Escherichia coli is arguably the most thoroughly characterized organism, and yet ∼20% of its essential genes serve completely unknown functions, and many core quantitative physiological principles remain unexplained, including the famous nutrient ‘growth law’ where cell volume seems to depend exponentially on growth rate. Here we develop a platform for massive, multi-generational Optical Pooled Screening (OPS), and apply it to link image-based phenotypes to genotypes for 133,000 CRISPRi knockdowns of essential genes, tracking tens of millions of lineages and analyzing 1.6 billion cells. Our multi-dimensional dynamic phenotypes correlate exceptionally well with known gene functions, allowing us to identify many unknown roles of essential genes. Quantifying the relation between growth and cell size in turn identifies three distinct variants of the bacterial growth laws, which we explain mechanistically by discovering a new role for (p)ppGpp and SpoT as a sensor of translation elongation. Finally, we propose and systematically test an exceedingly simple, passive mechanism for the nutrient growth law, based on triggering cell division through the accumulation of a protein that, unlike ribosomes, is not controlled by (p)ppGpp. The resulting hyperbolic growth law fits the data from our three variants even better than the previously proposed exponential relationship, and with fewer parameters.
