Colorectal cancers (CRCs) are composed of an amalgam of cells with distinct genotypes and phenotypes. Here we reveal a previously unappreciated heterogeneity in the biosynthetic capacities of CRC cells. We discover that the majority of ribosomal DNA transcription and protein synthesis in CRCs occur in a limited subset of tumor cells that localize in defined niches. The rest of the tumor cells undergo an irreversible loss of their biosynthetic capacities as a consequence of differentiation. Cancer cells within the biosynthetic domains are characterized by elevated levels of the RNA Polymerase 1 subunit A - POLR1A. Genetic ablation of POLR1A-high cell population imposes an irreversible growth arrest to CRCs. We show that elevated biosynthesis defines stemness in both LGR5+ and LGR5- tumor cells. Therefore, a common architecture in CRC is a simple cell hierarchy based on the differential capacity to transcribe ribosomal DNA and synthesize proteins.

Single-cell RNA sequencing (scRNA-seq) is the leading technique for charting the molecular properties of individual cells. The latest methods are scalable to thousands of cells, enabling in-depth characterization of sample composition without prior knowledge. However, there are important differences between scRNA-seq techniques, and it remains unclear which are the most suitable protocols for drawing cell atlases of tissues, organs and organisms. We have generated benchmark datasets to systematically evaluate techniques in terms of their power to comprehensively describe cell types and states. We performed a multi-center study comparing 13 commonly used single-cell and single-nucleus RNA-seq protocols using a highly heterogeneous reference sample resource. Comparative and integrative analysis at cell type and state level revealed marked differences in protocol performance, highlighting a series of key features for cell atlas projects. These should be considered when defining guidelines and standards for international consortia, such as the Human Cell Atlas project.