Files
Abstract
The CaaX pathway is a set of three post-translational modifications that has long been associated with increasing hydrophobicity and membrane association of proteins. These modifications occur to proteins containing a CaaX motif at the COOH terminus, where “C” is a Cysteine, “a” is an aliphatic amino acid and “X” is one of several residues. The 3-step pathway begins with isoprenylation of the Cysteine, followed by proteolysis of the -aaX tripeptide, and concludes with carboxyl methylation of the lipidated Cysteine. However, in recent years, a new branch of CaaX pathway termed the “shunt” pathway has been characterized. In this alternative outcome, proteins undergo prenylation and then are shunted out of the pathway before undergoing sequential proteolysis and methylation steps. Notably, these shunted sequences often lack the aliphatic amino acids associated with traditional CaaX motifs, which suggests that a broader range of sequences may be able to undergo prenylation than previously expected. In this study, we used a genetic screen utilizing Ydj1, a known shunt protein in yeast, to identify nearly 140 shunted sequences. These newly identified sequences, together with previously published data, were then used to train the machine learning algorithm, Support Vector Machine (SVM). Using SVM, we were able to predict prenylation for all 8000 possible Cxxx sequences. We then selected a subset of these predicted sequences to test experimentally, where we observed that the machine learning-based prediction method outperformed previously published methods. Lastly, we investigated the histone chaperone, Nap1, as a predicted shunt protein, and not only show that it is indeed prenylated, but also demonstrate that the prenylation status appears to impact Nap1 nuclear levels.