Extent and context dependence of pleiotropy revealed by high-throughput single-cell phenotyping

Pleiotropy - when a single mutation affects multiple traits - is a controversial topic with far-reaching implications. Pleiotropy plays a central role in ongoing debates about how complex traits evolve and whether biological systems tend to be modular or organized such that every gene has the potential to affect many traits. Pleiotropy is also critical to initiatives in evolutionary medicine that seek to trap infectious microbes or tumors by selecting for mutations that encourage growth in some conditions at the expense of others. Research in these fields, and others, would benefit from understanding the extent to which pleiotropy reflects inherent relationships among phenotypes that correlate no matter the perturbation (vertical pleiotropy), versus the action of genetic changes that impose correlations between otherwise independent traits (horizontal pleiotropy). We tackle this question by using high-throughput single-cell phenotyping to measure thousands of pairwise trait correlations across hundreds of thousands of cells representing hundreds of genotypes of the budding yeast, Saccharomyces cerevisiae. We map pleiotropic quantitative trait loci using genotypes derived from a cross between natural strains, and we separate vertical and horizontal pleiotropy by partitioning trait correlations into within- and between-genotype correlations. We investigate how pleiotropy can change by using genotypes from mutation-accumulation lines that experienced minimal selection, and by tracking trait correlations through the cell-division cycle. We find ample evidence of both vertical and horizontal pleiotropy, and observe that trait correlations depend on both genetic background and cell-cycle position. Our results suggest a nuanced view of pleiotropy in which trait correlations are highly context dependent and biological systems occupy a middle ground between modularity and interconnectedness. These results also suggest an approach to select pairs of traits that are more likely to remain correlated across contexts for applications in evolutionary medicine.