Science 340, 978–981 (2013)

Protein folding in the endoplasmic reticulum can be assisted by chaperones, but this process is energetically costly. Xu et al suspected that there should be a mechanism by which protein folding is terminated to avoid wasting energy in futile refolding cycles. To investigate this possibility, the authors used several variants of GFP to explore the impact of folding rate on folding success. The authors observed that GFP localized in the endoplasmic reticulum of wild-type cells folds very poorly. However, a fast-folding GFP was able to fold successfully. The authors speculated that protein mannosylation might serve as a covalent modification marking proteins for folding termination. Indeed, dol-P-Man:protein O-mannosyltransferases (Pmts) are known to mannosylate folding-defective proteins in the endoplasmic reticulum, but the purpose of this modification was unclear. Purification of the two GFP constructs using a concanavalin A (ConA) affinity column to separate glycosylated and nonglycosylated proteins revealed that the fraction of slow-folding GFP bound to the column was larger than the fraction of fast-folding GFP. However, deletion of the yeast mannosyltransferases Pmt1 and Pmt2 reduced or abolished binding of the slow-folding GFP to ConA in single and double mutants, respectively. The double mutant also showed increased binding of the chaperone Kar2 to another folding-defective GFP construct. These data support a model where Pmt-mediated protein mannosylation provides a new mechanism of quality control by directly terminating unproductive folding.