Mapping the Microenvironment: A Single-Cell Spatial Atlas of Diabetic Nephropathy

Background: The kidney is an architectural masterpiece, comprising nearly 100 distinct cell types. Despite this complexity, the cellular architectural principles of the kidney in both health and disease remain poorly understood. Our study aims to elucidate these principles by profiling healthy and diabetic kidney samples at true cellular resolution. We seek to uncover the molecular and spatial signatures that characterize the disease condition. Methods: We utilize the spatial transcriptomics platform CosMx, which enables the spatial profiling of 1,000 transcripts at true single-cell resolution. We applied deep learning-based algorithms to integrate our spatial transcriptomics dataset with human kidney single-nuclear gene expression information. This integration allows us to map annotations and cell states onto our spatial dataset, providing a comprehensive view of the kidney's cellular landscape in both health and disease. Results: We profiled gene expression and spatial information for 1,7 million cells across healthy, diabetic, and diabetic kidney disease samples. We identified 19 distinct major cell populations within these samples. Leveraging the spatial information of these cell populations, we identified 13 distinct kidney niches. In diabetic kidney disease, we found an enrichment of injured tubular epithelial niches, as well as fibroblast and immune niches. Further subclustering of these injured epithelial populations based on their spatial location revealed microenvironments almost exclusively associated with disease. Within these microenvironments, injured epithelial cells exhibited a distinct transcriptional signature, and colocalized with immune cells. To further analyze these immune cells, we integrated seven datasets comprising approximately 150 patients and 165,000 immune cells into a single kidney immune cell atlas. Using the cell states inferred from our integrations, we mapped cell-cell interactions within these disease-specific microenvironments. Conclusion: This study underscores the potential of spatial transcriptomics in dissecting the heterogeneity of kidney disease. By providing a detailed map of cellular and molecular landscapes, our findings pave the way for a better understanding of disease mechanisms and potential therapeutic targets. Funding: NIDDK Support, Government Support – Non-U.S.