Targeting Gamma-Related Pathophysiology in Autism Spectrum Disorder using Transcranial Electrical Stimulation: Opportunities and Challenges

Abstract

A range of scalp electroencephalogram (EEG) abnormalities correlate with the core symptoms of autism spectrum disorder (ASD). Among these are alterations of brain oscillations in the gamma-frequency EEG band in adults and children with ASD, whose origin has been linked to dysfunctions of inhibitory interneuron signaling. While therapeutic interventions aimed to modulate gamma oscillations are being tested for neuropsychiatric disorders such as schizophrenia, Alzheimer’s disease and frontotemporal dementia, the prospects for therapeutic gamma modulation in ASD have not been extensively studied. Accordingly, we discuss gamma-related alterations in the setting of ASD pathophysiology, as well as potential interventions that can enhance gamma oscillations in patients with ASD. Ultimately, we argue that transcranial electrical stimulation (tES) modalities capable of entraining gamma oscillations, and thereby potentially modulating inhibitory interneuron circuitry, are promising methods to study and mitigate gamma alterations in ASD.

Lay Summary

Brain functions are mediated by various oscillatory waves of neuronal activity, ranging in amplitude and frequency. In certain neuropsychiatric disorders, such as schizophrenia and Alzheimer’s disease, reduced high-frequency oscillations in the “gamma” band have been observed, and therapeutic interventions to enhance such activity are being explored. Here, we review and comment on evidence of reduced gamma activity in ASD, arguing that modalities used in other disorders may benefit individuals with ASD as well.

Introduction

Autism spectrum disorder (ASD) is a broad clinical term used to describe a group of complex neurodevelopmental disorders defined by two core behavioral symptoms, according to the DSM-5: (1) persistent deficiencies in social communication and social interaction, and (2) limited interest, repetitive behavior, and activity (Association, 2013). ASD and its common co-morbid disorders, which include anxiety, depression, attention-deficit/hyperactivity disorder (ADHD), and obsessive compulsive disorder (OCD), often lead to significant impairments in activities of daily living (ADLs), both in children and adults (Haertl, Callahan, Markovics, & Sheppard, 2013). According to a 2018 Autism and Developmental Disabilities Monitoring (ADDM) Network report, the United States prevalence of ASD in 2012 was 14.5 per 1,000 children aged 8 years, with prevalence estimates ranging among states from 8.2 to 24.6 per 1,000 (Christensen et al., 2018). The report urged research initiatives to respond to this growing health issue and described the need in particular to address the biological substrates leading to ASD’s core symptoms (Christensen et al., 2018).

Recent research into ASD pathophysiology, combined with active development of practical ASD biomarkers (e.g., Jannati et al., in press), has led to identification of numerous genetic, physiologic and anatomic variants that are reliably found in the ASD population. Among these is growing list of gene mutations that contribute to ASD, and (even in cases where a genetic abnormality is not evident) aberrant synaptic plasticity, and interneuron deficits (Figure 1). From the physiologic perspective, most EEG studies in ASD indicate deficient neural oscillatory patterns in the gamma frequency band (Maxwell et al., 2015; Rojas & Wilson, 2014; Stroganova et al., 2015; Sun et al., 2012; van Diessen, Senders, Jansen, Boersma, & Bruining, 2015). Notably, gamma-band EEG findings in ASD share similar characteristics with those in Alzheimer’s disease (AD), frontotemporal Dementia (FTD), and schizophrenia (SCZ), where novel therapies aimed at restoring gamma activity are currently being tested (Gandal, Edgar, Klook, & Siegel, 2012a; Iaccarino et al., 2016; Poza, Hornero, Abasolo, Fernandez, & Escudero, 2007; Sami et al., 2018; Spencer et al., 2003a; Suazo et al., 2012; Woo, Spencer, & McCarley, 2010). Thus, in ASD, as in other central nervous system (CNS) disorders, gamma band EEG metrics may provide both a disease biomarker and a measure of therapeutic target engagement.

Figure 1. Framework for ASD-related alterations.

ASD pathophysiology spans from genetic to global levels, with gamma abnormalities possibly playing a crucial role in linking cellular-level alterations involving inhibitory interneurons to E/I imbalance in ASD. Sources: (a) Depienne et al., 2009; (b) Iossifov et al., 2015; (c) Hulbert et al., 2016; (d) Devlin & Scherer, 2012; (e) Hashemi et al., 2016; (f) Wegiel et al., 2014; (g) Hutsler & Zhang, 2010; (h) Courchesne et al., 1997; (i) Dapretto et al., 2006; (j) Jou et al., 2011; (k) Casanova et al., 2010; (l) Brown et al., 2005; (m) Courchesne et al., 2001; (n) Oberman et al., 2016; (o) Zielinski et al., 2014; (p) Wegiel et al., 2010.

In the present commentary we will: (1) examine how gamma EEG band alterations may inform ASD symptomatology and (2) discuss plausible interventions to target gamma abnormalities in patients with ASD by transcranial electrical stimulation (tES), which we argue is a safe and feasible methodology to entrain and modulate gamma activity in the human brain.

The role of gamma EEG activity in the healthy and pathological brain

Endogenous brain oscillatory activity represents synchronized activity of neuronal ensembles. In humans, fast oscillatory EEG activity takes place in the 30–120 Hz frequency range, known as the “gamma” frequency band (Jia & Kohn, 2011). Recent studies have further identified gamma frequency sub-bands, characterized as slow/low gamma (30–50 Hz), mid-frequency gamma (50–90 Hz) or fast/high gamma (>90 Hz), although these frequency sub-band ranges vary between studies (Buzsáki & Wang, 2012a). Most notably, the discrete 40 Hz point has served as a neurophysiological frequency of interest, as it may be integral in supporting cortical arousal and sensory processing (McDermott et al., n.d.); thus, 40 Hz has served as a common EEG frequency for studies in gamma modulation and targeting.

Electrophysiological and behavioral evidence suggests that gamma activity mediates a range of essential neural functions, including sensory-motor integration, perceptual binding, working memory, attention-dependent stimulus selection, network synchronization, and higher-order cognition (Tallon-Baudry, 2009, 1999; Jia & Kohn, 2011; Buzsáki & Wang, 2012b; Hermes, Miller, Wandell, & Winawer, 2015; Uhlhaas & Singer, 2006). As measured by EEG and magnetoencephalography (MEG), cortical gamma activity is present during resting-state conditions as “spontaneous gamma” (Rojas & Wilson, 2014). Stimulus-related gamma perturbations can also be measured as “evoked” or “induced” gamma modulation (Amo, de Santiago, Barea, López-Dorado, & Boquete, 2017; Rojas & Wilson, 2014). Evoked responses are both time- and phase-locked to the onset of the stimulus, whereas induced responses are time-locked, but not phase-locked, to the stimulus onset (Cohen, 2014).

Gamma activity is thought to arise from populations of fast-spiking, parvalbumin (PV)-expressing interneurons, a family of gamma aminobutyric acid- (GABA-)ergic inhibitory neurons distributed throughout the cerebral cortex (Sohal, Zhang, Yizhar, & Deisseroth, 2009; Tamás, Buhl, Lörincz, & Somogyi, 2000; Whittington, Traub, & Jefferys, 1995; Ylinen et al., 1995). PV interneurons interact with local excitatory pyramidal neurons (PN) and other inhibitory cells to drive the excitation/inhibition (E/I) ratio in cortical and subcortical networks (Bartos, Vida, & Jonas, 2007; Hu, Vogt, Sandberg, & Rubenstein, 2017; Markram et al., 2004; Yizhar et al., 2011). PV interneurons generate feedforward and feedback inhibitory circuits and, notably, induce gamma rhythms as an emergent property of this interneuronal synchronization (Ferguson & Gao, 2018; Tamás et al., 2000; Whittington et al., 1995). Loss of PV density and disrupted PV-PN network interactions are associated with alterations of inhibitory control and correspond to abnormalities in gamma activity (Ferguson & Gao, 2018; Volman, Behrens, & Sejnowski, 2011). Although studies have generally reported reduced gamma power associated with aberrant PV interneuron dynamics (Sohal et al., 2009), there is likely a rich interplay of PV-PN connections, PV-PV autoconnections, and cortical state-dependent conditions involved in the network dynamics that predict how PV abnormalities affect gamma rhythms in disease states (Deleuze et al., 2019; Gandal et al., 2012a; Gonzalez-Burgos, Fish, & Lewis, 2011; Lodge, Behrens, & Grace, 2009).

In recent years, gamma activity has gained attention as a potential biomarker for AD, FTD and SCZ (Gandal et al., 2012a; Sami et al., 2018). While the nature of alterations in gamma activity varies among these disease states, the commonality of abnormal gamma activity indicates its physiological significance across neuropsychiatric disorders. For instance, AD and FTD patients exhibit loss of power/connectivity in the gamma band (Guillon et al., 2017; Koenig et al., 2005; Ribary et al., 1991; Stam et al., 2002), a finding that has been replicated in mouse models of AD (Gillespie et al., 2016; Iaccarino et al., 2016; Kurudenkandy et al., 2014; Verret et al., 2012). In contrast, SCZ patients exhibit irregular rhythmicity and elevated power in the gamma band (Gandal, Edgar, Klook, & Siegel, 2012b; Spencer et al., 2003b). A study on SCZ patients found the degree of abnormal elevation of gamma power to be correlated with the severity of negative symptoms in SCZ (Suazo et al., 2012), even though pathological studies indicate loss of PV+ interneuron function in SCZ (Do, Cuenod, & Hensch, 2015).

Because gamma EEG abnormalities have been reliably detected across these neuropsychiatric disorders, therapeutic applications aimed at normalizing aberrant gamma activity are being investigated in both animal models and exploratory clinical studies. In a seminal study on a mouse model of AD, optogenetic activation of hippocampal PV-interneurons at 40 Hz reduced levels of hippocampal amyloid-β (Aβ) isoforms (Iaccarino et al., 2016). Moreover, a noninvasive, 40-Hz light-flicker treatment targeting primary visual cortex (V1) significantly reduced Aβ isoforms and microglia response in V1 (Iaccarino et al., 2016). In a recent study on mouse models of neurodegeneration, 40-Hz light-flicker stimulation markedly preserved synaptic density, improved spatial learning, and enhanced neuroprotective factors in the visual cortex, hippocampus and prefrontal cortex (Adaikkan et al., 2019).

While gamma modulation for therapeutic purposes in AD remains a novel approach, preliminary studies using multisensory and noninvasive stimulation show promising results. For example, an exploratory study reported marked behavioral improvement following 40-Hz somatosensory stimulation in patients with mild to moderate AD (Clements-Cortes, Ahonen, Evans, Freedman, & Bartel, 2016). Similar gamma-modulating approaches in animal models and pilot clinical studies of SCZ have been investigated. For example, Lee and colleagues, (H. Lee, Dvorak, & Fenton, 2014), found applying the anticonvulsant ethosuximide normalized task-associated gamma activity in rats with neonatal ventral hippocampus lesions. In another study, repetitive transcranial magnetic stimulation (rTMS) of the dorsolateral prefrontal cortex normalized gamma hyperactivity and improved cognitive performance among SCZ patients (Farzan, Barr, Sun, Fitzgerald, & Daskalakis, 2012). Overall, converging lines of evidence suggest gamma-band abnormalities in AD, FTD, and SCZ may serve as valuable therapeutic targets and could lead to advances in treatment of neuropsychiatric patients.

Abnormal Gamma Activity in ASD

Recent studies indicate gamma-band activity significantly differs between healthy controls and individuals with ASD, in both pediatric and adult populations. These studies vary in methodology and gamma sub-bands of interest, as some measure spontaneous gamma activity while others assess evoked or induced gamma activity during sensory or cognitive tasks (Table 1). While the majority suggest alterations in low- and mid-frequency gamma in ASD, future research should establish a consensus regarding a taxonomy of methodologies for measuring and modulating gamma oscillations in ASD.

Table 1.

Evidence for Gamma-Related Alterations in ASD.

Paper	Sample Size	Age	Frequency of Interest	Gamma Type	Condition	Measured by	Gamma Findings in ASD compared to HC	Region/Activity of Gamma Alterations	Other Findings
Orekhova, 2007	n = 40 ASD (male); n= 40 HC	3–8 y/o	Gamma; Beta	Spontaneous Gamma	Sustained Visual Attention Task	EEG	+ Gamma	Midline, central and parietal electrodes	N/A
Orekhova, 2008	n=21 ASD; n=21 HC	3–8 y/o	Gamma; P50 supression	Spontaneous Gamma	Paired-click Paradigm	EEG	+ Gamma	Frontal, central and parietal electrodes	− P50 supression
Machado, 2015	n=11 ASD, n=14 HC	3–8 y/o	Gamma; Alpha; Delta; Beta	Spontaneous Gamma	Basal versus Visual+/−Audio	qEEG	+ Gamma	in Visual–Audio versus Visual with muted audio band, Right hemisphere	− Alpha; + Delta; +Beta
Cornew, 2013	n= 27 ASD, n= 23 HC	6–15 y/o	Gamma; Delta; Aplha	Spontaneous Gamma	Eyes-closed resting state	MEG	+Gamma (low, high and fast)	Anterior temporal, posterior temporal and occipital sites	+Alpha; +Delta
Sheikhani, 2009	n=15 ASD, n=11 HC	6–11 y/o	Gamma; Beta; Alpha	Spontaneous Gamma	Eyes-open resting state	qEEG	− Gamma	Temporal and frontal electrodes	− Alpha; +Beta
Maxwell, 2015	n= 15 ASD (all male), n=18 HC	9–18 y/o	Gamma	Spontaneous Gamma	Eyes-open resting state	EEG	− Gamma	Right lateral electrodes	N/A
Sheikhani, 2012	n=17 ASD, n=11 HC	6–11 y/o	Gamma; Beta; Alpha	Spontaneous Gamma	Eyes-open resting state	qEEG	− Gamma	Temporal and frontal electrodes	− Alpha; +Beta
Milne, 2009	n= 20 ASD, n=20 HC	8–18 y/o	Gamma; Alpha	Visually evoked gamma	Gabor patches	EEG	− Gamma	During Gabor patch visual perpection, Posterior cortical regions	− Low Alpha
Snijders, 2013	n= 12 ASD, n=12 HC	21–25 y/o	Gamma	Visually induced gamma	Gabor patches	EEG	− Gamma	During Gabor patch visual perpection, Posterior cortical regions	N/A
Grice, 2001	n=8 ASD, n= 8 WS, n=8 HC	30–36 y/o (mean ages)	Gamma	Visually evoked and induced gamma	Mooney face	EEG	− Induced Gamma	In response to upright Mooney faces relative to inverted Mooney faces	N/A
Sun, 2012	n= 13 ASD, n=16 HC	~30 y/o (mean age)	Gamma	Visually evoked and induced gamma	Mooney face	MEG	− Induced Gamma and phase-locking	To upright Mooney versus inverted Mooney; occipitoparietal sensors	N/A
Brown, 2005	n= 16 ASD, 8=HC	12–16 y/o (mean)	Gamma	Visually induced gamma	Prescence vs Absence of Kanizsa figures	EEG	+ Gamma	In response to Kanizsa-shape trials, parietal electrode sites	N/A
Stroganova, 2012	n=23 ASD (all male); n=23 HC (all male)	3–7 y/o	Gamma; Beta	Visually evoked gamma	Kanizsa squares versus non-Kanizsa squares	EEG	− Gamma	In response to Kanizsa-shape trials, V1 electrodes	− Beta
Rojas, 2008	n= 16 parents of ASD children; n = 11 ASD adults; n= 16 HC	20–50 y/o	Gamma	Auditory gamma-band responses	1 kHz sine-wave stimuli	MEG	+ Induced Gamma; − Evoked Gamma	In response to 1 kHz sine-wave stimulus	Parents of ASD children had same response to 1 kHz as ASD group
Gandal, 2010	n=25 ASD children, n=17 HC	8–12 y/o	Gamma	Auditory gamma-band responses	200, 300, 500, 1000 Hz	MEG	− Gamma-band phase-locking	To pure-tone stimulus presentations	Valproic acid (VPA) ASD mouse showed reduced gamma frequency phase-locking
Edgar, 2015	n=105 ASD children, n=36 HC	6–16 y/o	Gamma	Auditory gamma-band responses	200, 300, 500, 1000 Hz	MEG	+ Total Gamma Power (Induced+ Evoked)	In all frequency bands, left and right superior temporal gyrus	N/A

Note. Spontaneous, induced, and evoked gamma activity is generally reduced in ASD. However, some studies have obtained contradicting evidence, likely due to variability in experimental design, age group, and gamma-estimation techniques.

In one of the first studies on spontaneous gamma activity in ASD, young boys with ASD showed abnormally increased activity in the beta/gamma range (24.4–44.0 Hz) in midline, central and parietal EEG electrodes, during a sustained visual attention task (Orekhova et al., 2007), and the amount of beta/gamma activity was positively correlated with the degree of developmental delay. Because both PV- and somatostatin- (SOM-)expressing GABAergic interneurons contribute to the balance between gamma- and beta-band activity (Kuki et al., 2015), these results may indicate abnormal activity in both PV- and SOM-expressing interneurons in ASD. In a follow-up study that reproduced elevated spontaneous gamma activity in a different pediatric ASD cohort, the authors hypothesized this excess of high-frequency activity recorded in frontal, central and parietal EEG electrodes may reflect an underlying imbalance in cortical E/I tone (Orekhova et al., 2008). Similar results were obtained by Machado et al., (Machado et al., 2015), who found an abnormally increased activity in the 25–55 Hz range in the right hemisphere of children with ASD. In an MEG study, Cornew and colleagues, (Cornew, Roberts, Blaskey, & Edgar, 2012), found abnormally elevated spontaneous gamma power in temporal and occipital regions in high-functioning children with ASD. However, considering the critical role of PV interneurons in generating endogenous gamma oscillations (Cardin et al., 2009; Sohal et al., 2009) and the evidence for loss of PV-interneuron expression/function in ASD (Filice, Vörckel, Sungur, Wöhr, & Schwaller, 2016; Hashemi, Ariza, Rogers, Noctor, & Martínez-Cerdeño, 2016; Saunders et al., 2013; Wöhr et al., 2015), the gamma activity in ASD is expected to be abnormally reduced. Consistent with this hypothesis, several studies in ASD have found reduced spontaneous gamma activity in frontal, temporal, and right-lateral regions (Maxwell et al., 2015; Sheikhani, Behnam, Mohammadi, Noroozian, & Mohammadi, 2012; Sheikhani, Behnam, Noroozian, Mohammadi, & Mohammadi, 2009), reduced left-hemispheric MEG steady-state gamma responses (Wilson, Rojas, Reite, Teale, & Rogers, 2007), and reduced task-related gamma power and reduced long- and short-range gamma connectivity (Khan et al., 2013; Sun et al., 2012).

Abnormalities in evoked and induced gamma activity by sensory tasks have also been reported in ASD populations. In the visual domain, evoked gamma activity in both low-level sensory and high-level cognitive tasks has been found to be abnormally reduced in adolescents and adults with ASD (Milne, Scope, Pascalis, Buckley, & Makeig, 2009; Snijders, Milivojevic, & Kemner, 2013). Induced gamma activity during visual-perception tasks also has been found to be reduced in adults and adolescents with ASD (Grice et al., 2001; Stroganova et al., 2012; Sun et al., 2012). Similarly, studies in the auditory domain have found reduced evoked and induced gamma activity in both adults and children with ASD (Edgar et al., 2015; Gandal et al., 2010; Rojas, Maharajh, Teale, & Rogers, 2008).

While a majority of studies have found reduced evoked and induced gamma activity in individuals with ASD, some studies have obtained contradicting results in terms of localization, magnitude and direction of altered gamma activity (Rojas & Wilson, 2014)(see Table 1). Moreover, due to considerable variability in age range, method of gamma estimation, and stimulus type across studies (Rojas & Wilson, 2014), more robust and generalizable findings on gamma abnormalities may require establishing a consensus on techniques for quantifying spontaneous and induced/evoked gamma activity among ASD populations.

Gamma Activity and ASD Pathophysiology

In light of these lines of evidence, we argue that abnormal activity in the gamma range –either increased or decreased— should be considered part of the broad pathophysiological framework of ASD (Figure 1). Conventional models of ASD, based on decades of animal research and clinical studies, span from genetic to multi-system, global levels in an effort to better inform ASD etiology and pathophysiological mechanisms (Varghese et al., 2017). The genetic level of ASD pathophysiology points to mutations and genetic risk factors including inherited chromosomal abnormalities, rare genetic variants, and de novo mutations (Depienne et al., 2009; Devlin & Scherer, 2012; Grove et al., 2019; Hulbert & Jiang, 2016; Iossifov et al., 2014; Ramaswami & Geschwind, 2018). The cellular level focuses on abnormalities in neuronal morphology, growth, development, pruning, and cell-type densities (Courchesne, 1997; Hashemi et al., 2016; Hutsler & Zhang, 2010; Marchetto et al., 2017; Peter et al., 2016; Wegiel et al., 2014; Yip, Soghomonian, & Blatt, 2007). The framework then expands to systems and circuit-based alterations, commonly highlighted by regional cytoarchitecture and connectivity alterations (Bachevalier & Loveland, 2006; Casanova, El-Baz, Vanbogaert, Narahari, & Switala, 2010; Dapretto et al., 2006; Eigsti & Shapiro, 2003; Gotts et al., 2012; Jou et al., 2011; Just, Keller, Malave, Kana, & Varma, 2012; Marsh & Hamilton, 2011; Minshew, Luna, & Sweeney, 1999; Takarae, Minshew, Luna, & Sweeney, 2007; Zikopoulos & Barbas, 2013); of note, this level also captures the mirror neuron system (MNS), which has come under scrutiny in recent years as various studies and meta-analyses now argue against MNS dysfunction in ASD (Dapretto et al., 2006; Hamilton, 2013; Hamilton, Brindley, & Frith, 2007; Schulte-Rüther et al., 2017). It is at this systems level that we propose abnormal gamma activity be integrated into the framework of ASD pathophysiology. This could serve as a relevant, informative model as gamma alterations could play a role in linking cellular-level alterations involving inhibitory interneurons to E/I imbalances reported in ASD (Brown, Gruber, Boucher, Rippon, & Brock, 2005; Gogolla et al., 2009; Nelson & Valakh, 2015; Rubenstein & Merzenich, 2003). Finally, global-level characteristics, such as brain morphology, development and activity, serve to bridge higher-level dysfunction and behavioral deficits seen in core symptoms of ASD (Brambilla et al., 2003; Courchesne et al., 2001; Jannati et al., 2020; McAlonan et al., 2005; Oberman et al., 2016; Sparks et al., 2002; Wegiel et al., 2010; Zielinski et al., 2014).

While the potential pathophysiological role of gamma activity in ASD continues to gain attention, there is no robust initiative focused on gamma abnormalities in ASD to understand their etiology and to investigate therapeutic approaches to mitigate or normalize them. Thus, for the remainder of the paper, we will discuss the gamma-modulatory capabilities of transcranial electrical stimulation (tES) and its theoretical and practical considerations in ASD-related gamma dysfunction.

Transcranial Electrical Stimulation (tES)

tES is a broad term referring to a subset of noninvasive brain stimulation (NIBS) techniques, which deliver weak electric currents (typically 1–2 mA) transcranially via scalp electrodes (M. A. Nitsche, Liebetanz, Tergau, & Paulus, 2002). This electric current modulates spontaneous neuronal firing using subthreshold changes in resting-membrane potentials without generating action potentials (Brunoni et al., 2012). Through this mechanism, tES influences the ongoing neural activity and induces long-term modulatory effects leading to reproducible facilitatory or inhibitory plastic changes in the targeted neural systems (Yavari, Jamil, Mosayebi Samani, Vidor, & Nitsche, 2018; Yavari, Nitsche, & Ekhtiari, 2017). A notable feature of tES is its ability to apply electrical currents in various forms (e.g., direct current, random noise, alternating current) to almost any cortical region and concurrently influence a behavior of interest, yielding short-term and long-term behavioral outcomes (Reed & Cohen Kadosh, 2018; Yavari et al., 2018). Because of their valuable applications in experimental and clinical neuroscience research, tES techniques have gained popularity as affordable, low-risk, and user-friendly tools, particularly in double-blind, sham-controlled studies (Yavari et al., 2018).

tES modalities include transcranial direct current stimulation (tDCS), transcranial random noise stimulation (tRNS), and transcranial alternating current stimulation (tACS) (Figure 2A) (Santarnecchi et al., 2015; Yavari et al., 2018). tDCS uses a constant weak electrical current to induce bidirectional, polarity-dependent changes in cortical regions, enabling measuring effects at neuropsychological, physiological, and motor levels (Brunoni et al., 2012; Michael A Nitsche & Paulus, 2000). Because tDCS is well-tolerated and is associated with mild adverse effects in humans, it has been employed in clinical research to investigate major depressive disorder, addiction, chronic neuropathy, Parkinsonism, stroke, and traumatic spinal cord injury (Brunoni et al., 2012). In comparison, tRNS delivers a low-intensity, randomly alternating biphasic current directly to the scalp, with frequencies ranging from 0.1–640 Hz (Antal & Herrmann, 2016). The use of tRNS has been particularly advantageous in studying cortical excitability in motor learning and perceptual processing (Antal & Herrmann, 2016). Finally, tACS administers sinusoidal or biphasic currents at a chosen frequency to interact with the brain’s endogenous oscillatory activity, aimed at entraining large populations of neurons (Tavakoli & Yun, 2017). tACS has emerged as a promising neuromodulatory tool in the setting of cognition, neuroplasticity, and network dynamics (Ali, Sellers, & Fröhlich, 2013; Helfrich et al., 2014; Santarnecchi et al., 2016; Violante et al., 2017; Vossen, Gross, & Thut, 2015).

Figure 2. tES modalities and potential applications.

(A) The three major types of tES: tDCS, tRNS and tACS. Note, electrical current parameters, mechanism of action, and neural effects all differ across the three modalities, and in the case of tDCS stimulation can be used to either increase or decrease cortical excitability and/or specific brain oscillations. An example of a tACS montage aimed at synchronizing two brain networks (e.g. default-mode network – blue) is shown to highlight the potential of using multichannel tES to tap into network-level dynamics. (B) Some of the potential tES targets in ASD: the right temporoparietal junction (TPJ) and posterior superior temporal sulcus (pSTS) involved in Theory of Mind, social comprehension, and attention (Krall et al., 2015; Redcay, 2008; Santiesteban et al., 2012; Young et al., 2010); the right inferior frontal gyrus (IFG) involved in social impairments and communicative deficits (Bastiaansen et al., 2011; Dapretto et al., 2006; Lee et al., 2008; Shamay-Tsoory et al., 2009); and the dorsolateral prefrontal cortex (DLFPC) involved in comorbid depressive disorder and executive dysfunction (Just et al., 2004, 2007; Silk et al., 2006). (C-E) Three potential applications of inducing or modulating gamma oscillations. Specifically, stimulation can be used to modulate (C) spontaneous or (D) evoked gamma activity, measuring resting EEG before and after tES at gamma frequency (e.g., 40 Hz) delivered during resting state or a gamma-evoking task, respectively. The latter approach benefits from pre-activation of neural structures relevant to generating gamma oscillations, thus potentially amplifying the effect of tACS. Finally, (D) tES is delivered at gamma frequency concurrently with a gamma-inducing perceptual/cognitive task. (E) The same setup is used, but tACS is constantly informed by a closed-loop EEG-tES system triggered by, e.g., the phase of theta oscillations or spontaneous gamma bursts.

Of note, the induction effects of tES modalities are highly variable, with intra-individual and inter-individual responses varying widely in some studies (Chew, Ho, & Loo, 2015; Jacobson, Koslowsky, & Lavidor, 2012; Li, Uehara, & Hanakawa, 2015). As mentioned by Li and colleagues, (Li et al., 2015), it is likely that interindividual factors predicting tES induction effects, specifically tDCS, include baseline neurological state, neuroanatomy, age, and neuropathological states. To address these measured variabilities, multiple models have been proposed, including establishing a heuristic spatio-mechanistic framework (Yavari et al., 2017) and optimization and design of tES stimulation protocols (Opitz et al., 2016). Likely future investigations should investigate the value of personalized electrode montages, such as closed-loop tES/EEG stimulation paradigms.

While tDCS and tRNS have been effective in modulating cortical excitability and plasticity mechanisms, only tACS has been directly associated to frequency-specific modulation of oscillatory dynamics, with animal and human evidence of its effect on gamma activity. Therefore, while we consider the potential of each of the three major tES modalities for targeting gamma alterations in ASD, we note that tACS offers the greatest potential for neural oscillatory entrainment and modulation in humans to date. Thus, we will delve deeper into the theoretical and practical applications of tACS neuromodulatory approaches and how tACS can be utilized to target gamma-related ASD pathophysiology.

Transcranial Alternating Current Stimulation (tACS)

The development of non-invasive electrical stimulation techniques in the late 20^th century led to a subsequent interest in generating electric patterns capable of shaping network activity. Seminal in vitro work by Fröhlich and McCormick, (Fröhlich & McCormick, 2010), was among the first to demonstrate that active neocortical networks can be modulated by positive and negative electric feedback fields tuned at a specific frequency, suggesting a robust feed-back loop between neural activity and endogenous electric fields. These findings support the utility of tACS for delivering weak sinusoidal currents at biologically relevant frequencies that interact with the brain’s endogenous oscillations (Herrmann, Rach, Neuling, & Strüber, 2013; Vossen et al., 2015). tACS acts by synchronizing spiking activity to different driving frequencies, essentially entraining neuronal firing to the electrically applied field and exerting robust neuromodulatory effects at the systems level (Herrmann et al., 2013; Vossen et al., 2015). For these reasons, tACS has broad applications in human neuroscience research, with recent studies demonstrating the value of closed-loop tACS systems (Ketz, Jones, Bryant, Clark, & Pilly, 2018; Raco, Bauer, Tharsan, & Gharabaghi, 2016; Wilde, Bruder, Binder, Marshall, & Schweikard, 2015). For example, closed-loop techniques have been used to detect the onset of slow-wave sleep and to deliver tACS matching the frequency and the phase of the dominant oscillation during physiologically-relevant temporal windows (Ketz et al., 2018). Thus, closed-loop tACS systems could serve as a valuable electrophysiologic tool for measuring a desired frequency range and triggering stimulation matching the desired phase and frequency of the oscillations. Considering the heterogeneity of symptoms, differences in their severity, and, presumably, the inter-individual variability of their associated alterations in gamma activity among individuals with ASD, closed-loop tACS systems have the potential to individually tailor the gamma entrainment based on the paradigm (resting/active), task modality (visual/auditory/etc., if applicable), and the region/symptom of interest in certain ASD subpopulations (Figure 2C–2E).

Moreover, recent tACS applications have explored the possibility of modulating not just one oscillatory pattern but rather the coupling between multiple oscillations, e.g., cross-frequency coupling (CFC). CFC has been shown to be a key neural mechanism supporting complex brain functioning (Canolty et al., 2006; Canolty & Knight, 2010) and cognition (Aru et al., 2015; Chaieb et al., 2015; Gągol et al., 2018), suggesting the potential of tES techniques for modulating such interplay. Studies have shown the feasibility of simultaneous targeting of two frequencies both at rest and during a cognitive task, showing an advantage for combining multiple oscillatory targets (Alekseichuk, Turi, Amador de Lara, Antal, & Paulus, 2016). A form of CFC known as phase-amplitude coupling (PAC) indicates the modulation of amplitude of a higher-frequency signal by the phase of a concomitant lower-frequency neural signal, representing the integration of fast local circuits with slower long-range inputs. Both task-induced and resting-state alpha-to-gamma PAC are abnormally increased in ASD (Berman et al., 2015; Khan et al., 2013; Seymour, Rippon, Gooding-Williams, Schoffelen, & Kessler, 2019). Moreover, gamma-band PAC in animal models requires proper PV+ interneuron function (Port et al., 2019). Thus, tES modulation of gamma-related PAC may be an additional target of therapeutic investigation in ASD.

Gamma Induction and Modulation in Humans using tES

The gamma-frequency band has become a recent target of interest in tES modulation, with tACS being the most suitable modality for gamma-specific entrainment (Feurra et al., 2011; Kanai, Chaieb, Antal, Walsh, & Paulus, 2008; Reed & Cohen Kadosh, 2018; Santarnecchi et al., 2015). While the exact mechanisms of tACS-induced gamma entrainment remain unclear, its robust effects are thought to be mediated by spike timing-dependent plasticity (Dan & Poo, 2006) and driving the activity of local PV interneuron networks (Witkowski et al., 2016).

Studies investigating the human motor system were among the first to provide robust evidence for tACS-induced gamma modulation. Moisa and colleagues found gamma-tACS, but not beta-tACS or sham stimulation, of the primary motor cortex (M1) improved the velocity and acceleration of visually triggered movements (Moisa, Polania, Grueschow, & Ruff, 2016a). These gamma tACS-induced enhancements were significantly correlated with the changed blood oxygenation level-dependent (BOLD) activity in the stimulated M1 region (Moisa, Polania, Grueschow, & Ruff, 2016b). Guerra and colleagues, (Guerra, Bologna, et al., 2018), applied gamma and beta tACS stimulation to M1 in healthy controls during a repetitive finger tapping behavioral paradigm; gamma tACS (70 Hz) of M1 led to motor evoked potential (MEP) amplitude increment (i.e., progressive increase in amplitude) across the first ten movements of the finger tapping motor sequence, while beta tACS (20 Hz) led to amplitude decrement, with no significant effect on other movement parameters, nor any changes in MEPs (Guerra, Bologna, et al., 2018). Additionally, tACS targeting of M1 has been found to enhance force generation and motor reactivity in a visually guided movement-initiation task (Joundi, Jenkinson, Brittain, Aziz, & Brown, 2012). Miyaguchi et al., (Miyaguchi et al., 2018), demonstrated that M1-cerebellum gamma tACS applied during an isometric force task with a visuomotor control of the right index finger improved motor performance in healthy controls with poor motor performance at baseline. Similarly, enhanced motor performance during a visuo-motor coordination task has been demonstrated during high-gamma (80 Hz) stimulation (Santarnecchi et al., 2017).

tACS gamma-entrainment techniques also have been used to target higher-order behavioral processes. For example, Hoy and colleagues (Hoy, Bailey, Arnold, Windsor, et al., 2015) found selective improvement in working-memory performance following gamma-tACS, compared to tDCS and sham. Additionally, gamma-tACS applied to the right temporal lobe increased accuracy in a verbal insight task, as represented as a Eureka! moment (Santarnecchi et al., 2016), and was effective in improving abstract reasoning when applied over the left DLPFC (Emiliano Santarnecchi et al., 2013; Emiliano Santarnecchi, Rossi, & Rossi, 2015).These findings support the utility of tACS gamma-entrainment for improving cognitive and behavioral outcomes in humans.

There is strong evidence for the role of PV interneurons in endogenous gamma oscillations (Cardin et al., 2009; Sohal et al., 2009). Moreover, gamma activity has been found to drive plasticity throughout critical periods of development (Hensch & Bilimoria, 2012). Thus, enhancing the power and duration of endogenous gamma activity by tACS may enhance and strengthen these plasticity dynamics.

Recent studies have shown the specificity of gamma-tACS in modulating plasticity mechanisms in humans. For instance, Guerra et al., found tACS in the gamma band (but not in other frequency bands) enhanced the plasticity effects induced by intermittent theta-burst stimulation (iTBS) over the primary motor cortex, suggesting a possible link between gamma oscillations, interneuron activity, and long-term potentiation- (LTP) like plasticity (Guerra, Suppa, et al., 2018), a finding relevant in the context of LTP deficiency described in ASD models (Jung et al., 2013; Martin, Lassalle, Brown, & Manzoni, 2016). In a follow-up study, when continuous theta burst stimulation (cTBS) was paired with tACS to M1, cTBS-coupled with gamma tACS converted the expected after-effects of cTBS from inhibition to facilitation, while beta-tACS did not modify cTBS aftereffects (Guerra et al., 2019). Interestingly, this reversal of cTBS-induced plasticity effects by gamma-tACS suggests that gamma band activity may mediate long-term depression (LTD)-like plasticity in the human motor system (Guerra et al., 2019), which has been found to be abnormal in both children and adults with ASD (Jannati et al., 2020; Oberman et al., 2016). Additionally, Nowak and colleagues, (Nowak et al., 2017), utilized a combined tACS-TMS approach to drive gamma frequency oscillations, which resulted in an attenuation of GABA_A inhibition in the human motor cortex.

While the majority of tES gamma-entrainment studies have used the tACS modality, several studies suggest tDCS can also modulate gamma activity. In an MEG study, Wilson and colleagues, (Wilson, McDermott, Mills, Coolidge, & Heinrichs-Graham, 2018), found anodal tDCS applied to the occipital cortices enhanced gamma activity during a subsequent visual task in the right occipital, prefrontal and superior parietal cortices, compared to sham stimulation. Additionally, 2-mA tDCS applied to the prefrontal cortex was found to significantly increase gamma power and improve working memory in patients with SCZ compared to sham (Hoy, Bailey, Arnold, & Fitzgerald, 2015). While there is still a lack of consensus regarding optimal gamma-entrainment parameters for inducing robust and reproducible physiological and behavioral effects, the existing preliminary evidence suggests the utility of gamma-tACS, and possibly tDCS, for modulation of gamma-mediated functions and outcomes. The quest for detecting changes in gamma activity and their neurophysiological substrates should leverage multimodal imaging techniques, including, e.g., magnetic resonance spectroscopy (MRS) to look at GABAergic and glutamatergic function and assess E/I (Puts & Edden, 2012) or combined TMS-EEG for fine characterization of evoked responses indexing GABA_A and GABA_B activity (Farzan et al., 2010), or to understand the impact of tES on local and distributed dynamics (Romero Lauro et al., 2014).

tES and ASD: Opportunities and Challenges

Although current clinical interventions help alleviate various behavioral deficits of ASD, they are limited in their ability to affect the underlying pathophysiological mechanisms that give rise to the core symptoms of ASD. Therefore, targeting gamma band abnormalities in patients with ASD could serve as a non-invasive, non-pharmacological approach to address these shortcomings. Due to the heterogeneity of ASD populations, it is likely that tES gamma-modulation should target specific symptoms or characteristics of ASD patients rather than simply any individual with an ASD diagnosis. tES modalities, particularly frequency-specific tACS paradigms, can be investigated as therapeutic approaches to modulating gamma oscillations and restoring neurotypical gamma activity in ASD populations. While several pilot studies have explored the therapeutic potential of tES by means of tDCS, there is no robust initiative to explore tACS for gamma modulation in ASD (Kang et al., 2018; Rothärmel et al., 2019; Wilson et al., 2018). As such, tACS gamma-modulation techniques are a worthy area for future research and may serve as an integral part of a multi-modal clinical approach to ASD treatment.

Studies on gamma dysfunction in ASD tend to show reduced gamma power and frequency phase-locking across age groups, neural states, and sensory/perceptual tasks (Rojas & Wilson, 2014). While evidence suggests these patterns of gamma alterations in ASD are complex in nature (Rojas & Wilson, 2014), tACS gamma-modulation paradigms have the potential to target the relevant cortical areas through the use of appropriate stimulation paradigms. Previous studies have identified neuroanatomical correlates of ASD symptoms, which have largely focused on cortical regions implicated in social and executive functioning: (1) inferior frontal gyrus (IFG) mediating social impairments and communicative deficits including understanding emotion in others and mirror-neuron dysfunction (Bastiaansen et al., 2011; Dapretto et al., 2006; P. S. Lee et al., 2008; Shamay-Tsoory et al., 2009); (2) the right posterior superior temporal sulcus (pSTS) and temporoparietal junction (TPJ) mediating Theory of Mind, social comprehension, and attention (Krall et al., 2015; Redcay, 2008; Santiesteban et al., 2012; Young et al., 2010), indicating abnormality in the default-mode network (Padmanabhan et al., 2017), attributing intention to others (Costa et al., 2008), distinguishing ‘self-relevant’ from ‘other-relevant’ information (Heinisch et al., 2012), and predicting the behaviors of other individuals in social contexts (Carter et al., 2012); and (3) dorsolateral prefrontal cortex (DLFPC) mediating comorbid depressive disorder and executive dysfunction (Just et al., 2007, 2004; Silk, Rinehart, Bradshaw D Sc, et al., 2006). Moreover, additional evidence related to alterations in network-level functioning, including atypical white matter connectivity, most prominently characterized by impaired development of limbic connectivity and altered maturation of short-distance tracts (Ameis & Catani, 2015; Shukla et al., 2011), and aberrant cerebellar connectivity, which have been implicated in language development deficits and restrictive/repetitive behaviors (Stoodley et al., 2017; Verly et al., 2014), have been proposed in ASD populations. Finally, functional MRI network abnormalities in ASD populations have been identified as well, typically involving frontostriatal networks (Silk, Rinehart, Bradshaw, et al., 2006), the default-mode network (Cherkassky et al., 2006; Kennedy & Courchesne, 2008) and dorsal and ventral attention networks (Farrant & Uddin, 2016; Fitzgerald et al., 2015).

Based on these results, potential stimulation targets that can be explored include frontal (IFG, DLPFC), temporal, and parietal (pSTS, TPJ) regions to investigate improvements in social/communicative deficits, executive dysfunction and comorbid depressive disorder, as well as Theory of Mind, social comprehension, and attention, respectively (Figure 2B). These regions have shown generally consistent reproducible modulatory effects across spontaneous, induced, and evoked gamma activity (Table 1). Additionally, since the majority of studies find an overall reduction of gamma activity in ASD, applying gamma-tACS to these regions might be able to entrain and restore, at least partially, neurotypical gamma activity in the ASD population (Table 1). However, future studies must establish consensus tACS stimulation paradigms, including electrode montages, current amplitude, and perturbation-based/resting-state gamma protocols (Brennan, Peck, & Holodny, 2016; Chen, 2001).

If gamma entrainment is a valid objective in tES therapeutic strategies in ASD, then the implementation of tES in clinical settings will require prospective studies to demonstrate tACS is capable of modulating gamma in a heterogeneous patient population. Additionally, in a stepwise approach to tES deployment for management of core ASD symptoms, pilot studies can test whether isolated behavioral measures may improve with gamma entrainment. For example, future investigations could access the neuromodulatory effects of gamma tACS on motor learning in ASD, which is typically impaired in ASD rodent models and in ASD patients (Izadi-Najafabadi, Mirzakhani-Araghi, Miri-Lavasani, Nejati, & Pashazadeh-Azari, 2015; Marko et al., 2015; Piochon et al., 2014). Gamma oscillations may play an integral role in mediating motor learning mechanisms and gamma-tACS significantly enhances these learning processes (Guerra, Bologna, et al., 2018; Miyaguchi et al., 2018). In neurotypical subjects, 70 Hz tACS significantly increased the capacity for motor learning, compared to 10 Hz, 20 Hz, and sham stimulation (Sugata et al., 2018). Similar findings in healthy participants suggest low gamma-tACS applied to M1 improves implicit performance on a motor learning task (Giustiniani et al., 2019) and gamma-tACS improves acceleration of the practiced movement during motor task training (Bologna et al., 2019; E. Santarnecchi et al., 2017b). Interestingly, theta-gamma coupled tACS to right M1 in healthy participants enhanced motor skill acquisition, compared to sham and an active stimulation control, in a ballistic thumb abduction task of the left hand (Akkad et al., 2019). These data suggest gamma tACS can also be applied to M1 and measured in motor learning task performance in ASD populations, to probe for the extent of responsivity to these neuromodulatory effects in motor learning and motor control.

From a biomarker perspective, Nowak and colleagues, (Nowak et al., 2017), found the neurophysiological response to application of gamma-tACS to M1 in heathy controls is closely related to the ability to learn a motor task. This data suggests that probing M1 with gamma-tACS can serve as a tool to predict “responders” from “non-responders” in motor learning paradigms, although these robust findings in healthy participants would have to be studied and reproduced in ASD populations. Importantly, personalization of stimulation protocols/treatment is key, especially when targeting brain oscillatory activity which can differ due to state- and trait-dependent variability in individuals with ASD who also exhibit intrinsic individual variability due to developmental trajectories and clinical phenotypes. In this regard, our team has recently promoted the validation of so-called perturbation-based biomarkers (PBB), leveraging the combination of EEG recording and brain stimulation techniques (TMS, tES) to quantify brain responses to perturbation. These biomarkers have demonstrated high reliability and also a potential for “fingerprinting”, i.e. characterize each individual brain on the basis of patterns of activity propagation after, e.g., a TMS pulse (Ozdemir et al., 2020). Even more interestingly, this approach has been tested for targeting specific brain resting-state networks (e.g. default-mode network, dorsal attention network), showing the ability to capture network-specific responses that could be indeed used to tailor personalized interventions in patients with ASD. The same approach could be used to combine tES and EEG, leading to a portable and cheaper equivalent of TMS-EEG potentially deployable at home. All these methods could be used to characterize fine brain dynamics in the frequency domain, including gamma activity which is captured by TMS-EEG as the earliest local GABAergic evoked response after a TMS pulse (Ferrarelli et al., 2008). Overall, and beyond PBB, improved operational definition and understanding of gamma activity in ASD is a prerequisite for further biomarker discovery or therapeutic trials.

In the current commentary, we have argued for the application of gamma tACS entrainment in ASD. However, we acknowledge that gamma is not the only endogenous oscillatory frequency band that could serve as a suitable target for tACS neuromodulation in ASD. Disrupted oscillatory activity and synchronization across various frequency bands is recognized as a neurophysiological characteristic of ASD, most notably relating to sensory and perceptual functioning (Simon & Wallace, 2016); these altered frequencies include alpha, beta and theta activity. Preliminary studies investigating sensory processing in ASD suggest abnormalities in stimulus induced alpha activity (Spaak, de Lange, & Jensen, 2014), as well as reduction in the ability to phase lock alpha and beta oscillations to repetitive sensory stimulation (V. V. Lazarev, Pontes, Mitrofanov, & deAzevedo, 2010; Vladimir V. Lazarev, Pontes, & deAzevedo, 2009a). Additionally, Jochaut and colleagues utilized an EEG and fMRI study assessing cortical speech processing, demonstrating that gamma and theta cortical activity do not engage synergistically in response to speech in ASD subjects, compared to healthy controls (Jochaut et al., 2015). Future investigations should consider possible applications of targeting other frequency bands in ASD.

While tES-induced gamma modulation could serve as a useful therapeutic approach for ASD, this emerging technology faces various challenges. First, patients with ASD exhibit a wide range of capabilities in language expression, physical abilities, and intellectual functioning. Because many patients with ASD are typically dependent on care providers, any novel technique should be carefully assessed to ensure the well-being and safety of patients is upheld and prioritized. Therefore, it is important that investigations involving tES and ASD participants be aware of the special accommodations of the ASD community, as well the neuroethical issues this technology may present (Sarrett, 2016). While tES techniques have undergone robust developments in their current delivery and temporal precision, progress in refining the technology is still needed, especially with the closed-loop systems that deliver stimulation upon detection of set oscillatory thresholds. Of note, oscillatory gamma-band measurements in resting and active states are challenging and often criticized; EEG-recorded gamma activity is often contaminated by electromyographic activity and eye-movement artifacts (Yuval-Greenberg, Tomer, Keren, Nelken, & Deouell, 2008), making it difficult to isolate neural gamma activity from noise. However, this issue could potentially be circumvented by taking advantage of studies on patients with ASD and epilepsy who undergo evaluation for epilepsy surgery and have intracranial electrodes in place (Kokoszka et al., 2017). Another method to optimize investigating pathophysiological signatures of gamma activity would be to employ MEG, in which cerebral gamma activity can be more clearly and robustly identified (Mandal, Banerjee, Tripathi, & Sharma, 2018). Future investigations aimed at improving the temporal and spatial parameters of stimulation paradigms, as well as capturing the physiological effects of the stimulation, are needed for further development of this novel therapeutic application.

Conclusion

In summary, we present evidence suggesting that gamma-frequency alterations can serve as a potential biomarker of ASD pathophysiology and as a target for noninvasive neuromodulatory interventions currently under investigation for other neurological and psychiatric conditions. In particular, we argue that tACS, among other tES techniques, could be utilized to modulate and restore neurotypical gamma-band activity in patients with ASD.