Nuclear gene proximity and protein interactions shape transcript covariances in mammalian single cells

Single-cell RNA sequencing studies into gene co-expression patterns could yield important new regulatory and functional insights, but have so far been limited by the confounding effects of cell differentiation and the cell cycle. We apply a tailored experimental design that eliminates these confounders, and report >80,000 intrinsically covarying gene pairs in mouse embryonic stem cells. These covariances form a network with biological properties, outlining known and novel gene interactions. We provide the first evidence that miRNAs naturally induce transcriptome-wide covariances, and compare the relative importance of nuclear organization, transcriptional and post-transcriptional regulation in defining covariances. We find that nuclear organization has the greatest impact, and that genes encoding for physically interacting proteins specifically tend to covary, suggesting importance for protein complex stoichiometry. Our results lend support to the concept of post-transcriptional ‘RNA operons’ , but we further present evidence that nuclear proximity of genes on the same or even distinct chromosomes also provides substantial functional regulation in mammalian single cells.