The membrane insertion of eukaryotic proteins generally requires the assistance of other membrane proteins. But on page 767, Brambillasca et al. show that surprisingly long protein stretches get across membranes unassisted, as long as their transmembrane domains (TMD) are not too hydrophobic. Ancestral membrane proteins may have similarly self-inserted.

The group started with a variant of cytochrome b(5), whose 28-residue luminal domain was known to push its way through protein-free liposomes. They now find that b5 can push luminal domains of up to 85 residues through liposomes and into the ER of yeast cells lacking the normal translocation machinery. Unassisted translocation might also pick up the slack in normal eukaryotic cells when ER translocons are congested.

A membrane protein of similar topology, called Syb2, had no such power. The authors traced the disparity to the TMD: the lower hydrophobicity of b5's TMD was unexpectedly advantageous for insertion into liposomes. The authors imagine that more hydrophobic TMDs might aggregate in the cytoplasm as they are released from ribosomes, creating an unsuitable conformation for ER insertion. In vivo, however, chaperones might prevent this aggregation.

Based on its low TMD hydrophobicity, the authors identified the PTP1B phosphatase as another self-inserting protein. Unassisted insertion only works in low-cholesterol membranes, such as the ER and mitochondria, whose lipids are free to move around while proteins push through. The Golgi and plasma membrane, by contrast, are probably too rigid.