Cellular identity in complex multicellular organisms is strictly maintained over the course of life. This control is achieved in part by the organ structure itself, such that neighboring cells influence each other's identity. However, large-scale investigation of the cellular interactome has been technically challenging. Here, we develop CIM-seq, an unsupervised and highthroughput method to analyze direct physical cell-cell interactions between every cell type in a given tissue. CIM-seq is based on RNA sequencing of incompletely dissociated cells, followed by computational deconvolution of these into their constituent cell types using machine learning. We use CIM-seq to define the cell interaction landscape of the mouse small intestinal and colonic epithelium, uncovering both known and novel interactions. Specifically, we find that the general architecture of the stem cell niche is radically different between the two tissues. In small intestine, the stem-Paneth cell interaction forms an exceptionally strong and exclusive niche, in which Paneth cells provide Wnt ligands 1 . In colonic epithelium, no similar compartment exists to support stem cells, and Wnt signaling is provided by a mesenchymal cell layer 2,3 . However, colonic stem cells are supported by an adjacent, previously unrecognized goblet cell subtype expressing the wound-healing marker Plet1, which is also highly upregulated during regeneration of colon epithelium. These results identify novel cellular interactions specific for the colonic stem cell niche and suggest an additional level of structural control in the colon. CIM-seq is broadly applicable to studies that aim to simultaneously investigate the constituent cell types and the global interaction profile in a specific tissue.
MainCells in higher order multicellular organisms are diverse and highly specialized. Such a level of specialization requires strict control, which is in part encoded in their spatial organization 1,4 . Single-cell RNA-seq (scRNA-seq) methods have enabled rapid advances towards determining the full complement of cellular diversity in human and mouse tissue 5 . However, no similar high throughput and unbiased method exists to chart the fine-grained structural diversity of how these cells interact, or to profile interaction-dependent changes in gene expression. To this end, we developed Cell Interaction by Multiplet sequencing (CIM-seq), a method that allows largescale interaction profiling within a well-established scRNA-seq framework.