Characterizing cellular interactions in vivo and at high resolution remains challenging. Although single-cell RNA sequencing (scRNA-seq) approaches provide high resolution, the required single-cell suspension hampers the study of cell–cell interactions. A new technology, named PIC-seq, combines cell sorting of physically interacting cells (PICs) with scRNA-seq and computational modelling to profile cellular interactions and their impact on gene expression.

PIC-seq applies a mild tissue dissociation protocol that retains some of the in situ cellular structures. Fluorescence sorting by mutually exclusive cell type markers yields a mixture of cell aggregates, representing putative PICs, and single cells of the same cell types. Massively parallel scRNA-seq generates gene expression profiles for PICs that are then computationally deconvolved, by inferring their transcriptional states and contrasting those of non-interacting single cells.

Giladi, Cohen et al. assessed their method by interrogating interactions between T cells and dendritic cells (DCs), using TCRβ and CD11c as markers, respectively. The physical interaction between T cells and DCs represents a prototypical interaction within the immune system and is essential for the distinction of self and non-self antigens. Analyses of cells and their gene expression profiles grown in monoculture, and in co-culture as single cells or cells contributing to PICs, across three time points, showed that T cell–DC interactions were required for T cell activation, proliferation and differentiation.

To determine the in vivo applicability of PIC-seq, the team focused on physical interactions between immune cells and epithelial cells in mouse neonate lungs, which were distinguished using the pan-lineage surface markers CD45 and EPCAM, respectively. Using this model of lung development, the authors were able to assess the crosstalk in epithelial–immune PICs as well as downstream gene regulatory programmes. For example, PICs were enriched for premature alveolar macrophages, monocytes and DCs in neonate lungs, whereas monocyte-derived macrophages, neutrophils and lymphocytes showed little interaction with epithelial cells. Next, the authors looked at T cell–DC interactions and corresponding gene expression profiles in mouse draining lymph nodes during homeostasis and after infection, which highlighted a subset of regulatory T cells involved in these physical interactions, and revealed unique co-stimulatory genes induced in antigen-presenting PICs.

potential for the in vivo molecular characterization of cell–cell interactions

While caveats remain — for example, transcriptional profiles of interacting cells need to be sufficiently ‘distant’ for deconvolution by the PIC-seq algorithm — this technology shows potential for the in vivo molecular characterization of cell–cell interactions.