The accurate prediction of protein structures with artificial intelligence has been a spectacular success. Yet, how proteins fold into their native structures inside the cell remains incompletely understood. Of particular interest is to rationalize how proteins interact with the protein homeostasis network, an organism specific set of protein folding and quality control enzymes. Failure of protein homeostasis leads to widespread misfolding and aggregation, and thus neurodegeneration. Here, I present a comparative analysis of the folding across a protein family from a single organism, the Saccharomyces cerevisiae small GTPases. Using computational modelling to directly probe folding dynamics, this work shows how near identical structures from the same folding environment exhibit heterogeneous folding landscapes. Remarkably, yeast small GTPases are found to unfold along different pathways either via the N- or C-terminus initiated by structure-encoded predetermined breaking points. Moreover, degrons as recognition signals for ubiquitin-dependent degradation were systematically depleted from the initial unfolding sites, as if to protect from too rapid degradation upon spontaneous unfolding. In turn, Hsp70 Ssb chaperone binding correlated only with N-terminal hydrophobicity. The presented results highlight a direct coordination of folding pathway and protein homeostasis interaction signals across a protein family. A deeper understanding of the interdependence of proteins with their folding environment will help to rationalize and combat disease linked to protein misfolding and dysregulation. More generally, this work underlines the importance of understanding protein folding in the cellular context, and highlights valuable constraints towards a systems-level understanding of protein homeostasis.
