Connectome dysfunction in patients at clinical high risk for psychosis and modulation by oxytocin

Abnormal functional connectivity is a strong biological marker for psychosis. Alterations in the brain's functional network organization (functional connectome) are also observed before psychosis onset in individuals at Clinical High Risk (CHR-P). These individuals exhibit attenuated psychotic symptoms, emotional, cognitive, and functional impairments, with a 20% two-year risk of transitioning to frank psychosis. Currently, there are no licensed pharmacological treatments for CHR-P, highlighting a significant unmet clinical need. A better understanding of the connectomic abnormalities contributing to psychosis risk and the potential ameliorative effects of experimental therapeutics is crucial. Imaging studies have revealed brain-wide and regional dysfunction in CHR-P and established psychosis, particularly in the hippocampus, striatum, thalamus, and frontal cortex—core components of influential circuit models of psychosis pathophysiology. Aberrant resting-state functional connectivity between brain regions emphasizes that psychosis-related dysfunction isn't solely explained by spatially discrete differences in neural activation. Psychosis-related dysconnectivity is found within and between large-scale networks, particularly involving frontoparietal, default mode, and salience networks. This suggests that network-based approaches are necessary to fully understand brain dysfunction associated with psychosis. Graph theory applied to neuroimaging data offers new possibilities for exploring aberrant functional brain network organization. Brain regions are represented as nodes, connections as edges, and modules as communities of highly interconnected nodes. The brain's topology shapes its information transfer capacity and influences higher-order functions. Disruptions to the balance of integration and segregation can lead to information transfer loss, cognitive dysfunction, and psychotic phenomenology. Studies suggest that people with established psychosis show abnormalities in global and local topological properties, including reduced small-worldness, clustering, hubness, and modularity, as well as changes in global and local efficiency. However, the network dysfunction preceding psychosis onset in the CHR-P state is less well-characterized. Studies have shown that abnormal modular organization in CHR-P individuals at baseline is associated with a higher transition rate to frank psychosis. Further work demonstrates that CHR-P individuals who transition have altered topological centrality in frontal and anterior cingulate regions, reduced global efficiency and clustering (though global differences aren't always found), regional changes in nodal efficiency correlating with symptom severity, and extensive reorganizations of network community structure across most large-scale resting-state networks. These data suggest that further mapping of CHR-associated functional connectomic alterations—and their response to experimental therapeutics—would enrich understanding of the neurobiological mechanisms driving psychosis onset and illuminate novel treatment targets. Oxytocin, a neuropeptide, is a potential novel treatment with neurobehavioral effects beneficial for CHR-P individuals, including anxiolytic effects, modulation of social-emotional cognition, and hypothalamic-pituitary-adrenal axis regulation. Previous work demonstrated that oxytocin modulates frontal activation during mentalizing, anterior cingulate neurochemistry, and resting cerebral blood flow in the hippocampus and other regions in people at CHR-P. Evidence suggests that oxytocin modulates connectivity within resting-state networks in healthy volunteers and normalizes aberrant connectivity in several clinical populations. A recent study in healthy males showed that a single oxytocin dose modulated local functional network topology, including in regions and networks implicated in psychosis risk. This raises the possibility that oxytocin may ameliorate the connectomic dysfunction in CHR-P individuals. Preclinical work shows that oxytocin targets fast-spiking GABAergic interneurons to enhance spike transmission fidelity and temporal precision, augmenting signal-to-noise ratio in information transfer across brain networks. However, it's unclear whether oxytocin will have similar connectomic effects in CHR-P patients as in healthy controls. This study used a data-driven approach combining multi-echo resting-state blood-oxygen-level-dependent (BOLD) functional magnetic resonance imaging (fMRI) with graph-theory modeling to investigate differences in global and regional topology of the functional connectome related to CHR-P status, the effects of intranasal oxytocin vs. placebo, and group x treatment interactions to identify clinical status-specific effects of oxytocin.

The introduction extensively reviews the existing literature on connectome dysfunction in psychosis and CHR-P, citing numerous studies on altered functional connectivity, graph theoretical analyses of brain networks in psychosis, and the potential therapeutic role of oxytocin. The review highlights the lack of effective pharmacological treatments for CHR-P and the need for a better understanding of the underlying neurobiological mechanisms. Key papers are cited to support the hypothesis that aberrant functional connectivity plays a significant role in the development of psychosis and that oxytocin may offer a novel therapeutic approach by modulating these networks. The review establishes the context for the current study by summarizing previous research on altered brain network organization in CHR-P individuals and the known effects of oxytocin on brain function and connectivity in healthy individuals and various clinical populations.

This study combined data from two related studies: one involving 30 male CHR-P individuals and another with 17 healthy male controls. CHR-P status was determined using the Comprehensive Assessment of At-Risk Mental States (CAARMS) criteria. Both studies employed a randomized, double-blind, placebo-controlled crossover design with intranasal oxytocin (40 IU) and placebo challenges. Multi-echo resting-state fMRI data was acquired approximately 1 hour post-dosing. Functional connectivity matrices were generated by extracting BOLD time courses for each region of interest (ROI) using the Desikan-Killiany atlas (66 cortical regions) and the Harvard-Oxford atlas (subcortical regions), resulting in 83 nodes after removing two nodes due to signal dropout. Bivariate Pearson's correlation matrices were generated, and Fisher r-z transformations were applied. To control for differences in overall connectivity strength, mean functional connectivity was examined. Graph theoretical analysis using an undirected signed weighted approach was performed on the connectivity matrices. Equi-sparse networks were constructed by retaining a fixed percentage (5-34%) of edges. Global and nodal graph metrics (global efficiency, betweenness centrality, local efficiency, and node degree) were calculated for each sparsity level. The area under the curve (AUC) was used to summarize the values over the range of sparsity thresholds. Statistical analyses (t-tests and Network Based Statistics – NBS) examined group differences, treatment effects, and group x treatment interactions at global, node, and edge levels. False discovery rate (FDR) correction was used to control for multiple comparisons. Finally, the overlap between significant regions and large-scale resting-state networks (RSNs) from the Yeo et al. atlas was quantified using the Dice-kappa coefficient. This provided a qualitative contextualization of findings and facilitated comparison with previous literature.

The study yielded several key findings: **Group Effects (CHR-P vs. Controls):** No significant group differences were found in global metrics (mean functional connectivity and global efficiency). However, CHR-P individuals showed greater betweenness centrality and node degree in several frontal regions compared to controls, primarily within the frontoparietal network. These localized differences suggest a reorganization of brain network architecture in CHR-P individuals, without affecting global information transfer capacity. **Treatment Effects (Oxytocin vs. Placebo):** No significant effects were observed on global metrics. Oxytocin, compared to placebo, increased betweenness centrality in the brainstem and node degree in the left precentral gyrus across both groups, primarily within the ventral attention network. This suggests a general effect of oxytocin on these regions irrespective of clinical status. **Interaction Effects (Group x Treatment):** This is the study's most novel finding. Significant interaction effects were observed in numerous subcortical (thalamus, pallidum, nucleus accumbens) and cortical regions, mainly within the default mode network. In CHR-P individuals, oxytocin increased local graph metrics (betweenness centrality, local efficiency) in several regions where it decreased them in controls. This indicates that oxytocin's effects on network topology are distinct in CHR-P patients compared to healthy individuals. The effects in the thalamus, pallidum, and nucleus accumbens are particularly notable given their roles in psychosis pathophysiology. The overlap analysis with RSNs showed that group effects primarily overlapped with the frontoparietal network; treatment effects with the ventral attention and somatomotor networks; and interaction effects with the default mode network.

The findings demonstrate that CHR-P individuals exhibit altered local network topology, particularly within the frontoparietal network, suggesting a reorganization of network architecture rather than a global disruption of information processing. The effects of oxytocin on network topology, predominantly increasing local graph metrics, primarily within the ventral attention network, seem to be present in both groups, although the effects are considerably more pronounced in the clinical group. The most significant finding is the presence of widespread interaction effects, revealing that oxytocin modulates network properties in a clinical status-specific manner. This clinical status-specific response in key brain regions implicated in psychosis pathophysiology (thalamus, striatum, entorhinal cortex) suggests a potential therapeutic mechanism for oxytocin in CHR-P. The divergent effects of oxytocin in these regions might reflect a baseline network (dis)organization influencing the net effect of exogenous oxytocin. The results strengthen the rationale for further investigation into oxytocin's therapeutic potential for CHR-P.

This study provides novel insights into aberrant functional brain network organization in CHR-P and demonstrates that oxytocin modulates network topology in a clinical status-specific manner. The interaction effects observed in regions crucial for psychosis pathophysiology suggest a potential therapeutic role for oxytocin in this population. Future research should focus on larger sample sizes, longitudinal studies, and investigation of the underlying neurochemical mechanisms to further explore oxytocin's therapeutic potential for CHR-P.

The study has several limitations. First, data from CHR-P and control groups were from slightly different studies with variations in experimental design. This could influence the observed group differences. Second, the study was not pre-registered, although this is typical of exploratory data-driven studies. Third, certain clinical and demographic variables were not collected in the control group, meaning that the influence of such potential confounders cannot be entirely ruled out. Finally, while the sample size was relatively large for this type of study, some graph analytic results suggest even larger sample sizes are needed for optimal reliability.