Impact of Ketamine on Neuronal Network Dynamics: Translational Modeling of Schizophrenia‐Relevant Deficits

Summary

Subanesthetic doses of the psychomimetic, ketamine, have been used for many years to elicit behavioral effects reminiscent of schizophrenia in both healthy humans and in animal models of the disease. More recently, there has been a move toward the use of simple neurophysiological measures (event‐related potentials, brain oscillations) to assay the functional integrity of neuronal circuits in schizophrenia as these measures can be assessed in patients, healthy controls, intact animals, and even in brain slices. Furthermore, alterations of these measures are correlated with basic information processing deficits that are now considered central to the disease. Thus, here we review recent studies that determine the effect of ketamine on these measures and discuss to what extent they recapitulate findings in patients with schizophrenia. In particular, we examine methodological differences between human and animal studies and compare in vivo and in vitro effects of ketamine. Ketamine acts on multiple cortical and subcortical sites, as well as on receptors other than the N‐methyl‐d‐aspartate receptor. Acute ketamine models' changes correlated with psychotic states (e.g. increased baseline gamma‐band oscillations), whereas chronic ketamine causes cortical circuit changes and neurophysiological deficits (e.g. impaired event‐related gamma‐band oscillations) correlated with cognitive impairments in schizophrenia.

Contemporary views of schizophrenia have shifted from alterations in dopamine neurotransmission to impaired information processing. Cognitive impairment is regarded as the primary core deficit, resulting from dysfunction of neuronal microcircuits. The significance for pharmacology is an extension of potential therapeutic targets from neuromodulatory systems, that is, from the original D2 and the more recent combinations of aminergic receptor antagonists to neurotransmitters operating on the level of these microcicuits, that is, glutamate and GABA. Research focusing on the primary transmitters involved in neuronal network dynamics is conducted in the framework of the N‐methyl‐d‐aspartate (NMDA) receptor hypofunction hypothesis. This is one of the most powerful current models of schizophrenia, with a strong translational potential from rodents to primates to healthy subjects to patients with schizophrenia. This hypothesis originated from early observations of ketamine and other NMDA receptor (NMDAR) antagonists' ability to elicit (in healthy subjects 1) or exacerbate (in schizophrenics 2) positive psychotic symptoms. However, recent studies emphasize its validity to also explain cognitive deficits and negative symptoms 3, 4, 5, which has traditionally been the weak point of prior, mostly dopamine‐based, models. Positive, psychotic‐like symptoms elicited by ketamine (S‐Ketamine, in particular), the basis for its recreational abuse, are discussed in Chapter 10 of this Special Issue. There are also excellent recent reviews of behavioral and electrophysiological studies addressing cognitive symptoms and altered neuronal network dynamics in schizophrenia using different variants of the NMDAR antagonist model 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19.

This review will be limited to a critical comparison of the electrophysiological changes induced by ketamine and other NMDAR antagonists, for example, MK‐801 and phencyclidine, with those observed in patients with schizophrenia and animal models, which can help to understand the processes involved in cognitive impairment in schizophrenia. In the past several years, there has been a rapidly growing interest in electrophysiological markers of schizophrenia fueled by major technological advancements in clinical electrophysiology and by fundamental changes in understanding the function of cortical networks and in the interpretation of EEG. This trend is expected to continue, and ketamine, which can be used both in human and animals, may play an essential role in developing the right methods and standards for detecting schizophrenia‐relevant cortical dysfunction.

The Use of Ketamine as a Model of Schizophrenia‐Associated Neurophysiological Abnormalities

N‐methyl‐d‐aspartate receptor (NMDAR) is an ionotropic glutamatergic receptor widely distributed in the brain from spinal cord to the cortex and implicated in numerous functions from pain to learning and memory. It is a hetero‐oligomeric complex consisting primarily of two NR1 and two of several types of NR2 subunits (NR2A‐NR2D). NMDAR activation requires not only binding of glutamate but also a coagonist glycine and a voltage‐dependent removal of the Mg²⁺ ion blocking the channel at rest. There are numerous antagonists blocking the glutamate or the glycine binding sites, the ion pore, or allosteric binding sites. Ketamine belongs to this latter group of noncompetitive antagonists 20.

Acute administration of NMDAR antagonists to rodents impacts various CNS functions, from simple motor activity to high‐level cognitive tasks 21. In nonhuman primates, NMDAR antagonists evoke sensory, behavioral, and cognitive disturbances 22, 23, 24 similar to responses seen in humans 1. The dose of ketamine to induce schizophrenia‐relevant deficits is below the anesthetic dose, that is, 5–10 mg/kg subcutaneous injections 25, 26, 27 are used in rats and 0.3–0.5 mg/kg intravenous injections in human studies 1, 28, compared with 70–80 mg/kg and 1.0–4.5 mg/kg (2 mg/kg for 5–10 min anesthesia) used for anesthesia in rodents and human, respectively. It is worth noting that the ketamine dose found effective for treatment of depression is also 0.5 mg/kg i/v 29, 30. Thus, there are efforts underway to find a way to attenuate the ketamine‐induced psychotomimetic effect for the treatment of depression 31, 32, 33, 34.

The limitations of behavioral models of schizophrenia have been recognized, and it has been argued that animal modeling of inherently human disorders, like schizophrenia, might be impossible 18. Recently, an alternative approach emerged by recreating endophenotypes contributing to schizophrenia in preclinical animal models 9, 12, 35, 36, 37, 38, 39, 40, 41, 42, 43. Endophenotypes are heritable, state‐independent biomarkers associated and cosegregating with the illness 39. They allow reducing complex psychiatric phenotypes into components that can be modeled in animals and make genetic and neurobiological investigation of diseases with high heterogeneity of genetic etiology and behavioral manifestation more manageable 35. The symptoms of acute ketamine injection show a large overlap with many schizophrenia endophenotypes, but the analysis of the discrepancies is also important as it may provide additional information for understanding schizophrenia pathology. For example, one of the most firmly established schizophrenia endophenotypes, the deficit of prepulse inhibition on startle (PPI) 44, 45, 46, 47, is reproduced by NMDAR antagonism in rodents 48, 49, 50 and primates 51, 52 but not in healthy human subjects where ketamine enhances PPI 53, 54, 55.

Electrophysiological Signals

Electrophysiological signals recorded in patients can reflect abnormal neuronal functioning associated with a given disorder, genetic alteration, or clinical symptom and can be considered as endophenotypes. A number of neurophysiology‐based endophenotypes, or candidate endophenotypes, have been recognized in schizophrenia 8, 9, 12, 42, 43, 56, 57, 58, 59. Using neurophysiology‐based biomarkers, for example, in evaluating drug action, can also provide unique opportunities to establish translational measures both in preclinical models and in clinical studies. Some of these biomarkers are considered to be related to abnormal glutamate neurotransmission and can be elicited with an NMDAR antagonist in healthy subjects 60. In these studies, ketamine plays a pivotal role, as it has been approved as a drug for human administration. Many of the well‐defined electrophysiological biomarkers are closely linked to NMDAR function, and abnormal neurophysiologic signals characteristic for schizophrenia can be elicited by ketamine. Among these markers are auditory gating (P50 and N100), mismatch negativity (MMN) and the P300 event‐related potential.

Ketamine‐Induced Disruption of Auditory Gating

Acoustic stimulation elicits several auditory‐evoked potentials (AEPs) with various latencies, including a positive deflection at 50 ms (P50) and a negative deflection at 100 ms (N100) in healthy subjects. Repeating the same acoustic stimulation within sufficiently short intervals (typically at 0.5 s) significantly attenuates the AEPs associated with the second stimuli, a phenomenon known as auditory gating 61, 62. In various psychiatric and neurological disorders, auditory gating is impaired, although most likely due to different pathophysiological mechanisms. Impaired auditory gating has been described in schizophrenia using standard clinical EEG methodology, as well as by using magnetoencephalography, which confirmed the gating deficit and provided additional insight to neurological mechanisms contributing to normal and pathological gating 63, 64. In preclinical animal models, auditory gating is present, and it can be disrupted via genetic, epigenetic, and pharmacological manipulations 61. Importantly, auditory gating can be readily disrupted by ketamine in healthy subjects 65 as well as in rodents 9, 66, 67.

Mismatch Negativity and P300

Recording of auditory‐evoked potentials in response to deviant stimuli in an auditory “oddball” paradigm leads to a well‐characterized negative deflection at 100 ms, called the MMN, and a positive deflection at 300 ms (P300). P300 is elicited by attended oddball stimuli and is generated by both frontal and parietal cortical areas. In patients with schizophrenia, a reduction in P300 amplitude has been well characterized, although abnormalities of P300 are present in many psychiatric or neurological disorders. MMN is generated by the auditory cortex, but frontal brain regions are also crucially implicated in the generation of the MMN demonstrated by EEG 68, 69, 70and fMRI studies 71, 72, 73. The MMN can be elicited by unattended stimuli, and abnormal MMN is considered to be the most closely related endophenotype to schizophrenia 74. Impaired generation of both MMN and P300 has been observed after ketamine treatment in humans 75, 76, 77, 78, 79, resembling deficits present in patients with schizophrenic 80, 81. Importantly, the two populations showed the same relationship between the ketamine‐induced MMN memory trace effect and cognitive impairments 79, 80. It has also been shown recently, by using a model‐based approach, that ketamine affects synaptic plasticity during the encoding of the MMN as expressed by a reduced forward connection from left primary auditory cortex to superior temporal gyrus and that this model‐based estimate of ketamine effects on synaptic plasticity correlated significantly with ratings of ketamine‐induced impairments in cognition and control 82.

Cortical Oscillations

Cortical oscillations are essential for a variety of cognitive processes impaired in schizophrenia 83. Oscillatory synchronization organizes neuronal activity in local microcircuits and supports long‐range dynamic connections, also impaired in schizophrenia 83. They depend on a functioning network of fast‐firing interneurons that show structural abnormalities in human postmortem brains 84, 85, 86, 87, 88 of schizophrenics, suggesting that impaired neuronal oscillations serve as a mechanistic link between deficiencies of the interneuron network and cognitive dysfunction. Developmental abnormalities of parvalbumin‐positive interneurons are consistently observed in chronic animal models produced by a variety of techniques 27, 89, 90, 91, 92, 93, 94, 95, 96, 97, and in the past several years, it has also been firmly established that the essential features of human schizophrenia recapitulated by rodent models include abnormal oscillations in low‐ (delta, theta) and in the high‐(beta, gamma)frequency bands 25, 26, 27, 48, 89, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111.

Ketamine and other NMDAR antagonists cause severe perturbations in cortical oscillations at different frequencies, which, in general, resemble those in schizophrenia. Comparison of impaired oscillations in schizophrenia and those induced by ketamine administration in humans and in a variety of animal models, however, is not always straightforward and differs in important details. Some of the differences, however, may be due to differences in the recording techniques/experimental design and might thus point to inadequacies of the recording paradigms used for assessment of oscillatory impairment in schizophrenics rather than to limited validity of the animal model.

Gamma‐Band Oscillations

Significant alterations of the electrical activity in the gamma band (30–90 Hz) have been documented in patients with schizophrenia 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122 and in most animal models studied to date 89, 99, 106, 107 (Table 1). Extensive research in this area has clarified several key issues and led to the proposal that a “gamma oscillatory endophenotype” 8 underlies downstream phenotypic cognitive deficits characteristic of schizophrenia. Specific therapeutic targeting of gamma‐band deficits has also been suggested 8, and an intermediate oscillatory phenotype has also proven a more fruitful correlation target than behavioral measures for identifying genetic biomarkers in some human imaging studies 123.

Table 1.

Comparison of gamma‐band oscillations (GBO) in patients with schizophrenia, animal models of schizophrenia, and following applications of ketamine to healthy humans, intact animals, and rodent brain slices

	Enhanced GBO power	Attenuated GBO
Schizophrenia (human)	Background, psychosis Spontaneous 140, 141 Hallucinations 122, 134, 135, 180 Prestimulus baseline 112	Sensory‐evoked Task related (ASSR, visual stimuli, attention, working memory) 113, 114, 115, 116, 117, 118, 119, 120, 121 (but see 133 for ASSR GBO increase)
Chronic animal models of schizophrenia	Background Prenatal MAM rat 101 Amygdala picrotoxin rat 99 PV‐Cre/NR1 mice 136 Dysbindin‐1 mutant mice 107	Sensory evoked‐Task related Prenatal MAM rat (reverse learning 89) Neonatal VH lesion rat (ASSR 105, 106) PV‐Cre/NR1 mice (opto‐stim 136) Dysbindin‐1 mutant mice (ERP 107)
	Reduced GBO reaction to NMDAR antagonists Prenatal MAM 101, 104 Neonatal VH lesion rat 208 PV‐Cre/NR1 mice 136
Acute ketamine or other NMDA antagonists—human	Auditory ERP 28 ASSR 124	Magnetoencephalography 125
Acute ketamine or other NMDA antagonists – in vivo rodents	Background Spontaneous 25, 109, 127, 128, 142 (NR2A 48) Delayed in REM sleep (NR2B 100) Prestimulus baseline‐mice 26, 103	Sensory evoked Auditory ERP‐mice 26, 103
In vitro GBO in neocortex	Bath application of ketamine or other NMDAR antagonists 130, 131, 147 (but see 129 for negative finding)	Reductions in peak frequency with bath application of ketamine only 130, 131
		Reduction in peak frequency ex vivo following 5 daily i.p. injections 209

Gamma‐band oscillations (GBO) alterations also appear in the acute state after administration of NMDAR antagonists to healthy humans 28, 124, 125 or patients with schizophrenia 125 (Table 1). They are also apparent in in vivo 25, 26, 27, 48, 100, 109, 126, 127, 128 and in vitro 129, 130, 131 preclinical studies in rodents, but there are inconsistencies between studies using different experimental paradigms. The most prominent discrepancy is that whereas NMDA blockade induces a massive increase in gamma activity, the current human schizophrenia EEG literature is dominated by reports of GBO deficits (revs. 8, 83, 132, 133). Until recently, animal models have also been focused on decreased GBO. These studies were designed to evaluate oscillations during cognitive tasks and postulated that a decrease in GAD67 and PV expression, the most consistent postmortem histological marker of schizophrenia, leads to attenuated GBO activity. During the past decade, a number of experimental paradigms and signal analysis methods were developed and standardized, which produced a wealth of data demonstrating the link between cognitive deficits and impaired synchrony in schizophrenia. Animal models generated by a variety of interventions also revealed a final common pattern of anatomical abnormalities of parvalbumin‐expressing interneurons, cognitive deficits, and decreased gamma power in schizophrenia‐relevant tasks or in sensory paradigms directly replicating those performed in humans 89, 106.

However, the recent demonstration of increased gamma‐band power after administration of subanesthetic doses of ketamine associated with schizophrenia‐relevant behavior in rats 25 turned attention to earlier sporadic reports of increased, that is, rather than decreased, GBO activity in patients with schizophrenia. Most of these earlier studies recorded baseline activity in the high beta and gamma range in default mode or resting‐state paradigms (rev.8), but a few also showed increased gamma activity directly associated with positive symptoms 122, 134, 135. The relevance of these data is further underscored by verification of the gamma‐enhancing effect of ketamine in humans 28, 124 and the demonstration of increased background gamma activity in different chronic rodent models of schizophrenia, some of which showed both increased baseline and decreased task‐related GBO 99, 101, 107, 136, 137. This prompted revisiting the issue of background GBO, which should be expanded in the future using today's advanced recording and analysis techniques 138, 139. In fact, an overall increase in high‐frequency EEG activity (>30 Hz) in schizophrenics compared with healthy controls was demonstrated in the dawn of computerized EEG 40 years ago, using analogue filters and the first versions of digital spectral analysis 140, 141. The pattern revealed in these 24‐h EEG recordings (increased high‐frequency activity along with decreased theta/alpha and increased delta‐band activities) corresponds with that seen in extended EEG recordings in rodent models during the animals' natural states and behaviors 99, 101, 107, 136, 137 and with acute changes in EEG after ketamine 25, 27, 48, 109, 142. There are current efforts to solve technical‐related problems including the difficulties of the elimination of the microsaccade artifact or the problem related to the practice of standardizing evoked responses using prestimulus “background” activity. This latter was recently shown to increase in the acute NMDAR antagonist model in mice 26 and in a reanalysis of earlier data of human ASSR 112. It is important that these studies followed similar protocols in human and animals, but the prestimulus segments they used for evaluation of background GBO remain tied to repetitive sensory test stimuli and thus might have been compromised by increased neuronal activity induced in sensory cortex by extended stimulation, especially at short interstimulus intervals 134, 143, 144. Stimulus parameters may strongly affect the level of this activation as well as the ASSR itself; when using longer stimulus duration and interstimulus interval, gamma‐band ASSR was shown to increase (rather than decrease) in schizophrenics relative to healthy controls 133.

Investigation of Ketamine Effect on GBO In Vitro

Although GBO are generally considered to be generated within specific cortical areas, GBO abnormalities in schizophrenia or with application of NMDAR antagonists can be due to changes occurring locally in the cortical area underlying the EEG electrode or due to impaired connections with other cortical or subcortical sites involved in their state‐ or event‐related modulation. Thus, in vitro studies can provide vital information, unavailable from in vivo preparations, regarding the cellular mechanisms of GBO generation within local cortical regions. In particular, studying oscillations in brain slices has several advantages: (1) the cellular and molecular basis for changes in the oscillations can be studied using recordings from specific cell types generating the oscillations, (2) novel genetic technologies can be used to identify particular cell types by the introduction of fluorescent markers and modulate their activity to model pathological changes observed in schizophrenic brains 130, (3) changes in oscillations can be ascribed to local changes in neuronal circuitry rather than to changes in afferent inputs, and (4) the elicited oscillations tend to be more specific for the gamma band, allowing a more precise determination of peak frequency and power. Recently, several groups, including our own, have investigated GBO alterations with ketamine or other NMDAR antagonists in vitro, in the hope of developing a simple translational assay for potential therapeutic agents. It remains to be seen, however, whether such experiments will have predictive value in developing effective treatments.

In vitro experiments also have several disadvantages. They are obviously not suitable for modeling psychotic symptoms in schizophrenia or any of the behavioral symptoms induced by ketamine and cannot recapitulate deficits due to impaired long‐range communication between cortical and subcortical sites 145. Thus, it is important to combine and compare the information obtained from these studies with in vivo animal models and human studies, as we do here.

Several studies 129, 130, 131, 146, 147 have examined the effect of acute, bath application of ketamine or other NMDAR antagonists on in vitro GBO and have identified some of the molecular and circuit changes, which may underlie both suppression and enhancement of GBO in schizophrenia. The first studies using interface‐type slice chambers reported that GBO are either unaffected or reduced in the hippocampus and neocortex 129. However, since then, several studies using different technology have shown that NMDAR antagonist administration strongly increases the power of GBO in the visual cortex, auditory cortex, and prefrontal (prelimbic) cortex 130, 131, 146, 147. These findings are more consistent with in vivo findings of increased power and imaging studies indicating increased activity in prefrontal cortex with ketamine (e.g. 148). While in vitro studies have a number of advantages (described above), there are important methodological issues related to the generation of GBO in vitro which make it important to compare in vitro findings to each other and to the findings associated with in vivo application of ketamine.

The major difficulty of slice studies of oscillations is that in most in vitro preparations, spontaneous GBO do not occur, as the afferent inputs are severed during the slicing process. Therefore, some means of increasing network activity is required to elicit the oscillations. Pioneering early work by the Whittington group in the hippocampus used tetanic electrical stimulation or application of metabotropic glutamate receptor agonists 149. However, most groups now use application of the glutamate receptor agonist, kainate 150, application of the cholinergic receptor agonist carbachol 151, or a combination of the two. Application of kainate mimics the increased glutamatergic input caused by the activity in other cortical areas and/or from the thalamus. The application of carbachol mimics the release of acetylcholine in the hippocampus and neocortex produced by the activity of basal forebrain cholinergic neurons, during waking and rapid‐eye‐movement sleep 152. Rhythmic release of GABA from perisomatic, parvalbumin‐positive interneurons is required for GBO elicited by both kainate and carbachol in vitro 153, 154.

Until recently, the majority of slice oscillation studies were performed in the hippocampus using the so‐called interface or Haas‐type recording chambers, together with bath application of low concentrations of kainate (typically in the high nanomolar range) or carbachol. In interface‐type chambers, the upper surface of the slice is exposed to a humidified carbogen (95% O₂/5% CO₂) gas mixture, which provides excellent oxygenation. An advantage of this approach is that once induced, oscillations can be studied for many hours. While this approach has produced a large volume of high‐quality and important data, it does not allow visually guided, whole‐cell recordings from specific neuronal subtypes labeled using fluorescent markers, as this requires the use of high magnification, water‐immersion lenses. Recent studies in recording chambers where the slices are completely submerged in artificial cerebrospinal fluid suggest that GBO require a high oxygen level 155, consistent with the correlation found between the blood‐oxygenation level (BOLD) and GBO in human imaging studies 156. Thus, improving oxygen flow to the slice by increasing the flow of artificial cerebrospinal fluid to both surfaces of the slice was found to facilitate GBO in submerged‐type slice chambers. However, even with such modifications, we could not reliably induce GBO in neocortical slices 130. Thus, we and others 157 have used a modified method involving brief, focal application of a higher (1 mM) concentration of kainate. This generates a relatively brief (10 s of seconds) burst of GBO, which can be reproducibly elicited provided sufficient time is allowed between applications. These differences in methodology (kainate vs. carbachol, interface vs. submerged and prolonged vs. brief focal application) should be born in mind when comparing different in vitro studies as well as between in vitro and in vivo findings.

In addition to increasing the power of GBO, we also recently found that ketamine slows the peak frequency from the gamma to the beta range 130. These findings are intriguing, given the gamma→beta shift described in patients with schizophrenia 158 and modeling studies which suggest that beta oscillations are more effective in synchronizing activity over longer distances 159, 160. This slowing of the peak frequency was not reproduced by more selective NMDAR antagonists and was found to be an off‐target effect of ketamine on GABA_A receptors, causing a slowing of the decay time of inhibitory postsynaptic currents 130.

Low‐Frequency Oscillations

Although recent research is primarily focusing on high‐frequency oscillations, there is also evidence of disturbances in slow rhythms in the delta and theta bands in schizophrenia 161, 162, 163, 164, 165. Functional deficits are most likely the consequence of parallel impairments of fast and slow rhythms as oscillations at different frequencies have overlapping neuronal substrates 166, 167 and are hierarchically organized such that slow rhythms drive coordinated shifting of excitability in local neuronal ensembles and optimize gamma dynamics 168, 169. The oscillatory hierarchy operating across multiple spatial and temporal scales 170, 171 is important for long‐range synchronization between cortical areas and plays a critical role in various cognitive processes 168, 172. Theta–gamma cross‐frequency coupling was recently found impaired in mice with genetically induced chronic NMDAR hypofunction 137, 173 and in rats after acute NMDAR blockade 48. This effect also showed subunit specificity; that is, acute elevated GBO induced by NR2A‐preferring antagonists was associated with a severe impairment of low‐frequency theta modulation similar to nonselective NMDAR blockade 48. Theta‐band deficits were also found in sensory gating paradigm in schizophrenics and their first‐degree relatives compared with healthy controls showing a significantly higher heritability than the commonly used P50 gating endophenotype 56. Enhanced activity in the low‐frequency delta range is a common observation in schizophrenia 140, 174 and is replicated by NMDAR antagonists in preclinical studies in vivo 110, 111 and in vitro 175 as well as in the NMDAR hypomorphic mice model 176.

Cellular Mechanisms of Ketamine‐Induced and Schizophrenia‐Associated Changes in Cortical Oscillations

As discussed above, impaired oscillations in schizophrenia have been linked to GABA pathology, that is, to reductions in GAD67 and PV expression 11, 84, 85, 86, 87, 88, 177. PV+ interneurons are involved in the generation of oscillations 178, 179, suggesting a causal relationship between this histological marker and decreased oscillations documented in human schizophrenia 83, 180 and in a variety of animal models 26, 89, 106, 107, 136, 173, 181. NMDAR hypofunction was proposed to affect oscillations through changes brought about in the interneuron network of the schizophrenic brain 11, 177. PV+ interneurons are specifically vulnerable to NMDAR blockade. NMDARs play a specific role in the maintenance of the phenotype of PV+ interneurons. Exposure of cultured PV+ neurons to ketamine induced time‐ and dose‐dependent decrease in PV and GAD67 182. Thus, NMDAR activation of genomic programs and intracellular signaling pathways 182, 183, 184, 185 may contribute to impairment of neuronal synchronization in schizophrenia in human 83 and in NMDAR hypofunction‐based chronic animal models 27, 137, 186, 187. In contrast to acute application of ketamine, which causes psychosis, chronic application of ketamine and other NMDAR antagonists has been used to more closely model deficits in executive function. In rats receiving daily injections of 30 mg/kg ketamine for 5 days 186, there was a decrease in GAD67+ and PV+ cell number 27, 188 and a decrease in GBO power 27, which contrasts the increases in power produced by acute ketamine. In preliminary studies, we have also found reduction in peak frequency of oscillations and a trend‐level reduction in power in vitro ²⁰⁷.

In our view, NMDARs on PV+ interneurons are relatively unimportant for normal synaptic transmission (see discussion of acute effects below) but serve as a sensor for the level of network activity. Thus, GAD67 and PV may be downregulated in response to the increased network activity produced by repeated ketamine applications. This mechanism might also be relevant in producing some of the subacute effects of administration of a single dose of NMDAR antagonists, such as the delayed GBO increase observed during REM sleep, 4–5 h after injection of MK‐801 100. In contrast to the relatively short lasting (<1 h) effect of ketamine, MK‐801 elicits longer‐lasting aberrant GBO elevation accompanied by stereotypic (waking) behavior, which is then followed by a second type of enhanced GBO, only occurring during REM sleep. The effect of NMDAR antagonists preferably blocking NR2A subunit‐containing receptors resembles this two‐component pattern 48, whereas selective NR2B blockade does not disrupt the sleep–wake cycle and elicits REM sleep‐related GBO enhancement at short latencies, that is, starting right from the first REM sleep episode after injection 100. Thus, the delayed, REM sleep‐related (apparently NR2B‐dependent) GBO elevation induced by MK‐801 may involve subacute changes in the composition of NMDARs in interneurons generating a relative NR2B hypofunction. This is because NMDARs are regulated by receptor activity in a subtype‐specific manner. Compensatory upregulation of NR2A but not NR2B subunit‐containing receptors can change the NR2A/NR2B ratio as early as 4 hrs after MK‐801 application to generate a relative NR2B deficit 189, 190. Acute application of selective NR2B antagonists also enhanced kainate‐induced GBO in vitro, similar to general antagonists of the NMDAR 130. Importantly, the NR2B subunit has been implicated as a schizophrenia‐susceptibility gene 191.

Ketamine‐induced increased power of GBO also has often been ascribed to a selective block of NMDAR on interneurons 192, resulting in increased excitability of pyramidal neurons, as the NMDARs on interneurons are reported to have a higher affinity for these agents than those on pyramidal neurons in the hippocampus 193. However, at the concentrations used in in vitro GBO studies, most NMDARs in the slice are likely to have been blocked. Furthermore, whole‐cell recordings from fast‐spiking interneurons in the hippocampus and neocortex have revealed a relatively minor contribution from NMDARs to synaptic currents 194. Thus, block of the more prominent NMDAR on low‐threshold spiking, dendrite‐targeting interneurons may be more important in increasing the excitability of the pyramidal neurons. Furthermore, the acute potentiating effect of NMDAR antagonism may also be due to block of the NMDAR currents on pyramidal neurons reducing jitter of synaptic currents and thereby enhancing synchrony. Selective block of NMDA on different types of neurons will ultimately be required to resolve this question. Interestingly, an additional mechanism that may account for increased power is the collapsing of high and low‐gamma oscillators in deep and superficial cortical layers 131.

N‐methyl‐d‐aspartate receptor blockade on dendritic‐targeting interneurons, for example, those located in the stratum oriens of the hippocampus and projecting to lacunosum‐moleculare (OLM), may also play a key role in changes of theta rhythm after ketamine administration. These cells are very sensitive to NMDA blockade 195. They are not synchronized with gamma oscillations 196 but play a critical role in generating theta rhythm 196, 197. Theta–gamma coordination is important for cognitive processes 168, 169, and OLM interneurons were shown to critically contribute to the formation of gamma‐coherent cell assemblies at long distances by entrainment at theta frequency 198. A recent computational study replicating essential features of hippocampal oscillatory activity, including cross‐frequency theta–gamma modulation, explored the effect of different combinations of NMDA blockade in two pyramidal cell domains and on two types of interneurons, that is, slow‐firing OLM cells and parvalbumin‐positive basket cells firing at high rates and in synchrony with gamma field potentials 199. They found that the in vivo pattern of reduced theta and enhanced GBO could only be elicited by selective NMDAR blockade on OLM cells, whereas NMDAR blockade on basket cells led to a decrease in gamma power.

Furthermore, since the administration of ketamine also modulates other receptors and neurotransmitter systems such as AMPA and dopamine, the effect of ketamine on cortical oscillations might include manipulations of these systems as well, for example, an excessive AMPA receptor stimulation due to glutamate spillover 200, 201.

Finally, changes in cortical and hippocampal oscillations may also develop due to mechanisms potentially related to NMDARs outside of the cortex or hippocampus. Hippocampal theta power and frequency is regulated by a number of subcortical structures, including most importantly the pontine reticular formation, median raphe, supramammillary nucleus, and medial septum 202, 203. Medial septum input also amplifies GBO 199, and the involvement of medial septum and supramammillary nucleus in NMDAR antagonist‐induced gamma and PPI impairment has been demonstrated 67, 128. Similarly, more caudally located basal forebrain neurons may modulate neocortical theta and gamma power 152. In the brainstem, injection of MK‐801 or AP‐7 in the theta‐promoting pontine reticular formation had no effect, whereas NMDAR blockade of the median raphe nucleus, normally inhibiting theta, was shown to produce theta rhythm in the hippocampus at short latencies and long duration, in urethane‐anesthetized rats 204. Local field potentials showed enhanced GBO activity in several subcortical structures, including accumbens, basalis, striatum, and thalamus 109. An NMDAR mechanism in the thalamus was also implicated in altered cortical delta oscillations in schizophrenia 110, 111, 175. In particular, the slowing and enhancement of delta waves in the prefrontal cortex observed after systemic NMDAR antagonists were replicated by local microinjections of MK‐801 into the mediodorsal thalamus but not into the prefrontal cortex 110.

Conclusion

Both acute and chronic NMDAR blockade have been used in previous studies to validate the NMDAR hypofunction model. These investigations found that alterations in fast‐spiking interneurons 182, 188 and in cognition 50, 186, 205, 206, 207 were observed in both types of models and were consistent with human data. Measurement of EEG oscillations, which may serve as a link between the pathology of GABA networks and cognition deficits, presents a unique challenge, however, as the ketamine‐induced acute increase has to be reconciled with more commonly reported chronic decreases of gamma activity in schizophrenics. We believe that some of the differences between human EEG data and animal models may be due to differences in the recording techniques/experimental design and might thus point to inadequacies of the recording paradigms used for assessment of oscillatory impairment in schizophrenics rather than to limited validity of the animal model. Critical comparison of the electrophysiological changes induced by ketamine and other NMDAR antagonists with those observed in patients with schizophrenia and animal models (Table 1) suggests that in schizophrenia, GBO increases and decreases may both be a consequence of NMDAR hypofunction, that is, schizophrenia and NMDAR blockade on interneurons may be associated with increased baseline GBO and decreased evoked GBO. Ketamine, which can be used both in human and animals, may play thus an essential role in developing the right methods and standards for detecting schizophrenia‐relevant cortical dysfunction as well as testing novel therapeutic agents targeting GBO.

Conflict of Interest

The authors declare no conflict of interest.