High Resolution Slide-seqV2 Spatial Transcriptomics Enables Discovery of Disease-Specific Cell Neighborhoods and Pathways

Abstract

High resolution spatial transcriptomics is a transformative technology that enables mapping of RNA expression directly from intact tissue sections; however, its utility for the elucidation of disease processes and therapeutically actionable pathways remain largely unexplored. Here we applied Slide-seqV2 to mouse and human kidneys, in healthy and in distinct disease paradigms. First, we established the feasibility of Slide-seqV2 in human kidney by analyzing tissue from 9 distinct donors, which revealed a cell neighborhood centered around a population of LYVE1+ macrophages. Second, in a mouse model of diabetic kidney disease, we detected changes in the cellular organization of the spatially-restricted kidney filter and blood flow regulating apparatus. Third, in a mouse model of a toxic proteinopathy, we identified previously unknown, disease-specific cell neighborhoods centered around macrophages. In a spatially-restricted subpopulation of epithelial cells, we also found perturbations in 77 genes associated with the unfolded protein response (UPR). Our studies illustrate and experimentally validate the utility of Slide-seqV2 for the discovery of disease-specific cell neighborhoods.

Principles of Spatial Transcriptomics Analysis: A Practical Walk-Through in Kidney Tissue

“…As a result of either the NMFreg or Seurat transfer learning methodologies, not only is each bead informed by a gene expression profile and spatial location, but also its overarching cell type identity. In Marshall et al (2021) , Seurat transfer learning presented higher consistency with physiologically known structures than NMFreg for all cell types in human tissue and in collecting ducts (CD), vascular smooth muscle cells (vSMC), and distal tubules in mouse tissue. An overview of mapped cell types in mouse and human tissue is shown in Figure 1 .…”

“…Our group used Slide-seqV2 to develop a spatial transcriptomic atlas of human and mouse kidney tissue in health and disease ( Marshall et al, 2021 ). We profiled two mouse models of disease, early diabetic kidney disease (DKD, BTBR ob/ob ) and uromodulin autosomal dominant tubulointerstitial kidney disease (ADTKD, UMOD-KI).…”

“…In our work, we have implemented automated methods to identify and filter out isolated podocyte or distal tubule calls that do not occur within these denser structures. Specifically, cell type calls that have less than k nearest neighbors with the cell type identity of interest are filtered out ( Marshall et al, 2021 ).…”

“…Not only are cell type calls trusted based on the agreement between their spatial distribution and known morphology of kidney tissue, but we can also verify their identities by looking at the relative expression of known cell type marker genes. Furthermore, cell type proportions can be computed and compared between Slide-seqV2 arrays and Hybridization Chain Reaction (HCR) images ( Choi et al, 2018 ; Marshall et al, 2021 ).…”

“…That is, what cell types often neighbor a cell type of interest. In our work, we target cell type neighborhoods of Trem2 + and LYVE1 + macrophages using k nearest neighbors ( Marshall et al, 2021 ). Lastly, we can look at the morphology of cell type structures.…”