The basal forebrain complex comprising the medial septum, horizontal and vertical diagonal band (VDB) of Broca and nucleus basalis of Meynert (NBM), is an important structural basis for cholinergic projections (Schliebs and Arendt, 2011). Aging leads to moderate degenerative changes in basal forebrain cholinergic neurons (BFCNs) located in this complex, accompanied by loss of the neurotransmitter ACh, resulting in reduction of cholinergic projections in the cerebral cortex and hippocampus (Schliebs and Arendt, 2011; Ruan et al., 2018). Considerable advances have been made in our understanding of the roles of the cholinergic system in the development of AD since the 1970s and 1980s (Bowen et al., 1976; Davies and Maloney, 1976; Whitehouse et al., 1982; Coyle et al., 1983). Abnormal cholinergic activity and function have been extensively observed in both AD animal models (Zhu et al., 2017; Xhima et al., 2020) and human patients (Fernandez-Cabello et al., 2020; Machado et al., 2020). Notably, the loss of cholinergic neurotransmission in AD is mainly due to the dysfunction of BFCNs and a decrease in physiological levels of ACh at the cholinergic synapse. The degeneration of BFCNs in the NBM and VDB induced by Aβ deposition and tau hyperphosphorylation is an important pathological mechanism underlying cognitive deficits in AD patients (Vana et al., 2011; Baker-Nigh et al., 2015; Gonzalez et al., 2021). Thus, there is an urgent need for therapeutics and delivery methods that slow or reverse the degeneration of cholinergic neurons in AD.

ACh, an important neurotransmitter in cholinergic transmission, participates in a range of cognitive activities including attention, learning, and memory (Hasselmo, 2006; Haam and Yakel, 2017). The cholinergic hypothesis of AD has inspired research into the roles of ACh in this disease, resulting in a widely held view that restoring levels of ACh may be useful in treating AD. ACh can promote the soluble Aβ peptide conformation rather than the aggregation-prone β-sheet conformation (Grimaldi et al., 2016; Polverino et al., 2018), and regulate tau phosphorylation (Rubio et al., 2006), to combat Aβ and tau pathology. Of concern, the synthesis, transport, release, and metabolism of ACh are multi-step processes that need to be finely modulated by choline, acetyl coenzyme A, choline acetyltransferase (ChAT), vesicular acetylcholine transporter, AChE, and choline transporters, among others (Ferreira-Vieira et al., 2016). However, the progressive dysfunction of certain key components during aging results in decreased ACh levels and subsequent memory deficits (Bartus et al., 1982). Currently, much attention is paid to changes in the activity of ACh-synthesizing enzyme ChAT and ACh-degrading enzyme AChE in AD. Biochemical examinations of brain tissue samples obtained from AD animal models and human patients have revealed reduced ChAT activity and increased AChE activity in multiple brain areas (Bailey and Lahiri, 2012; Atukeren et al., 2017; Zheng et al., 2018). Supplementation with exogenous ChAT and/or AChE inhibitors (AChEIs) therefore represents a potential therapeutic strategy against AD.

The role of cholinergic receptors, i.e., nicotinic and muscarinic receptors in cholinergic neurotransmission, is not negligible. Muscarinic acetylcholine receptors (mAChRs) belong to the GPCR family and comprise the M1 and M2 subfamilies with a total of five subtypes. M1 receptors (M₁, M₃, and M₅ mAChRs) are located in the postsynaptic membrane and couple to the Gα_q/11 protein (Berizzi et al., 2018), whereas M2 receptors (M₂ and M₄ mAChRs) act as auto-receptors of the presynaptic membrane to inhibit ACh synthesis and release and couple to the Gα_i/o protein (Santiago and Abrol, 2019). M₁ mAChR is highly expressed in the frontal cortex and hippocampus, and reduction of its levels exacerbates AD-like pathology and cognitive decline (Shiozaki et al., 2001; Tsang et al., 2006). These findings have more recently been confirmed in animal studies. Data from experiments in 3 × Tg-AD and Tg-SwDI mice indicated that ablating M₁ mAChR promoted tau hyperphosphorylation and amyloidogenic processing, which were attributed to changes in PKC and GSK-3β activities, as well as increasing the astrocytic and microglial response associated with Aβ plaques (Medeiros et al., 2011). Consistent with this, the loss of M₁ mAChR also resulted in increased levels of brain Aβ and greater accumulation of amyloid plaques in APP/PS1 transgenic mice (Davis et al., 2010). By contrast, activation of M₁ mAChR was found to regulate Aβ neurotoxicity and tau pathology and reverse cognitive deficits through activation of PKC and inactivation of GSK-3β in a Gα_q/11-dependent manner (Farias et al., 2004; Caccamo et al., 2006). Recently, based on knowledge of M₁ mAChR, a variety of agonists (e.g., HTL9936, Brown et al., 2021) and positive allosteric modulators (e.g., VU0486846, Abd-Elrahman et al., 2021) have been designed and synthesized to alleviate the pathology of AD and improve cognitive function (Table 1).

Similar to the results obtained for M₁ mAChR, data from experimental in vitro studies indicated that M₃ mAChR might be involved in coupling with PLC-mediated hydrolysis of phosphatidylinositol-4, 5-bisphosphate and PKC activation to stimulate the release of sAPP and reduce levels of Aβ (Buxbaum et al., 1992; Nitsch et al., 1992). This suggested that the activity of M₃ mAChR was associated with amyloidogenic processing. Furthermore, M₃ mAChR may be involved in improving learning and memory in a manner dependent on receptor phosphorylation/arrestin rather than via G protein signaling (Poulin et al., 2010). Abnormalities of M2 auto-receptors in AD brains have been mainly found in the cortex and hippocampus (Mulugeta et al., 2003; Bellucci et al., 2006). Notably, ACh levels can be increased by blockade of presynaptic M2 receptors that regulate ACh release; treatment with an M2 receptor antagonist, i.e., SCH-72788, can increase ACh contents (Lachowicz et al., 2001). The role of M2 receptors in AD has been partially established using a GRK5-deletion animal model. GRK5 is a serine/threonine kinase whose dysfunction selectively impairs desensitization of presynaptic M2 receptors, causes M2 receptors hyperactivity, and inhibits ACh release, resulting in cognitive impairment and AD-like pathology (Liu et al., 2009). In APP transgenic mice, inhibition of hyperactivity of presynaptic M2 receptors by antagonist methoctramine has been shown to be an effective therapy for eliminating the increase in Aβ and tau pathology induced by GRK5 deficiency and promoting ACh release (Cheng et al., 2010; Zhang et al., 2014; Table 1). Collectively, these findings provide evidence that activation of M1 receptors and inhibition of M2 receptors have potential benefits in ameliorating neuropathological features resembling those of AD.

Exercise increases cholinergic input in the cortex and hippocampus. In the cerebral cortex, significantly increased ACh contents were observed in rats following walking for as little as 5 min (Kurosawa et al., 1993). This suggests that ACh is released in response to exercise stimulation, even if the exercise volume is not large. Teglas et al. (2019) and Xu et al. (2019) found that chronic exercise programs, i.e., months of treadmill running [40 min/session, 3 sessions/week at 60% maximal oxygen intake (VO_2max)] or voluntary wheel running (1 h/session, 5 sessions/week), attenuated age-related reduction of cholinergic fibers, reduced malformations of cholinergic forebrain innervation, and prevented the loss of cholinergic inputs in the hippocampus. Concomitant with these alterations, improvements in cognitive and motor behaviors was recorded. Similarly, Itou et al. (2011) reported that voluntary wheel running could reverse age-associated impairments of cholinergic innervation in the hippocampus of nestin promoter-GFP transgenic mice. In addition, Hall and Savage (2016) and Hall J. M. et al. (2018) observed re-emergence of the cholinergic/nestin neuronal phenotype (i.e., ChAT⁺/nestin⁺ neurons) within the medial septum/diagonal band (MS-DB) following exercise training; this improvement may in part have been mediated by nerve growth factor (NGF). These findings represent possible mechanisms by which the cholinergic system could promote cognitive function in response to exercise.

Several studies have reported regulation of the cholinergic system by exercise in AD animal models. In aged APP/PS1 transgenic mice, 4 weeks of continuous non-shock treadmill running (20–60 min/session, 5 sessions/week) improved learning and memory in association with an increased number of cholinergic neurons in the MS-DB (Ke et al., 2011). In THY-Tau22 mice, long-term voluntary wheel running for 9 months reversed pathological reduction of ChAT⁺ neurons in the MS-DB; this was accompanied by improvements in tau pathology and neuroinflammation within the hippocampus (Belarbi et al., 2011). Moreover, where comparisons were possible, the effect size for exercise was generally comparable with that of donepezil (Pisani et al., 2021). In Aβ_25–35-induced rats, both chronic aerobic and resistance exercise have similar effects to those of AChEIs, reducing AChE activity and reversing recognition memory deficits (Farzi et al., 2019). However, conflicting results have been obtained regarding changes in AChE activity after exercise intervention; some investigators found that chronic exercise could increase AChE activity altered by Aβ neurotoxicity to maintain cholinergic system activity (Schimidt et al., 2019, 2021; Dare et al., 2020). Such contradictory results clearly warrant further study.

Several researchers have described the effects of combined interventions on the cholinergic system in AD. Donepezil hydrochloride combined with swimming exercise (7 sessions/week for 4 weeks with no weight bearing) improved learning and memory in Aβ_1–40-induced rats, and the mechanism was related to increased ChAT activity and decreased AChE activity in the cortex and hippocampus (Jiangbo and Liyun, 2018). Probiotics can modulate the inflammatory process, counteract oxidative stress, and modify gut microbiota and are considered to be among the best preventive measures against cognitive decline in AD (Naomi et al., 2021). A study found that mono or combined progressive treadmill running (5 sessions/week for 8 weeks) and probiotic (e.g., Bifidobacterium bifidum and Lactobacillus plantarum) treatment significantly increased levels of ACh in the brains of Aβ_1–42-induced rats and reversed spatial learning impairment (Shamsipour et al., 2021). Moreover, combined interventions could modulate the activity of M₁ mAChR in the brain. Grape seed proanthocyanidin extract (GSPE) has been shown to have a strong antioxidant effect, can protect the central nervous system from oxidative stress damage, and may have a role in alleviating AD-related cognitive impairment (Sun et al., 2019; Gao et al., 2020). A study demonstrated that administration of GSPE and swimming training (2 h/session, 5 sessions/week for 14 weeks with 3% weight bearing) either individually or in combination led to improvements in learning and memory with reduced AChE activity in the medial prefrontal cortex and hippocampus of adult and middle-aged rats. Moreover, both mRNA and protein levels of M₁ mAChR were increased in the cortex and hippocampus, and activation of the ERK/CREB/BDNF pathway was observed following swimming training with GSPE treatment (Abhijit et al., 2017). Although other mechanisms may be involved, these findings in individual models of AD indicate the importance of regulation of the cholinergic system (especially cholinergic neurons, ChAT, AChE, ACh, and M₁ mAChR) by physical exercise.

The locus coeruleus (LC) is the norepinephrine (NE)-containing nucleus in the brainstem and innervates into widespread brain regions. It is composed of noradrenergic neurons that project to different brain regions and supplies NE to the cortex, hippocampus, striatum, amygdala, cerebellum, and basal forebrain, among others (Schwarz and Luo, 2015). The integrity of the LC-NE system is critical for attention, arousal, and specific aspects of learning and memory, and its activation across the lifespan is a crucial determinant of later-life cognitive reserve (Sara, 2009; Wilson et al., 2013; Mather and Harley, 2016). However, the LC-NE system is especially vulnerable to toxins and infection (Mather and Harley, 2016). During aging, decline of the LC-NE system is associated with reduced cognitive abilities relating to episodic memory and reduced cognitive reserve (Betts et al., 2019). Cumulative evidence suggests that aberrant tau accumulation in the LC and noradrenergic dysfunction are critical early events in the progression of AD (Mravec et al., 2016; Rorabaugh et al., 2017; Matchett et al., 2021). Aβ aggregation can also cause axonal degeneration in LC neurons (Sakakibara et al., 2021). Ablation of the LC-NE system, in turn, further exacerbates Aβ and tau pathology and the resulting cognitive deficits (Chalermpalanupap et al., 2018; Jacobs et al., 2021) and neuroinflammation (Jardanhazi-Kurutz et al., 2011; Cao et al., 2021), setting up a vicious cycle. Consequently, targeting the LC-NE system may have significant therapeutic potential in AD.

Notably, NE is widely regarded as a mediator of cognitive regulation in multiple neurodegenerative diseases, including AD (Holland et al., 2021). Animal studies suggest a progressive reduction of NE levels within the hippocampus, cortex, and cerebellum during the development of AD (Francis et al., 2012; Rorabaugh et al., 2017); these changes have been found to coincide with altered expression of BDNF and to precede the onset of cognitive and behavioral impairments (Francis et al., 2012). NE also has a central role in regulating Aβ production and Aβ-related pathologies. At the molecular level, it can interact with Aβ and inhibit Aβ generation (Liu et al., 2017; Zou et al., 2019). Furthermore, NE can reduce Aβ-induced neurotoxicity via activating tyrosine kinase receptor B (TrkB) (Liu et al., 2015), alleviate neuroinflammation by downregulating the expression of inducible nitric oxide synthase and interleukin 1β (IL-1β) (Heneka et al., 2002), and attenuate oxidative stress through limiting production of reactive oxygen species (ROS) (Jhang et al., 2014). By contrast, deficiency of NE increases the Aβ burden and activation of microglia and astroglia, and decreases expression and activity of the Aβ degrading-enzyme neprilysin (Kalinin et al., 2007). Thus, NE may improve a wide range of physiological and pathophysiological processes in AD.

To exert its neuroprotective effect, NE binds to adrenergic receptors (ARs), which can be classified into three main categories (i.e., α₁, α₂, and β) and all belong to the GPCR family. In general, the α₁-AR subtypes (i.e., α_1A, α_1B, and α_1D) couple to Gα_q/11 (Chen and Minneman, 2005) and Gα_12/13 (Maruyama et al., 2002) proteins, whereas the α₂-AR subtypes (i.e., α_2A, α_2B, α_2C, and α_2D) couple to the Gα_i/o protein (Reynolds et al., 2005). Several studies have suggested that inhibition of α₁-AR and/or α₂-AR activity may represent a new strategy for anti-Aβ therapy. Previous studies showed that treatment with an α₁-AR antagonist, i.e., prazosin or doxazosin, could reduce the generation of Aβ in N2a cells, alleviate neuroinflammation, and prevent memory deficits in APP23 transgenic mice (Katsouri et al., 2013), as well as protecting hippocampal slices from Aβ neurotoxicity through prevention of GSK-3β activation and tau hyperphosphorylation in an in vitro model of AD (Coelho et al., 2019). Similar to α₁-AR, treatment with α₂-AR antagonists including fluparoxan (Scullion et al., 2011), dexefaroxan (Francis et al., 2012), and mesedin (Melkonyan et al., 2017) was found to be beneficial for improving AD-like pathological mechanisms (Table 2). In addition, increased activity of α_2A-AR was observed in AD patients and mouse models (Zhang F. et al., 2020). APP enhances the surface retention of α_2A-AR and the intensity and duration of its signaling (Zhang et al., 2017); then, α_2A-AR reduces Golgi localization of APP and concurrently promotes APP distribution in endosomes and cleavage by BACE1 (Chen et al., 2014). It follows that α_2A-AR could serve as a key biological target for Aβ generation. Moreover, α_2A-AR may represent a bridge between Aβ and tau pathology. One study demonstrated that oAβ bind to an allosteric site on α_2A-AR to redirect NE-elicited signaling and thus to increase tau hyperphosphorylation, which depends on GSK-3β signaling (Zhang F. et al., 2020). The aforementioned results indicate that activation of α_2A-AR may aggravate the pathological development of AD. It is necessary to further verify whether its genetic or pharmacological blockade could reduce Aβ production and Aβ-related neuropathology.

β-ARs are crucial targets for increasing synaptic plasticity and maintaining learning and memory (Goodman et al., 2021); however, the total number of β-ARs is significantly reduced in AD brains (Kalaria et al., 1989). NE may exert pleiotropic neuroprotective effects via its action on β-AR signaling in various cell types. Activation of β-AR signaling may be responsible for the activation of CREB and the induction of NGF and BDNF to protect against Aβ neurotoxicity in neurons (Counts and Mufson, 2010). In microglia-like cells, NE suppresses Aβ-induced toxicity and production of monocytic chemotactic protein-1, a pro-inflammatory chemokine, through activating β-AR signaling, accompanied by activation of the cAMP/PKA/CREB pathway (Yang et al., 2012). Furthermore, as shown in Table 2, accumulating evidence from pharmacological intervention studies confirms that β-ARs mediate distinct functions related to different aspects of AD pathology. Some previous studies, for instance, have reported that application of isoproterenol, a non-selective β-AR agonist, worsened Aβ and tau pathology; however, other studies have demonstrated the opposite effect. Hence, the molecular mechanism of the differential effect of the same β-AR agonist in different models of AD needs to be studied. The regulatory roles of different β-AR subtypes in AD pathology should also be considered.

There are three β-AR subtypes (β₁, β₂, and β₃) in the brain that couple to the Gα_s protein (Liu et al., 2019; Lee et al., 2020; Su et al., 2020). Recently, the contribution of the β₁- and β₃-ARs to AD pathology and cognitive function has been a subject of interest. Animal studies have revealed that selective activation of β₁-AR activates the cAMP/PKA/CREB pathway to rescue social memory deficit in APP mice (Coutellier et al., 2014), as well as inhibiting the expression of neuroinflammatory markers (e.g., ionized calcium binding adapter molecule 1) and reducing Aβ and tau pathology in 5 × FAD mice (Ardestani et al., 2017). Moreover, treatment with CL-316243, a β₃-AR agonist, could reduce Aβ pathology and reverse memory loss (Gibbs et al., 2010; Tournissac et al., 2021; Table 2). Therefore, the current consensus is that pharmacological activation of β₁- and β₃-ARs is beneficial in AD. However, prior research on β₂-AR in AD resulted in seemingly contradictory findings. Some studies showed that activation of β₂-AR accelerated amyloid plaque formation, and that this beneficial effect could be reversed by antagonist ICI-118551 (Ni et al., 2006; Yu et al., 2010). Conversely, other studies have demonstrated that treatment with highly selective β₂-AR agonists such as clenbuterol can reduce Aβ levels, promote hippocampal neurogenesis, enhance synaptic plasticity, and improve neuronal death and microglial inflammation (Chai et al., 2016, 2017), whereas application of ICI-118551 exacerbates Aβ and tau neuropathology and cognitive deficits (Branca et al., 2014; Wu et al., 2017). Further studies are warranted to confirm and explain these contradictory observations.

Previous studies in animals have shown that central noradrenergic neurons are activated in response to exercise training and participate in thermoregulation during exercise (Ohiwa et al., 2006; Rodovalho et al., 2020). Acute running increases noradrenergic activity, and the longer the running time, the longer the duration of central activation in the recovery period (Pagliari and Peyrin, 1995a,b). In the course of chronic exercise intervention, rodents with long-term exercise training experience showed significant increases (approximately 26%) in levels of NE in the brain (Ostman and Nyback, 1976). Consistent with this, progressive treadmill running for 8 weeks in rodents was accompanied by brain noradrenergic adaptations and increases in NE levels in the areas of NE cell bodies and the spinal cord (Dunn et al., 1996). Based on this information, a single bout of exercise appears to temporarily increase central noradrenergic activity, and the cumulative effect of long-term regular exercise leads to a significant increase in the levels of NE.

Salivary α-amylase (sAA), a non-invasive biomarker of central noradrenergic activity, is a promising avenue for characterizing the arousal-mediated effects of exercise on cognition (Weiss et al., 2019). A study demonstrated that 6 min of stationary bicycle exercise at 70% VO_2max significantly enhanced memory consolidation in both patients with amnestic MCI and cognitively normal individuals through activating the noradrenergic system (as determined by measuring sAA) (Segal et al., 2012). In another study, patients with MCI who underwent a chronic mind-body exercise program [60 min/session, 3 sessions/week for 24 weeks at 55–75% heart rate reserve (HRR)] showed a significant increase in intrinsic functional connectivity in the LC-NE system and improvements in cognitive performance, as measured by magnetic resonance imaging (MRI) scans and the Montreal Cognitive Assessment (Liu et al., 2021). These findings suggest that activation of the noradrenergic system by exercise improves cognitive performance of individuals in the prodromal stage of AD.

Studies in animal models found that exercise could also effectively reduce the activity of α₂-AR and increase the activity of β-ARs. Spontaneously-Running-Tokushima-Shikoku rats, an animal model for high levels of wheel-running activity, showed decreased hippocampal monoamine oxidase A levels and increased extracellular NE levels, and the elevation of NE levels caused downregulation of α₂-AR (Morishima et al., 2006). Furthermore, the affinity of this receptor in the nucleus tractus solitarius was reduced in trained rats compared with sedentary animals (De Souza et al., 2001). These results indicate that long-term exercise may lead to reduced affinity of α₂-AR in multiple brain regions. Considering the anti-AD effect produced by inhibiting α₂-AR, the potential of α₂-AR-mediated exercise to improve AD deserves future exploration. In addition to the findings for α₂-AR, Ebrahimi et al. (2010) confirmed that exercise enhanced learning and memory through β-AR-dependent pathways by administering propranolol to mice. Indeed, previous studies have proposed that an intact noradrenergic system, especially activation of β-ARs by NE, serves as a vital link in the upregulation of BDNF expression by exercise (Garcia et al., 2003; Ma, 2008). BDNF can induce the expression of thioredoxin-1 (TRX-1) via the TrkB/Akt/CREB pathway (Bai et al., 2019). TRX-1 is a disulfide-reducin-system low-molecular-weight protein with redox properties, levels of which are significantly reduced in AD brains (Akterin et al., 2006). Increased TRX-1 levels can alleviate endoplasmic reticulum stress, oxidative stress, and apoptosis in AD (Guo et al., 2021). Recently, an experimental study showed that a treadmill running program (60 min/session, 6 sessions/week for 3 weeks) increased the content of TRX-1 in the hippocampus of mice and activated the ERK1/2/β-catenin/T-cell factor pathway, which in turn promoted hippocampal cell proliferation and neurogenesis (Kim and Leem, 2019). Notably, Kim and Leem (2019) proposed a hypothesis regarding the signaling that links exercise, β₂-AR, BDNF, and TRX-1; that is, exercise may promote the expression and interaction of BDNF and TRX-1 through activating the β₂-AR/cAMP/PKA pathway. A study further to test this hypothesis subjected an animal model of cognitive impairment induced by a high-fat diet to a treadmill running program (30 min/session, 5 sessions/week for 23 weeks at 40–50% VO_2peak); activation of the β₂-AR/cAMP/PKA pathway, increased expression of TRX-1 and BDNF, inhibition of microglial activation, decreased expression of inflammatory markers, and reduction of oxidative stress markers in the dentate gyrus of the hippocampus were observed (Han et al., 2019). These results suggest that exercise alleviates neuroinflammation and oxidative stress potentially through a signaling cascade involving β₂-AR, BDNF, and TRX-1. However, the relationship among the these three factors needs to be validated in AD animal models and patients.

Serotonergic neurotransmission is dependent on the synthesis and release of the neurotransmitter serotonin [5-hydroxytryptamine (5-HT)], and the serotonergic projections from the dorsal raphe nucleus (DRN) have widespread ramifications throughout the brain, including the frontal cortex, temporal cortex, and hippocampus (Vertes, 1991; Vertes et al., 1999). Aging exerts complex effects on the central serotoninergic system. Impaired serotonergic neurotransmission and altered expression of 5-HT transporter (5-HTT) and 5-HT receptors (5-HTRs) have been observed in multiple brain regions, although the number of serotonergic neurons did not change significantly (Rodriguez et al., 2012). Several studies have demonstrated that dysfunction of the serotonergic system is linked to the development of AD pathology (Rodriguez et al., 2012; Tajeddinn et al., 2016b; Jankowska et al., 2018). The number of serotonergic neurons in the DRN (Aletrino et al., 1992; Lyness et al., 2003) and the contents of 5-HT, 5-HTT, and its metabolite 5-hydroxyindoleacetic acid (5-HIAA) in the cortex and hippocampus have been shown to be significantly reduced in AD brains (Palmer et al., 1987; Thomas et al., 2006; Vermeiren et al., 2014). Furthermore, lower concentrations of 5-HT in cerebrospinal fluid (Lanctot et al., 2002) and platelets (Prokselj et al., 2014; Tajeddinn et al., 2016a) have been observed in patients with AD compared with controls. Treatment studies aim to increase serotonergic tone; selective serotonin reuptake inhibitors including escitalopram (Ehrhardt et al., 2019), citalopram (Porsteinsson et al., 2014), and fluoxetine (Xie et al., 2019) have been found to have beneficial effects on psychiatric symptoms and cognitive impairment in patients with AD.

Increasing attention has been paid to the functions of 5-HTRs and their impact on the pathophysiology of AD. In general, 5-HTRs constitutes seven subfamilies, which can be split into a total of 14 subtypes (i.e., 5-HT_1A–1FR, 5-HT_2A–2cR, 5-HT₃R, 5-HT₄R, 5-HT_5A–5BR, 5-HT₆R, and 5-HT₇R). Except for 5-HT₃R, these receptors belong to the GPCR family (Sharp and Barnes, 2020). The 5-HT₁R and 5-HT₅R subfamilies couple to the Gα_i/o protein and mainly inhibit PKA signaling (Francken et al., 2000; Rojas et al., 2017). 5-HT₁R members are expressed in large quantities in the hippocampus and have a significant role in the regulation of memory processes (Ogren et al., 2008). 5-HT_1AR is a well-studied member of this subfamily and shows overexpression under Aβ stimulation (Verdurand et al., 2011, 2016). Treatment with 5-HT_1AR antagonists (e.g., NAD-299 and WAY-100635) has been shown to reduce amyloid plaque deposition, increase levels of hippocampal BDNF, alleviate neuroinflammation and oxidative stress, and improve cognitive deficits in individual animal models of AD (Afshar et al., 2018, 2019; Wang et al., 2020; Table 3). These results imply that 5-HT_1AR, in response to specific ligands, is involved in the regulation of AD pathology through multiple pathways.

Unlike the above two subfamilies, members of the 5-HT₂R subfamily couple to the Gα_q/11 protein (Odagaki et al., 2017) and are strongly associated with Aβ pathology, especially 5-HT_2AR and 5-HT_2CR. An imaging study has revealed prominent reductions in neocortical 5-HT₂R in patients with AD (Blin et al., 1993). The binding of 5-HT_2AR is significantly decreased in the brains of AD animal models (Holm et al., 2010) and human patients (Lai et al., 2005; Marner et al., 2012). Molecular biochemistry studies have disclosed that 5-HT could induce the release of sAPP through activation of 5-HT_2AR and 5-HT_2CR (Nitsch et al., 1996). Although there is evidence to suggest that the 5-HT_2AR and 5-HT_2CR modulate sAPP secretion in vitro and in vivo (Table 3), further studies are required to determine whether this effect is mediated by a change in α- or β-secretase activity and whether this effect correlates with a change in Aβ generation. Furthermore, studies have found that administration of 5-HT_2AR-selective ligands stimulated the autophagy process of microglia, enhanced the phagocytosis of Aβ (Lu et al., 2021), and reduced amyloid plaques (Afshar et al., 2018). Hence, 5-HT_2AR may also be a key target for regulating Aβ clearance and degradation.

The remaining subfamilies (5-HT₄R, 5-HT₆R, and 5-HT₇R) couple to the Gα_s protein and activate PKA signaling (Fisher J. R. et al., 2016). 5-HT₄R and 5-HT₆R are mainly located in brain regions involved in cognitive processes (King et al., 2008). Neurochemical and behavioral studies have demonstrated that activation of 5-HT₄R or blockade of 5-HT₆R improved cognitive performance (Quiedeville et al., 2015). Recently, 5-HT₄R agonists and 5-HT₆R antagonists have attracted interest with respect to AD treatment and have been widely investigated from a drug discovery perspective. Studies in vitro and in vivo suggest that activation of 5-HT₄R by agonists (e.g., ML10302, prucalopride, and RS-67333) increases levels of neuroprotective sAPPα, reduces amyloid plaque deposition, and rescues cognitive deficits through the classical cAMP/PKA pathway and non-classical pathway. In addition, inhibition of 5-HT₆R by antagonists (e.g., SB-258585, SB-399885, and SB-271046) can inhibit Aβ generation, and protect neurons from Aβ-induced neurotoxicity, neuroinflammation, oxidative stress, and apoptosis (Table 3). Preclinical and early clinical studies of 5-HT₄R agonism and 5-HT₆R antagonism are being conducted to investigate the ability of these approaches to alleviate cognitive deficits associated with AD (Ivachtchenko et al., 2016; Nirogi et al., 2021). 5-HT₇R, the most recently discovered receptor of 5-HT, has been shown to regulate individual cognition (Gasbarri and Pompili, 2014). Chronic treatment with AS-19, a selective 5-HT₇R agonist, could prevent cognitive deficits by alleviating Aβ plaque accumulation and neuronal apoptosis and improving neuronal plasticity (Hashemi-Firouzi et al., 2017; Shahidi et al., 2018). Furthermore, a recent study revealed that 5-HT₇R is an important target for the treatment of tauopathy (Labus et al., 2021). Collectively, 5-HT₄R, 5-HT₆R, and 5-HT₇R may represent novel therapeutic targets for the treatment and prevention of AD.

The positive effects of physical exercise on emotional and cognitive performance through activation of the serotonergic system can be summarized as follows. First, exercise enhances the function of the serotonergic system via increases in numbers and activity of serotonergic neurons in the DRN. Treadmill running exercise, whether performed acutely for 30 min or chronically for 3 or 8 weeks, increased the number and activity of serotonergic neurons in the DRN of rodent models (Hong et al., 2015; Otsuka et al., 2016; Ge and Dai, 2020). In particular, 4 weeks of progressive treadmill running (20–60 min/session, 5 sessions/week) selectively improved spatial learning and memory in association with an increase in numbers of serotonergic neurons in the DRN of aged APP/PS1 transgenic mice (Ke et al., 2011); this exercise paradigm also reduced Aβ levels and abnormal microglia activation, but not enough to reduce the plaque loading in the hippocampus.

Second, exercise increases the availability of 5-HT precursor tryptophan (TRP) and TRP hydroxylase (TPH), and increases levels of 5-HT and 5-HTAA. Studies in humans suggest that a bout of exercise, i.e., 35 min of graded exercise or 60 min of treadmill exercise, can increase serum TRP and 5-HT levels and exert pro-cognitive and antidepressant effects (Melancon et al., 2012; Zimmer et al., 2016). Additional studies found that 5-HT released during yoga and meditative practices activated an alternate cleavage of APP to produce a fragment with known neurotrophic effects, giving it the unique ability to inhibit the oAβ production cycle in an in vitro AD model (Hassan et al., 2018). Data from animal studies indicate that exercise significantly increases the synthesis and metabolism of central 5-HT (Dey et al., 1992), and is required for the exercise-induced neurogenic response, especially adult hippocampal neurogenesis (Klempin et al., 2013, 2018). Furthermore, saffron combined with endurance exercise increased levels of hippocampal 5-HIAA, and this change was associated with improved short-term memory (Akbari-Fakhrabadi et al., 2021). Regulation of central 5-HT synthesis and metabolism by exercise may represent a therapeutic opportunity in depression and age-related cognitive decline. Isolation stress accelerates the onset of AD (Huang et al., 2015; Peterman et al., 2020), reducing TPH and 5-HT expression in the DRN and promoting apoptosis in the hippocampus, leading to anxiety and memory decline during old age, whereas a swimming exercise program (30 min/session, 5 sessions/week for 4 weeks) reversed these changes (Park et al., 2020). Moreover, three different forms of exercise (i.e., treadmill exercise, involuntary exercise and voluntary exercise, 5 sessions/week for 4 weeks) could all improve cognitive and behavioral functions by increasing levels of hippocampal 5-HT in a vascular dementia rat model (Zhang L. et al., 2020).

Third, exercise can regulate the expression of the serotonergic receptor and the activity of its downstream signaling pathway to improve impaired cognitive ability and abnormal emotion. Notably, as the hippocampal neurons mainly express type II AC, which is not regulated by the Gα_i/o subunit but is activated by the Gβγ subunit, the cAMP/PKA pathway is activated by 5-HT_1AR (Albert and Vahid-Ansari, 2019). Exercise improves cognitive function by increasing the expression of 5-HT_1AR in hippocampal neurons, which is potentially associated with BDNF-related signaling. Results from several animal studies are consistent with this notion. For instance, Kim et al. (2015) reported that 4 weeks of treadmill running (30 min/session, 7 sessions/week) enhanced CREB phosphorylation and increased the expression of BDNF and TrkB via activation of 5-HT_1AR in rat hippocampus neurons. Consistent with this, the improvements in learning and memory after rats underwent a chronic and progressive treadmill running program (30–60 min/session, 3 sessions/week for 14 weeks at 60–70% VO_2max) were found to be due to increased expression of 5-HT, 5-HTT, 5-HT_1AR, and BDNF in the hippocampal CA1 area (Pietrelli et al., 2018). In addition, treadmill exercise training (45 min/session, 3 sessions/week for 32 weeks) could ameliorate anxious/depressive-like behavior and attenuate fear-avoidance behavior deficits in TgF344-AD rats in the early stage of Alzheimer’s pathogenesis by increasing the expression of 5-HT and 5-HT₆R in the cortex and hippocampus (Wu et al., 2020). Existing studies on the regulation of individual cognition by exercise through the regulation of serotonergic receptor function are still lacking, and more detailed evidence is necessary.

Dopamine (DA) is a major catecholamine neurotransmitter that projects dopaminergic signaling to the prefrontal cortex, hippocampus, striatum, nucleus accumbens, amygdala, and other areas from the ventral tegmental nucleus area (VTA) of the midbrain; it is mainly involved in emotion, behavior, and cognition, and in regulation of synaptic plasticity (Jay, 2003; Bjorklund and Dunnett, 2007). Aging is associated with a loss of dopaminergic function, which may originate from defects on multiple components, including loss of dopamine-producing neurons, atrophy of projection brain regions, and reduced density of dopamine receptors. These alterations result in the efficiency of dopaminergic projecting systems declines slowly during physiological aging (Ciampa et al., 2021; Gasiorowska et al., 2021). Cumulative evidence suggests that impaired dopaminergic neurotransmission is also involved in the pathological development of a variety of neurological disorders, including AD (Martorana and Koch, 2014; D’Amelio et al., 2018). In particular, loss of dopaminergic neurons in the VTA and/or substantia nigra pars compacta (SNpc) (Mann et al., 1987; Nobili et al., 2017) and significantly decreased levels and availability of DA in the cortex and hippocampus have been observed in AD animal models and human patients, where they led to severe synaptic dysfunction and cognitive deficits (Reinikainen et al., 1988; Trabace et al., 2007). In a recent animal study, Aβ decreased cortical DA levels and caused profound impairment of both long-term potentiation (LTP) and long-term depression, as well as recognition memory (Moreno-Castilla et al., 2016). In the Tg2576 mouse model of AD, dopaminergic neuron loss was shown to begin before Aβ plaque formation, resulting in reduced hippocampal DA outflow, which decreased neuronal synaptic plasticity and excitability and contributed to memory and reward dysfunction; nevertheless, these defects could be partially reversed with DA precursor levodopa (L-DOPA) supplementation (Nobili et al., 2017; Cordella et al., 2018). Thus, increasing DA levels moderately may ameliorate synaptic dysfunction and cognitive decline in AD.

In addition to improvements in neuroplasticity, the effects of DA on AD neuropathology include disruption of Aβ protofibril and inhibition of Aβ aggregation, as well as partial alleviation of neuroinflammation, and oxidative stress. On the one hand, DA can inhibit Aβ aggregation and disrupt Aβ fibrils in a dose-dependent manner (Huong et al., 2010; Liu et al., 2016). Recently, Chen et al. (2021) reported that DA disrupted Aβ protofibrils and prevented Aβ dimerization at the molecular level mostly through π-π stacking interactions with residues F4, H6, and H13; hydrogen-bonding interactions with negatively charged residues D7, E11, E22, and D23; and cation-π interactions with residue R5. This may be an important mechanism by which DA interferes with Aβ generation. On the other hand, DA and its derivatives significantly diminish neuroinflammation and oxidative stress triggered by lipopolysaccharides (LPS) and Aβ through decreasing levels of inflammatory mediators and upregulating expression of heme oxygenase-1, the enzyme responsible for production of antioxidants (Nam et al., 2018). Further research should determine the optimal dose-effect relationship for DA regulation of AD-like pathology.

Dopaminergic receptors, which are widely distributed in various brain regions, belong to the GPCR family and can be divided into D1-like receptors (i.e., DA₁R and DA₅R) and D2-like receptors (i.e., DA₂R, DA₃R, and DA₄R) according to their biological and pharmacological properties (Beaulieu and Gainetdinov, 2011). D1-like receptors are widely distributed in the brain, couple to Gα_s and Gα_q/11 proteins (Himmelreich et al., 2017), and regulate PKA and PLC signaling, whereas D2-like receptors couple to the Gα_i/o protein (Conley and Watts, 2013). Interestingly, a recent systematic review and network meta-analysis indicated that DA₁R and DA₂R levels were decreased in patients with AD compared with controls; the dopaminergic receptors were ranked as follows according to their correlation with AD from highest to lowest: DA₂R, DA₃R, DA₄R, DA₅R, and DA₁R (Pan et al., 2019). As shown in Table 4, application of D1-like receptor agonists (e.g., L-stepholidine, L-theanine, and SKF-38393) reduced Aβ and tau pathology and significantly improved synaptic dysfunction and cognition. These results are largely consistent with the activation of PKA signaling. Furthermore, in vitro and in vivo experiments confirmed that application of selective DA₁R agonist A-68930 significantly ameliorated neuroinflammation and mitochondrial dysfunction through adenine monophosphate activated protein kinase (AMPK)-related signaling pathways (Cheng et al., 2020, 2021). Together, these results indicate that activation of D1-like receptors represents an important strategy for prevention and treatment of AD-like pathology.

Accumulating evidence over the past decade suggests that D2-like receptor agonists can prevent multiple pathological features found in AD. For instance, rotigotine, a DA_2/3R agonist, could increase cortical excitability and restore central cholinergic transmission in patients with AD (Martorana et al., 2013) and reduce symptoms associated with frontal lobe cognitive dysfunction, thereby delaying impairment of activities of daily living (Koch et al., 2020). However, distinct ligands may allow D2-like receptors to exert a dual role in pathobiological activity of Aβ and tau. In vitro, both L-DOPA and piribedil promoted the generation of Aβ and increased the activity of γ-secretase, mediated by the activation of the DA₂R and β2-arrestin signaling pathway in neuronal cells (Lu et al., 2017), whereas pretreatment with higher concentrations of DA₂R agonist quinpirole protected neurons from Aβ toxicity (Melief et al., 2015). In addition, Koppel et al. (2016) reported that a tau mouse model of AD treated with DA₂R antagonist haloperidol showed a significant reduction in tau phosphorylation associated with an inactivation of the tau kinase AMPK, whereas a study by Ziu et al. (2020) suggested an effective role of DA_2/3R agonists in inhibiting tau aggregation. The underlying mechanism of this dual effect requires further verification.

The effects of exercise on the dopaminergic system in PD patients and animal models has been widely reported (Hou et al., 2017; Sacheli et al., 2019). The dopaminergic system is also known to be involved in the effects of exercise on AD. Studies have shown that exercise can increase the contents of DA in the hippocampus. Using a microdialysis technique, Goekint et al. (2012) found that a bout of treadmill running for 60 min induced a twofold increase in hippocampal DA release in rats. In agreement with this, 30 min of treadmill running at 60–70% VO_2max rather than strength exercise ameliorated Aβ neurotoxicity by increasing hippocampal DA levels and promoted recognition learning in Aβ-induced rats (Rossi Dare et al., 2020). A chronic exercise intervention program consisting of 4 weeks of treadmill running (60 min/session, 5 sessions/week) before intraperitoneal LPS injection prevented LPS-induced loss of dopaminergic neurons in the SNpc, reduction in DA levels, and dysfunction of motor coordination (Wu et al., 2011). Furthermore, after swimming exercise for 4 weeks (30 min/session, 7 sessions/week), increases in the DA contents of the brain were found to be associated with improvements in learning and memory in AD rats induced by LPS, with the best effects for combined vitamin D and exercise treatment (Medhat et al., 2020). Thus, exercise appears to markedly improve LPS-induced cognitive and motor deficits by rescuing dysfunction of the dopaminergic system.

D1-like receptors are important participants in the improvements in cognition that occur under exercise stimuli. One study reported promotion of the persistence of object recognition memory and induction of the release of DA in the hippocampus as a result of 30 min of treadmill running at an intensity of 60–70% VO_2max, whereas this effect was blocked by treatment with D1-like receptor antagonist SCH-23390 (Vargas et al., 2020). A recent study by Ramires Lima et al. (2021) identified specific mechanisms and found that a similar protocol activated D1-like receptors in rat hippocampus and improved memory persistence; however, the administration of SCH-23390 or inhibition of PKA but not PKC impaired the effect of acute aerobic exercise on memory persistence. In another study, 1 month of voluntary wheel running was shown to activate the DA₁R/cAMP/PKA pathway, induce differentiation of hippocampal neurons, and enhance neurogenesis via the AMPK/CREB pathway in mice (Zang et al., 2017). Thus, signaling events mediated by PKA are critical for exercise to improve cognition. However, the effects of exercise on cognition mediated by D1-like receptors need further validation in models of AD in the future.

In AD brains, expression of D2-like receptors has been shown to be reduced in the cortex, striatal, and hippocampus regions (Pizzolato et al., 1996; Kemppainen et al., 2003; Kumar and Patel, 2007). Although research has proven that regular exercise can increase DA₂R levels and improve dopaminergic signaling (Bauer et al., 2020), the relationships among exercise, DA, D2-like receptors, and cognition are not understood in sufficient detail. Several human studies offer some insight. In a cross-sectional study, the intensity of habitual physical activity of elderly individuals was found to be positively correlated with episodic memory and the availability of DA_2/3R in the striatum, but the frequency of physical activity was not related to the availability of DA_2/3R (Kohncke et al., 2018). In another study, elderly participants underwent an aerobic exercise intervention for 6 months, which led to significant increases in DA contents and improvements in working memory, compared with an active control; unexpectedly, DA₂R levels decreased with exercise, and there was no relationship between DA₂R and working memory at baseline or following exercise (Jonasson et al., 2019). Thus, the regulation of cognitive function by exercise through D2-like receptors is complicated, and it is still difficult to determine whether D2-like receptors are involved in the improvements in cognitive function of AD patients that are associated with exercise.