Previous crystal structures of membrane-spanning enzymes called ATPases have revealed that the enzymes undergo complex movements. The movements, it now emerges, involve rocking in place in the membrane. See Article p.193
Enzymes called transport ATPases use the energy stored in ATP molecules to pump ions across cell membranes against steep concentration gradients. The enzymes do this in a slow, milliseconds-long dance of sequential conformation changes that cause membrane-spanning parts of the protein to slide up or down, changing their positions relative to one other. An enduring mystery has been how protein segments buried in the membrane's phospholipid bilayer can move without thrusting their hydrophobic portions into an aqueous environment that is unfavourable for them. On page 193, Norimatsu et al.1 pin down the interactions between a calcium-ion ATPase and the phospholipids that surround it in a series of crystal structures. In doing so, they demonstrate that the enzyme keeps its hydrophobic domain buried by rocking back and forth in the membrane.
Calcium-ion (Ca2+) ATPases pump Ca2+ from the cell's cytosolic fluid, either out of the cell or into organelles. The enzymes contain three globular domains (A, N and P) that project into the cytosol, and a hydrophobic domain formed by ten transmembrane α-helices (M1–M10). The Ca2+-binding pocket is located in the hydrophobic domain. Since 2000, many of the conformational changes that Ca2+-ATPases undergo during the pump cycle have been captured in crystal structures2,3,4,5,6; together, these reveal how the pump works.
Large-scale movements occur around the enzyme's ATP-binding active site, which is tucked into the centre of the three globular domains. First, two Ca2+ ions from the cytosol become bound in the hydrophobic pocket. ATP is then bound to the N domain, which rotates towards the P domain while the A domain rotates away. Next, the third phosphate group of the ATP is transferred to an aspartate amino-acid residue in the P domain, and a conformational change occurs using the energy of the aspartate-bound phosphate, allowing Ca2+ ions to leave the enzyme from the other side. The aspartyl phosphate is then hydrolysed by rotation of the A domain back into its starting position.
These large rotations and twists of the globular domains drive movement of the transmembrane spans relative to one other. M1 and M2 move in a piston-like manner when ATP binding occurs, and M4 moves by the equivalent of one helix turn when ADP (formed when the third phosphate is cleaved from ATP) leaves the enzyme and the bound Ca2+ is released. Other transmembrane helices are extended, shortened or kinked, and some pairs of helices change their elevation and separation. These movements guarantee that Ca2+ moves in only one direction.
However, this model does not explain how the hydrophobic transmembrane segments stay buried in lipid7. If the ATPase were stationary in the membrane, the vertical movements of the spans would either expose hydrophobic side chains to an aqueous environment, or distort the lipid bilayer by pulling phospholipids out of the membrane plane. Investigation of this apparent paradox has been difficult, because the lipid bilayer cannot normally be visualized in crystal structures.
Norimatsu et al. tackled this problem using a technique called solvent contrast modulation, in which the crystallized protein–lipid complex is placed in iodine-containing contrast media of various concentrations, each of which diffracts X-rays differently. A bilayer of lipid and detergent fills the hydrophobic spaces between proteins in the crystal lattice. Analysis of X-ray diffraction intensities reveals variations in membrane thickness and identifies phospholipid headgroups, which face outward on either side of the membrane.
The authors analysed crystals of the Ca2+-ATPase SERCA1a at four stages of the pump cycle, mapping the positions of phospholipid headgroups to reveal specific protein–lipid interactions. They found that residues of arginine and lysine (basic amino acids, which have positive charge under physiological conditions) were anchored to negatively charged phospholipid headgroups on each side of the protein. The researchers' observations confirmed that, if the ATPase were stationary, some lipids would move away from the membrane with M1 and M3, and, at the other end of the protein, be forced into the membrane with M10 — an unlikely scenario.
Norimatsu et al. angled each structure such that the phospholipid headgroups and basic residues remained in two parallel planes throughout the pump cycle. In this new conceptual framework, the ATPase is not stationary, but tilts back and forth (Fig. 1). A belt of tryptophan amino-acid residues also maintain parallel positions in the membrane, acting as a float to sense the water–lipid boundary8. At the M10 side of the protein, two tryptophans and a lysine–phospholipid interaction might serve as a pivot point from which the protein can tilt.
Norimatsu et al. include videos (see Supplementary Videos S1 and S3) that model the way consecutive conformations cause the protein to rock, surrounded by stable belts of amino-acid residues. The tilting allows large perpendicular movements of transmembrane helices relative to one another without forcing hydrophobic residues out of the bilayer.
The work also revealed that specific basic residues perform one of two tasks. Some 'snorkel' up from within the bilayer to form salt bridges with fixed phospholipid headgroups, orienting the protein. Others extend down from outside the membrane to interact with different headgroups at different stages of the pump cycle, possibly functioning as 'catches' to stabilize transient conformational states. Such interactions could affect the kinetics of conformation changes.
The discovery of these two roles for basic residues adds to our understanding of the interactions between membrane proteins and surrounding lipids. A few lipids can be resolved in conventional crystallography, and four have been documented9 in functionally crucial regions of SERCA1a. A different experimental approach in a related enzyme, the sodium–potassium ATPase, identified two interactions between lysine residues and tightly associated membrane lipids10. Mutation of these lysines revealed one stabilizing and one activity-related function of lysine–lipid interactions. Lipid–protein interactions are therefore emerging as crucial to protein function.
It remains to be seen whether mutation of phospholipid-binding amino-acid residues can cause disease by impairing protein function. Norimatsu et al. cite an example11 in which mutation of one basic residue in a Ca2+-ATPase halted a major conformational transition. Such mutations might not be predicted to be damaging in the computer algorithms that are currently used to interpret the effects of human genetic variants, and could fall through the cracks in patient diagnosis12.
The elegance of SERCA1a's rocking motion might indicate that other transport ATPases behave in a similar way, but this hypothesis awaits further research. Other examples would allow more lipid–protein interactions to be discerned, providing a better picture of how transport occurs. Norimatsu and colleagues' contrast-modulation method could be extended to many crystallized membrane proteins, facilitating this process.Footnote 1
Norimatsu, Y., Hasegawa, K., Shimizu, N. & Toyoshima, C. Nature 545, 193–198 (2017).
Toyoshima, C., Nakasako, M., Nomura, H. & Ogawa, H. Nature 405, 647–655 (2000).
Sørensen, T. L.-M., Møller, J. V. & Nissen, P. Science 304, 1672–1675 (2004).
Olesen, C. et al. Nature 450, 1036–1042 (2007).
Møller, J. V., Olesen, C., Winther, A.-M. L. & Nissen, P. Q. Rev. Biophys. 43, 501–566 (2010).
Toyoshima, C. & Cornelius, F. Curr. Opin. Struct. Biol. 23, 507–514 (2013).
Sonntag, Y. et al. Nature Commun. 2, 304 (2011).
Killian, J. A. & von Heijne, G. Trends Biochem. Sci. 25, 429–434 (2000).
Drachmann, N. D. et al. FEBS J. 281, 4249–4262 (2014).
Habeck, M., Kapri-Pardes, E., Sharon, M. & Karlish, S. J. D. Proc. Natl Acad. Sci. USA 114, 2904–2909 (2017).
Yamasaki, K., Daiho, T., Danko, S. & Suzuki, H. J. Biol. Chem. 279, 2202–2210 (2004).
Dong, C. et al. Hum. Mol. Genet. 24, 2125–2137 (2015).