Abstract
The prevalence and significance of schizophrenia-related phenotypes at the population-level are debated. This study assesses whether two recently reported neuroanatomical signatures of schizophrenia—Signature 1 (widespread reduction of gray matter volume) and Signature 2 (increased striatal volume)—could be replicated and detected at the population level. Using independent samples of schizophrenia patients, typically developing youth, and healthy adults, the researchers quantified signature expression using machine learning. Signature 1 was significantly more common in youth with psychosis-spectrum symptoms and was associated with poorer cognitive performance and elevated polygenic risk scores. Signature 2 frequency was similar between typically developing and symptomatic youth and did not show elevated genetic risk. These findings suggest distinct neurobiological vulnerabilities underlying schizophrenia risk in the general population.
Introduction: Addressing Neurobiological Heterogeneity
Schizophrenia is a highly heterogeneous disorder. Clinical symptoms, cognitive profiles, and brain structures vary significantly between patients, often confounding traditional case-control studies that treat "all patients" as a single group. This study sought to replicate two previously identified neuroanatomical subtypes (signatures) in independent samples and investigate their presence in the general population to determine if they signal biological vulnerability to psychosis before clinical diagnosis.
Methods and Datasets
The researchers employed a semi-supervised machine learning pipeline to identify and quantify brain structural patterns:
- HYDRA (Heterogeneity through Discriminative Analysis): Used to find distinct patterns within patient groups that differentiate them from controls.
- Signature Quantification: Support Vector Machines (SVM) estimated the expression strength of two previously identified signatures:
- Signature 1 (S1): Characterized by widespread reductions in gray and white matter volume.
- Signature 2 (S2): Characterized by an increase in the volume of the striatum and internal capsule.
Key Results
Successful Replication and Population Prevalence
The two-subtype model was successfully replicated in the independent patient sample (58% S1, 42% S2). Notably, these signatures are not exclusive to clinical diagnoses; approximately 23-24% of typically developing youth express one of these signatures, supporting the concept of "schizotaxia"—a biological state of risk existing without overt symptoms.
Psychosis and Cognitive Divergence
In the youth cohort (PNC), S1 was significantly more prevalent in individuals reporting psychosis-spectrum symptoms (39.8%) compared to typically developing peers (23.03%). S2 frequency remained similar across groups. Cognitively, S1 was consistently associated with poorer efficiency and accuracy in both youth and adults (z-stat = -4.64 for efficiency). In contrast, S2 expression in adults showed a positive correlation with fluid intelligence (ρ = 0.17), suggesting it may represent a distinct pathological or even compensatory pathway.
Genetic Validation
Adults in the UK Biobank expressing S1 exhibited significantly higher polygenic risk scores for schizophrenia (z-stat = 2.69). Signature 2 did not show this elevated genetic burden, indicating that S1 may more closely align with the inherited genetic risk typically associated with the disorder.
Discussion and Future Directions
While these signatures provide critical clues into the pathophysiology of schizophrenia, several limitations remain. The cross-sectional nature of the data prevents definitive conclusions about whether individuals with these signatures will eventually develop clinical schizophrenia. Furthermore, while S1 is linked to volume loss, the functional implications of S2's enlarged striatum require further investigation via fMRI. Future longitudinal studies are necessary to determine the predictive value of these neuroanatomical markers for clinical outcomes.
Summary updated