Computational Modeling of C-Terminal Tails to Predict the Calcium-Dependent Secretion of Endoplasmic Reticulum Resident Proteins

Abstract

The lumen of the endoplasmic reticulum (ER) has resident proteins that are critical to perform the various tasks of the ER such as protein maturation and lipid metabolism. These ER resident proteins typically have a carboxy-terminal ER retention/retrieval sequence (ERS). The canonical ERS that promotes ER retrieval is Lys-Asp-Glu-Leu (KDEL) and when an ER resident protein moves from the ER to the Golgi, KDEL receptors (KDELRs) in the Golgi recognize the ERS and return the protein to the ER lumen. Depletion of ER calcium leads to the mass departure of ER resident proteins in a process termed exodosis, which is regulated by KDELRs. Here, by combining computational prediction with machine learning-based models and experimental validation, we identify carboxy tail sequences of ER resident proteins divergent from the canonical “KDEL” ERS. Using molecular modeling and simulations, we demonstrated that two representative non-canonical ERS can stably bind to the KDELR. Collectively, we developed a method to predict whether a carboxy-terminal sequence acts as a putative ERS that would undergo secretion in response to ER calcium depletion and interacts with the KDELRs. The interaction between the ERS and the KDELR extends beyond the final four carboxy terminal residues of the ERS. Identification of proteins that undergo exodosis will further our understanding of changes in ER proteostasis under physiological and pathological conditions where ER calcium is depleted.