This work reports that correlations between mRNA and protein abundance of the same gene across thousands of single cells are often quite high, and that single cells that are outliers of such linear correlations are, moreover, the result of a gene-specific adaptation of mRNA-to-protein ratios to properties of the phenotypic state and microenvironment of individual cells. Accounting for this allows an accurate prediction of protein abundance in single cells. Using gene induction of JUN illustrates that these properties of single cells can influence mRNA-protein relationships at multiple levels in sometimes non-intuitive ways, involving adaptation of transcription, nuclear export, protein translation, and mRNA decay. Finally, the work shows that also subcellular patterns of JUN transcripts vary non-randomly between single cells and are adapted to the extent of local cell crowding, which may contribute to determining cell-to-cell variability in mRNA-to-protein ratios.