Device-Based Modulation of Neurocircuits as a Therapeutic for Psychiatric Disorders

Abstract

Device-based neuromodulation of brain circuits is emerging as a promising new approach in the study and treatment of psychiatric disorders. This work presents recent advances in the development of tools for identifying neurocircuits as therapeutic targets and in tools for modulating neurocircuits. We review clinical evidence for the therapeutic efficacy of circuit modulation with a range of brain stimulation approaches, including subthreshold, subconvulsive, convulsive, and neurosurgical techniques. We further discuss strategies for enhancing the precision and efficacy of neuromodulatory techniques. Finally, we survey cutting-edge research in therapeutic circuit modulation using novel paradigms and next-generation devices.

1. INTRODUCTION

The use of devices to modulate the functioning of neural circuits in the brain is emerging as a promising new approach in the study and treatment of psychiatric disorders. Whether they are surgically implanted in the brain or noninvasively applied to the scalp, devices for modulating circuit function have advanced from the preclinical and first-in-human testing stages to becoming Food and Drug Administration (FDA)-approved treatments such as transcranial magnetic stimulation (TMS) and vagus nerve stimulation (VNS).

The success of device-based approaches depends heavily on advances in neuroscientific understanding of how to deploy these devices to effectively target the circuitry implicated in the disorder and how to personalize the intervention by selecting those most likely to respond. Advances in the fields of neuroimaging, neurophysiology, and computational neuroscience have yielded new understanding of psychiatric disorders as arising from dysfunction in distributed networks in the brain (1). This has ushered in what some have called the era of circuit psychiatry (2).

While the traditional method for circuit discovery took a categorical approach to identifying commonalities within patient groups that distinguish them from healthy controls, more modern methods take a dimensional approach to discover and identify data-driven structure across and within diagnostic groups. The latter approach, based on the National Institute of Mental Health (NIMH) Research Domain Criteria (RDoC) (3) paradigm for target discovery, seeks circuits underlying domains of function (such as reward sensitivity), which may be expected to be better aligned with the function of identifiable circuits than traditional DSM diagnoses which are based on symptom lists that span multiple domains of function. Identifying differences in circuit expression within diagnostic categories, an approach called biotyping (4), represents a promising means of personalizing treatment by selecting and/or modifying the treatment to be optimally effective for the individual patient’s needs.

The experimental therapeutics framework is well adapted to device-based circuit modulation. In that framework, the device is used to first engage the targeted circuit, so-called target engagement. Then, when the circuit is engaged, one can evaluate whether the predicted change in clinical presentation occurs. The experimental therapeutics approach is a powerful way to evaluate the therapeutic potential of modulating specific circuits and is a rapid way of disqualifying ineffective therapies due to their failure to engage the circuit. While the experimental therapeutics approach is well known in the development of drugs, it is a relatively recent advance in other therapeutic modalities.

The availability of devices that can modulate the functioning of circuits has transformed our ability to both identify and test their causal role in psychiatric disorders and, ultimately, in disorders that are unresponsive to traditional pharmacological approaches. In this review, we focus on circuits at the meso- and macroscale, given that it is the level of resolution that currently available devices can engage in the human (see Figure 1). Next-generation devices, in development through initiatives such as the National Institutes of Health (NIH) BRAIN Initiative (5), are anticipated to enable circuit manipulation at the microscale in the future.

Figure 1.

Research Domain Criteria (RDoC) units of analysis, including the genetic, molecular, cellular, circuit, physiological, behavioral, subjective, and clinical levels. While pharmacology engages at the molecular level, and psychosocial interventions engage at the behavioral level, device-based neuromodulation tools currently engage at the level of circuits. Figure adapted with permission from Reference 159.

2. TOOLS FOR IDENTIFYING NEUROCIRCUITS AS THERAPEUTIC TARGETS

2.1. Neuroimaging

One of the key advantages of device-based interventions over pharmacological interventions is that devices offer the ability to precisely target specific sites within the brain. The use of neuroimaging is rapidly becoming a key component in efforts to identify neural circuits as therapeutic targets for psychiatric disorders (6). However, the concept of a target is itself a moving target. The field has shifted from thinking of direct targeting of sites toward transsynaptic targeting of circuits and temporal tuning of neural dynamics (Figure 2).

Figure 2.

Approaches to targeting device-based intervention. Targets for intervention with neuromodulation can be sites (identified by structural or functional neuroimaging), circuits (identified by structural or functional connectivity), or neural dynamics (identified via neurophysiology measures of neural oscillations). If sites are superficial, they may be reached via direct targeting employing stereotaxy; however, if they are deeper or distributed, transsynaptic targeting may be needed, informed by connectivity measures. If targets are dynamic, the temporal aspects of the neuromodulation may need to be tuned accordingly to engage them.

The conventional approach has been to focus on direct targeting of sites, in other words, placing the stimulation electrode or coil over the structural or functional brain region implicated in the disorder. An example is the FDA approval of TMS delivered over the left dorsolateral prefrontal cortex (dlPFC) for the treatment of depression (7). Decades of neuroimaging and physiology work have identified dysfunction of the left dlPFC in depression (8–12). The dlPFC, which is part of the frontoparietal attention network (FPN), plays a key role in the regulation of emotions (13). Direct targeting of the dlPFC, while effective, typically achieves only modest effect sizes.

More recent work has demonstrated that the transsynaptic connectivity of the targeted brain region may be even more impactful in driving clinical outcomes than direct targeting. An example is the failure of deep brain stimulation (DBS) targeting the subcallosal cingulate cortex, which, like the dlPFC, had substantial support for involvement in depression (14–18). Based on these findings, Mayberg and colleagues (18) targeted this region with DBS in severely depressed patients with some success, though the pivotal trial failed (19). Subsequent work suggested that connectivity of the rostral cingulate target is related to outcome and that connectome-informed targeting could be more effective (20).

Finally, emerging evidence implicating abnormalities in neural oscillations is providing clues as to how devices may target the dynamics of information flow within distributed circuits. So-called temporal tuning can use information from high-temporal-resolution physiology tools such as electroencephalography (EEG) and magnetoencephalography and the somewhat lower temporal resolution but higher spatial resolution of functional magnetic resonance imaging (fMRI) to identify not only where but when to stimulate. Advanced analytical techniques like machine learning and global brain connectivity are advancing our understanding of the neural dynamic mechanisms mediating psychopathology. For instance, recent researchers have used machine learning to link patterns of intracranial EEG coherence in the amygdala and hippocampus to negative mood states (21, 22). Similarly, another group used resting-state fMRI to quantify the global connectivity within the FPN and showed that decreases in FPN global connectivity were linked with depression symptoms (8, 23).

2.2. From Precision Psychiatry to Precision Neurostimulation

Precision medicine seeks to match the right patient to the right treatment, informed by individual characteristics such as genetics and physiology. Bringing the promise of precision medicine to psychiatry has been challenging given the heterogeneity within and comorbidity across psychiatric disorders that are classically defined via symptom-based diagnosis. However, the RDoC platform employs imaging and physiological measurements of circuit function and applies a data-driven approach to identify biotypes that can guide treatment selection and optimization (Figure 3). Understanding underlying neurocircuit dysfunction offers the opportunity to cluster disorders at the brain level rather than relying exclusively on symptom-based categorical diagnosis (24).

Figure 3.

(a) Precision psychiatry and (b) precision neurostimulation. Symptom-based diagnostic groups have high degrees of heterogeneity with respect to phenotypic expression and neurobiological measures of circuit function. They also have significant diagnostic comorbidities. The concept of precision psychiatry re-sorts patients into data-driven biotypes, informed by biological measures such as neuroimaging, neurophysiology, and neurocognitive measures. These data-driven clusters are more homogenous with respect to underlying neurobiology and therefore may be more precisely targeted by interventions that selectively engage that neurobiology. Extending the concept of precision psychiatry to precision neurostimulation, we can adapt the stimulation to individual patient needs by achieving specificity in the key elements of dosing (spatial, temporal, and contextual). Spatial precision can be achieved through individual image guidance, electric field modeling of the current induced in the brain, and focal coils/electrodes. Temporal precision can be achieved by optimizing the pulse waveform, frequency, and train characteristics and by coupling it to endogenous oscillations via closed-loop techniques. Contextual precision can be achieved by controlling brain state during stimulation via online cognitive task performance, synchronizing to endogenous neural oscillations, or combination therapies such as simultaneous cognitive behavioral therapy and pharmacotherapy.

A data-driven approach using fMRI to measure connectivity has elucidated four clusters of depressed patients that exhibit distinct patterns of abnormal functional connectivity (25, 26). These clusters correlate with distinct clinical symptom combinations. Clustering according to functional connectivity yielded stable clinical groupings, whereas clustering according to symptoms yielded low overall longitudinal stability within the clusters. Dunlop & Mayberg (27) have suggested that there are two main types of biomarkers that predict treatment response in depressed patients. Pre-treatment brain states can be used to select for treatment approaches that would be the most likely to be beneficial for individual patients, including responsiveness to TMS. For example, Weigand et al. (28) found that anticorrelation in resting-state connectivity between dlPFC stimulation and the subgenual cingulate accurately predicted TMS treatment response. In addition, the cumulative brain engagement index (cBEI) is an electrophysiological marker that can be used to predict treatment response to repetitive transcranial magnetic stimulation (rTMS) (29) and antidepressant medications (30). Both studies found that a difference in cBEI correlated with a difference in response to treatment. A second kind of biomarker allows for the identification of patients who are unlikely to respond to the standard antidepressant treatments, indicating that these approaches should be skipped and interventions typically used for treatment-resistant patients should be employed. However, a recent meta-analysis found that electroencephalographic biomarkers are not yet completely reliable for the use of predicting treatment response (31).

Moving from precision psychiatry to precision neurostimulation requires gaining not only specificity of the individual patient’s circuitry as discussed above but also specificity in how the neurostimulation device is dosed across the three dimensions of space, time, and context (Figure 3). Spatial targeting can be informed using frameless stereotaxy systems that can navigate the electrode or the stimulating coil to the individual’s brain circuit as derived by neuroimaging. Electric field modeling can inform the intensity and focality of the stimulation being administered to each patient, and multifocal electrodes or coils can be designed to best match the patients’ spatial targeting needs. Temporal precision can be achieved by optimizing the waveform of the stimulation (32), employing specific temporal patterns (either mono-frequencies or burst patterns), and by employing closed-loop systems that stimulate during specific phases of neural oscillations. The response of the brain depends on what the brain is doing at the time of stimulation, i.e., the context or brain state. Context can be controlled by having the subject engaged in a neurocognitive task, by administering concomitant medications, or by delivering simultaneous or sequential cognitive behavioral therapy (CBT).

3. TOOLS FOR MODULATING NEUROCIRCUITS

3.1. Pharmacology

Pharmacological treatments can provide relief for patients with psychiatric disorders by targeting the brain at the molecular to cellular levels. For example, current first-line antidepressant treatment options include pharmacological agents primarily directed at monoaminergic targets such as monoamine oxidase inhibitors, selective serotonin reuptake inhibitors, serotonin-norepinephrine reuptake inhibitors, and tricyclic antidepressants (33, 34). Yet, many patients remain pharmacoresistant, as documented in the Sequenced Treatment Alternatives to Relieve Depression (STAR*D) study (35, 36). Unfortunately, discovery of novel compounds remains an expensive, lengthy, and high-risk process with a low rate of success. To reduce the cost of early-phase psychiatric drug development, NIMH initiated a series of Fast-Fail Trials (FAST) that aim to quickly identify candidate pharmacological agents by evaluating their ability to engage a target in the brain and measurably alter a biomarker of brain function. The FAST Mood and Anxiety Spectrum Disorders (FAST-MAS) program recently successfully implemented this approach. This first proof-of-mechanism trial aimed to assess the potential of a selective kappa-opioid receptor antagonist for the treatment of anhedonia in a cross-diagnostic fashion (37).

3.2. Behavior

Simple behavior changes have been shown to induce plasticity in the brain, which can lead to improvement in symptoms. The most commonly used behavioral approach to treat psychiatric disorders is CBT, an empirically supported treatment that aims to transform an individual’s cognitions to mitigate maladaptive thoughts and behaviors (38). Studies showing CBT-related neuroplasticity typically use pre-CBT and post-CBT assessments of structural MRI and fMRI to track changes in gray matter volume, neural activation, or functional connectivity of an area of interest (39–43). This pre/post design has been used to show evidence of CBT-induced plasticity in brain regions that are critical for mood regulation (44), fear conditioning (41), inhibition (39), and cognitive reappraisal (45).

One alternative approach to CBT is behavioral modification of the memory reconsolidation process (46, 47). When memories are acquired, synapses in the brain are remodeled through the degradation of old proteins (48) and the synthesis of new proteins (49–52). This process of memory consolidation lasts between 10 minutes and 24 hours (53), a window in which the memory is considered labile. This same consolidation process is also engaged when a previously stable long-term memory is reactivated through experience (53), allowing it to be modified or erased by pharmacological (46, 47) or behavioral manipulations (54–56). Recent research on fear conditioning shows that reactivation of a memory prior to extinction can eliminate conditioned amygdala responses to the previously feared stimulus (56) and that this modification is long lasting (55). This approach may offer new treatment options for trauma-related disorders [especially in combination with noninvasive neuromodulation techniques (57)].

3.3. Devices

Device-based neuromodulation approaches encompass a wide spectrum of strategies that can be categorized into subthreshold, subconvulsive, convulsive, and neurosurgical implantation therapies. The spatial focality of these stimulation modalities depends greatly on stimulation setup and parameters, such as electrode number/size/placement, coil geometry, and stimulation intensity. Figure 4 summarizes the typical spatial resolution and invasiveness profile of the various neuromodulation therapies.

Figure 4.

Spatial resolution and invasiveness profile of various neuromodulation therapies for mood disorders. Neuromodulation tools vary in their degree of invasiveness (x axis) and spatial resolution (y axis). Surgical approaches are the most invasive but also the most focal. Transcranial magnetic and electrical approaches tend to be the least invasive but also the least focal. New approaches to improve the focality of transcranial approaches include high-definition electrode arrays and novel approaches for focal deep brain stimulation such as temporal interference and transcranial focused ultrasound. Abbreviations: DBS, deep brain stimulation; dTMS, deep transcranial magnetic stimulation; ECT, electroconvulsive therapy; HD, high-definition; LFMS, low-field magnetic stimulation; MST, magnetic seizure therapy; sTMS, synchronized TMS; tES, transcranial electrical stimulation; tFUS, transcranial focused ultrasound; TI, temporal interference; TMS, transcranial magnetic stimulation; VNS, vagus nerve stimulation. Figure adapted with permission from Reference 160.

Subthreshold therapies involve the application of electromagnetic fields at levels below action potential threshold. Such subthreshold modalities include a family of transcranial electrical stimulation (tES) techniques that differ in their stimulus waveform (58): transcranial direct current stimulation (tDCS), transcranial alternating current stimulation (tACS), random noise stimulation, pulsed current stimulation, and slow oscillatory stimulation. Regardless of stimulus waveform, low-intensity current is typically injected into the head via scalp electrodes, which can be large, saline-soaked sponge pads or a multielectrode–conductive gel configuration (so-called high-definition tES). The applied current intensity is typically less than 4 mA, producing an electric field in the brain on the order of 0.5 V/m. Subthreshold magnetic stimulation includes low-field magnetic stimulation (LFMS) and synchronized transcranial magnetic stimulation (sTMS).

Subconvulsive therapies involve the application of electromagnetic fields at levels above action potential threshold but below seizure threshold. An example of such subconvulsive techniques is TMS, in which a stimulation coil is placed over the scalp and a coil current pulse induces a time-varying magnetic field, which in turn induces an electric field in the brain via electromagnetic induction. The induced electric field strength is on the order of 100 V/m.

The convulsive therapies involve the induction of a therapeutic generalized seizure in patients under anesthesia, either via the direct injection of electrical current pulse trains through scalp electrodes, as in the case of electroconvulsive therapy (ECT), or via electromagnetic induction, as in the case of magnetic seizure therapy (MST). The neurosurgical implantation therapies, such as DBS and VNS, involve the implantation of battery-powered devices to deliver chronic or intermittent electrical stimulation.

3.4. Combinatorial Therapy

It is of tremendous interest to explore whether the use of brain stimulation devices can augment the efficacy of pharmacological and behavioral interventions. There have not been many studies to date investigating the effectiveness of combining devices with drugs for acute therapeutic response (59). For example, in the use of ECT in depression, only a single randomized controlled trial combining ECT and antidepressants has been done to examine acute effects of the combination, with the finding that concomitant ECT with nortriptyline increased the effectiveness of ECT alone, while venlafaxine tended to worsen cognitive side effects (60). Combining ECT with antipsychotic drugs has been used to remediate the symptoms of schizophrenia, for example, in an open-label study in which combined use improved some cognitive deficits without worsening function (61) and in an earlier randomized study demonstrating that augmenting clozapine with ECT was safe and effective in treating psychosis (62). Similarly, in the use of tDCS in depression, a factorial study showed that the combination of tDCS with sertraline led to a synergistic response that was superior to tDCS or sertraline alone and to sham stimulation and placebo (63). For TMS in depression, the labeled indication is for patients who have failed to respond to prior antidepressant medication. Consequently, the TMS–drug interaction is less well studied. There remains the possibility that targeting noninvasive brain stimulation on a brain network known to be related to a drug’s therapeutic action may be key to making a previously ineffective drug effective. The groundwork for examining this possibility, at least for TMS, has been laid out by Ziemann et al. (64), who reviewed what is known about the interaction of drugs with TMS, focusing on the effects of central nervous system drugs on brain plasticity.

Combining the use of devices with behavioral therapy is another promising avenue of research. Brain stimulation is more effective—and generates more profound circuit changes—in neural circuits that are already active. Once a network involved in a given illness is singled out, if a particular set of neurons within it is activated via a behavioral therapy, then that neural system that is already therapeutically active can be specifically targeted for additional modulation by brain stimulation. A proof-of-concept study for this sort of functional targeting was reported using TMS combined with simultaneous working memory training to remediate working memory deficits caused by sleep deprivation (65). An adaptation of this approach to use TMS to augment the action of CBT in depression was laid out in a review by Luber et al. (66), and its feasibility was demonstrated in an open-label clinical study (67). There is a growing body of work on the simultaneous use of TMS and tDCS with psychotherapy (such as CBT or cognitive control therapy) in depression, schizophrenia, anxiety disorders, obsessive-compulsive disorder (OCD), post-traumatic stress disorder (PTSD), and autism (see 68). Unfortunately, a recent double-blinded, randomized, controlled trial combining tDCS with CBT for the treatment of depression did not find tDCS enhancement over and above CBT (69).

4. MECHANISMS OF CIRCUIT MODULATION

Devices modulate circuit function through the interaction of the applied electromagnetic fields and neuronal activity. How these fields interact with the brain can be examined at multiple spatial and temporal scales. Spatially, changes can be observed at the synaptic level; at the level of individual neurons; in local networks of neurons; and at the macro, systems-level brain network scale. Temporally, acute effects occur in the millisecond to second range, and changes can be seen over minutes to hours and over weeks to months. The latter type of long-lasting change is of particular interest in the therapeutic context.

4.1. Mechanisms Underlying Acute Effects of Stimulation

Electric fields affect individual neurons acutely by depolarizing or hyperpolarizing cell membranes. While the exact mechanism is not known, theoretical models have been suggested (70), for example, examining how the electric field might alter the conformational states of the neuronal membrane’s protein channels (71). A strong enough field can cause an individual neuron to fire. However, given the tight coupling of neurons in local networks, even much smaller fields can result in action potentials. In general, the interaction of electric fields and neuronal membranes is complex and depends on the part of the neuron involved, as well as its orientation relative to the field and the waveform of stimulation used. These interactions have, to some extent, been computationally modeled (e.g., 72) and tested in vitro (73). For example, radial currents parallel to the cortex are induced by electric stimulation and influence depolarization of the synapse, whereas tangential currents result in hyperpolarization (74). Overall, animal research has lagged behind the growing use of devices for noninvasive magnetic and electrical stimulation; however, recent developments have begun to fill that gap. For example, Mueller et al. (75) reported a methodology that successfully allowed recording of artifact-free, single-cell electrophysiology in awake behaving primates within 1 ms of a TMS pulse from a coil placed immediately above the stimulated cortex. Similarly, there have been few computer simulations studying effects of external stimulation. One such model, using 33,000 model neurons to demonstrate the effects of single and paired pulses of TMS on a local network, illustrated their potential utility (76). It should also be pointed out that, even in the small amount of work done to understand mechanisms using animal research and computational modeling, almost no attention has been paid to the effects on glial and endothelial cell function caused by stimulation (77).

4.2. Mechanisms at the Network Level: Engaging Cortical Oscillatory Behavior

In terms of modulating local circuit dynamics, recent work has suggested that understanding and manipulating endogenous brain oscillations with external stimulation may be a key to controlling and optimizing its effect. It is becoming increasingly clear that neural oscillations play a role in both local cortical processing and in coordinating that processing among multiple regions. Technologies such as TMS and transcranial alternating current stimulation (tACS) have shown promise in their ability to drive these oscillations (78, 79). Some examples include a study showing the effect of rTMS on altering alpha-band oscillations (80), another showing the propagation of traveling waves caused by TMS-induced gamma oscillations to adjacent brain regions (81), and another demonstrating the induction of slow waves similar to those involved in consolidating memories during sleep when TMS stimulated the sensorimotor cortex (82). There has also been some suggestion that engagement of cortical oscillatory activity with brain stimulation is behind the growing reports of cognitive enhancement and remediation of cognitive deficits (83). Furthermore, the many clear examples of departures from normal oscillatory behavior in psychiatric illnesses, for example, in depression where patients have increased synchrony in oscillations that is thought to result in an impairment in brain circuits (84, 85), point to possible mechanisms behind the therapeutic action of brain stimulation.

4.3. Mechanisms Related to Longer-Lasting Effects of Neuromodulation

One of the earliest effects found with brain stimulation was its ability to up- and downregulate cortical excitability, with such changes lasting minutes to hours beyond the end of a period of stimulation, typically applied for 10–20 min. The heuristics involving TMS and tDCS, for example, are that low-frequency trains of rTMS or cathodal tDCS lower excitability, while high-frequency rTMS and anodal tDCS raise it, although actual effects in specific circumstances depend on a number of other factors related to other stimulation parameters, the network stimulated, and individual variability (for TMS, see 86; for tDCS, see 87). Excitability changes have traditionally been measured as changes pre-/post-stimulation intervention in the amplitude of electromyographic potentials due to contractions in peripheral muscles evoked by the stimulation of the motor cortex—motor-evoked potentials. Other ways to measure these changes are with scalp EEG and performance on behavioral tasks (88, 89). A more recently developed method is to observe changes in blood-oxygen-level-dependent fMRI response and resting-state connectivity using MRI (90). In all of these cases, changes lasting days and weeks, or longer, have been observed when stimulation sessions were repeated daily over one or more weeks. For example, using TMS, observations of decreased levels of depression severity after weeks of daily sessions of 10-Hz stimulation to dlPFC ultimately led to its approval as a therapy for treatment-resistant depression by the FDA in 2008. A week of daily 5-Hz stimulation of the parietal cortex led to increased resting-state connectivity to the hippocampus and corresponding increases in memory performance (91). The suggested mechanisms behind the temporary changes in cortical excitability and the longer-lasting changes linked to repeated stimulation sessions are short-term and epigenetic synaptic changes, respectively. These mechanisms are linked to electrically induced long-term potentiation (LTP) and long-term depression (LTD), initially studied using hippocampal slices (92). Some work has been done directly linking brain stimulation with LTP/LTD. For example, tDCS was shown to bias modulation of LTP in hippocampal slices (93, 94). In rats, TMS has also been shown to affect expression of LTP in rats (95) and expression of genes related to LTP induction (96). There are also suggestive similarities with LTP/LTD: For example, NMDA receptor antagonists can prevent TMS plasticity effects (97). In addition, a great deal of work has been done linking a specific TMS effect involving temporally pairing TMS to the somatomotor cortex with afferent somatic stimulation with Hebbian-type neuroplasticity (98). Overall, there is much evidence linking the plasticity effects of brain stimulation–induced neuroplasticity with synaptic mechanisms underlying LTP/LTD.

5. EVIDENCE FOR THERAPEUTIC EFFICACY OF CIRCUIT MODULATION

5.1. Subthreshold

Here we discuss evidence for the therapeutic efficacy of subthreshold forms of neurostimulation, including tES, LFMS, and sTMS. To date, none of these has reached the level of evidence needed for FDA approval.

5.1.1. Transcranial electrical stimulation.

Several open-label studies and randomized controlled trials have been conducted investigating the antidepressant efficacy of tDCS (99–101). Early studies generally found active tDCS delivered to the left dlPFC to be more effective than sham for the reduction of depression severity (102). However, a large international, randomized, controlled trial of tDCS found no antidepressant difference between active and sham stimulation for unipolar or bipolar depression (103). Brain-derived neurotrophic factor genotype was not associated with response to tDCS (103). One of the questions raised as a result of this trial was whether sham tDCS was biologically active. During sham stimulation, a steady nonzero current (0.034 mA) was delivered. Furthermore, various device manufacturers implement the sham protocol differently, possibly contributing to inconsistencies in clinical trial results (104).

Another modality that is difficult to sham is tACS. A pilot trial using tACS to treat depression was recently completed (105), showing preliminary efficacy of 10-Hz tACS compared to 40-Hz tACS and sham at the 2-week follow-up, although the groups did not separate at the completion of the treatment trial at 4 weeks. There was also a significant reduction in EEG alpha power over the left frontal brain regions for patients who received 10-Hz tACS. A recent question related to the mechanism of action of tACS is whether its effects are mediated by transcutaneous stimulation of peripheral nerves in the scalp and not direct electrical stimulation of cortical neurons (106). Entrainment of cortical rhythmic activity by tACS has also been a topic of recent debate (107, 108).

5.1.2. Low-intensity magnetic stimulation.

LFMS administers a magnetic field that creates a low-intensity electric field in the brain using an MRI gradient coil (109). It was observed that patients had an improvement in mood following LFMS treatment compared to sham (109). The electric field of LFMS was observed to have penetrated to the brain’s cortical regions with a field strength of approximately 0.25 V/m. The electric field strength in the subcortical regions was measured at about 0.05 V/m (110). In a multicenter, randomized, controlled trial in treatment-resistant depression, LFMS did not produce any improvements in outcome after two days of stimulation compared to sham stimulation (111). Only a slight, nonsignificant improvement in the sad mood rated on the visual analog scale occurred after patients received active LFMS compared to sham (111). However, it was observed in a different study that LFMS did appear to induce an effect only after the third treatment session (112). Subsequently, while the effect of LFMS may be short lived, there is evidence that shows that a longer course of treatment could potentially improve the effects of LFMS.

5.1.3. Synchronized transcranial magnetic stimulation.

In line with the motivation to entrain with network neural oscillations, sTMS was developed. The sTMS device is composed of a configuration of three cylindrical neodymium magnets mounted over the midline frontal polar region, the superior frontal gyrus, and the parietal cortex. The speed of rotation for the magnets was set to the patient’s individualized resting-state peak alpha frequency of neural oscillations, as obtained by pretreatment EEG recorded from prefrontal and occipital regions (113). Computational modeling showed that the maximum induced electric field strength at the level of the cortex was approximately 0.02 V/m, which is an order of magnitude lower compared to those delivered by tES and LFMS. In a double-blind, sham-controlled depression treatment trial using sTMS, there was no difference in efficacy between active and sham in the intent-to-treat sample (113). However, in the per-protocol population that was more pharmacoresistant and received sTMS with the correct individualized alpha frequencies, active stimulation produced greater reductions in depressive symptoms compared to sham. Subsequent exploratory analysis showed significant increases in EEG alpha current spectral density from baseline to final treatment that were positively correlated with improvement on a self-rated inventory of depressive symptomatology scores in anterior midline brain regions in the active sTMS but not sham (114).

5.2. Transcranial Magnetic Stimulation

Since focal TMS targeted to one brain region has network-wide effects, there has been an emerging emphasis on using circuit-based approaches to target TMS to nodes implicated in psychiatric diseases such as depression, bipolar disorder, addiction, and many others (59, 115). For example, TMS has been applied most successfully to the left dlPFC, a key node within the central executive network. The central executive network and default mode network are consistently dysregulated in depressed patients, and stimulation to the left dlPFC has been shown to normalize connectivity both between and within both networks (116). TMS is also FDA approved for treating OCD, which involves dysregulation of the cortico–striato–thalamic–cortical circuit. Specifically, high-frequency deep TMS over the medial prefrontal cortex has been shown to reduce OCD symptoms (117).

While TMS is a promising treatment with a well-established safety profile, it is less effective than convulsive techniques such as ECT for some psychiatric conditions. To improve clinical efficacy through better targeting of circuit nodes of interest, several methods are being investigated to enhance the spatial precision of TMS. The original TMS spatial targeting method, which involved moving the TMS coil 5 cm anterior to the motor cortex, did not account for variations in head shape and functional anatomy. Neuronavigational systems can be used to individualize targeting based on structural and functional neuroimaging, which has significantly improved the spatial precision of TMS. A robotic system that allows for precise TMS coil placement and tracking has recently received FDA 510(k) clearance. Another ongoing effort inspired by electric field modeling involves changing the coil design itself. Figure-eight coils, which tend to have better focality at superficial depths, and H-coils, which enable slightly deeper stimulation, are currently FDA approved for treating depression and OCD, respectively (118). Multilocus TMS transducers, which include several overlapping figure-eight coils, are currently being developed to allow for more flexibility in choosing sites of stimulation and adjusting the induced electric field direction to optimize modulation of relevant circuits (119).

5.3. Convulsive

Here we present evidence for the therapeutic efficacy of convulsive approaches, including ECT and MST. Of these, only ECT is currently FDA approved.

5.3.1. Electroconvulsive therapy.

ECT is the longest-surviving somatic treatment in psychiatry, dating back to the 1930s, and is widely accepted as the gold-standard treatment for severe depression. The FDA recently reclassified ECT from preamendments class III (high risk) to class II (moderate risk) for the treatment of catatonia, or a severe major depressive disorder or bipolar disorder in patients aged 13 years and older who are treatment-resistant or who require a rapid response due to the severity of their psychiatric or medical condition (120). For the treatment of major depressive disorder in adults, ECT has a sustained response rate of approximately 80% and remission rate of 75% (121). Over half of the patients achieve an initial response by the first three sessions within the first week, and approximately a third of patients achieve remission by the sixth session within the second week of treatment (121). In the depressed elderly receiving right unilateral ECT treatment, approximately 60% of the patients met remission criteria, with a mean number of treatments to remission of 7.3 (122).

The understanding of the mechanisms of action of ECT remains incomplete. Neuroimaging and electrophysiological studies over the past decade have demonstrated support for two seemingly contradictory hypotheses: the anticonvulsant hypothesis, which posits that ECT enhances GABAergic tone, suppresses neural metabolic activity, increases seizure threshold, and decreases seizure duration over the course of treatment (123); and the neurotrophic hypothesis, which postulates that structural brain plasticity and cellular morphological changes—including an increase in the growth and production of transcription factors, neurogenesis, synaptogenesis, gliogenesis, and dendritic arborization—contribute to the therapeutic effects of ECT (124). For example, postictal EEG suppression is correlated with clinical response as well as reduced cerebral blood flow in the PFC post-ECT, supporting the anticonvulsant theory (125). Findings of post-ECT increases in cortical excitability, as measured by TMS-evoked potentials in the PFC (126), as well as volumetric increases in the hippocampus observed in neuroimaging studies (127) are both consistent with the neurotrophic hypothesis.

5.3.2. Magnetic seizure therapy.

MST involves inducing a series of generalized seizures under anesthesia using high-intensity and high-frequency rTMS. Computational electric field models show that the ECT-induced electric field is nonfocal and variable due to the high electrical impedance of the skull, current shunting in the scalp, and variation of head tissue anatomy, while MST offers more superficial and less intense stimulation (128). In vivo preclinical studies with nonhuman primates demonstrated the safety of MST and supported the hypothesis that the MST-induced current and resultant seizure are more focal than ECT (129). To date, a few dozen clinical trials and case reports have been published on the antidepressant efficacy of MST, which has been found to be similar to that of ECT. It has also been shown that MST has a faster reorientation and return to cognition compared to ECT (130, 131).

There has been very little neuroimaging with MST. In a case report of a single depressed patient, hexamethylpropylene amine oxime–single-photon emission-computed tomography (HMPAO-SPECT) was used at baseline and after 12 sessions of MST (132). The patient demonstrated antidepressant response to MST, and SPECT revealed an increase in blood flow in the frontoparietal cortex and basal ganglia. In an open-label study of 10 unipolar depressed patients undergoing a course of high-frequency (100 Hz) MST, fluorodeoxyglucose positron emission tomography (FDG-PET) showed that MST induced a relative increase in metabolism in a number of regions, including the basal ganglia, orbitofrontal cortex, medial frontal cortex, and dorsolateral prefrontal cortex (133). Sparing frontal and temporal regions from ECT-induced functional impairments may be key to the superior cognitive outcomes seen with MST.

5.4. Surgical

Here we present evidence for the therapeutic efficacy of surgical approaches, including DBS, VNS, and epidural stimulation. Of these, DBS is FDA approved in certain movement disorders and has a humanitarian use exemption for OCD, and VNS is approved for treatment-resistant epilepsy and also treatment-resistant depression.

5.4.1. Deep brain stimulation.

DBS is a neurosurgical intervention that uses electrodes delivering high-frequency current in cortical or subcortical areas. The FDA granted a Humanitarian Device Exemption for the use of DBS targeting the ventral capsule/ventral striatum in medication-refractory OCD. DBS is under study for other psychiatric disorders such as depression and PTSD. Initial open-label trials for treatment-resistant depression showed promising results, but inconsistent results were found in pivotal randomized clinical trials targeting the subcallosal cingulate (19) and the ventral capsule/ventral striatum (134). The reasons for these failures are not known, in part because these pivotal trials did not include measures of target engagement. However, clues can be found in secondary analyses where the connectivity of the surgical site to the distributed circuitry involved in depression may have been key in determining response. Furthermore, no optimization or individualization was done to select the frequency of stimulation or its duty cycle. While chronic high-frequency stimulation delivered continuously may be effective for movement disorders, those temporal dynamics may not be optimal for a complex behavioral condition such as depression, where mood state waxes and wanes during the day and in which there are classically observed disruptions in sleep architecture and circadian rhythm. A discussion of the possible reasons for these failures and potential ways forward can be found in Reference 135. Other targets for depression are actively being explored, including the median forebrain bundle, nucleus accumbens, inferior thalamic peduncle, and habenula.

5.4.2. Vagus nerve stimulation.

VNS is an implanted device that delivers electrical pulses to the left cervical vagus nerve. As with DBS, the pulse generator is implanted in the chest wall, but instead of targeting the brain directly, the leads attach to the left vagus nerve in the neck. VNS parameters such as current, frequency, and duty cycle are adjusted over the course of treatment using a programming wand connected to a portable computer. The FDA has approved VNS for the treatment of severe, recurrent unipolar and bipolar depression in patients that failed at least four antidepressant treatments. Already on the market for medication-refractory epilepsy, the serendipitous observation that epilepsy patients with depression also had improvements in their depression after VNS was followed up with a randomized, controlled trial on depression. While the pivotal trial failed to meet its end point (136), VNS subsequently received FDA approval on the basis of a comparison with a treatment-as-usual group in an unblinded design. Unlike other depression treatments where a pattern of relapse is common, VNS demonstrates excellent long-term sustained benefits in responders (137, 138).

VNS is thought to act via a bottom-up approach that stimulates vagal afferent fibers in the neck, targeting the nucleus of the solitary tract, the locus coeruleus, and the dorsal raphe nucleus (139). Other proposed mechanisms of antidepressant action implicate an increase in inhibitory neurotransmission (140). Supporting that hypothesis, motor cortex excitability by TMS was measured in major depressive disorder patients before, during, and after treatment with VNS, and results showed that after 10 weeks, VNS modulates the central nervous cortical inhibitory pathways (141). Clinical uptake for VNS has been slowed by the lack of insurance coverage, but the US Centers for Medicare and Medicaid Services have recently announced coverage for VNS through the Coverage with Evidence Development framework. Starting in the summer of 2019, 1,000 patients will be enrolled in a sham-controlled trial of VNS designed to evaluate the evidence for safety and efficacy and to inform future coverage decisions (https://clinicaltrials.gov/ct2/show/NCT03887715). Other companies are exploring the potentially more cost-effective approach of wearables targeting the vagus nerve transcutaneously (142).

5.4.3. Epidural cortical stimulation.

Epidural prefrontal cortical stimulation (EpCS) is a surgically implanted device that delivers electrical stimuli to cortical regions such as the prefrontal cortex. EpCS has been identified as a potential treatment for patients with treatment-resistant depression but is currently only used experimentally for this purpose. This surgical approach is arguably safer and less invasive than DBS because the electrodes do not penetrate the brain parenchyma. EpCS offers stimulation advantages with a wide range of stimulation configurations that vary in their pulse width, intensity, and frequency parameters. Several groups are trying to evaluate whether EpCS could be an efficient and safe antidepressant therapy. Kopell et al. (143) found a durable antidepressant effect in treatment-resistant depression patients and are investigating the dlPFC as a target region for EpCS, while Williams (144) targeted the frontopolar cortex, and Nahas et al. (145) used bilateral anterior pole and midlateral EpCS.

Different targeting approaches are being used during EpCS surgery for electrode placement. Kopell et al. (143) used MRI before the placement of the stimulation electrode to identify the dlPFC region, transferring the imaging data to a frameless stereotactic neuronavigational system that was used by the neurosurgeon during the placement of the electrode. Nahas et al. (145) used a similar approach with the addition of a three-dimensional head surface with the brain regions selected for the stimulation in distinctive color and an external wand for easier recognition and intraoperative fluoroscopy.

6. FRONTIERS OF RESEARCH ON THERAPEUTIC CIRCUIT MODULATION

Many more interesting studies could be done if brain stimulation devices could noninvasively modulate deep brain regions without affecting the superficial cortex. Unfortunately, at the frequency ranges that are typically used with conventional techniques for tES and TMS, one could not achieve focal deep stimulation using any superposition of extracranial current sources (146). The induced electric field is generally more intense near the surface of the brain and attenuates in depth, following a depth–focality tradeoff (118). Recent proposals to overcome this physical limitation include the following: the use of temporally interfering waveforms (147), in which two high-frequency sine-wave electric field waveforms with a small frequency difference are used to form a beat signal with a low-frequency envelope deep in the brain; and the use of intersectional short pulse stimulation (148), in which fast pulses (<10-μs pulse width) are delivered via multielectrode pairs in different vector directions, exploiting the neurons’ ability to temporally summate electric field inputs of different spatial orientations. These techniques remain to be validated and replicated; a preliminary study using computational neuron models failed to support the experimentally reported effects of interferential stimulation (149).

Transcranial focused ultrasound (tFUS) is a new and promising noninvasive neuromodulatory technique with spatial and temporal precision on the order of millimeters and milliseconds and adjustable focus in deep brain regions (150, 151). In humans, Legon et al. (152) first demonstrated that low-intensity focused ultrasound (<50 W/cm, 0.5 MHz) delivered to the somatosensory cortex transiently suppressed somatosensory evoked potentials and modulated sensory detection thresholds. Recent studies in nonhuman primates showed that tFUS can alter local functional connectivity, with effects lasting up to several hours beyond the stimulation period (153, 154). Another exciting application of ultrasound is focal uncaging of neuromodulatory drugs from nanoparticle drug carriers (155). A recent rodent study showed that ultrasonic anesthetic propofol uncaging directed to the primary visual cortex can selectively and reversibly suppress visual evoked potentials (156). Another recent development of potential consequence is the report that tying the timing of TMS stimulation to a particular phase of an ongoing functional oscillation, such as alpha, increased the effect of the TMS pulse (157). Such closed-loop control, tied to an immediate electrophysiological state, serves both to help explore the function of endogenous oscillations and to potentially optimize the therapeutic delivery of stimulation (158).

7. AREAS WHERE WE NEED TO KNOW MORE

As exciting as these developments in device-based modulation of neurocircuits for treatment in psychiatry are, there are many unanswered questions regarding the mechanisms of action of these interventions at a biophysical level, which is the topic of a currently available funding opportunity in the NIH BRAIN Initiative. More complete understanding of mechanisms, across levels of analysis, will be key to our ability to optimize these promising treatments. Perhaps the greatest challenge to the field of device therapeutics in psychiatry, though, is the one that is shared with the larger field of intervention development: insufficient knowledge of the mechanisms underlying psychiatric disorders themselves and their neurodevelopmental origins. When we understand the developmental processes during infancy and early childhood that lead to the development of disabling psychiatric disorders in adolescence and adulthood, we could leverage that knowledge to design targeted prevention strategies, be they pharmacological or device based. Tools to safely and precisely modulate circuit function have the exciting potential of translating emerging knowledge about the neurodevelopmental origins of circuit dysfunction leading to severe mental illness into effective prevention strategies. At present, we have limited knowledge of the impact of brain stimulation on the course of brain development, and the impact of such knowledge could be truly transformative.

Glossary

TMS

transcranial magnetic stimulation uses brief, intense pulses of electric current delivered to a coil placed on the head to generate an electric field in the brain via electromagnetic induction

VNS

vagus nerve stimulation involves the surgical implantation of a pulse generator to deliver intermittent electrical stimulation to the left vagus nerve via helical electrodes

DBS

deep brain stimulation involves the surgical implantation of a pulse generator and electrodes to deliver chronic electrical stimulation to focal deep brain regions

tDCS

transcranial direct-current stimulation uses constant, low-amplitude direct current (typically 1–2 mA) delivered via scalp electrodes to modulate cortical excitability

LFMS

low-field magnetic stimulation employs very weak magnetic fields applied uniformly throughout the brain

ECT

electroconvulsive therapy is administered by delivering electricity to the brain via scalp electrodes to induce a generalized tonic–clonic seizure

MST

magnetic seizure therapy involves inducing a generalized seizure under anesthesia using high-dose repetitive TMS