Deficits in mediodorsal thalamus-dorsolateral prefrontal cortex (MDT-DLPFC) resting-state functional magnetic resonance imaging (rs-fMRI) connectivity and prefrontal sleep spindles have been reported in chronic and early course schizophrenia. However, the presence of these alterations in clinical high-risk for psychosis (CHR), alongside their relationships with underlying neurotransmission and cognitive function, remains to be established. Thirty-one CHR and thirty-two HC underwent: 1) 7 T rs-fMRI
2) 7 T magnetic resonance spectroscopy imaging (MRSI)
and 3) sleep electroencephalography (EEG). Rs-fMRI connectivity was analyzed by seeding the whole thalamus (WT) and seven thalamic subsections. Spindle duration was computed across all EEG channels. GABA/creatine (Cr) and glutamate/Cr were calculated in DLPFC and MDT. Relative to HC, CHR showed WT-DLPFC hypoconnectivity (p-FDR = 0.001), especially involving MDT-DLPFC (p-FDR <
0.001) and reduced prefrontal spindle duration (t-stat = -2.64, p = 0.010), while no differences were found for MRSI neuro-metabolites. We then performed clustering analysis using rs-fMRI connectivity and spindle duration to identify CHR and HC subgroups and predict their working memory (WM) performance. A cluster with intact rs-fMRI and spindle duration included mostly HC (83.33% purity), while a cluster with both measures altered involved almost entirely CHR (91.66% purity) and showed worse WM performances. We also examined MRSI metabolites' contribution to spindles and rs-fMRI connectivity with a within-group multivariable regression analysis. In HC, but not in CHR, MDT glutamate/Cr negatively predicted spindle duration and positively predicted MDT-DLPFC connectivity. Combined, these findings indicate that a multimodal neuroimaging approach can identify distinct thalamocortical dysfunctions in CHR individuals, thus informing future research aimed at developing personalized interventions in these individuals.