Cell compartment is a predictor of protein rate of evolution, but not in the manner expected: evidence against the extended complexity hypothesis

What accounts for the variation between proteins in their rate of evolution per synonymous substitution (i.e. dN/dS, alias ω)? Previous analyses suggested that cell location is predictive, with intracellular proteins evolving slower than membrane proteins, a result considered supportive of the extended complexity hypothesis. However, as they occur in 3D space, cytoplasmic proteins are expected to be more abundant. As the level of gene expression is the strongest predictor of ω, and many predictors of protein rate variation are explained by covariance with it, here we ask whether the cell compartment effect is explained by covariates. We employ two single-celled species for which there exist exceptional data, the bacterium (Escherichia coli) and the eukaryote (Saccharomyces cerevisiae). For both, we establish informative species trios to determine branch-specific ω values. In both species, in the absence of covariate control, cytoplasmic proteins evolve relatively slowly, while membrane proteins evolve fast, as originally claimed. After controlling for protein abundance, however, membrane proteins have the lowest rates, this inversion being resilient to multiple alternative abundance control methods. The effect size of the cell compartment as a predictor is of a comparable magnitude to the essentiality effect and remains when allowing for essentiality. We conclude that the effects of the cell compartments are real, but their direction is dependent on the presence or absence of abundance control. These results question any model, such as the extended complexity hypothesis, that claims support from a slower evolution of cytoplasmic proteins and underscore the importance of covariate control.