Introduction: Pleiotropy (i.e. when one gene influences multiple phenotypes) is central to major problems in biology. It contributes to difficulty predicting the fitness effect of mutations and also can lead to development of drugs with undesired side effects. Despite the importance of pleiotropy, there are fundamental disagreements about its prevalence, mechanism, and definition. One particularly intractable issue when studying pleiotropy has been the definition of independent phenotypes, especially in high-dimensional studies in which many related and potentially correlated phenotypes are surveyed. To achieve a fuller understanding of pleiotropy’s extent and molecular underpinnings, we develop a novel approach to describe the network of correlated phenotypes that contribute to pleiotropy in yeast single-cell morphology.
Methods: We used high-throughput microscopy and image analysis to quantify hundreds of single-cell morphological features in 374 lines from a cross between two genetically diverged yeast strains. We identified 12 pleiotropic loci (QTL) that each contribute to variation in 2 or more traits. But does each QTL influence multiple unrelated traits or does some fraction of pleiotropy result because genetic effects cascade through a network of phenotypes that exert influence on one another? To determine which phenotypes are correlated in the absence of genetic effects, we leveraged hundreds of single cell measurements from within each of the 374 clonal yeast populations. We used advanced statistical methods that tease apart phenotypic correlations present within strains from those that arise between strains.
Results: Within clonal yeast populations, over 80% of phenotype pairs are more correlated than after random permutation, suggesting that a connected network of phenotypes underlies cell morphology. Correlations increase between genetically distinct yeast strains for ~35% of phenotype pairs, suggesting that genetic effects contribute additional pleiotropy beyond what would occur due to the underlying structure of the phenotype-phenotype network. As for QTL effects, pleiotropy is achieved through indirect effects cascading through the phenotype-phenotype network roughly 2/3 of the time. A QTL on chromosome 13 is the exception. It affects the most phenotypes in this study (43); experiments confirm these effects are due to a single gene, CYK2. Our analysis suggests that CYK2 influences 22 phenotypes directly and 21 indirectly through the phenotype-phenotype network. In sum, our novel approach can enhance knowledge of (1) phenotype-phenotype networks, and (2) the extent and nature of pleiotropy. Both types of knowledge further our understanding of the relationship between genotype and phenotype.