Neuron, Vol. 3, 665-676, December, 1989, Copyright 0 1989 by Cell Press The Structure of Ion Channels Review in Membranes of Excitable Cells Nigel Unwin The purpose of this review is to discuss ilnplications MRC Laboratory of Molecular Biology of some of the recent findings in relation to three- Hills Road dimensional structure, drawing on information derived Cambridge Cl32 2QH from studies of channels, other bilayer-spanning pro- England teins, and soluble proteins by electron microscopy and X-ray diffraction. It is likely that channels have parallels with soluble allosteric enzymes. The fact that both types Ion channels play a central role in the transmission of of molecules switch states reversibly in an all-or-none electrical signals along the membranes of neurons and fashion, the cooperative behavior of many channels, other excitable cells; they also mediate communication and the conformational changes detected in channel between cells at synapses and participate in a multitude proteins at low resolution all suggest the existence of of regulatory processes involving signal transduction at close structural analogies. Obviously, the lipid bilayer is the membrane surface. For a single molecule these rep- also important, and must restrict the conformational resent tasks that are both diverse and complex. Yet, on changes that would be energetically acceptable, since it the face of it, channels are remarkably uncomplicated. is a separate phase that the protein surfaces are con- First, they appear to share the same, simple architectural strained to match. plan. The building blocks are identical or homologous Ion channels, like soluble proteins, often take part in polypeptide units, which are assembled symmetrically a range of actions. Not only do they detect and transmit within the membrane (at least where structural evidence the effect of ligand binding or shifts in membrane poten- is available), delineating a pore down their center. Sec- tial, but they interact with various cytoplasrnic and ex- ond, channels have just two essential states-open or tracellular molecules as well. These extra interactions closed. Either the pore is sealed, making a permeability may involve intimate associations with other proteins or barrier across the membrane, or it forms a continuous, necessitate separate structural specializations of the water-filled pathway, exposing the surfaces needed to channel protein itself. Such individual properties and discriminate between the ions passing through. the classification of channels according to their gating This is the elementary picture. At a molecular level, characteristics, or to the superfamilies in which they be- the detailed structure of channels has not been resolved, long, have been discussed in several recent reviews (e.g., and so fundamental questions remain unanswered. What Hille, 1984; Catterall, 1988; Miller, 1989). The emphasis is the design that makes channels so efficient, facilitating here is on features that channels may have in common. selective transport across the membrane typically of Since these molecules have all evolved with the same thousands of ions in a millisecond? What are the chemi- task of mediating passive transport across the mem- cal and physical origins of ion transport specificity that brane, it is possible that they are constructed and work allow channels to hinder or prevent diffusion of some by a similar set of underlying rules, whatever the differ- ions but not of others across the membrane? What con- ences in subunit composition, in the fine detail, or in the formational changes are involved in gating-the rapid specializations they incorporate. switching between open and closed states? Do channels In keeping with a unifying theme, I shall show that make use of the same general rules that apply to soluble putative pore-lining a-helices of channels in different su- enzymes or multisubunit complexes, or does the pres- perfamilies share identical alignments of small polar and ence of the lipid bilayer enforce something basically large hydrophobic residues. This implies that there is a different? common packing principle involved in buildung the wall Of course, a really sound understanding will have to around a pore and suggests a possible general way by await the three-dimensional framework furnished by which channels may open and close. structures solved crystallographically to atomic resolu- tion. In the meantime, however, channels are becoming Channels as Oligomeric Proteins much better understood physiologically and biochemi- cally as a result of powerful new techniques such as Channels are essentially oligomeric membrane proteins patch-clamping, molecular cloning, and site-directed possessing cyclic or pseudo-cyclic symmetry (Figure l), mutagenesis, which enable single-channel behavior to and one way to classify them is in terms of the number be recorded, amino acid sequences to be derived, and of subunits, or structural units, around the ring. Al- specific parts of the sequence to be related to precise though in some cases this number has been established physiological functions. New channel proteins are being directly by experiment, in others it has been deduced in- discovered and characterized, Superfamilies of voltage- directly from the amino acid sequence. Either there are and ligand-gated channels are being identified and ex- homologous bilayer-spanning domains, representing the tended. And experiments are now even pin-pointing structural units, within the sequence of a single polypep- certain amino acid residues as lining the walls of pores, tide (in which case it is not strictly an oligomer), or the as affecting rate of ion transport, and as involved in sens- protein shows homologies with a channel protein in ing voltage shifts across the membrane. which the number of structural units has already been junction channel and synaptophysin, have been deter- mined by structural methods (Makowski et al., 1977; Un- win and Zampighi, 1980) or cross-linking studies (Thomas et al., 1988) to be hexamers of identical subunits. Quaternary Structures None of the above channels has yet been analyzed at 4 5 6 anywhere near atomic resolution, and some we know Figure 1. ion Channels Are Bulk from Four, Five, or Six Structural about only through single-channel conductance mea- Units surements and deduced amino acid sequences. How- The channels discussed in this review are all thought to be integral ever, it has been possible to discover details about two membrane proteins possessing cyclic or pseudo-cyclic symmetry, in of these channels at low resolution by three-dimensional which the central symmetry axis delineates a gated, water-filled electron image analysis of molecules crystallized in pathway for the ions. The structural units are identical or homolo- membranes as planar or he!ical arrays. The best images gous protein subunits, or homologous domains comprlslng a single polypeptidechain. The most selective channels, with the narrowest have been obtained by rapidly freezing the specimens pores (e.g., the voltage-gated sodium and calcium channels), have in a thin film of solution and observing the specimens four structural units; the least selective channels, with the widest at a temperature low enough (<-130°C) for the ice to pores (e.g., the gap junction channel), have six units; and the chan- remain stable. A defined lipid and ionic environment is nels with intermediate propertIes (e.g., the nicotinic acetylcholine captured by the freezing, providing closer ties with the receptor) have five units. physiology than had been feasible previously using negative stains to preserve shape and provide contrast. Furthermore, the protein is seen directly, contrasted against the lipid and water molecules, making it possible determined. The tertiary folding and overall three-di- to detect and analyze small conformations/ changes. mensional form are the same in proteins whose poly- What are the significant features in three-dimensional peptide chains are homologous over their length, and maps from the cryo-images? so inferences from homologies concerning oligomeric First, some background on the gap junction channel- state are probably correct. one of the channels studied in this way. This channel is There are now several well-studied examples of chan- built from a ring of six identical, -75 A long subunits, nels in the four, five, and six structural unit class. The called connexins. Several homologous polypeptides be- voltage-gated sodium channel and calcium channel (di- longing to the connexin family have been discovered, hydropyridine receptor) are both composed of four ho- the best characterized of which are the 32 kd and 43 kd mologous bilayer-embedded domains, making up a polypeptides from, respectively, liver and heart (Hen- large fraction of a single polypeptide chain, and so pre- derson et al., 1979; Paul, 1986; Kumar and Gilula, 1986; sumably consist of four structural units in a ring (Noda Beyer et al., 1987; Young et a!., 1987). Connexins are un- et al., 1984; Tanabe et al., 1987). The same may be true usual because of their dual role: they associate together for the voltage-gated A-type potassium channels, whose within the plasma membrane of one cell to make the polypeptides have some homologies with these do- channei, and they link to connexins in the plasma mem- mains (Tempel et al., 1987; Timpe et al., 1988). And four- brane of another cell to make a continuous com- fold symmetry of the ryanodine receptor (a calcium municating pathway between the two interiors (Loewen- channel in sarcoplasmic reticulum) has been demon- stein, 1981). The central pathway, when open, is wide strated by electron microscopy (Lai et al., 1988; Wagen- (~16 A minimum diameter; Schwarzmann et al., 1981) knecht et al., 1989). and relatively nonspecific, allowing the passage of signal A five-subunit stoichiometry of the muscle nicotinic molecules such as CAMP as well as ions, between con- acetylcholine receptor was found originally by prepara- nected cells. Caicium and hydrogen ions are ligands that tive gel electrophoresis (Lindstrom et al., 1979) and by regulate the transport, but the channel closes, rather quantitative amino-terminal analysis (Raftery et al., 1980).