TL;DR
This study reveals that specific patterns of brain functional organization, measured by "functional gradients," are consistently altered in individuals with schizophrenia. These altered patterns are not only reproducible across different datasets but can also reliably predict the severity of both clinical and cognitive symptoms, offering a potential biomarker for the disorder.
What Problem Does This Paper Address?
Schizophrenia (SZ) is a complex mental disorder characterized by a wide range of symptoms, including positive (hallucinations, delusions), negative (apathy, social withdrawal), and cognitive deficits. Despite decades of research, the precise neurobiological mechanisms underlying these diverse symptoms remain unclear. Traditional neuroimaging approaches often rely on defining discrete brain regions, which may oversimplify the continuous and hierarchical organization of the brain. This paper aims to address this gap by using a novel approach, "functional gradient," to characterize brain organization and investigate whether disruptions in this organization are consistently present in SZ and linked to symptom severity.
Methods
The researchers employed a technique called functional gradient mapping, derived from resting-state functional magnetic resonance imaging (rs-fMRI) data. This method moves beyond discrete brain regions to represent the brain's organization as continuous axes of variation. Specifically, they calculated functional gradients across the cerebral cortex for each individual. They then performed vertex-wise comparisons between individuals with SZ and healthy controls (NC) to identify differences. To ensure the reliability of their findings, they conducted reproducibility and robustness analyses, including using different methodological choices (e.g., global signal regression) and control analyses. Finally, they used generalized additive models (GAMs) to build predictive models linking the identified abnormal gradient patterns to the severity of clinical (PANSS scores) and cognitive (CFT scores) symptoms in SZ.
Datasets
The study utilized two independent, large-scale, multi-site case-control datasets:
- Discovery Dataset: 736 subjects (400 SZ, 336 NC)
- Replication Dataset: 541 subjects (279 SZ, 262 NC)
These datasets were collected from multiple hospitals in China, enhancing the generalizability of the findings.
Key Insights and Ideas
- Brain organization is continuous, not discrete: The functional gradient approach treats brain organization as a spectrum rather than distinct regions, offering a more nuanced view of brain function.
- Functional gradients are reproducible in SZ: The study found consistent disruptions in functional gradients in individuals with SZ, which were not only present in the discovery dataset but also replicated in an independent dataset.
- Abnormal gradients predict symptoms: The identified gradient abnormalities are not just markers of SZ but can also predict the severity of specific clinical and cognitive symptoms, suggesting a direct link between brain organization and symptom expression.
- Negative and general psychopathology symptoms are predictable: The findings highlight that negative and general psychopathology symptoms, along with cognitive deficits, are more reliably predicted by functional gradients than positive symptoms.
Key Results
- The first functional gradient explained 25% of the connectome variance and showed highly consistent patterns between SZ and NC groups.
- Ten significant clusters of between-group differences in gradient were identified in the discovery dataset. These findings were largely reproducible in the replication dataset, with six clusters remaining significant even after stringent correction.
- Robustness analyses confirmed the reliability of these findings across different methodological choices.
- Cluster-based gradient features significantly predicted PANSS total scores (discovery: r=0.28, replication: r=0.26) and subscale scores for negative (discovery: r=0.22, replication: r=0.22) and general psychopathology (discovery: r=0.26, replication: r=0.23).
- CFT scores (cognitive processing speed) were also predicted (discovery: r=0.25, replication: r=0.17).
- Interestingly, PANSS positive scale scores could not be predicted, suggesting different underlying neurobiological mechanisms for positive symptoms.
- Gradient in the dorsolateral prefrontal cortex (DLPFC) was consistently identified as the most important feature for predicting PANSS domains.
Limitations and Future Work
The authors acknowledge that the cross-sectional design of the study limits their ability to establish causal relationships between neuroimaging findings and clinical assessments. They suggest that longitudinal studies are needed to track how gradients change with ongoing symptoms and to further test their predictive power over time. Future work could also explore the specific neurobiological mechanisms underlying these gradient alterations and their implications for developing targeted interventions.
Summary updated