Bridging the Gap between Structural Models of Nicotinic Receptor Superfamily Ion Channels and Their Corresponding Functional States

Abstract

Aromatic–aromatic interactions are a prominent feature of the crystal structure of ELIC [Protein Data Bank (PDB) code 2VL0], a bacterial member of the nicotinic receptor superfamily of ion channels where five pore-facing phenylalanines come together to form a structure akin to a narrow iris that occludes the transmembrane pore. To identify the functional state of the channel that this structure represents, we engineered phenylalanines at various pore-facing positions of the muscle acetylcholine (ACh) receptor (one position at a time), including the position that aligns with the native phenylalanine 246 of ELIC, and assessed the consequences of such mutations using electrophysiological and toxin-binding assays. From our experiments, we conclude that the interaction among the side chains of pore-facing phenylalanines, rather than the accumulation of their independent effects, leads to the formation of a nonconductive conformation that is unresponsive to the application of ACh and is highly stable even in the absence of ligand. Moreover, electrophysiological recordings from a GLIC channel (another bacterial member of the superfamily) engineered to have a ring of phenylalanines at the corresponding pore-facing position suggest that this novel refractory state is distinct from the well-known desensitized state. It seems reasonable to propose then that it is in this peculiar nonconductive conformation that the ELIC channel was crystallized. It seems also reasonable to propose that, in the absence of rings of pore-facing aromatic side chains, such stable conformation may never be attained by the ACh receptor. Incidentally, we also noticed that the response of the proton-gated wild-type GLIC channel to a fast change in pH from pH 7.4 to pH 4.5 (on the extracellular side) is only transient, with the evoked current fading completely in a matter of  seconds. This raises the possibility that the crystal structures of GLIC obtained at pH 4.0 (PDB code 3EHZ) and pH 4.6 (PDB code 3EAM) correspond to the to the (well-known) desensitized state.

Introduction

Although the number of high-resolution structural models of membrane proteins continues to grow, our ability to associate structures with well-defined functional states lags far behind. Frequently, this is due to an incomplete understanding of the conformational dynamics of the protein in question (even in its native membrane microenvironment), a problem that is aggravated by the possibility that experimental artifacts (such as the effect of detergents1, 2 or the effect of forces developing within the crystal³) shift the equilibrium among conformations in hard-to-predict ways. In some other cases, structural data come from a given protein, whereas functional information is available for a distantly related ortholog, raising the inevitable question as to the extent to which the two sets of data can be compared.

The crystal structure of ELIC, a bacterial member of the nicotinic receptor superfamily⁴ whose conformational dynamics are still uncharacterized, was recently solved⁵ [3.3 Å resolution; Protein Data Bank (PDB) code 2VL0] and proposed as a model for the entire superfamily. Perhaps because the transmembrane portion of the ion permeation pathway is occluded near its extracellular end (minimum pore radius ≅ Na⁺ radius ≅ 1.0 Å; Fig. 1), this structural model was identified as the closed-channel conformation and was subsequently used to infer a mechanism for the closed ⇄  open gating conformational change.9, 10 But another structural model of the closed conformation of a nicotinic receptor, generated based on 4.0-Å-resolution cryo-electron microscopy data from the acetylcholine receptor (AChR) of Torpedo's electric organ (a muscle-type AChR) in the absence of activating ligands, features a much wider pore with a minimum radius of ∼ 2.5–3.0 Å at the center of the membrane⁷ (Fig. 1a). Thus, while the occluded pore of ELIC suggests that the closed ⇄  open gate poses a steric barrier to the permeation of ions, the model of Torpedo's AChR posits that this gate presents a desolvation penalty instead11, 12, 13 (radius of Na⁺ with the innermost hydration shell ≅ 3.0 Å¹⁴)—clearly two very different mechanisms for the same phenomenon.

Evidently, the resolution of the crystal structure (3.3 Å) is much better than that of the cryo-electron microscopy images (4.0 Å), but equally evident is the fact that the amino acid sequence of the AChR from Torpedo is much closer to the sequences of the functionally best-characterized members of the superfamily (such as the mouse muscle or human muscle AChR) than is any of the sequences from bacterial or archaeal origin described to date. So which model provides a more appropriate structural framework for interpreting the existing wealth of functional data?

The ion permeation pathway of ELIC is occluded because five pore-facing phenylalanines (phenylalanine 246), one per subunit, come together in a nearly T-shaped edge-to-face arrangement¹⁵ to form a structure akin to a narrow iris (Fig. 1c; Supplementary Fig. 1). Two other rings of pore-facing residues, the leucines at position 239 and the asparagines at position 250 (Supplementary Fig. 1), flank the phenylalanines at position 246 and contribute two additional constrictions (Fig. 1b), but the pore of ELIC is narrowest at the level of the central ring of phenylalanines (Fig. 1a; note that, throughout the article, we use the term “ring” to denote the arrangement of pore-facing atoms around the long axis of the channel rather than to refer to the benzene-ring-like properties of aromatic side chains). Owing to the low identity, there is some ambiguity in the alignment of primary sequences, but we favor the notion that positions 239, 246, and 250 of ELIC correspond to positions 9′, 16′, and 20′, respectively, of the second transmembrane (M2) segment of the better-understood metazoan members of the superfamily (Supplementary Fig. 1; others, however, have proposed that the ring of phenylalanines at position 246 aligns with position 20′; e.g., Hilf and Dutzler⁵).

Importantly, an inspection of sequenced genomes reveals that full rings of aromatic residues are not predicted to occur in vertebrate members of the superfamily at position 16′ or at any other pore-facing position, and that even the GLIC channel (a bacterial proton-gated homolog¹⁶ whose structure has also been determined by X-ray crystallography)9, 10 has an aliphatic residue at the aligned position⁴ (Supplementary Fig. 1). On the other hand, the amino acids forming the other two constrictions are not unique at all. Indeed, the leucine at position 239 of ELIC aligns with the very well-conserved leucine 9′ of the metazoan members of the superfamily, and the polar asparagine at position 250 aligns with the “extracellular ring of charge” (first identified in the AChR from Torpedo),¹⁷ which mainly contains asparagine, glutamine, and ionizable residues.

Admittedly, despite the very low sequence identity between ELIC and the AChR from Torpedo, many structural and functional features are remarkably well conserved; after all, ELIC is a bona fide member of the Cys-loop receptor superfamily even when lacking the signature cysteine loop. But it is conceivable that some sequence differences may have a more profound impact on structure and function than others. Therefore, we decided to focus on the central phenylalanines. Specifically, we asked how relevant the “aromatic plug” of ELIC and its mechanistic implications are to members of the superfamily that lack full rings of aromatic side chains at pore-facing positions.