Ion channel proteins provide a pathway for movement of ions across the hostile, low-dielectric environment of the membrane. New work on crystalline bacterial K+ channels shows that this process is optimized for the natural substrate, K+. An additional high resolution structure of the channel illustrates how the protein provides prosthetic solvation for the ion — and also a beautiful picture of how water itself solvates the ion.
A combination of crystal enzymology and high resolution structures for bacterial K+ channels, reported recently in two Nature articles1,
2 from Rod MacKinnon's laboratory at Rockefeller University, sheds new light on the energetics of potassium selectivity and permeation through these channels and also offers a taste of functional conformational changes. By using protein crystallography to observe how occupancy of the ion binding sites changes with ion concentration, Morais-Cabral et al.1 develop a clear picture of the ion permeation cycle and how its energetics are optimized for potassium. These energetics, of course, depend on the ability of the channel protein to make the K+ ions feel completely at home — the normal physiological home of ions being bulk aqueous solution. In addition, a 2.0 Å resolution structure of the bacterial KcsA channel protein obtained by Zhou et al.2 not only reveals how the channel protein keeps K+ happy, but also gives a remarkable view of how water itself hydrates a K+ ion.
Ion site occupancy from titration in the crystal In the original report of the 3.2 Å resolution KcsA structure3, the basic outlines of the ion permeation pathway were clear. The narrowest region of the pore is near the extracellular surface, where the four subunits each provide an extended structure that forms the selectivity filter (Fig. 1). Roughly in the middle of the membrane is a water-filled cavity, lined by hydrophobic amino acids, and at the bottom of the cavity is another constriction that is likely to be involved in channel gating4. Previously the ion binding positions have been inferred from difference maps of the channel protein equilibrated with different permeant ions such as K+ and Rb+(ref. 3). The new work uses both the old crystal form1,
3 and a higher resolution form2 (see below). Relative electron density at the ion binding sites is used to estimate their occupancy, which can be varied either by altering the free ion concentration around the crystal or by varying the permeant ion species. The initial conclusion is that although four ion binding sites are seen in the selectivity filter (Fig. 1), there are likely to be only two potassium ions bound in the filter at a given time.
Figure 1. A composite stereo view of the structure of the KcsA bacterial K+ channel, highlighting the ion binding sites.
The four-fold symmetric structure contains a selectivity filter (top center) supported by four pore helices (pink cylinders); the view is from within the membrane, perpendicular to the fourfold axis, with the extracellular surface at the top. The selectivity filter carbonyl oxygens are shown as spacefilling, as are the four ion binding sites, depicted here in the (2,4) configuration with K+ ions at positions 2 and 4 and water at positions 1 and 3. The grey cross section through the center of the protein shows the molecular surface smoothed with a 1.4 Å radius water molecule (calculated from PDB entry 1J95). This highlights the overall molecular profile (truncated at the sides), including the central cavity. Dotted lines indicate the probable boundaries of the lipid membrane, judged from the positions of outward facing tryptophan side chains. The K+ ion in the central cavity is surrounded by eight water molecules, evident as distinct electron density in the 2.0 Å resolution structure. Note how each ion in the structure (both in the filter and in the cavity) is surrounded by eight oxygen atoms in a configuration approximating a square antiprism.
A steady gait works best The conduction cycle of the channel thus primarily involves shuttling between two different two-ion occupancy states of the selectivity filter, one with sites 1 and 3 occupied, and the other with sites 2 and 4 occupied. The interconversion can occur in two ways: either the ions can simply move back and forth, or a new ion can enter from one side, expelling one of the resident ions to the opposite side and producing net transport. The kinetics of transport are dictated by the energetic terrain, which becomes least corrugated when the two stable intermediates — (1,3) and (2,4) — have equal energy. Any substantial imbalance makes channel throughput slower. For instance, with Rb+ as the conductive species, the two states are very unequal in energy, as judged from the strong preference seen in the Rb+-containing crystal for the (1,3) occupancy state. Just as a person with a very uneven gait is unable to walk quickly, the imbalance between the two permeation steps for Rb+ ions hinders their movement through the channel. An imbalance can also result when a mixture of two permeant ion species is present — again, the channel limps along. This example is the well-known 'mole fraction' effect, in which a mixture of two species conducts more poorly than either ion alone, and which was one of the principal pieces of evidence for multi-ion pores in potassium channels5.
The potassium ions, for which this channel is intended, exhibit no limp at all. All four ion binding sites exhibit nearly equal occupancy with potassium — which, based on the inference that there are usually two ions in the pore, means that the (1,3) and (2,4) states are approximately equally occupied. This indicates that potassium ion channels are optimally tuned for their natural substrate, accounting for their remarkable physiological performance: total transmembrane flux through a single, very narrow channel can be extremely high (>108 M-1 s-1), approaching the diffusion limit6.
Eight oxygens per ion What is the structural basis for conduction in this pore with such finely-tuned potassium selectivity? Some of the answers to this question could be hypothesized based on the original lower resolution structure3, but a spectacular picture from the new 2.0 Å structure2 provides both confirmation and new insights. This structure was obtained from a complex of the bacterial KcsA channel protein with a monoclonal Fab that recognizes a peripheral extracellular loop of the channel. The crystal lattice has tetragonal symmetry, with one channel subunit and one Fab per asymmetric unit; the channel is 'assembled' from the proteins in the asymmetric unit based on the crystallographic symmetry. In the structure, the channel subunits interact with each other only within the tetrameric channel, with all of the other crystal contacts involving only the Fab. This approach of cocrystallizing membrane proteins with Fab is similar to that used by Michel7,
8. The result is remarkable, giving a clear view of the chemical basis of permeation as the potassium ion interacts with the channel protein and with some very interesting water molecules.
Besides the selectivity filter, the original view of the structure revealed the presence of a water-filled cavity located roughly in the middle of the membrane (Fig. 1). The cavity is apparently designed to make an ion feel at home even in the potentially least hospitable place — the low dielectric center of the membrane9. The piéce de resistance of the higher resolution structure is the clearly visible water structure surrounding a potassium ion located in the center of this cavity. A complete hydration shell of eight water molecules is arranged in a square antiprism, a distorted cube with the top twisted 45° relative to the bottom. This unprecedented view of a hydrated potassium ion is apparently made possible by the ability of the tetragonally symmetric protein to keep it from tumbling to invisibility. Anchoring of the hydration shell appears to be accomplished by specific hydrogen bonding with threonine hydroxyls at the bottom of the cavity, and may be assisted by the four pore helices, with the negative ends of their helix dipoles pointed toward the cavity9. The eight-coordinate ion−water complex is surrounded by roughly a single layer of unresolved second-shell waters, which directly contact the hydrophobic side chains lining the cavity walls.
As suspected from the original low resolution structure, the backbone carbonyl oxygens of the selectivity filter surround each ion binding site, with four oxygens above and four below each site (Fig. 1). These eight oxygen ligands — also organized approximately as a square antiprism — are now seen as a beautiful mimic of the inner hydration shell of the potassium ion. A chain of pre-formed binding sites stretches from the extracellular surface of the protein, through the selectivity filter, and eventually reaches the safe harbor of the water filled cavity. Completing the passage to the intracellular solution presumably requires a conformational change that allows an opening at the base of the cavity4,
10.
A conformational switch The K+ channel protein thus provides substantial stabilization to the K+ ions, to compensate them for the loss of their very favorable interactions with the bulk water outside the membrane. The converse is also true: the presence of potassium ions in the selectivity filter stabilizes the conducting conformation, in which the backbone carbonyls point toward each other, instead of participating in hydrogen bonds with solvent or with other protein groups. Consequently, when the protein is partially depleted of potassium, the selectivity filter adopts an alternative conformation, as seen in a second high resolution structure2 in low K+ concerntration (Fig. 2). The backbone carbonyls twist away from the center of the channel and point more tangentially, sometimes recruiting a water molecule into the body of the structure. This structural perturbation propagates slightly through the region surrounding the selectivity filter, with the outward facing side chains moving by 1 Å, but it does not much affect the lower half of the protein. The ion binding sites are clearly disrupted, indicating that this is a nonconducting conformation. This raises the possibility that the conformation switch seen in low [K+] might allow the selectivity filter to function not just as a exquisitely tuned ion permeation pathway, but also as a gate to regulate the channel open state.
Figure 2. Alternative conformations of the selectivity filter.
The structure of the selectivity filter backbone atoms is shown in the normal high [K+] (left) and low [K+] (right) forms. View is from the extracellular side, along the central axis.
What is the physiological correlate of this conformational switch? Biologically, K+ channels are quite diverse, and a single class of channel can have many different nonconducting conformations, or closed states. For the many voltage-dependent K+ channels found in animal cells, channel closure is coupled to a sensor of the transmembrane voltage. Channels open when the voltage changes from the normal resting voltage (negative inside the cell) toward zero. After voltage-dependent opening, other nonconducting states can come into play: a short-lived 'flicker' closed state (entered about once per ms and lasting 0.1−1 ms), or various longer-lived inactivated states (entered once per 20 or more ms and lasting for 100's of ms to many seconds). The inactivated states in particular are crucial to the channels' roles in neuronal signaling.
So which of these closed or nonconducting states corresponds to the low [K+] form of KcsA? From the crystal structures we know nothing about the time scale on which the deformation of the selectivity filter occurs. It probably occurs at different rates for different members of the K+ channel family, and corresponds to different physiological processes. For the well-studied voltage-dependent Shaker channels from Drosophila, the change bears similarity to the slow C-type inactivation process, which is promoted by K+ depletion11 and involves conformational changes detectable at the upper entrance to the selectivity filter12; the stability of this state can be altered by changing the aromatic residues surrounding the selectivity filter, or other nearby residues. The clearest functional correlate of the low [K+] KcsA structure is probably found in the 'inward rectifier' class of K+ channels, in which backbone mutagenesis of the selectivity filter, using unnatural amino acid substitutes, modifies the fast 'flicker' process seen in single channel measurements13. The physiological consequence of this is unknown, but it may correspond to the main gating of the related G-protein-gated and ATP-gated K+ channels.
In the Shaker channel10, and possibly for KcsA14, the main closing process occurs at the bottom of the cavity, toward the intracellular side. The narrow constriction seen here in all the known KcsA structures has a hydrophobic lining that is unlikely to permit ion passage. Closure of this intracellular gate will deprive the selectivity filter of a steady supply of intracellular K+ ions; depending on the extracellular K+ concentration and the channel affinity, this may indirectly promote the deformed state of the selectivity filter as seen in the low [K+] crystal structure. In other K+ channel relatives, the constriction at the base of the cavity may not be so narrow, and the selectivity filter may play the role of primary gate15 by closing in the manner seen for KcsA. In cases where the physiological trigger for channel closing and opening comes from the C-terminal intracellular domains (for example, ref 16), the four 'inner helices' that line the cavity may either act as a gate themselves (by constricting the bottom of the cavity), or they may convey a signal to close the selectivity filter, as seen in the low [K+] structure.
In any case, the high and low [K+] structures are the first examples of high resolution structures for a channel protein in two different conformations. They illustrate how a rather local and small conformational change can functionally gate the channel. We can look forward eagerly to seeing more conformational pairs, and learning how these changes are coupled to physiological stimuli such as voltage and ligand binding.
1 Å, but it does not much affect the lower half of the protein. The ion binding sites are clearly disrupted, indicating that this is a nonconducting conformation. This raises the possibility that the conformation switch seen in low [K+] might allow the selectivity filter to function not just as a exquisitely tuned ion permeation pathway, but also as a gate to regulate the channel open state.
