Using computational approaches to estimate cortical electric fields induced by TMS in combination with simulations of the effects of electric fields on neurons, some insights into TMS effects on neural tissue have been gained (e.g., Basser 1994; Opitz et al. 2011; Rotem and Moses 2008; Rusu et al. 2014). Nevertheless, it has remained largely unknown how TMS affects individual neurons within distinct cortical networks (Dayan et al. 2013).

A major limitation in this field of research has been the challenge to record from individual neurons during stimulation, due to the strong electromagnetic field induced by TMS. Recent technical advances, however, have made it possible to assess neural activity during stimulation using electrophysiological (Muller et al. 2014; Pashut et al. 2014) or functional optical imaging techniques (Kozyrev et al. 2014). These studies provide experimental evidence that single-pulse magnetic stimulation initiates action potentials preferentially in low-threshold interneurons (Pashut et al. 2014), resulting in a suppression of the stimulated cortex for about 200 ms after stimulation (Kozyrev et al. 2014). Conversely, high-frequency repetitive magnetic stimulation (10 Hz; or single-pulse stimulation with higher intensity) seems to shift the balance between excitation and inhibition toward excitation (Kozyrev et al. 2014). Additional work is now required to better understand how structural and functional properties of individual neurons and specific network architectures influence the outcome of single-pulse and repetitive magnetic stimulation.

In this context, recent work has also indicate that rMS may assert its effects by simultaneously depolarizing pre- and postsynaptic neuronal compartments, i.e., through the induction of both anterograde and backward propagating action potentials (Lenz et al. 2015). Hence, simultaneous recordings of distinct cells or dual recordings from individual neurons, e.g., somato-dendritic recordings, are expected to provide new important insights into the effects of rTMS during stimulation at the single-cell level. The impact of rTMS on non-neuronal cell types in the brain (e.g., astrocytes, microglia, oligodendrocytes, endothelial cells, immune cells) remains to be determined (Fig. 2.1a).

Early reports in human subjects have demonstrated that rTMS can increase or decrease cortical excitability beyond the stimulation period (Chen et al. 1997; Ziemann et al. 2008). It was noted that stimulus intensity, frequency, and the state of the stimulated network influence the duration, direction, and extent of rTMS-induced changes in cortical activity (for details, see Chap. 1). These after-effects of rTMS have been assumed to represent changes in synaptic efficacy and were therefore termed “long-term potentiation and depression (LTP/LTD)-like” phenomena, respectively. Hence, it was proposed that rTMS could assert its beneficial effects in the context of neurological and psychiatric disease by interfering with Hebbian forms of plasticity, e.g., LTP/LTD, which is considered to underlie learning and memory processes. Accordingly, animal studies have been employed to assess the effects of rTMS on synaptic plasticity. Initial experimental evidence for rTMS-induced synaptic activity was derived from immunostainings for immediate early gene (IEG)-encoded proteins, such as c-fos and zif-268 (e.g., Barth 2007; Loebrich and Nedivi 2009; Okuno 2011; Smeyne et al. 1992), which are recruited in the early stage of synaptic plasticity. Although robust experimental evidence has been provided that rTMS recruits IEG-encoded proteins, increased levels of c-fos and zif-268 were observed independent of stimulation frequency and pattern (Aydin-Abidin et al. 2008; Hausmann et al. 2000, 2001; Hoppenrath and Funke 2013; Volz et al. 2013). Yet, it was noted that rTMS may activate distinct brain regions and specific neurons within stimulated networks (Ji et al. 1998). Likewise, immunostainings for presynaptic (Vlachos et al. 2012; Volz et al. 2013) and postsynaptic markers (Gersner et al. 2011; Lenz et al. 2015; Ma et al. 2013; Vlachos et al. 2012; Fig. 2.1b) provided evidence that synaptic changes may underlie rMS-induced plasticity. More recent work in organotypic slice cultures was able to provide direct experimental evidence at the single-cell level that rMS is capable of inducing long-lasting functional and structural synaptic plasticity, that is an N-methyl-D-aspartate receptor (NMDAR)-dependent, Ca²⁺-mediated enlargement of dendritic spines and strengthening of excitatory synapses (Vlachos et al. 2012; Lenz et al. 2015). These studies are in line with earlier in vivo and in vitro work (e.g., Levkovitz et al. 1999; Tokay et al. 2009) supporting the notion that rTMS of the human cortex may induce Hebbian-type synaptic plasticity, i.e., LTP of excitatory synapses.

Although rMS has been shown (1) to require the activation of voltage-gated sodium channels (VGSC); (2) to be Ca²⁺-dependent, i.e., requiring the activation of both NMDAR and L-type voltage-gated calcium channels (L-VGCC) (Vlachos et al. 2012; Lenz et al. 2015); (3) to recruit intracellular signals such as cAMP-CREB (Hellmann et al. 2012); and (4) to depend on BDNF-TrkB signaling (Fig. 2.1c; Wang et al. 2011; Ma et al. 2013), the precise downstream signaling pathways leading to LTP of excitatory synapses following rTMS, such as phosphorylation and/or accumulation of α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptors (AMPAR) at excitatory postsynapses (Gersner et al. 2011; Vlachos et al. 2012; Lenz et al. 2015), warrant further investigation (Fig. 2.1b, c). Future studies employing (opto-)genetic, pharmacologic, computational, and other experimental approaches will help in delineating similarities and differences between LTP mechanisms recruited by electromagnetic vs. local electric stimulation and may help in defining specific stimulation parameters for the effective induction of structural, functional, and molecular plasticity of distinct synapses in defined cortical networks by rTMS.

In addition to its effects on excitatory synapses, rTMS is expected to also modulate inhibitory neurotransmission. A variety of activity markers and calcium-binding proteins of inhibitory interneurons have been assessed in this context (e.g., Labedi et al. 2014; Mix et al. 2014, 2015; Trippe et al. 2009). For instance, it has been shown that intermittent theta-burst stimulation reduces parvalbumin (PV)-expression in fast-spiking interneurons, while continuous theta-burst stimulation and 1 Hz rTMS predominantly affect calbindin (CB)-expression in cortical areas (Benali et al. 2011; Trippe et al. 2009; Volz et al. 2013). As PV-expressing interneurons primarily control pyramidal cell output, i.e., somatic inhibition, whereas CB-expressing interneurons are considered to regulate pyramidal cell input, i.e., dendritic inhibition (c.f., Fig. 2.1a), these findings imply that distinct rTMS protocols may affect specific aspects of inhibition and hence network activity and function (Funke and Benali 2011; see also Mix et al. 2014, 2015). In line with this notion, TMS-EEG experiments in humans demonstrate that GABAergic inhibitory neurotransmission has a major impact on cortical excitability and connectivity (Premoli et al. 2014). However, direct experimental evidence for the effects of rTMS on inhibition is still missing, since to date no studies are available assessing rTMS-induced structural and functional changes of GABAergic synapses on principle neurons (or excitatory synapses on inhibitory interneurons). Similarly, a comprehensive analysis of rTMS effects on passive and active intrinsic cellular properties, e.g., voltage-gated sodium, potassium, chloride, and calcium currents, is required to better understand the effects of rTMS on excitation and inhibition (E/I) balance in neural circuits and their relevance for plasticity.

Structural changes, such as axonal sprouting and pruning, remodeling of the dendritic tree, dendritic spine turnover, and the formation or loss of excitatory and inhibitory synapses, continuously modify connectivity in the CNS. Since structural plasticity is known to depend on neural activity, it is conceivable that rTMS could assert long-lasting effects on neural networks by inducing the structural remodeling of neural networks. However, so far only one published study exists, which has employed in vitro live-cell microscopy to assess the dynamics of rMS-induced structural plasticity (Vlachos et al. 2012). In this study an increase in the volume of dendritic spines was reported to occur predominantly in small spines, while no effects on spine numbers were observed after high-frequency (10Hz) rMS (Vlachos et al. 2012). These findings are consistent with recent data on spine densities obtained from fixed tissue in vivo (Sykes et al. 2013). Since synapses on small spines are known to constitute weak synapses with low numbers or even no AMPARs (so-called silent synapses containing mainly NMDARs; e.g., Hanse et al. 2013; Kerchner and Nicoll 2008), it is possible that rTMS could modulate network connectivity by recruiting these weak or silent synapses without the need of additional spino- or synaptogenesis. It is tempting to speculate that a simultaneous depolarization of pre- and postsynaptic compartments, i.e., rTMS-induced anterograde (aAP) and backward propagating action potentials (bAP), may recruit silent synapses by increasing the probability of presynaptically released glutamate to activate postsynaptic NMDARs in the absence of AMPARs (see “bAP-aAP theory” in Lenz et al. 2015). Apparently, more work is required to clarify the contribution of “synaptic unsilencing” in rTMS-induced plasticity (see also Rodger et al. 2012) and to determine the effect of single vs. repeated rTMS sessions on structural properties (i.e., axons, dendrites, spines, synapses) of individual neurons, and other cells in the CNS.

Experimental evidence indicates that rTMS can modify gene expression profiles relevant for neural plasticity (Müller et al. 2000; Stock et al. 2012; Okada et al. 2002). However, it remains to be shown how rTMS-induced changes in gene expression affect proteostasis (i.e., the balance between biogenesis, folding, trafficking, and degradation of specific proteins; for review on proteostasis, see, e.g., Mardones et al. 2014), in distinct neural compartments, and how the observed effects influence the ability of neurons to express plasticity (Fig. 2.1). Neuroprotective, e.g., expression of neurotrophic factors such as BDNF (e.g., Gersner et al. 2011), but also toxic effects (Fang et al. 2010; Fujiki and Steward 1997; Okada et al. 2002) of rTMS must be considered in this context as well.

Neuromodulation is another relevant aspect to consider in the context of rTMS-induced plasticity (e.g., Vahabzadeh-Hagh et al. 2012). It is plausible that dopamine, serotonin, acetylcholine, adrenaline, and other neuromodulators may affect the outcome of rTMS. In turn, it is possible that rTMS may act on these neuromodulatory systems to influence plastic properties of neuronal networks beyond the stimulation period.

Indeed, human studies disclose that rTMS-induced LTP- and LTD-like plasticity in the primary motor cortex depends on neuromodulation (Korchounov and Ziemann 2011; Thirugnanasambandam et al. 2011; for review, see Ziemann et al. 2015). Similarly, alterations in rTMS-induced motor cortex plasticity were reported in a rat model of Parkinson’s disease, which correlated with behavioral deficits and neuronal cell loss in the substantia nigra (Hsieh et al. 2015), therefore pointing toward a role of dopamine in rTMS-induced plasticity. On the other hand, several animal studies (in vitro and in vivo) indicate stimulus- and site-specific rTMS effects on the expression of neuromodulators, their receptors, and transporters (e.g., Ben-Shachar et al. 1999; Erhardt et al. 2004; Ikeda et al. 2005; Keck et al. 2002; Kole et al. 1999; Zangen and Hyodo 2002). A better understanding of the role of neuromodulation in rTMS-induced plasticity may thus support the development of novel means in early diagnosis, prognosis, and therapy of brain diseases, e.g., by combining pharmacological neuromodulation with specific rTMS protocols.

As outlined in this book, numerous clinical studies have investigated and confirmed the therapeutic potential of rTMS in various brain diseases (see also Lefaucheur et al. 2014). However, our knowledge of the cellular and molecular mechanisms underlying rTMS-based therapies remains limited. Considering experimental advances in this field of research during the past decade, a biologically driven attempt to improve the use of rTMS in clinical practice has started to emerge, which may also help to better understand the considerable degree of inter- and intraindividual variability of rTMS effects seen in human subjects (see Chap. 1). However, this attempt can only go hand in hand with a better understanding of the role of neural plasticity under pathological conditions (Maggio and Vlachos 2014). For example, it remains unclear through the induction/modulation of which form(s) of plasticity (i.e., Hebbian plasticity, homeostatic plasticity, metaplasticity) rTMS could assert its beneficial effects in the course of a neurological or psychiatric disease (Müller-Dahlhaus and Vlachos 2013). In this context, rTMS effects on non-neuronal cell types need to be considered as well. To successfully transfer knowledge on the cellular and molecular mechanisms of repetitive magnetic stimulation into more effective therapies in neurological and psychiatric patients, it will be also important to study rTMS effects in animal models of brain diseases (e.g., by using genetic mouse and rat models of depression; Barkus 2013). We are confident that these studies will help building evidence-based frameworks for the clinical use of rTMS in the future (for review, see Nitsche et al. 2012).