Deregulation of Synaptic Plasticity-Related MicroRNAs After Repetitive Transcranial Magnetic Stimulation in Alzheimer’s Disease

Abstract

Repetitive transcranial magnetic stimulation (rTMS) is an emerging non-invasive therapeutic approach to slow down cognitive and functional decline in Alzheimer’s disease (AD), potentially through plasticity-related mechanisms. MicroRNAs (miRNAs) play a crucial role in synaptic plasticity, and their deregulation contributes to AD-related cognitive impairment. In the present study, we first used a dosimetric model to translate rTMS field applied in AD patients to an in vitro system, identifying miRNAs as potential biomarkers responsive to rTMS. We found that rTMS induced in vitro deregulation of miR-26b, miR-125b, miR-181c, and miR-146a. Then, we investigated the effects of rTMS over precuneus during a 3-week, randomized, sham-controlled trial in AD patients. In patient serum, miR-26b, miR-30b, and miR-125b were significantly modulated in AD patients compared to healthy controls, though no significant modulation emerged between sham and rTMS groups before or after stimulation. Subsequently, the correlation analyses, which incorporated patients’ cognitive scores, revealed that reduced miR-25 levels were significantly associated with cognitive improvement. However, no significant differences emerged between Real- and sham-rTMS correlation coefficients, likely due to the limited sample size, indicating that miR-25 may represent a general prognostic marker rather than a treatment-specific indicator. Furthermore, the ability of this miRNA to discriminate responders from non-responders, shown by ROC analysis, highlights its potential as a promising predictor of rTMS treatment efficacy to be validated in a larger patient cohort. Altogether, our findings suggest, for the first time, that rTMS may modulate specific miRNAs in AD patients, with miR-25 representing a pivotal key target for future validation studies.

Supplementary Information

The online version contains supplementary material available at 10.1007/s12035-025-05620-y.

Introduction

Alzheimer’s disease (AD) is the most common form of dementia, affecting over 40 million people globally. The key pathological features of AD are the accumulation of beta-amyloid (amyloid-β) plaques and tau tangles, which trigger a cascade of neuropathological events leading to synaptic plasticity impairment and brain cell degeneration [1, 2]. Notably, these changes start many years before the onset of symptoms, which usually begin with mild memory deficits and later affect other cognitive domains and daily living activities.

Currently, there is no effective cure for AD, and available treatments only manage symptoms without slowing disease progression [3]. Repetitive transcranial magnetic stimulation (rTMS) is a non-invasive brain stimulation tool that has been shown to improve cognition in AD patients [4–6], by modulating neuronal excitability and promoting synaptic plasticity.

Specifically, rTMS has been applied to the default mode network (DMN), which has a key role in autobiographical and episodic memory [7]. The precuneus (PC), the most important hub of the DMN, is active during memory processes [8, 9] and is the first region where amyloid plaques and neurofibrillary tangles accumulate. Recently, PC rTMS has been shown to improve long-term memory [7] and to slow down cognitive and functional decline [10].

Despite clinical studies consistently demonstrating beneficial outcomes following rTMS in AD patients [11–13], little is known about the underlying molecular mechanisms driving these outcomes. MicroRNAs (miRNAs) are small non-coding RNA molecules (about 20 nucleotides long) that, by pairing with target mRNAs, induce mRNA silencing, thereby profoundly influencing cellular behavior at the epigenetic level [14]. Circulating miRNAs, secreted into biological fluids, can be easily quantified, making them promising biomarkers for therapy response [15–17]. In AD, miRNAs contribute to key pathological processes, including amyloid-β plaque formation, tau hyperphosphorylation, neuroinflammation, and synaptic dysfunction [18]. Their dysregulation has been linked to neurodegenerative diseases, including AD [19, 20], particularly with plasticity-related miRNAs [15] which are involved in Long-Term Potentiation (LTP) mechanisms, known to be impaired in AD animal models and patients [15, 21, 22], driving cognitive decline [23, 24]. Recent evidence highlights the roles of miR-125b-5p and miR-146a as pro-inflammatory factors that can exacerbate AD pathology [25]. Elevated levels of miR-26b have also been implicated in neuronal dysfunction and degeneration in AD models [27]. In addition, Kumar et al. [28] identified six circulating miRNAs—miR-9, miR-125b, miR-146a, miR-181c, let-7g-5p, and miR-191-5p—as promising biomarker candidates. Furthermore, the expression of miR-25, implicated in neuronal survival [29], synaptic regulation, and neuroplasticity, was markedly upregulated in the ischemic cortex following rTMS [30]. Taken together, these findings underscore the pathogenic contribution of dysregulated miRNA expression to AD progression and highlight their potential as both biomarkers and therapeutic targets.

The selection of the miRNAs, in this study, was based on a comprehensive review of the existing literature on blood-based miRNA biomarkers in AD, with an emphasis on those associated with neuronal plasticity.

For the first time, this study demonstrates a modulation of miRNAs’ expression by rTMS treatment, highlighting their role in driving plasticity mechanisms. To this extent, we first investigated the effects of rTMS treatment on miRNA expression in an in vitro AD model. As a complementary approach, we then assessed changes in circulating extracellular miRNAs in the serum of AD patients enrolled in a double-blind, sham-controlled PC-rTMS trial. Notably, it is important to highlight that miRNAs can be modulated differently in cellular versus serum environments [28, 31]. Together, these approaches would provide a more comprehensive understanding of miRNA dynamics in AD and the effects of rTMS treatment.

Materials and Methods

Chemicals and Reagents

All chemicals and reagents were of molecular biology grade and obtained from commercial suppliers. Amyloid-β protein fragment 1–42 (A9810), retinoic acid (RA, R2625), brain-derived neurotrophic factor (BDNF, B3795), poly-D-lysine (P6407), and DAPI (32670) were purchased from SIGMA (St. Louis, MO, USA). Vacutainer® blood tubes treated with EDTA (SKU: 366643 GTIN: 00382903666430) were obtained from BD Biosciences (Franklin Lakes, NJ, USA). All cell culture chemicals were purchased from Corning (New York, NY, USA): mixture of Dulbecco’s Modified Eagle’s Medium and Ham’s F12 (DMEM/F12 50/50) (10–090-CVR), fetal bovine serum (FBS) (35-015CV), penicillin–streptomycin solution (30-002-CI). Primary antibodies were purchased from different commercial suppliers: anti-GAP43 (Ab277627), Abcam, Cambridge, UK), anti-βIII-TUBULIN (2G10, Invitrogen, Maltham, MA, USA), anti-Heat Shock Protein 70 (HSP70, H5147, SIGMA, St. Louis, MO, USA), horseradish peroxidase-conjugated secondary antibody (sc-516102, Santa Cruz Biotechnology, Dallas, TX, USA). Secondary antibodies conjugated to Alexa Fluor 488 (A11001) were obtained from Life Technologies, Thermo Fisher Scientific (Waltham, MA, USA). The ECL Super Signal West Pico PLUS Chemiluminescent Substrate (West PICO 34577) was purchased from Thermo Fisher Scientific (Waltham, MA, USA). For RNA extraction, miRNeasy Tissue/Cells Advanced Kits (ID.217684) and miRNeasy Serum/Plasma Kit (ID 217184) were obtained from QiaGen GmbH (Hilden, Germany), while for reverse transcription and amplification, TaqMan Universal PCR Master Mix (UNG 4364341), TaqMan microRNA Assay (assay ID are reported in Table S1), High-capacity cDNA Reverse Transcription kit (4374966), and Power-Up SYBR Green master mix (A25742) were purchased from Thermo Fisher Scientific (Waltham, MA, USA). Finally, the BDNF ELISA assay kit (BT-DBD00-1KT) was obtained from R&D Systems (Minneapolis, MN, USA).

In Vitro Model

Dosimetry of rTMS

To ensure reproducibility and to align the in vitro stimulation with the electromagnetic exposure observed in patients undergoing rTMS, we refer to the dosimetric analysis previously conducted and described in detail in [32].

Briefly, 16 individualized head models were generated from MRI scans of treated patients, whose stimulation intensities were known. Simulations were performed using SimNIBS v4.0.0, an open-source software that integrates MRI-based segmentation, mesh generation, and quasi-static finite element modeling of the induced electric field. Standard conductivity values provided by SimNIBS were adopted, and the built-in model of a 70 mm figure-of-8 coil (Deymed 70BF), commonly used in clinical practice, was selected for the simulations.

Based on this analysis, we identified both the optimal coil positioning and the stimulation intensity required to reproduce, within the in vitro system, an electromagnetic dose comparable to that received by patients. Since individual patients received different levels of stimulation [32], we selected the lowest induced dose among the cohort as the reference value for the in vitro setup. The coil was positioned tangentially to the surface of the multiwell plate, at a fixed distance, in a configuration that ensured a homogeneous electric field distribution across the well floor (Fig. 1). This orientation was chosen over a parallel configuration, which—although capable of generating higher peak field values—would have concentrated the field along the well edges, reducing uniformity [32]. The selected intensity was calibrated to match the minimum electric field magnitude observed in the cortical regions of interest. For the in vitro setup, a six-well polystyrene plate (7.4 × 11.4 cm) was modeled, with each cylindrical well measuring 1.7 cm in height and 1.6 cm in radius, and a wall thickness of 2 mm. Wells were filled with an isotonic buffer (PBS dilute solution containing 8.8% sucrose) with conductivity adjusted to 0.3 S/m. A volume of 10 ml per well was selected to optimize energy transfer and ensure consistent exposure across the cellular layer. Full methodological details can be found in [32].

Fig. 1.

Representation of the in vitro setup

Cell Culture

The human neuroblastoma cell line, SH-SY5Y, was obtained from the European Collection of Authenticated Cell Cultures (ECACC, Porton Down, England). Cells were grown in a humidified atmosphere at 37 °C and 5% CO₂ in a 1:1 mixture of DMEM/F12 50/50 supplemented with 10% FBS, 100 U/ml penicillin, and 100 U/ml streptomycin. Cells were sub-cultured once they reached 80% confluence and used before the fifteenth culture passage.

Amyloid-β Oligomerization Protocol

Soluble amyloid-β oligomers were prepared with a protocol as in [33]. Amyloid-β protein fragment 1–42 was resuspended at 1 mg/ml concentration in 1% NH₄OH, sonicated for 5 min, divided into four aliquots (250µl each), and incubated at room temperature for 3 h. The protein was completely desiccated and stored at −20 °C. One aliquot at a time was resuspended in DMSO at 1 mM, sonicated for 10 min, and stored at −20 °C. At the time of use, amyloid-β was diluted 1:1 (500µM) in phosphate-buffered saline (PBS), incubated at 4 °C for 3 h, then further diluted to the working solution.

AD Cell-Like Model

The differentiation protocol for cholinergic neurons is reported in Fig. 2. Briefly, cells were seeded at a density of 8000 cells/cm² in DMEM/F12 with 10% FBS (day 0) in the first two wells of the short side of a six-well plate to reach the best stimulation conditions [32]. Neuronal differentiation was initiated as described in [34], lowering FBS in culture medium to 1% and supplementing with 10 μM retinoic acid (RA) on the first day. The medium was replaced with a fresh one supplemented with 50 ng/ml of brain-derived neurotrophic factor (BDNF) and RA on the fourth day. To mimic the conditions of AD, 10 nM of amyloid-β oligomers were added to the culture medium on day 6 after the start of differentiation [34], and a dilution of DMSO equal to that used for amyloid-β administration was added to the HC condition.

Fig. 2.

Timeline of SH-SY5Y differentiation to AD-like cell model and rTMS exposure. Experimental scheme (A). Representative phase contrast image of proliferative human neuroblastoma cell line SH-SY5Y cells in culture medium with 10% of FBS (B, left). Representative phase contrast image of differentiated SH-SY5Y cell line cultured for 7 days with 10 µM RA in culture medium with 1% of FBS, with the addition of 50 ng/ml of BDNF on the fourth day (B, right). Note the abundance of neurites in differentiated cells (×20 magnification, scale bar = 50 μm)

rTMS Protocol

As shown in Fig. 2, on day 7, differentiated cells and treated with amyloid-β, mimicking AD, underwent rTMS (Real) or placebo treatment (Sham), in which the coil was positioned as in the treated samples, but no stimulation was delivered. Additionally, differentiated cells but untreated with amyloid-β were used as a healthy control (HC) model. To ensure consistency with the dosimetric model, the coil (Deymed 70BF) was positioned tangentially to the surface of the multiwell plate at a fixed distance. The coil orientation was selected to maximize electric field homogeneity across the well floor, avoiding edge-concentrated hotspots. The stimulation intensity was calibrated to match the minimum dose induced in the patient cohort, as determined by the dosimetric analysis, i.e., the 60% of MSO. For all conditions, the cell medium was removed and replaced, during the exposure, with 10 ml of PBS dilute solution containing 8.8% sucrose. The used buffer reduced the electrical conductivity of the medium while maintaining physiological osmolarity to obtain induced current densities comparable to those achieved in patients’ brains (for complete details see [32]). Cells were exposed to a single-shot protocol of rTMS consisting of a single sequence of magnetic pulses lasting in total 20 min (the sequence was composed of 40 trains of 40 pulses, each train lasted 2 s, and between one train and another, there were 28 s without signal), as described in [7, 10]. The pulses in one train were biphasic and delivered at a repetition frequency of 20 Hz. After treatment, cells were incubated in their culture medium for 24 h before proceeding with the collection of cell pellets for subsequent analyses.

Clinical Study

AD Patients’ Sample

Thirty AD patients, diagnosed through cerebrospinal fluid (CSF) biomarker analysis, and ten age-matched healthy controls (HC) were recruited for the study. All AD patients, diagnosed according to the IWG-2 criteria [35], underwent a comprehensive clinical investigation, including interviews regarding their medical history, a complete neurological examination, Mini-Mental State Examination (MMSE), complete blood screening, neuropsychological and neuropsychiatric evaluations, and magnetic resonance imaging. We recruited AD patients based on the criteria shown in Table 1.

Table 1.

Inclusion and exclusion criteria for patients’ recruitment

Inclusion criteria	Exclusion criteria
Aged between 60 and 85 years at the time of signing the informed consent form	Inability to provide informed consent
CSF-specific biomarker profile or positive Amyloid PET scan consistent with amyloid pathology	Participation in other studies involving treatment, whether pharmacological or non-pharmacological
Mini-Mental State Examination (MMSE) score > 18	History and/or evidence of any other neurological disorder that could be interpreted as a cause of dementia (e.g., structural or developmental abnormalities, epilepsy, infectious, degenerative, or inflammatory/demyelinating diseases such as Parkinson’s disease and frontotemporal dementia)
Stable intake of cholinesterase inhibitors for at least 3 months prior to study start	History of a significant psychiatric disorder that, in the investigator’s judgment, would interfere with study participation
Compliance with the inclusion criteria for magnetic resonance imaging and transcranial magnetic stimulation	History of alcohol or other substance abuse, according to DSM-IV criteria, or recent/past history of abuse that could contribute to dementia
Availability of a caregiver able to provide the necessary patient information and present during the signing of informed consent	Dermatitis, extensive scarring, eczema, or psoriasis on the scalp
	Ongoing treatment with psychotropic drugs
	Presence of cardiac pacemakers, artificial heart valves, electronic prostheses, biostimulators, metal implants, or electrodes implanted in the brain, skull, or spine
	Claustrophobia

Of the 30 patients recruited, 10 dropped out of the study. The final cohort was composed of 20 patients: 14 received real stimulation treatment (Real), while the remaining 6 received the sham treatment (Sham). To evaluate the cognitive, clinical, neurophysiological, and blood changes related to the treatment, the patients underwent specific evaluation at the beginning (AD T0) and at the end of the treatment (AD T1). The study was approved by the review board and the local ethics committee in accordance with the principles of the Declaration of Helsinki and the International Conference on Harmonization Good Clinical Practice guidelines. All patients or their relatives or legal representatives provided written informed consent. Patients could withdraw at any point without prejudice. Clinical trial number: not applicable.

rTMS Treatment

The rTMS treatment consisted of 15 sessions in total (once a day, 5 days per week, for an overall duration of 3 weeks) applied daily from Monday to Friday to the PC, in either real or sham condition as described in [7, 10]. Each daily stimulation session consisted of forty trains of 2 s delivered at 20 Hz, spaced out by 28 s of no stimulation (total number of stimuli: 1600). So, the entire session lasted approximately 20 min. The rTMS coil was oriented parallel to the midline to induce a posterior–anterior directed current. The TMS coil position was constantly monitored using a neuronavigation system (Softaxic, EMS) coupled with an infrared camera. We used the individual structural MRI previously performed for diagnostic purposes to accurately position the coil over the target area through the neuronavigation system. This ensured that the same spot was reached across different sessions performed days or weeks apart. rTMS sham treatment was applied with a sham coil positioned in correspondence to the target area. The intensity of stimulation was based on a previous study [10].

Plasma Collection

Whole blood samples were collected using Vacutainer® blood tubes treated with EDTA. After 1 h, cells were removed from plasma samples by centrifugation for 10 min at 1900 × g and 4 °C. The supernatant, obtained by centrifuging the plasma samples at 16,000 × g for 15 min, was depleted of platelets and stored at −80 °C until use.

Protein Extraction and Western Blot

To extract protein, the cells were lysed in ice-cold lysis buffer containing 50 mM Tris (pH 7.4), 5 mM EDTA, 250 mM NaCl, 0.1% Triton X-100, 50 mM NaF, 0.1 mM sodium orthovanadate, 1 mM phenylmethylsulfonyl fluoride (PMSF), and 1 × protease inhibitor cocktail. Equal amounts of proteins extracted from HC, Sham, and Real treated cells (10 µg) were resolved using SDS-PAGE gels (NuPage; Thermo Fisher Scientific) and transferred to a polyvinylidene difluoride (PVDF) Immobilon-p membrane (Bio-Rad, Hercules, CA, USA). Membranes were saturated with Tris-buffered saline with Tween-20 (TBS-T) containing 5% non-fat dry milk for at least 1 h at room temperature and then incubated overnight at + 4 °C under constant shaking, with the following primary antibodies: anti-GAP43 (1:2000), anti-βIII-TUBULIN (1:10,000), and anti-HSP70 (1:10,000). All antibodies were diluted in TBS-T containing 3% non-fat dry milk. Membranes were next incubated for 1 h with horseradish peroxidase-conjugated secondary antibody (1:4000), washed in TBS-T, and visualized using ECL SuperSignal West Pico PLUS Chemiluminescent Substrate. Densitometric analysis was performed using the iBright Imaging System (Thermo Fisher Scientific).

Immunofluorescence

Cells were differentiated in a six-well plate pretreated with poly-D-lysine, fixed for 10 min in PBS with 4% paraformaldehyde, permeabilized for 10 min in PBS-0.01% Triton X-100, and then blocked for 60 min in PBS-1% bovine serum albumin (BSA). Samples were incubated with anti-GAP43 (1:200) overnight at 4 °C and then washed and incubated with species-specific secondary antibodies conjugated to Alexa Fluor 488 for 60 min. Cells were counterstained with DAPI (1 μg/mL).

Real-Time qPCR

Total RNA, including miRNAs, was isolated from cells using the miRNeasy Tissue/Cells Advanced Kits and from serum using the miRNeasy Serum/Plasma Kit. RNA purified from all samples exhibited a 260/280 absorbance ratio of approximately 2.0, indicating high purity. For miRNAs’ analysis, 10 ng of RNA was reverse transcribed by TaqMan MicroRNA Reverse Transcription Kit and amplified with TaqMan Universal Master Mix II and TaqMan microRNA Assay (Table S1).

For mRNA expression analysis, 800 ng of RNA were reverse transcribed by a High-Capacity cDNA Reverse Transcription kit and amplified using PowerUp SYBR Green master mix. The sequences of the primers utilized in this study are presented in Table S2. The total amount of mRNAs and miRNAs was calculated by the 2^−ΔΔCt method relative to the reference genes: GAPDH for mRNAs, U6 for in vitro, and the mean between miR-21 and miR-16 for ex vivo analysis. The qPCR amplifications were obtained by a StepOne Plus Real-Time PCR System (Thermo Fisher Scientific).

Pathway Analysis

A sub-network of human miRNA–Protein interaction, specific for miRNAs modulated by rTMS, was selected starting from a database of experimentally validated has-miRNA-target interactions (hsa-MTIs, miRTarBase), using the software Cytoscape 3.10.2 (UC, San Diego, California, USA) [36]. The Reactome Cytoscape Plugin (v8.0.9) was used to access the REACTOME pathways stored in the database and perform the pathway enrichment analysis. ClueGO plug-in (v2.5.10) was used to cluster pathways (REACTOME Pathways 25.05.2022) as functionally grouped networks, using a two-sided hypergeometric test with a p-value cut-off of 0.05 and the Bonferroni correction method.

Enzyme-Linked Immunosorbent Assay (ELISA)

Serum samples collected from AD patients before and after rTMS exposure, AD patients subjected to placebo treatment, and from healthy controls were diluted tenfold. Quantitative measurement of BDNF protein was performed in duplicate by ELISA assay according to the manufacturer’s instructions.

Statistical Analysis

Statistical tests were performed with GraphPad Prism software v.7 (GraphPad, San Diego, CA). For the in vitro analysis, p-values were determined using a two-tailed t-test. In vitro experiments were performed five times, and data were expressed as mean ± standard error of the mean (SEM).

Statistical analyses of all ex vivo experiments were performed using a linear mixed-effects model (LMM) to test for Group × Time interactions. Prior to group comparisons, data distributions were assessed for normality using the Shapiro–Wilk test and for homogeneity of variances using Levene’s test. As both assumptions were violated and the sample sizes were unequal (Real n = 14; Sham n = 6), we employed an LMM, which is more robust under these conditions than a traditional repeated-measures ANOVA test.

The model included the expression level of miRNAs in serum as the outcome, with Time (T0, T1) as a within-subject factor and Group (Sham, Real) as a between-subject factor, with random intercepts for subjects. Given only two time points, sphericity was not applicable, and no correction (e.g., Greenhouse–Geisser) was necessary.

To explore the presence of correlation between rTMS effects in the ADAS-Cog score and in the miRNA levels, we computed the Kendall’s rank coefficient between the T1 and T0 score difference in the two different groups receiving Real or Sham rTMS. The resulting data were visualized using non-parametric locally estimated scatterplot smoothing (LOESS) plots. In addition to the separate Kendall’s tau correlations conducted for each group, we performed a formal statistical comparison of the correlation coefficients between the Real-rTMS and Sham-rTMS groups. Specifically, we applied a permutation test to assess whether the difference between the two tau coefficients was statistically significant. A total of 10,000 permutations were generated to create a null distribution of the difference between group-specific tau values, and the exact p-value was calculated from this distribution. Receiver operating characteristic (ROC) curve analysis was performed to evaluate the potential role of miRNAs in discriminating a group of “responder,” i.e., patients who received rTMS and showed an increase in cognitive abilities, from a group of “non-responder,” i.e., patients who received rTMS and did not show an increase in cognitive abilities, as assessed by ADAS-Cog. The area under the ROC curve (AUC) was calculated as a measure of overall accuracy. An AUC value of 0.5 was considered indicative of no discrimination; values closer to 0.7 indicated sufficient discriminatory power; values closer to 1.0 indicated excellent discriminatory power.

Results

rTMS-Modulated miRNAs in In Vitro AD Model

To identify potential markers of rTMS therapeutic efficacy, miRNAs involved in the regulation of synaptic plasticity and known to be deregulated in the early stages of AD were selected and analyzed in the in vitro AD model exposed to rTMS protocol as described above. It is worth noting that the in vitro model aligns with findings reported in the literature [34, 37] confirming its capability of recapitulating aspects of AD pathology.

We analyzed 18 miRNAs involved in synaptic plasticity in differentiated AD-like SH-SY5Y cells; 2 of them were excluded for low levels of expression (Table S3) and 3 for absence of rTMS modulation (Figure S1). Figure 3 shows rTMS-modulated miRNAs. In the condition that mimics AD exposed to placebo treatment (Sham), all miRNAs (miR-26b, miR-125b, miR-181c, miR-146a, miR-30b, miR-132, miR-25, miR-124) were upregulated compared to the normalized value of healthy cells, confirming their involvement also in the in vitro AD model. Following a single exposure to rTMS (Real), their expression decreased, recovering HC values 24 h after treatment. The decrease was statistically significant for miR-26b (p = 0.007), miR-125b (p = 0.027), miR-181c (p = 0.023), and miR-146a (p = 0.018), which are involved in synaptic maturity, neurotransmitter release, and immunomodulation, respectively [16, 27]. For other analyzed miRNAs (miR-30b, miR-132, miR-25, miR-124), their decreases were close to statistical significance. Overall, the results demonstrate that a single session of rTMS treatment can restore values of the miRNAs analyzed to control healthy cells in our in vitro AD models. The recovery of miRNA-regulated synaptic plasticity signaling pathways, disrupted by amyloid-β accumulation, suggests a potentially active role of identified miRNAs, which could represent markers of rTMS treatment efficacy.

Fig. 3.

Expression of miRNAs modulated by rTMS in differentiated SH-SY5Y cells treated with amyloid-β and exposed to Sham or Real rTMS. The black line indicates the normalized value of HC. Data are shown as mean ± SEM of n = 4/5 independent experiments. Differences were analyzed using a two-tailed t-test *p < 0.05; **p < 0.01

Direct and Indirect Target Genes of miRNAs Involved in the Regulation of Synaptic Plasticity

To understand the molecular mechanisms activated following rTMS treatment, we analyzed direct or indirect target genes of the identified miRNAs. As direct targets of miR-26b, we selected Retinoblastoma (Rb-1), neprilysin (NEP) [27, 38], and insulin-like growth factor 1 (IGF-1) [39] involved in cell death, neurite outgrowth, and amyloid-β plaque formation. Additionally, the ionotropic AMPA receptor subunit 2, GluA2 [40], relevant for synaptic transmission and plasticity, was investigated as a direct target of miR-30b and miR-124 (Figure S2). As indirectly modulated genes, we selected synaptic proteins such as postsynaptic density protein (PSD95) and synaptopodin (SYNPO), implicated in synaptic plasticity processes. Finally, we quantified the expression of βIII-TUBULIN as a marker of neuronal differentiation (Figure S2). We observed no significant modulations in the expression of either direct or indirect target genes following exposure to Real or Sham rTMS. Interestingly, the level of mRNA of PSD95, the main synaptic scaffold molecule regulating synaptic maturation, was significantly decreased compared to control samples (black lines in the graphs), respectively, in cells exposed to Sham (p = 0.046) and to rTMS (p = 0.007) treatment, indicating the accuracy of our experimental model of AD. The lack of PSD95 recovery may reflect the need for longer or repeated rTMS sessions to observe transcriptional or translational effects at the synaptic level.

Additionally, we tested the level of one of the most important presynaptic proteins involved in synaptic plasticity and directly regulated by miR-125b [41], the growth-associated protein 43 (GAP43), and the expression of βIII-TUBULIN, through immunofluorescence and Western blotting, 24 h after rTMS exposure (Figure S3, panels A, B, and C). Analysis of GAP43 revealed no significant differences between the AD mimicking group (Sham) compared to HC, while following rTMS exposure, there was a slight but not statistically significant increase (Figure S3B, C).

Pathway Enrichment

Attempting to better understand the mechanisms of rTMS stimulation, we carried out an extended pathway analysis of all the identified rTMS-responsive miRNAs. This approach offers a more comprehensive view compared to the analysis in the section, which was limited to a few target genes. Figure 4 shows the key pathways identified through the complex interactions among all target genes of all deregulated miRNAs (miR-26b, miR-125b, miR-181c, and miR-146a). Of these target genes, as shown in Fig. 4, 22.5% are associated with growth factor receptor signaling and second messengers. This is followed by 18.5% of genes involved in immune system signaling pathways, particularly those linked to Toll-like receptors. Additionally, around 15% of the genes are implicated in regulating cell proliferation and differentiation processes (Signaling by PTK6, Fig. 4), and 13.91% are involved in the aberrant regulation of mitotic G1/S transition. Finally, several minor signaling pathways were also identified. This analysis points to a broad range of pathways that deserve further exploration and understanding, covering various functions and activated mechanisms.

Fig. 4.

Pathway analysis by chart diagram

rTMS-Modulated miRNAs in AD Patients

The same rTMS-modulated miRNAs, identified in an in vitro AD model, were analyzed in the serum of patients. Firstly, we compared their expression in AD patients to age-matched HC to underline their potential role in the disease. As shown in Fig. 5A, we found a significant increase of miR-26b (p = 0.006), miR-30b (p = 0.030), and a significant decrease of miR-125b (p = 0.031) in AD patients at baseline (T0) compared to HC. No significant changes were observed in miR-25 and miR-146a between AD patients and HC (Fig. 5A). To assess the impact of rTMS treatment, we then evaluated changes from baseline (T0) to post-treatment (T1) in the Real and Sham groups using a linear mixed-effects regression model with random intercepts for subjects (fixed effects: Group, Time, and their interaction). This analysis revealed no significant Group × Time interaction for any of the analyzed miRNAs (Fig. 5B). Other miRNAs were analyzed in AD patients; however, they were excluded due to minimal target expression levels in all human samples (Table S3).

Fig. 5.

Expression of miRNAs in the serum of AD patients. A The graphs indicate the expression levels of miRNAs in serum patients before rTMS treatment (AD T0) compared to HC. B Expression of miRNAs in the serum of AD patients before (T0) and after (T1) Real or Sham rTMS treatment. The data are shown as mean ± SEM. Differences were analyzed using the Mann–Whitney test (AD T0 versus HC) and the LMM (to test Group × Time interactions) *p < 0.05; **p < 0.01

Cognitive Evaluation and Its Correlation with the Biological Effects of rTMS

To evaluate the therapeutic efficacy of the rTMS treatment, patients underwent a cognitive evaluation through the Alzheimer’s Disease Assessment Scale–Cognitive Subscale (ADAS-Cog11) [42] before (T0) and after (T1) Real rTMS treatment (Table 2). AD patients were classified as responders, those who received rTMS and showed improvement or stability in cognitive performance, or non-responders, defined as patients who did not show cognitive improvement as assessed by ADAS-Cog. Overall, 5 out of 14 patients (35%) showed cognitive improvement, and an additional 5 patients (35%) maintained stable performance. Only 4 patients (28.5%) were classified as non-responders, showing a worsening in cognitive performance, as indicated by an increase in ADAS-Cog scores. To explore possible relationships between the cognitive and biological effects of rTMS, we computed correlations between T1 and T0 differences of the ADAS-Cog of both Real-rTMS and Sham-rTMS groups and miRNAs. This analysis revealed a significant correlation between the (T1–T0) changes of ADAS-Cog score and miR-25 levels (τ = 0.331; p = −0.049) in the Real-rTMS group and not in the Sham-rTMS group (τ = 0.200; p = 0.360, Fig. 6A). In this study, we found that patients who showed improvement in cognitive ability demonstrated a corresponding decrease in miR-25 expression. To ensure the robustness of our statistical results, we carried out the permutation test comparing the Kendall’s tau coefficients between the Real-rTMS and Sham-rTMS groups, which revealed no significant difference (p = 0.787). This indicates that the correlation between miR-25 and cognitive change does not differ statistically between the two groups. No other miRNA displayed significant correlations (all p > 0.05) (Fig. 6A). Finally, to evaluate the potential capacity to discriminate a group of responders from a group of non-responders, and to quantitatively assess the potential biomarker’s performance, we performed a ROC curve analysis. ROC curve analysis evidenced that only miR-25 achieved an AUC of 0.950 with an accuracy of 85.7% and sensitivity of 90% and p < 0.001, indicating very good discriminatory performance (Fig. 6B). None of the other miRNAs were detected as significant predictors of therapeutic efficacy (all p > 0.05, data not shown). These results indicate that further investigations into a larger experimental cohort of patients will enable a more accurate assessment of miRNAs as suitable biomarkers for therapeutic efficacy.

Table 2.

The table shows the demographic and clinical characteristics of subgroups of patients showing improvement (*), maintenance (^), or non-responders with worsened performance ^(#) of cognitive functioning. MMSE Mini-Mental Score Examination, ADAS-Cog Alzheimer’s Disease Assessment Scale–Cognitive Subscale. The last column indicates the individual change in ADAS-Cog score (T1–T0)

Demographic characteristics				Clinical characteristics
Responders to rTMS
Patients	Age	Education	MMSE score	ADAS-Cog score T0	ADAS-Cog score T1	T1-T0
1	78	13	23	17	13.9	−3.1 *
2	72	13	18	43.3	36.6	−6.7 *
3	76	8	18	26.9	22.9	−4 *
4	78	13	18	29.2	24	−5.2 *
5	72	5	23	8.6	9.9	1.3 ^
6	71	13	23	25.9	25.6	−0.3 ^
7	56	18	21	27	28	1 ^
8	60	11	20	18.2	19	0.8 ^
9	69	17	24	11.9	12.3	0.4 ^
10	70	13	22	15.3	13.6	−1.7 *
Demographic characteristics				Clinical characteristics
Non responders to rTMS
Patients	Age	Education	MMSE score	ADAS-Cog score T0	ADAS-Cog score T1	T1-T0
11	71	8	23	12	16.6	4.6#
12	64	13	22	17	18.3	1.3#
13	75	17	22	13.5	19.6	6.4#
14	78	5	21	7.9	22.3	14.4#

Fig. 6.

Association between molecular changes and clinical outcomes. A Correlation analysis between clinical outcomes changes (T1–T0) in ADAS-Cog (x-axis) and ΔmiRNAs expression level (T1–T0). Black circles and red diamonds depict cases receiving Real or Sham-rTMS, respectively. B ROC curve for miR-25 to discriminate responder vs. non-responder patients. Sensitivity is plotted against 1-specificity. The p-value associated with our AUC is p < 0.001. The 95% confidence interval of our AUC is [0.836–1]

Evaluation of the BDNF Response Induced by rTMS Treatment

Brain-derived neurotrophic factor (BDNF) is a key upstream regulator of LTP in brain regions. We measured serum levels in AD patients at baseline (AD T0) and HC, showing a slight but not statistically significant increase in AD patients (Fig. 7A). This aligns with findings by [43], who suggested BDNF as a potential biomarker of compensatory mechanisms in mild cognitive impairment. Comparing BDNF level pre- and post-rTMS (Real or Sham), no significant differences were observed between groups (Fig. 7B) by LMM analysis.

Fig. 7.

BDNF levels in the serum of AD patients. A Comparison between AD patients and HC at baseline (AD-T0), B BDNF levels in AD patients before (T0) and after (T1) Real or Sham treatment. Differences were analyzed using the Mann–Whitney test (AD T0 versus HC) and the LMM (to test Group × Time interactions). Data are presented as mean ± SEM

Discussion

In this study, we investigated miRNAs’ modulation following rTMS treatment in AD patients by using two different approaches that combined an in vitro model and human AD patients. We firstly selected 18 miRNAs known to be impaired in AD pathology and measurable in serum. Using an AD cellular model exposed to the same rTMS dose as patients [32], we observed that a single rTMS session significantly reduced the expression of miR-26b, miR-125b, miR-181c, and miR-146a compared to the AD Sham group, restoring levels closer to controls. Additionally, miR-30b, miR-132, miR-25, and miR-124 showed a near-significant decrease. No modulation was detected in other miRNAs. Notably, pro-inflammatory microglia release exosomes enriched with pathogenic miRNAs, including miR-30b, miR-125b, and miR-146a, which promote neuroinflammation, enhance amyloid genesis, and disrupt neuron-specific phosphoproteins involved in neurotrophic support and synaptic signaling [15, 44].

To clarify molecular mechanisms activated by rTMS, we analyzed targets modulated by the identified miRNAs at both gene and protein levels and involved in cell death, neurite outgrowth, amyloid-β plaque formation, synaptic transmission, plasticity, and neuronal differentiation. No significant changes were observed in the expression of downstream genes (RB-1, NEP, IGF-1, and GLUA2) and proteins (GAP43, PSD95, and βIII-TUBULIN) following rTMS, and it may be due to the short time between treatment and evaluation (24 h) fixed for the experimental setting. Moreover, since each miRNA interacts with multiple target genes, regulation could be complex to elucidate and not immediately detectable. Interestingly, the level of mRNA of PSD95, the main synaptic scaffold molecule regulating synaptic maturation, was significantly decreased compared to control samples in both cells exposed to Sham and to rTMS treatment. Although PSD95 level was not rescued, this may indicate an early engagement of neuroplasticity-related pathways, observed by miRNA modulation, that precedes detectable changes in gene or protein expression. Slight, non-significant increases in GAP43 and βIII-TUBULIN in the rTMS group suggest a potential enhancement of neuronal plasticity and cytoskeletal remodeling, possibly reflecting a compensatory mechanism to counteract AD-related synaptic deficits. While further studies with extended observation periods are required to confirm this hypothesis, these findings support the notion that rTMS may promote neuronal plasticity even without detectable changes in direct miRNA targets. To further investigate the biological pathways and mechanisms activated by rTMS exposure, we carried out, starting from all in vitro rTMS-responsive miRNAs, a pathway analysis through bioinformatics tools. This allowed us to correlate all target genes associated with the deregulated miRNAs. Our results indicated that the primary signaling pathway involved is associated with growth factor receptor signaling and second messengers. This suggests that in our in vitro AD model, rTMS targets neuronal structures at the membrane level, and this can be reflected in positive or negative modulation of receptor signaling in the central nervous system, as also observed by Madnani and colleagues [2]. As a next step, we aim to employ NGS-based screening to characterize the full spectrum of rTMS-responsive miRNAs, thereby facilitating deeper insights through pathway enrichment analysis.

Our model, as described by de Medeiros [34], represents a well-established neuronal model of AD, but it does not include other cell types such as microglia or astrocytes. Since miRNA regulation is highly complex [31, 45, 46], we complemented this neuronal model with the analysis of circulating extracellular miRNAs in the serum samples for patients treated with rTMS. Together, these approaches provide a broader understanding of miRNA dynamics in AD. Upon analyzing patient samples, we observed notable changes in miRNA expression levels. Specifically, we found a significant increase in miR-26b and miR-30b, alongside a significant decrease in miR-125b in AD patients before rTMS treatment (AD T0), compared to age-matched HC, confirming an active role of these miRNAs in AD disease. In fact, miRNAs can act within cells or be released via exosomes to influence gene expression at a distance, and their stability in circulation may indicate a role in early disease progression [31]. Altered levels of circulating miRNAs are common in neurodegenerative diseases such as AD [47]. However, assessing the impact of rTMS treatment in our samples, we found no significant changes in the expression levels of the analyzed miRNAs post-rTMS (AD T1 Real) compared to baseline. This finding suggests that their roles could be limited in the regulation inside the neuron cells, as shown in our in vitro results. As expected, the placebo group (AD T1 Sham) showed no significant differences in miRNA expression compared to baseline (AD T0 Sham) across all the analyzed miRNAs. Further investigations into larger cohorts are warranted to confirm and better elucidate the potential rTMS’s effect. Furthermore, to evaluate a potential correlation between the modulation of miRNAs and the therapeutic efficacy of rTMS treatment in AD patients, we compared the expression changes of previously analyzed miRNAs with cognitive outcomes conducted before and after rTMS treatments. Interestingly, our analysis showed that the decrease in miR-25 levels was significantly correlated with a reduction in ADAS-Cog scores, an indicator of cognitive improvement, in the rTMS-treated group, indicating that changes in miR-25 expression may be linked to the cognitive benefits in AD rTMS treated-patients. However, when we carried out a direct comparison of the correlation coefficients of the Sham and Real-TMS groups, we found no statistically significant difference between these groups, probably due to the small cohort of patients available. This finding indicates that this miRNA can be considered as a reliable prognostic marker of cognitive improvement rather than highlighting a treatment specific effect. Furthermore, the ROC curve analysis demonstrated that miR-25 can reliably distinguish responders from non-responders, possessing a discriminatory power to identify patients likely to benefit from rTMS and highlighting its potential relevance as a valuable candidate marker of therapeutic efficacy, to be validated in a larger cohort of patients. None of the other miRNAs showed significant correlations or predictive capacity, highlighting the unique relevance of miR-25 in this context. Notably, our results are consistent with the findings of Duan and colleagues [29], who demonstrated that elevated levels of miR-25 exacerbate hippocampal neuron injury in AD by downregulating the target gene expression, thereby inhibiting neuronal proliferation and promoting apoptosis. In our study, a decrease in miR-25 levels after rTMS treatment could relieve the downregulation of key target genes, contributing to cognitive improvement in AD. Furthermore, Guo and colleagues [30] observed that rTMS treatment markedly increased the expression of miR-25 in the ischemic cortex, supporting its potential as a biomarker to guide personalized rTMS interventions. The next step will be to transition our findings into an in vivo AD animal model and/or cell co-cultures to evaluate and further validate current findings from a single system. This new step will allow us to assess the effects of rTMS not only on blood biomarkers but also on brain tissue. This would help gain a more comprehensive understanding of molecular mechanisms, particularly focusing on regions involved in memory, such as the hippocampus. Altogether, our findings suggest, for the first time, that rTMS may modulate specific miRNAs in AD patients, with miR-25 representing a pivotal key target for future validation studies in a larger patient’s cohort.

Finally, we investigated the plasma levels of BDNF, providing valuable insights into the role of this neurotrophin in the context of rTMS treatment for AD patients. Elevated BDNF levels may reflect a neuroprotective mechanism in the early stages of AD, where increased BDNF represents a compensatory response to synaptic dysfunction and neuronal loss [43]. Indeed, in our cohort, we observed a slight but not statistically significant increase in BDNF levels in AD patients compared to HC. However, the comparison of BDNF levels between Sham and rTMS-treated AD patients before and after stimulation showed a trend of decreasing BDNF levels in both groups, suggesting a potential deterioration of BDNF signaling. Previous studies have shown that rTMS can elevate BDNF levels and reduce oxidative stress in conditions such as treatment-resistant depression, stroke, and AD [48–50]. This effect is crucial as BDNF is an important upstream regulator of LTP, which is essential for learning and memory processes [51]. Although research specifically focusing on AD is limited, future studies in a larger cohort of patients will be necessary to clarify BDNF expression patterns after rTMS treatment to assess its possible involvement in the cognitive improvement observed in rTMS-treated patients.

Conclusions

In summary, this study highlights, for the first time, the modulation of miRNA networks by rTMS exposure. We found a significant correlation between cognitive improvement and a decrease of miR-25, suggesting the prognostic value of this molecular indicator in AD. In addition, ROC curve analysis revealed that miR-25 has a discriminatory power to identify patients likely to benefit from rTMS, supporting its potential as a valuable biomarker of rTMS efficacy. Overall, our findings provide initial evidence that miRNAs may reflect the rTMS response, warranting validation in a larger cohort of patients. Even though we used an innovative approach with an in vitro experimental setting based on the dosimetric translation of the rTMS-induced electric field in AD patients, which drove an ex vivo investigation in those patients (diagnosed with evidence of positive AD biomarkers), the major limitation of the current study is the paucity of patients enrolled in the rTMS protocol, which affected the statistical significance of our results. Nonetheless, the findings observed encourage further investigations to better clarify the relationship between miRNAs’ expression in AD patients after rTMS treatment and cognitive improvement.

Supplementary Information

Below is the link to the electronic supplementary material.

Author Contribution

Conceptualization: A.C., B.T., C.M., and G.K.; biological experiments and data analysis (including Bioinformatics): A.C., E.C. (Colantoni), F.C., B.T., E.F., and C.M; patients recruitment and cognitive assessment: L.M., S.B., F.D.L., E.C. (Casula), and G.K.; writing—review and editing: A.C., E.C. (Colantoni), B.T., and C.M.; funding acquisition: C.M. All authors contributed to the discussion and interpretation of the results, participated in editing and proofreading, and approved the final version of the manuscript for publication.

Funding

This work was supported by “Progetti gruppi di ricerca 2020” Regione Lazio POR FESR Lazio 2014-2020 n. A0375-2020-36546, Dosimetria di un nuovo trattamento rTMS in colture 3D della malattia di Alzheimer per l’identificazione di marcatori di efficacia terapeutica - DTA - CUP I8F21000950009, and partially supported by the Italian Ministry of Health (Ricerca Corrente 2026).

Data Availability

No datasets were generated or analysed during the current study.

Declarations

Ethics Approval

The clinical trial providing patients’ MRI has been approved by the Ethics Committee of the Santa Lucia Foundation CE/PROG.716, Rome, Italy. All patients entering the study provided written informed consensus.

Competing interests

The authors declare no competing interests.