Altered Brain-Behavior Association During Resting State is a Potential Psychosis Risk Marker

Abstract

Alterations in cognitive and neuroimaging measures in psychosis may reflect altered brain-behavior interactions patterns accompanying the symptomatic manifestation of the disease. Using graph connectivity-based approaches, we tested the brain-behavior association between cognitive functioning and functional connectivity at different stages of psychosis. We collected resting-state fMRI of 204 neurotypical controls (NC) in two independent cohorts, 43 patients with chronic psychosis (PSY), and 22 subjects with subthreshold psychotic symptoms (STPS). In NC, we calculated graph connectivity metrics and tested their associations with neuropsychological scores. Replicable associations were tested in PSY and STPS and externally validated in three cohorts of 331, 371, and 232 individuals, respectively. NC showed a positive correlation between the degree centrality of a right prefrontal-cingulum-striatal circuit and total errors on Wisconsin Card Sorting Test. Conversely, PSY and STPS showed negative correlations. External replications confirmed both associations while highlighting the heterogeneity of STPS. Group differences in either centrality or cognition alone were not equally replicable. In four independent cohorts totaling 1,203 participants, we identified a replicable alteration of the brain-behavior association in different stages of psychosis. These results highlight the high replicability of multimodal markers and suggest the opportunity for longitudinal investigations that may test this marker for early risk identification.

Keywords: at risk mental states; brain networks; brain‐behavior relationship; chronic psychosis; executive function; neuropsychology.

Figure 1

Outline of the study. The main phases of the study are shown in the figure. For each cohort, the arrows between the brain and the cognitive test indicate the directionality of the association (pFDR < 0.05) between the degree centrality of right prefrontal‐cingulum‐striatal brain regions and cognitive functioning. Arrows pointing in the same direction represent a positive brain‐behavior association, whereas arrows pointing in opposite directions represent a negative brain‐behavior relationship. Abbreviations: MI: Main investigation of the study using UNIBA dataset (n = 269); MI‐NCD: Neurotypical controls in discovery cohort of the main investigation (n = 117); MI‐NCR: Neurotypical controls in within‐site replication cohort of the main investigation (n = 87); MI‐PSY: Patients with psychosis of the main investigation (n = 43); MI‐STPS: Individuals with subthreshold psychotic symptoms of the main investigation (n = 22); PNC: Philadelphia Neurodevelopmental Cohort (n = 331); PNC‐TD: Individuals with typical development in PNC dataset (n = 57); PNC‐STPS: Individuals with a developmental trajectory toward psychotic disorders in PNC dataset (n = 45); PNC‐OD: Individuals with a developmental trajectory toward other psychiatric disorders in PNC dataset (n = 229); PRONIA: Personalised Prognostic Tools for Early Psychosis Management dataset (n = 371); PRONIA‐NC: Neurotypical controls in PRONIA dataset (n = 202); PRONIA‐PSY: Patients with psychosis in PRONIA dataset (n = 80); PRONIA‐STPS: Individuals with subthreshold psychotic symptoms in PRONIA dataset (n = 89); LIBD: Lieber Institute for Brain Development (n = 232); LIBD‐NC: Neurotypical controls in LIBD dataset (n = 149); LIBD‐PSY: Patients with psychosis in LIBD dataset (n = 54); LIBD‐SIB: Unaffected siblings of patients with schizophrenia in LIBD dataset (n = 29).

Figure 2

The 25 meta‐ROIs obtained through cluster analysis in MI‐NCD. Hierarchical Ward clustering analysis was performed using the coordinates of 144 ROIs from the Dosenbach atlas (x, y, z based on the MNI atlas), along with the mean values of ROI‐betweenness centrality and ROI‐degree centrality, calculated in MI‐NCD cohort (n = 117). More details are reported for meta‐ROI 3 of the right hemisphere, i.e., the prefrontal‐cingulum‐striatal meta‐ROI. Capital letters refer to brain regions included in the prefrontal‐cingulum‐striatal meta‐ROI: A. Anterior cingulate cortex; B. Superior frontal gyrus; C. Middle frontal gyrus; D. Cingulate gyrus; E. Caudate. Meta‐ROIs displayed using BrainNet Viewer. Abbreviation: MI‐NCD: Neurotypical controls in discovery cohort of the main investigation.

Figure 3

Association between the degree centrality of the prefrontal‐cingulum‐striatal meta‐ROI and the executive performance. A) Association between the meta‐ROI degree centrality and the standardized residuals WCST‐TE in main investigation (pFDR < 0.05). B) Association between the meta‐ROI degree centrality and the standardized residuals PCET‐IR in PNC dataset (P < 0.05). Individuals with typical development (PNC‐TD) are plotted in the first box on the left, in the NC column. Individuals with a developmental trajectory toward psychotic disorders (PNC‐STPS) are plotted in the third box on the left, in the STPS column. C) Association between the meta‐ROI degree centrality and the standardized residuals DSST‐ICM in PRONIA dataset (P < 0.05). D) Association between the meta‐ROI degree centrality and the standardized residuals of WCST‐TE in LIBD dataset (P < 0.05). In graphs are plotted the 95% confidence intervals. Abbreviations: NC: Neurotypical controls; PSY: Patients with psychosis; STPS: Individuals with subthreshold psychotic symptoms; OD: Individuals with a developmental trajectory toward psychiatric disorders other than psychosis in PNC dataset (n = 229); SIB: Unaffected siblings of patients with schizophrenia in LIBD dataset (n = 29); MI‐NCD: Neurotypical controls in discovery cohort of the main investigation (n = 117); MI‐NCR: Neurotypical controls in within‐site replication cohort of the main investigation (n = 87); SCZ: Patients with schizophrenia (n = 24); BD: Patients with bipolar disorder (n = 19); PNC: Philadelphia Neurodevelopmental Cohort (n = 331); PNC‐TD: Individuals with typical development in PNC dataset (n = 57); PNC‐STPS: Individuals with a developmental trajectory toward psychotic disorders in PNC dataset (n = 45); PRONIA: Personalised Prognostic Tools for Early Psychosis Management dataset (n = 371); LIBD: Lieber Institute for Brain Development (n = 232); DC: Degree centrality; WCST‐TE: Number of total errors at Wisconsin Card Sorting Test; PCET‐IR: Proportion of incorrect responses at Penn Conditional Exclusion Test; DSST‐ICM: Inverted T‐score of correct symbol‐number matching at Digit Symbol Substitution Test.

Figure 4

Graphical representation of the interpretative model. In NC, the reduction in connectivity reflected a functional response to the rest condition, indicating an efficient disengagement of task‐dependent regions. In contrast, clinical populations exhibited reduced connectivity, suggesting pathological dysconnectivity of networks, where individuals with less compromised brain networks showed better performance compared to those with global dysconnectivity.

Figure 5

Degree and betweenness centrality mean values for each Dosenbach's ROIs within the MI‐NCD cohort. A) Panel shows the average betweenness centrality value of each ROI. B) Panel shows the average degree centrality value of each ROI. The color bars indicate the betweenness and degree of each ROI in the whole brain. Two lateral and two axial views are showed. Abbreviations: MI‐NCD: Neurotypical controls in discovery cohort of the main investigation (n = 117); L: Left; R: Right.