The beta component of gamma-band auditory steady-state responses in patients with schizophrenia

Abstract

The mechanisms underlying circuit dysfunctions in schizophrenia (SCZ) remain poorly understood. Auditory steady-state responses (ASSRs), especially in the gamma and beta band, have been suggested as a potential biomarker for SCZ. While the reduction of 40 Hz power for 40 Hz drive has been well established and replicated in SCZ patients, studies are inconclusive when it comes to an increase in 20 Hz power during 40 Hz drive. There might be several factors explaining the inconsistencies, including differences in the sensitivity of the recording modality (EEG vs MEG), differences in stimuli (click-trains vs amplitude-modulated tones) and large differences in the amplitude of the stimuli. Here, we used a computational model of ASSR deficits in SCZ and explored the effect of three SCZ-associated microcircuit alterations: reduced GABA activity, increased GABA decay times and NMDA receptor hypofunction. We investigated the effect of input strength on gamma (40 Hz) and beta (20 Hz) band power during gamma ASSR stimulation and saw that the pronounced increase in beta power during gamma stimulation seen experimentally could only be reproduced in the model when GABA decay times were increased and only for a specific range of input strengths. More specifically, when the input was in this specific range, the rhythmic drive at 40 Hz produced a strong 40 Hz rhythm in the control network; however, in the 'SCZ-like' network, the prolonged inhibition led to a so-called 'beat-skipping', where the network would only strongly respond to every other input. This mechanism was responsible for the emergence of the pronounced 20 Hz beta peak in the power spectrum. The other two microcircuit alterations were not able to produce a substantial 20 Hz component but they further narrowed the input strength range for which the network produced a beta component when combined with increased GABAergic decay times. Our finding that the beta component only existed for a specific range of input strengths might explain the seemingly inconsistent reporting in experimental studies and suggests that future ASSR studies should systematically explore different amplitudes of their stimuli. Furthermore, we provide a mechanistic link between a microcircuit alteration and an electrophysiological marker in schizophrenia and argue that more complex ASSR stimuli are needed to disentangle the nonlinear interactions of microcircuit alterations. The computational modelling approach put forward here is ideally suited to facilitate the development of such stimuli in a theory-based fashion.

Figure 1

Network response to ASSR stimuli of different frequency. Simulated MEG signal of the control and IPSC-SCZ-like network in response to click-train stimuli with drive frequencies of 20, 30, and 40 Hz, replicating earlier studies using this model^,.

Figure 2

Power spectra of network responses to ASSR stimuli of different frequency. Power spectral densities of the simulated MEG signals from Fig. 1, again replicating earlier studies using this model^,.

Figure 3

Input dependence of the 20 Hz component in the ‘IPSC-SCZ-like’ model. (a) Power at 40 Hz in response to 40 Hz drive as a function of the input strength. (b) Power at 20 Hz in response to 40 Hz drive as a function of the input strength. (c–e) Simulated MEG signals for three different input strengths: (c) $I=0.4$ Input strength too low to drive synchronization. (d) $I=1.0$ Input strength high enough to drive synchronization and to allow for a beat-skipping behaviour. (e) $I=1.4$ Input strength too strong for beat-skipping behaviour, external 40 Hz drive dominates recurrent effects. (f–h) Power spectral densities for the signals from (c–e).

Figure 4

Input strength dependence of the 20 Hz component in the ‘IPSC+gGABA-SCZ-like’ model. (a) Power at 40 Hz in response to 40 Hz drive as a function of the input strength. (b) Power at 20 Hz in response to 40 Hz drive as a function of the input strength. In both plots the network model has increased IPSC decay times (from 8 to 28 ms) and the I-E and I-I synaptic strength ($g_{\mathit{ie}}$ and $g_{\mathit{ii}}$, respectively) is varied from 100% (black) to 10% (lightest grey) in steps of 5%.

Figure 5

Input strength dependence of the 20 Hz component ‘IPSC+bInh-SCZ-like’ model. (a) Power at 40 Hz in response to 40 Hz drive as a function of the input strength. (b) Power at 20 Hz in response to 40 Hz drive as a function of the input strength. In both plots the network model has increased IPSC decay times (from 8 to 28 ms) and the interneuron excitability $b_{\mathit{inh}}$ is varied from − 0.01 (black) to − 0.1 (darkest grey) and then in steps of − 0.05 to − 0.6 (lightest grey).

Figure 6

Input strength dependence of the 20 Hz component ‘Full-SCZ-like’ model. (a) Power at 40 Hz in response to 40 Hz drive as a function of the input strength. (b) Power at 20 Hz in response to 40 Hz drive as a function of the input strength. In both plots the network model has increased IPSC decay times (from 8 to 28 ms) and now both the I-E and I-I synaptic strength ($g_{\mathit{ie}}$ and $g_{\mathit{ii}}$, respectively) is varied from 100% (black) to 10% (lightest grey) in steps of 10% and simultaneously the interneuron excitability $b_{\mathit{inh}}$ is varied from − 0.01 (black) to − 0.1 (darkest grey) and then in steps of − 0.05 to − 0.6 (lightest grey).

Figure 7

Input strength dependence of the response to 20 Hz drive in the three different models: (a) and (b) ‘IPSC+gGABA-SCZ-like’, (c) and (d) ‘IPSC+bInh-SCZ-like’ and, (e) and (f) ‘Full-SCZ-like’. (a), (c) and (e) Power at 40 Hz in response to 20 Hz drive as a function of the input strength. (b), (d) and (f) Power at 20 Hz in response to 20 Hz drive as a function of the input strength. In all plots the network model has increased IPSC decay times (from 8 to 28 ms). In (a) and (b) the I-E and I-I synaptic strength ($g_{\mathit{ie}}$ and $g_{\mathit{ii}}$, respectively) is varied from 100% (black) to 10% (lightest grey) in steps of 5%. In (c) and (d) the interneuron excitability $b_{\mathit{inh}}$ is varied from − 0.01 (black) to − 0.1 (darkest grey) and then in steps of − 0.05 to − 0.6 (lightest grey). In (e) and (f) now both the I-E and I-I synaptic strength ($g_{\mathit{ie}}$ and $g_{\mathit{ii}}$, respectively) is varied from 100% (black) to 10% (lightest grey) in steps of 10% and simultaneously the interneuron excitability $b_{\mathit{inh}}$ is varied from − 0.01 (black) to − 0.1 (darkest grey) and then in steps of − 0.05 to − 0.6 (lightest grey).

Figure 8

(a) Network schematic showing the two neural populations (excitatory pyramidal cells and inhibitory basket cells) and their connectivity. Additionally, both populations receive periodic ASSR input drive and random background noise. (b) Three potential microscopic changes underlying gamma ASSR deficits were implemented: Increased IPSC decay times at inhibitory synapses onto PCs, decreased GABA levels resulting in reduced IPSC amplitudes at inhibitory synapses onto PCs, NMDAR hypofunction resulting in decreased excitability of GABAergic interneurons. (c) Depiction of a 40 Hz click-train stimulus, where tones (synchronous inputs to the cells of the network) are presented with an inter-click interval of 25 ms resulting in a drive frequency of 40 Hz. (d) Example simulated MEG signal of the network in response to a 40 Hz click-train stimulus.