Learning how amino acid sequences define protein structure has been the major challenge for molecular biology since the first protein structures were determined in the 1960s. While staggering progress has been made with soluble proteins, investigations of membrane protein folding have long been hampered by the technical challenges owed to the properties of membrane environment that poses enormous complexity matched only by equally complex properties of the proteins that reside there. Consequently, mechanisms of maintaining stability in membrane proteins are largely unknown. For instance, the major contributor for soluble proteins is the hydrophobic effect associated with desolvation during side-chain burial, but how membrane proteins cope with the loss of desolvation expected in the bilayer has not been experimentally characterized. Here, to develop better understanding of membrane protein folding, I have quantitatively determined for the first time the molecular interplay of hydrogen bonding, van der Waals packing and desolvation contributions in the context of side-chain interactions in a large natural membrane protein.

Contrary to the widely accepted notion that hydrogen bonds in membrane proteins are extremely strong, my experimental evaluation of eight hydrogen-bonded interactions in bacteriorhodopsin using double mutant cycles indicates a surprisingly small contribution of only 0.6 kcal mol⁻¹ on average, which is very similar to the contributions in soluble proteins. Thus, hydrogen bonding does not appear to be adequate to compensate for the loss of desolvation. Alternately, denser packing in membrane proteins has been proposed to provide added stability, but this remains controversial. I set out to experimentally measure the packing density, as well as desolvation, by observing the energetic and structural response to cavity creating mutations made in bacteriorhodpsin. I find that the desolvation cost due to removing three methyl groups from the bacteriorhodopsin core was minimal as expected (0.3 ± 0.4 kcal mol⁻¹ to 0.6 ± 0.9 kcal mol⁻¹), while van der Waals packing contributes significantly by 30 ± 6 cal mol⁻¹ Å⁻³, or 18 ± 10 cal mol⁻¹ Å⁻², which is also similar to the strength measured in a soluble protein, suggesting similar packing density in membrane and soluble proteins. Taken together with our recent finding that shows a larger fraction of residues is buried in membrane proteins compared to soluble proteins, I propose that packing is indeed dominant in membranes and that membrane proteins further optimize packing by burying the residues more extensively.