Resting state functional hyperconnectivity within a triple network model in paranoid schizophrenia

Abstract

BackgroundThree large-scale intrinsic connectivity brain networks, the salience network (SN), the central executive network (CEN), and the default mode network (DMN), have fundamental roles in higher cognitive processes. In particular, SN activity is crucial in the process of salience mapping—ie, allocating attentional resources to external and internal stimuli and attributing salience to the stimuli. Recent functional neuroimaging studies suggest that the core psychopathological changes in schizophrenia could arise from aberrant connectivity (dysconnectivity) within large scale networks in the brain. However, studies have yielded inconsistent results and have largely been unsuccessful at mapping these abnormalities to core psychopathology. The triple network model of aberrant saliency mapping suggests that the dysconnectivity underlying the psychopathology could lie within these three core brain networks. We hypothesised that intrinsic connectivity within these brain networks is altered in schizophrenia when compared with healthy adults. MethodsWe assessed 37 medicated patients with paranoid schizophrenia (Diagnostic and Statistical Manual of Mental Disorders IV-295.3) and 37 age-matched and sex-matched controls from the Center for Biomedical Research Excellence dataset. We did a group-level independent component analysis on resting-state functional MRI to extract 30 group-level spatial intrinsic connectivity (IC) maps (FSL, FMRIB Software Library). From the group-level IC maps, we identified maps corresponding to the SN, DMN, and CEN. Then, we estimated a version of the group-level spatial map for each individual using the dual regression method and established between-group differences using permutation tests. Using a sparse logistic regression method, we also assessed whether IC maps could predict group membership (patient vs healthy controls) and, with a similar method, whether the IC maps could predict psychotic symptomatology (the positive and negative syndrome scale [PANSS]). FindingsThe schizophrenia group showed stronger functional connectivity within the left CEN, SN, and DMN than did controls. When we used leave-one-out cross-validation, the SN map best discriminated patients from controls with 82·4% accuracy, 94·6 % sensitivity, and 70% specificity. In addition, the SN maps predicted the severity of positive symptoms score on the PANSS (r2=0·95, p<0·0001), particularly delusions (r2=0·38, p=0·003). DMN hyperconnectivity predicted negative symptoms score (r2=0·98, p=0·0005). InterpretationThe SN, CEN, and DMN in paranoid schizophrenia are characterised by significant hyperconnectivity. SN hyperconnectivity predicted positive symptoms, especially delusions. This hyperconnectivity seems to be a prominent feature of paranoid schizophrenia and could represent the pathophysiological change underlying the psychopathology. The cross-sectional nature of our data means that causal associations are speculative. Nonetheless, our results suggest that hyperconnectivity within the SN could be a potential biomarker of schizophrenia. FundingJim Gatherall Travelling Fellowship awarded to RK to attend Stanford Cognitive and Systems Neuroscience laboratory.