Lioba C S Berndt1,2,5 Rosina M Diebel1 Nicholas A Donnelly2,4 Jeremy Hall3 Marianne B M van den Bree3 Rick A Adams5,6 Alexander D Shaw1 Matthew W Jones2
1 Department of Psychology, Faculty of Health & Life Sciences, University of Exeter, UK. 2 School of Physiology, Pharmacology and Neuroscience, University of Bristol, UK. 3 Neuroscience and Mental Health Innovation Institute and Centre for Neuropsychiatric Genetics and Genomics, Cardiff University, Cardiff, UK. 4 Avon and Wiltshire Mental Health Partnership NHS Trust, Bristol, UK. 5 Hawkes Institute, University College London, UK. 6 Institute of Cognitive Neuroscience, University College London, UK.
Background: 22q11.2 deletion syndrome (22q11.2DS) is associated with a spectrum of psychiatric outcomes and is the strongest known genetic risk factor for schizophrenia. Sleep disturbances and sleep EEG alterations (including in thalamocortical slow-wave and spindle oscillations) feature early in 22q11.2DS, correlating with psychopathology and potentially reflecting the neurodevelopmental circuit dysfunction mediating psychiatric risk. However, scalp EEG analyses alone lack mechanistic resolution, limiting the understanding required for precision treatment development. Computational modelling offers a framework to overcome this by inferring underlying synaptic mechanisms from EEG, bridging genetic risk with circuit-level signatures and predicting therapeutic targets.
Methods: We applied conductance-based thalamocortical Dynamic Causal Modelling (DCM) to sleep and wake (resting state) EEG from young people with 22q11.2DS (n=28, M=14.6, SD=3.4) and sibling controls (n=17, M=13.7, SD=3.4). We aimed to: (a) identify key synaptic parameters distinguishing 22q11.2DS neurophysiology across vigilance states, utilising hierarchical Bayesian inference (Parametric Empirical Bayes, PEB) for robust group comparisons (parameters exceeding 99% posterior probability considered significant); (b) evaluate state-dependent associations between these synaptic parameters and a range of clinical/cognitive measures using LASSO regression followed by appropriate statistical models with FDR correction; and (c) predict which synaptic adjustments may shift 22q11.2DS EEG signatures towards those of controls by simulating both parameter-specific perturbations and system-wide scaling of AMPA, NMDA, GABA-A, and GABA-B receptor properties.
Results: Bayesian model comparison across 15 architectures confirmed that optimally capturing state-dependent EEG spectral dynamics required incorporation of all four major receptor types (AMPA, NMDA, GABA-A, GABA-B). This model achieved the highest posterior probability (0.76) and the lowest Bayesian Information Criterion (iBIC = 24.3). It provided substantially stronger support compared to alternatives, particularly those relying on single receptor systems (e.g., NMDA-only, iBIC = 679.7; GABAB-only, iBIC = 573.8), highlighting the necessity of modeling multiple interacting receptor dynamics. Synaptic dysfunction in 22q11.2DS was more pronounced during sleep, with alterations (identified via PEB, Pp > 0.99) becoming progressively more extensive from wakefulness through NREM to REM sleep. We identified distinct, state-dependent clinical correlates (e.g., stronger N3 thalamocortical connectivity linked to reported sleep problems, r = 0.57, pFDR < 0.05). Parameter-specific perturbation analyses highlighted that adjustments in specific connections, particularly increasing gain in NMDA-mediated superficial pyramidal connections, were critical for aligning simulated EEG spectra with control patterns across multiple sleep stages. System-wide perturbation analysis consistently revealed that increasing overall NMDA system gain produced the most effective alignment of simulated EEG spectra with control patterns across all vigilance states, with the strongest effect observed during N3 sleep (ES=0.45, p=0.001).
Conclusion: Computational modelling across all arousal states predicts state-dependent synaptic pathophysiology in 22q11.2DS. This highlights sleep EEG as a sensitive window into underlying circuit dysfunctions, underscoring the importance of sleep neurophysiology in potential biomarkers of complex neurodevelopmental disorders. Our findings provide circuit-level evidence that putative cortical NMDA hypofunction contributes to this syndrome's sleep EEG alterations and clinical symptoms, predicting therapeutic avenues that warrant preclinical and clinical experimentation.