A data-driven single-cell and spatial transcriptomic map of the human prefrontal cortex

INTRODUCTION The cortical layers of the human neocortex were classically defined by histological distinction of cell types according to size, shape, and density. However, emerging single-cell and spatially resolved transcriptomic technologies have facilitated the identification of molecularly defined cell populations and spatial domains that move beyond classic cell type definitions and cytoarchitectural boundaries. RATIONALE Given the close relationship between brain structure and function, assigning gene expression to distinct anatomical subdivisions and cell populations within the human brain improves our understanding of these highly specialized regions and how they contribute to brain disorders. We sought to create a data-driven molecular neuroanatomical map of the human dorsolateral prefrontal cortex (DLPFC) at cellular resolution using unsupervised transcriptomic approaches to identify spatial domains associated with neuropsychiatric and neurodevelopmental disorders. RESULTS We generated complementary single-cell and spatial transcriptomic data from 10 adult, neurotypical control donors across the anterior-posterior axis of the DLPFC. Unsupervised spatial clustering revealed fine-resolution data-driven spatial domains with distinct molecular signatures, including deep cortical sublayers and a vasculature-enriched meninges layer. Cell type clustering of single-nucleus RNA-sequencing (snRNA-seq) data revealed 29 distinct populations across seven broad neuronal and glial cell types, including 15 excitatory subpopulations. To add cellular resolution to our data-driven molecular atlas, we took two complementary approaches to integrate single-cell and spatial transcriptomics data. First, we used our previously developed spatial registration framework to map the paired snRNA-seq data to specific unsupervised spatial domains, providing anatomy-based laminar identities to excitatory neuron subpopulations. Second, we used three existing spot-level deconvolution tools to computationally predict the cell type composition of spatial domains on the basis of the paired snRNA-seq reference data. These tools were rigorously benchmarked against a newly generated gold-standard reference dataset acquired with the Visium Spatial Proteogenomics assay, which enabled us to label and quantify four broad cell types across the DLPFC on the basis of protein marker expression, including neurons, oligodendrocytes, astrocytes, and microglia. Using these approaches, we identified the proportion of cell types in each spatial domain and showed that these proportions were consistent across individuals and the DLPFC anterior-posterior axis. We demonstrated the clinical relevance of our highly integrated molecular atlas using cell-cell communication analyses to spatially map cell type–specific ligand-receptor interactions associated with genetic risk for schizophrenia (SCZ). For example, we mapped the interaction between ephrin ligand EFNA5 and ephrin receptor EPHA5 to deep-layer excitatory neuron subtypes and spatial domains. To leverage the rich single-cell data generated by PsychENCODE Consortium companion studies, we spatially registered eight DLPFC snRNA-seq datasets collected across the consortium in the context of different neuropsychiatric disorders and demonstrated a convergence of excitatory, inhibitory, and non-neuronal cell types in relevant spatial domains. Using PsychENCODE Consortium and other publicly available gene sets, we further demonstrated the clinical relevance of our data-driven molecular atlas by mapping the enrichment of cell types and genes associated with neuropsychiatric disorders—including autism spectrum disorder, posttraumatic stress disorder, and major depressive disorder—to discrete spatial domains. CONCLUSION Our study identified high-resolution, data-driven spatial domains across the human DLPFC, providing anatomical context for cell type–specific gene expression changes associated with neurodevelopmental disorders and psychiatric illness. We provide a roadmap for the implementation and biological validation of unsupervised spatial clustering approaches in other regions of the human brain. We share interactive data resources for the scientific community to further interrogate molecular mechanisms associated with complex brain disorders.