We combine data of a group of schizophrenia and HCs across three scales, including blood transcriptome, functional neuroimaging-derived brain phenotypes, and clinical symptoms, showing cross-scale associations that are key for understanding the biological mechanisms of schizophrenia. Integrating multiple abnormalities in brain function, pairwise interaction between MRI and transcriptomics have provided a way to establish the transcriptomic correlates of cerebral alterations in relation to schizophrenia [22]. Our study suggests that blood-sample gene expressions might provide an informative way to study participant-level transcriptome-neuroimaging associations in schizophrenia.
Our reported relationship between DEGs in blood samples and brain functional alterations in schizophrenia implies blood transcriptomic data to be an informative source to study transcriptome-neuroimaging associations in schizophrenia. Genetic association between schizophrenia and cerebral MRI has been demonstrated by means of independent study data (Psychiatric Genomics Consortium, CLOZUK, and UK Biobank) [23]. However, one of the obstacles in identifying transcriptome-neuroimaging associations of the disorder is that transcriptome and MRI data are commonly derived from different sources. Our previous study thus attempted to address this issue by collecting cross-scale data from one cohort of schizophrenia [22], and our observed correspondence between blood DEGs and brain DEGs is in line with the correlation of genetic effects between human blood and brain [24]. Our results of associations between blood DEGs and schizophrenia GWAS further suggests blood DEGs were significantly regulated by common genomic variants in schizophrenia, regardless of the examined ethnics [22]. Therefore, examinations of the participant-level associations between blood gene expression and imaging-derived brain phenotypes in schizophrenia might provide evidence of the disease-involved biological pathway that advances our understanding of the disease mechanisms.
Our RNA-seq analysis of blood from individuals with schizophrenia and demographically matched HCs has revealed a striking imbalance in gene expression, with 73 genes upregulated compared to 921 genes downregulated in schizophrenia. This preponderance of downregulated genes echoes a recent report [25], pointing to several possible etiological factors. Primarily, schizophrenia may involve dysregulation of transcriptional networks, culminating in the suppression of select gene cohorts [22, 26, 27]. Additionally, the chronic inflammatory and oxidative stress states observed in these patients likely contribute to the altered gene expression patterns, a hypothesis supported by our pathway enrichment analyses [28]. Furthermore, neuropsychiatric conditions such as schizophrenia are frequently associated with dysfunctional neuronal circuits. Aberrations in neurotransmitter systems, for example, dopamine and glutamate, may underlie reduced neuronal activity and subsequent downregulation of genes essential for neuronal integrity, synaptic plasticity, and communication. This is corroborated by our findings of a suppressed state in functional brain connectivity among schizophrenia patients, indicative of neuronal injury, neurodegenerative processes, and transcriptional imbalance. These observations align with recent pivotal studies that underscore the complex interplay between gene expression perturbations, neuroinflammation, and neurodegeneration in the pathophysiology of schizophrenia.
These biological underpinnings are compatible with polygenic contribution to the brain disorder of schizophrenia patients in vivo, involving neurochemical disturbance and neurodevelopment [29]. A recent meta-analysis identified brain structure in the cognitive networks, amygdala, hippocampus, and cerebellum typically showing associations with conceptually related cognitive domains in schizophrenia [30]. Meanwhile, declined cognitive ability is significantly associated with schizophrenia polygenic risk scores [31]. Cognitive dysfunction of schizophrenia is related to the altered whole blood gene expression of immunological process [32].
Cross-scale associations between gene expression and brain function were presented, showing that lower expression of DEGs in schizophrenia was associated with the reduced functional activity in the dorsolateral prefrontal cortex and anterior cingulate cortex. These regions have been broadly reported in previous neuroimaging studies in schizophrenia, where decreased neural activity and connectivity were observed in schizophrenia patients compared to healthy controls, underscoring the convergence of genetic and neuroimaging findings in the disorder. Within these regions (e.g., the anterior cingulate cortex and superior temporal cortex), downregulated expressions of immune-associated genes have been also demonstrated in schizophrenia [33]. Moreover, our reported transcriptome-neuroimaging associations at the participant level corroborate recent findings linking transcriptome and brain on the basis of brain spatial variations [34, 35], for instance, genes related to functions in the central nervous system development show overlapped pattern with brain volume changes in schizophrenia [36]. Cortical thickness alterations are in relation to heterogeneity between individuals with schizophrenia, with higher polygenic risk for schizophrenia associated with a greater number of regions with infra-normal deviations in cortical thickness [37]. Such associations between transcription and brain volume might be related to the crucial regulatory role of schizophrenia risk genes in human brain development [38]. In line with this, neuron- and astrocyte/oligodendrocyte progenitors-linked genes were identified based on the transcriptomic and polygenic manifestations of cortical thickness heterogeneity in schizophrenia [39]. Our observed participant-level relationship between gene transcription and brain function in schizophrenia show a great potential of integrating multi-omics features in future works on predictive models in schizophrenia.
Similarly, we showed that the white matter connectivity strength of brain hubs, such as superior frontal and superior parietal regions, is associated with blood gene expression of DEGs in schizophrenia [22]. This finding is in line with previous studies showing brain hubs to show higher heritability compared to the rest of the brain regions and to be related to a tight coupling of transcriptional profiles [40]. Schizophrenia risk genes specifically showed a brain transcriptional profile overlapped with the spatial pattern of disruptions in white matter connectivity [41]. Such an association between schizophrenia gene expression and brain dysconnectivity might reflect the genetic origins of widespread connectivity disruptions observed in schizophrenia, in particular for hubs and the rich club [42].
To elucidate genes concurrently associated with both genetic markers and clinical symptoms of schizophrenia, we intersected the two gene sets, identifying six overlapping genes. Notably, GPAT2, glycerol-3-phosphate acyltransferase 2, exhibited significant correlations with both sets of indicators. GPAT2 is a member of the GPAT family involved in the biosynthesis of triglycerides and phospholipids, predominantly active in the endoplasmic reticulum and mitochondrial outer membrane [43]. While no prior studies have linked GPAT2 to schizophrenia, its role in generating phosphatidic acid -- a precursor to membrane phospholipids essential for cell membrane structure, function, and inter-neuronal interactions -- is intriguing. The downregulation of GPAT2 in schizophrenia suggests neuronal/glial damage and lipid metabolism dysregulation, aligning with schizophrenia pathophysiology. Consequently, GPAT2 may serve as a potential biomarker for aiding in the diagnosis of schizophrenia.
Recent advances in functional neuroimaging have demonstrated how intrinsic connectivity networks can reveal the systems-level mechanisms underlying schizophrenia [44, 45]. Our findings using multiple rs-fMRI metrics align with this perspective - the observed ALFF abnormalities in prefrontal regions correspond to emerging evidence of default mode network dysregulation in schizophrenia [44], while the ReHo disruptions in sensory-motor areas may reflect local connectivity deficits associated with neuroinflammatory processes [46]. Furthermore, the DC alterations we identified in thalamocortical hubs resonate with studies linking these network properties to both genetic risk factors [45] and clinical symptom severity [47]. Together, these converging lines of evidence strengthen our integrated approach combining transcriptomic and neuroimaging data to map schizophrenia's complex pathophysiology.