Structure and gating mechanism of theacetylcholine receptor poreAtsuo Miyazawa*, Yoshinori Fujiyoshi† & Nigel Unwin‡* RIKEN Harima Institute, 1-1-1 Kouto, Mikazuki-cho, Sayo, Hyogo 679-5148, Japan† Department of Biophysics, Faculty of Science, Kyoto University, Oiwake, Kitashirakawa, Sakyo-ku, Kyoto 606-8502, Japan‡ MRC Laboratory of Molecular Biology, Hills Road, Cambridge CB2 2QH, UK...........................................................................................................................................................................................................................The nicotinic acetylcholine receptor controls electrical signalling between nerve and muscle cells by opening and closing a gated,membrane-spanning pore. Here we present an atomic model of the closed pore, obtained by electron microscopy of crystallinepostsynaptic membranes. The pore is shaped by an inner ring of 5a-helices, which curve radially to create a tapering path for theions, and an outer ring of 15a-helices, which coil around each other and shield the inner ring from the lipids. The gate is aconstricting hydrophobic girdle at the middle of the lipid bilayer, formed by weak interactions between neighbouring inner helices.When acetylcholine enters the ligand-binding domain, it triggers rotations of the protein chains on opposite sides of the entrance tothe pore. These rotations are communicated through the inner helices, and open the pore by breaking the girdle apart.The propagation of electrical signals between nerve cells and theirtargets takes place at the chemical synapse through the action oftransmitter-gated ion channels. These fast-acting molecularswitches are oligomeric proteins composed of two main functionalparts: an extracellular, ligand-binding domain, and a gated, mem-brane-spanning pore. Neurotransmitter released from the nerveterminal enters the ligand-binding domain on the surface of thetarget cell, and triggers a transient conformational change thatopens the gate in the membrane-spanning pore. Ions then flowselectively through the pore down their electrochemical gradients,giving rise to a change in membrane potential.The acetylcholine (ACh) receptor, at the nerve–muscle synapse, isa member of a superfamily of transmitter-gated ion channels, whichincludes the serotonin 5-HT3,g-aminobutyric-acid (GABAAandGABAC) and glycine receptors1. It has a cation-selective pore,delineated by a ring of five subunits (a,a,b,gor1,d), thatopens upon binding of ACh to distant sites in the twoa-subunitsat or near the subunit interfaces2–4.Therearefourpredictedmembrane-spanning segments, M1–M4, in each subunit. Thesecond membrane-spanning segment, M2, shapes the lumen ofthe pore, and forms the gate of the closed channel. Although muchinformation has been obtained about the roles of individual aminoacids in affecting ion transport, and about their relative positions onthe membrane-spanning segments, their detailed three-dimen-sional arrangement has not yet been visualized in this receptor, orin any other transmitter-gated ion channel. Nor is it known how theACh-triggered conformational change is communicated throughthe membrane to open the pore.The (muscle-derived) electric organ of the Torpedo electric ray ishighly enriched in ACh-receptor-containing membranes, and hasbeen a valuable source of tissue for physiological and biochemicalstudies of neurotransmission for more than 50 years. The isolatedpostsynaptic membranes are also amenable to structure analysis byelectron microscopy. They convert readily into tubular crystals,having receptors and intervening lipid molecules organized like theyare in vivo5,6(Fig. 1), and enable different functional conformationsto be investigated under near-physiological ionic conditions. Adescription of the amino-terminal ligand-binding domain of thereceptor has been obtained by fitting theb-sheet core structure froma homologous pentameric ACh-binding protein, AChBP7, to thethree-dimensional densities determined from electron images8.However, the quality of these images and distortions of the crystallattice limited previous descriptions of the pore, revealing somea-helical folding, but only in the pore-lining segments9. By record-ing exceptional images at liquid-helium temperatures10and apply-ing a computational method to correct for the distortions11,wehavenow extended the resolution to 4 A˚. We report here an atomic modelof the pore domain based on the 4-A˚density map, and propose anexplanation of how the pore opens when ACh binds to the receptor.Structure determinationTubular crystals, having helical symmetry, were grown from Torpedomarmorata membranes, and imaged in thin films of amorphous ice(see Methods). Altogether 359 images (, 106receptors), involvingfour helical families, were analysed.The tubes were too small to yield accurate amplitudes by electrondiffraction, so the amplitude, as well as the phase terms, had to bemeasured from Fourier transforms of the images, and then addedvectorially to enhance the signal-to-noise ratio. The amplitudesshowed resolution-dependent fading, which was compensated byscaling the measured values against values calculated from a modelFigure 1 Cross-section of a tubular crystal, at low resolution. The receptor proteinprojects from either side of the membrane, visible as two concentric rings of density, 30 A˚apart. A single receptor, cut centrally, is shown at the top. The membrane-spanning poreand the N-terminal ligand-binding domain, shaping a large central vestibule, are outlinedby red and green rectangles, respectively. The surfaces encompassing the hydrophobiccore of the membrane are assumed to lie along the centres of the rings of density48.articlesNATURE | VOL 423 |26 JUNE 2003 | www.nature.com/nature 949© 2003 Nature PublishingGroupbased on the structure of AChBP (see Methods). The phases weresignificant to 4-A˚resolution after the terms from many images ofthe same helical family had been added (Table 1). Structures weresynthesized from the amplitude and phase terms in each family andcombined, by averaging in real space, to obtain the final three-dimensional density map.The real-space averaging brought about a substantial improve-ment in signal/noise ratio (Fourier shell correlation coefficient of0.5 (refs 12, 13) at 4.0 A˚; Supplementary Fig. 1), and enabled thepolypeptide chains to be fitted to the densities (see Methods) afterassigning the four helical segments as in ref. 8, with
View Full Document
Unlocking...