Size matters in activation/inhibition of ligand-gated ion channels

Abstract

Cys-loop, glutamate, and P2X receptors are ligand-gated ion channels (LGICs) with 5, 4, and 3 protomers, respectively. There is now growing atomic-level understanding of their gating mechanisms. Although each family is unique in the architecture of the ligand-binding pocket, the pathway for motions to propagate from ligand-binding domain to transmembrane domain, and the gating motions of the transmembrane domain are similar. Here we review common features among the LGICs. In particular, agonists and competitive antagonists apparently induce opposite motions of the binding pocket. A simple way to control the motional direction is ligand size. Agonists, usually small, induce closure of the binding pocket, leading to opening of the channel pore, whereas antagonists, usually large, induce opening of the binding pocket, thereby stabilizing the closed pore. A cross-family comparison of the gating mechanisms of the LGICs, focusing in particular on the role played by ligand size, provides new insight on channel activation/inhibition and design of pharmacological compounds.

Common threads of LGICs

As with change in transmembrane voltage, ligand binding is a common stimulus for ion channels. The term ligand-gated ion channels (LGICs) often specifically refers to three families of ionotropic receptors: Cys-loop receptors [in particular, nicotinic acetylcholine receptors (nAChRs)], ionotropic glutamate receptors (iGluRs), and P2X receptors (P2XRs). The functional units of these receptors are all oligomers comprised of identical or homologous protomers, but the numbers of protomers differ (Figure 1) [1–3]. Cys-loop receptors, iGluRs, and P2XRs have 5, 4, and 3 protomers, respectively. In each family, the minimum construct for channel function consists of an extracellular domain (to be referred to as the ligand-binding domain, or LBD) that harbors the ligand binding sites, and a transmembrane domain (TMD) that contains the pore for ion permeation, with additional domains that play other functional roles such as regulation of channel activity and trafficking. The ligand binding sites are located in the inter-protomer interfaces for Cys-loop receptors and P2XRs, but in the cleft between two lobes of each LBD protomer for iGluRs (Figure 1). With a significant amount of structural and mechanistic knowledge accumulated for each family of LGICs, it now seems appropriate to look for common lessons.

Figure 1.

Structures of three families of ligand-gated ion channels. The protomers in each receptor are shown in different colors; ligands are shown as spheres. (a) Muscle nAChR in the apo form (PDB entry 2BG9) [1]. The M3–M4 linker and M4 are not shown. Strychnine, an antagonist, is built into the binding site by superimposing against PDB entry 2XYS [9]. (b) GluR2 AMPA receptor bound with antagonist ZK 200775 (PDB entry 3KG2) [2]. In chain A (front right), the two lobes of the ligand binding domain are shown as dark (D1) and light (D2) blue surface. Chain B (front left) is in green. (c) Zebra fish P2X4R in the apo form (PDB entry 3H9V) [3]. ATP in the “distal” orientation, which was suggested to stabilize the closed state by molecular dynamics simulations [4], is built into the putative binding site.

A common set of questions that define the gating mechanisms of all the LGICs is: What rearrangement of the binding site does ligand binding induce? How are the motions of the LBD propagated to the TMD? And what are the motions of the pore-lining helices that are responsible for pore opening/closing? Given their distinct molecular architectures, the three families of LGICs are expected to have different solutions to these questions. However, as suggested recently [4], the different LGICs could have some common elements in their gating mechanisms. The gating mechanisms provide the basis for understanding ligand actions and designing pharmacological compounds. This review focuses on ligand size as a common factor in affecting ligand actions on the three families of LGICs.

Significant advance in the understanding of the gating mechanisms of Cys-loop receptors and iGluRs was made by comparing the structures of the LBDs bound with various agonists and antagonists. For Cys-loop receptors, this was made possible by the fact that the LBD is homologous to a water-soluble, homopentameric acetylcholine binding protein (AChBP) [5]. For iGluRs, this was made possible by constructs in which the TMD is removed, and the open ends are then connected by a short peptide linker [6, 7]. By comparing the iGluR LBD structures bound with agonists, partial agonists, and competitive antagonists, Armstrong et al. [7] observed that agonists induce significant cleft closure, partial agonists do so to a lesser extent, and antagonists keep the LBD in the open-cleft conformation. Hansen et al. [8] similarly observed that the AChBP C loop closes down on a nAChR agonist but opens up when an antagonist is bound. This observation was confirmed recently by a comparison of AChBP structures bound with various agonists and antagonists [9].

In developing a model for nAChR activation and inhibition, it was further observed that small ligands can fit better into a closed-down binding pocket, whereas larger ligands can fit better into an opened-up binding pocket, leading to the conclusion that agonists tend to be small and antagonists tend to be large [10]. As shown below, a systematic analysis of the available AChBP and iGluR LBD structures generally supports this conclusion. Although no antagonist-bound structure is yet available for any P2XR, structural models obtained from molecular dynamics simulations [4] suggest that the same conclusion may extend to P2XRs as well. It seems that ligand size is a key determinant in the activation/inhibition of all the three families of LGICs. Agonists, usually small, induce closure of the binding pocket, leading to opening of the channel pore, whereas competitive antagonists, usually large, induce opening of the binding pocket, thereby stabilizing the closed channel pore. This review pays only passing attention to inverse agonists, which close spontaneously open channels, and allosteric modulators, which bind the receptors at sites distinct from the agonist binding sites.

Cys-loop receptors

In addition to nAChRs, the Cys-loop family also includes 5-hydroxytryptamine (5-HT), γ-aminobutyric acid, and glycine receptors, and glutamate-gated chloride channels (GluCl); the first two are cation-selective, whereas the last three are anion-selective. In the LBD, each protomer is comprised of two β-sheets (Box 1). The C loop, connecting β9 and β10 in the outer sheet, moves a considerable distance to accommodate the bound ligand (Figure 2a). At the apex of the C loop is a pair of adjacent Cys residues (at positions 189 and 190 in the homomeric neuronal α7 nAChR) that form a disulfide bond. Among 31 AChBP structures in the Protein Data Bank (PDB), the distances between the first C-loop Cys residues of two neighboring protomers (d_Cys-Cys) span a wide range, from 35.5 to 48.4 Å (Table 1). A shorter inter-protomer d_Cys-Cys signifies closing down of the C loop, whereas a longer d_Cys-Cys signifies opening up of the C loop. Confirming previous observations [8, 9], histograms in d_Cys-Cys (Figure 2b) show that agonist-bound structures tend to have the C loop closed down, antagonist-bound structures tend to have the C loop opened up, and partial agonist-bound structures have intermediate opening of the C loop.

Figure 2.

Ligand-induced LBD conformational changes of Cys-loop receptors. (a) Superposition of an agonist-bound AChBP structure (PDB entry 1UW6) [15] (red) and an antagonist-bound AChBP structure (PDB entry 2BYP) [8] (blue). The inner β-sheets of the pentamer are used for superposition. Two of the five protomers are shown as cartoon; the remaining ones are shown as gray surface. C_α atoms of two Cys residues in neighboring C loops used for defining d_Cys-Cys are shown as yellow spheres. (b) Histograms of nAChR agonists, partial agonists, and antagonists binned according to d_Cys-Cys. (c) Correlation between d_Cys-Cys and ligand molecular weight. The PDB names of three outliers are given. (d) Compound 6 (shown as surface) bound to Aplysia califonica AChBP (PDB entry 2XNV) [11]. The C loop is on the left and in red; β2 is on the right and in blue. (e) Lobeline bound to Aplysia califonica AChBP (PDB entry 2BYS) [8].

Table 1.

Ligands bound to two types of ligand-binding domains

Ligand	molecular weight	PDB	ligand type	distance (Å)	reference
AChBPsa
Acetylcholine	146.2	2XZ5	agonist	36.4	[80]
Carbamylcholine	147.2	1UV6	agonist	37.6	[15]
Anabaseine	160.2	2WNL	agonist	36.7b	[81]
Nicotine	162.2	1UW6	agonist	36.5	[15]
Epibatidine	208.7	2BYQ	agonist	35.9	[8]
CAPS	221.3	2BJ0	buffer	35.5	[82]
Designed compound 1	231.3	2Y54	agonist	39.1	[58]
HEPES	238.3	1I9B	buffer	37.8	[5]
Clothianidin	249.7	2ZJV	agonist	37.2	[83]
Thiacloprid	252.7	3C84	agonist	37.6	[84]
Imidacloprid	255.7	3C79	agonist	38.1	[84]
Compound 18 (in silico screening)	261.8	2XNT	antagonist	44.5	[11]
Tropisetron	284.4	2WNC	partial agonist	38.7	[81]
4-OH-DMXBA	294.4	2WN9	partial agonist	42.0	[81]
Compound 6 (in silico screening)	304.4	2XNV	antagonist	45.0	[11]
DMXBA	308.4	2WNJ	partial agonist	38.0	[81]
Strychnine	334.4	2XYS	antagonist	40.7	[9]
Designed compound 4	335.4	2Y57	agonist	38.1	[58]
Lobeline	337.5	2BYS	agonist	36.3	[8]
In silico compound 31	348.5	2W8F	antagonist	43.3	[57]
Designed compound 3	351.4	2Y56	agonist	39.9	[58]
Designed compound 6	366.5	2Y58	agonist	38.2	[58]
In silico compound 35	410.6	2W8G	antagonist	44.9	[57]
Gymnodimine A	509.7	2X00	antagonist	41.7	[85]
d-Tubocurarine	610.7	2XYT	antagonist	43.0	[9]
Methyllycaconitine	682.8	3SIO	antagonist	44.8	[86]
13-Desmethyl spirolide C	694.0	2WZY	antagonist	45.0	[85]
α-Conotoxin IMI	12 resi	2BYP	antagonist	48.4	[8]
α-Conotoxin PnIA(A10L D14K)	16 resi	2BR8	antagonist	48.3	[87]
α-Cobratoxin	71 resi	1YI5	antagonist	44.6	[88]
		3SQ9	apo	41.9	[89]
GluR2 LBDc
Glutamate	147.1	2GFE	agonist	25.6	[90]
(S)-Des-Me-AMPA	172.1	1MQD	agonist	24.9	[91]
(R)-5-HPCA	184.2	3PD9	agonist	25.7	[92]
(S)-7-HPCA	184.2	3PD8	agonist	25.3	[92]
AMPA	186.2	1P1Q	agonist	26.0	[93]
Quisqualate	189.1	1MM7	agonist	25.6	[94]
(R)-TDPA	189.2	3BFU	agonist	24.8	[95]
(S)-TDPA	189.2	3BFT	agonist	25.4	[95]
Willardiine	199.2	1MQJ	partial agonist	25.7	[34]
Kainate	213.2	1GR2	partial agonist	26.5	[6]
ACPA	214.2	1M5E	agonist	26.1	[96]
Fluoro-willardiine	217.2	1MQI	partial agonist	25.8	[34]
(S)-ATPA	227.2	1NNP	agonist	26.5	[97]
Chloro-willardiine	233.6	3RTF	partial agonist	26.3	[98]
(S)-4-AHCP	238.2	1WVJ	agonist	26.2	[99]
(S)-CPW399	239.2	1SYH	agonist	25.5	[100]
Nitro-willardiine	244.2	3RTW	partial agonist	26.2	[98]
(S)-Thio-ATPA	244.3	2AIX	agonist	26.4	[101]
DNQX	250.1	1LB9	antagonist	28.0	[32]
Br-HIBO	251.0	1M5C	agonist	24.7	[96]
2-Me-Tet-AMPA	254.2	1M5B	agonist	25.9	[96]
Bromo-willardiine	278.1	1MY3	partial agonist	26.2	[102]
(S)-ATPO	322.3	1N0T	antagonist	28.1	[103]
Iodo-willardiine	325.1	1MQG	partial agonist	26.1	[34]
2-Bn-Tet-AMPA	330.3	2P2A	agonist	27.0	[104]
Compound 28	366.3	3R7X	antagonist	31.0	[105]
ZK 200775	409.3	3KG2	antagonist	30.1	[2]
(S)-NS1209	518.6	2CMO	antagonist	28.6	[106]

^aThe protein in 3SQ9 is a chimera of Lymnaea stagnalis AChBP and the LBD of α7 nAChR; the protein in 3SIO is a chimera of Aplysia califonica AChBP and the LBD of α7 nAChR. The proteins in the remaining entries are AChBPs from Lymnaea stagnalis, Aplysia califonica, or Bulinus truncatus. Seven other PDB entries containing AChBPs were not used for a variety of reasons [noncompetitive interactions (2PGZ and 2PH9); C loop disorder (2Y7Y) or asymmetry (2BYN, 2XZ6, 3GUA, and 3T4M)]. Distance listed in column 5 is d_Cys-Cys, calculated as the average over the pentamer. Nine other PDB entries were discarded because they contained ligands identical to those listed above; the retained entries have representative d_Cys-Cys, values.

^bOnly two of the five ligand binding sites in 2WNL are bound with the cyclic form of anabaseine, which acts as a nAChR agonist. The inter-protomer Cys-Cys distance was calculated after replicating one of the two protomers, whose C loops are in contact with a cyclic ligand, to the positions of the other three protomers.

^cDistance listed in column 5 is d_K410-K695, calculated on the first chain (typically chain A) of a dimer. Thirty-five other PDB entries were discarded because they contained ligands identical to those listed above; the retained entries have representative d_K410-K695 values.

The information in Table 1 also provides support to the notion that small ligands fit better into a closed-down binding pocket whereas larger ligands fit better into an opened-up binding pocket [10]. Among 27 structures bound with non-peptide ligands, there is a moderate correlation (R² = 0.48) between d_Cys-Cys and ligand molecular weight (Figure 2c). The peptide ligands have much higher molecular weights (though in each case much of the molecular weight may be positioned outside the ligand binding pocket) and correspond to larger d_Cys-Cys values, further buttressing the trend.

The outliers in the correlation plot of Figure 2c are interesting. In PDB 2XNV, the ligand is 2-[2-(4-phenylpiperidin-1-yl)ethyl]-1H-indole (compound 6 from an in silico screening) (Figure 2d), an α7 nAChR antagonist [11]. The bound ligand adopts a wedge-shaped conformation, with the indole and phenyl rings forming the two adjoining faces. The indole ring is stacked against the C loop of the protomer on the “principal” side, whereas the phenyl ring is stacked against strand β2 (in particular residue Y53, corresponding to α7 W54) of the protomer on the “complementary” side. [On the principal side, ligands are commonly found to contact a number of aromatic residues, including Y92, W148, Y187, and Y194, as well as C190 (α7 numbering) at the apex of the C loop (see Box 1).] The wedge shape, with the tips of the two apposing rings ~7 Å apart, allows the ligand to expend a minimal number of atoms to achieve a significant opening of the C loop. The ligand in PDB 2XNT, (2s)-2-[(4-chlorobenzyl)oxy]-2-phenylethanamine, behaves in much the same way. Again, a wedge-shaped ligand is positioned as if to pry open the inter-protomer interface.

By contrast, the situation is very different in the structure of AChBP bound with lobeline (PDB 2BYS) [8], an agonist (Figure 2e). The bound lobeline adopts an extended conformation, with the two terminal phenyl rings maximally separated from each other. The slender ligand is positioned nearly vertically, with the mid-section surrounded by the C loop on the principal side and by β2 Y53 on the complementary side. The two terminal phenyl rings form additional interactions: one positioned at the top interacting with the complementary side and the other positioned at the bottom interacting with the principal side. Lobeline thus distributes its atoms away from the C loop, explaining why a relatively high molecular-weight ligand can still reside entirely in a binding pocket with a closed C loop. Clearly, ligand size alone has its limitation in predicting ligand action. Reinforcing this point, a ligand can change from a partial agonist for one receptor subtype to a full or “super” agonist for another receptor subtype [12], or even from a competitive antagonist for the wild-type channel to an agonist for a mutant channel [13]. In addition, some competitive antagonists can also act, perhaps with the same mechanism of action, as inverse agonists for constitutively active mutant channels [14].

Although the most noticeable conformational change induced by agonists and antagonists is the movement of the C loop, in some agonist-bound structures, the rest of the LBD also seems to undergo conformational change (Figure 2a). In particular, the outer β-sheet appears to undergo a clockwise rotation (top view) around an axis that divides the two β-sheets. Based in part on a comparison between the 4-Å electron microscopy structure of the heteromeric muscle nAChR in the apo form and the structure of carbamylcholine-bound AChBP (PDB 1UV6; [15]), Unwin [1] proposed that the clockwise rotation is involved in channel activation. Recently, the structures of two bacterial homologues, ELIC and GLIC, of Cys-loop receptors were determined [16–18]. Putatively ELIC is in the closed state and GLIC is in the open state. Because ELIC and GLIC have low sequence identity (~20%), there is no unique way to align their structures, and different alignments have led to varying speculations on the motions of the extracellular domain that may trigger channel opening [17–20]. In particular, Zimmermann and Dutzler [19] suggested that a counterclockwise rotation, opposite to what was proposed by Unwin [1], of the extracellular domain leads to channel opening. Computational studies have also produced an assortment of models for LBD motions [10, 20–26].

The structures of ELIC and GLIC do unequivocally show that the external end of the pore-lining M2 helix undergoes a outward tilt to open the pore [17, 18]. That an M2 tilt constitutes the gating motion is supported by functional studies [27, 28] and by a recent structure of a GluCl channel in the putative open state [29]. So for Cys-loop receptors, there is now broad consensus that agonist-induced channel activation begins with the closing down of the C loop and ends with the outward tilt of the M2 helix. However, what additional motions are involved in between and how the motions are propagated remain unsettled [30, 31] (Box 2).

Glutamate receptors

iGluRs are divided into three main sub-types: α-amino-3-hydroxyl-5-methyl-4-isoxazolepropionic acid (AMPA), N-methyl-D-aspartate (NMDA), and kainate receptors. In 2009, the structure of a nearly full-length AMPA receptor (GluR2 homomeric) in the antagonist-bound form was determined [2]. This structure reveals that the LBD is a dimer of dimers, with limited contact between the dimers, whereas the TMD has four-fold symmetry, with helix M3 lining the pore (Figure 1b). All other iGluR structures in the PDB are for the isolated LBD (in dimeric form), with the most entries for AMPA receptor (GluR2). In each protomer, the LBD consists of two lobes, referred to as D1 and D2 that form a clamshell structure (Figure 3a). Depending on the ligand bound at the inter-lobe cleft, the opening of the clamshell, as measured by the distance (d_K410-K695) between residue K410 near the tip of D1 and residue K695 near the tip of D2 (GluR2 numbering; Box 1), varies widely, from 24.7 to 31.0 Å (Table 1). As observed by Armstrong et al. [7], agonists close the clamshell to a greater extent than partial agonists, whereas antagonists keep the clamshell open (Figure 3a, b).

Figure 3.

Ligand-induced LBD conformational changes of glutamate receptors. (a) Superposition of an agonist-bound GluR2 LBD structure (PDB entry 1M5C) [96] (red) and an antagonist-bound structure (PDB entry 3R7X) [105] (blue). The D1 lobe of a protomer is used for superposition. C_α atoms of K410 in D1 (dark red) and K695 in D2 (light red) used for defining d_K410-K695 are shown as yellow spheres. (b) Histograms of AMPA receptor agonists, partial agonists, and antagonists binned according to d_K410-K695. (c) Correlation between d_K410-K695 and ligand molecular weight.

Here again ligand size appears to be a determinant for the extent of the binding pocket opening. There is reasonable correlation (R² = 0.58) between d_K410-K695 and ligand molecular weight (Figure 3c). The trends shown by the histograms of Figure 3b and the correlation plot of Figure 3c for the GluR2 AMPA receptor are also seen for kainate receptors (there are too few NMDA receptor LBD structures in the PDB for statistical analysis). For both nAChRs and iGluRs, it takes ~70 Dalton of ligand molecular weight to increase the binding pocket opening by 1 Å.

Based on a comparison of the GluR2 LBD dimer structures bound with either an agonist (AMPA) or an antagonist (DNQX), Sun et al. [32] speculated that agonist-induced closure of the clamshell moves the tip of the D2 lobe away from the central axis, and this opens the channel pore. The structure of the nearly full-length GluR2 AMPA receptor in the resting state [2] showed that the outward movement of the D2 tip would drag the M3–D2 linker, which in turn pulls the M3 helix outward. Moreover, at the level of the M3–D2 linkers, the diagonal pair of protomers (chains B and D) distal to the central axis moves much more than the proximal pair (chains A and C), and hence contributes more to channel activation. Atomistic details of how agonist binding induces channel opening are revealed by targeted molecular dynamics simulations (see Glossary) [33]. The resulting activation model explains many observations in functional studies, in particular the correlation between the extent of ligand-induced LBD closure and the degree of channel activation [7, 34] (Box 2).

LGICs become desensitized in the sustained presence of an agonist. For iGluRs, significant structural information is available for the desensitized state (this is in contrast to the lack of such information for Cys-loop receptors). Desensitization involves the disruption of the D1-D1 dimeric interface [35–37]. A model for the LBD dimer in the desensitized state was created by introducing a disulfide bond in the D2-D2 dimeric interface (which leads to separation of the D1-D1 dimeric interface) [36]. Targeted molecular dynamics simulations showed that the separation of the D1-D1 dimeric interface pushes the D2–M3 linkers inward, thus largely reversing the motions seen during channel activation and leading to the closing of the pore [33].

P2X receptors

In mammals there are seven subtypes of P2XRs, denoted as P2X1 to P2X7 receptors. From a structural point of view, P2XRs have been the least well characterized among the three families of LGICs, but the recent determination of the zebrafish P2X4 receptor (zP2X4R) in the apo form [3] (Figure 1c) is changing the situation [38, 39]. Previous mutational studies [40–45] identified four conserved charged residues, K70, K72, R298, and K316 (zP2X4R number; Box 1), as part of the ATP-binding site. These four residues are clustered at the inter-protomer interface, with K70 and K72 on strand β1 of one protomer, and R298 on strand β13 and K316 on strand β14 of another (Figure 4); β1 and β14 are connected to TMD helices M1 and M2, respectively, the latter lining the pore (Figure 1c). Recent functional studies have illuminated the gating mechanism of P2XRs. The channel gate is located at the external portion of the M2 helix [46, 47]. Permeant ions likely access the transmembrane pore via the lateral fenestrations located just above the TMD [48, 49].

Figure 4.

Two opposite ATP binding orientations in P2X4R, suggested by NCS-ATP labeling [50] and refined by molecular dynamics simulations [4]. (a) ATP bound in the “proximal” orientation. Sidechains of the four-charge cluster, interacting with the triphosphate group, are shown. The backbone carbonyl of G294 forms a hydrogen bond with the adenine group. L191 corresponds to a position where NCS-ATP was labeled [50]. (b) ATP bound in the “distal” orientation. D145 corresponds to a second position where NCS-ATP was labeled [50] The antagonist suramin perhaps binds in a similar mode. D141 corresponds to a position implicated in suramin binding [56].

By cross-linking 8-thiocyano-ATP (NCS-ATP) with single cysteine substitutions around the putative binding site, Jiang et al. [50] have now obtained evidence that NCS-ATP can bind in two opposite orientations (Figure 4). One has the adenine ring on the TMD-proximal side of the four-charge cluster (FCC) and the other has the ring on the TMD-distal side. (A precedent for opposite ligand orientations was reported for an AChBP [51].) Interestingly, ATP-gated currents were potentiated by labeling NCS-ATP in the “proximal” orientation but inhibited in the “distal” orientation.

To elucidate a possible gating mechanism, Du et al. [4] carried out a normal mode analysis of the apo zP2X4R structure. A normal mode involving coupled motions of β1, β13, β14, and M2 was identified (Figure Ic in Box 2). In particular, the pre-M2 loop expanded outward. The resulting widening of the fenestrations above the TMD and opening of the transmembrane pore bear signatures of channel activation. The authors then docked ATP near the FCC in two opposite orientations, with the adenine either proximal or distal to the TMD, and followed with parallel molecular dynamics simulations (Figure 4). In the simulations with the proximal orientation, the ATP molecule became wedged between β13 and β14 of one protomer and β1 of the neighboring protomer, acting as a fulcrum that turned modest closure of the FCC into significant outward expansion of the pre-M2 loop. The simulations thus hint at how ATP binding triggers channel activation. The gating mechanism suggested by the computational studies explains the potentiation effect of NCS-ATP labeled in the “proximal” orientation as well as many other observations in functional studies (Box 2). Very recently, Hattori and Gouaux [52] determined a structure for zP2X4R in the ATP-bound form. This structure confirms key elements of the proposed gating mechanism, including an outward expansion of the pre-M2 loop. ATP is bound to a location similar to that in the simulations, but adopts a different conformation, with the adenine positioned to the outside, not inside, of the triphosphate group. In the molecular dynamics simulations with the distal orientation, ATP appeared to stabilize the tightening of the surrounding trimeric interface and the closure of the pre-M2 loop, possibly representing a desensitized state.

Box 2 Figure I.

Gating models for three families of ligand-gated ion channels. (a) Gating mechanism of Cys-loop receptors as suggested by superimposing the structures of ELIC (PDB entry 2VL0) [16] (in gray) and GLIC (PDB entry 3EHZ) [17] (in color). The M1 helices of the pentamer are used for superposition. (b) Gating mechanism of iGluRs suggested by targeted molecular dynamics simulations [33]. The four protomers have the same colors as in Figure 1c. Agonist-induced motions of the LBD and TMD are indicated by arrows. The D2 lobes move outward, pulling apart the outer portions of the M3 helices and listing up the TMD. (c) Gating mechanism of P2XRs developed from normal mode analysis and molecular dynamics simulations [4]. The resting state is shown in gray; bound ATP is shown as spheres, and the activated state is shown in color. M2 becomes less bent in the activated state.

Many P2XR antagonists have been identified [53–55], including a subtype non-selective, competitive antagonist called suramin. A mutational study [56] identified K138 in the human P2X1 receptor (corresponding to D141 in zP2X4R) as a possible suramin-interacting residue. In the zP2X4R structure, this residue is located on the distal side of the FCC (Figure 4b). This lends support to the suggestion that competitive antagonists may occupy the wider inter-protomer ridge above the FCC (similar to ATP in the distal orientation), thereby maintaining the closure of the pre-M2 loop. Many other P2XR antagonists are described as noncompetitive, and their binding sites remain unknown [53–55].

Insight for designing pharmacological compounds

The three families of LGICs are involved in many human diseases and are known drug targets. Structural information and mechanistic understanding are undoubtedly valuable for antagonist design. For both Cys-loop receptors and iGluRs, it seems clear that competitive antagonism requires sufficient opening of the ligand binding pocket, which in turn requires ligands with sufficient sizes. There are hints that the same may apply to P2XRs as well.

However, for ligands to become drugs, molecular weight (a measure of molecular size) is a premium. Compound 6, an α7 nAChR antagonist [11], points a way for eliciting maximal binding pocket opening with a minimal molecular weight. The wedge shape of the molecule is ideally suited to pry open the binding pocket. It will be interesting to see whether the same design can be adapted for iGluR and P2XR antagonists.

With more and more structures available for the LGICs, in silico screening has become increasingly useful for identifying high-affinity ligands [11, 57, 58]. Experimental verification is essential. In silico screening by Akdemir et al. [11] successfully identified compound 6, but not the docking pose in the crystal structure (PDB 2XNV). In the docking simulation the ligand adopted an extended conformation, instead of the wedge-shaped folded conformation in the crystal structure.

Subtype selective antagonists are more useful as drug leads but are more challenging to design. Subtype specific residues lining the ligand binding pocket will have to be taken into consideration.

Concluding remarks

In this review, common lessons among the three families of LGICs have been emphasized. In all these LGICs, agonists induce closure of the binding pocket, whereas competitive antagonists seem to reverse the direction of motion; ligand size appears to be a key determinant of binding pocket motional direction. The gates of these LGICs all seem to be located at the outer portion of a helix, with outward tilting the likely gating motion (Box 2). For Cys-loop receptors, the mechanism by which the binding pocket motion is transmitted to the gating motion has been controversial among many studies. Mechanistic models have recently been developed for the activation and desensitization of iGluRs [33] and P2XRs [4]. These likely will prompt additional studies and debate. The fact that the agonist binding pocket can accommodate larger ligands (usually antagonists) means that some residues that could interact with ligands do not participate in interactions with agonists, and hence do not directly contribute to channel activation.

To move the field forward, integration of structure determination, functional study, and modeling and simulation will be important. Structures will continue to be the basis for modeling and simulation, and be indispensible for guiding functional studies. Modeling and simulation will be necessary to cover the chemical space of ligands and the subtype space of receptors. They can also be used to yield atomistic information on the activated state, which usually is too transient for structure determination. Structural comparison between the resting state and the activated state is valuable in many respects, in particular for understanding the action of allosteric modulators. Of course verification by functional studies remains essential.

Box 1. Sequences and functionally important motifs in three families of LGICs.

The sequence alignment of several AChBPs and the LBDs of several Cys-loop receptors is as follows:

The LBD has 10 β-strands per protomer. β1, β2, β3, β5, and β6 form the inner sheet, with the remaining four strands (β4, β7, β9, and β10) forming the outer sheet. The loop connecting β9 and β10 is commonly referred to as C loop; the loop connecting β6 and β7 has a conserved pair of disulfide-bonded Cys residues (orange letters), and is referred to as the Cys-loop, which is the namesake for the protein family. Ligands are commonly found to contact C190 (orange letter; α7 numbering) at the apex of the C loop and a number of aromatic residues, including Y92, W148, Y187, and Y194 on the principal side and W54 on the complementary side (green letters; α7 numbering).

The TMD of a Cys-loop receptor has four α-helices per protomer; M2 lines the pore. The β1–β2 loop, Cys-loop, and M2–M3 linker are positioned at the LBD-TMD interface [1, 16–18], and are involved in propagating motions from the LBD to the TMD (Box 2).

The sequence and domain separation of the GluR2 AMPA receptor are shown below:

The LBD consists of two lobes, D1 and D2, as indicated by dark and light blue letters in the sequence. In the resting-state structure [2] (Figure 1b), the pore-lining M3 helices are longer by four residues in the “proximal” (chains A and C) protomers than in the “distal” (chains B and D) protomers. The C-terminal portion of the M3 helix contains the conserved SYTANLAAF motif and the channel gate.

The sequence alignment of zP2X4R with rat P2X1 to P2X3 receptors is given below:

Four conserved charge residues (red letters) putatively interact with the triphosphate group of bound ATP. A number of other positions implicated in ligand interactions are highlighted in green. The M2 helix lines the channel pore.

Box 2. Gating mechanisms of three families of LGICs.

The understanding of ligand actions on the three families of LGICs requires an understanding of their gating mechanisms. To provide the basis for rationalizing the proposed correlation between ligand size and ligand action, the current knowledge on the propagation of motions from the LBDs to the TMDs and the gating motions of the TMDs is summarized below.

Many studies of Cys-loop receptors have suggested that the β1–β2 loop, Cys-loop, and M2–M3 linker, positioned at the LBD-TMD interface [1, 16–18] (Figure Ia), are involved in propagating motions from the LBD to the TMD in triggering channel opening [59–70]. Using mutant cycle analysis, Sines and co-workers [64, 69, 70] identified a pathway, consisting of strongly coupled residues, e.g., between V46 in the β1–β2 loop and P272 in the M2–M3 linker (muscle nAChR α-protomer numbering), across the LBD-TMD interface. Based on rate-equilibrium free-energy relations, Auerbach and coworkers [59, 63, 68] further suggested that agonist binding generates a wave that propagates from the ligand binding site to the β1–β2 loop and Cys-loop, then to the M2–M3 linker, and finally down the M2 helix. The substituted cysteine accessibility method (SCAM) indicates that, upon channel activation, the lower end of β1 in α7 nAChR becomes less accessible but the lower end of β2 becomes more accessible [71]. Change in the propensity of substituted cysteines to form disulfide bonds between the β1–β2 loop and M2–M3 linker of a 5-HT₃ receptor suggests that these two elements move apart during channel activation [66].

Still unclear is the mechanism by which the motions at the ligand binding site are propagated across the LBD-TMD interface to drive the outward tilt of the external end of the pore-lining M2 helix. An idea that has been repeatedly raised is that the resting-state conformation is strained; agonist binding acts to relieve the strain and opens the pore as a result. Unwin [1] suggested that the strain is manifested by lack of five-fold symmetry in the LBD of the muscle nAChR. Specifically, the two α protomers, whose C loops provide the ligand binding sites, are rotated relative to the three non-α protomers. Lee et al. [69] considered the apo LBD as a brake that maintains a closed channel pore; ligand binding disengage the LBD from the TMD, allowing the pore to open. Similarly, Hilf and Duztler [17] envisioned breaking an unidentified critical interaction in the LBD-TMD interface.

Figure Ib illustrates the mechanism of GluR2 AMPA receptor activation developed from targeted molecular dynamics simulations [33]. As previously suggested by Sobolevsky et al. [2], the outward motion of each “distal” D2 tip drags the M3–D2 linker, which in turn pulls outward the external end of the pore-lining M3 helix, opening the pore and distorting the four-fold symmetry. The whole TMD also moves upward, such that the D2 lobes stay anchored on top of the cell membrane. Moreover, the diagonal pairs of M3–D2 linkers rotate relative to each other in the lateral plane, so that the four linkers are positioned more in line with four-fold symmetry. It is as though the abrupt transition between the two-fold symmetry of the M3–D2 linkers and the four-fold symmetry of the M3 helices in the resting state builds up a strain, which, upon agonist binding, drives the relative rotation between the proximal and distal pairs of M3–D2 linkers.

This activation model explains many observations in functional studies, including the pattern of agonist-induced accessibility changes from SCAM measurements of the M3 external portion [72]; the two-fold symmetry of the outer pore in the activated state [73]; and the effect of M3–D2 linker charge mutations on channel desensitization [74]. One could now also see why the extent of ligand-induced LBD closure is correlated with the degree of channel activation [7, 34]. As illustrated in Figure Ib, the closure of the D2 lobe is directly translated into the outward movement of the M3 external end and channel opening. Therefore the greatest LBD closure, as induced by agonists, opens the channel to the greatest extent. Moderate closure of the LBD clamshells, as induced by partial agonists, opens the channel to a moderate extent. The open LBD conformation, as maintained by antagonists, keeps the channel closed.

By combining normal mode analysis and molecular dynamics simulations, Du et al. [4] developed a gating mechanism for P2X4R (described in main text; Figure Ic). The atomistic model produced both a pattern of ATP-induced accessibility changes in M1 and M2 that is consistent with SCAM measurements [46, 75] and the observed expansion of the lateral fenestrations [48]. It also explains why an inter-protomer disulfide bond (presumably leading to interface closure) above the FCC potentiates the ATP-activated current [48], whereas a disulfide bond below the FCC has a reduced rate of formation in the presence of ATP [76].

A lipophilic compound ivermectin acts as a positive allosteric modulator of P2X4R. Ivermectin, extending ~20 Å in length, was suggested by mutational studies [77, 78] to be inserted between the M1 and M2 helices. As shown in Figure Ic, the spacing between M1 and M2 increases upon channel opening, providing room for ivermectin. Therefore, by preferentially binding, ivermectin may stabilize the open state, as previously proposed [79]. A similar effect of ivermectin was established by the putative open structure of a GluCl channel with ivermectin wedged between the M1 and M3 helices of two neighboring protomers [29].

Glossary

Mutant cycle analysis:

a way to obtain information on the contribution of the interaction between two neighboring residues to channel gating, by studying the effects of mutating these residues. The perturbations of the two single mutations and the double mutations on the free-energy difference between open and closed states of the channel (from measuring single-channel dwell times) are obtained. A significant deviation of the double mutation's effect from the sum of the two single mutations' effects indicates the importance of the interaction between the two residues in channel gating.

Normal mode analysis:

a modeling approach to obtain functionally important conformational changes of a protein. The motions of the protein are modeled as harmonic. By solving an eiganvalue problem, harmonic modes are obtained. Typically, some of the low-frequency modes, which involve collective motions of a large fraction of the atoms, are assumed to be associated with biological function.

Rate-equilibrium free-energy relation:

a way to assess whether the environment of a residue in the transition state for channel opening is more similar to that in the closed state or that in the open state. Mutations on the residue are made and their effects on the activation energy for channel opening and on the free-energy difference between open and closed states are obtained. The former are then expressed as a fraction of the latter. A fraction close to 0 indicates that the environment of the residue in the transition state is similar to that in the closed state; a fraction close to 1 indicates that the environment of the residue in the transition state is similar to that in the open state.

Substituted cysteine accessibility method:

a given residue in a channel protein is mutated into cysteine and a cysteine modifying agent is then applied in the absence and presence of an agonist. The change in modification rates is interpreted as indicating whether the residue becomes more or less accessible upon channel activation.

Targeted molecular dynamics simulation:

a part of a protein is forced to move in a given direction, and the goal is to see how the rest of the protein responds in the simulation. In particular, a part of the ligand-binding domain may be moved to mimic the motion induced by an agonist, and the response of the transmembrane domain may suggest a gating mechanism.