In this project, we have used advanced molecular techniques to measure the expression (usage) of different genes across different brain cell types (scRNAseq) in an area called the prefrontal cortex. This area has been for many years shown to be crucial for many symptoms of schizophrenia. A majority of risk genes for schizophrenia are involved in the function of synapses (connections between neurons) and accordingly, when we analyze which brain cell types that express the risk genes we see that neurons are enriched. We have also shown that different cell types have largely non-overlapping sets of risk genes. One outstanding question was how different neurons might be implicated in schizophrenia although the signal comes from synaptic genes that were often thought to be common. In one study from this project, we resolve this by showing that a large majority of synaptic genes are specifically expressed and that there are three different modes of variability (class-dependent, cell type-dependent, and continuous variation). Along these lines, we have within SCHIZTYPE also refined the risk-enrichment analysis (which showed that specific types of neurons were enriched) by analyzing recent large-scale human data sets of the entire brain. This has allowed us to perform a data-driven prioritization for cell types and brain regions which are important for the development of the disorder. This analysis confirmed previous findings of neuronal enrichment but also indicated that, in addition to the prefrontal cortex, the amygdala and hippocampus are important for schizophrenia etiology. The findings from this unbiased approach were confirmed using two independent datasets and methods in a deep neural network analysis of functional Magnetic Resonance Imaging (fMRI) of patients.
To identify convergent risk in excitatory cells in the prefrontal cortex we have performed scRNAseq of cells from five different mouse models of schizophrenia risk. Each mouse model carries a mutation (copy number variation; CNV) that in humans greatly increases the risk for schizophrenia. When comparing the changes in gene expression between the different mouse lines we observed no significant overlap. This is very much in contrast to our hypothesis and we conclude that for large-effect size schizophrenia risk variants the behavioural convergence observed in patients cannot be observed on a molecular level. This suggests that perhaps convergence rather happens on a network level.
Within SCHIZTYPE we have also performed a large-scale scRNAseq study of postmortem prefrontal cortex tissue from patients who suffered from schizophrenia. First, we observed near-to-no overlap with the findings from the five mouse models. On the other hand, we show that neurons and astrocytes (a type of support cell) have perturbations in the energy supply (mitochondrial function) and that the genetic programs that are involved also contain synaptic genes. Importantly, these genetic programs are also enriched for schizophrenia risk arguing that these changes are casual, rather than due to the disease. We have also performed transcriptomic analysis on tissue sections from human postmortem material which confirmed the findings from the scRNAseq study that we do not see any gross changes in the number of different cell types or their location in patients.
We are currently perturbing the expression of important genes that we have identified in mouse and human samples analyzed to test the role of those genes in the cell and how perturbation of one cell can affect the function of neighboring cells.
