Improving Spatial Transcriptomics with Membrane-Based Boundary Definition and Enhanced Single-Cell Resolution.

Abstract

Accurately defining cell boundaries for spatial transcriptomics is technically challenging. The current major approaches are nuclear staining or mathematical inference, which either exclude the cytoplasm or determine a hypothetical boundary. Here, a new method is introduced for defining cell boundaries: labeling cell membranes using genetically coded fluorescent proteins, which allows precise indexing of sequencing spots and transcripts within cells on sections. Use of this membrane-based method greatly increases the number of genes captured in cells compared to the number captured using nucleus-based methods; the numbers of genes are increased by 67% and 119% in mouse and axolotl livers, respectively. The obtained expression profiles are more consistent with single-cell RNA-seq data, demonstrating more rational clustering and apparent cell type-specific markers. Furthermore, improved single-cell resolution is achieved to better identify rare cell types and elaborate spatial domains in the axolotl brain and intestine. In addition to regular cells, accurate recognition of multinucleated cells and cells lacking nuclei in the mouse liver is achieved, demonstrating its ability to analyze complex tissues and organs, which is not achievable using previous methods. This study provides a powerful tool for improving spatial transcriptomics that has broad potential for its applications in the biological and medical sciences.