Edinburgh Research Explorer Association of Magnetoencephalographically Measured High-Frequency Oscillations in Visual Cortex With Circuit Dysfunctions in Local and Large-scale Networks During Emerging Psychosis

IMPORTANCE Psychotic disorders are characterized by impairments in neural oscillations, but the nature of the deficit, the trajectory across illness stages, and functional relevance remain unclear. OBJECTIVES To examine whether changes in spectral power, phase locking, and functional connectivity in visual cortex are present during emerging psychosis and whether these abnormalities are associated with clinical outcomes. DESIGN, SETTING, AND PARTICIPANTS In this cross-sectional study, participants meeting clinical high-risk criteria for psychosis, participants with first-episode psychosis, participants with affective disorders and substance abuse, and a group of control participants were recruited. Participants underwent measurements with magnetoencephalography and magnetic resonance imaging. Data analysis was carried out between 2018 and 2019. MAIN OUTCOMES AND MEASURES Magnetoencephalographical activity was examined in the 1-to 90-Hz frequency range in combination with source reconstruction during a visual grating task. Event-related fields, power modulation, intertrial phase consistency, and connectivity measures in visual and frontal cortices were associated with neuropsychological scores, psychosocial functioning, and clinical symptoms as well as persistence of subthreshold psychotic symptoms at 12 months. RESULTS The study participants included those meeting clinical high-risk criteria for psychosis (n 32 26 (mean 24 [4.2] 16 38 participants with affective disorders and substance 49 age [SD], 23 [3.6] Clinical first-episode

N eural oscillations are a crucial aspect of normal brain functioning owing to their role in facilitating communication between neuronal populations, 1 a process that is closely linked to the integrity of sensory and cognitive processes. 2,3 There is emerging evidence that psychotic disorders with pronounced cognitive impairments, such as schizophrenia, involve aberrant neuronal oscillations, 4 but the nature of the impairment, the onset of deficits, and clinical relevance remain unclear.
β/γ-Band oscillations [5][6][7] but also lower frequencies are impaired during sensory 8 and cognitive tasks 9 in schizophrenia. 10,11 During normal brain functioning, inhibition of excitatory pyramidal cells through different classes of aminobutyric acid (GABA)ergic interneurons lead to the emergence of neural oscillations. [12][13][14][15] Converging evidence from genetics, 16 postmortem data, 17,18 and brain imaging have 19 highlighted that GAB-Aergic as well as glutamatergic neurotransmission is impaired in schizophrenia, supporting the possibility that measurements with electro/magnetoencephalography (EEG/MEG) could be important for translational research aimed at identifying circuit mechanisms in the disorder. 10 Critical questions concerning the role of neural oscillations in the pathophysiology of schizophrenia are the onset of abnormalities, the nature of the deficit, and functional relevance. Early signs of psychosis as well as associated cognitive deficits are already present several years prior to the full emergence of schizophrenia, 20 and thus, research efforts have shifted the focus toward identifying circuit abnormalities and biomarkers in participants who are at risk for the development of psychotic disorders that could allow for early intervention and clinical staging. 21,22 There is only limited evidence available on alterations of neural oscillations in individuals meeting clinical high-risk criteria for psychosis (CHR-P). 23,24 To address this fundamental question, we applied a state-of-the-art MEG approach to examine lowfrequency and high-frequency oscillations during a visual paradigm in CHR-P participants, patients with first-episode psychosis (FEP), and participants with substance-related and affective disorders. Magnetoencephalography is characterized by an improved signal-to-noise ratio for measurements of highfrequency oscillations compared with EEG 25,26 and is ideally suited for source reconstruction, allowing the identification of anatomical layout of generators with high spatial resolution. 27 Based on models of developing psychosis that have highlighted the central role of visual deficits during the early stages of psychosis 28,29 that predict transition to psychosis 30 as well as the importance of high-frequency oscillations for the integrity of visual perception, 1,31,32 we predicted that CHR-P participants would be characterized by a circumscribed dysfunction of β/γ-band oscillations in visual cortex that would be linked to clinical outcomes. Specifically, we focused on the persistence of attenuated psychotic symptoms (APS) because there is evidence to suggest that persistent APS are associated with poor outcomes 33 and cognitive deficits in CHR-P populations. 34 Patients with FEP, on the other hand, would involve largescale dysfunctions of induced oscillations and effective connectivity between frontal and visual areas, consistent with a disconnection syndrome. 35,36 Methods Participants Four groups of participants (total n = 232) were recruited: (1) participants meeting CHR-P criteria (n = 119) from the ongoing Youth Mental Health Risk and Resilience (YouR) Study 37 ; (2) 38 participants who did not meet CHR-P criteria (CHR-N) and were characterized by nonpsychotic disorders, such as affective disorders (n = 11), anxiety disorders (n = 16), eating disorders (n = 1), and/or substance abuse (n = 10); (3) 26 patients with FEP (13 antipsychotic-naive); and (4) 49 healthy control individuals (HC) without an axis I diagnosis or family history of psychotic disorders. Data from 10 patients with FEP and 10 HC have been published previously. 38 The CHR-P status was confirmed by ultrahigh-risk criteria according to the Comprehensive Assessment of At Risk Mental States (CAARMS) interview 39 (Table) 41 and with the Positive and Negative Symptom Scale. 42 For all groups except patients with FEP, neurocognition was assessed with the Brief Assessment of Cognition in Schizophrenia (BACS). 43 The study was approved by the ethical committees of University of Glasgow and the National Health Services Research Ethical Committee Glasgow and Greater Clyde. All participants provided written informed consent. the stimulus, randomly occurring between 750 and 3000 milliseconds ( Figure 1).

Magnetoencephalography Data Analysis
Magnetoencephalography data were analyzed with MAT-LAB using the open-source Fieldtrip Toolbox. 45 Preprocessing included correct trials only with nonoverlapping 3.8second segments (1-second baseline), time locked to the onset of the visual grating. Line noise was attenuated with a discrete 50-Hz Fourier transform filter, and faulty sensors with large signal variance or flat signals were removed. Data were denoised relative to MEG reference channels and downsampled to 300 Hz. Artifact-free data were created by removing trials with excessive transient muscle activity,  SDs of RTs, ms CHR-N CHR-P FEP HC a A, Inward moving grating task: participants report, by button press, the onset of a change in velocity of inward motion of the visual stimulus (correct response window, 200-1200 milliseconds). Feedback on performance was provided on every trial, shortly after the response onset terminated stimulus presentation. B, Histograms of group means and standard errors for accuracy (% correct), mean reaction times (RTs), and behavioral variability (intraindividual standard deviation of RTs). CHR-N indicates clinical high risk negative; CHR-P, clinical high risk positive; FEP, first-episode psychosis; HC, healthy control individuals. a Indicates significant group differences (Welch F tests, α = .05, 2-sided, 1000 samples bootstrapping, Games-Howell corrected for multiple comparisons).  46 Whole-brain source estimation of γ-band power (57-67 Hz) between 250 and 750 milliseconds was computed using the Dynamic Imaging of Coherent Sources beamforming approach 47 (eMethods in the Supplement). γ-Band source data were statistically tested across groups to determine the location of main effects ( Figure 2). These were then used to guide selection of the main regions of interest (ROIs) for more fine-grained virtual channel analyses (Figure 2 and Figure 3C) (eMethods in the Supplement). Virtual channel time series were used for the analysis of event-related fields (ERF), TFR, inter-trial phase coherence (ITPC), 48 baseline FFT, and Granger causality (GC).
Granger-causality estimates were computed using a nonparametric approach, including spectral density matrices estimated directly from FFT-data (250-750 milliseconds; DC-149 Hz; Hanning tapered; 5-Hz frequency smoothing; 1-Hz resolution; data zero-padded to 4 seconds), followed by matrix factorization and variance decomposition. Epochs were split into 2 × 250millisecond segments to increase trial numbers (see Michalareas et al 32 ). The middle occipital gyrus and cuneus ROIs were not used in the GC analyses to minimize overlap between primary visual, dorsal, and ventral stream connectivity estimates. Grangercausality data from each pair were averaged over hemisphere to create 4 main ROI pairs for statistical testing. To determine the alterations in feedforward (FF) vs feedback (FB) GC activity, we also computed the directed asymmetry index (DAI; see Michalareas et al, 32 Bastos et al, 49 and eMethods in the Supplement).

Research Original Investigation
Association of High-Frequency Oscillations With Circuit Dysfunctions During Emerging Psychosis

Statistical Analysis
Group differences in trial numbers, γ-band peak frequency, behavioral performance, demographic, and clinical data were assessed with 1-way Welch analysis of variance (ANOVA); 2-sided α level of .05. Brief Assessment of Cognition in Schizophrenia data were first z-normalized to the HC data. Bootstrapping (n = 1000) and Games-Howell correction were used to control type I errors in post hoc pairwise group comparisons. Statistical testing of group differences in MEG virtualchannel data included nonparametric Monte-Carlo-based per- mutation (n = 2000) independent F test (main group effect) and post hoc t test statistics 45 for ERFs (0-750 milliseconds); TFRs and ITPC power (1-90 Hz, 0-750 milliseconds for TFR power, 0-350 milliseconds for ITPC, and dB change from a 500millisecond baseline); baseline FFT spectra (1-90 Hz); and GC data. Type I errors were controlled by cluster statistics across time and/or frequency (eMethods in the Supplement). Finally, binary logistic regression and receiver operating characteristic curve (ROC) analyses were used to examine the association between MEG parameters and clinical outcomes in CHR-P participants (eMethods in the Supplement).

Follow-up Outcomes
We examined persistence of APS up to 12 months in CHR-P participants who met APS criteria at baseline (n = 84). For 75 CHR-P participants, at least 1 follow-up assessment was available. Thirty-nine CHR-P participants continued to meet APS criteria (APS-persistent group) while 36 CHR-P participants were characterized by a remission of APS-criteria (eResults in the Supplement). Moreover, 9 of 119 CHR-P participants made a transition to psychosis (mean follow-up period, 17.3 months). Eight transitions occurred in the APS-persistent group.

Sensor-Level Analysis
Modulation of spectral power was characterized by early evoked activity (<approximately 250 milliseconds), which is phase locked and time locked to the onset of the stimulus and sustained induced activity that represents non-phase-locked oscillations (>250 milliseconds) (Figure 2). Task-induced γ peak frequency across participants was approximately 62 Hz. A main group effect (cluster F 3,228 = 341.7; P < .001; 95% CI range, −0.0004 to 0.002) for 57-to 67-Hz power was found over occipital and parietal-temporal regions (Figure 2A), with no differences in any other frequency range. Post hoc test results revealed significantly increased γ power for CHR-N vs HC over superior occipital-parietal regions (cluster t 85 = 48.4; P = .03; 95% CI range, 0.0147-0.0453) and significantly decreased γ power (cluster t 73 = −50.7; P = .02; 95% CI range = 0.0143-0.0257) over inferior occipital regions for FEP compared with HC (eFigure 1 in the Supplement).

Behavioral and Magnetoencephalographical Parameters Associated With APS Persistence in the CHR-P Group
Intertrial phase coherence data (30-50 Hz; 125-200 milliseconds) from 10 occipital ROIs, accuracy, RTs, and RT variability were entered into a regression model to predict persistence of APS criteria in the CHR-P group. Only γ-band ITPC  activity contributed significantly to the model. Specifically, ITPC data from the left/right cuneus and left middle occipital gyrus ROIs led to a significant model (χ 2 3 = 14.4; P = .002) that explained 22.2% of the variance (Nagelkerke R 2 = 0.222). The associated ROC curve was also significant ( Figure 4A: area under the curve, 0.728; 95% CI, 0.612-0.841; P = .001) (eMethods and eResults in the Supplement).

Regions of Interest: Baseline Power Spectra and ERF Responses
No group differences in baseline spectral power  or ERF amplitudes were observed in any visual cortex ROI (eFigures 3 and 4 in the Supplement).

Local and Long-range Oscillations in CHR-P Subgroups
We examined differences between CHR-P subgroups (CAARMS n = 34; SPI-A n = 30; CAARMS/SPI-A n = 55) in MEG activity and behavioral and clinical parameters (eResults, eFigure 6, and eTables 1-3 in the Supplement

Discussion
This study examined neural oscillations during visual processing with a state-of-the art MEG approach to investigate whether emerging psychosis is associated with aberrant oscillatory activity in visual cortex as well as the functional relevance of impaired neural oscillations. Specifically, our data highlight a reduction of phase locking of high-frequency oscillations, a measure of the variability of an ongoing oscillation across trials, 48 in visual cortices as well as impaired longrange connectivity in CHR-P participants. Importantly, ITPC deficits were also associated with persistent APS, providing important evidence for the role of high-frequency oscillations in clinical staging of emerging psychosis. Further evidence for the functional relevance of β/γband phase locking are significant correlations with RTs, severity of APS, and the combination of SPI-A/CAARMS criteria as well as GAF-scores across participants. In addition, β/γband ITPC was associated with induced γ-band power, highlighting the importance of the integrity of early visual processing for large-scale cognition and functioning. These data are consistent with previous findings that have identified associations between compromised sensory processing, impaired functioning, and cognitive deficits in schizophrenia. [50][51][52] Comparisons between FEP and CHR-P groups revealed overlapping and distinct oscillatory signatures. Induced γ-band oscillations were prominently impaired in the FEP group in visual areas, which was not observed in CHR-P participants. Both groups were characterized by impaired long-range connectiv-ity between visual and frontal cortices while the FEP group also showed reduced visual cortex connectivity. An influential model in schizophrenia has been the disconnectivity hypothesis 36 as well as the notion of reduced cognitive control mediated by frontal cortices. 53 Our GC data are consistent with these hypotheses, suggesting a shared feature of both FEP and at-risk participants is the presence of impaired connectivity between sensory regions and frontal cortices.
Impairments in high-frequency oscillations showed a considerable degree of specificity. First, ITPC impairments were only found for activity in the β/γ-band range but not for lower frequencies. Together with the large reductions in induced γ-band activity in the FEP group, these data highlight the unique contribution of high-frequency oscillations toward circuit impairments in emerging psychosis. Second, the CHR-N group showed intact behavioral task parameters as well normal power and phase of high-frequency oscillations in visual cortices and long-range connectivity with only evidence for a circumscribed impairment in local connectivity in visual cortex.
Thus, EEG/MEG readouts could potentially inform clinical decision-making and search for novel treatment opportunities. The search for biomarkers that have prognostic utility and could guide treatments in emerging psychosis is an important objective of research. 22 This study highlights that impaired γ-band ITPC differentiates between CHR-P individuals who have a high likelihood of persistent APS and transition to psychosis vs CHR-P individuals who show more benign APS. On the other hand, reductions in induced γ-band power emerged as a specific signature of FEP, suggesting that impaired γ-band oscillations could serve as a biomarker for established psychosis that warrants more aggressive treatments, such as antipsychotic medications.
Our observations of increased variability in the timing of β/γ-band oscillations is consistent with formulations that have implicated aberrant glutamatergic and GABAergic neurotransmission as key mechanism for circuit dysfunctions in psychotic disorders. 17,54,55 Specifically, an increase in variability of neuronal responses can be elicited by N-methyl-Daspartate receptor hypofunction, 56 suggesting that elevated excitability in sensory regions during the early stages of psychosis may lead to favorable conditions for altered network dynamics to emerge. Moreover, deficits in high-frequency oscillations highlight the contribution of specific GABAergic interneurons, such as parvalbumin or somatostatinexpressing interneurons 13,57 that are impaired in visual areas in schizophrenia. 58 In addition to the pharmacologic correction of aberrant circuit dynamics, it is also conceivable that interventions that improve the fidelity of sensory processing through cognitive remediation 59 or brain stimulation 60 could potentially prevent the progression of circuit dysfunctions from sensory areas to more extended networks.

Limitations
This study has several limitations. Although we can predict the persistence of subthreshold psychotic symptoms through MEG data in our CHR-P cohort, further follow-up data are required to test whether abnormalities in high-frequency oscillations can predict transition to psychosis as well as the

Conclusions
In summary, this advanced MEG analysis provides, to our knowledge, the first comprehensive investigation into the oscillatory signatures during different stages of early psychosis. Specifically, we can show that the timing of high-frequency oscillations in visual cortices is the first impairment to emerge in CHR-P participants in combination with abnormal longrange connectivity. Patients with FEP were characterized by a pronounced reduction in the power of induced γ-band oscillations in combination with reduced β/γ-band ITPC as well as local and long-range connectivity. Importantly, impaired γ-band IPTC-values were associated with the persistence of subthreshold psychotic experiences, suggesting that γ-band oscillations could constitute a possible biomarker for clinical staging of emerging psychosis. Future studies and preclinical research should therefore focus on the circuit-mechanisms mediating precise coordinated neural responses that could offer targets for preventive approaches.