Are most transporters and channels beta barrels?

Given the sequence of transporters or channels of unknown secondary structure, it is usual to predict their putative transmembrane regions as alpha-helical. However, recent evidence for a facilitative glucose transporter (GLUT1) appears inconsistent with such predictions, which has led us to propose an alternative folding model for GLUTs based on the 16-stranded antiparallel beta-barrel of porins. Here we apply the same predictive algorithms we used for GLUTs to several other membrane proteins. For some of them, a high-resolution structure has been derived (beta-barrels: Rhodobacter capsulatus and Escherichia coli porins; multihelical: colicin A, bacteriorhodopsin, and reaction center L chain); we use them to test the prediction procedures. The other proteins we analyze (GLUT1, CHIP28, acetylcholine receptor alpha subunit, lac permease, Na(+)-glucose cotransporter, shaker K+ channel, sarcoplasmic reticulum Ca(2+)-ATPase) are representative of classes of similar membrane proteins. As with GLUTs, we find that the predicted transmembrane segments of these proteins are consistently shorter than expected for transmembrane spanning alpha-helices, but are of the correct length and number for the proteins to fold instead as porin-like beta-barrels.