Target Engagement and Brain State Dependence of Transcranial Magnetic Stimulation: Implications for Clinical Practice

,

https://doi.org/10.1016/j.biopsych.2023.09.011 Transcranial magnetic stimulation (TMS) is a noninvasive technique that delivers magnetic pulses to a targeted brain region, inducing electric currents that can depolarize neurons and stimulate action potentials.Repetitive TMS (rTMS) involves the rhythmic delivery of magnetic pulses with a specific frequency (measured in Hz) to a target region in the brain to induce changes in brain activity.By adjusting the frequency of rTMS, it is possible to either increase or decrease the excitability of targeted brain regions (1)(2)(3).When rTMS is applied repeatedly across multiple sessions, clinically relevant, longer-lasting neuroplastic brain changes can be triggered that have the potential to result in the normalization of aberrant network activity.In the last 3 decades, TMS has worked its way into the established range of treatment options for depression.Currently, the gold-standard TMS approach to treat depression is daily high-frequency rTMS, applied to the left dorsolateral prefrontal cortex (DLPFC) for several weeks (4).This depression TMS protocol is assumed to stimulate the left DLPFC and, importantly, also affect connected limbic system regions, among which the connection between the DLPFC and the subgenual anterior cingulate cortex (ACC) seem most relevant in the treatment success of TMS (5)(6)(7)(8).Response rates and remission rates vary between clinical trials, but conservatively estimated, approximately 30% to 40% of patients with treatment-resistant depression respond to this TMS treatment (9).
Although TMS has demonstrated clinical benefits on a group level, a significant proportion of patients still do not respond to the treatment.The main problem identified is the high variability in how different patients respond to the TMS treatment.Recent findings suggest that this variability is not related to the heterogeneity of depression itself, but instead is caused by interindividual differences in whether or not the rTMS intervention effectively modulates the depression network in the brain (10)(11)(12).This is also referred to as target engagement.
Importantly, successful target engagement depends on the individual's brain state at the moment of stimulation (13)(14)(15).This currently poses one of the greatest challenges for establishing and optimizing clinical TMS treatments.In this review, we systematically describe how target engagement in TMS interventions depends on 1) the cognitive brain state, 2) the oscillatory brain state, and even 3) the recent history of brain state of patients (Figure 1).It is surprising to note that despite these fundamental principles, clinical TMS protocols continue to largely ignore this critical factor, which could potentially explain the limited efficacy and variability of TMS treatments.It is imperative to develop stimulation protocols that can reliably induce the desired brain modulation effects on a single-subject level, by incorporating state-dependent brain stimulation protocols in clinical practice.

COGNITIVE STATE DEPENDENCE
When applying TMS to a patient during a given treatment session, the magnetic pulses (external stimulation) actively interact with the current internal brain state.With cognitive state dependence, we here explicitly also include emotional and perceptual states.The brain network effect of TMS is thus always the result of an interaction between the internal cognitive, emotional, and/or perceptual brain state and the external brain stimulation parameters.Sleep, rest, wakefulness, attention, learning, memory, and other cognitive functions that patients may engage in during brain stimulation have all been shown to qualitatively and quantitatively affect the stimulation-related changes observed in the brain (13,16,17).
One of the earliest and simplest documentations of brain state dependency of TMS is the modulation of TMS-induced motor evoked potentials (MEPs), recorded during stimulation of primary motor cortex (M1), during active muscle contraction as compared with rest (18).In an active motor state, MEPs induced with the same TMS intensity are significantly enlarged because of an interaction between external stimulation and internal brain state.This interaction between external stimulation and internal brain state has been observed in various domains, including the visual domain.Silvanto et al. (19) even proposed to increase the functional specificity of TMS by taking advantage of these state dependencies.By using neural adaptation, in which a visual stimulus is presented prior to TMS, the authors were able to systematically change the perceptual color quality of TMS-induced phosphenes.The idea is as simple as it is intriguing: A visual stimulus is presented to tune specific subsets of cells, thereby inducing an activation imbalance between differing neuronal populations, and in this perceptually or cognitively induced brain state, TMS will produce very specific behavioral or neural changes.This paradigm has been used to demonstrate specialized areas for grip and goal integration (20), affective body movements (21), quantity and social concepts (22), belief updating (23), and others.
Zokaei et al. (24) showed that motion items held in working memory are differently affected by a disruptive TMS pulse over the motion area, based on the memory state that the items were in at the time of stimulation.Items in a privileged state caused either by cueing or by recency were more susceptible to TMS disruption, whereas nonprivileged items were spared.Ezzyat et al. (25) showed that the encoding state of an item, either high or low as decoded from intracranial electroencephalography (EEG) data, dictated the effects of TMS on the recall accuracy for that specific item.It has also been shown that the presence or absence of distracting stimuli can modulate the global network effects of TMS applied to the DLPFC during a delayed-retention working memory task (26).This internal brain state dependence is also relevant for clinical applications of TMS.Borgomaneri et al. (27), for example, showed that rTMS given during the reactivation of a fearrelated memory successfully abolished physiological fear responses.It may therefore not be surprising that many TMS monotherapy studies for obsessive-compulsive disorder, posttraumatic stress disorder, and substance use disorders apply TMS concurrent with cue-triggered symptom provocation in order to capitalize on the internal brain state dependence of TMS (27)(28)(29)(30).This emotional state dependence of rTMS in the context of obsessive-compulsive disorder and posttraumatic stress disorder, however, seems much less straightforward in the context of depression.Here, in contrast, it has even been found that depressive patients who engaged  It is crucial to understand that these internal brain statedependent TMS effects can be observed not only on a behavioral or psychological level, but also on a neural network level.This means that the same TMS protocol in the same patient can induce measurably different network effects in the brain, depending on the cognitive, emotional, and/or perceptual state at the time of stimulation.The most direct and compelling evidence for this notion stems from studies applying TMS inside a functional magnetic resonance imaging (fMRI) scanner, stimulating the brain with TMS in different cognitive brain states, while simultaneously imaging TMSinduced neural network changes and TMS-induced changes in cognition and behavior.Using such concurrent TMS and fMRI technology, our group demonstrated how the local and remote TMS network effects in the brain are modulated by the cognitive brain state at the moment of TMS (16).Similarly, it has been shown that the long-range network effects of TMS crucially depend on the task that participants perform during stimulation.In a feature-based attention task with two simultaneously presented stimulus categories, stimulation to the frontal eye field resulted in TMS-induced activity changes in the fusiform face area when participants attended the face but in the motion area when attending the array of moving dots (36).In conclusion, TMS can affect local and long-range network nodes, causing change in behavior and clinical symptoms.However, the brain's cognitive, emotional, and/or perceptual state during stimulation can significantly influence its network effects, including interactions with stimulation parameters and changes to neural and behavioral outcomes.This has been underappreciated in clinical settings, in which psychological, cognitive, emotional, and/or perceptual state could be altered in various ways in a relatively fast, cheap, and efficient manner to increase clinical efficacy and reduce variability.

OSCILLATORY STATE DEPENDENCE
In the previous section, we described how a specific perceptual and/or cognitive brain state determined the TMS-induced neural network effects in the brain.But what is the exact neurophysiological substrate of a certain cognitive state?Or even more importantly, can this neurophysiological brain state that constitutes a specific cognitive state also fluctuate from moment to moment within the same cognitive state?The simple answer to both of these questions is yes.
Human cognition arises from information exchange within and between functionally connected brain networks.One way of communication is the continuous coupling and uncoupling of brain areas within functional networks, resulting in ongoing rhythmic fluctuations in brain states, known as neural oscillations.These oscillatory brain states can change rapidly to promote or prevent the spread of information (propagation of signals) along specific routes (networks).Neural oscillations thus reflect a cyclic waxing and waning of excitation in brain regions, so that a brain region will be temporarily more susceptible to inputs during a subset of excitable phases of the oscillatory cycle (37,38).Because different neural assemblies show fluctuations over time in their excitability, communication among those assemblies is promoted when excitable phases in the oscillatory cycle are synchronized, or phase locked, at a proper phase difference (to compensate for transmission delays) (39)(40)(41)(42)(43)(44).In this regard, neural oscillatory rhythms contribute to the communication between neuronal populations that are simultaneously engaged in cognitive processes (45,46).This fundamental idea has driven a large amount of neuroscientific research and has received encouraging support particularly from invasive animal recordings (37,38,(47)(48)(49)(50)(51) and noninvasive human studies (47,49,52,53) (Figure 2).
Our recent research has revealed that the commonly used rTMS protocols for depression are associated with limited intrasubject reliability and significant intersubject variability in TMS-induced network effects in the brain (54).To overcome this limitation, we propose the use of personalized and precision TMS interventions.Recent research has shown that TMSinduced effects on cortical excitability are directly related to the phase of underlying intrinsic neuronal oscillations at the time and site of TMS pulse application, indicating that ignoring the role of neural temporal communication mechanisms, such as brain oscillations, may contribute to the variability in TMS efficacy.We have revealed this by simultaneous TMS-EEG-fMRI, which integrates noninvasive brain stimulation (TMS), online EEG data collection, and online monitoring of fMRI activity in brain networks.It allows measuring the current oscillatory brain state on a local (power, phase) and global (coherence, longrange synchronization) scale, while introducing a temporally and spatially precise signal to a predetermined network node using TMS (55).This technology has revealed how TMS signals are propagated through the brain, depending on the momentary oscillatory brain state and in the context of a certain cognitive task.
This technology revealed that trial-by-trial, pre-TMS, EEG alpha and low-beta power fluctuations influenced signal propagation and network communication in the corticosubcortical motor system (55).The crucial finding in the experiment was that the inserted TMS signal spread effectively throughout this motor network, including signal propagations from cortical to subcortical (thalamic) brain regions, only when alpha power measured over the motor system was low.These findings provide direct, causal evidence for the hypothesized function of alpha oscillations in gating information flow through a pulsed inhibition mechanism, orchestrated by activated thalamocortical circuits.This spread of activity was blocked when alpha power was high.In contrast to alpha activity, strong pre-TMS beta power did not impede TMS signal propagation, but instead tended to facilitate signal propagation in motor circuits.Such facilitatory effects corroborate and extend recent evidence for the key role of beta activity in coordinating communication in the human motor system (56,57).These data reveal how TMS as a probe can elicit (sub) cortical network responses based on EEG-indexed oscillatory brain state.Using a similar setup, Pantazatos et al. (58) recently showed that the phase of the endogenous alpha rhythm also modulates TMS-evoked DLPFC-to-subgenual-ACC fMRIderived effective connectivity in healthy volunteers.Interestingly, these authors also reported, for depressed patients specifically, prefrontal EEG alpha phase-dependent effects on TMS-evoked fMRI blood oxygen level-dependent activation of the rostral ACC (58).This opens the door to studying dynamic brain circuits and their dysfunction in neuropsychiatric disorders and developing personalized state-dependent TMS protocols for clinical use.

HISTORY STATE DEPENDENCE
We have now established the complexity of the network effects of TMS in the brain.These effects not only are state dependent, but also are influenced by spontaneous momentto-moment fluctuations in oscillatory brain activity.However, in addition, the history of stimulation and/or the previous neural ).This can be observed through the measured differences in the activity of those connected regions.For example, alpha oscillations have a broadly inhibitory influence on network communication.If a transcranial magnetic stimulation (TMS) pulse arrives during a moment of low endogenous alpha amplitude, then the flow of signaling from the target area dlPFC to other connected nodes might be more efficient and the resulting signal might be stronger.Alternatively, when alpha amplitude is high, it inhibits the spread of TMS-related network activity, leading to a lower measurable change in activity in its connected nodes.(Bottom) (A) On a more global scale, the brain can be in different cognitive states, for example a task-related state (left) or a resting state (right).(B) These global cognitive states can also either facilitate or inhibit the spread of TMS-related activity in the functional network connected to the stimulation target.(C) Additionally, they may interplay in a complex and nonlinear manner with the ongoing endogenous oscillatory states, leading to differences in the measured signal at connected nodes.This is important, as the global cognitive state of a patient during stimulation in the clinic might inadvertently impede or facilitate treatment with noninvasive brain stimulation without the knowledge of the clinician.Data shown in boxes in top and bottom panel (C) are not real and are only meant to represent a potential scenario involving dlPFC stimulation in the context of oscillatory and cognitive state dependence.BOLD, blood oxygen level-dependent; fMRI, functional magnetic resonance imaging.activity prior to TMS can also affect the local and remote network effects and even reverse the intended changes in excitability.
To explore this, researchers have turned to the Bienenstock-Cooper-Munro model of metaplasticity, which describes changes in the magnitude or direction of synaptic plasticity due to prior neural activity (59).Unlike Hebbian-like synaptic plasticity, metaplasticity regulates changes in synaptic and neuronal states independently of overall network excitability (60,61).Metaplastic changes occur based on the previous integrated activity of the postsynaptic neuron, which results in a dynamic threshold for activity-dependent synaptic plasticity (Figure 3).This highlights the complexity of TMSinduced effects on the brain and the importance of considering both the current state and history of neural activity.Metaplasticity has both nonhomeostatic and homeostatic effects.Nonhomeostatic or additive metaplasticity promotes synaptic strengthening through repeated excitatory stimulation.Homeostatic metaplasticity ensures neuronal and network stability by preventing excessive long-term potentiation/long-term depression (LTP/LTD) expression.This complements synaptic plasticity by stabilizing neuronal activity within a physiological range, therefore adjusting the ability of a neuron to induce LTP/LTD (Figure 3).The sliding threshold theory for bidirectional synaptic plasticity explains how the threshold for LTP/LTD induction dynamically adjusts based on the integrated level of prior postsynaptic activity (62).Recent developments in NIBS have enabled the study of homeostatic plasticity in the intact human brain, utilizing a priming test design (Figure 4).
Plasticity changes have primarily been assessed in fastconducting corticospinal projections by targeting the M1 with NIBS such as TMS, transcranial direct current stimulation (tDCS), or paired associative stimulation, thereby quantifying the mean amplitude of MEPs.When using a facilitatory anodal tDCS priming protocol prior to a 15-minute 1 Hz low-intensity rTMS test protocol, an LTD-like effect was observed.An inhibitory cathodal tDCS priming protocol, however, resulted in an LTP-like effect, reflected in the increased corticomotor excitability while the rTMS frequency remained the same.Replacing the priming protocol with sham stimulation resulted in no altered effects on the test protocol (63).From this and other studies, a pattern of homeostatic reversal of excitability emerged, showing that a priming protocol may enhance the effects of a test protocol if the opposite effect on excitability is induced (61)(62)(63)(64).Similarly, if the priming protocol has the same effect on excitability as the test protocol, subsequent effects of excitability are weakened or reversed.More so, this reversal was also demonstrated when the same NIBS protocol was applied for priming and test stimulation (65).For instance, paring the same priming and test theta burst stimulation (TBS) protocols (intermittent TBS/intermittent TBS, continuous TBS/continuous TBS) resulted in a weakening of the nonprimed TBS effect.Pairing different TBS protocols for priming and test stimulation enhanced the effect of the test TBS protocol (66,67).Doubling the duration of the stimulation, as it has  been shown for instance with TBS (68,69), also followed this polarity-reversing pattern.However, other findings contradict this distinction in inducing a polarity flip by simply changing the number of pulses in comparison with sham stimulation (70).It is not clear what precisely drives this polarity reversal or when exactly such a flip occurs in a given subject.TBS-induced aftereffects have repeatedly shown a significant portion of variability (70), which may be attributed to the methods of measuring the modulatory effects of TBS, the inter-and intrasubject variability commonly observed in response to TBS (71)(72)(73)(74), or intrinsic biological factors (75).When excluding the breaks between 2 blocks of a standard 5 Hz rTMS protocol, this continuous protocol led to a change in the direction of the aftereffect, from facilitation to inhibition of corticospinal excitability (76).Interestingly, the same NIBS protocol can either induce LTP-or LTD-like effects in the same individual, dependent on whether neuronal activity was preconditioned with a priming protocol.This highlights the importance of history-dependent activity, but more so, it puts the rigid distinctions of facilitatory and inhibitory NIBS protocols into question.
A recent review and meta-analysis by Wittkopf et al. (71) compared the duration and timing effects of priming-test NIBS protocols on corticospinal excitability.The timing between these protocols represents a crucial component in the direction and magnitude of the aftereffects.This critical time window is thought to be fundamental in inducing homeostatic responses.Animal research has highlighted the importance of sufficiently spaced pauses between excitatory sessions of rTMS/TBS for the additive (LTP) plastic effects to occur, emphasizing once more the timing-dependent nature of metaplasticity effects.A comparison of MEP amplitudes pre-and poststimulation indicated that an interval of 10 minutes or less between priming and test stimulation may be able to induce a homeostatic response (71).However, no conclusive evidence was found for homeostatic plasticity induction using 2 NIBS protocols with such intervals, mainly due to the heterogeneity and poor quality of primary studies available.Still, within the sliding threshold theory of homeostatic plasticity (59), this critical window may indicate the importance of the temporal relationship between priming and test stimulation in shifting the threshold in a homeostatic manner.
The clinical implications of TMS's history state dependence are clear: there is not only potential for priming or preconditioning rTMS protocols, but also a risk.Homeostatic metaplasticity mechanisms stabilize network activity but can interfere with the plasticity effects of rTMS.In clinical applications, the goal is to promote additive, increased plasticity effects, not stabilization.To address this, Thomson and Sack (77) published guidelines for designing optimal accelerated rTMS protocols that promote beneficial metaplasticity while avoiding counteracting effects.

TARGET ENGAGEMENT
What do the different forms of state dependence described in previous sections all have in common?They all determine which neural network effects are induced by TMS, and as such they all directly affect what has been referred to as target engagement in the field of (clinical) TMS applications.
One way to directly assess state-dependent target engagement is to stimulate the brain with TMS while at the same time measure and visualize the whole-brain network responses induced by TMS.Such concurrent TMS and brain imaging approaches have been described in detail elsewhere (7,10,78,79).Concurrent TMS-fMRI, TMS-EEG, or now even TMS-EEG-fMRI combinations are available to unravel which exact neural network effects are induced by TMS, and how those network effects may be modulated and optimized when applying TMS at a particular brain state (be it a cognitive or perceptual state, an oscillatory state, or a preconditioned or primed state).
An obvious limitation of such multimodal stimulation approaches is the reduced practicability and accessibility for clinical practice.An important practical question is how knowledge of functional networks in depression may lead to a well-informed selection of the TMS target that allows to somehow read out successful target engagement in the absence of neuroimaging.The depression network shares overlapping nodes with the frontal-vagal network, including the DLPFC, subgenual ACC, and vagus nerve.The vagal network is responsible for parasympathetic regulation and, when activated, leads to heart rate (HR) deceleration, increases in HR variability, and decreases in blood pressure [see (12) for a detailed review].
A recently developed target engagement method, called neurocardiac-guided TMS, utilizes this principle, in which successful engagement of the frontal-vagal network is at least indirectly supported by target-specific HR deceleration during TMS stimulation (12,78,79).First studies demonstrated such specific HR decelerations on the group level confined to 2 locations corresponding to classical TMS depression targets (Beam F3 and 5 cm anterior to M1), whereas TMS applied to other locations resulted in HR accelerations.Further research has refined this method into a potential approach to evaluate target engagement based on TMS-induced heart-brain coupling (80).Here, various TMS stimulation protocols can be used that differently modulate the HR depending on their stimulation pattern.By assessing TMS-induced heart-brain coupling in the electrocardiogram, in the specific corresponding TMS frequency band, it could potentially be determined whether you have successfully engaged the frontalvagal network.However, it remains to be directly validated, using neuroimaging, whether TMS targeting a location on the scalp with measurable HR deceleration indeed affects the frontal-vagal network as now assumed.In addition, prospective studies are needed to evaluate the added value for clinical efficacy, independent from whether HR deceleration is a consequence of successful modulation of the frontal-vagal network or not.

CONCLUSIONS AND IMPLICATIONS FOR CLINICAL PRACTICE
It has been demonstrated that TMS has state-dependent effects in the human brain, whereby both the immediate and long-term effects of TMS, including plasticity, are contingent upon the cognitive, emotional, and/or perceptual state of the individual at the time of stimulation, as well as on the spontaneously fluctuating oscillatory brain state.Additionally, the efficacy of TMS therapy can be further influenced by the previous history of neural activity prior to stimulation, potentially leading to a reversal in polarity of the intended effects.This means nothing less than that the target engagement of our TMS intervention cannot be controlled reliably without systematically considering and co-controlling the different brain state dependencies of our TMS therapy.Surprisingly, however, clinical TMS protocols are largely ignoring this fundamental principle, which may explain the large variability and often still limited efficacy of TMS treatments.
We argue here that it is time to change this practice and replace standard clinical TMS by personalized statedependent TMS protocols.There are several approaches available to this end, including the combination of TMS with a simultaneously applied cognitive intervention (e.g., cognitive engagement) (24)(25)(26)54).Li et al. (81) for example, combined TMS over the DLPFC for the treatment of depression, with a computerized rostral ACC-engaging cognitive task.They revealed that this manipulation of patients' cognitive brain state augmented the antidepressant effects to rTMS treatment with more reduction in total depression scores, more responders, and more remitters as compared with standard rTMS (81).Donse et al. (82) demonstrated high response and remission rates when simultaneously combining rTMS over right DLPFC with psychotherapy.
In addition to cognitive state-dependent interventions, EEG-informed or even EEG-triggered closed-loop TMS protocols are now available, capable of stimulating at particular amplitudes or phases of simultaneously recorded and analyzed EEG-indexed oscillatory states (83)(84)(85).Zrenner et al. (86) showed in 2020 that oscillatory state-dependent rTMS of the left DLPFC is feasible and capable of inducing fast neuromodulatory effects in patients with antidepressant-resistant depression.Our own group has also developed a userfriendly, hardware-software system for oscillatory statedependent TMS neuromodulation (84), which allows the user to bring oscillatory state under experimental control using transcranial alternating current stimulation, and to then apply the TMS pulse at predetermined oscillatory parameters (phase or power) of the endogenous brain oscillations (84,85).
Finally, priming or preconditioning TMS with a preceding tDCS (or rTMS) session could be a clinically powerful means to capitalize on the described processes of homeostatic plasticity, leading to more stable and consistent rTMS-induced neuroplastic changes within and between patients.In a meta-analysis, Brunoni et al. (87) indeed already concluded 5 years ago that the estimated relative ranking of effective TMS depression treatments suggests that priming rTMS may be among the most efficacious of all current rTMS strategies.
Still, none of these state-dependent TMS approaches listed above are widely accepted and/or adopted in clinical practice today.In fact, many clinicians focus solely on the best TMS protocol or the best parameters for stimulation, assuming that the state of the brain of their patients during TMS plays no role for the efficacy of the treatment.Clinicians even often introduce various ways of making patients most comfortable during the TMS sessions, allowing them to listen to music, watch a movie, or even fall asleep during TMS.It is vital to be aware that all of these different states could also affect the efficacy of TMS therapy.We cannot yet recommend which exact state is desired or should be avoided for a given TMS therapy (e.g., depression) in a given patient, but this will hopefully change in the next 5 years.

ACKNOWLEDGMENTS AND DISCLOSURES
ED is the owner of Neurowave.ATS is chief scientific advisor of Plato-Science; founder and CEO of Neurowear Medical; and Director of the International Clinical TMS Certification Course (www.tmscourse.eu),receiving equipment support from MagVenture, MagStim, and Deymed.MA is the unpaid chairman of the Brainclinics Foundation; holds equity/stock in neurocare; serves as consultant to neurocare, Roche, and Numinous; and is named an inventor on neurocare-owned patent and intellectual property related to neurocardiac-guided TMS but receives no royalties.The Brainclinics Foundation received research and consultancy support from neurocare and equipment support from neuroconn and Deymed.All other authors report no biomedical financial interests or potential conflicts of interest.

Figure 1 .
Figure 1.Brain state dependence of noninvasive brain stimulation.Spontaneous, moment-tomoment fluctuations in oscillatory parameters like power or phase (upper left), endogenous cognitive and perceptual state (upper right), and changes in the magnitude or direction of synaptic plasticity from to prior transcranial electrical stimulation (lower left) or transcranial magnetic stimulation (lower right) can all enhance or disrupt the spread of both local and remote network effects of noninvasive brain stimulation (middle).

Figure 2 .
Figure 2. Infographic explaining oscillatory and cognitive state dependence.Mechanisms of oscillatory (top) and cognitive (bottom) state dependence, with the example of stimulating the dorsolateral prefrontal cortex (dlPFC).(Top) (A) Endogenous oscillations at the moment of stimulation.(B) Oscillations can differ in their content, for example in their frequency, amplitude, or phase.(C) The differences in amplitude may lead to a facilitation or inhibition of the spread of information from the targeted area (in this case the dlPFC) to other connected nodes (e.g., subgenual anterior cingulate cortex [sgACC], ventromedial prefrontal cortex [vmPFC]).This can be observed through the measured differences in the activity of those connected regions.For example, alpha oscillations have a broadly inhibitory influence on network communication.If a transcranial magnetic stimulation (TMS) pulse arrives during a moment of low endogenous alpha amplitude, then the flow of signaling from the target area dlPFC to other connected nodes might be more efficient and the resulting signal might be stronger.Alternatively, when alpha amplitude is high, it inhibits the spread of TMS-related network activity, leading to a lower measurable change in activity in its connected nodes.(Bottom) (A) On a more global scale, the brain can be in different cognitive states, for example a task-related state (left) or a resting state (right).(B) These global cognitive states can also either facilitate or inhibit the spread of TMS-related activity in the functional network connected to the stimulation target.(C) Additionally, they may interplay in a complex and nonlinear manner with the ongoing endogenous oscillatory states, leading to differences in the measured signal at connected nodes.This is important, as the global cognitive state of a patient during stimulation in the clinic might inadvertently impede or facilitate treatment with noninvasive brain stimulation without the knowledge of the clinician.Data shown in boxes in top and bottom panel (C) are not real and are only meant to represent a potential scenario involving dlPFC stimulation in the context of oscillatory and cognitive state dependence.BOLD, blood oxygen level-dependent; fMRI, functional magnetic resonance imaging.

Figure 3 .
Figure 3. Simplified concept of metaplasticity according to the Bienenstock-Cooper-Munro theory.The modification threshold (M 0 ) indicates the crossover from long-term depression (LTD) to long-term potentiation (LTP) and varies as a function of postsynaptic activity.Applying an LTD-like prime decreases the change in synaptic strength and shifts the modification threshold (M 0 0 ) toward an LTD-like response.When using an LTP-like prime, the threshold (M 0 0 0 ) is moved toward an LTP-like response.Adapted with permission from Karabanov et al. (88).

Figure 4 .
Figure 4. Basic concept of the priming-testdesign.Metaplasticity mechanisms (the modulation of Hebbian-like plasticity) prior to neural activity (priming) can occur in absence of excitability changes in the stimulated neuronal group.Mechanisms of metaplasticity and especially homeostatic plasticity effects have been primarily investigated through the use of 2 subsequent blocks of transcranial magnetic stimulation or transcranial electrical stimulation, separated by a variable interval of no stimulation.The priming (left), or preconditioning protocol, is aimed at triggering a homeostatic response, followed by a test, or stimulation protocol (right), to capture the homeostatic response to the stimulation.