Metabolic zonation and characterization of tissue slices with spatial transcriptomics

Abstract

The exchanges of small molecules between cells and their environments are essential for the formation of functioning tissues. To study them at scale, we developed Harreman (Basque for "receive and give"), an algorithm for identifying metabolic crosstalk from spatially resolved transcriptomics of intact tissues. Unlike previous methods, which primarily focus on the secretion or reception of protein signals, Harreman reconstructs molecular metabolic crosstalk based on the co-localized expression of metabolite transporters. By utilizing a series of increasingly detailed models for testing spatial correlation, Harreman provides insight at multiple levels: a) coarse partition of the tissue into regions sharing metabolic characteristics; b) identification of metabolic exchanges within each region; and c) inference of the cell subsets involved in those exchanges. Harreman identified a sodium/calcium exchange at the tumor boundary in human lung metastases of human renal cancers, and associated it with nearby pro-inflammatory macrophages. In the mouse model of DSS-induced colitis, Harreman identified vitamin A and lysophosphatidylcholine transport at the interface of the epithelial monolayer as major signals associated with regeneration. Harreman is available at https://github.com/YosefLab/Harreman.