Long–Evans rats (Charles River Laboratories), used for the ex vivo and in vivo experiments, were individually housed and maintained under specific pathogen-free conditions on a 12-h light/dark cycle at the National Institute on Aging/National Institute on Drug Abuse (NIA/NIDA) animal facilities in the Biomedical Research Center (Baltimore, MD). Standard rat chow and water were available ad libitum throughout the experiments. All procedures were approved by the Animal Care and Use Committee of the Intramural Research Program of the NIA.

To establish the baseline cognitive status of the animals used in the ex vivo experiment, rats were tested in a ‘place’ version of the Morris water maze task, as previously described (Gallagher et al., 1993). The Morris water maze is a widely recognized tool for investigating spatial memory and learning in rodents, and the procedure used here has been extensively validated as a test for neurocognitive aging (Rapp et al., 1987; Rapp and Gallagher, 1996; Haberman et al., 2012; Gallagher et al., 2015; Tomás Pereira and Burwell, 2015).

Training continued over 8 consecutive days, three training trials per day. Every other day, the third trial was a probe in which the platform was inaccessible for 30 s. A learning index score was calculated for each animal from their average proximity to the escape platform during training; lower scores indicate better task performance. Aged rats that performed on par with young animals were denoted aged unimpaired (AU), while rats that scored greater than the young were classified as aged impaired (AI; Spiegel et al., 2013; Supplementary Figure 1). To control for non-mnemonic deficits, rats were tested in a single session of a hippocampus-independent cued water maze protocol the following day. No animals that performed outside the normal range on this version of the task were included in the present experiments.

rTMS was applied using a Magstim Rapid² stimulator with a 70-mm figure-eight coil (The Magstim Company, Whitland, Dyfed, UK). 1 Hz stimulation was applied in 5 blocks of 600 pulses lasting 10 min repeated at 15-min intervals for a total of 3,000 pulses in 70 min. iTBS was administered in 5 blocks repeated every 15 min with each block consisting of 20 trains of 3 50-Hz pulse bursts repeated at 5 Hz for 2 s with a 10-s inter-train interval as described elsewhere (Huang et al., 2005). Each iTBS block consisted of 600 pulses in 192 s for a total of 3,000 pulses delivered in 63.2 min.

As shown in Table 1, stimulation resulted in hundreds of significant transcriptional changes between rTMS and sham-treated controls in all experimental models, with both increases and decreases in each treatment group. While there are considerable overlaps between rTMS treatments, each sample-treatment combination produced a complex pattern of transcriptional response (Figure 2). This was evident in the in vitro, ex vivo, and in vivo experimental models. The transcriptional response was quite dynamic, often with both increases and decreases in the same gene and gene families with time, stimulation protocol, and relative to brain region. The complete set of statistically significant transcriptional changes due to rTMS can be found here (Supplementary Table 1).

Table 2 (in vitro), Table 3 (ex vivo), and Table 4 (in vivo) show groups of selected genes with altered transcription (|z-ratio| > 1.5) implicated in complex neuronal processes in all three models tested. The categories of these genes include neurotransmission, synaptic function, inflammation, myelination, and cell–cell adhesion among other functional gene classes.

With the in vitro model using rat hippocampal neuronal cultures, genes involved in inhibitory and excitatory neurotransmission were altered, both upregulated and downregulated (Table 2), which included multiple GABA (Figure 3A) and glutamate receptors (Figure 3B) such as Gabbr1,2; Grik1,4; Grm3-7, and Gabra4 (Heidelberg et al., 2013) as well as genes involved in learning and memory-related plasticity. In addition, the transcriptional response was altered in oxidative phosphorylation pathways including shared genes involved in Huntington’s, Alzheimer’s, and Parkinson’s disease (Figures 4A,B). In the ex vivo and in vivo models (Tables 3, 4) altered transcription of additional neuronal genes was shown to include genes involved in neurotransmission [Ptk2b (Brys et al., 2013), Slc6a13 (Christiansen et al., 2007)]; choline transport [Slc5a7 (Ribeiro et al., 2006), Ryr2 (Abu-Omar et al., 2017), Chrna5 (Proulx et al., 2014)]; synaptic function [Grin3a (Glun3a; Perez-Otano et al., 2016)]; plasticity (Arc; Tomas Pereira et al., 2015, Cnp; Barmashenko et al., 2014); learning and memory [Arc; Morin et al., 2015; Tomas Pereira et al., 2015), Grin2d, (Glun2d; Jacobs et al., 2014)], and cognition (Ryr2; Liu et al., 2012, Arc; Morin et al., 2015, Sorl1; Li et al., 2017), among others.

Coordinate transcriptional upregulation occurred in a cohort of myelin regulatory genes at 2 h post-rTMS stimulation in the ex vivo model (Table 3). These include increases in Mobp, Mog, Cnp, Erbb3 Klk6, Cadm4, and transferrin (Tf). However, the myelin regulatory factor Sox10 was downregulated in the iTBS-stimulated ex vivo samples. Myelin-related transcriptional induction was not statistically significant at 48 h for in vivo or at 2 h for in vitro cultures. Moreover, gene expression from nine Claudin family members was both increased and decreased in the ex vivo model (Table 3). Claudins mediate cell–cell contact and blood–brain barrier integrity. A shared protein domain network showing interrelationships between the TMS transcriptionally altered Claudins is shown in Figure 5.

rTMS produced a substantial reduction in transcripts involved in immune and inflammatory processes in the ex vivo and in vivo samples. Most striking was the coordinate downregulation of many members of the classical complement pathway in the ex vivo rTMS model with both 1 Hz and iTBS (Table 3) stimulation after 2 h. In addition, the expression of complement factors C1s and C2 was reduced in the in vivo model in the young hippocampus at 48 h (Table 4). In contrast, C1qb, C1qc, C2, C3, C4a, and C4b were increased in the AI hippocampus using iTBS in the ex vivo model (Table 3). Although generally thought of as mediators of innate immunity, members of the complement cascade have also been shown to be important in synaptic remodeling (Presumey et al., 2017), and have been associated with schizophrenia (Nimgaonkar et al., 2017), age-related macular degeneration (Lu et al., 2018), and Alzheimer’s disease (Torvell et al., 2021).

Additional inflammatory genes that were transcriptionally repressed by rTMS included Il1b, Il1rn, Cd74, and Il12b in the ex vivo model following iTBS (Table 3), and Tlr3, Cd74, and RT1-Da in the in vivo model (Table 4). Moreover, the gene for S100b, a marker of neuroinflammation and neuronal damage, was markedly downregulated by rTMS in all in vivo samples (Table 4). Importantly, immune regulatory molecules as a group were not significantly altered by rTMS in the in vitro model using purified neuronal cultures suggesting that modulation of immune transcripts in both ex vivo and in vivo models (both complex tissue intact samples) were from non-neuronal cells including resident glial cell populations.

Notably, expression was altered among multiple genes that have been studied in disorders for which rTMS is used clinically, including depression [Slc6a4 (Lam et al., 2018), S100b (Schroeter et al., 2014), Il18 (Prossin et al., 2011; Bufalino et al., 2013; Kim et al., 2017), Il1b (Bufalino et al., 2013), Htr2c (Brummett et al., 2014), Gabra6 (Inoue et al., 2015)], epilepsy [Grin2d (GluN2D; Li D. et al., 2016), Gabra6 (Prasad et al., 2014), Scn1b (Ramadan et al., 2017), Scn3a (Lamar et al., 2017), Kcnt1 (Evely et al., 2017)], schizophrenia [Rgs4 (Schwarz, 2017), Grin1 (Zhao et al., 2006), Grm5 (Matosin et al., 2017), Pde10a (Boden et al., 2017), C4a (Nimgaonkar et al., 2017)], bipolar disorder [Pde10a (McDonald et al., 2012); S100b (da Rosa et al., 2016)], Parkinson’s disease [Nr4a2 (Liu et al., 2017)], Alzheimer’s disease and other dementias [Ptk2b (Li Y. Q. et al., 2016), Ttr (Silva et al., 2017), Nr4a2 (Montarolo et al., 2016), Ryr2 (Briggs et al., 2017), Mobp (Irwin et al., 2014), Cd40 (Giunta et al., 2010), Grn (Sudre et al., 2017), Hmo1 (Sung et al., 2016), Arc (Bi et al., 2017)], stroke [Lgals3 (Gal3; He et al., 2017), Ace (Wei et al., 2017), Cd40 (Huang et al., 2017), Cxcl12 (Shen et al., 2017), Hmgb1 (Choi et al., 2017)], as well as substance abuse [Slc6a4 (Bauer et al., 2015), Penk (Moeller et al., 2015), Rgs4 (Ho et al., 2010), Chrna5 (Lassi et al., 2016; Olfson et al., 2016)]. Overlap between the complex functional processes, disorders, and genes mentioned here highlights a central role of shared fundamental neuronal pathways in multiple processes and distinct neurological disorders (Brainstorm et al., 2018; Gandal et al., 2018).

Moreover, TMS altered the expression of genes studied in other disorders, suggesting additional potential clinical applications of TMS. These include hereditary motor neuropathy [Slc5a7 (Barwick et al., 2012)], congenital myasthenic syndrome (Slc5a7; Bauche et al., 2016), age-related hearing loss (Gabra6; Sun et al., 2014), Canavan disease (Gabra6; Surendran et al., 2003), gout (Abcg2; Yu et al., 2017), pancreatitis (Cldn2; Giri et al., 2016), as well as numerous genes involved in inflammatory and immune disorders. Additionally, TMS alters the transcription of many genes of unknown function. These can be found in Supplementary Table 1.

Gene set analysis is based on gene expression changes in functionally related groups of genes, as opposed to individual genes, resulting in statistically significant aggregate scores for each gene group. Figure 6 displays a marked reduction in Gene Ontology (GO) inflammatory gene sets in the ex vivo young hippocampus using 1 Hz stimulation versus the unstimulated sham. These include reductions in Innate Immune Response, Complement pathways, as well as Chemokine and Cytokine GO gene sets. Broad-based immune suppression was found in the ex vivo young hippocampus model using iTBS versus sham as well, including gene sets for Inflammatory Response, Chemotaxis, Antigen Processing and Presentation, Immunoglobulin Mediated Immune Response, among others (Figure 7). In addition to marked immune suppression, upregulated GO gene sets in the iTBS ex vivo model included a theme of mitochondrial and energy-related GO gene groups. Immune suppression is evident in the in vivo model in the cortex after iTBS stimulation: GO gene sets such as Immune Response, Complement Activation, and Chemokine activity, among others, were downregulated, and upregulated gene sets included Synaptic Vesicle, Synaptic Transmission, and Neurite Development (Figure 8). In the in vivo model in the hippocampus, a theme of translation and ribosomal-related gene sets were upregulated after iTBS (Figure 9). For the in vivo iTBS model, alterations in GAD sets for inflammatory disorders such as Scleroderma, Crohn’s disease, and Lupus, were found in both cortex and hippocampus, as well as disorders specifically relevant to TMS treatment including mood disorders, seasonal affective disorders, attention-deficit/hyperactivity disorder, schizophrenia, substance use related disorders, among others (Supplementary Figures 2–5).

Using rat hippocampal neuronal cultures, we observed that rTMS regulated the expression of genes involved in inhibitory and excitatory neurotransmission mediated in part by glutamatergic and GABAergic synapses. The evidence supporting this argument can be found in Table 2 and Figures 3A,B. In Table 2, we present results highlighting the specific genes that exhibit alterations following rTMS. These findings demonstrate that rTMS had a notable impact on the expression of various glutamatergic and GABAergic genes. In addition, Figures 3A,B depict the changes in gene expression induced by rTMS. These figures illustrate the specific genes in the glutamatergic and GABAergic pathways that are significantly modulated by rTMS. The observed alterations support the suggestion that rTMS is positioned to potently influence inhibitory and excitatory neurotransmission.

Alterations in the glutamatergic system can promote excitotoxic cell death, comprising a potential mechanism of neurodegeneration in Alzheimer’s disease (Mufson et al., 2008; Wang and Reddy, 2017). Early aberrant excitatory neurotransmission is frequently observed in both animal models of Alzheimer’s disease (Palop et al., 2007) and patients (Scott et al., 2002), and blocking the action of glutamate and targeting excitatory synapses has been exploited as a potential pharmacological treatment for the disease. Altered oscillatory rhythmic activity and network hypersynchrony are also features of Alzheimer’s disease (Koch et al., 2022) and may contribute to cognitive impairment (Busche and Konnerth, 1700; Palop and Mucke, 2016). However, while a major emphasis in prior research has focused on dysfunction of the glutamatergic system, accumulating evidence links inhibitory GABAergic interneurons to excitatory/inhibitory imbalance as a potential early contributor to cognitive impairment in both aging (Oh et al., 2010; Gallagher et al., 2019) and disease (Li Y. et al., 2016; Govindpani et al., 2017).

Disruption in excitatory/inhibitory balance has also been associated with other neuropsychiatric and neurological conditions such as chronic stress (Varga et al., 2017), major depression disorder (Pehrson and Sanchez, 2015; Fee et al., 2017; Fogaca and Duman, 2019), autism spectrum disorder (Schur et al., 2016), bipolar disorder (Bialer, 2012), epilepsy (Ye and Kaszuba, 2017; Losi et al., 2019) and schizophrenia (Metzner et al., 2019), and normalizing altered inhibitory function represents a promising target for therapeutic intervention. rTMS has been identified as a modifier of GABAergic (Trippe et al., 2009; Jazmati et al., 2018) and glutamatergic (Yang et al., 2014) systems previously and preventing or reversing disease-related imbalance by targeting either system with rTMS could be of both symptomatic and disease-modifying value.

A major insight from our analysis is the broad transcriptional change related to immunosuppression, observed as early as 2 h after stimulation in hippocampal slices. Neuroinflammation has emerged as a key feature in the pathogenesis of Alzheimer’s disease, potentially playing a causative role rather than simply as a secondary consequence of the pathological cascade (Sheng et al., 2003). Given the failure of amyloidogenic drugs to provide therapeutic benefit, scientific interest has shifted to other features of neurodegeneration including neuroinflammation (Castello et al., 2014). Similarly, major depressive disorder has been associated with increased activation of the immune system (Lee and Giuliani, 2019), and many front-line pharmaceutical treatments for depression have been shown to reduce inflammatory activation and lower circulating cytokine levels (Galecki et al., 2018).

rTMS has proven effective in animal models as a therapeutic tool targeting the inflammatory response (Sasso et al., 2016). Our results are consistent with a potential anti-inflammatory benefit, demonstrating that stimulation-induced downregulation of genes related to immunosuppression in young hippocampal slices 48 h following treatment. Strikingly, however, while decreased neuroinflammatory gene transcription was the predominant effect in young, rTMS had the opposite effect in hippocampal slices from aged animals with cognitive impairment (AI), inducing a potential proinflammatory transcriptional response. Given that aging itself is associated with chronic increases in circulating levels of inflammatory markers (Singh and Newman, 2011) – a phenomenon exacerbated by age-associated diseases (Chung et al., 2009) – our results suggest that rTMS might exacerbate age-related pathological immune activity. Thus, rTMS treatments with demonstrated anti-inflammatory benefits in young adults may have unanticipated effects in older recipients. Future studies focusing on the neuroinflammatory effects of rTMS specifically in the context of aging will be needed to move this technology forward in the clinical setting.

Although the mechanisms that mediate neuroinflammation are not fully understood, it is well-accepted that microglia play a key role (DiSabato et al., 2016; Dokalis and Prinz, 2019). In our study, the comparative analysis of purified hippocampal neuronal culture preparations versus the ex vivo or in vivo complex tissue samples, where glial cells are present, suggests that the observed influence on inflammatory response genes likely arises from the non-neuronal compartment. Taking advantage of the unique dataset generated in these experiments, future analyses directly comparing the transcriptional response to rTMS as a function of experimental preparation, temporal kinetics, age, and cognitive status will provide a rich source of insight into the complex molecular consequences underlying the phenotypic response to intervention.

Although FDA-approved as a stand-alone treatment for non-responsive major depressive disorder, in clinical practice rTMS is typically used as a supplementary therapy, together with psychotropic medication. A therapeutic benefit of combined rTMS and adjunctive drug treatment for major depression has been confirmed previously (Wei et al., 2017), but evidence fully exploring the interaction between rTMS and anti-depressants is scarce. Psychotropic drugs affect cortical excitability and plasticity (Minzenberg and Leuchter, 2019) and medications commonly used to treat neurological conditions change neural circuit and network activation (Borchert et al., 2016; Zheng et al., 2017). The use of concomitant medication may impact rTMS treatment outcome favorably or unfavorably, depending on the drug category or mechanism of action (Hunter et al., 2019). For example, while still in their infancy, studies in pharmaco-TMS have shown that medications that block voltage-gated sodium channels increase the evoked motor threshold to TMS, i.e., a common metric used to normalize and titrate stimulation intensity across individuals (a readout for TMS dosage; Ziemann, 2004; Ziemann et al., 2015).

The idea that the effects of rTMS vary depending on the use of concurrent medication aligns with the ‘state-dependency’ concept of TMS (Silvanto and Pascual-Leone, 2008). Among the factors that might influence the outcomes of TMS, the pharmacological ‘state’ of the brain has received relatively limited attention, likely because patients treated with rTMS for neuropsychiatric conditions are typically receiving concurrent pharmacological treatment. In this context, our ex vivo experiments point to one potentially useful approach for research aimed at identifying drug-rTMS interactions. Overall, while recognizing that the barriers to successful translation are substantial, properly designed basic research can nonetheless inform therapeutic development toward safer and more effective TMS application in a variety of conditions.

It has been over a decade since the first rTMS application was approved by the FDA for the treatment of major depressive disorder, and since then the range of potential applications under investigation with rTMS has skyrocketed. The research on rTMS as a therapy includes patients with a variety of disorders and pathological signatures, from all ages and different cognitive statuses. Recent in vivo data points to the potential translational relevance of findings in animal models demonstrating effects on behavior and structural plasticity (Cambiaghi et al., 2022) and the potential implication for clinical application in neurodegenerative diseases (Weiler et al., 2020). In our results, we observed that rTMS induced unique gene expression profiles in each experiment, all of which reflected a variety of biological conditions and different stimulation parameters. In addition, the gene expression response following rTMS was highly complex. Whereas previous studies investigating the effects of rTMS have focused on alterations in only a handful of priori-selected genes (Lee et al., 2014; Wang et al., 2014; Grehl et al., 2015; Cirillo et al., 2017; Legrand et al., 2018; Cui et al., 2019; Wu et al., 2022), our large-scale analysis of gene expression detected several complex different pathways that are altered after stimulation.

Any non-invasive brain stimulation protocol intended to target a specific gene set or mechanism in the context of treating a given condition may have unexpected consequences in a different context, e.g., in other brain regions, or as a function of the cognitive status and age of the recipients. For example, clinical trials of rTMS (Turriziani et al., 2012) have reported that the same stimulation protocol that yields cognitive improvement in some subjects at risk for Alzheimer’s disease has detrimental effects in cognitively healthy individuals. Likewise, stimulation aiming at targeting a specific pathway or mechanism may induce changes in other genes and unintended downstream pathways. The findings reported here also highlight that the transcriptional response can be temporally dynamic, resulting in both increases and decreases in the same gene families across time, suggesting that a protocol chosen for its acute effects may lead to unpredictable long-term changes in expression.

Taken together, our results suggest that patterns of gene expression following rTMS are complex, dynamic, and dependent on the brain region, age, cognitive status, and potentially many other subject variables not examined here. Given the current and growing clinical application of rTMS, it is timely that greater attention turns to basic research aimed at understanding the underlying basis of reported therapeutic benefits, with the goal of optimizing treatment in the context of the individual patient.

The limitations of research on non-invasive brain stimulation in experimental animal models are significant. A primary limitation from a translational perspective is that, although we analyzed multiple stimulation protocols, we only studied the effects of one stimulation session, while clinical rTMS approaches employ treatments over the course of days or weeks. Further studies investigating the effects of rTMS on a long-term basis should be conducted. Another limitation is that although in some cases we used a diverse parametric setting such as different ages and cognitive states, our initial report is predominantly focused on the effects of rTMS itself. Here, each rTMS-stimulated sample group was statistically compared to its own unstimulated sham control, as opposed to comparisons across age and between cognitive groups. Detailed results of the transcriptional changes due to rTMS treatment in the context of age, cognition, and brain region will be presented elsewhere.