Background
Alzheimer disease (AD), the most common form of dementia, is a progressive neurodegenerative disease affecting elderly populations worldwide. AD is characterized by extra- and intra-cellular amyloid-beta (Aβ) plaques deposition and formation of neurofibrillary tangles (NFT) inside the neurons, synaptic loss, and severe cognitive decline [
1‐
3]. Several non-genetic risk factors such as diabetes, obesity, hypertension, brain injuries, depression, or physical inactivity are associated with increased risk of AD worldwide [
4‐
7]. Available pharmacological therapies provide only brief symptomatic relief [
8]. Several epidemiological and clinical studies revealed that education, occupation, and physical activity can improve cognitive ability in healthy older people and provide protection against the development and progression of AD [
6,
9‐
13].
Several epidemiological studies have already reported that physical activity significantly reduced the risk of dementia [
14‐
16]. For example, people involved in mentally challenging activity and physical exercise during young and middle adulthood have four times less chance of getting AD compared with control subjects not engaging in such activities [
11]. Additionally, people participating in leisure-time physical activity during middle age for at least twice a week showed reduced risk of dementia and AD, when compared with people not exercising at this age [
17,
18]. Furthermore, several experimental studies have reported that physical activity prevents AD progression and improved cognitive functions by reducing Aβ pathology and amyloid angiopathy [
19‐
23].
Mice exposed to social, physical, and cognitive training show a protective effect against cognitive impairment, decreased brain Aβ burden, and enhanced hippocampal synaptic immunoreactivity [
24,
25]. Additionally, physical exercise in isolation has been shown to reduce AD pathology and improve memory in various murine models of AD [
26,
27].
These and other animal studies show the potential benefits of physical exercise in reducing AD pathology and associated cognitive impairments. However, there are some caveats associated with this work including issues surrounding how accurately the mouse models of AD used in these studies mimic the brain changes in human AD. A second issue is the limited range of AD pathology assessed in these experiments.
The first issue, as noted, is that most mouse models of AD have overexpressed amyloid precursor protein (APP), or APP and presenilin1 (PS1) which leads to the accumulation of unusual fragments generated by α-secretase, such as C-terminal fragment-β (CTF-β). CTF-β is more toxic than Aβ and CTF-β does not accumulate in human AD brains. A recent study estimates that the neuropathological features of these mouse models are due to artifacts related to APP overexpression [
28] and may explain the lack of translational success in clinical trials. We have been using a second-generation AD model recently developed at the Riken [
28] which has a modified APP gene that has humanized Aβ sequence with three mutations in APP
NL-G-F. This mouse model produces robust age-related spread of Aβ aggregates and cognitive problems with endogenous levels of APP. It is possible that the beneficial effects of exercise in traditional AD mouse models are due to reductions in these unusual protein fragments not seen in human AD.
A second issue we want to address in the present experiments is the focus of earlier studies on the effects of exercise on the two commonly studied pathologies associated with AD, Aβ plaques, and neurofibrillary tangles (NFT) [
2,
3]. However, other brain pathologies are linked to the etiology of AD, including activated microglia and reactive astrocytes and other neuroinflammatory markers such as interleukin [
29,
30]. Additionally, cholinergic cell death in nucleus basalis was also considered as one of the main markers of AD [
31]. Previously, it has also been reported that AD patients have low levels of acetylcholine in the brain [
32,
33]. Cholinergic dysfunction in the cortex and hippocampus regions is highly correlated with cognitive decline in AD [
34]. Accordingly, we will employ a wider panel of AD pathology than most of the other studies in this area of research.
In the present study, we investigated the potential beneficial effects of long-term voluntary exercise on cognitive functions and brain pathology found in the APPNL-G-F mouse model of AD. For the voluntary exercise experiment, individual mice (3 months old) were kept in a cage with a running wheel for 9 months. At 12 months of age, the subjects were tested on the Morris water task (MWT), novel object recognition (NOR), and a fear conditioning (FC) task to assess learning and memory functions mediated by brain networks centered on the hippocampus (HPC), perirhinal cortex (PRhC), and the amygdala (AMYG) respectively. After completion of behavioral tests, histology was completed to assess any changes in amyloid pathology, a microglial marker for neuroinflammation, and cholinergic cells.
Discussion
Regular physical exercise is associated with sustained cognitive functions as people age and may slow down or prevent the progression of various neurodegenerative diseases including AD [
55]. In the present experiments, we tested the effects of long-term voluntary exercise for 9 months in a new generation knock-in mouse model of AD, the APP
NL-G-F model [
28]. At the age of 12 months, the effects of this lifestyle treatment on learning and memory function and brain pathology associated with AD were evaluated. APP
NL-G-F mice allowed to voluntarily exercise showed an improvement in cognitive functions associated with various learning and memory networks [
50,
56]. Brain pathology associated with AD was also impacted by long-term exercise including significant reductions in amyloid load, microgliosis, and preservation of ChAT
+ cells (cholinergic function) in the brain of APP
NL-G-F mice. These profound reductions in brain pathology associated with AD are likely responsible for the observed improvement of learning and memory functions following extensive and regular exercise. These findings suggest potential of voluntary physical exercise to mitigate the cognitive deficits found in humans suffering from AD.
Use of an AD mouse model that might more closely replicate the brain pathology found in human AD
Our lab has characterized the APP
NL-G-F mouse in several experiments and found that these mice show significant Aβ plaque through regions of neocortex (NC) and HPC, and indicate cognitive impairments at 6 months, but not earlier [
37]. We showed that learning and memory functions associated with HPC, PRhC, and AMYG were compromised. The APP
NL-G-F mice also showed increased astrocytosis in the HPC, NC, MSDB, and other brain areas. Other brain changes in APP
NL-G-F mice included cholinergic and norepinephrine dysfunction [
37]. These brain and behavioral changes are consistent with changes found in human AD patients supporting the use of APP
NL-G-F mice as an important improvement in available AD mouse models.
In the present study, we demonstrated the efficacy of long-term physical activity in reducing AD brain pathology and associated cognitive impairments in the APPNL-G-F mice, a mouse model of AD that probably produces results that are more translatable to humans because of its similarity to human AD pathology.
Effects of long-term voluntary exercise on various neural networks implicated in learning and memory functions
The APP
NL-G-F mouse model shows significant learning and memory impairments, and the pattern of impairments suggests that several memory networks are compromised [
37]. These impairments include spatial learning and memory abilities dependent on a memory network centered on the HPC; novel object recognition dependent on a memory network centered on the PRhC [
57,
58]; and fear conditioning processes dependent on a memory network centered on the AMYG [
59‐
61]. We also showed that these same central brain regions of these networks exhibit many of the pathologies of human AD including Aβ pathology, microglial activation, and cholinergic dysfunction.
In the present experiment, we used this knowledge base about the mouse model of AD and evaluated the impacts of long-term voluntary exercise. This study was inspired by epidemiological and clinical studies showing that lifestyle changes like physical activity are a viable preventative approach for this form of age-related cognitive decline.
The potential of exercise to preserve cognitive functions in aging and age-associated brain diseases including dementia and AD has been reported in previous studies [
14,
18,
20,
21]. These clinical findings were supported by several preclinical studies showing that physical activity improved the learning and memory functions in experimental models of AD [
62‐
65].
The findings from the present study are consistent with other work assessing the benefits of exercise in mouse models of AD. One study reported that voluntary exercise for 16 weeks improved the memory function of Tg2576 mice and evidence also suggests that voluntary exercise over forced exercise was more effective [
27] (using treadmill). In another study, results showed the beneficial effect of exercise in the Tg4-42 mouse model of AD on MWT and NOR tests, although it should be noted that these experiments combined exercise with an enriched environment [
26]. Similarly, Adlard and colleagues also reported that exercise improved the learning and memory of TgCRND8 mice in the MWT [
63]. The results from the present experiments combined with previous research suggest that long-term voluntary exercise reverses learning and memory functions dependent on neural networks centered on the HPC and PRhC. We also showed that long-term voluntary exercise improved fear conditioning to context in the APP
NL-G-F mice, although this improvement did not extend to cued fear conditioning. These results suggest that long-term voluntary exercise reverses, at least partially, the learning and memory network functions dependent on a neural network centered on the amygdala.
Effects of long-term voluntary exercise on brain pathology associated with AD
In the present experiment, we found decreases in Aβ pathology in various brain regions such as MSDB complex, HPC, RSC, PRhC, and CCA of 12 months old APP
NL-G-F mice following long-term voluntary exercise. These brain regions have been implicated in different types of learning and memory functions [
56,
57,
66‐
68]. The reduction in the amyloid burden in these areas may be responsible for improvement in the various behavioral tasks.
Other work using different mouse models of AD show similar effects [
62,
64,
65,
69‐
73]. For example, physical exercise decreases levels of Aβ
40 and Aβ
42 in the HPC and AMYG of APP/PS1 transgenic mice [
71]. In another experiment, Adlard and colleagues also reported that TgCRND8 mice with wheel running access for 5 months showed reduction in amyloid pathology which they argued might be responsible for the cognitive improvements observed [
63]. Other studies are also consistent with this claim that exercise can reduce amyloid load [
22,
27,
63].
Glial cell dysfunction observed in the postmortem human AD brain has been documented in various clinical studies [
74,
75]. Additionally, several preclinical studies reported that physical exercise decreased astrocytes and microglial activation in experimental models of AD [
70,
72,
73,
75]. Interestingly, in the present study, the long-term voluntary exercise manipulation only reduced microgliosis in specific brain regions like the HPC, PRhC, and MSDB. The latter effect is of specific interest as the MSDB complex provides cholinergic inputs to HPC and progressive deterioration of these projections are found in aging and neurodegenerative diseases including AD [
32,
76‐
78], and we found the same effect in APP
NL-G-F mice [
37]. In the present study, long-term voluntary exercise prevented the loss of ChAT
+ cells in MSDB complex in 12-month-old APP
NL-G-F mice. A previous study also reported the beneficial effect of exercise on ChAT
+ neurons in the MSDB of THY-Tau22 mice, a model of AD [
69]. These findings indicate that long-term voluntary exercise aids in reducing inflammation in the AD brain. In the present study, the pattern of results suggests that the HPC, PRhC, and MSDB benefit the most from regular physical activity.
APP mice, anxiety, and exercise
There is a significant amount of disagreement concerning the APP
NL-G-F mouse and whether they show an anxiety phenotype. Some researchers report that this AD mouse model does not show anxiety [
79‐
82] and others do report an anxiety phenotype on some but not other measures [
83,
84]. This is potentially relevant to the present study as it has been shown that exercise can reduce anxiety in humans although it is controversial in rodents [
85]. If the APP
NL-G-F mice are anxious it is possible that the exercise treatment is reducing anxiety which could improve learning and memory performance separate from potential effects on plasticity and learning and memory function. For example, anxious mice might show elevated levels of thigmotaxia (not venturing away from the pool wall) which can impair MWT performance [
45,
86]. If the exercise treatment reduces anxiety, this could be a reason for the improvements on the MWT. To assess this idea, we analyzed thigmotaxia during MWT training and found evidence that the APP
NL-G-F mice that were not given the exercise manipulation showed elevated thigmotaxic behavior compared to the controls or the APP
NL-G-F that were given the exercise treatment.
As noted above, thigmotaxia can emerge for a variety of reasons including, but not limited to, impaired learning and memory ability, anxiety, sex hormones, and executive functions. It is unclear in the present study, as in all other studies of this nature, what is driving the increase in thigmotaxia in the APP mice although impairments in learning ability and increased anxiety seem to be the most likely. One idea is that the positive effects of exercise on spatial learning and memory might be on neural networks important for controlling anxiety that are separate from the central structures of neural networks important for learning and memory functions examined in the present study. Another hypothesis is that impaired learning and memory functions render the subjects less confident in finding an escape route and they stay near the pool wall waiting to be removed by the experimenter [
44]. This view emphasizes that the relationship between thigmotaxia and learning is not unidirectional. That is, elevated thigmotaxia can impair learning in the MWT but impaired learning can also increase thigmotaxia. A final hypothesis, one that we favor, is that the neural networks assessed in the present study that are implicated in learning and memory functions are also implicated in controlling fear responses and general anxiety [
87‐
89] particularly ventral hippocampus, amygdala, and various parts of the prefrontal cortex. These systems are thought to control fear and anxiety by constraining fear responses to predictive contexts/cues predictive of aversive events, in other words via learning and memory functions [
50]. According to this view, it is likely that the impacts of exercise on anxiety are via reduced pathology and dysfunction of the hippocampus, amygdala, and prefrontal cortex and associated cognitive functions. This analysis is consistent with the claim that voluntary exercise improved learning and memory functions at least partially via reductions in anxiety and reduced various brain pathologies associated with AD in medial temporal lobe brain regions thought to be central in complex neural networks supporting various forms of memory that control fear and anxiety responses.
Further research focused on disentangling the different hypotheses about the relationship between memory impairments associated with AD, anxiety, and treatment effects is required to fully understand this complex issue. This work would include an assessment of the effects of treatments in AD models, on dorsal versus ventral hippocampal pathology, and on their relationship to improved memory and/or reduced anxiety.
Strengths of the current approach
The current experiment is unique and may represent an advancement in our understanding of the efficacy of long-term exercise in preventing the rapid descent into dementia in AD. First, the current approach used a new generation mouse model of AD. Second, this study focused on long-term physical activity in isolation versus in combination with other strategies like environmental enrichment and cognitive training. Third, we employed a battery of learning and memory tasks as functional assays for different learning and memory networks. Finally, we assessed a wider panel of pathologies associated with AD in these brain networks.
Limitation of the study
In the present study, we have provided strong evidence for the therapeutic potential of non-pharmacology strategies for the management of AD. Although, we have revealed promising results showing protective effect of long-term exercise against cognitive impairment and pathology of AD, still, there are few limitations to the present study. First, negative littermates of APP
NL-G-F mice have not been used as a normal control. We used C57BL/6 as a normal control as APP
NL-G-F mice are generated on a C57BL/6 background. Looking at the research literature using this AD mouse model, a lot of researchers use C57’s as controls for APP
NL-G-F mice. However, if you look at the paper by Nilsson, Saito, and Saido (2014), the authors suggest that because APP
NL mice exhibit the same levels of CTF-beta as APP
NL-F and APP
NL-G-F mice, they are the proper negative controls [
90]. In our view, the use of the APP
NL mice as the negative controls might not be appropriate as they have one of the mutations associated with AD pathology. For the present study, these negative controls were not available in our lab at the time. We currently have these mice in house now and will use them in future experiments with these caveats in mind. Still, we found that non-pharmacological (long-term exercise) manipulation reduced brain pathology and improved cognitive functions of this knock-in mouse model of AD. Second, normal control and APP-NE groups were not provided with a static wheel in same cage for 9 months. Ideally, this experimental group should be included in future studies to rule out the potential additive effect of environmental enrichment due to the presence of the wheel in the APP-Ex group. We predict that this type of experimental design would replicate the present results showing the beneficial effect of exercise only in the APP-Ex group and no contribution of the presence of an inoperable running wheel in the home cage. Additionally, we did not record the extent of exercise in the present study, so we could not correlate the level of exercise of an individual with the behavioral parameters and pathological makers studied in the present study. Third, another potential caveat associated with the design of the present study is that the subjects that were given the exercise treatment were socially isolated during this time. It is possible that social isolation might have contributed to the reductions in brain pathology [
91]. However, we think that this is unlikely as our reading of the literature suggests that social isolation should impair hippocampal plasticity and learning and memory function [
92‐
94]. Our view is that these effects would have worked against the beneficial effects of exercise and yet we still found profound impacts of the exercise manipulation on brain pathology, anxiety, and learning and memory functions. Finally, we selected 4–5 mice from each group for histopathological assessment. This means that not all the brains from subjects included in the behavioral analysis were assessed. We have consistently (37) found that this number of subjects for pathological assessments is large enough and the patterns of effects, in the AD mouse model we use, over many studies (onset of pathology, maximum threshold levels, location). However, it is important for the reader to note this when interpreting the data, as well as that this design made it difficult to correlate long-term exercise-induced improvement in behavioral outcomes and brain pathology.
Future work should consider these limitations when designing future experiments assessing the impacts of exercise on AD pathology and associated cognitive impairments.
Summary
The present experiments provide strong causal evidence that long-term voluntary exercise significantly reverses severe cognitive impairments, anxiety, and associated pathological changes in the brain of a new generation knock-in model of AD. Several possible mechanisms involved in protective effect of exercise against AD include reduction in oxidative stress [
95,
96], neuroinflammation [
70,
72,
73,
75], amyloid pathology [
22,
27,
63,
71], and improvement in cerebral blood flow [
97,
98]. Physical exercise enhanced neurogenesis, synaptogenesis, and cholinergic cell’s function may be responsible for improvement in various cognitive functions including learning and memory [
64,
69,
97‐
100]. Although, out of these several mechanisms, we have only studied the neuroinflammation, amyloid pathology, and cholinergic cells integrity, however, involvement of other mechanisms in the protective effect of long-term exercise cannot be ruled out. The present results combined with our analysis of the existing research literature suggests that implementing exercise in combination with other lifestyle factors like cognitive training or diet could be effective non-pharmacological approaches to prevent or delay the progression of AD. Future work will be directed at assessing the effects of other lifestyle preventative measures alone or in combination with voluntary exercise.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.