Blockade of glutamate N-methyl-D-aspartate receptors (NMDAR) by a subanesthetic dose of ketamine in humans transiently induces negative, positive and cognitive symptoms similar to those found in schizophrenia (SZ). Together with findings in animals of behavioral, biochemical, and electrophysiological alterations with NMDAR antagonists, convergent genetic, molecular and neuroimaging evidence points to NMDAR hypofunction in SZ (Javitt, 2012; Javitt et al., 2012). This supports use of the ketamine challenge as a pharmacological approach to generating insights into SZ symptoms and their underlying pathophysiologic processes (Frohlich and Van Horn, 2014).

Oscillatory brain activity is considered a fundamental process in the temporal coordination of circuits linked to perceptual, cognitive, and emotional functions. This brain activity, as measured by the amplitude/power of electroencephalographic (EEG) rhythms, is increasingly seen as an indicator of intrinsic brain function when assessed during stimulation/task-free states (Spencer, 2014). Further, this activity has been found to be altered in SZ and to vary with symptomatology (Boutros et al., 2008, 2014; Moran and Hong, 2011; Galderisi et al., 2014). Aberrant “resting-state” spectral EEG profiles in psychosis have invariably shown increased activity in low frequency (delta, theta) EEG and magnetoencephalographic (MEG) rhythms (Fehr et al., 2003; Boutros et al., 2008; Venables et al., 2009; Uhlhaas and Singer, 2010; Moran and Hong, 2011; Ranlund et al., 2014) which have been found specifically evidenced in chronic SZ patients (Sponheim et al., 1994, 2000; Narayanan et al., 2014), their first-degree relatives (Narayanan et al., 2014) and first-episode SZ patients (Sponheim et al., 2000), but not in individuals at-risk for psychosis (Ranlund et al., 2014). Spontaneous alpha activity, which predominates in healthy individual’s EEG, is significantly diminished in chronic SZ patients (though unaffected in some studies), with varying findings that depend on scalp region, and whether slower or faster frequencies are examined within this band (Sponheim et al., 1994, 2000; Knyazeva et al., 2008; Hong et al., 2012; Narayanan et al., 2014; Goldstein et al., 2015; Kim et al., 2015). Similar inconsistencies are evident with beta rhythms, with chronic patients showing either no abnormalities (Wada et al., 1994; Kam et al., 2013; Kim et al., 2015) or increased activity (Sponheim et al., 1994; Venables et al., 2009; Narayanan et al., 2014). As with delta/theta findings (Galderisi et al., 2009), these inconsistencies can be partially attributed to the effects of chronicity, medication and patient symptomatology (Czobor and Volovka, 1992; Merlo et al., 1998; Knott et al., 2001; Cerdán et al., 2005; Venables et al., 2009; Moran and Hong, 2011). The recent interest in resting gamma activity, increased in some (Czobor and Volovka, 1992; Merlo et al., 1998; Knott et al., 2001; Cerdán et al., 2005; Kam et al., 2013; Kocsis et al., 2013; Tikka et al., 2013, 2014, 2015) but not all EEG or MEG studies (Sponheim et al., 1994) of SZ patients (Rutter et al., 2009; Fuggetta et al., 2014; Hirano et al., 2015) is of particular interest both because of its purported role in feature binding and coordination of local neuronal populations, and because these rhythms are in part dependent on the activity of Parvalbumin-containing GABAergic cortical inhibitory interneurons, which are diminished in SZ patients and subject to NMDAR modulation from excitatory pyramidal cells (Lee et al., 2003; Herrmann and Dermirlap, 2005; Sun et al., 2011; Gandal et al., 2012; Lewis et al., 2012; Cohen et al., 2015). Integrative brain functions, however, are modulated through multiple oscillatory processes in different frequency bands, each with differing neuropharmacological sensitivities. As such, the study of the full frequency spectrum and its pharmacological basis is critical for deciphering the complex electrophysiologic abnormalities in SZ patients (Moran and Hong, 2011).

Electrophysiological recordings in animals with NMDAR antagonists such as ketamine and MK-801 recreate some of the background spontaneous EEG features of SZ (Hunt and Kasicki, 2013). Consistent with patient findings, increases in delta are dose-dependent and evident in cortical and subcortical (hippocampus, thalamus) regions. Theta frequency activity also shows these dose related changes, with power in this band consistently increased in cortex but reduced in the hippocampus. Ketamine’s effects on these rhythms in rodents, although generally augmenting beta and gamma by its actions on PV-interneurons in multiple cortical regions (Pinault, 2008; Carlén et al., 2012), vary depending on dose, acute vs chronic treatments, subcortical region and within-band (slow vs. fast) frequencies (Pinault, 2008; Roopun et al., 2008; Hakami et al., 2009; Carlén et al., 2012; Kittelberger et al., 2012; Hunt and Kasicki, 2013; Kocsis et al., 2013). In humans administered relatively high (anesthetic) doses of ketamine (~2 mg/kg), activity is increased in theta and beta bands and decreased in alpha (Schultz et al., 1990; Bojak et al., 2013). Sub-anesthetic psychotomimetic bolus doses of ketamine (0.2–0.5 mg/kg) have also increased theta and decreased alpha in baseline (pre-stimulus) EEGs (Kochs et al., 1996). In the only two human resting-state EEG studies to date, reductions in delta, alpha, and beta (gamma not assessed) were found with an ultra-low ketamine dose (Knott et al., 2006), and increases in gamma (other bands not assessed) were observed with a psychotogenic dose, but relationships with induced dissociative or psychotic-like symptoms were not examined (Sanacora et al., 2014). Of the two human resting MEG studies, subanesthetic ketamine increased gamma in both while reducing either alpha (Muthukumaraswamy et al., 2015) or beta (Rivolta et al., 2015).

Ketamine-modulated neuroelectric activity in rodent models suggest complex-spatiotemporal effects, which vary with frequency band and cortical area (Lee et al., 2003; Herrmann and Dermirlap, 2005; Pinault, 2008; Roopun et al., 2008; Moran and Hong, 2011; Carlén et al., 2012; Gandal et al., 2012; Lewis et al., 2012; Hunt and Kasicki, 2013; Cohen et al., 2015). Similar systematic EEG studies on healthy humans of the effects of subanesthetic, psychotomimetic doses of ketamine on the generation and cortical distribution of intrinsic brain rhythms across the frequency spectrum have not been conducted. Thus, one of our objectives has been to profile the acute effects of ketamine on the generation and topography of sensor-level EEG in low and high frequencies using reference-free current source density (CSD) measures (Nunez and Srinivasan, 2006). Such reference-free measures avoid problems associated with reference-dependent EEG indices (Kayser and Teake, 2010); reference-independent CSD measures reduce volume conduction from distal sites, sharpen spatial resolution, and are more closely related to neuronal activity and the strength of underlying current generators (Nicholson, 1973; Nunez and Srinivasan, 2006; Kayser and Teake, 2010).

A second objective has been the evaluation of the effects of NMDAR blockade on source-level, electric activity within Nekovarova et al.’s (2014) recent triple network (TN) model of SZ psychopathology. Largely investigated with fMRI (Fox and Raichle, 2007), but also assessed with PET (Raichle et al., 2001) and EEG (Chen et al., 2008; Yuan et al., 2012), brain regions have been shown to be dynamically organized into functional networks of intercorrelated areas (or “nodes”) that act together to perform specific tasks (Bressler, 1995). Of the full repertoire of brain networks, all of which are continuously and dynamically “active” not only during cognition but even when at “rest” (Smith et al., 2009), during sleep (Fukunaga et al., 2006) and under anesthesia (Vincent et al., 2007), there are three large-scale domain-specific networks that make up the model which exhibits abnormalities in SZ (Bressler and Menon, 2010) and depression (Mulders et al., 2015): the default mode network (DMN), the central executive network (CEN), and the salience network (SN). This TN model relies on accumulating neuroimaging evidence in healthy humans indicating that SN, which is involved in the orientating of attention to the most homeostatically relevant (salient) of ongoing extrapersonal (sensory) and intrapersonal (limbic driven) events (Bressler and Menon, 2010), causally influences anticorrelated activation of DMN and CEN. Clinical and cognitive symptoms of SZ are in part attributed to aberrant switching by the SN between internal processes (i.e., self-referential, autobiographical functions) supported by the DMN and external processes (i.e., attention and processing of exogenous stimuli) supported by the CEN as a result of a dysfunctional SN, a system that enables the switch between various dynamic brain states (Sridharan et al., 2008; Palaniyappan et al., 2012).

Adopting a region of interest (ROI) approach with exact low resolution brain electromagnetic tomography (i.e., eLORETA), which is an electrophysiological neuroimaging method that allows a reliable source localization of surface level electrical signals (Pascual-Marqui, 2011; Pascual-Marqui et al., 2011), we assessed ketamine-induced changes in two key regional nodes anchoring each of the three large-scale networks: DMN [ventromedial prefrontal cortex (VMPFC) and posterior cingulate cortex (PCC)]; CEN [dorsolateral prefrontal cortex (DLPFC) and posterior parietal cortex (PPC)]; and SN [anterior cingulate cortex (ACC) and anterior insula (AI)]. Based on limited human EEG studies with NMDAR antagonists, we hypothesized that relative to placebo, ketamine administration in healthy humans would produce SZ-like scalp surface CSD changes (↑ delta, beta and gamma, and ↓ alpha). Ketamine has been shown to alter resting SN and DMN function (Bonhomme et al., 2016) and given the pivotal role of the SN in the TN model and the available neuroimaging evidence indicating a primary role of dopamine in the interaction of the SN with subcortical sites, whilst the within-network activity of the SN and its interaction with other large-scale networks is thought to predominantly depend on glutamate/gamma-aminobutyric acid (GABA) neurotransmission (Palaniyappan et al., 2012), we expected ketamine to result in band-dependent CSD alterations within the cortical networks. We further hypothesized that these sensor- and source-level EEG changes would be accompanied by dissociative symptoms characteristic of SZ.

Although there are plenty of single blind EEG/MEG studies, this is the first randomized, placebo-controlled, double-blind crossover study of healthy humans administered a sub-anesthetic, psychotomimetic dose of ketamine that demonstrates changes in resting-state EEG across the frequency spectrum. The work builds on extensive animal research on the electrophysiological effects of ketamine and well-studied EEG correlates of anesthetic doses of ketamine. As key findings, we report CSD alterations in low and high frequency rhythms analyzed from scalp EEG recordings and these were accompanied by similar current density changes in source-localized cortical ROIs within the three large-scale networks of the schizophrenia model, namely the DMN, CEN and SN. These electrocortical effects of ketamine were paralleled by changes in dissociative states, the degree of which was significantly correlated with scalp and SZ network alpha current density changes.

N-methyl-D-aspartate receptor antagonism in preclinical models has most frequently been associated with increases in low (delta/theta) and high (beta/gamma) frequencies (Hunt and Kasicki, 2013). These in vivo electrocortical findings with NMDA receptor blockade, paralleled to some extent by oscillatory changes in in vitro (Kim et al., 2015) recordings and intracortical networks (Hakami et al., 2009), vary considerably, particularly with respect to theta rhythms, which have also been shown to be reduced both with acute and chronic ketamine treatment (Kittelberger et al., 2012). Our EEG observations in healthy volunteers of increased gamma and reduced non-gamma activity with a ketamine dose typically used to model SZ in part confirm previous independent study findings of EEG alpha and gamma changes seen with anesthetic and sub-anesthetic dosing in humans (Schultz et al., 1990; Kochs et al., 1996; Knott et al., 2006; Bojak et al., 2013; Sanacora et al., 2014) and they correspond with ketamine-induced alpha/beta and gamma alterations in recent resting MEG studies (Muthukumaraswamy et al., 2015; Rivolta et al., 2015). These emerging electrocortical patterns with NMDA receptor blockade seen in our study and in other investigations are reminiscent of some of the reports of aberrant EEG activity in SZ. Deficits in resting spontaneous alpha band activity reported across the clinical course of the disease (chronic SZ, first episode psychosis, prodromal SZ, relatives of SZ probrands) (Boutros et al., 2008; Goldstein et al., 2015; Kim et al., 2015) were mimicked in our EEG recordings across scalp regions following NMDA antagonist treatment. They were also seen in recent MEG recordings (Muthukumaraswamy et al., 2015), and were concurrent with diffuse increments in gamma also described in some recent resting state studies of SZ (Kam et al., 2013; Kocsis et al., 2013; Tikka et al., 2013, 2014, 2015). These parallel observations with alpha and gamma spectral frequencies contrast with the response of delta-theta and beta activities to NMDA receptor blockade as the reduction in current density with ketamine, particularly in the low frequencies, is contrary to the observation that increased local and global EEG delta/theta is one of the more consistent findings reported in unmedicated, first episode, and chronic SZ patients (Boutros et al., 2008; Moran and Hong, 2011; Kim et al., 2015).

Thus, acute NMDA receptor hypofunction may mediate some of the EEG disturbances associated with psychosis, particularly those relating to high frequency gamma perturbations, which in this study likely result from ketamine inhibition of NMDA receptors on GABA interneurons and disinhibition of pyramidal neurons (Grunze et al., 1996; Maccaferri and Dingledine, 2002; Jackson et al., 2004; Homayoun and Moghaddam, 2007) subpopulations, some of which act to increase cortical gamma during inhibition (Lovett-Barron et al., 2012; Royer et al., 2012). Considering ketamine’s multitude of effects, with actions at muscarinic, opioid, and adrenergic receptors and actions blocking serotonin and norepinephrine transporters (Bergman, 1999; Chen et al., 2009; Stahl, 2013), any one or more of these mechanisms [and/or the purported increase in non-NMDA glutamate (i.e., AMPA and kainate) receptor neurotransmission resulting from release of GABA restraint during impaired NMDA receptor function (Olney and Farber, 1995; Moghaddam et al., 1997)] may underly changes in non-gamma rhythms seen in SZ and with acute NMDA receptor blockade. Given that oxidative stress is thought to play a crucial role in several brain disorders including psychosis (Sorce and Krause, 2009; Schiavone et al., 2012, 2016), is involved in the modulation of neuronal activity (Infanger et al., 2006), and is increased with acute subanesthetic doses of ketamine and controls the resultant prefrontal glutamate release accompanying NMDA receptor antagonism (Sorce et al., 2010), brain oxidative stress may be one novel mechanism underlying the SZ-like electrophysiological changes with ketamine administration.

Contributing cortical sources to our observed ketamine modulated EEG profile were examined within the TN model of SZ. Although there is increasing interest in the pathophysiology of brain networks in SZ, to our knowledge this is the first human study to explore CSD in nodes of three well defined large-scale neural networks during acute NMDA receptor hypofunction. Directional changes in surface-level EEG current density induced by ketamine within each frequency band were mirrored by CSD alterations across the networks and, whereas only with alpha CSD were all three networks shown to be affected by NMDA receptor blockade, nodes of at least one or two of the networks evidenced ketamine-induced changes in delta (DMN), theta (DMN, CEN), beta (SN), and gamma (DMN, SN) rhythms. Activity of individual high and low frequencies is associated with both overlapped and distinct sensory and cognitive functions but given the general observation that reductions in low frequency and increases in high frequency rhythms are associated with increased arousal and behavioral activation (Moran and Hong, 2011), our frequency specific oscillatory changes with ketamine generally suggest that affected nodes within the respective large scale networks are hyperactivated during NMDA antagonist treatment. The exception is seen with the ACC, which showed alpha reductions and gamma increases (activation) along with decreases in beta (suppression) during ketamine administration. This general pattern of cortical activation during waking states is very similar to that reported in healthy volunteer studies utilizing fMRI and PET to image resting cerebral blood flow and brain metabolic response to ketamine’s blockade of NMDA receptors (Lahti et al., 1995; Breier et al., 1997; Vollenweider et al., 1997; Deakin et al., 2008; De Simoni et al., 2013; Doyle et al., 2013; Scherbinin et al., 2015). Although there are discrepancies in these neuroimaging studies, ketamine has typically activated frontal and temporal lobe structures as well as the AI and PCC but has suppressed the VMPFC while exhibiting mixed effects on the ACC, which has been found to be either hyperactive or hypoactive with acute antagonist treatment.

Accumulating neuroimaging evidence in healthy individuals indicates that an imbalance between the normally anti-correlated DMN (internal, self-related processing) and CEN (external environment/task-related processing) may underlie clinical and cognitive features of various psychiatric disorders (Menon, 2011), including SZ (Northoff, 2015). Although aberrant glutamatergic signaling in either the DMN or CEN may disrupt this balance, the AI-ACC nodes of the SN network, and the AI in particular, play a critical and causal role in the switch between activation and deactivation of the two large-scale brain networks (Sridharan et al., 2008). Structural and functional abnormalities across the different stages of SZ occur in the nodes of SN (Palaniyappan et al., 2012) and dopaminergic dysfunction in the SN is viewed as having a central role in aberrant motivational salience (Kapur, 2003) and clinical symptoms (Palaniyappan and Liddle, 2012) in individuals with psychosis. GABA/glutamate coupling is also thought to mediate SN functions and subanesthetic doses of ketamine have been specifically shown to increase glutamate turnover in the ACC, consistent with studies in SZ in which glutamatergic levels are elevated in the ACC of early-stage, drug naïve or drug-free patients (Merritt et al., 2013; Poels et al., 2014).

Our findings of ketamine-induced increases in gamma in the ACC and AI nodes, likely resulting from reduced excitation of GABAergic interneurons and the subsequent disinhibition of glutamatergic pyramidal neurons, may represent elevated noise at pyramidal cell assemblies (Moran et al., 2015), which may act to reduce high frequency signal-to-noise ratio in response to sensory input (Saunders et al., 2012) and disrupt information processing in these networks. Given that activity in the SN typically precedes and predicts activity in both DMN and CEN (Sridharan et al., 2008), implying that the SN coordinates multi-networking activity, then alterations in gamma in the SN may act to decrease the inherent anti-correlation between the DMN and CEN, as suggested by the co-activation of these networks, which is indexed by reductions in delta/theta and alpha activity. Decreases in anti-correlation between the DMN and CEN are thought to diminish the boundaries or distinction between internally and externally oriented cognitions, confusing internal and external mental contents and resulting in “self-environment blurring” (Vollenweider et al., 1997) – thus providing a mechanistic path that may explain several core symptoms of SZ.

Oscillations generated by thalamocortical circuits are thought to be responsible for the synchronization of neural activity between different cortical regions and, depending on the frequency range of the most prominent oscillation, are associated with the appearance of specific mental states. Generally, thalamocortically generated gamma is thought to be a potential explanation for coherence of perception in the brain while alpha serves to route information to downstream regions by inhibiting neuronal processing in task-irrelevant regions (Jensen and Mazaheri, 2010; Foxe and Snyder, 2011; Klimesch, 2012), thus allowing task-relevant regions to communicate. Accordingly, the reduction in alpha and increase in gamma seen in the SN with acute ketamine may be associated with reduced salience processing capabilities (due to impaired sensory/perceptual functions) and diminished registration of salience below the threshold that is needed to allow switching between the triple networks.

Study analysis of the subjective response to NMDA antagonist treatment assessed with CADSS showed significant dissociative effects, with ketamine infusion resulting in marked increases in amnesia, depersonalization, derealisation, and objective rater subscale scores. These dissociative phenomena, which have been consistently reported in previous ketamine studies with healthy volunteers and particularly in males (Morgan et al., 2006), are part of a spectrum of transient SZ-like psychoactive effects (including positive and negative symptoms) produced with acute subanesthetic ketamine (Krystal et al., 1994; Malhotra et al., 1996; Adler et al., 1998) and can be attenuated by inhibiting the reuptake glycine, a co-agonist at the strychnine-insensitive glycine-β site on the NMDA receptor (D’Souza et al., 2012).

Of the perceptual effects we observed with ketamine infusion, only changes in depersonalization scores were related to the EEG alterations. Increases in depersonalization ratings were associated with reductions in alpha current density across the scalp and in specific nodes of the DMN (bilateral PCC), CEN (right PPC), and SN (right AI). Altered self-other boundary, experienced as depersonalization (feeling detached as if one’s body is unreal), is one of several so-called “basic symptoms” of prodromal SZ (Maggini et al., 2002; Raballo, 2012). There is considerable overlap in midline cortical structures comprising the neural network implicated in self-specificity with those showing high resting state in the DMN, and while evidence suggests that the pregenual ACC is specifically involved in self-processing (Qin and Northoff, 2011), several neuroimaging studies have also pointed to the AI as being the representational cortex for the sense of self (Craig, 2002, 2009). Alpha rhythms are considered among the most important building blocks for functioning, association and communication in the brain (Basar, 2012; Basar and Guntekin, 2012; Bazanova and Vernon, 2014). Alpha is prominent in the EEG DMN profile (Chen et al., 2008), overlaps with resting-state networks identified in fMRI (Jann et al., 2007; Mantini et al., 2007), and in addition to being the main neural oscillation of self-agency (a person’s feeling that his action is generated by himself; Kang et al., 2015), correlates with both perception of self-related stimuli and resting state glutamate concentration in the pregenual ACC (Bai et al., 2016). Given this context, our findings with ketamine suggest that glutamatergic modulated alpha activity may be one mechanism underlying pathological self-processing in SZ and may serve as a viable target for novel treatment interventions.

In summary, resting-state EEG appears to be a useful and efficient method for investigating the neuropharmacology of altered brain rhythms suspected to underlie perceptual/dissociative symptoms of SZ. In the present study, sensor-level and surface projected neural network electrocortical activity was investigated in a NMDA receptor hypofunction model of SZ and intravenous ketamine was shown to partly mimic the aberrant EEG activity observed in SZ patients and to alter oscillations in large-scale resting-state networks implicated in psychosis. Neuroelectric changes correlated with the severity of dissociative symptoms induced by NMDA receptor antagonist treatment. Together, these findings provide additional information on how the modulation of the glutamatergic system may regulate brain electric activity, supporting its potential utility both as a biomarker of NMDA receptor dysfunction and a possible target for novel treatments in psychiatric disorders such as SZ which involve glutamatergic deficits.

Each of the authors participated in this research by contributing to the conception and design of the project (VK), participant screening (SdS, JC, DS, HB, and VI), performance of the experiment (SdS, JC, DS, HB, and JM), electrophysiology and statistical analysis (SdS), and interpretation (VK and SdS).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.