The function of mitochondria to regulate healthy neurons is mainly affected by aging and pathological status. Mitochondria are not only the leading bioenergetic powerhouse, but are also the primary generators of reactive oxygen species (ROS). Normally, there is a dynamic balance between production and neutralization of ROS, and low-level ROS will not produce toxic effects. As a regulator for the development and function of neurons, ROS can help to establish neuronal polarity for growth cone pathfinding and connect the synaptic transmission to the tuning of neuronal networks (Oswald et al., 2018a). Nevertheless, when there is electron leakage, ROS is mainly produced in electron transport chain (ETC) complexes I and III of the mitochondria, thus leading to the generation of superoxide anions (O^2–) and hydrogen peroxide (H₂O₂). ROS dysregulation with the extension of age may interfere with the adaptive adjustment of neurons, thereby resulting in network failure and synapse loss (Oswald et al., 2018b).

Recent evidence has suggested that oxidative stress also causes mitochondrial dysfunction is associated with the development of AD-related pathology (Yu et al., 2018). Due to the limited glycolytic activity of neurons, it is entirely dependent on the energy production of mitochondria. Driven by energy, neurons establish membrane potential, and synthesize, secrete and recycle neurotransmitters, thereby maintaining intracellular Ca²⁺ homeostasis and effectively transmitting neuronal signals. Meanwhile, several studies have shown that mitochondrial respiratory chain and TCA cycle enzymes are significantly reduced (Butterfield et al., 2006). Correspondingly, impaired mitochondrial respiratory chain will further disorder the electron transfer efficiency, increase ROS content, and lead to cellular injury. Also, mtDNA located near the mitochondrial respiratory chain is susceptible to mutations, thus resulting in faulty proteins and ROS (Grimm et al., 2016). In addition, oxidative stress is one of the most prominent hallmarks of the aging process, characterized by the combination of the reduced antioxidant defense system and impaired mitochondrial oxidative phosphorylation (OXPHOS) activity (Grimm and Eckert, 2017). Neurons that rely heavily on OXPHOS to meet their energy demands are particularly prone to energy hypometabolism. Similarly, due to the non-dividing characteristics, neurons could not be replaced in life except that the hippocampus continuously generates new neurons during the adulthood period (Sanchez-Varo et al., 2012). In this case, more and more evidences confirm that energy failure and oxidative stress caused by mitochondrial respiratory dysfunction play a crucial role in the pathogenesis of AD (Marques-Aleixo et al., 2015).

Several clinical evidences have documented that patients with mild cognitive impairment (MCI) have already experienced oxidative stress-induced damage, and this symptom represents the early clinical stage of AD (Aluise et al., 2011). At the same time, the level of biomarkers associated with oxidative stress is directly related to the severity of cognitive impairment and the disease progression from MCI to AD (Ansari and Scheff, 2010). Oxidative stress not only occurs in vulnerable regions of the brain, but also in non-degenerative tissues, such as fibroblasts and lymphocytes, and can precede the Aβ plaque formation and the onset of symptoms of AD. As described, ROS can mainly increase the levels of lipid peroxidation markers such as thiobarbituric acid-reactive substances (TBARS) and 4-hydroxynonenal (4-HNE), thereby accelerating the damage of the brains in AD patients (Butterfield et al., 2002). Similarly, other data have also confirmed that mitochondrial-derived ROS is a sufficient trigger to induce the formation of Aβ (Leuner et al., 2012). In this case, the β-site APP cleaving enzyme 1 (BACE1) activity, which Aβ itself enhances due to the increased ROS level, can accelerate mitochondrial dysfunction and ROS production, thereby enhancing Aβ production in the progression of SAD (Leuner et al., 2012). Furthermore, recent reports have verified that increased production of ROS and susceptibility to the opening of mitochondrial permeability transition pore (mPTP) in APP/PS1/Tau triple transgenic (3xTg) AD model can lead to the increased brain barrier permeability and abnormal brain mitochondria (Carvalho et al., 2013).

Another critical factor for AD pathology is neuroinflammation that is strongly associated with oxidative stress in the brain. For example, Aβ also induces oxidative stress in astrocytes by activating NADPH oxidase, thereby producing chronic neuroinflammation, secreting proinflammatory cytokines, and further exacerbating oxidative stress-induced damage (Mostafa et al., 2016; Wang et al., 2016). At the same time, Aβ can induce increased mitochondrial fragmentation, membrane potential loss and impaired respiratory function of astrocytes (Sarkar et al., 2014). In summary, mitochondrial disorders can be induced by increased oxidative stress and neuroinflammation, thereby contributing to the accumulation of Aβ, and leading to the damage to the structure and function of neurons and synapses as well as astrocytes in the brain tissues, which plays a vital role in the pathogenesis of AD.

Mitochondrial biogenesis, that is, the growth and division of the pre-existing mitochondria, is a process that requires the coordinated synthesis of proteins encoded by nuclear DNA (nDNA) or mtDNA. Peroxisome proliferator-activated receptor gamma coactivator 1-alpha (PGC-1α) is a central inducer of mitochondrial biogenesis and co-activates the transcription of nuclear respiratory factor 1 (NRF1), which can regulate the transcription of mitochondrial transcription regulator A (TFAM) (Palikaras and Tavernarakis, 2014). When cell energy requirements exceed mitochondrial ATP production, mitochondrial biogenesis is initiated, and mitochondrial regulatory factors will enter the mitochondria to initiate mtDNA replication and transcription and trigger the expansion of the mitochondrial network to meet the demand of the body’s energy metabolism. There is accumulating evidence linking NDs to the defects or mutations of mtDNA. Increasing evidence suggests that an increase in the number of mtDNA mutations in the frontotemporal lobes of the brain in AD patients is consistent with the increase in Aβ42 and hyperphosphorylated Tau in the cerebrospinal fluid, which further confirms the increased mtDNA mutation is associated with the deposition of Aβ in the brain (Podlesniy et al., 2013). When mtDNA is damaged under normal physiological conditions, oxoguanine glycosylase 1 (OGG1) will work with other repairing genes to maintain mtDNA stability. Furthermore, in other clinical studies, the decreased OGG1 activity reveals the early occurrence in the progression of AD, which is possibly mediated by 4-HNE inactivation and may contribute to the elevated 8-hydroxydeoxyguanosine (8-OHdG) in the brain associated with MCI and late-stage AD (Shao et al., 2008).

PGC-1α, the primary regulator of mitochondrial biogenesis, mediates different transcription factors to promote mitochondrial biogenesis. As previously reported, PGC-1α and its significant downstream target genes are down-regulated in the AD brain (Qin et al., 2009) and the M17 cells over-expressing familial AD-causing APP mutant (APPswe) (Sheng et al., 2012), suggesting mitochondrial biogenesis is impaired in AD. PGC-1α is regulated by NRF1 and NRF2, and TFAM encodes, translates and transcribes mitochondrial proteins. Previous studies have found that the expression of PGC-1α, NRF1, NRF2, and TFAM is decreased in cells transfected with APP, and the deposition of Aβ is increased. In parallel, the expression of PGC-1α is negatively correlated with the level of Aβ deposition, indicating that deposition of Aβ can suppress mitochondrial biogenesis (Sheng et al., 2012). Following this line of thought, in vitro cultured Tg2576 (an AD mouse model) mouse hippocampal neurons, increased expression of PGC-1α can inhibit the production of Aβ (Porcellotti et al., 2015). Also, the overexpression of TFAM in SH-SY5Y cells can protect them from Aβ-induced mitochondrial dysfunction (Xu et al., 2009), while the overexpression of PGC-1α can effectively inhibit the secretion of Aβ and increase the levels of soluble Aβ in neuroblastoma (N2a) cells (Katsouri et al., 2011). It is also worth noting that the mortality is reportedly increased in APP/PS1 double transgenic mice with specific knockout of SIRT1 that exhibits the reduced activity of SIRT1 deacetylase and increased production of Aβ (Zhang et al., 2011). Further studies have confirmed mitochondrial biogenesis can be promoted by increased SIRT1, which deacetylates and activates PGC-1α to increase its transcriptional activity (Fanibunda et al., 2019). Thus, PGC-1α can regulate mitochondrial biogenesis through the deacetylation of SIRT1 and plays an essential role in the pathogenesis of AD. In addition, Tau also can be deacetylated by SIRT1 (Min et al., 2010), and the level of SIRT1 in the brain is decreased during the progression of AD patients (Julien et al., 2009). Since acetylation inhibits the binding of Tau to microtubules, enhances Tau accumulation by suppressing Tau degradation, and affects the structure and function of neurons in Tau overexpression models such as C. elegans single-copy transgenic model (Guha et al., 2020a,b), the activation of SIRT1 can promote the deacetylation of Tau protein, suggesting that targeting SIRT1 activation can serve as an effective strategy for the prevention and mitigation of AD.

As the highly dynamic organelles, mitochondria present a constant change in shape and status under the regulation of fusion and fission to promote dynamic balance and maintain the health of the mitochondrial network (Mishra and Chan, 2016). On the one hand, the fission/fusion dynamics of mitochondria promote the distribution of mitochondria along the axon to the synapse (Chen et al., 2007), regulate mitochondrial transport and neurotransmitter release, thereby maintaining the transmission of mitochondrial membrane potential (ΔΨm) and calcium buffering (Szabadkai et al., 2006). On the other hand, defective mitochondrial components can be effectively separated during the aging process, thereby preventing damage from ROS (Westermann, 2008). In short, the mitochondrial dynamic is essential for meeting high energy demands and promoting neuroprotection (Chen and Chan, 2009). In the process of regulating mitochondrial dynamics within the neuronal environment, dynamin-related protein 1 (Drp1) and mitochondrial fission protein 1 (Fis1) mainly regulate the process of fission (Su et al., 2010); meanwhile, the fusion process is mediated by mitofusin 1 (Mfn1), mitofusin 2 (Mfn2), and optic atrophy protein 1 (OPA-1) (Youle and van der Bliek, 2012). The morphology and function of mitochondria are regulated dynamically by conserved fission and fusion proteins, and are closely related to cellular homeostasis and aging status (Picard et al., 2013). Additionally, previous studies have also confirmed that mitochondrial fragmentation caused by the dysregulated mitochondrial dynamics, especially the excessive mitochondrial fission, seems to be a vital indicator of neuronal damage (Gao et al., 2017). Another relevant study has shown that abnormal mitochondrial dynamics caused by increased fission is an important early event observed in AD cells (Bonda et al., 2010). In this regard, as previously mentioned, compared with the age-matched control group, the number of mitochondria in neurons of AD mice is reduced, and the proportion of mitochondria with broken cristae is relatively higher (Nixon et al., 2005).

Under normal circumstances, mitochondria dynamically regulate intracellular calcium balance and ATP production, and keep ROS at an appropriate level through continuous fusion and fission to eliminate and repair mutant mtDNA. During the aging process, increased ROS leads to imbalanced mitochondrial homeostasis, and accumulated Aβ and hyperphosphorylated Tau protein cause mitochondrial damage, thereby resulting in pathological changes of AD (García-Escudero et al., 2013). For instance, studies have found that the mitochondrial transition of fission state characterized by increased levels of Drp1 in MCI and AD patients (Silva et al., 2013). Similarly, studies have shown that the increase of Aβ can increase the content of nitric oxide (NO) in the brain and activate the abnormal expression of neuronal Drp1, thus resulting in increased mitochondrial fission and the damage of neurons (Archer, 2013). The most important genetic risk factor for about 90% of SAD is ApoE4. In SAD, ApoE4 can down-regulate the expression levels of the proteins associated with mitochondrial dynamics and ETC (Yin et al., 2020). In the hippocampus of the ApoE4 vector and ApoE4 transgenic mice, the expression of Mfn1, Mfn2, and Drp1 is decreased (Simonovitch et al., 2019). Similarly, the defects in mitochondrial dynamics have been reported in different AD mouse models with Tau overexpression. For example, the expression of P301L mutant Tau can modify mitochondrial dynamics, thereby reducing the expression of both fission and fusion protein regulators (Schulz et al., 2012). In the brain tissues of various AD mouse models with the transfection of APP, APP/PS1, and 3xTg, the binding of Drp1 to hyperphosphorylated Tau protein is enhanced, thereby aggravating mitochondrial dysfunction and synaptic damage (Manczak and Reddy, 2012). Therefore, abnormal mitochondrial dynamics is a contributing factor to AD, and rescuing abnormal mitochondrial dynamics is a valuable therapeutic target for AD.

PINK1/Parkin is one of the components in the classic signaling pathway of mitophagy. Under the conditions of stress, injury or aging, mitochondrial serine/threonine kinase PINK1 can sense mitochondrial polarization status, an E3 ubiquitin-protein ligase (Parkin). Under normal conditions, mitochondria have a low membrane potential, and PINK1 is imported into the inner membrane of mitochondria (IMM) and cleaved by mitochondrial-specific protease presenilin-associated rhomboid-like (PARL) (Kane et al., 2014). However, in compromised and depolarized mitochondria, PINK1 is accumulated in the outer membrane of mitochondria (OMM) (Cadete et al., 2019). Mitochondrial protein is phosphorylated by PINK1, thus leading to the recruitment of autophagy cargo adapters and degrading autophagolysosomes by binding to autophagosomes and lysosomes through LC3 (Springer and Macleod, 2016). In order to amplify this signal, PINK1 recruits Parkin from the cytosol by the phosphorylation of Mfn1 and Mfn2. Besides, PINK1 phosphorylates ubiquitin and ubiquitin-like domain (Ubl) of Parkin, which activates Parkin E3 ligase and facilitates its recruitment to OMM (Hamacher-Brady and Brady, 2016). Once in OMM, Parkin can trigger the ubiquitination of Mfn1 and Mfn2, prevent the fusion of mitochondria, and can act as an isolation mechanism when these organelles are damaged, followed by isolation and degradation through selective autophagy (Gomes and Scorrano, 2013). Then, Parkin ubiquitinates OMM protein, such as voltage-dependent anion channel (VDAC1), to generate polyubiquitin chains, recruit autophagy receptor p62, and bind to LC3, thereby degrading and clearing damaged mitochondria (Novak, 2012).

This type of autophagy driven by Nix/Bnip3, two major pro-apoptotic proteins, is named mitophagy induced by various potential mechanisms (Marinković et al., 2021). Unlike the signal pathway controlled by PINK1/Parkin that cannot directly bind to autophagosome receptors, Bnip3, and Nix can directly bind to autophagy machinery components to mediate mitophagy, including mitochondrial depolarization, PTP aperture, or interference in fission-fusion machinery (Matsuda, 2016). Indeed, the phosphorylation of Bnip3 at Ser17 and Ser24 can promote its binding to LC3-II and Golgi-associated ATPase enhancer of 16 kDa (Gate-16), and facilitate subsequent mitophagy. Recently, Nix has been suggested to induce mitophagy by its interaction with LC3. LC3 interacts with gamma-aminobutyric acid receptor-associated protein (GABARAP) to form the LC3/GABARAP complex, which mediates the mobilization of autophagosomes to mitochondria, for accomplishing the elimination of damaged mitochondria (Hamacher-Brady and Brady, 2016). Besides, Nix can also interact with the Beclin1/Bcl-2 complex to release Beclin1, thus generating the free Beclin1 to induce autophagy (Mishra and Chan, 2016). Furthermore, Bnip3 and Nix are phosphorylated and form homodimers, thereby integrating with OMM and binding to LC3 to induce mitophagy (Zhang and Ney, 2011).

In addition to classical PINK1/Parkin and Nix/Bnip3-mediated mitophagy, other mitophagy receptors have been discovered, including FUNDC1, and activating molecule in BECN1-regulated autophagy protein 1 (Ambra1) in mammals (Novak, 2012). Under hypoxic conditions, the mitochondrial membrane protein FUNDC1 interacts with LC3 to promote the clearance of mitochondria (Fu et al., 2013). In neurons, the externalization of cardiolipin and its interaction with LC3 has been confirmed to mediate mitophagic degradation of malfunctioning mitochondria (Chu et al., 2013). Additionally, Ambra1, an upstream autophagy regulator of mitochondria, and its pro-autophagy activity are inhibited by Bcl-2 under normal conditions. Upon the induction of mitophagy, Ambra1 binding to LC3 via LC3-interacting region (LIR) motif can increase mitochondrial clearance (Strappazzon et al., 2015). The uncertainty of these results stimulates the necessity for further exploring cellular mechanisms of mitochondrial turnover.

There is a complex interplay between Aβ and autophagy, as autophagy controls both the clearance and generation of Aβ. On the one hand, the induction of autophagy can reduce the accumulation of amyloid plaques and rescue neurodegeneration in various systems (Boland et al., 2008). On the other hand, Aβ may also be generated in autophagosomes that appear to contain both the precursor APP and the enzyme presenilin-1 (PS1) responsible for the cleavage of APP to Aβ (Nixon et al., 2005). Furthermore, autophagy is critical for the extracellular secretion of Aβ peptides, as validated by the conditional knockout of the key autophagy-related gene 7 (Atg7), in the APP (Swedish mutation) transgenic mice. However, these events are associated with toxic intracellular accumulation of Aβ, which likely causes neurodegeneration together with amyloidosis and memory deficit (Nilsson et al., 2013).

As a negative regulator of autophagy, mTOR activation will inhibit the initiation of autophagy. The expression of mTOR in the brain of APP/PS1 mice is increased, accompanied by the blocked initiation of autophagy and the increased deposition of Aβ (Wang et al., 2014). However, the injection of mTOR inhibitor, rapamycin, into the brain of APP-mutant AD mouse model can activate autophagy, reduce Aβ pathology, and improve cognitive capacity (Spilman et al., 2010). Besides, the expression of Beclin1 is used as a sign of autophagosome initiation. The expression of Beclin1 in the brain of APP transgenic AD mouse is reduced, and the deposition of Aβ is enhanced (Komatsu et al., 2006). After the initiation of autophagy, the membrane vesicles gradually expand to form the closed autophagosomes. Previous studies have demonstrated that many autophagosomes accumulate in the brain of APP/PS1 mice before Aβ deposits, and many autophagic vesicles containing APP or Aβ accumulate in the brain tissue (Jiang et al., 2014). Consistent with this, a report has described that LC3-II/LC3-I ratio reveals a decrease and p62 expression exhibits an increase in the cortex of AD mouse at the late stage of the disease, and the autophagosomes containing APP and Aβ are gathered in the axon and cannot be transported normally, thus disrupting the fusion of autophagosomes and lysosomes, and leading to the failure of Aβ degradation (Son et al., 2012; Bordi et al., 2016). Alternatively, the excessive accumulation of Aβ can also lead to the destruction of bilayer membrane structure of autophagolysosomes and the abnormal increase of hydrolytic enzymes, thereby causing neuron loss in AD patients (Usenovic and Krainc, 2012). However, the application of lysosomal activators can improve the lysosomal activity in the brain tissues of AD mouse, thereby effectively reducing the deposition of Aβ, and improving learning and memory capacity (Yang et al., 2011).

Besides Aβ, there is also a link between hyperphosphorylated Tau and mitophagy. Despite being substrates of the ubiquitin-proteasome system, Tau proteins are delivered to the autophagosome-lysosomal system for degradation, so impaired autophagy can lead to the formation of Tau oligomers and the aggregation of Tau insoluble species (Li et al., 2017). Furthermore, autophagy impacts Tau phosphorylation: autophagy-deficient mice display hyperphosphorylated Tau, while autophagy induction reduces Tau phosphorylation (Piras et al., 2016; Uddin et al., 2019). Interestingly, Tau self-modulates the signal pathways of autophagy, as shown in disturbed integrity of lysosome membranes and retarding autophagosome maturation by inhibiting the activity of histone deacetylase 6 (HDAC6) in the presence of Tau aggregation (Uddin et al., 2019). On the other hand, previous studies have reported autophagy in brain cells is declined before the occurrence of AD. Similarly, insufficient Beclin1 in neurons cultured in vitro also can cause the accumulation of intracellular Tau hyperphosphorylation. The injection of Beclin1 can reduce the accumulation of hyperphosphorylated Tau protein (Nixon, 2017). In addition, in primary neurons with trehalose, an enhancer of autophagy, the lysosomal activity is significantly increased, and the accumulation of hyperphosphorylated Tau protein is reduced (Krüger et al., 2012).

Genome-wide association studies in AD have identified the variants of phosphatidylinositol-binding clathrin assembly protein (PICALM), which are associated with NFTs and co-localized with conformational abnormalities and hyperphosphorylated Tau proteins in late-onset AD (Ando et al., 2013). PICALM is a clathrin adaptor protein, whose loss-of-function impairs the endocytosis of VAMP2 and VAMP3 (soluble NSF attachment protein receptors-SNAREs involved in the fusion of autophagosome precursors), as well as VAMP8 (involved in the fusion of autophagosomes with lysosomes/late endosomes). Thus, PICALM depletion in transgenic Zebrafish models can inhibit autophagy at multiple steps: from autophagosome biogenesis to maturation and fusion steps, thus ultimately leading to Tau aggregation (Moreau et al., 2014; Menzies et al., 2015). In addition, the hyperphosphorylation of Tau protein also affects mitochondrial function. On the one hand, mitochondrial changes may be secondary to microtubule depolymerization, thus resulting in the inability to transport healthy mitochondria to axons and dendrites, and dysfunctional mitochondria can be removed through autophagy. On the other hand, hyperphosphorylated Tau proteins induce the defects of mitophagy by their inserting into the mitochondrial membrane and impair the mitochondrial residence of PINK1/Parkin (Oliver and Reddy, 2019).

To date, approximately 30% of adults worldwide are underactive, which becomes a more significant risk factor for abnormal ROS. At the same time, the biomarkers associated with low antioxidant capacity and high-level oxidative stress can be determined in AD patients at the early stage (Schrag et al., 2013; Brown et al., 2017). It has been confirmed in different biological models that exercise is closely correlated with improved antioxidant capacity of the brain and reduced oxidative stress-induced damage, thereby preventing redox changes associated with aging and NDs. In addition, 4-week treadmill exercise can significantly promote the neurogenesis of streptozotocin (STZ)-induced AD model rats, accompanying with reduced 4-HNE, DNA double-strand breaks (p-H2AX), and 8-OHdG, as well as reduced Aβ aggregation and hyperphosphorylated Tau protein in the brain (Lu et al., 2017). Similarly, in NSE/hTau23 mice following 3 months of treadmill exercise, the increased activity of antioxidant enzymes such as SOD and CAT in the brain is detected, while the expression of hyperphosphorylated Tau protein is decreased (Leem et al., 2009). Another previous study through 3xTg-AD mice for 3-month voluntary wheel running has documented that exercise can significantly reduce Aβ deposition and Tau hyperphosphorylation, reduce oxidative indicators such as glutathione disulfide (GSSG), and improve SOD activity in the hippocampal tissue, suggesting that exercise can improve redox status, reduce Aβ accumulation and Tau hyperphosphorylation and enhance brain memory function (García-Mesa et al., 2016). However, exercise regimens with different intensities, durations and loads can lead to distinct physiological and functional adaptation of the brain. Moreover, age and animal characteristics also affect the variability of experiments, which may explain the controversial results between oxidative stress in the brain and physical exercise. For example, some studies have found that endurance training can decrease lipid peroxidation and oxidative DNA damage, while acute exercise cannot cause any significant alteration of mitochondria in the brain (Liu et al., 2000). The above findings indicate that mitochondrial adaptation may be affected by the bioenergetic properties of exercise.

In addition, moderate exercise combined with antioxidant supplementation can synergistically reduce the age-dependent risk of protein and lipid oxidation modification in the brain, thereby significantly activating the endogenous antioxidant system (Sakr et al., 2015). In the study of AD mouse models, moderate exercise can inhibit the elevated level of oxidative stress in the brain and maintain redox homeostasis (Kim et al., 2019). ROS generation during exercise may play an important role in regulating mediator molecules to affect adaptive responses. ROS also can modulate the expression of antioxidant proteins and uncoupling proteins (U) (Andrews et al., 2005), as well as heat shock proteins (HSPs) (Calabrese et al., 2000). This combinatorial up-regulation mediated by feedback mechanisms can prevent the extent of oxidative stress and apoptosis, ultimately promoting neuroprotection. Overall, exercise training can be used as a potentially effective strategy for enhancing antioxidant enzyme activity of neurons, reducing the release of mitochondrial ROS, improving the excessive oxidation of lipids, proteins and DNA, and reducing the level of oxidative stress and apoptosis with subsequent attenuation of AD progression.

During exercise, energy requirements reveal a sharp increase, and mitochondria can increase the source of available energy. Mitochondrial biogenesis is the basis for maintaining the number of mitochondria and a guarantee for meeting the needs of energy metabolism in neurons. Damaged mtDNA is one of the critical factors for leading to abnormal mitochondrial biogenesis in the AD brain. Recent studies have confirmed that l6-month treadmill exercise can stimulate a significant increase in the expression of oxidative factors such as GSSG, and an improved metabolic level of neurotransmitters such as acetylcholine, glutamate, and aspartate in the brain of 3-month-old mtDNA mutant mice (Clark-Matott et al., 2015). Similarly, another study has reported that the combinatorial therapy with metformin and 6-week treadmill exercise can significantly reduce Aβ accumulation and Tau hyperphosphorylation, up-regulate the activities of SOD and CAT, and down-regulate damaged mtDNA in the brain of the mice, thereby improving learning and memory capacity (García-Mesa et al., 2012). Recently, an animal model study has found that 20-week treadmill exercise can significantly increase manganese superoxide dismutase (MnSOD), ATP production and OGG1, with a significant decrease in Aβ deposition and ROS production in the brain of APP/PS1 mice, suggesting that exercise can reduce mtDNA damage and increase the amount of mtDNA by increasing the activity of mtDNA-repairing enzymes (Bo et al., 2014).

The dynamic balance between mitochondrial biogenesis and mitochondrial clearance and renewal is critical in maintaining mitochondrial function in neurons. Exercise can stimulate mitochondrial biogenesis and maintain the integrity of neuronal structure and function. Recent studies have confirmed that, in 6-month-old SAMP8 female rapidly aging mice, 2-week voluntary wheel running can promote mitochondrial biogenesis by increased SIRT1 and OXPHOS substrates in the hippocampal tissues of mice (Bayod et al., 2015). Similarly, 8-week exercise can decrease Aβ deposition via the SIRT1/PGC-1α signaling pathway, thereby improving learning, and memory capacity (Koo et al., 2017). Furthermore, the mice subjected to 4 weeks of treadmill training exhibit higher learning and memory capacity, higher expression levels of AMPK, PGC-1α and other regulatory factors of mitochondrial biogenesis, and lower levels of Aβ deposition in the brain of AD rats (Azimi et al., 2018). Collectively, exercise can improve the brain function by increasing mitochondrial biogenesis, improving mitochondrial metabolic function and reducing neuronal Aβ deposition via the AMPK/SIRT1/PGC-1α signaling pathway in the brain of AD subjects.

In order to stabilize the normal status of cells, mitochondria urgently need to reorganize through fusion and fission cycles so that damaged or dysfunctional mitochondria can be specifically repaired or eliminated, thereby maintaining the integrity of mitochondrial structure and function. According to a previous report, resistance exercise has almost no effect on mitochondrial biogenesis, but it can change the activation status and total amount of mitochondrial-related fusion/fission proteins (Kitaoka et al., 2015). Additionally, in comparison with untrained mice, the mice subjected to swimming training exhibit the better cognitive function, stronger rejuvenated mitochondria, higher expression levels of PGC-1α, Mfn1/2, and Drp1 proteins, and lower level of oxidative stress (Luo et al., 2017). Similarly, according to current studies, swimming training can significantly improve the learning and memory capacity of D-galactose-induced AD model rats, and induce autophagy by suppressing miR-34a, as well as promote the expression of Mfn2 to improve mitochondrial quality control (Kou et al., 2017). A recent study has reported 12-week treadmill exercise and voluntary wheel running can increase mitochondrial fusion proteins such as Mfn1/2, and suppress the fission-related proteins such as Drp1 to increase mitochondrial plasticity in the cortex and cerebellum of the brain (Marques-Aleixo et al., 2015).

However, another study has demonstrated exercise training can improve mitochondrial function in the brain by improving the function of ETC and promoting a shift in the mitochondrial fission-fusion balance toward fission (increased Drp1) in aged mice (Gusdon et al., 2017). As brain energetic stress can induce mitochondrial fission arrest in AD, the up-regulation of Drp1 may reverse aging-induced mitochondria-on-a-string (MOAS) formation and preserve residual mitochondrial function in the brain (Zhang et al., 2016). Thus, all above results have confirmed that exercise can enhance the dynamic balance of mitochondrial fusion and fission, and maintain brain health.

Under normal physiological conditions, mitochondria eventually split into two uneven offspring. Mitophagy can selectively reduce the membrane potential and degrade mitochondrial progeny that exceeds their repairing capability. Usually, mitochondria responding to depolarization under stress stimuli (such as ROS, nutrition deficiency and exercise), are wrapped into autophagosomes and fused with lysosomes to degrade the damaged mitochondria and misfolded proteins (such as Aβ, Tau protein) in the brain through mitophagy as a form of selective autophagy, thereby maintaining mitochondrial metabolism in the brain. Exercise-induced autophagy caused by treadmill running can improve the learning and memory capacity, which is first discovered in the cerebral cortex of normal mice in 2012 (He et al., 2012).

Currently, increasing evidence confirms the beneficial effects of different exercise regimens on the functional status of autophagy in brain tissues of rodent models. The mice subjected to treadmill running and voluntary wheel running exhibit a higher expression level of LC3-II protein, and a lower expression level of p62 protein in the cerebral cortex of aged mice (Rocchi and He, 2017). The latest study has demonstrated 12-week voluntary wheel running and treadmill exercise can also increase the expression of Beclin1 and LC3-II, and reduce the expression of p62 in the brain of Sprague-Dawley (SD) rats, suggesting that both active and passive exercise can promote the formation of autophagosomes, and accelerate the degradation of damaged organelles and mis-folded proteins by autolysosomes to maintain normal brain functions (Marques-Aleixo et al., 2015).

Similar findings have been reported that 10-week swimming exercise in young SD rats can significantly up-regulate the expression of mitophagy biomarkers LC3-II/LC3-I ratio and Parkin, while the expression of p62 is decreased in the brain of rats, further confirming that exercise can activate mitophagy (Luo et al., 2017). However, the mechanism by which exercise mode achieves this effect needs to be further explored. Furthermore, another study has verified that voluntary wheel running for 5 months can increase the expression of Beclin1 and Atg5 in the cerebral cortex and hippocampus of both 7- and 1-month-old TgCRND8 (expressing a mutant human APP695 gene) mice, thereby suppressing the Aβ deposition in the brain of AD mice (Herring et al., 2016). These studies suggest that exercise can increase the level of mitophagy at different stages of AD, accelerate the clearance rate of Aβ in the manner of mitophagy, and play a certain role in preventing and relieving AD. Additionally, numerous studies have demonstrated that exercise activates the PI3K/Akt/mTOR signaling pathway to increase synaptic plasticity, control neuron survival, suppress apoptotic cell death, and improve the learning and memory capacity of mice (Chae and Kim, 2009; Fang et al., 2013). Further studies on exercise-induced autophagy/mitophagy will be required to assess exercise as a critical built-in mechanism for promoting mitochondrial renewal cycle and functional improvement in AD subjects.