σ Q(DG) are the total number of neurons that were counted in the CA1 field, CA2-3 fields and DG of each hemisphere, respectively.

Counting newborn cells

A group of tissue sections was randomly selected from four additional groups of tissue sections per animal. All sections were used for anti-BrdU immunohistochemistry [31]. The immune reagents were washed and diluted with PBS-T. The sections were pretreated with 2 mol/l HCL at room temperature for 20 min for antigen retrieval, followed by two 30-min washes with PBS-T. Next, the sections were pretreated with 0.3% H₂O₂ at room temperature for 20 min to reduce endogenous peroxidase activity, followed by two 15-min washes with PBS-T. The sections were subsequently treated with mixed serum (1.5 ml 5% bovine serum albumin + 1.5 ml 10% rabbit serum albumin+ 0.5% fetal bovine serum albumin) at 37°C for 2 h to prevent non-specific binding. The samples were then incubated with a primary antibody (rat monoclonal antibody for BrdU, Santa Cruz Biotechnology, Inc., Product code sc-56258) diluted with PBS-T (0.05% final concentration) for 96 h at 4°C, followed by four 15-min washes with PBS-T. Primary antibodies were detected with a Strept Avidin Biotin Complex Kit (Boster, SA1025). The sections were then incubated at 37°C for 2 h with biotin-conjugated rabbit anti-rat IgG, followed by four 15-min washes with PBS-T. Then, the sections were incubated at 37°C for 2 h with a chain enzyme avidin-biotin peroxidase complex, followed by four 15-min washes with PBS-T. Staining was developed with a DAB Horseradish Peroxidase Color Development Kit (Beyotime, P0202). After immunostaining, the sections were mounted on gelatin-coated slides, stained with hematoxylin, air-dried, dehydrated in an ethanol gradient, vitrified with dimethylbenzene and mounted with neutral balsam.

After anti-BrdU immunohistochemistry, all sections were observed under an Olympus BX51 microscope (Olympus, Tokyo, Japan). The methods used to count newborn cells in the CA1 field, CA2-3 fields and DG of each section were the same as those used to count neurons (Fig. 1C). However, the set values of related parameters for counting newborn cells and those used for counting neurons differed. The set parameters for counting newborn cells were as follows: the asf was 40%, and the height of the counting zone was 15μm, and the average thickness of the sections was 23.1μm; so, the tsf was 0.65.

Analyzing newborn neurons

A group of tissue sections was randomly selected from three additional groups of tissue sections per animal. All sections were used for anti-BrdU and anti-NeuN immunofluorescence staining [23, 31]. The immune reagents were washed and diluted with PBS-T. The sections were pretreated with 2 mol/l HCL at room temperature for 20 min for antigen retrieval, followed by two 30-min washes with PBS-T. Next, the sections were treated with mixed serum (1.5 ml 10% goat serum albumin + 1.5 ml 10% rabbit serum albumin+ 0.5% fetal bovine serum albumin) at 37°C for 2 h to prevent non-specific binding. The samples were then incubated with primary antibodies (rat monoclonal antibody for BrdU, Santa Cruz Biotechnology, Inc., Product code sc-56258 and rabbit monoclonal antibody for NeuN, a neuronal marker, Abcam, ab177487) diluted with PBS-T (anti-BrdU final concentration: 0.01% and anti-NeuN final concentration: 0.05%) for 96 h at 4°C, followed by four 15-min washes with PBS-T. Primary antibodies were detected with fluorophore-conjugated IgG. The sections were incubated at 37°C for 45 min with rabbit anti-rat IgG DyLight 649 (Abbkine) and goat anti-rabbit IgG DyLight 488 (Abbkine) in the dark, followed by three 10-min washes in the dark with PBS-T. Then, the sections were mounted on gelatin-coated slides with anti-fade solution to reduce fluorescence quenching. After immunofluorescence staining, all sections were observed under a laser scanning confocal microscope.

Statistics

Fig.2

A) The mean escape latency of the WT control, AD control and AD runner mice during the hidden platform task. The error bars indicate SEM. ^** p < 0.01 versus WT control. ^# # p < 0.01 versus AD runner. B) The mean swimming speed in the removed platform task. The error bars indicate SD. C) The swimming paths in the target quadrant in the probe trial task. The error bars indicate SD. ^* p < 0.05 versus WT control. ^# p < 0.05 versus AD runner. D). The time spending in the target quadrant in the probe trial task. The error bars indicate SD. ^** p < 0.01 versus WT control. ^# p < 0.05 versus AD runner. E) The times crossing the platform location in the probe trial task. The error bars indicate SD. ^** p < 0.01 versus WT control.

The hidden platform task results are presented as the mean±SEM, and other results are presented as the mean (SD). Statistical analyses were performed using SPSS (ver. 19.0, SPSS Inc., Chicago, USA). The Shapiro-Wilk test indicated that the group means of the data from the Morris water test and the stereological data were normally distributed. Repeated-measures analysis of variance (ANOVA) was used to compare the data from the hidden platform task among the three groups. A one-way ANOVA was used to compare the data from the probe trial task and stereological data among the three groups. Following the ANOVA, Fisher’s least significant difference (LSD) test was used to detect significant differences. A p-value <0.05 was considered statistically significant. Observed coefficient of variation (OCV) values and observed coefficient of error (OCE) values were calculated according to Gundersen et al. [34].

RESULTS

Spatial learning and memory ability

Fig.3

A) Left pictures show the amyloid plaques of hippocampus detected with Thioflavine S staining in the WT control mice, the AD control mice and the AD runner mice. Bar = 300μm. Right pictures show the amyloid plaques stained using anti-beta Amyloid antibody in the WT control mice, the AD control mice and the AD runner mice. Bar = 45μm. Arrows show the amyloid plaques of hippocampus. B) The number of amyloid plaques per mm² in the hippocampus of the AD control mice and the AD runner mice. The error bars indicate SD. ^# p < 0.05 versus AD runner.

During the hidden platform task, there were significant differences in escape latency among WT control, AD control and AD runner mice (F = 9.643; p < 0.001). The escape latency of AD control mice was not only longer than that of WT control mice but also longer than that of AD runner mice (p < 0.001, p = 0.002; Fig. 2A).

In the removed platform task, there were no significant differences in swimming speed among WT control, AD control and AD runner mice (F = 0.102; p = 0.903; Fig. 2B), but there were significant differences in swimming paths in the target quadrant among the three groups (F = 3.439; p = 0.030; Fig. 2C). The swimming path in the target quadrant of AD control mice was not only shorter than that of WT control mice but also shorter than that of AD runner mice (p = 0.028, p = 0.022; Fig. 2C). In addition, there were significant differences in time spent in the target quadrant among WT control, AD control and AD runner mice (F = 6.247; p = 0.004; Fig. 2D). Compared with WT control and AD runner mice, AD control mice spent significantly less time in the target quadrant (p = 0.004, p = 0.005, respectively; Fig. 2D). The times of platform location crosses was significantly different among WT control mice, AD control mice and AD runner mice (F = 5.5705; p = 0.006; Fig. 2E). AD control mice crossed the platform location significantly less times than WT control mice, but there was no significant difference between AD control mice and AD runner mice (p = 0.002, p = 0.050, respectively; Fig. 2E).

Amyloid plaques

As shown in Fig. 3A, no amyloid plaques were observed in the hippocampi of the WT control mice. In contrast, many large amyloid plaques were observed in the hippocampi of AD control and AD runner mice. There were fewer amyloid plaques in the hippocampi of AD runner mice than those of AD control mice (p = 0.017; Fig. 3B).

Hippocampal volume

For the WT control group, 465 points hitting CA1 field, 271 points hitting CA2-3 fields and 263 points hitting DG were counted per mouse hemisphere. For the AD control group, 465 points hitting CA1 field, 265 points hitting CA2-3 fields and 258 points hitting DG were counted per mouse hemisphere. For the AD runner group, 489 points hitting CA1 field, 257 points hitting CA2-3 fields and 267 points hitting DG were counted per mouse hemisphere. In current study, the observed variance of the individual estimate (OCE²) was less than half of the observed inter-individual variance (OCV²), which indicated that sampling was considered optimal (Table 1).

Table 1

Stereological results for hippocampal volume of the WT control, AD control and AD runner groups

	WT control group			AD control group			AD runner group
	CA1	CA2-3	DG	CA1	CA2-3	DG	CA1	CA2-3	DG
Mean (mm³)	8.94	4.90	4.10	8.58	4.86	4.14	9.09	4.65	4.72
SD	0.68	1.10	0.76	0.60	0.50	0.34	0.45	0.99	0.48
OCV	0.076	0.225	0.186	0.069	0.102	0.082	0.047	0.213	0.101
OCE	0.019	0.022	0.020	0.023	0.024	0.024	0.021	0.025	0.016
OCE²/OCV²	0.063	0.010	0.012	0.113	0.056	0.087	0.191	0.014	0.026

^*The mean volume of hippocampal sub-regions (Mean), standard deviation (SD), observed coefficient of variation (OCV) and observed coefficient of error (OCE) are provided.

After four months of treadmill running exercise, there were no significant differences in the volumes of the CA1 field, CA2-3 fields and DG between the groups (F = 1.025, p = 0.388; F = 0.106, p = 0.900; F = 1.987, p = 0.18, respectively; Table 1).

The number of neurons

For the WT control group, 648 neurons in CA1 field, 371 neurons in CA2-3 fields and 652 neurons in DG were counted per mouse hemisphere. For the AD control group, 476 neurons in CA1 field, 304 neurons in CA2-3 fields and 471 neurons in DG were counted per mouse hemisphere. For the AD runner group, 521 neurons in CA1 field, 329 neurons in CA2-3 fields and 608 neurons in DG were counted per mouse hemisphere. In current study, the observed variance of the individual estimate (OCE²) was less than half of the observed inter-individual variance (OCV²), which indicated that sampling was considered optimal (Table 2).

Table 2

Stereological results for hippocampal neurons of the WT control, AD control and AD runner groups

	WT control group			AD control group			AD runner group
	CA1	CA2-3	DG	CA1	CA2-3	DG	CA1	CA2-3	DG
Mean (×10⁴)	44.13	25.54	41.49	34.52	24.09	27.50	39.07	26.58	44.28
SD	3.61	7.30	13.63	5.70	8.74	8.79	8.80	7.13	6.65
OCV	0.082	0.286	0.328	0.165	0.363	0.240	0.320	0.268	0.150
OCE	0.048	0.065	0.052	0.054	0.065	0.070	0.064	0.060	0.047
OCE²/OCV²	0.347	0.051	0.025	0.107	0.032	0.084	0.040	0.051	0.100

^*The mean number of hippocampal neurons (Mean), standard deviation (SD), observed coefficient of variation (OCV), and observed coefficient of error (OCE) are provided.

Fig.4

A) The total number of neurons in field CA1 of the WT control, AD control and AD runner mice. B) The total number of neurons in field CA2-3 of the WT control, AD control and AD runner mice. C) The total number of neurons in the DG of the WT control, AD control and AD runner mice. The error bars indicate SD. ^* p < 0.05 versus WT control. ^# # p < 0.01 versus AD runner.

After four months of treadmill running exercise, there were significant differences in the numbers of neurons in the CA1 field, and DG among WT control, AD control and AD runner mice (F = 4.066, p = 0.045; F = 5.688, p = 0.018, respectively; Fig. 4A, C), but there were no significant differences in the numbers of neurons in the CA2-3 fields among the three groups (F = 0.193; p = 0.827; Fig. 4B). The number of neurons in the CA1 field was significantly decreased in AD control mice compared with that in WT control mice (p = 0.015), but there were no significant differences between AD control and AD runner mice (p = 0.202; Fig. 4A). There were no significant differences in the number of neurons in the CA2-3 fields between WT control and AD control mice or between AD control and AD runner mice (p = 0.728, p = 0.547; Fig. 4B). The number of neurons in the DG was significantly decreased in AD control mice compared with that in WT control mice (p = 0.022; Fig. 4C); conversely, the number of neurons in the DG was significantly increased in AD runner mice compared with that in AD control mice (p = 0.008; Fig. 4C).

Newborn cells and newborn neurons

For the WT control group, 218 newborn cells (BrdU⁺ cells) in DG were counted per mouse hemisphere. For the AD control group, 163 newborn cells in DG were counted per mouse hemisphere. For the AD runner group, 222 newborn cells in DG were counted per mouse hemisphere. In current study, the observed variance of the individual estimate (OCE²) was less than half of the observed inter-individual variance (OCV²), which indicated that sampling was considered optimal (Table 3).

Table 3

Stereological results for newborn cells of the WT control, AD control, and AD runner groups

	WT control group	AD control group	AD runner group
Mean (×10³)	10.15	7.49	10.15
SD	1.30	1.15	1.04
OCV	0.128	0.154	0.103
OCE	0.068	0.080	0.068
OCE²/OCV²	0.282	0.268	0.437

^*The mean number of newborn cells in the DG (Mean), standard deviation (SD), observed coefficient of variation (OCV) and observed coefficient of error (OCE) are provided.

After four months of treadmill running exercise, there were significant differences in the number of newborn cells (BrdU⁺ cells) and density of newborn neurons (BrdU⁺/NeuN⁺ cells) per unit area in the DG among WT control, AD control and AD runner mice (F = 8.662, p = 0.005; F = 3.754, p = 0.031, respectively; Fig. 5). The number of newborn cells and density of newborn neurons per unit area in the DG were significantly decreased in AD control mice compared with those in WT control mice (p = 0.004, p = 0.037, respectively; Fig. 5). Conversely, the number of newborn cells and density of newborn neurons per unit area in the DG were significantly increased in AD runner mice compared with those in AD control mice (p = 0.004, p = 0.001, respectively; Fig. 5).

Fig.5

A) The newborn neurons (Brdu⁺/Neun⁺ cells) in the DG of the WT control, AD control and AD runner mice. Bar = 60μm. B) The total number of newborn cells (Brdu⁺ cells) in the DG of the WT control, AD control and AD runner mice. The error bars indicate SD. ^** p < 0.01 versus WT control. ^# # p < 0.01 versus AD runner. C) The number of newborn neurons (Brdu⁺/Neun⁺ cells) pur unit area in the DG of the WT control, AD control and AD runner mice. The error bars indicate SD. ^* p < 0.05 versus WT control. ^# # p < 0.01 versus AD runner.

DISCUSSION

Previous studies have shown spatial learning and memory dysfunction in APP/PS1 mice at 7–10 months of age [39 , 41]. In the current study, we subjected 12-month-old APP/PS1 mice with spatial learning and memory impairments to a mild, regular treadmill running protocol. The mice ran on a treadmill at a speed of 10 m/min, 20 min/day, 5 days/week for 4 months. After four months of running exercise, APP/PS1 mice performed significantly better on the hidden platform task and probe trial task of the Morris water maze than APP/PS1 mice not subjected to exercise. The results indicate that running exercise can reduce learning and memory impairments in middle-aged APP/PS1 mice. The current results are consistent with most previous findings showing that treadmill running exercise improves cognitive and brain functions in aged AD mice [27, 28]. In a study by Um et al., 24-month-old Tg-NSE/PS2 mice were subjected to treadmill running exercise at a speed of 12 m/min, 60 min/day, 5 days/week, for 3 months, and exercise enhanced learning and memory abilities in the Morris water maze test [27]. In our previous study, we subjected 6-month-old APP/PS1 mice with intact spatial learning and memory to the same running paradigm and observed that running can delay the decline of learning and memory abilities during the early disease stage in APP/PS1 mice [26]. The current results and these previous findings indicate that treadmill running exercise not only delays the progression of spatial learning and memory decline during the early stage of AD in mice but also improves spatial learning and memory dysfunctions during the late stage of AD in mice. Therefore, regular and moderate intensity running exercise might serve as a potential preventative measure and treatment for AD.

Spatial learning and memory abilities are closely associated with the hippocampus. Accelerated hippocampal atrophy has been associated with the loss of memory and cognitive abilities during AD, and hippocampal atrophy has been considered an early change in AD [4]. Surprisingly, we did not find significant differences in the volume of hippocampal subregions between 16-month-old APP/PS1 and WT mice. Similarly, Schmitz et al. have also shown no significant differences in hippocampal volumes between 16-month-old APP/PS1 and WT mice using a stereological method [42]. van de Pol et al. have also reported no significant differences in hippocampal volume between healthy elderly subjects and elderly patients with AD [43]. Previous studies have shown that hippocampal atrophy also occurs in normal aging brains [44, 45]. Thus, we speculate that the lack of significant differences in hippocampal volume between 16-month-old WT and AD mice might be due to age-related hippocampal atrophy, but the exact cause needs to be further investigated.

Evidence has shown that exercise delays hippocampal atrophy, which could explain how exercise protects against learning and memory loss in AD [36, 46]. In our previous study, we also found that four months of exercise can delay DG atrophy in 6-month-old APP/PS1 mice [26]. However, in the present study, running exercise did not affect the volume of hippocampal subregions in middle-aged APP/PS1 mice. We speculated that exercise could protect against hippocampal atrophy in early AD mice but could not reverse the reduced hippocampal volume in middle-aged AD mice. As we expected, running exercise had no protective effect on hippocampal volume in middle-aged AD mice.

Next, we explored the structural basis for the protective effect of running exercise on spatial learning and memory abilities in middle-aged AD mice. Amyloid plaques and neurofibrillary tangles are the characteristic pathological changes in AD [47, 48]. Evidence has indicated that exercise can reduce amyloid load in AD [36, 49]. In addition, previous studies have suggested that exercise not only reduces amyloid load but also decreases tau phosphorylation near amyloid plaques [23]. In this study, we found that exercise can reduce amyloid plaques in the hippocampus of middle-aged AD mice. Our results further suggest that exercise could alleviate the characteristic pathological changes of AD.

Previous reports have suggested that the interaction between Aβ oligomers and tau proteins might cause neuronal cell death in the brains of those with AD [50 –52]. In the current study, we quantified neuronal numbers in a three-dimensional space within the hippocampus using a stereological method and found that 16-month-old APP/PS1 mice had fewer neurons in the hippocampus than WT mice of the same age, especially in the CA1 field and DG. To determine whether running exercise had effects on neurons in the hippocampus of AD subjects, Tapia-Rojas et al. used a semi-quantitative method and found that running exercise reduced neuronal cell loss in the DG but had no effects on neurons in the CA1 field in young APP/PS1 mice [23]. In the present study, we observed that running exercise reduces neuronal cell loss in the DG but has no effect on neurons in the CA1 field of middle-aged APP/PS1 mice (Fig. 4). The present results suggest that running exercise protects neurons in the DG of AD mice.

We also studied the potential underlying mechanisms of the protective effects of running exercise on neurons in the DG of AD mice. NSCs are located in the subgranular zone of the DG [5, 12]. Adult neurogenesis can be divided into the four following phases: precursor cell phase, early survival phase, postmitotic maturation phase, and late survival phase [53, 54]. Perry et al. have suggested that neurogenesis abnormalities in AD would differ based on these phases and areas of neurogenesis and stages of AD [55]. Mirochnic et al. have reported that voluntary running exercise increases the number of newborn granule cells in the DG of APP23 mice at 18 months but not at 6 months [29]. The effects of running exercise on neurogenesis in AD mice might be associated with the phases and areas of neurogenesis and stages of AD. Moreover, Mu and Gage have summarized the neurogenic changes in transgenic mouse models of AD; they believe that neurogenesis deficits vary among different AD mouse models [12]. In the early disease stage, APP23 mice exhibit increased neuronal proliferation and differentiation, but the newborn neurons fail to mature [56 , 58]. In contrast, APP/PS1 mice show decreased neuronal proliferation, differentiation and survival [59, 60]. In the current study, we injected BrdU for the first 7 days of treatment, which labels actively proliferating nonradial NSCs, and tracked NSC proliferation, migration, and differentiation for four months. The number of BrdU+ cells could reflect the proliferation of nonradial NSCs and survival of newborn cells. Therefore, we quantified BrdU⁺ cells in the DG using a stereological method and found that 16-month-old APP/PS1 mice had fewer BrdU⁺ cells in the DG than WT mice of the same age. In addition, we also found that 16-month-old APP/PS1 mice subjected to exercise had more BrdU+ cells in the DG than APP/PS1 not subjected to exercise. However, actively proliferating nonradial NSCs can generate both glial cells and neuroblasts [12]. Thus, these BrdU⁺ cells could be newborn neurons or glial cells. To address this issue in the present study, we identified newborn neurons with both anti-BrdU and anti-NeuN antibodies (a mature neuronal marker). Previous studies have suggested that granule cells in the DG are the only type of neurons generated in the adult hippocampus [54]. On the other hand, Rietze et al. have reported that most new neurons are located in the granule cell layer but have also detected new neurons in the hilus [61]. In current study, we observed new neurons in the granule cell layer and hilus of the DG in all mice. Rietze et al. have reported that the number of newborn neurons declines as survival time increases from the 3rd to the 24th week after BrdU injections; meanwhile, the number of newborn glial cells remains constant in the DG over time [61]. In the present study, after BrdU injections and the 4-month treatment period (16 weeks), the lower numbers of new neurons labeled by anti-BrdU and anti-NeuN antibodies (the mature neuron marker) in the granule cell layer might be due to the decreased number of newborn neurons as survival time increases. We observed more newborn neurons in the DG of WT and exercising APP/PS1 mice than in that of APP/PS1 mice that did not exercise. These results might be attributed to the following two factors: 1) treadmill running might delay the decline of the newborn neurons in the DG of middle-aged APP/PS1 mice, and 2) treadmill running might increase the proliferation of nonradial NSCs in the DG of middle-aged APP/PS1 mice. In addition, Tapia-Rojas et al. subjected 7-month-old APP/PS1 mice to running exercise, with BrdU labeling during the last 3 days of treatment and the double-labeling of newborn neurons using anti-BrdU and anti-doublecortin antibodies (an immature neuronal marker) [23]. The authors observed that voluntary running exercise increases the number of new neurons in the granule cell layer in APP/PS1 mice. We believe that these data indicate that moderate treadmill running exercise might improve neurogenesis deficits in middle-aged APP/PS1 mice. Kempermann et al. have observed that newborn dentate granule cells gradually develop elaborate dendritic trees in the molecular layer to receive inputs from the entorhinal cortex and project to CA3 pyramidal neurons and the interneurons in the hilus [54]. Treadmill running exercise could delay the decline in newborn neurons, which might underlie the exercise-induced improvement in spatial learning and memory ability.

Conclusion

Moderate treadmill running exercise has a positive effect on the learning and memory ability of aged APP/PS1 mice. Moreover, the present results suggest that running exercise can reduce neuronal loss in the DG by enhancing the proliferation, differentiation or survival of new neurons in AD, which might provide a structural basis for the exercise-induced improvement in the spatial learning and memory abilities of aged APP/PS1 mice.

Footnotes

ACKNOWLEDGMENTS

This study was supported by the National Natural Science Foundation of China (NSFC, 81671259; 81501101); Research Foundation for 100 Academic and Discipline Talented Leaders of Chongqing, P. R. China; The Foundation for “the Excellent Young Scholars of Chongqing Medical University” in 2015 (CYYQ201509); The 2016 Supporting Excellent Ph.D Projects of Chongqing Medical University; The 2016 Supporting the Innovative Projects of Graduate Students of Chongqing.

Authors’ disclosures available online ().

References

West

, Coleman

, Flood

, Troncoso

(1994) Differences in the pattern of hippocampal neuronal loss in normal ageing and Alzheimer’s disease. Lancet 344, 769–772.

Price

, Ko

, Wade

, Tsou

, McKeel

, Morris

(2001) Neuron number in the entorhinal cortex and CA1 in preclinical Alzheimer disease. Arch Neurol 58, 1395–1402.

Andrade-Moraes

, Oliveira-Pinto

, Castro-Fonseca

, da Silva

, Guimarães

, Szczupak

, Parente-Bruno

, Carvalho

, Polichiso

, Gomes

, Oliveira

, Rodriguez

, Leite

, Ferretti-Rebustini

, Jacob-Filho

, Pasqualucci

, Grinberg

, Lent

(2013) Cell number changes in Alzheimer’s disease relate to dementia, not to plaques and tangles. Brain 136(Pt 12), 3738–3752.

Barnes

, Fox

(2014) The search for early markers of AD: Hippocampal atrophy and memory deficits. Int Psychogeriatr 26, 1065–1066.

Eriksson

, Perfilieva

, Björk-Eriksson

, Alborn

, Nordborg

, Peterson

, Gage

(1998) Neurogenesis in the adult human hippocampus. Nat Med 4, 1313–1317.

Gage

(2000) Mammalian neural stem cells. Science 287, 1433–1438.

Brinton

, Wang

(2006) Therapeutic potential of neurogenesis for prevention and recovery from Alzheimer’s disease: Allopregnanolone as a proof of concept neurogenic agent. Curr Alzheimer Res 3, 185–190.

Ziabreva

, Perry

, Minger

, Ekonomou

, Przyborski

, Ballard

(2006) Altered neurogenesis in Alzheimer’s disease. J Psychosom Res 61, 311–316.

, Yamamori

, Tatebayashi

, Shafit-Zagardo

, Tanimukai

, Chen

, Iqbal

, Grundke-Iqbal

(2008) Failure of neuronal maturation in Alzheimer disease dentate gyrus. J Neuropathol Exp Neurol 67, 78–84.

10.

Taupin

(2009) Adult neurogenesis, neural stem cells and Alzheimer’s disease: Developments, limitations, problems and promises. Curr Alzheimer Res 6, 461–470.

11.

Adamowicz

, Mertens

, Gage

(2015) Alzheimer’s disease: Distinct stages in neurogenic decline?. Biol Psychiatry 77, 680–682.

12.

, Gage

(2011) Adult hippocampal neurogenesis and its role in Alzheimer’s disease. Mol Neurodegener 6, 85.

13.

Fitzsimons

, van Bodegraven

, Schouten

, Lardenoije

, Kompotis

, Kenis

, van den Hurk

, Boks

, Biojone

, Joca

, Steinbusch

, Lunnon

, Mastroeni

, Mill

, Lucassen

, Coleman

, van den Hove

, Rutten

(2014) Epigenetic regulation of adult neural stem cells: Implications for Alzheimer’s disease. Mol Neurodegener 9, 25.

14.

Daviglus

, Bell

, Berrettini

, Bowen

, Connolly

Jr , Cox

, Dunbar-Jacob

, Granieri

, Hunt

, McGarry

, Patel

, Potosky

, Sanders-Bush

, Silberberg

, Trevisan

(2010) National institutes of health state-of-the-science conference statement: Preventing Alzheimer disease and cognitive decline. Ann Intern Med 153, 176–181.

15.

Norton

, Matthews

, Barnes

, Yaffe

, Brayne

(2014) Potential for primary prevention of Alzheimer’s disease: An analysis of population-based data. Lancet Neurol 13, 788–794.

16.

Laurin

, Verreault

, Lindsay

, MacPherson

, Rockwood

(2001) Physical activity and risk of cognitive impairment and dementia in elderly persons. Arch Neurol 58, 498–504.

17.

Lautenschlager

, Cox

, Flicker

, Foster

, van Bockxmeer

, Xiao

, Greenop

, Almeida

(2008) Effect of physical activity on cognitive function in older adults at risk for Alzheimer disease: A randomized trial. JAMA 300, 1027–1037.

18.

Makizako

, Liu-Ambrose

, Shimada

, Doi

, Park

, Tsutsumimoto

, Uemura

, Suzuki

(2015) Moderate-intensity physical activity, hippocampal volume, and memory in older adults with mild cognitive impairment. J Gerontol A Biol Sci Med Sci 70, 480–486>.

19.

Teri

, Gibbons

, McCurry

, Logsdon

, Buchner

, Barlow

, Kukull

, LaCroix

, McCormick

, Larson

(2003) Exercise plus behavioral management in patients with Alzheimer disease: A randomized controlled trial. JAMA 290, 2015–2022.

20.

Cotman

, Berchtold

(2007) Physical activity and the maintenance of cognition: Learning from animal models. Alzheimers Dement 3(2 Suppl), S30–37.

21.

Nichol

, Deeny

, Seif

, Camaclang

, Cotman

(2009) Exercise improves cognition and hippocampal plasticity in APOE epsilon4 mice. Alzheimers Dement 5, 287–294.

22.

Hirsch

, ACP Journal

Club

(2013) A 12-month, in-home exercise program delayed functional deterioration in Alzheimer disease.JC. Ann Intern Med 159, 10.

23.

Tapia-Rojas

, Aranguiz

, Varela-Nallar

, Inestrosa

(2016) Voluntary running attenuates memory loss, decreases neuropathological changes and induces neurogenesis in a mouse model of Alzheimer’s disease. Brain Pathol 26, 62–74.

24.

Wolf

, Kronenberg

, Lehmann

, Blankenship

, Overall

, Staufenbiel

, Kempermann

(2006) Cognitive and physical activity differently modulate disease progression in the amyloid precursor protein (APP)-23 model of Alzheimer’s disease. Biol Psychiatry 60, 1314–1323.

25.

Richter

, Ambrée

, Lewejohann

, Herring

, Keyvani

, Paulus

, Palme

, Touma

, Schäbitz

, Sachser

(2008) Wheel-running in a transgenic mouse model of Alzheimer’s disease: Protection or symptom?. Behav Brain Res 190, 74–84.

26.

Chao

, Zhang

, Luo

, Xiao

, Lv

, He

, Zhou

, Zhang

, Jiang

, Gu

, Tang

(2015) Running exercise reduces myelinated fiber loss in the dentate gyrus of the hippocampus in APP/PS1 transgenic mice. Curr Alzheimer Res 12, 377–383.

27.

, Kang

, Koo

, Kim

, Jin-Lee, Kim

, Yang

, An

, Cho

(2011) Treadmill exercise represses neuronal cell death in an aged transgenic mouse model of Alzheimer’s disease. Neurosci Res 69, 161–173.

28.

Kang

, Kwon

, Koo

, Kim

, Lee

, Yang

, Lee

, Cho

(2013) Treadmill exercise represses neuronal cell death and inflammation during Aβ-induced ER stress by regulating unfolded protein response in aged presenilin 2 mutant mice. Apoptosis 18, 1332–1347.

29.

Mirochnic

, Wolf

, Staufenbiel

, Kempermann

(2009) Age effects on the regulation of adult hippocampal neurogenesis by physical activity and environmental enrichment in the APP23 mouse model of Alzheimer disease. Hippocampus 19, 1008–1018.

30.

Bilkei-Gorzo

(2014) Genetic mouse models of brain ageing and Alzheimer’s disease. Ther 142, 244–257.

31.

van Praag

, Shubert

, Zhao

, Gage

(2005) Exercise enhances learning and hippocampal neurogenesis in aged mice. J Neurosci 25, 8680–8685.

32.

Chen

, Chen

, Knox

, Inglis

, Bernard

, Martin

, Justice

, McConlogue

, Games

, Freedman

, Morris

(2000) A learning deficit related to age and beta-amyloid plaques in a mouse model of Alzheimer’ s disease. Nature 408, 975–979.

33.

Gundersen

, Bendtsen

, Korbo

, Marcussen

, Møller

, Nielsen

, Nyengaard

, Pakkenberg

, Sørensen

(1988) Some new, simple and efficient stereological methods and their use in pathological research and diagnosis. APMIS 96, 379–394.

34.

Gundersen

, Jensen

, Kiêu

, Nielsen

(1999) The efficiency of systematic sampling in stereology–reconsidered. J Microsc 193(Pt 3), 199–211.

35.

Bomba

, Ciavardelli

, Silvestri

, Canzoniero

, Lattanzio

, Chiappini

, Piantelli

, Di Ilio

, Consoli

, Sensi

(2013) Exenatide promotes cognitive enhancement and positive brain metabolic changes in PS1-KI mice but has no effects in 3xTg-AD animals. Cell Death Dis 4, e612.

36.

Yuede

, Zimmerman

, Dong

, Kling

, Bero

, Holtzman

, Timson

, Csernansky

(2009) Effects of voluntary and forced exercise on plaque deposition, hippocampal volume, and behavior in the Tg2576 mouse model of Alzheimer’s disease. Neurobiol Dis 35, 426–432.

37.

Hao

, Janssen

WGM

, Tang

, Roberts

, Mckay

, Lasley

, Allen

, Greengard

, Rapp

, Kordower

, Hof

, Morrison

(2003) Estrogen increases number of spinophilin-immunoreactive spines in the hippocampus of young and aged female rhesus monkeys. J Comp Neurol 465, 540–550.

38.

Zhao

, Shi

, Zhang

, Wu

, Chao

, Huang

, Gao

, Qiu

, Chen

, Lu

, Tang

(2013) Enriched environment induces higher CNPase positive cells in aged rat hippocampus. Neurosci Lett 555, 177–181.

39.

Garcia-Alloza

, Robbins

, Zhang-Nunes

, Purcell

, Betensky

, Raju

, Prada

, Greenberg

, Bacskai

, Frosch

(2006) Characterization of amyloid deposition in the APPswe/PS1dE9 mouse model of Alzheimer disease. Neurobiol Dis 24, 516–524.

40.

Reiserer

, Harrison

, Syverud

, McDonald

(2007) Impaired spatial learning in the APPSwe+ PSEN1DeltaE9 bigenic mouse model of Alzheimer’s disease. Genes Brain Behav 6, 54–65.

41.

Toledo

, Inestrosa

(2010) Activation of Wnt signaling by lithium and rosiglitazone reduced spatial memory impairment and neurodegeneration in brains of an APPswe/PSEN1DeltaE9 mouse model of Alzheimer’s disease. Mol Psychiatry 15, 272–285.

42.

Schmitz

, Rutten

, Pielen

, Schäfer

, Wirths

, Tremp

, Czech

, Blanchard

, Multhaup

, Rezaie

, Korr

, Steinbusch

, Pradier

, Bayer

(2004) Hippocampal neuron loss exceeds amyloid plaque load in a transgenic mouse model of Alzheimer’s disease. Am J Pathol 164, 1495–1502.

43.

van de Pol

, Hensel

, Barkhof

, Gertz

, Scheltens

, van der Flier

(2006) Hippocampal atrophy in Alzheimer disease: Age matters. Neurology 66, 236–238.

44.

Small

, Chawla

, Buonocore

, Rapp

, Barnes

(2004) Imaging correlates of brain function in monkeys and rats isolates a hippocampal subregion differentially vulnerable to aging. Proc Natl Acad Sci U S A 101, 7181–7186.

45.

Rutten

, Brasnjevic

, Steinbusch

, Schmitz

(2010) Caloric restriction and aging but not overexpression of SOD1 affect hippocampal volumes in mice. Mech Ageing Dev 131, 574–579.

46.

Smith

, Nielson

, Woodard

, Seidenberg

, Durgerian

, Hazlett

, Figueroa

, Kandah

, Kay

, Matthews

, Rao

(2014) Physical activity reduces hippocampal atrophy in elders at genetic risk for Alzheimer’s disease. Front Aging Neurosci 6, 61.

47.

Goedert

, Spillantini

(2006) A century of Alzheimer’s disease. Science 314, 777–781.

48.

Mandelkow

, Mandelkow

(2012) Biochemistry and cell biology of tau protein in neurofibrillary degeneration. Cold Spring Harb Perspect Med 2, a006247.

49.

Adlard

, Perreau

, Pop

, Cotman

(2005) Voluntary exercise decreases amyloid load in a transgenic model of Alzheimer’s disease. J Neurosci 25, 4217–4221.

50.

Crews

, Masliah

(2010) Molecular mechanisms of neurodegeneration in Alzheimer’s disease. Hum Mol Genet 19, R12–20.

51.

Ittner

, Gotz

(2011) Amyloid-beta and tau–a toxic pas de deux in Alzheimer’s disease. Nat Rev Neurosci 12, 65–72.

52.

Morris

, Maeda

, Vossel

, Mucke

(2011) The many faces of tau. Neuron 70, 410–426.

53.

Steiner

, Klempin

, Wang

, Kott

, Kettenmann

, Kempermann

(2006) Type-2 cells as link between glial and neuronal lineage in adult hippocampal neurogenesis. Glia 54, 805–814.

54.

Kempermann

, Song

, Gage

(2015) Neurogenesis in the adult hippocampus. Cold Spring Harb Perspect Med 5, a018812.

55.

Perry

, Johnson

, Ekonomou

, Perry

, Ballard

, Attems

(2012) Neurogenic abnormalities in Alzheimer’s disease differ between stages of neurogenesis and are partly related to cholinergic pathology. Neurobiol Dis 47, 155–162.

56.

Jin

, Peel

, Mao

, Xie

, Cottrell

, Henshall

, Greenberg

(2004) Increased hippocampal neurogenesis in Alzheimer’s disease. Proc Natl Acad Sci U S A 101, 343–347.

57.

Lopez-Toledano

, Shelanski

(2007) Increased neurogenesis in young transgenic mice overexpressing human APP(Sw, Ind). J Alzheimers Dis 12, 229–240.

58.

Sun

, Halabisky

, Zhou

, Palop

, Yu

, Mucke

, Gan

(2009) Imbalance between GABAergic and glutamatergic transmission impairs adult neurogenesis in an animal model of Alzheimer’s disease. Cell Stem Cell 5, 624–633.

59.

Verret

, Jankowsky

, Xu

, Borchelt

, Rampon

(2007) Alzheimer’s-type amyloidosis in transgenic mice impairs survival of newborn neurons derived from adult hippocampal neurogenesis. J Neurosci 27, 6771–6780.

60.

Demars

, Hu

, Gadadhar

, Lazarov

(2010) Impaired neurogenesis is an early event in the etiology of familial Alzheimer’s disease in transgenic mice. J Neurosci Res 88, 2103–2117.

61.

Rietze

, Poulin

, Weiss

(2000) Mitotically active cells that generate neurons and astrocytes are present in multiple regions of the adult mouse hippocampus. J Comp Neurol 424, 397–408.