The synchronization of cortical circuits
One possible functional correlate of the aberrant connectivity observed in TRS is disturbances in cortical network oscillations. Oscillations in network activity include the theta (~4–8 Hz), alpha (~8–13 Hz), and gamma (~30–80 Hz) bands. These oscillations are measurable by electroencephalography (EEG) and magnetoencephalography (MEG) and are thought to be reflective of cortical information processing and integration (79, 81) Importantly, they reflect the synchronous activity of large populations of neurons that integrate information across multiple brain regions. With regard to schizophrenia, specific interest has been paid to the gamma band oscillation (GBO) (82–85). The GBO plays an important role in a variety of cognitive tasks including sensory processing, working memory, attention, and cognitive control–all of which are disturbed in the illness (86–91). More generally, it is thought to be critical to the process of feature binding, in which sensory information of a variety of modalities is integrated coherently into a unified representation (92). Fittingly, it has been suggested that the underlying dysfunction in schizophrenia is the inability to integrate the activity of distributed neuronal networks. These disturbances in the GBO could underlie the dysfunctional communication observed between disparate brain regions in the illness.
The GBO has been shown to be disrupted in schizophrenia patients during the performance of a wide variety of tasks, including simpler perceptual tasks and more complex and cognitively demanding tasks (93–96). In patients diagnosed with schizophrenia, EEG studies have shown that the GBO is impaired in working memory tasks at frontal and posterior sites, as well as in the frontal cortex during cognitive control tasks (97–100). Performance of these tasks is typically associated with increase in GBO activity in healthy subjects. However, in subjects with schizophrenia this demand-related modulation of the GBO is absent or diminished. The deficit in task-related modulation of the GBO is also present in first-episode patients, suggesting that this is driven by the underlying disease process rather than illness chronicity or long-term use of antipsychotic medications (99). Several of these studies have also shown that deficits in cognitive control in patients with schizophrenia are correlated with their deficits in GBO activity (91, 98). Convergent evidence from fMRI studies has also shown a lack of task-demand related modulation of activity in the PFC in schizophrenia patients (101). These findings suggest that for cognitive tasks, particularly those that may depend on integration of information, the GBO is a reflection of disturbed functional connectivity between communicating brain regions.
Multiple models have been generated to describe the underlying neural circuitry that gives rise to the GBO. Two prominent ones include the Interneuron Network Gamma (ING) model and the Pyramidal Interneuron Network Gamma (PING) model (102). In the ING, pyramidal cells are synchronized by the activity of interneurons, but pyramidal cells themselves are not directly involved in the generation of the GBO. In PING, oscillations are generated via the recurrent synaptic connectivity between the excitatory activity of pyramidal cells and feedback inhibition of interneurons. While this process is still not fully understood, experimental observations favor the PING model of GBO generation. In this case, synaptic inhibition via GABAergic interneurons defines the timing and firing rate of pyramidal neurons, creating precise windows within which large groups of excitatory cells can fire synchronously (103–105). In turn, excitatory cells also provide input onto GABAergic interneurons, creating a loop for entrainment of cortical networks across brain regions. Support for the PING model comes from findings that interneuron activity follows pyramidal cell activity by a short delay, consistent with pyramidal cell excitatory drive as the main stimulus for interneuron excitation in the model (106, 107). Within excitatory cells, α1-containing GABAA receptors post-synaptic to a subset of inhibitory interneuron processes produce currents with decay periods fitting for the production of gamma oscillations (84). Lastly, it has been shown that with genetic knockout of AMPA glutamate receptors within specific populations of inhibitory interneurons, synaptic excitation of these inhibitory interneurons is diminished and the power of the gamma oscillation severely reduced (108). These findings support the theory that the GBO arises from a complicated interplay between excitatory pyramidal cells and inhibitory interneurons.
Consistent with the PING model, there is ample evidence to suggest that both excitatory glutamatergic and inhibitory GABAergic activity are disturbed in schizophrenia (84, 85). Deficits in excitatory glutamatergic signaling have been identified as a possible core feature behind the pathophysiology of schizophrenia that gave rise to the NMDA receptor hypofunction hypothesis (71). This hypothesis arose from the observation that NMDA receptor antagonists (e.g., ketamine, PCP) can reproduce some of the symptoms of schizophrenia. Subsequent studies have identified widespread dysfunction of NMDA receptors in schizophrenia. Interestingly, given that the GBO is thought to be generated by the activity of inhibitory interneurons, much of the observed dysfunction in NMDA receptors has been specific to inhibitory interneurons themselves. For example, post-mortem analysis of the PFC of schizophrenia patients has shown a 50% reduction in the expression of the NR2A subunit within inhibitory interneurons that express parvalbumin, a calcium binding protein (109). Moreover, chronic NMDA receptor antagonist administration in rodent models reduces the expression of the parvalbumin protein and GAD67 (the primary GABA-synthesizing enzyme glutamic acid decarboxylase) in parvalbumin-positive(+) inhibitory interneurons (110, 111). Acute administration of NMDA receptor antagonists has also been shown to decrease the activity of interneurons with a corresponding increase in the activity of pyramidal cells (70). Thus, NMDA receptor antagonism may reduce the function of inhibitory interneurons which subsequently disinhibits the activity of pyramidal cells. Within the context of schizophrenia, NMDA receptor hypofunction may result in the diminished excitation of inhibitory interneurons within cortical networks.
Inhibitory interneurons are particularly sensitive to NMDA receptor antagonists (70, 112, 113). In combination with findings of altered expression of NMDA receptors within these interneurons, it is well-supported that inhibitory interneurons, particularly those expressing the calcium-binding protein parvalbumin, are a locus for dysfunction in schizophrenia (shown in Figure 2) (84, 114, 115). A number of studies have shown that parvalbumin+ cells are critical to the generation and maintenance of the GBO (106, 113, 116, 117). These interneurons have extremely fast-spiking properties and their rapid synaptic activation is consistent with the frequency required for entrainment of the GBO (118). Parvalbumin+ cells also show the strongest coupling to the gamma oscillation cycle relative to other interneuron types (e.g., calbindin, calrentin) (119, 120). Parvalbumin+ interneurons are typically fast-spiking and provide perisomatic inhibition onto excitatory pyramidal cells. Parvalbumin+ interneurons can present morphologically as either basket (project to the soma and proximal dendrites of neurons) or chandelier cells (project to the initial axon segment of neurons) as illustrated in Figure 3. While both parvalbumin+ basket and chandelier cells are active during GBO, parvalbumin+ basket cell activity is more strongly coupled with the GBO (121). Studies have also shown that GBO power is markedly reduced by opioid receptor activation, which dampens the activity of synaptic inputs from parvalbumin basket cells onto pyramidal neurons but does not affect chandelier neurons (122). These findings emphasize the critical importance of parvalbumin+ basket cells specifically to the generation of the GBO and their dysfunction in schizophrenia. In support of this, it has been shown that reductions in the firing rate of parvalbumin+ interneurons via optogenetics can reduce the power of GBO (114). Conversely, non-rhythmic stimulation provided to parvalbumin+ interneurons can increase the power of the GBO.
Figure 2  Parvalbumin interneurons contribute to the inhibitory dysfunction in schizophrenia. Parvalbumin interneurons are fast-spiking inhibitory interneurons characterized by the calcium binding protein parvalbumin. These interneurons are innervated by excitatory glutamatergic cells and in turn their projections target the cell soma of excitatory pyramidal cells. This excitatory-inhibitory interplay is thought to give rise to the GBO, which is reflective of parvalbumin interneurons role in synchronizing large populations of excitatory cells. The GBO is disturbed in schizophrenia, and dysfunction within parvalbumin interneurons is thought to be central to these abnormalities.
Figure 3  Schematic diagram of perisomatic inhibition of cortical pyramidal cells by parvalbumin+ basket cells and chandelier interneurons. Adapted from Lewis et al. (115). Parvalbumin+ cells have been extensively studied in schizophrenia and evidence of their dysfunction extends well beyond their contribution to the GBO (85, 115, 123). Parvalbumin+ cells have a reduction in mRNA and protein levels of parvalbumin itself despite unaltered neuronal density in patients with schizophrenia observed post-mortem (124–126). Parvalbumin+ cells also have reduced protein and mRNA levels of GAD67 and up to 50% of parvalbumin+ cells are wholly devoid of GAD67 (124). This loss of GAD67 represents a significant decrease in the strength of inhibitory inputs on the pyramidal cells they target (115). Moreover, this deficit has been observed in parvalbumin+ cells across multiple cortical regions including the DLPFC and ACC (127–129). Two hypotheses have been generated to account for the convergent evidence of dysfunction localized to parvalbumin+ basket cells (84). One hypothesis emphasizes the inhibitory contribution to this network interplay and the other excitatory activity. First, lower GAD67 levels in parvalbumin+ basket cells could result in a disinhibition of pyramidal cells. Alternatively, the loss of GAD67 in parvalbumin+ basket cells could be a development disruption due to lack of excitatory input onto these cells to drive their activity. Consistent findings of dendritic spine loss on pyramidal cells in areas like the DLPFC and dysfunction within glutamatergic channels (e.g., NMDA, AMPA) could contribute to this loss of excitatory input onto parvalbumin+ basket cells in schizophrenia (68). These findings support the central importance of parvalbumin+ inhibitory interneurons in schizophrenia pathophysiology but whether this is a primary pathology or homeostatic mechanism in response to diminished pyramidal cell input is still unclear.
Despite an improved understanding of the underlying pathophysiology of schizophrenia, particularly with regard to cellular mechanisms contributing to the GBO, a multitude of questions remain. Of utmost importance to the current review is the validity of these findings, many of which have been garnered from animal models of schizophrenia, to TRS. Current cellular and animal models have significant limitations in modeling the illness and few, if any, attempts have been made to replicate the treatment-resistant presentation of the disorder. Secondly, further investigation is required to understand the complex interplay between excitatory glutamatergic cells and inhibitory interneurons in the dysfunctional circuitry of schizophrenia. Specifically, a better understanding of the cellular properties that give rise to the GBO are necessary to better understand approaches for treatment. And lastly, novel treatments and therapeutics need to be designed to target the pathophysiological functioning of GBO circuitry. These approaches may include pharmacological stimulation of the neural circuitry or might be targeted through novel non-pharmacological approaches, such as rTMS which can directly stimulate the GBO.