A significant challenge in understanding the genetic basis of brain disorders lies in the heterogeneity of the human brain. The majority of disease-associated genetic variants are located in non-coding regions, impacting gene regulation in complex ways. Previous studies have attempted to map gene regulatory networks in the brain, but cellular heterogeneity has hampered progress. This study aims to overcome this limitation by focusing on two major cell classes: neurons and glia. By using fluorescence-activated nuclear sorting (FANS) to isolate these cell types, the researchers can generate high-resolution chromosome conformation capture (Hi-C) data to map three-dimensional chromatin architecture within each cell type. This allows for a more precise understanding of how chromatin structure influences gene expression and contributes to disease risk. The integration of this data with transcriptomic and enhancer profiles will provide a more complete picture of cell-type-specific gene regulation and its implications for brain disorders.

The authors review previous studies that have integrated multi-dimensional datasets (transcriptomic, epigenomic, and higher-order chromatin interactions) to identify gene regulatory relationships in the human brain. However, they point out the significant challenge posed by cellular heterogeneity, which obscures the underlying regulatory networks. They mention studies that have used Hi-C in iPSC-derived neurons and astrocytes, but these are limited by their in vitro nature. More recent work has provided promoter interaction profiles from neurons, oligodendrocytes, and microglia in the postnatal human cerebral cortex, but lack the comprehensive 3D chromatin architecture detail. This current study aims to advance this knowledge by applying Hi-C to purified neurons and glia from the adult human brain, providing a more realistic and complete picture of gene regulation in these cell types.

The researchers used FANS to isolate nuclei from postmortem dorsolateral prefrontal cortex (DLPFC) samples. They categorized the nuclei into neurons (NeuN+) and glia (NeuN−) and generated genome-wide Hi-C data for each cell type from four individuals. This Hi-C data was analyzed to identify various aspects of 3D chromatin architecture including compartments, topologically associating domains (TADs), frequently interacting regions (FIREs), and gene loops. To refine the neuronal analysis, they integrated the Hi-C data with H3K27ac peak data from glutamatergic and GABAergic neurons. They compared FIREs and super-FIREs between NeuN+ and NeuN− cells and assessed their relationship to cell-type-specific gene expression using single-cell RNA sequencing (scRNA-seq) data. To investigate disease relevance, they overlapped their cell-type specific Hi-C maps with published data on H3K27ac changes in Alzheimer’s disease (AD) and performed H-MAGMA analysis to integrate this with genome-wide association studies (GWAS) data for AD, schizophrenia (SCZ), and bipolar disorder (BD). This analysis mapped genetic risk variants to specific genes via chromatin interactions within the relevant cell types.

The study identified numerous cell-type-specific differences in 3D chromatin architecture. They found extensive compartment switching between NeuN+ and NeuN− cells, with genes in switched compartments showing cell-type-specific expression patterns. They identified thousands of FIREs, with a subset showing cell-type specificity. NeuN+ FIREs were enriched for genes related to neuronal function (synaptic function), while NeuN− FIREs were enriched for genes involved in myelination and glial differentiation. Similarly, super-FIREs displayed cell-type specificity, further emphasizing the distinct regulatory landscapes of neurons and glia. Analysis of promoter-anchored chromatin loops revealed complex enhancer-promoter interactions, and the number of interacting enhancers correlated with gene expression levels. In AD, they observed that hyperacetylated peaks were enriched in NeuN− cells (glia), while hypoacetylated peaks were enriched in NeuN+ cells (neurons). Gene ontology (GO) analysis revealed that the hypoacetylated genes in AD were enriched for synaptic function, while hyperacetylated genes were linked to catalytic activity and glycoprotein binding. GWAS analysis for AD showed enrichment of heritability in NeuN− enhancers, and H-MAGMA analysis identified 181 AD risk genes, many of which were expressed in microglia and enriched in microglial co-expression modules upregulated in AD. For SCZ and BD, the study found heritability enrichment in both NeuN+ cells and, more specifically, in both glutamatergic and GABAergic neurons. However, detailed analysis revealed both shared and distinct cellular targets for SCZ and BD. Parvalbumin-expressing interneurons showed a shared risk signal, while distinct upper-layer neurons were implicated for BD and deeper-layer neurons for SCZ.

This study provides high-resolution maps of chromosome conformation in neurons and glia, demonstrating the importance of cell-type-specific analysis in understanding brain disorders. The finding that AD-associated epigenetic dysregulation affects neurons and oligodendrocytes while genetic risk factors point to microglia suggests distinct mechanisms driving disease pathogenesis. This highlights the complexity of AD, where different glial cells may contribute to the disease through different pathways. Similarly, the refined cellular etiology of SCZ and BD, showing both shared and distinct neuronal subtypes, provides critical insights into the biological basis of these disorders. The shared involvement of parvalbumin-expressing interneurons supports previously observed pathological findings in both conditions. This study emphasizes the importance of integrating high-resolution chromatin interaction data with genetic risk factors to pinpoint cellular mechanisms underlying brain disorders.

This study provides a significant advance in our understanding of cell-type-specific gene regulation in the human brain and its relevance to brain disorders. The high-resolution maps of 3D chromatin architecture in neurons and glia, combined with the integration of GWAS data, reveal distinct cellular etiologies for AD, SCZ, and BD. Future studies using neuronal subtype-specific Hi-C data and CRISPR-Cas9 genome engineering could further validate these findings and explore causal relationships. This work lays a strong foundation for future research aiming to develop targeted therapies for these complex diseases.

The study relies on postmortem brain tissue, which may introduce limitations due to potential artifacts from tissue handling and death. The analysis for SCZ and BD uses NeuN+ Hi-C data with Glu and GABA neuron-specific H3K27ac peaks, while ideally neuronal subtype-specific Hi-C data would be used for a more accurate analysis. Furthermore, the causal relationships inferred from the association between genetic variants and gene expression require further investigation using experimental methods. Finally, the study is limited to a small number of individuals, limiting the generalization of the findings.