Engineering differential charge selectivity from a single structural template

For any ion channel, both the identity and fundamental physiological roles are largely defined by ion selectivity. It is the regulated passage of some ions, but not others, across a cell membrane that determines specific membrane potential (V_m) changes that occur with channel activation. For example, opening of potassium (K⁺)-selective channels typically drives V_m to negative values, generally reducing excitability, whereas activation of voltage-dependent sodium (Na⁺)- or calcium (Ca²⁺)-selective channels contributes to strong depolarization. For the superfamily of tetrameric, strongly cation-selective channels, the determinants of selectivity arise in part from favored occupancy by permeant ions at one or more specific sites within the permeation pathway, sites readily identified in available structures (1, 2). Specific coordination of permeant cations by carbonyl oxygens or carboxylic acid side chains favors occupancy by the permeating species over other cations of slightly different radii (3, 4). However, other ion channels are not so explicitly selective and have evolved alternative filtering strategies. A remarkable case is that of members of the pentameric ligand-gated ion channel (pLGIC) superfamily, in which each member shares similar structural features, but some are cation-selective whereas others are anion-selective. How such functional distinctions arise among channels with highly conserved features has been a long-standing question. Previous work has focused on some specific charged loci in one or a few pLGIC members to identify potential influences on charge selectivity (5–8). However, as noted in a paper by Cymes and Grosman in PNAS (9), a detailed analysis of the literature reveals that a large number of discrepancies exist and that no consensus mechanism has yet been attained. For example, the elementary question as to whether charge selectivity is governed by interactions of the permeating ions with backbone atoms or side-chain atoms had not been answered conclusively—for the pLGIC superfamily—before this work.

Taking an approach that is really one of classic comparative physiology, but applied at the molecular level, Cymes and Grosman (9) undertake an exhaustive mutational analysis of various elements that might influence charge selectivity, defining K⁺/Cl⁻ permeability ratios using a traditional dilution-condition evaluation of ion selectivity. They compare six distinct pLGIC templates, including four cation-selective channels [the adult-muscle (α1)₂βδε nicotinic acetylcholine receptor (nAChR), the homomeric serotonin type 3A receptor (5-HT_3AR), the heteromeric serotonin type 3A–3B receptor (5-HT_3A-3BR) from mammals, and the bacterial ELIC channel] and two anion-selective channels [the α1 glycine receptor (α1GlyR) from mammals and a chimeric chick–Caenorhabditis elegans α7-AChR–β-GluCl receptor channel]. By selecting ion gradients to set E_K at −50.6 mV and E_Cl at +46.2 mV, agonist applications across voltages spanning the likely reversal potentials allowed determination of selectivity changes arising from any introduced mutation.

With this approach, the analysis leads to the conclusion that charge selectivity in the pLGIC family does not arise from any simple set of determinants. Although some ionizable residues located in the first turn of the cytosolic end of the M2 helix strongly influence charge selectivity, the impact of any given ionizable residue depends critically on other features of a given channel. Specifically, nonionizable residues at particular positions in the first turn of the M2 α-helix, differences in the length of the M1–M2 linker, and, in some cases, also other ionizable residues more distant from the M1–M2 region can all impact on charge selectivity. As the authors state, “The effect of mutations is context dependent.” For those seeking a simple explanation of charge selectivity in the pLGIC family, the results of this paper may seem disappointing. However, the work resolves a long-standing puzzle: why neutralization of the intracellular mouth glutamates (−1′ position) in some pLGICs failed to alter charge selectivity. On balance, Cymes and Grosman provide a tour-de-force example of the extent to which specific complex biophysical features of a protein family can be teased apart with appropriate mutational analyses. In this case, the importance of comparing an extensive set of related proteins cannot be underestimated.

So what elements do influence charge selectivity? The tested structural elements (some highlighted in Fig. 1A) include the following: (i) the first turn of the M2 transmembrane helix (positions 0′ to –3′); (ii) the M1–M2 linker (positions –4′ to –8′); (iii) residues at the 19′ and 20′ positions of M2; (iv) some charged residues in the extracellular domain; and (v) the cytosolic M3–M4 linker. Each of these elements in one or another study had been suggested to have some influence on some aspect of channel conductance or selectivity. Here, neither charged residues in the extracellular domain nor the M3–M4 linker influenced charge selectivity in wild-type channels (Fig. 1A, gray residues). In contrast, primary charge selectivity determinants in pLGICs reside in the first turn of the M2 transmembrane α-helix near the intracellular entrance of the ion permeation pathway, most notably at positions 0ʹ and −1′ (highlighted in Fig. 1B). Specifically, an acidic residue at position −1′ is important for cation selectivity. Neutralization of glutamates at −1′ (−1′E) in cation-selective pLGICs, such as 5-HT_3AR, 5-HT_3A-3BR, or ELIC channels, reduces the cation selectivity by more than 10-fold, whereas an alanine-to-glutamate substitution at position −1′ in α1 GlyR converts a channel from anion selective to cation selective. Interestingly, position 0′, which contains a basic residue in the anion-selective pLGIC subset, also contains a basic residue in most cation-selective pLGICs. Loss of expression with mutations at position 0′ has hindered its study. However, Cymes and Grosman find that the anion-selective α7-AChR–β-GluCl chimeric receptor is still functional following replacement of the 0′ arginine with either glutamine, asparagine, or glycine. With neutralization of the 0′ position, anion selectivity is almost abolished. The fact that basic residues are also present in cation-selective channels may reflect the fact the net effect on selectivity arises from combined influences of both 0′ and −1′ positions, such that, in cation-selective channels, the acid residue in position −1′ perhaps acts to counterbalance the basic residue at position 0′. Overall, the results show that the permeation free-energy landscape of pLGIC channels is influenced by electrostatic interactions between passing ions and ionized side chains at both −1′ and 0′ positions.

Fig. 1.

Key residues for pLGIC charge selectivity. (A) Side view of the crystal structure of GLIC cation-selective channel [PDB ID code 4HFI (12)] with front subunit removed. Extracellular domain and transmembrane/intracellular domain are colored in magenta and orange, respectively. Segments other than M1 and M2 are rendered semitransparent. Sites shown to have minor effect on charge selectivity of wild-type pLGICs are colored in gray. Three acidic residues (−5′, −1′, and 20′), which, when simultaneously neutralized, reduce AChR cation selectivity, are shown as blue spheres. (B) The crystal structures of GLIC channel (B1) and GluCl anion-selective channels (B2) [PDB ID code 3RIA (13)] viewed from intracellular side. Residues at positions −2′, −1′, and 0′ are rendered as ball-and-chain in white, blue, and red, respectively. The ion permeation pathway is denoted by the central Na⁺ (orange sphere in the GLIC structure). B3 shows a zoom-in view of these three residues in a GLIC M2, with homologous residues from the GluCl structure superimposed and rendered semitransparent. Images were prepared using UCSF (University of California, San Francisco) Chimera (14).

An important point regarding the cytosolic end of M2 is that, based on crystal structures of the cation-selective GLIC and the anion-selective α1 GluCl, the side chains of residues at positions −1′ and 0′ are not oriented toward the central permeation pathway (Fig. 1B). Thus, effects of ionizable groups on selectivity do not require that they be pore-facing. Indeed, from previous single-channel recordings, it was shown that the conserved 0′ basic residue is partially buried in open AChRs (10, 11). Other work from these same authors has also established that the formal charge of ionizable groups engineered on M1 and M3 α-helices—far from the pore axis—can also profoundly influence ion permeation (10). It is therefore not unexpected that the contributions of ionizable residues at positions 0′ and −1′ to charge selectivity

On balance, Cymes and Grosman provide a tour-de-force example of the extent to which specific complex biophysical features of a protein family can be teased apart with appropriate mutational analyses.

are sensitive to local microenvironment and to changes in the conformation of these side chains. Correspondingly, in the present paper, the influence of ionizable residues at positions 0′ or −1′ is shown to be affected by nearby nonionizable residues, insertions, or deletions that do not change the net charge. For example, whereas most anion-selective pLGICs contain a rigid proline at position −2′, small neutral residues such as glycine at this position favor cation permeation. In α1 GlyR, substitution of the −2′ proline with glycine, and shortening the M1–M2 linker by one residue, largely abolishes anion selectivity, whereas −2′ glycine-to-proline mutation in a mutant 5-HT_3AR converts this moderately cation-selective channel into an anion-selective channel. Thus, the mutational results can largely be explained by the idea that charge selectivity of pLGICs is determined by partially buried ionizable residues in the first turn of M2, and that side-chain conformation is likely to be influenced by neighboring nonionizable M2 residues and the associated structure of the M1–M2 linker.

Intriguingly, the classic pLGIC, the muscle nAChR, is remarkably resistant to mutation of charged residues near the first helical turn of M2, with neutralization of the −1′E having minimal effect. Unveiling perhaps additional complexity in the determinants of charge selectivity in pLGICs, simultaneous neutralization of three rings of acidic residues in the AChR pore—positions −5′E, −1′E, and 20′E—moderately reduces the cation selectivity of AChR (Fig. 1A). This indicates that charge selectivity in pLGICs can be influenced by additional ionizable residues other than those in the M2 primary selectivity filter. Cymes and Grosman therefore suggest that there is a “backup” system in pLGICs that influences the control of charge selectivity only after key ionizable residues in the primary selectivity filter are neutralized.

Although this work argues that charge selectivity is an emergent aspect of multiple components of a given channel sequence and its structure, interpretation of how specific mutations impact on selectivity requires some cautionary remarks. Some mutations that alter charge selectivity also produce substantial changes in channel-gating kinetics, suggesting influences on conformational dynamics. This raises the possibility that side-chain swapping may not be the only consequence of a mutation, but that there might be changes in helix properties that might reposition some side chains and alter their influence on ion permeation. For example, the strong shifts in selectivity that arise in the α1 GlyR through replacement of native proline residues with glycine might be explained by the idea there are changes in M2 helix rotation. However, arguing against this idea, the authors aligned the presumed open-channel structural models of a cation-selective pLGIC (GLIC) and an anion-selective pLGIC (α1 GluCl), for which the GLIC’s M1–M2 linker is two residues shorter than in α1 GluCl. Remarkably, the rotation, tilt, and length of each M2 α-helix were largely the same. This supports the idea that it is an effect of side-chain conformation that influences charge selectivity. However, the marked effect of charge selectivity-changing mutations on the time course of channel activation and deactivation remains to be understood. An attractive aspect of this work is that it should be possible to obtain structures for both mutated and parent constructs, enabling explicit comparison of functional and structural changes associated with particular sequences.

That channel-to-channel variation may occur in the specific determinants of a biophysical mechanism as fundamental as selectivity provides an interesting perspective pertinent to efforts to understand other biophysical mechanisms. For example, in voltage-gated channels, commonly we seek a simple unitary, evolutionarily conserved description of the molecular interactions and conformational changes that may underlie how voltage sensor movement couples to channel opening. Given the present example, it would suggest that the basic energetic determinants of any fundamental mechanism may involve channel-specific elements that will require considerable effort to unravel.