Repetitive transcranial magnetic stimulation in Alzheimer’s disease: effects on neural and synaptic rehabilitation

Abstract

Alzheimer’s disease is a neurodegenerative disease resulting from deficits in synaptic transmission and homeostasis. The Alzheimer’s disease brain tends to be hyperexcitable and hypersynchronized, thereby causing neurodegeneration and ultimately disrupting the operational abilities in daily life, leaving patients incapacitated. Repetitive transcranial magnetic stimulation is a cost-effective, neuro-modulatory technique used for multiple neurological conditions. Over the past two decades, it has been widely used to predict cognitive decline; identify pathophysiological markers; promote neuroplasticity; and assess brain excitability, plasticity, and connectivity. It has also been applied to patients with dementia, because it can yield facilitatory effects on cognition and promote brain recovery after a neurological insult. However, its therapeutic effectiveness at the molecular and synaptic levels has not been elucidated because of a limited number of studies. This study aimed to characterize the neurobiological changes following repetitive transcranial magnetic stimulation treatment, evaluate its effects on synaptic plasticity, and identify the associated mechanisms. This review essentially focuses on changes in the pathology, amyloidogenesis, and clearance pathways, given that amyloid deposition is a major hypothesis in the pathogenesis of Alzheimer’s disease. Apoptotic mechanisms associated with repetitive transcranial magnetic stimulation procedures and different pathways mediating gene transcription, which are closely related to the neural regeneration process, are also highlighted. Finally, we discuss the outcomes of animal studies in which neuroplasticity is modulated and assessed at the structural and functional levels by using repetitive transcranial magnetic stimulation, with the aim to highlight future directions for better clinical translations.

Introduction

Alzheimer’s disease (AD) is a progressive disorder characterized by cognitive impairments including executive function deficits and continuous loss of memory (Unnisa et al., 2023). The hallmarks of AD are the accumulation of intracellular hyperphosphorylated tau protein provoking the formation of neurofibrillary tangles, and extracellular amyloid-beta (Aβ) plaques which may, in turn via synaptic dysfunction (Franzmeier et al., 2020), precipitate the propagation of tau pathology. Aβ can not only bring about notable synaptic loss in the hippocampal region (Pereira et al., 2021), involving apoptosis or necroptosis (Chi et al., 2018), but can also activate glial cells in the AD brain, inducing a chronic inflammatory response by releasing neurotoxic cytokines (Hampel et al., 2021). Oligomeric Aβ also induces abnormal excitability as evidenced by the hyperexcitation of sensitive neurons in mouse models of amyloidosis (Zott et al., 2019). Resting motor threshold (RMT) is a neurophysiological parameter that mirrors the cortical excitability of the primary motor cortex region (M1), and studies on patients with AD have shown that RMT decreases significantly at the pre-clinical stage of AD, indicating the role of increased cortical excitability (Mimura et al., 2021). Activity-dependent synaptic plasticity, including long-term potentiation (LTP) and long-term depression (LTD) in the hippocampus, is acknowledged as a cellular correlate of learning and memory and is significantly impaired in patients with AD (Fá et al., 2016). A study investigating multiple cognitive domains in 75 patients showed that the extent of damage to LTP-like cortical plasticity was correlated with the inefficiency of long-term and working memory (Li et al., 2021b). Furthermore, oligomeric Aβ (Lei et al., 2016) or amyloid precursor proteins (APP) (Martín-Belmonte et al., 2020) are detrimental to hippocampal LTP and cortical excitation and inhibition balance (E/I balance) by disturbing glutamatergic and gamma-aminobutyric acid (GABA)-receptor dependent function. Consequently, neuronal apoptosis, inflammatory responses, hyperexcitability, reduced cortical plasticity, glutaminergic excitotoxicity, and impairment of inhibitory systems are critical components of AD pathogenesis in the neurodegenerative cascade.

Repetitive transcranial magnetic stimulation (rTMS) is a non-invasive tool that delivers rhythmic, repeated trains of magnetic stimuli through the scalp, creating a magnetic field that enables temporary modulation of neuronal activity and plasticity in targeted cortical tissue (Xie et al., 2021; Huang et al., 2022; Moretti and Rodger, 2024). However, the magnetic field created by the coil is transient and decays exponentially, which explains why only superficial cortex tissue (approximately 2 cm depth) under the focal region can be directly activated. To resolve the problem of limited stimulation depth, one approach is the use of deep H-coils, which enables bilateral hubs in the lateral prefrontal cortex (PFC) to be stimulated directly and simultaneously, because it allows the penetration depth of the magnetic field to reach up to 3 cm, slowing the drop-off in magnetic field intensity (Roth et al., 2007; Deng et al., 2013). Generally, high-frequency rTMS (HF-rTMS, > 5 Hz) results in a long-lasting increase in cortical excitability in a manner similar to that of LTP while low-frequency rTMS (LF-rTMS) seems to exert LTD-like effects (Hoppenrath et al., 2016). Therefore, capitalizing on rTMS may allow clinical researchers to modulate cognitive and motor functions and explore a potential new strategy to ameliorate the symptoms of neuropsychiatric diseases such as AD (Lefaucheur et al., 2020), Parkinson’s disease (PD) (Li et al., 2022), and depression (Blumberger et al., 2018). In patients with AD, rTMS significantly prevents neuronal loss, enhances functional integration, and strengthens the dynamic connectivity between different brain regions (Liu et al., 2022). Amyloid deposition, a mainstream hypothesis, can precede the onset of clinical symptoms; therefore, the development of drugs that target Aβ peptides is one of the leading therapeutic concepts. Although clinical trials have rarely shown their effectiveness or exacerbated the deterioration of cognition (Honig et al., 2018; Egan et al., 2019; Selkoe, 2019), rTMS may be an early intervention to inhibit the production of these pathogenic proteins with enhanced cognitive function (Chen et al., 2020b). Alternatively, probing the influence of rTMS on Aβ elimination via a clearance system such as astrocytic and microglial uptake (Knopman et al., 2021), meningeal lymphatics (Da Mesquita et al., 2018), glymphatic pathway (Rasmussen et al., 2018), and blood–brain barrier (Nation et al., 2019) may provide alternative explanations for neural rehabilitation. Microglia mediate the identification, pruning, formation, and stability of synapses, connections, and processes associated with synaptic structural plasticity. They also enhance memory quality and strengthen remote memory (Cornell et al., 2022), potentially playing a key role in rTMS-induced cognitive enhancement. Given the pathological distinctions of AD, rTMS exhibits neuroprotective properties against synaptic loss by influencing the proliferation and apoptosis of different cells (Zhao et al., 2019). Furthermore, one of the most pivotal mechanisms underlying rTMS as an add-on treatment for AD is its ability to induce structural plasticity (i.e., the formation of new spines and enlargement of axon terminals) and functional plasticity, which is usually accompanied by changes in neural excitability, neurotransmitter levels, and the number of receptors on the post-synaptic membrane. Therefore, targeting structural or functional plasticity may be of great significance for neuroprotection and neurorestoration.

Although there have been studies (Chou et al., 2020; Lefaucheur et al., 2020) on the clinical effects of rTMS on behavior after treatment, its neural mechanisms remain to be elucidated. This review aimed to explore the impact of rTMS on synaptic structural and functional plasticity from a neurophysiological perspective. In this review, we delve into the various neurobiological effects of rTMS, such as Aβ clearance, glycogenesis, neurogenesis, and apoptosis as well as its influence on macroscopic morphometric changes.

Search Strategy

In this narrative review, we used the following key words to search on PubMed for articles published until 2023: Alzheimer’s disease, repetitive transcranial magnetic stimulation, Aβ, glial cells, apoptosis, gene expression, neurotrophic factors, miRNA, structural plasticity, and functional plasticity. Inclusion criteria: (1) original articles or reviews in English; (2) rTMS studies conducted in original or modified animal models of AD to explore its effects on pathological alterations; (3) rTMS studies with different stimulation targets presenting the effects on the cognitive functioning in patients with AD. Exclusion criteria: (1) conference abstracts and theoretical papers; (2) non-English publications; (3) studies using single pulse TMS; (4) studies using TMS-EEG only for measurement instead of treating; and (5) studies including patients on medications. A search for additional articles included references to those selected papers. After reviewing titles and abstracts, 173 papers were included in the final analysis.

Effects of Repetitive Transcranial Magnetic Stimulation on Pathological Alterations

Aβ generation and deposition

Exploring how rTMS promotes AD rehabilitation by blocking Aβ neurotoxicity is one of the most important prerequisites for gaining a deeper insight into the underlying substrates. Several rat AD models with different characteristics have been successfully generated and are useful adjuncts to human studies as they mimic various cognitive and pathological alterations discovered in AD. Commonly used models of cerebral β-amyloidosis include APP/PS1, 3×Tg, and 5×FAD transgenic mice. APP23 transgenic mice show overproduction of total Aβ and first develop deficits in spatial memory at 3 months, which worsen with age. As for the 5×FAD model, it was designed to accelerate amyloid deposition in vivo and overexpresses two important human proteins harboring familial AD (FAD) mutations: the human APP 695 isoform with Swedish (K670N/M671L), Florida (I716V), and London (V717I) mutations and the human presenilin-1 (PS1) with M146L and L286V mutations (Han et al., 2011). 5×FAD mice are particularly suitable to explore Aβ generation during the AD course as they exhibit evident amyloid and gliosis at an early age, which subsequently leads to neurodegeneration and neuronal loss (Oakley et al., 2006). Many aspects of their cognitive abilities and motor functions are also progressively impaired after 3 months. These models recapitulate sporadic AD-like pathology and therefore, allow us to explore the mechanisms of the disease, while evaluating the effectiveness and safety of rTMS before simultaneously translating it to clinical use.

In a study using a 5×FAD transgenic mouse model of presenile AD, it was discovered that 20 Hz rTMS exerted an inhibitory effect on the deposition of Aβ and decreased the levels of intraneuronal and plaque-like Aβ deposits (Lin et al., 2021). These results were similar to those of another study which showed that 20 Hz rTMS inhibited Aβ aggregation in the hippocampus of 5×FAD mice and enhanced the expression of synaptically associated proteins, which may help restore LTP. Expect for Aβ deposition, a study using an APP23/PS45 mouse model demonstrated that Aβ generation is also disrupted in the brain following LF-rTMS (1 Hz; Huang et al., 2017). rTMS significantly reduced APP and its C-terminal and site APP-cleaving enzyme 1 (BACE1), a critical enzyme that mediates the generation of amyloid oligomers, thereby ameliorating cognitive and synaptic dysfunction. BACE1 also interferes with hippocampal LTP and a reduction in its expression contributes to the enhancement of synaptic plasticity. These results are congruent with those implicating that rTMS-induced alleviation of memory and LTP defects in AD is concomitant with the reduction of Aβ deposition (Wang et al., 2015; Assogna et al., 2022). Activation of the catenin signaling pathway has also been reported to exert anti-AD effects by regulating GSK-3β and tau (Chen et al., 2019), which are downstream effectors of mutant human PS1.

To further investigate the mechanisms underlying Aβ reduction in rTMS-treated patients with AD, Tao et al. (2022) delivered 20 Hz rTMS at an intensity of 100% RMT to the left dorsolateral prefrontal cortex (DLPFC) for 20 minutes. After rTMS intervention, the serum levels of soluble ectodomain of p75 neurotrophin receptor, a protein that suppresses Aβ accumulation, were higher than that of the sham group. Levels of Aβ₄₀, Aβ₄₂, and total Aβ were measured. Polymorphic apolipoprotein E (ApoE) alleles dominate AD genetic risk factors associated with both Aβ aggregation and synaptic dysfunction. Tao et al. (2022) reported that ApoE seemed to have little impact, whereas Chen et al. (2020b) detected a considerable decrease in the expression of ApoE using 5 Hz burst trains at an intensity of 120% average RMT for 2 weeks. The precise mechanisms that account for this discrepancy are unknown, but they suggest that different rTMS treatment parameters may be required for individuals with different ApoE genotypes.

Although these studies provide a foundation for a comprehensive understanding of the potential mechanisms of rTMS in reducing pathological damage in the early stages, further research is necessary to elucidate in greater detail how these mechanisms interact at the network level to reverse mediated cognitive deficits. In addition, the effects of rTMS on cognitive status differ depending on stimulation frequency. Hence, it would be intriguing to determine what differences exist in the effects of high-frequency compared to low-frequency rTMS on Aβ generation and the expression of associated proteins. Moreover, Aβ accumulation activates glial cells to surround the plaques and release inflammatory factors, resulting in neurodegeneration and cognitive decline (Kempuraj et al., 2016). Thus, it is surmised that cognitive improvement detected in the mouse model and patients with AD is also related with rTMS-induced impacts on glial cells and Aβ clearance process.

Glial cells and Aβ clearance

Aβ species can alter the microenvironment, which triggers “priming” of glial cells. Overactivation of glial cells in AD underlines the contribution of proinflammatory cytokines and the resultant potent inflammatory response in local brain regions to neurological deterioration and disease process (Hampel et al., 2021). The primary function of astrocytes is to regulate the neuroinflammatory response, neuronal formation, and neuronal signaling, and to drive the synaptic and network systems back to homeostasis (Cekanaviciute and Buckwalter, 2016; Allen and Eroglu, 2017). Microglia are immune cells in the central nervous system responsible for homeostasis maintenance, Aβ removal, and synaptic pruning. They also play an important role in memory and synaptic function, especially in AD conditions (Cornell et al., 2022). However, the clearance ability of these cells is overwhelming. They are initially of the anti-inflammatory type (M2 for microglia and A2 for astrocytes), but later polarize to a pro-inflammatory phenotype (M1 and A1, respectively) after activation during the advanced stage of AD (Kwon and Koh, 2020). There is a vicious cycle of this polarization between neuro-inflammation and the progression of neurodegenerative disease (Figure 1). Microglia also affect the genesis and differentiation of neural stem cells, thereby promoting neuronal and astrocytic genesis (Luo et al., 2022). Therefore, proper activation of glial cells, promotion of M1/A1 reversion to beneficial phenotypes, and inhibition of the initiation of local immune responses are the key to addressing neuroinflammation in AD (Table 1). Recent studies have indicated that rTMS can serve as a useful therapeutic tool to facilitate Aβ clearance through cellular mechanisms and boost autophagy levels (Chen et al., 2020b).

Figure 1.

Potential vicious circle between AD progression and glial cell polarization.

This figure shows the relationships between glial cells and AD. There are aggregated pathogenic proteins in the AD brain, which activate the microglia and astrocytes. These cells then shift from the anti-inflammatory type to the pro-inflammatory type, resulting in the increase of pro-inflammatory mediators. Local inflammation leads to degeneration of synapses, dysfunction of the blood–brain barrier, and suppresses neuronal proliferation, which ultimately exacerbates neurodegenerative process. Created using Adobe Photoshop. AD: Alzheimer’s disease; Aβ: amyloid-beta; IL: interleukin; mSOD1: misfolded superoxide dismutase protein; TNF: tumor necrosis factor.

Table 1.

Neurobiological changes and biomarkers associated with glial cell activation, anti-inflammation, Aβ clearance, and cognitive improvement in rTMS

Study	Experimental model	Aim	Biomarker or gene	rTMS parameter	Result	TMS outcome
Chen et al., 2020b	APP/PS1 transgenic mice model	Explore mechanisms of neuroprotective effects	p-Tau, APP, Aβ, PP2A, BDNF	5 Hz rTMS	Higher level of BNDF, TrkB, reduced PP2A, and ApoE expression	Improvement of learning and memory impairment, and increased autophagy level
Li et al., 2021a	5xFAD transgenic mice	Explore the effect of HF-rTMS on the pathology	Aβ, SYN, PSD95, NR2B, IL-6, TNF-α, Akt, NF-κB	20 Hz rTMS	Reduced Aβ deposition, IL-6, TNF-α, NF-κB, lower ratio of p-Akt/Akt	Improvement in cognitive performance via PI3K/Akt/NF-κB pathway, decreased activation of microglia, and expression of proinflammatory factors
Cao et al., 2022	3xTG-AD mice model	Explore the effect of HF-rTMS	Akt, GLT-1, Aβ1–42, ROS, SOD, GSH, IL-6, IL-1β, TNF-α, PSD95, SYN, Iba1	25 Hz rTMS	Reduced hippocampal ROS, SOD, Aβ1–42, IL-6, IL-1β, and TNF-α levels, lower ratio of p-Akt/Akt	Reduction of microglial activation, neuroinflammatory response and cell apoptosis, improvement in cognitive function, synaptic plasticity, and oxidative stress via PI3K/Akt/GLT-1 pathway
Clarke et al., 2021	Purified neonatal cortical astrocyte culture	Test whether biology of astrocytes can be influenced in vitro	mRNA: Stim1, Orai3, Kcnmb4, Ncam1 Protein: STIM1 and ORAI3	1 and 10 Hz rTMS	Both reduced: Stim1, Kcnmb4, and Orai3 mRNA; 10 Hz reduced: STIM1, ORAI3, KCNMB4, NCAM1; 1 Hz increased: STIM1 and ORAI3	Frequency-dependent effects on the expression of astrocytic genes and proteins, Influence on calcium signalling and inflammation
Clarke et al., 2017	Primary neonatal astrocyte cultures	Examine the rTMS influence on microglial polarization	Intracellular calcium GFAP- and BrdU-positive cells	1 and 10 Hz rTMS, 50 Hz continuous TBS and BHFS	No effects on cell culture composition and migration, proliferation of astrocytes, reduced astrocyte hypertrophy	Significant rise in intracellular calcium in the cytoplasmic and nuclear compartments of the cultured astrocytes following 1 Hz stimulation
Hong et al., 2022	Ischemic stroke animal model and microglia culture	Evaluate the effect on neuronal activity and gene expression	iNOS, CD206, TNF-α, IL-10, top six miRNAs	10 Hz rTMS	Increased iNOS and IL-10, declined CD206 and TNF-α, increased let-7b-5p	Inhibition of microglial M1 polarization via let-7b-5p/HMGA2/NF-κB pathway
Zorzo et al., 2019	Young Wistar rats	Investigate the effect on behavioral deficits and mechanisms	COx, c-Fos, GFAP, and Iba-1	100 Hz rTMS	No difference in GFAP+ and Iba1+ cells, higher COX activity in RSG, RSA and PAR, higher c-Fos in RSA	Enhanced metabolic activity in RSC and PAR, no alteration in number of glial cells or inflammatory responses
Zong et al., 2020	Rat photothrombotic stroke model	Explore the changes in the phenotype and function of microglia	NADPH oxidase, MnSOD, 3-NT, caspase-9, caspase-3, pro-inflammatory cytokines	3 pulses of 50 Hz TBS	Polarization switch of glial cells, suppression of pro-inflammatory cytokines, and superoxide production	Reduced synaptic loss, neuronal degeneration and cerebral infarct volume, suppression of neuroinflammation, and oxidative neuronal damage
Luo et al., 2022	Ischemic stroke animal model	Explore the changes in the phenotype and function of microglia	TNF-α, IL-4, IL-1β, IL-10, DCX, BrdU, and Iba-1	10 Hz rTMS	Higher level of BrdU+ and DCX+ NPCs, increased IL-4 and IL-10, decreased TNF-α and IL-1β	Improvement in neural function recovery and neurogenesis, reduced polarization of Iba-1 positive microglia

This table summarizes studies investigating the effects of rTMS on glial cells. It indicates that rTMS may exert neuroprotective effects by inhibiting the activation of glial cells, neuroinflammatory response, and microglial M1 polarization. ApoE: Apolipoprotein E; APP: amyloid precursor protein; Aβ: amyloid-β; BDNF: brain derived neurotrophic factor; BHFS: biomimetic high frequency stimulation; BrdU: 5-bromo-2-deoxyuridine; COx: cytochrome oxidase; DCX: doublecortin; GFAP: glial acidic fibrillary acidic protein; GLT-1: glutamate transporter-1; GSH: glutathione; Iba-1: ionized calcium-binding adapter molecule 1; IL: interleukin; MnSOD: Mn-superoxide dismutase; NADPH: nicotinamide adenine dinucleotide phosphate; NPCs: neural precursor cells; NR2B: N-methy-D-aspartate receptor subunit 2B; PAR: parietal cortex; PP2A: protein phosphatase 2A; PSD95: postsynaptic density protein 95; ROS: reactive oxygen species; RSA: agranular retrosplenial cortex; RSG: granular retrosplenial cortex; rTMS: repetitive transcranial magnetic stimulation; SOD: superoxide dismutase; Stim1: stromal-interacting molecule 1; SYN: synaptophysin; TBS: theta-burst stimulation; TNF-α: tumor necrosis factor-α.

However, few studies have investigated the interaction between rTMS and astrocytic activity. In vitro research has found that 1 Hz rTMS can help reduce astrocyte hypertrophy. However, migration or proliferation is not affected, as determined by the number of glial fibrillary acidic protein (GFAP)-positive cells (Clarke et al., 2017). This biomarker, which examines astrogliosis accompanied by morphological changes, increases in AD and shows the potential for differential diagnosis (Oeckl et al., 2019). Notably, by measuring GFAP levels and using RNA sequencing, Yu et al. (2023) demonstrated that 10 Hz rTMS reduced A1-related genes and proteins back to basal levels and blocked A1 neurotoxin in the hippocampus, the mechanism for which remains unknown. Consistent with this idea, rTMS has been shown to promote a detrimental A1 to a beneficial A2 shift, which may result from the microglial phenotype switch (Zong et al., 2020), implicating the underlying interaction between glial cells.

In terms of microglia, a previous in vitro study reported that HF-rTMS (10 Hz) suppressed the uncontrolled activation of glial cells by inducing a shift from the neurotoxic M1 to the reparative M2 state (Hong et al., 2022). This finding was supplemented by studies in different animal models showing that 10 and 15 Hz also reduced the excessive activation of microglia and astrocytes, thereby reducing the secretion of pro-inflammatory mediators (Luo et al., 2022; Zuo et al., 2022). rTMS has been proven to effectively promote anti-inflammatory responses through the promotion of M2 polarization and the activation of calcium signaling or TLR4/NF-κB/NLRP3 signaling pathways. This results in the release of interleukin (IL)-4 and IL-10, while simultaneously reducing the expression of astrocyte genes and proteins (Clarke et al., 2021; Zuo et al., 2022; Table 2). These results agree with those of previous studies that showed that HF-rTMS (20 and 25 Hz) could block Aβ neurotoxicity in an AD mouse model by suppressing the activation of microglia in the dentate gyrus (DG) region and inhibiting the production of neuroinflammatory factors such as IL-6, IL-1β, and tumor necrosis factor-alpha (TNF-α) (Li et al., 2021a; Cao et al., 2022). TNF-α, regarded as M1-related, leads to a preponderance of A1 astrocytes. In terms of Aβ clearance, decreasing TNF-β is of particular importance, as TNF-β genetic deletion in the AD model ameliorates plaque formation by inhibiting Aβ generation through the reduction of functionally active PS1 and secretase without affecting phagocytic activation. IL-1β also suppresses neuronal proliferation, while IL-4 and IL-10 exert anti-apoptotic effects in AD (Kiyota et al., 2012), suggesting that the neuronal regeneration effects of rTMS may be boosted after morphological transformation of microglia.

Table 2.

Signaling pathways and gene expression associated with neuro-protection, Aβ clearance, and cognitive improvement of rTMS

Study	Experimental model	Biomarker or gene	Signaling pathway	rTMS parameter	Effect	Conclusion
Chen et al., 2019	AD model mice	BDNF, NGF, DCX, β-catenin, cleaved caspase-3, Bax, Bcl-2	Wnt/β-catenin pathway	1 and 10 Hz rTMS	Increased BDNF, NGF, doublecortin, cleaved caspase-3 and Bax, decreased β-catenin, Bcl-2	Improved cognitive function and suppressed neuron apoptosis
Stekic et al., 2022	TMT-induced Alzheimer’s-like disease model	IL-1β, IL-10, p-Akt	PI3K/Akt/mTOR Signaling Pathway	iTBS	Decreased IL-1β and p-Akt/t-Akt, increased IL-10	Improved cognition, reduced inflammation, anxiety-related and aggressive behavior, increased anti-inflammatory molecules
Bao et al., 2021	SAMP8 AD mouse	Caspase-3, Bcl-2, Bax, cAMP, PKAc, and CREB	cAMP/PKA/CREB pathway	5 Hz rTMS	Reduced caspase-3 and Bax, increased Bcl-2, cAMP, PKAc, and CREB	Improved spatial learning and memory and morphological abnormalities, reduced neuronal apoptosis
Capelli et al., 2017	PBMCs from peripheral blood of 13 AD patients	BACE1	miR-335-5p, miR-26b-5p	75 Hz PEMF	Increased miR-107 and decreased BACE1	Modulation of miRNA expression, alleviation of brain tissue destruction, and neuronal death
Pang and Shi, 2021	MCI mice	BDNF, NGF, miR-567, NEUROD2, PSD95	miR-567/NEUROD2/PSD95 Axis	1 and 10 Hz rTMS	Suppressed miR-567 expression, up-regulated BDNF, NGF, NEUROD2 and PSD95	Improved the cognitive function and brain neuron activity
Baek et al., 2018	Cultured Neuro-2a cells	NMDA1, CaMKIIδ, CaMKIIα, CaMKIIγ, BDNF	Ca2+–CaMKII-CREB pathway	0.5 and 10 Hz rTMS	Enhanced BDNF, NMDA1, CaMKIIδ and CaMKIIα, decreased CaMKIIγ	Enhancement of BDNF level by activating Ca2+–CaMKII–CREB pathway
Ma et al., 2019	Aged mice	CREB, SYN, BDNF, PSD95	BDNF/CREB pathway	5 and 25 Hz rTMS	Increased BDNF, SYN, PSD95, pCREB	Improved spatial learning and memory impairment, enhanced synaptic structural plasticity
Zuo et al., 2020	Mouse model of depression	Ki67, DCX, p11, BDNF, Homer1a, and p-TrkB proteins	p11/BDNF/Homer1a pathway	15 and 25 Hz rTMS	Increased Ki67 and DCX-positive cells, enhanced levels of p11, BDNF, Homer1a, and p-TrkB/TrkB	Significant antidepressant effects, enhanced neurogenesis and synaptic plasticity, reduced neuronal loss
Yan et al., 2022	Mice model of depression	GFAP, FGF2, Bax, Bcl-2	FGF2/FGFR1/p-ERK pathway	10 Hz rTMS	Reduced Bax, increased Bcl-2, GFAP and FGF2 expression	Improvement of depression-like behavior, alleviation of neural apoptosis
Luo et al., 2022	Middle cerebral artery occlusion to create focal cerebral ischemia model	IL-1β, IL-17A, TNF-α, IFN-γ, caspase1, IL-10, EdU, ASC, GSDMD,Iba1 positive cells	TLR4/NFκB/NLRP3 pathway	iTBS	Reduced IL-1β, IL-17A, TNF-α, and IFN-γ, elevated level of IL-10, fewer Iba1+/CD86+ microglia, more Iba1+/CD206+ microglia	Protection against locomotor deficits and neuronal damage, inhibition of neuronal pyroptosis, promotion of motor functional recovery
Luo et al., 2017	Middle cerebral artery occlusion to create focal cerebral ischemia model	Ki67, Nestin, DCX, NeuN GFAP, BDNF and p-TrkB	BDNF/TrkB pathway	20 Hz rTMS and iTBS	Increase of Ki67/DCX, Ki67/Nestin, and Ki67/NeuN-positive cells, increased BDNF and p-TrkB levels	Improved neurological function and neurogenesis, reduced infarct volume
Guo et al., 2014	Middle cerebral artery occlusion to create focal cerebral ischemia model	BrdU, Nestin, p57 and PTEN	miR-25/p57 pathway	10 Hz rTMS at 120% resting motor threshold	Increased expression of PTEN and miR-25, suppressed p57, increased number of BrdU+/Nestin+ positive cells	Improved proliferation of adult NSCs and neurological function

This table shows different signaling pathways concerning rTMS-induced effects acting against neuronal degeneration, suppressing neuronal apoptosis, and modulating miRNA expression. Aβ: Amyloid-β: BACE1: β-site amyloid precursor protein (APP) cleaving enzyme-1; BDNF: brain derived neurotrophic factor; CaMKII: calmodulin-dependent protein kinase II; CREB: cAMP-response element-binding protein; DCX: doublecortin; DLPFC: dorsolateral prefrontal cortex; FGF2: fibroblast growth factor 2; GFAP: glial fibrillary acidic protein; GSDMD: gasdermin D; iTBS: intermittent theta burst stimulation; LTL: lateral temporal lobe; NEUROD2: neuronal differentiation 2; NGF: nerve growth factor; NMDA1: N-methyl d-aspartate 1; NSCs: adult neural stem cells; PBMCs: peripheral blood mononuclear cells; PEMF: pulsed electromagnetic field; PSD95: postsynaptic density protein 95; PTEN: phosphatase and tensin homolog; rTMS: repetitive transcranial magnetic stimulation; TMT: trimethyltin; TrkB: tropomyosin-related kinase.

rTMS may promote the clearance of Aβ by glial cells through the synchronization of oscillations. Moreover, entraining 40 Hz gamma oscillations is associated with both Aβ generation and its clearance via the inhibition of APP cleavage and state alteration of microglia in CA1 sections, respectively (Iaccarino et al., 2016). rTMS can regulate brain oscillations in patients with AD by resetting the oscillatory activity (Liu et al., 2022), so it is reasonable to assume that rTMS-evoked gamma oscillations may help to clear excessive amounts of Aβ deposition. Another study stressed that the effects of rTMS on Aβ clearance and glial activation were associated with the brain glymphatic system in the parenchyma and meningeal lymphatics in the dura mater (Lin et al., 2021). Impairment of meningeal lymphatics has been documented to promote plaque formation in the brain tissue of 5×FAD mice, reinforcing the notion that improvement of brain drainage efficiency following rTMS contributes to parenchymal Aβ clearance, and its malfunction exacerbates the progression of Aβ-related pathology. Furthermore, this can lead to adverse transition of the microglial phenotype (Da Mesquita et al., 2018, 2021).

Glial cell function is inextricably linked to neuroinflammation inhibition, promotion of Aβ clearance, and neuroprotective effects triggered by rTMS. They interact with each other, and modulating microglia via rTMS may significantly strengthen their effects. Additionally, as an indirect consequence of rTMS on glial cells, these modulatory effects may in turn suppress glial activity in AD mice, forming a negative feedback mechanism. These studies provide strong evidence for the central mediator role of microglia in promoting AD recovery using rTMS. At the same time, they demonstrate that glial cells may act together in response to rTMS and may possibly be key cellular effectors of this therapy. Furthermore, glial cells also likely serve as pivotal markers for early detection and monitoring of the therapeutic qualities of rTMS throughout the course of AD. However, there have been few attempts to explore this drainage system as a promising therapeutic target for rTMS treatment. Future investigations should focus on exploring how glial cells interact under rTMS and how this mutual effect and anti-inflammatory factors complement each other to achieve Aβ clearance and enhance cognitive function at the macroscopic level. In addition, the specific potency of rTMS in glial cells, their morphological transformation, and their mechanisms at the microscopic level are poorly understood. Studies showing that endocytic uptake and lysosomal degradation in microglia and astrocytes are responsible for the clearance of toxic molecules from the brain may provide some insightful clues (Knopman et al., 2021). Notably, the exact holistic efficacy of rTMS on glial cells is largely context-dependent (Cullen and Young, 2016). Outcomes can vary depending on the different animal models used; however, to date, studying glial cells in response to rTMS in AD conditions has been largely underexplored.

Effects of Repetitive Transcranial Magnetic Stimulation on Apoptosis and Gene Expression Modulation

One of the most important mechanisms of rTMS as a suitable therapeutic technique for neuro-restoration and regeneration is its regulatory effect on gene expression and signaling pathways (Table 2). Here, we discuss the main pathways and protein expression induced by rTMS that may potentially serve as therapeutic targets for the treatment of AD. Notably, Ji et al. (1998) first showed that rTMS influenced immediate early gene expression and cyclic adenosine monophosphate (cAMP)-responsive element-binding protein (CREB) phosphorylation, suggesting that rTMS-induced neuroprotective effects are possibly associated with neurobiological changes on a molecular level. To date, few studies have been conducted in different animal models (Table 2); however, the limited number of studies using AD animal models call for more systematic research investigating various signaling pathways and the expression of genes and proteins, respectively, and as a whole, especially in AD models.

Neural apoptosis and regeneration

The therapeutic effects of deep brain magnetic stimulation, including promotion of hippocampal neurogenesis and neural progenitor cell proliferation in AD, have been verified (Luo et al., 2017; Zhen et al., 2017). As an alternative therapy to deep brain magnetic stimulation, rTMS also shows a certain potency in reducing neuronal loss and apoptosis in AD (Cao et al., 2022), enhancing adult rat endogenous neural stem cells in both the hippocampal region and subventricular zone (Arias-Carrión et al., 2004; Ueyama et al., 2011). The 5 Hz rTMS delivered to healthy mice also showed an increased volume of the granular cell layer and an enhanced concentration of microglial markers (Ramírez-Rodríguez et al., 2022). Along with neurogenesis, when applied in an intermittent theta burst pattern, LI-TMS promotes the viability and differentiation of oligodendrocyte precursors in the superficial and deep layers of the cortex (Cullen et al., 2019). These oligodendrocytes, which have received limited attention, are highly specialized cells that produce myelin to wrap around axons and are involved in the progression from a healthy brain to mild cognitive impairment (MCI), and finally to AD, as evidenced by a transcriptomic study (Chen et al., 2020a). Additionally, oligodendrocytes, along with other neurons in the central nervous system, are sensitive to rTMS stimulation. More systematic studies are required to determine the extent to which they respond to rTMS treatment and the exact manner in which rTMS promotes proliferation and neural rehabilitation.

Similarly, on a more granular aspect, rTMS has the potential to improve the pathological changes associated with AD by regulating molecules and proteins related to neural growth and apoptosis. The Bcl-2 gene family includes various subgroups depending on its primary function: (i) apoptosis-stimulated proteins such as caspase-3, Bad, and Bax, which significantly contribute to cell death and neuronal loss in AD; (ii) anti-apoptotic proteins such as Bcl-2 and Bcl-XL; and (iii) pro-apoptotic BH3-only proteins (Kale et al., 2018). By regulating Bax and Bcl-2 expression, LF-rTMS exerts neuroprotective effects in AD by restoring synaptic plasticity and reducing cell death (Yang et al., 2015). This effect may be related to rTMS intensity. When the intensity was set at 1.26 T, both HF-rTMS (10 Hz) and LF-rTMS (1 Hz) significantly enhanced neuronal viability in the hippocampal regions in a mouse model of AD, accompanied by reduced levels of caspase-3 and Bax, and increased expression of Bcl-2 (Chen et al., 2019). This result is consistent with a previous study that reported that the level of caspase-3 was reduced after high-frequency rTMS in ischemia, which was related to the recovery of brain function. Caspase-3 is a pro-apoptotic protease in AD that causes axonal transport deficits and neural degeneration, as it leads to DNA cleavage and tau formation, the activation of which may be mediated by Aβ plaques (Chen et al., 2013). Therefore, the downregulation of its expression strongly indicates the positive effects of rTMS against synaptic structural damage and brain tissue loss induced by AD.

CREB pathway

rTMS can induce synaptic and nerve regeneration through the upregulation of cAMP and CREB, which is beneficial for synaptic plasticity. cAMP is a second messenger, and deficits in CREB-mediated gene transcription are associated with Aβ deposition in patients with AD (Rosa and Fahnestock, 2015). The postmortem brains of both mice and humans with AD show that a reduction in CREB, along with its activated form, pCREB, mainly appear in the hippocampus and PFC, underscoring the importance of CREB transcription in the degeneration process (Pugazhenthi et al., 2011; Bartolotti et al., 2016). Enhancement of CREB was then linked to the reversal of cognitive function in AD by increasing the levels of intracellular cAMP and cGMP (Saura and Valero, 2011). An in vitro model of human-derived neuronal cells, but not rat cells, was consistent with this result (Hellmann et al., 2012). It has also been reported that the upregulation of cAMP and CREB phosphorylation following rTMS, which stimulates transcriptional activity, can be harnessed to induce synaptic and neuroregeneration, and is beneficial to synaptic plasticity (Hellmann et al., 2012). The cAMP/PKA/CREB signaling pathway is also critical for modulating neuronal metabolism (Wang et al., 2019b), as rTMS-induced activation of cAMP/PKA/CREB abrogates neurotoxicity in response to decreased Bcl-2 protein expression and increased Bax protein levels in the AD brain (Bao et al., 2021). Furthermore, as stated, there is a positive correlation between CREB and Aβ toxicity. As a result, the expression of CREB in the PFC and pCREB in peripheral blood mononuclear cells (PBMCs), which indicates the level of pCREB in the PFC, could become novel molecular targets for tracking disease progression and evaluating the cognitive state of patients with AD following rTMS treatment (Bartolotti et al., 2016).

PI3K/Akt pathway

The phosphatidylinositol 3-kinase/Akt (PI3K/Akt) pathway modulates transcription and apoptosis and may be an important target of rTMS. The downstream signaling protein, NF-α B binds to DNA and influences the production of TNF-α, leading to neuronal damage. In AD, rTMS may reduce neuroinflammation and enhance neuronal survival by reducing the pathway disturbance (Li et al., 2021a). It also promotes the expression of Akt protein, resulting in a decreased p-Akt/Akt ratio. This finding concurs with a study that applied rTMS to mouse neuroblastoma cells, suggesting a role for rTMS in the regulation of neuronal proliferation (Lee et al., 2015). The activation of PI3K/Akt expressed in microglia, which modulates its proliferation, is also beneficial to Aβ_1–42 inhibition, cognitive improvement, and synaptic plasticity (Li et al., 2018; Saw et al., 2020; Yoon et al., 2021).

The mammalian target of rapamycin (mTOR) plays a crucial role in protein synthesis, axonal regeneration, and remodeling (Zareen et al., 2018). When the intermittent protocol of theta-burst stimulation (iTBS) is at a frequency of 50 Hz, rTMS attenuated the abnormal regulation of mTOR expression in the cerebral cortex, which may also be mediated via the PI3K/Akt pathway (Stekic et al., 2022). This idea is reinforced by a recent study showing that rTMS failed to protect neurons from inflammation and apoptosis after the administration of the PI3K inhibitor LY294002 in 3xTg-AD mice (Cao et al., 2022). Additional studies indicate that prolonged exposure to HF-rTMS modulated autophagy in human bone mesenchymal stromal cells in vitro through mTOR regulation. mTOR can also enhance the levels of brain-derived neurotrophic factor (BDNF) and growth factors by activating CREB and Akt (Baek et al., 2018; Wang et al., 2019a). High-frequency rTMS activates the phosphorylation of ribosomal protein S6, a downstream marker of the mTOR signaling pathway. The level of activation determines how effectively the motor cortical neurons can regenerate their axons (Fujiki et al., 2020). Moreover, the enhancement of motor cortical activity and activation of mTOR may be the neuromechanism underlying rTMS-induced improvement in the acquisition of motor skills and learning. (Yang et al., 2021). Based on these findings, mTOR activation may serve as a plausible neural mechanism by which rTMS enhances cognitive and motor abilities. Further studies are required to confirm and expand upon these findings.

Expression of neurotrophic factors

The roles of endogenous neurotrophic factors, i.e., pro-neurogenic signals such as BDNF and vascular endothelial growth factor, and their related signaling pathways in inhibiting apoptotic cell death have been extensively discussed over the past 20 years. The expression of these neuroplasticity markers, which are critical for neurodegenerative processes, is dramatically decreased in patients with AD and animal models (Miranda et al., 2019). Therefore, concentrating on the promotion of these neurotrophic factors is a possible target for AD treatment.

A conspicuous upregulation in the level of BDNF can be produced with both 15 and 25 Hz rTMS via activation of the BDNF tropomyosin-related kinase (Trkb) pathway, which effectively rehabilitates cortical function by regulating cortical excitability (Zuo et al., 2020). The CREB-BDNF pathway is disturbed in the AD brain, which is a key instrument in mediating Aβ toxicity and memory impairment (Amidfar et al., 2020); therefore, activation of cAMP/PKA/CREB by rTMS and drugs such as rolipram may reverse cognitive defects and restore cAMP levels, hence increasing BDNF levels (Saura and Valero, 2011; Bao et al., 2021). Increased expression of vascular endothelial growth factor from baseline following a 10-Hz rTMS session is also involved in the enhancement of neurogenesis (Fukuda et al., 2020). In addition, increased BDNF is associated with improvements in the resting network connections between the left parietal region and hippocampus and memory performance in patients with AD, as shown by functional magnetic resonance imaging evaluation and Wechsler Memory Scale-visual reproduction test recognition scores (Velioglu et al., 2021). With the intensity remaining unchanged, relatively lower frequencies (5 and 1 Hz) applied over the motor cortex in vascular dementia model rats or AD mice resulted in the enhancement of LTP amplitude at CA3-CA1 synapses, spatial learning, memory ability, and synaptic plasticity, which also depend on BDNF and vascular endothelial growth factor elevation (Fu et al., 2012; Yang et al., 2015). A similar response to rTMS treatment was previously documented in a study that recorded the release of BDNF, neurotrophin-3, and glial cell line-derived neurotrophic factor in cultured neuroblastoma cells that could be reversed by administration of anti-growth factor blockers (Lee et al., 2015). In summary, these studies demonstrate that the potency of rTMS in cell proliferation and cognitive improvement is partly mediated by the activation of growth factors.

MicroRNA expression

MicroRNAs (miRNAs), which are regulators of human gene translation, have been shown to play a crucial role in pathogenic Aβ toxicity, abnormal APP processing (Amakiri et al., 2019), and neural formation and maturation (Song, 2020). The roles of miRNAs in AD in terms of cognitive function, neuroinflammation, and disease progression have been verified (Soreq and Wolf, 2011). miRNAs can also be affected by rTMS. HF-rTMS can promote neurogenesis in the dentate gyrus, i.e., by proliferation of neural progenitor cells and neural stem cells, by elevating the level of miR106b and downregulating the expression of p21 (Liu et al., 2015, 2018). This result is consistent with that of a previous study showing that miR-25, which belongs to the miR-106b-25 cluster, is involved in the proliferation of these cells in the subventricular zone (Guo et al., 2014). The inhibition of miR-567 and upregulation of post-synaptic density protein (PSD95) and NEUROD2 induced by rTMS may also yield clinical improvements in cognition and brain neuron viability with minimal complications, as well as pronounced production of BDNF and NGF (Pang and Shi, 2021). In addition, BACE1 can be negatively downregulated by hsa-miR-107, whose expression is reduced in AD. Using human in vivo peripheral PBMCs, Capelli et al. (2017) identified that rTMS could increase the expression of miR-107, thereby counteracting APP cleavage. They also reported the modulatory effects on miR-335-5p, which regulates differentiation, transcription, proliferation, and synaptic plasticity, as well as on miR-26b-5p, which is involved in LTP induction. Furthermore, by transfecting an miRNA antagomir to reduce let-7b-5p expression, the effect of rTMS on beneficial M2 microglial phenotype polarization was reversed, along with a decreased concentration of rTMS-induced anti-inflammatory IL-10 (Hong et al., 2022). This study implies that the let-7b-5p/HMGA2/NF-κB signaling pathway is involved in the attenuation of microglia-associated neuroinflammation through inhibiting HMGA2 protein expression and NF-κB activation. Other possible pathways concerning rTMS against neuronal loss, inflammation, or Aβ deposition include miR-195a-5p/CREB or miR-409-3p/PDHB axis (Wang and Gao, 2021; Sun et al., 2022). Whether such downstream genes and indices exert profound anti-AD effects on pathogenesis remains to be further elucidated. Expression of miRNA does change the disease course, and the miRNA-based regulatory system may serve as a diagnostic biomarker for neurodegenerative disease.

These signaling cascades, which mediate the transduction pathways involved in neuronal activity, are considered important processes of synaptic modification, and rTMS can both affect brain signaling and counteract neurodegeneration. Prior research suggests that rTMS rescues pathological alterations and exhibits subsequent neuro-protective function in AD by eliminating Aβ deposition, balance proliferation, and apoptosis of different cells and by regulating gene expression. Limited research has demonstrated the effects of rTMS on miRNA, and it could be a feasible future direction to probe the specific role of rTMS at the molecular level and its neuromodulatory mechanisms in AD. However, studies on synaptic plasticity showing that rTMS induces morphological alteration of synapses may offer different insights into neural regeneration and rehabilitation for future directions.

Effects of Repetitive Transcranial Magnetic Stimulation on Synaptic Plasticity

One of the rationales for the clinical application of rTMS is its ability to trigger particular structural or functional changes in the synapse and, therefore, influence the intrinsic capacity of the brain to refresh new circuits or recover from perennial damage (Figure 2). Longitudinal microscopy enables the direct measurement of neural plasticity and connectivity in the living brain. Rat rTMS models serve as an effective adjunct to human studies as they allow for direct measures of neural plasticity. Research on rTMS mechanisms in humans primarily involves measuring changes in motor-evoked potentials (MEPs), a method that has also been successfully applied to rodents.

Figure 2.

Mechanisms underlying rTMS-induced changes in plasticity.

This figure summarizes changes in the structural and functional plasticity induced by rTMS. Release of plasticity-related proteins lay the foundation for morphological changes. Increase in spine density and elimination of redundant spines help to form more efficient neuronal circuits. Functional plasticity is associated with long-term potentiation, which is modulated by rTMS and a more balanced interaction between excitatory and inhibitory networks is an important mechanism. The role of neurotrophic factors and neurotransmitters in rTMS-induced plasticity are also pivotal. Created using Adobe Photoshop. LTP: Long-term potentiation; PSD95: post-synaptic density protein; rTMS: repetitive transcranial magnetic stimulation; SYN: synaptophysin.

Change in structural plasticity

Structural plasticity-related protein

Synaptophysin (SYN), a common protein marker of pre-synaptic nerve terminals, and PSD95 are not only related to structural plasticity but are also involved in learning and memory. They play a key role in the formation of synapses and their connections and are both repressed in AD conditions. LF-rTMS (1 Hz) and suprathreshold intensity rTMS (110% average resting motor threshold, 110%arMT) dramatically upregulate PSD95 levels and synapse density (Li et al., 2019). It also activates CaMKII, which plays a key role in inducing LTP and facilitating learning and memory (Zhang et al., 2015). In SAMP8 mice, a model characterized by marked dendritic spine loss and Aβ deposition, HF-rTMS (5 Hz) enhanced the levels of SYN/PSD95 proteins, accompanied by the alleviation of cognitive deficits (Ma et al., 2017). This implies the fundamental modification impacts of rTMS on synaptic structure and synaptic protein levels, through which cognitive function is subsequently improved. However, this effect may vary among the studies. An in vitro study showed upregulation of SYN, thickening of the PSD, and increased synaptic density following 1 Hz stimulation at around 1T intensity. In contrast, high-intensity (150%arMT) stimulation is deleterious, resulting in structural destruction and an increased rate of synaptic apoptosis (Ma et al., 2014). In terms of stimulation frequency, 5 Hz rTMS induced a longer-lasting and more distinct remodeling of the synaptic structure with respect to the higher frequency (25 Hz) group, decreasing the width of the synaptic cleft and increasing the expression of these proteins (Ma et al., 2019). This was consistent with the western plotting and polymerase chain reaction experiments. The expression levels of PSD95 and SYN at 1.14 T stimulation appeared to be higher than at 1.55 T stimulation (Ma et al., 2013). The underlying substrate remains to be clarified, but this shows that appropriate stimulation is conducive to neuronal growth, while the therapeutic effects of excessively high intensity or frequency are harmful and may lead to lesions in synapses and neurons.

Formation of dendritic spines and refreshment of neuronal circuits

Ample evidence has demonstrated abnormalities in the excitability of the motor cortex and corticospinal tract (Freitas et al., 2011). Dendritic spines are places where neurons interact with each other, form synaptic connections, and are the post-synaptic components of most excitatory synapses. Changes in synaptic plasticity between neurons are often accompanied by changes in spine morphology and dynamism (Matsuzaki et al., 2004), such as spine density and rate of loss and gain, which serve as good quantitative metrics to reflect synaptic structural plasticity.

The high-frequency (15 Hz) rTMS protocol on pyramidal neurons in layers II/III of the M1 resulted in a predominant upregulation of the number and morphogenesis of thin dendritic spines (and subsequent excitatory synapses) and enhanced the complexity of dendritic processes (Cambiaghi et al., 2021). Furthermore, 10 Hz of stimulation can successfully drive the post-synaptic structural reconstruction of apical dendrites of CA1 pyramidal neurons in vitro, primarily by increasing the number of small spines, while those of the side branches remain unchanged (Vlachos et al., 2012). In contrast, a recent study using cranial window implantation revealed a transient increase in the rate of dendritic spine loss and a delayed decrease in spine density by delivering a single session of high-frequency (5 Hz) but subthreshold (0.12 T) iTBS treatment over M1 (Tang et al., 2021). Multiple sessions of subthreshold rTMS not only maintained the acute effect, but when in association with repetitive rehabilitative training, proved to improve the rate of learning and motor memory in mice (Tang et al., 2018). Given that stimuli below the threshold cannot elicit neuronal electrical activity, this suggests that a subthreshold stimulation intensity can still drive structural plasticity without directly firing action potentials. The mechanism may be illustrated by an increase in the concentration of Ca²⁺, which are released from intracellular stores rather than the extracellular milieu (Hulme et al., 2012; Grehl et al., 2015).

Additionally, following motor training, prompt formation of thin spines and the ensuing synaptic modifications in M1 have also been detected (Raffin and Siebner, 2019). The process from new spinogenesis to maturity and stabilization is highly correlated with motor learning (Albarran et al., 2021). These findings are in line with those of a study that imaged the same apical dendritic layer V pyramidal neurons in mice over different periods during prolonged training (Xu et al., 2009). The results also showed a delayed enhancement in the removal of spines following rapid enlargement of spines (morphogenesis); however, the total spine number was maintained at an almost steady level as that of the control subjects. Despite the constant synaptic density, skilled motor training acts concurrently with brain stimulation to form abundant spines on pyramidal neurons, increase (structural plasticity) the efficiency of layer V synapses in the stimulated cortical area, and induce functional alterations after brain damage (Adkins et al., 2008).

These studies indicate that the connection between synapses and the genesis of synaptic microscopic structures tends to be strengthened by repeated experiences (i.e., brain stimulation, learning, or training). It is conceivable that continuous remodeling of the synaptic configuration during learning, which may be the foundation of long-lasting memory, involves the selective stabilization and strengthening of newly developed spines while eliminating or weakening redundant pre-existing connections and inaccurate ectopic axon terminals, rather than simply adding new spines (Xu et al., 2009; Rodger et al., 2012). Taken together, rTMS presumably helps re-establish a more efficacious circuitry in the M1 in this mode, and this training-induced structural plasticity (and correspondingly increased complexity) improves neural integrity and information integration ability. The combination supports the potential clinical utility of rTMS in the reinstatement of motor memory via better coordination with other synaptic subgroups. The effects of rTMS may extend beyond synaptic structural modulation, and further studies on circuit reorganization are needed. The functional consequences of subthreshold stimulation have rarely been explored; hence, the modulation of these changes at the cellular level remains unclear.

Enhancement in functional plasticity

Functional neuroplasticity in vitro, particularly in hippocampal neurons, is assessed by LTP or LTD (Malenka and Bear, 2004), which can be induced by regular rTMS pulses (Hong et al., 2021) and is acknowledged as a cellular substrate for learning processes and the formation of long-lasting memory. The possible mechanisms leading to cognitive impairment may involve pyramidal neuron hyperexcitability, inhibitory network impairment, inhibitory receptors, and synapse dysfunction. Therefore, the adjustment of cortical excitability and its modulation is crucial in treating neurodegenerative diseases. Similar to synaptic connections, repeated elicitation of electrical activity leads to persistent strengthening or weakening of neural and cortical excitability, and their underlying mechanisms are biologically reminiscent of the LTP/LTD modulation of excitatory synapses.

rTMS can interact with excitatory and inhibitory neurons (Hoppenrath and Funke, 2013), but the aftereffects of rTMS on neurons are associated with neuroplasticity and vary from facilitation to inhibition, depending mainly on the frequency and intensity. They are characterized by Hebbian synaptic plasticity. The mechanism underlying rTMS-induced corticomotor excitability is reminiscent of classical LTP and LTD, but its subsequent influence can be profound, ineffective, short term, or long term, with different stimulation paradigms (Fitzgerald et al., 2006). HF-rTMS (5–10 Hz) is acknowledged to drive LTP-like plasticity, neural excitability, and facilitation (Medina and Túnez, 2013) and is generally considered to benefit patients with cognitive impairments. rTMS at 5 Hz with hundreds of pulses results in incremental excitability in the motor cortex, even hours after the stimulation session is completed (Ziemann et al., 2008). These findings are also in accordance with subsequent studies reporting that HF-rTMS can create long-lasting outcomes (about 3 days) in awake animals, whereas the aftereffects of low frequency appear to be notably shorter in either an anesthetic or awake state (Gersner et al., 2011). Alternatively, LF-rTMS (usually < 1 T) typically depresses synaptic efficiency in the stimulated area and induces LTD-like plasticity, which may aid in the treatment of seizures. TMS at 1 Hz for 5 minutes can induce immediate (within 8 minutes) but persistent (up to 35 minutes) LTD-like modulation of corticomotor excitability in anesthetized rats, as determined by the MEP amplitude (Muller et al., 2014). It also negatively affects the excitatory connectivity of cortical circuits (Grehl et al., 2015). Similarly, in a human study, 1 Hz of M1 led to transient downregulation of motor excitability without affecting basic motor behavior (Muellbacher et al., 2000). Notably, stimulation in specific regions may evoke compensatory responses, such as triggering functional connections in relevant or remote areas to counteract this cortical regulatory inhibition aftereffect (Beisteiner et al., 2020). In support of this idea, 1 Hz TMS of stimulation resulted in a durable (15 minutes) enhancement of excitability in the contralateral M1 in healthy humans when the stimulation time was prolonged to 30 minutes (Schambra et al., 2003).

One study delivered different stimulation loads (pulses per unit time), while other experimental parameters remained identical and reported that there was only a small increase in the concentration of intracellular calcium when the number of pulse trains was 600 pulses (1 Hz for 10 minutes). When the frequency reached 10 and 100 Hz, 10 minutes of stimulation resulted in a pronounced enhancement in Ca²⁺ concentration and a decreased rate of neuronal survival. However, delivering pulses similar to endogenous patterns (TBS: 7000 and BHFS: 120,000 pulses) contributed to an increase in intracellular calcium levels, whereas the rate of neuronal survival remained stable (Grehl et al., 2015). This concurs with the general idea that the determinant of the outcome of rTMS is the stimulation frequency. With the intensity remaining unchanged at an 80% average RMT, both 2 and 6 Hz stimulation suppress the magnitude of MEP (Todd et al., 2006). At stimulus intensities near the RMT (90% RMT), high frequencies (> 5 Hz) drive cortical excitability (Peinemann et al., 2004), whereas low frequencies (1 Hz) result in the opposite effect (Romero et al., 2002) as discussed above. In addition, TBS emphasizes the importance of the temporal structure, a possibly efficient parameter, with intermittent TBS (iTBS) increasing excitability and continuous TBS (cTBS) suppressing excitability.

Studies have reported that stimulation intensity is as important as stimulation frequency for modulating cortical excitability and driving plasticity (Ma et al., 2013; Makowiecki et al., 2014; Tang et al., 2018). The application of LF-rTMS was more effective when the intensity was above the RMT. High stimulation intensities (usually > 1 T, suprathreshold rTMS, HI-rTMS) delivered to M1 in humans can drive neural activity and trigger action potentials, which can be observed as MEPs. Low-intensity rTMS (subthreshold rTMS, LI-rTMS) can induce neuroplasticity and has specific clinical applications without directly eliciting neuronal electrical activity. In mice with AD, low-frequency (1 Hz) stimulation at an intensity of 100% RMT alleviates LTP impairment and improves neuronal plasticity, thereby improving spatial memory (Huang et al., 2017). To further understand the effect of LF-rTMS, different intensities were observed in AD mice. High intensity (1.55 T) reduces neuronal viability and induces apoptosis, accompanied by neurite regression and structural destruction, while low intensity (1.14 T) has no substantial effect on neuronal survival, but indeed facilitates positive changes in synaptic structure (Ma et al., 2013).

These findings indicate that the rhythm or intensity of pulses is the most important outcome of rTMS-drive cortical excitability modulation. These findings provide pivotal insights for clinical applications. Challenges remain if the rTMS is clinically utilized owing to the nonlinear and complex association between its influence and stimulation parameters. Therefore, it is crucial to normalize the therapeutic rTMS extensively.

Repetitive Transcranial Magnetic Stimulation and Neural Plasticity in Alzheimer’s Disease Patients

Based on lasting alterations in MEP amplitude, rTMS shows promise for determining the extent to which neuroplasticity, M1 excitability, and neural circuits are altered in patients with AD or MCI, particularly when coupled with simultaneous functional magnetic resonance imaging, electromyography, or EEG recording (Pascual-Leone et al., 2011; Smailovic and Jelic, 2019; Khedr et al., 2020). These have been utilized in non-invasive, real-time recording of brain electrical activity at the cellular level. The size of the electromyographic response is widely used to assess excitability of the motor cortex and experimental effects on stimulated muscle, whereas a combination of EEG and MRI shows the influence of rTMS in nonmotor regions. Interleaving EEG with rTMS is also useful in probing changes of brain oscillations, neural inhibition, and excitation (Joseph et al., 2021), plasticity of cortical circuits, and damage to communication within different brain regions (Tremblay et al., 2019) in patients with AD and MCI, during resting state or task performance. Moreover, progress on the disease can be traced and therapeutic effects in response to stimulation can be monitored (Ferreri et al., 2021). When combined with drugs administration, TMS-EEG can also be an efficient option to investigate the influence of various medicines on inhibitory GABAergic circuitry or neural plasticity (Tremblay et al., 2019).

It has been revealed that rTMS can generate long-lasting neuro-modulatory influence on the structural plasticity of patients, enhance the cortical-hippocampal networks (Wang et al., 2014), and strengthen functional connectivity between medial frontal regions and the precuneus, with boosted EEG oscillatory activity of the precuneus (Koch et al., 2018). The precuneus, part of the posteromedial parietal cortex, is a key node for the default mode network. It is generally activated when doing tasks related to visuospatial working and episodic memory. Prolonged rTMS treatment targeting the precuneus can slow down cognitive decline, as evidenced by the Clinical Dementia Rating Scale-Sum of Boxes, Alzheimer’s Disease Cooperative Study-Activities of Daily Living, and Mini-Mental State Examination scores at 24 weeks from baseline (Koch et al., 2022). The DLPFC is a core part of the PFC that underlies the complicated and rich nature of cognitive function in humans. In patients with AD, rTMS treatment helps to enhance LTP-like plasticity in DLPFC and modulate corticospinal excitability, which are closely associated with improvement in cognitive and executive function (Kumar et al., 2017; Li et al., 2021b). The DLPFC and its extensive neural circuit may also be affected by rTMS. Ahmed et al. (2012) reported that 20 Hz rTMS over DLPFC for five daily sessions significantly promoted cognitive functioning in individuals with mild-to-moderate AD, which is considered to result from enhancement in activity of intrinsic network and efficiency of synaptic transmission. The duration of transcallosal inhibition was also shortened, indicating positive effects of rTMS on functional plasticity and communication between hemispheres. Apart from PFC, Anderkova et al. (2015) reported that stimulating the superior temporal gyrus and right inferior frontal gyrus also induces short-term modulatory effects on neural plasticity, relieving disturbances of attention, and accelerating psychomotor speed in patients with MCI and AD.

Furthermore, studies in both the mouse model and patients with AD revealed that early anatomical alterations of the ventral tegmental area and degeneration of dopaminergic neurons are possible causes of brain dysfunction in the prodromal stages of the disease (Nobili et al., 2017; De Marco and Venneri, 2018). rTMS may also activate the PFC and decrease the severity of clinical symptoms by modulating the mesocorticolimbic dopaminergic circuitry (D’Amelio and Di Lazzaro, 2023). Functional synaptic plasticity may also be enhanced via increasing the release of endogenous dopamine in the ipsilateral striatum (Strafella et al., 2001; Martorana et al., 2009). A study on human participants reported that high frequency rTMS of the cerebral cortex induced an increase in dopamine concentration and upregulated the level of its receptor in the striatal projection of the targeted region (Strafella et al., 2003), thereby reducing Aβ deposition, since this process may expend dopamine. rTMS has the most significant effect on dopamine-receptor 4, which has a role in cognition, memory, and restoring LTP function at the hippocampus (Guo et al., 2017). This also confirms that modulation of dopamine levels may underlie the therapeutic effects of rTMS on cortical plasticity and cognitive function (Choung et al., 2021). Other neurotransmitters are also involved. Unlike electrically evoked potentiation, which is mediated presynaptically, changes in structural and functional plasticity can be induced by magnetic stimulation in an postsynaptic N-methyl-D-aspartate receptor (NMDAR)-dependent mode (Tang et al., 2017). With regard to the GABAergic system, two isoforms of GABA receptors—short-interval intracortical inhibition and long-interval intracortical inhibition—are supposed to be vital in mediating cortical inhibition by numerous rTMS studies (Di Lazzaro et al., 2021). Modulation of NMDAR and GABAergic system activity also helps to slow the disease progression.

Variability and Future Directions

Application of the TBS protocol to induce transient changes in cortical excitability has long been considered effective, reproducible, controllable, and relatively consistent. Indeed, substantial aftereffects exist in response to TBS at the group level, but these effects tend to decrease during the second evaluation. Using linear regression analyses and intraclass correlations, Schilberg et al. (2017) found a low reliability of iTBS effects within individuals (intra-individual variability) across two follow-up visits using prolonged measures of corticospinal excitability. This hinders the interpretation of the effects of the observed group iTBS intervention in a single assessment, particularly at the individual level.

Moreover, variability between individual outcomes is a major problem limiting the application of rTMS. After specific TBS protocols, particularly those developed for cortical facilitation, less than half of the participants displayed the anticipated plasticity-inducing effects such as enhancement of MEP amplitude (Hamada et al., 2013), excitability (Hinder et al., 2014), and behavioral and executive function (Nicolo et al., 2015). The significant relationship between the effects of TBS and the latency of MEPs evoked by single pulses of TMS with directed current (anterior-posterior[a-p] and lateral-medial [l-m]), which preferentially recruits different interneurons, indicates that differences in the populations of activated interneuron networks may partly account for the variability in response to rTMS (Hamada et al., 2013). Hamada et al. (2013) concluded that the relative recruitment of late versus early indirect waves, that are considered to originate from a monosynaptic excitatory connection to pyramidal cells, influenced the effects of this new intracortical spike-timing-dependent plasticity protocol. Therefore, it is reasonable to assume that the complex interplay between different mechanisms regulating excitability and stimulation parameters lies at the root of the considerable inter- or intra-individual variability.

The M1 functional state, which can be assessed by TMS-evoked EEG oscillations, has also been reported to influence the effects of rTMS (Corp et al., 2020). This could be explained by the individual differences in the oscillatory activity of specific cortical neural circuits activated by iTBS, particularly in the beta range (Leodori et al., 2021). Wang et al. (2020) reported that stimulation factors at the protocol level, such as the number of stimulation targets, stimulation frequency, or concurrent cognitive training are influential. Other putative reasons for outcome variability at the sample level include regular activity, attention, activation state, cortical thickness, and educational level (Conde et al., 2012; Goldsworthy et al., 2014; Wang et al., 2020). However, few TMS-EEG studies reveal iTBS-evoked EEG oscillations. New approaches are needed to individually optimize rTMS protocols, and further systematic studies determining the reliability of rTMS and its mechanisms in the field of AD research are needed. Furthermore, cytoarchitecture, macrostructure, and even functional organization vary between individual brains; therefore, personalized stimulation approaches should be developed.

Precise location or personalization of the stimulation area may reduce such wild inter-subject variability, which is a crucial factor for treatment effects. However, currently, the site of stimulation in clinical practice is subjective and ambiguous, hindering the second location. Furthermore, dislocation may arise from coughing, sneezing, and fatigue during the intervention. The more consistent the stimulation effects of a specific region, the easier it is to reduce the incidence of ineffective stimulation and lessen errors between treatments, to some extent. Many methods exist to precisely locate the stimulation site. For example, the 5-cm rule means the coil should be placed 5 cm anterior to the motor cortex to locate DLPFC. However, it chooses a stimulation site relying on gross anatomical landmarks regardless of interindividual differences of skull size and cortical tissue. Another generalized measurement is the 10–20 electroencephalography (EEG) electrode cap disposition. Some scholars believe the “10–20 EEG system” has an advantage over the 5-cm method in reducing inter-individual variability (Rusjan et al., 2010). More personalized target selection procedures involve neuro-navigated TMS. It is a visualization technique for the cortical location that maximizes the anatomical information provided by structural or functional MRI to replicate the 3D model of the brain. It significantly improves the episodic memory of MCI patients and the Montreal Cognitive Assessment score (Yang et al., 2022). Individualized neuro-navigation can be adapted based on the distance from the scalp to the cortex. Combined with an infrared camera, this guarantees the same stimulation point for each subject during different sessions. It may elicit positive clinical effects in depressive disorder, with inaccurate location resulting in low anti-depressive effects (Johnson et al., 2013), and it is even thought to be superior to the standard “10–20 EEG system” localization in dealing with auditory hallucinations in schizophrenic patients (Klirova et al., 2013). Alternatively, a recent study argued that neuro-navigated iTBS failed to justify its clinical superiority over the non-neuro-navigated F3-EEG positioning approach (Hebel et al., 2021). Despite this discrepancy, the neuro-navigated coil provides an approach for accurate targeting both for the first time and repeatedly, especially after efficient mapping of related brain regions; however, its inconvenience and inefficiency limit its practical use. Better navigation of the coil to the targeted sites to guarantee accuracy remains a research focus. Other personalization measures include symptom-specific targets (or symptom-response map; Siddiqi et al., 2020) and a functional connectivity-based targeting approach (Fox et al., 2013). However, whether patients with neurodegenerative diseases would benefit from such precise coil placement and would experience better cognitive outcomes remain to be further investigated. A study concluded that personalized interventions did not hold better outcomes. This may attribute to the patient’s personal profiles that most studies used when analysing individualized stimulation site, because they relied more on anatomical neuroimaging data instead of functional connectivity profile (Menardi et al., 2022). Functional connectivity profiles are generally characterized by greater individual specificity (Finn et al., 2015), with moderate variations from task-state and day-to-day variability (Gratton et al., 2018). Therefore, they were more valuable than anatomical MRI when aiming at the personalization to better characterize individual brain organization and its targeting. Further studies are warranted to explore new and reliable ways to both reduce inter-study heterogeneity and effectively promote residual cognitive functioning, with conclusive results.

Limitations

This review has some limitations. First, most of the above-mentioned discoveries and conclusions were drawn according to the first observations after the stimulation and there is a lack of studies investigating long-term effects of rTMS on neural protection and rehabilitation. Second, this review mainly focuses on cognitive outcomes in terms of clinical application, and influence of rTMS on behavioral and psychological symptoms of patients still warrants further exploration. Third, we did not discuss studies including sufficient individual patient data, so more correlation studies and systematic studies based on the personal data are required to evaluate validity and reliability of rTMS, and tailor its parameters to suit individual patients to reduce variability. Finally, since clinical studies usually stimulate DLPFC, non-significant effects of rTMS on other brain areas need to be interpreted with caution.

Conclusion

This review aimed to provide a comprehensive understanding of rTMS-derived effects in the context of the AD brain, particularly those related to pathology, gene expression, and neural regeneration. Furthermore, an appropriate frequency can influence synaptic structure and functional plasticity, promote the growth of dendritic spines, and reorganize synaptic connections. It also improves LTP impairment, regulates excitation and inhibition imbalances in cortical and neural networks, and therefore improves cognitive and memory function in individuals. Ultimately, our goal was to gain a deeper understanding of the multifaceted role of rTMS in shaping brain plasticity and AD therapy. However, current research on non-invasive stimulation in patients with AD involves multiple brain regions such as the frontal, parietal, and temporal lobes, and accurate screening of stimulation sites remains the focus and difficulty in the clinical application of rTMS. A thorough understanding of the relationship between stimulation target, parameters, and cognitive enhancement remains to be established. At this stage, more experiments including inflammatory, apoptotic, and functional brain imaging are needed to verify how rTMS effectively triggers changes in synaptic plasticity in AD models, whether these changes vary between different brain regions, and to what extent they contribute to cognitive improvement in patients with AD. Taken together, the experimental outcomes discussed herein support the application of rTMS as an adjunctive therapy for disease rehabilitation, or at least as a potential tool to distinguish between AD and MCI. This approach is non-invasive and more convenient, and safer than other existing treatments and has the potential for wider applications if treatment options are tailored to individual patients.

Funding Statement

Funding: This work was supported by the Hefei Comprehensive National Science Center Hefei Brain Project (to KW); the National Natural Science Foundation of China, Nos. 31970979 (to KW), 82101498 (to XW); and the STI2030-Major Projects, No. 2021ZD0201800 (to PH).