Structural Basis for Allosteric Coupling at the Membrane-Protein Interface in Gloeobacter violaceus Ligand-gated Ion Channel (GLIC)*

Background: Allosteric mechanisms in ligand-gated ion-channels (pLGIC) that couple neurotransmitter binding to channel opening are poorly understood. GLIC is an important prokaryotic surrogate. Results: EPR studies at the junctional interface of GLIC reveal structural changes during desensitization. Conclusion: The closed conformation is characterized by extensive intrasubunit interactions at the junctional interface that weaken during desensitization. Significance: These studies elucidate the role of structural dynamics in pLGIC function. Ligand binding at the extracellular domain of pentameric ligand-gated ion channels initiates a relay of conformational changes that culminates at the gate within the transmembrane domain. The interface between the two domains is a key structural entity that governs gating. Molecular events in signal transduction at the interface are poorly defined because of its intrinsically dynamic nature combined with functional modulation by membrane lipid and water vestibules. Here we used electron paramagnetic resonance spectroscopy to delineate protein motions underlying Gloeobacter violaceus ligand-gated ion channel gating in a membrane environment and report the interface conformation in the closed and the desensitized states. Extensive intrasubunit interactions were observed in the closed state that are weakened upon desensitization and replaced by newer intersubunit contacts. Gating involves major rearrangements of the interfacial loops, accompanied by reorganization of the protein-lipid-water interface. These structural changes may serve as targets for modulation of gating by lipids, alcohols, and amphipathic drug molecules.

The Cys-loop superfamily of pentameric ligand-gated ion channels (pLGICs) 2 play a fundamental role in fast synaptic transmission. Members within the family are known to have evolved from a single ancestral gene and share a common overall architecture with significant homology. Ligand binding within the extracellular domain (ECD) initiates conformational changes that traverse 50 -60 Å across the channel and result in pore opening at the transmembrane domain (TMD). Although this allosteric transition is believed to involve a large-scale global change in the quaternary structure (1)(2)(3)(4), the molecular details of protein motions remain unclear. In the absence of a high-resolution structure of human pLGIC, valuable structural insights have been derived from closely related prokaryotic homologues from Gloeobacter violaceus (GLIC) and Erwinia chrysanthemi (ELIC) (3)(4)(5). Crystal structures of ELIC and GLIC have been proposed to represent the closed and the open states of the channel, respectively (3)(4)(5). However, significant controversies surround the functional states of the crystallographically trapped conformations (6). Interestingly, the structure of the agonist-bound glutamate-gated chloride channels from Caenorhabditis elegans closely resembles GLIC in its overall conformation.
The prokaryotic channels are emerging to be important surrogates for human pLGIC because of conservation of their overall structural scaffold and also of key functional properties such as selectivity, permeation, and desensitization (3,4,(7)(8)(9)(10). Furthermore, their remarkable sensitivity to alcohols, general anesthetics, and other clinically relevant compounds (11,12) make them attractive targets for drug design. Remarkably, chimeras of GLIC with other eukaryotic members of the LGIC family retain the functional properties of the individual domains, strongly suggesting that the pathway for allosteric communication is essentially conserved across prokaryotic and eukaryotic channels (13,14).
The long-range allosteric communication between the ECD and the TMD is mediated by the interfacial region between the two domains. The ECD-TMD interface is characterized by three main players: a network of interfacial loops, lipid molecules closely associated with the TMD, and water vestibules that extend from the ECD-TMD interface into the protein core. The interfacial loops consist of ␤1-␤2, ␤6-␤7, and ␤8-␤9 segments (within the ECD) and the pre-M1, M2-M3 linker and the C-terminal end of M4 (within the TMD) (Fig. 1, A and B). The functional consequences of mutational perturbations in these loops have been investigated extensively (Refs. 15-17 and references therein), and it is known that this region contributes to both the lifetime of the open state and the kinetics of desensitization (18). The interface also contains several charged residues, and gating models have been proposed on the basis of the interactions between specific charged pairs (19,20). However, it is to be noted that these interactions are not conserved across the family (21). The strategic location of the ECD-TMD interface also makes this region a likely target for lipid-mediated effects on channel function (22)(23)(24). Interestingly, sequence analysis predicts the presence of a potential cholesterol-binding motif in GLIC at the pre-M1 region of the interface (25). GLIC crystal structures also reveal associated lipid molecules, suggesting a tight interaction between the two (3,4). The packing of helical bundles within the TMD supports distinct water pockets, referred to as the intrasubunit and intersubunit cavities. These water vestibules form the binding site for several allosteric ligands and therapeutic agents (11,26,27) that modulate channel gating. Although these independent pieces of information point toward the indisputable role of the ECD-TMD region in allosteric mechanisms of gating and modulation, the mechanistic details of the interplay between each of these components are unknown.
A comparison of GLIC and ELIC structures shows small differences in the relative positioning of their interfacial loops (Fig.  1A). Such small changes in conformation are unexpected, given the dramatic functional impact of perturbations in this region in other members of the family (Refs. 15-17 and references therein). Interestingly, an overlap of GLIC structure with the nAChR model highlights large scale differences in the ECD-TMD interface between the two structures (Fig. 1B) despite a very similar M2 conformation (3,4). However, loops in the GLIC and ELIC crystal structures may be altered by the lack of a membrane environment. The dynamic nature of the interaction between these loops and water vestibules makes it difficult to draw direct conclusions about their gating motion.
We showed recently that EPR spectroscopy, in combination with functional measurements, is a powerful approach to elucidate protein motions that govern key gating transitions in GLIC. Membrane-reconstituted GLIC undergoes rapid pH-dependent activation, followed by transitions to a stable desensitized state (9). EPR measurements under steady-state conditions, therefore, conveniently allow us to probe the two end-state conformations: the closed state and the desensitized state. Structurally, at the level of M2, this transition involves a major conformational change at the extracellular end of M2. This change involves an outward translational movement leading to an increase in the pore diameter, whereas the intracellular end remains relatively immobile. In addition, the polar residues in the middle of M2 may contribute to desensitization by moving closer and occluding ion permeation (28). Here we investigated the structural changes at the ECD and TMD interface during transition to the desensitized state. We analyze these findings in the light of available gating models on the basis of x-ray structures and electron micrographs.

EXPERIMENTAL PROCEDURES
Protein Expression and Spin Labeling-The GLIC gene cloned into a modified pET26b vector was expressed as a fusion construct with N-terminal maltose binding protein as described previously (3,4). Briefly, BL21 (DE3) Escherichia coli cells transformed with the construct were grown in terrific broth medium containing 50 g/ml kanamycin at 37°C to A 600 of 1.0. Cells were induced with 0.2 mM isopropyl 1-thio-␤-D-galactopyranoside overnight at 20°C. Membranes were prepared by homogenizing the cells in buffer A (150 mM NaCl, 20 mM Tris base (pH 7.4)) with protease inhibitors and centrifuged at 100,000 ϫ g for 1 h. Membranes were solubilized in buffer A using 40 mM n-dodecyl-␤-D-maltopyranoside (Anatrace) at 4°C. GLIC was purified by binding to amylose resin and eluting with 20 mM maltose. The maltose binding protein tag was cleaved with human rhinovirus 3C protease (GE Healthcare), and the GLIC protein was separated using size exclusion chromatography on a Superdex 20/200 column (GE Healthcare).
Membrane Reconstitution and Functional Measurements-Electrophysiological measurements were made by patch clamp recordings in channel-reconstituted liposomes prepared as described earlier (9). Macroscopic currents for Cys mutants in this study were made in the absence of a spin label. However, it is likely that gating kinetics are further altered in the presence of the spin label. Purified protein was reconstituted into preformed asolectin vesicles by diluting in reconstitution buffer (150 mM NaCl, 10 mM HEPES (pH 7.0)). Detergent was removed by incubating the proteoliposome suspension with Biobeads (Bio-Rad). The suspension was centrifuged at 100,000 ϫ g for 1 h, and the pellet was resuspended in reconstitution buffer. A drop of the proteoliposome was placed on a glass slide and dried overnight in a desiccator at 4°C. The sam- ple was then rehydrated with 20 l of buffer, which yielded giant liposomes. GLIC was reconstituted in 1:20,000 protein: lipid (molar ratio) for macroscopic currents. Currents were measured using an inside-out patch clamp of proteoliposomes in symmetrical NaCl. All experiments were performed at room temperature. Recording pipettes were pulled from thin-walled borosilicate glass, heat-polished to a resistance of 1.5-2 M⍀, and filled with 150 mM NaCl, 10 mM HEPES (pH 8.0). Low pH was obtained using 10 mM sodium citrate buffer. Currents were elicited in response to pH jumps (8.0 to 2.5) using an RCS-200 fast solution exchanger (switch time, 2 ms) fed by gravity (Biologic). Currents were measured using Axopatch 200B, digitized at a 10-kHz sampling frequency, and analyzed using Clampfit 10.2.
Site-directed Spin Labeling-The native Cys (Cys-27) was mutated to Ser, and single Cys mutants in M2 were generated using the Cys-free construct (C27S) as the template. Purified protein was labeled with a methanethiosulfonate spin probe (1-oxyl-2,2,5,5-tetramethylpyrrolidin-3-yl) methyl methanethiosulfonate) (Toronto Research) at a 10:1 label:protein molar ratio. Mutants exhibiting dipolar coupling were underlabeled by protocols, as described previously (30). Briefly, mutants were labeled with a 0.5 molar excess of methanethiosulfonate label and incubated on ice for 30 min, after which a 20-fold molar excess of diamagnetic spin label (1-acetoxy-2,2,5,5-tetramethyl-⌬3-pyrroline-3-methyl) methane thiosulfonate was added and incubated for 2 h. The labeled protein was purified by size exclusion chromatography on a Superdex 20/200 column (GE Healthcare). Spin-labeled samples were reconstituted at a 1:3000 protein:lipid molar ratio in a mixture of asolectin, incubated with Biobeads, and centrifuged to obtain a pellet of the proteoliposomes. To determine the position of interfacial residues, a mixture of asolectin:DOGS-NTA[Ni(II)] lipids in a 3:1 molar ratio was used (31). EPR measurements were made for the closed and desensitized states at pH 8.0 and 2.5, respectively. The pH changes were made by equilibrating the liposomes at 42°C with appropriate buffers in a water bath. The sample was centrifuged, and the process was repeated multiple times to ensure complete buffer exchange. Reversibility of structural changes was ensured by switching back to pH 8.0.
EPR Spectroscopy and Analysis-Continuous wave EPR measurements were performed at room temperature on a Bruker EMX X-band spectrometer equipped with a dielectric resonator and a gas-permeable TPX plastic capillary. First derivative absorption spectra were recorded at an incident microwave power of 2.0 milliwatt, a modulation frequency of 100 kHz, and a modulation amplitude of 1.0 gauss. Our analyses were centered on three types of dynamic EPR structural information (32,33). The first is the mobility of the spin probe, calculated as the inverse of the central line width of the first derivative absorption spectra (⌬H o

Ϫ1
). This parameter is governed both by the local steric contacts in the immediate vicinity of the probe and by the flexibility of the backbone to which it is attached (34). As the frequency of nitroxide rotational motion is reduced, as witnessed during the formation of tertiary or quaternary contacts, the ⌬H o Ϫ1 value decreases for any particular motional geometry. On the contrary, structural motions leading to an increase in the freedom of movement of the probe is reflected as an increase in ⌬H o

Ϫ1
. The second is the proximity between intersubunit spins, estimated from amplitudes of the EPR signal. The spectral line shapes and the amplitudes of the signal are affected by the extent of through-space, short-range, spin-spin dipolar coupling between spin labels within the multimeric protein. The third is spin probe solvent accessibility, evaluated by collisional relaxation methods. Here, nonpolar molecular oxygen (⌸O 2 ) serves as a contrast agent to evaluate membrane accessibility, whereas polar Ni(II) ethylenediaminediacetic acid (⌸NiEdda) reports the extent of aqueous exposure (31,32). The accessibility to NiEdda (50 mM) for the desensitized state was measured at pH 3.0 instead of 2.5 because of the instability of the complex at very acidic pH. The accessibility parameter (⌸) is estimated from power saturation experiments in which the vertical peak-to-peak amplitude of the central line of the first derivative EPR spectra is measured as a function of increasing incident microwave power (32).

RESULTS
Single Cys mutations (on a Cys-free background, C27S) were generated along the entire length of ␤1-␤2, ␤4-␤5, ␤6-␤7, and ␤8-␤9 loops; the pre-M1 segment; the M2-M3 linker; and the C-terminal end of M4. The vast majority of Cys mutations (69 mutants) were well tolerated and did not alter protein folding, the oligomeric state, or the hydrodynamic property of the channel. All of the analyzed mutants in this study were stable, monodisperse pentameric populations with similar profiles to the WT on size exclusion chromatography ( Fig. 2A). Functional activities of the Cys mutations at several positions in each of the loops were confirmed by patch clamp measurements in proteoliposomes (Fig. 2B). As is expected of perturbations to key gating regions, some of the mutants altered the kinetics of activation and desensitization. Because our EPR measurements were carried out under steady-state conditions, they are less likely to be impacted by differences in the fast transitions. The Cys mutants were reacted with a thiol-specific nitroxide spin probe (1-oxyl-2,2,5,5-tetramethylpyrrolidin-3-yl) methyl methanethiosulfonate) and reconstituted in asolectin liposomes. EPR environmental parameters were determined for both the closed and pH-activated desensitized conformations (referred to as the desensitized state from here on). Changes in probe dynamics were evaluated from line shape differences (the inverse of the width of the central resonant line, ⌬H o Ϫ1 ), and the accessibility to either membrane soluble molecular O 2 (⌸O 2 ) or watersoluble NiEdda (⌸NiEdda) was determined from power saturation experiments (31,33).
Structural Changes at the TMD Loops-The ECD-TMD interface is lined by three critical regions from the TMD, namely, the pre-M1 segment, the M2-M3 linker, and the C-terminal end of M4. The pre-M1 region, which links the ␤-strands to the first TM segment, contains conserved arginines that are suggested to be important for gating, and mutations in the equivalent positions in all members of the family are shown to alter gating (35)(36)(37)(38). The M2-M3 linker is also a key transduction element, and mutations here have been implicated in human diseases (39 -43). The C-terminal end of M4 is relatively less conserved, but its role in gating is predicted by mutational studies and computational analysis (2,44,45). At the structural level, the nAChR model substantially differs from GLIC and ELIC in the location and extent of interactions of these regions. In particular, in nAChR, the M2-M3 linker fits snugly between the ECD and TMD within the subunits and is involved in extensive intrasubunit interactions with the ␤1-␤2 and ␤6-␤7 loops. On the other hand, both in ELIC and GLIC, the M2-M3 linker appears to be at the intersubunit interface. As a consequence, the M2-M3 linker in GLIC and ELIC is associated with intersubunit contacts with the external face of pre-M1 in the adjacent subunit, whereas, in the nAChR model, the pre-M1 region is relatively unconstrained (Fig. 1, A and B).
Similarly, the C terminus of M4 in GLIC and ELIC is positioned farther away than in nAChR and, thereby, contributes less to the intrasubunit ECD-TMD interactions.
We studied the relative positions and the extent of interactions between these regions in the membrane-bound closed (measured at steady-state conditions at pH 8.0) and desensitized conformations (measured at steady-state conditions at pH 2.5). Fig. 3 shows spectral line shapes, mobility, and accessibility parameters for pre-M1, the M2-M3 linker, and the C terminus of M4 in the two conformations. In the closed state, spin labels attached to positions in pre-M1 are in two contrasting environment (Fig. 3A, black traces). Residues that line up along the external face of M1  are mobile. Residues that point within the subunit core  show motional restriction (Fig.  3A, arrows). Among positions with high mobility, labels at Ser-191 and Gln-193 show high accessibility to water-soluble NiEdda and low exposure to membrane-soluble O 2 , suggesting that these side chains are surface-exposed and placed above the membrane (Fig. 3B). Positions Phe-195 and Ile-198 are more accessible to O 2 than NiEdda, indicating that their location is at the membrane boundary. In contrast, residues pointing within the subunit core  show limited accessibility to NiEdda and O 2 . Because the outward face of the pre-M1 region is more dynamic, with minimal protein contacts in comparison to the subunit-facing side, we conclude that the pre-M1 region in the closed state interacts less closely with the adjacent subunit and is more tightly associated with intrasubunit contacts.
Spin labels at the M2-M3 linker exhibit highly immobile spectra indicative of a substantially restricted environment (Fig. 3A). In particular, residues from Thr-244 to Lys-248 show very low O 2 and NiEdda accessibility values, suggesting that this region of the M2-M3 linker is completely buried within the protein (Fig. 3B). The C-terminal end of the linker is exposed to either NiEdda (Tyr-251) or O 2 (Thr-249 and Met-252). In addition, the EPR line shapes at several sites, most notably for residues Asn-245 (21Ј) and Lys-248 (24Ј), have extensive broadening (Fig. 3A, arrows). We reported previously such broadening in the closed conformation for pore-lining M2 residues (particularly at the C-terminal hydrophobic end, Leu-241 (17Ј)) (28). Because the line shape is affected both by the motional freedom of the probe and by dipolar coupling, we underlabeled Lys-248 (24Ј) by colabeling with a diamagnetic spin label (see "Experimental Procedures"). The underlabeled and fully labeled spectra for the Lys-248 (24Ј) position was compared with the spectra measured at the Leu-241 (17Ј) position in M2 (Fig. 4). The extent of spectral broadening in these spin-labeled mutants decreases when the channels are underlabeled, which clearly indicates an interaction between spin labels among different subunits. Such a close proximity between labels and a buried environment would suggest that the M2-M3 linker lies close to the M2 segment, making extensive contacts with the rest of the protein. This finding was somewhat unexpected, given its peripheral location in both GLIC and ELIC structures.
At the C terminus of M4, the line shapes at Phe-312, Leu-313, Gly-316, and Phe-317 reflect highly mobile spin labels as expected at lipid-exposed sites (Fig. 3A, black traces). On the opposite face, line shapes characteristic of motionally restricted spin labels occur at sites Leu-310, Phe-314, and Phe-315 and define the M4 surface that packs against the other TM helices (Fig. 3A, arrows). In addition, the membrane-facing residues show high O 2 exposure, whereas the residues on M4 facing the rest of the helices are less exposed to O 2 (Fig. 3B). The residues at the tip of M4 (Gly-316 and Phe-317) are accessible to NiEdda, indicating that this region is at the membrane-water interface. Remarkably contrasting dynamics and O 2 accessibility of residues on the two faces of M4 suggest that the C-terminal end in the closed conformation is very close to the helical bundle.
There are widespread pH-dependent changes in spin label dynamics (Fig. 3A, red traces) and accessibility patterns for the residues within the interfacial loops (B). Mapping the differences in the EPR parameters between the closed and desensitized states on the GLIC structure reveals the location and the magnitude of structural changes underlying the transition between the two conformations (Fig. 5). The most notable difference is in the M2-M3 linker, where residues that are fully buried in the closed conformation (Thr-244 to Lys-248) are now clearly more mobile and exposed to the lipids (Figs. 3B and 5). Additionally, the spin-spin coupling observed in the closed conformation disappears (Fig. 4). This suggests that M2-M3 moves away from a region of extensive protein-protein contact toward a membrane-exposed peripheral location.
In the pre-M1 segment, the residues pointing into the intrasubunit helical bundle  undergo an increase in motional freedom (Fig. 3A, arrows) and an enhanced accessibility to both NiEdda and O 2 (Figs. 3B and 5). A decrease in steric restrain for these residues is consistent with an outward movement of pre-M1 during desensitization. Such a motion would predict that the opposite face of the pre-M1 segment would be more constrained in this conformation. Indeed, several residues both on the external face of pre-M1 (Gln-193 and Phe-195) and in the M2-M3 linker (Tyr-251 and Met-252) show a decrease in either the mobility/water accessibility (Figs. 5 and 6), suggesting that the proximity between the M2-M3 linker and the pre-M1 segment from the adjacent subunit increases during desensitization.
At the C-terminal end of M4, residues facing the intrasubunit helical bundle (particularly Phe-314 and Phe-315, which are highly constrained in the closed conformation) undergo dra-  . Changes in dipolar spin-spin coupling at residues in M2 and the M2-M3 linker. EPR spectra of residues exhibiting spin-spin interactions. The red and black traces represent line shapes from the channel in the closed and pH-activated, desensitized conformations, respectively. The spectra marked by a dashed line were from underlabeled channels (in the presence of a diamagnetic label), and those marked by a solid line were from fully labeled channels. The inset shows an overlay of amplitude-normalized spectra. Broadening under fully labeled conditions is highlighted by arrows. The scan width is 150 G. matic increases in mobility (Fig. 3A, arrows) as well as accessibility to O 2 and NiEdda (Figs. 3B and 5). In comparison, membrane-facing residues undergo very little change in dynamics. The residues at the tip of M4 (Gly-316 and Phe-317 are not resolved in the structure) are less water-exposed and more membrane-buried upon desensitization. This implies that desensitization involves the C-terminal end of M4 moving outward away from the rest of the channel so that the residues at the very end of M4 move into the membrane.
To determine the relative positions of the interfacial residues to the membrane-aqueous interface, we reconstituted selected spin-labeled mutants in asolectin membranes incorporated with DOGS-NTA[Ni(II)] lipids. These synthetic lipids have a chelated Ni(II) bound on their head groups and, thereby, localize the paramagnetic relaxing agent at the membrane-aqueous interface within a region of 14 Å above the interface (in the lipid-extended configuration) (31). At positions Thr-244, Pro-247, and Thr-249 in the M2-M3 linker, the spin labels in the closed conformation are not accessible to DOGS-nickel, but upon desensitization they undergo rearrangement that exposes them to nickel on the lipid head groups (Fig. 7A). This is also consistent with a completely buried conformation for the M2-M3 linker in the closed state and a relatively unconstrained lipid-exposed orientation for this region in the desensitized state (Fig. 7B). In contrast, residue Ser-191 in the pre-M1 segment moves from a position highly exposed to DOGS-nickel (also accessible to NiEdda) in the closed conformation to a region of lesser accessibility in the desensitized state (less accessible to NiEdda but more to membrane O 2 ), which is in agreement with an outward movement of the pre-M1 region movement (Fig. 7, A and B).
As a consequence of these movements, there is an overall increase in dynamics and accessibility to NiEdda and O 2 of residues lining the intrasubunit helical bundle (namely,  (Pre-M1); Asn-245, Thr-249, and Thr-253 (M2-M3 linker); and Phe-314 and Phe-315 (M4)) (Figs. 3B and 5). These results indicate that the channels incorporated into membranes exhibit a tightly packed structural organization of the interface in the closed conformation. Notably, the channel opening (followed by desensitization) weakens the intrasubunit interactions and increases the solvent accessibility at positions that project into the intrasubunit cavity.
Structural Changes at the ECD Loops-The major loops from the ECD that line the interface are ␤1-␤2, ␤6-␤7 from the principal face, and ␤8-␤9 from the complementary face of the adjacent subunit. The ␤1-␤2 loop is highly variable in sequence but is conserved in length across the family of pLGIC with an over-  all negative charge. Mutations in this loop have been shown to effect EC 50 in the nAChR, GlyR, GABAR, and 5-HT 3 R channels and have, thereby, been implicated in channel gating (19 -21, 35, 46, 47). The ␤6-␤7 (or the Cys-loop) carries the canonical cysteine residues in the eukaryotic members of the family. Mutations in this region affect gating and also the expression of the channels on the membrane (19,35,46,48,49). Along with the ␤1-␤2 loop, the ␤6-␤7 loop is tightly engaged with the M2-M3 linker in nAChR (Fig. 1B). Both loops are found to be in an extended conformation and tilted away from the M2-M3 linker in GLIC and ELIC structures (Fig. 1A). In the nAChR channel, the tip of the ␤1-␤2 loop is buried within the hydrophobic pocket at the top of M2. A Val in this loop has been proposed to be involved in a "pin-into-socket" type of interaction in the resting state and is pulled away from the membrane during desensitization (50). ELIC and GLIC structures suggest a downward movement of ␤1-␤2 toward the M2-M3 linker during desensitization (3)(4)(5). The ␤8-␤9 loop, on the other hand, is long and unstructured and relatively less conserved among members of the family. This loop lies at the interface of the ECD and TMD and also between the principal and complimentary faces of the adjacent subunits (Fig. 1, A and B). Although crystal structures have not revealed significant changes in the conformation of the ␤8-␤9 loop, there is evidence of agonist-dependent movements in this region both in AChBP and in members of pLGIC (46,(51)(52)(53). In the open GluCL channel, ␤8-␤9 residues contact the M2-M3 linker of the adjacent subunit (54).
The EPR analysis of the three loops in the closed and desensitized conformation is shown in Fig. 8, and the ⌬ changes in parameters between the two conformations are highlighted in Fig. 9. In the closed conformation, residues in the ␤1-␤2 loop are highly constrained with very little accessibility to either O 2 or NiEdda, which suggests that this region is buried in a protein environment (Fig. 8, A and B). Interestingly, the ␤6-␤7 residues (except Asp-115 and Arg-118; Fig. 8A, arrows) also exhibit limited mobility. Although residues Asp-115 and Arg-118 are accessible to NiEdda, the other residues within the loop are occluded from both NiEdda and O 2 (Fig. 8B). This pattern of low accessibility and motional restriction is indicative of extensive protein-protein interactions at the interface between the ␤1-␤2 and ␤6-␤7 loops, presumably with the M2-M3 linker. In contrast, the spin labels attached to several positions in ␤8-␤9 are highly dynamic (Gly-150, Asp-154, Val-155, and Thr-158) (Fig. 8A, black traces). This is consistent with the intrinsically unstructured and flexible nature of the ␤8-␤9 loop. In addition, there is substantially high accessibility to water in this region, particularly at residues Val-149, Gly-150, Lys-151, Val-155, and Thr-158, suggesting minimal constraints from the adjacent subunit (Fig. 8B).
Upon desensitization, the ␤1-␤2 loop undergoes an increase in mobility and NiEdda accessibility. This effect is more pronounced at residue Ala-34, which is at the tip of this loop and, in the GLIC structure, is away from the interface (Figs. 8 and 9). Structural changes at this position suggest that this side chain is highly constrained, buried in a water/lipid-inaccessible pocket in the closed state and exposed during desensitization. Indeed, in the closed conformation, Ala-34 is not exposed to membrane-incorporated DOGS-nickel, and accessibility increases in the desensitized conformation (Fig. 7, A and B). Similarly, the ␤6-␤7 loop in this conformation is far less constrained in comparison to the closed state (Fig. 8A). The C-terminal end of this loop shows the most dramatic effect in terms of membrane exposure (Fig. 8B). The residues Leu-114, Phe-116, Tyr-119, and Phe-121 undergo enhanced flexibility and O 2 accessibility. In the GLIC structure, these side chains point toward the intrasubunit cavity and away from the interface between subunits (Fig. 9). The Phe-121 side chain in the GLIC structure interacts with a lipid molecule within this cavity (7). Overall, these changes in the dynamics and O 2 accessibility suggest that the ␤6-␤7 loop moves away from a buried environment (with extensive protein-protein contact at the interface) to more closely interact with the intrasubunit cavity (enhanced interaction with lipid molecules). Several of these residues also show a small increase in NiEdda accessibility, reflecting the amphipathic nature of this cavity. Consistent with this movement, membrane-facing Arg-118 shows a prominent increase in DOGS-nickel accessibility in the desensitized conformation, whereas Asp-115 is less exposed in this state (Fig. 7, A and B). Interestingly, the tightly associated lipid molecule in the GLIC structure closely interacts with Arg-118 through its polar head group (7). It is likely that this interaction plays a role in the open/desensitized conformations. At the level of ␤8-␤9, residues facing the adjacent subunit undergo a marked decrease in water accessibility (Asp-154, Phe-156, and Thr-158) and mobility, whereas those facing away from the interface (Asn-152, Leu-157, and Gly-159) show a marginal increase (Fig. 8). These changes in environmental parameters suggest that ␤8-␤9 is less flexible and interacts more closely with the adjacent subunit. Consistent with this idea, position Glu-35 in the ␤1-␤2 loop (which interacts with Thr-158 in the ␤8-␤9 loop in the GLIC structure) is also more restricted upon desensitization (Fig. 9).
Structural Changes at the ␤4-␤5 Loop-Conformational changes underlying activation originate near the ligand-binding site of pLIGIC. We have shown previously that the ␤9-␤10 loop (loop C), upon activation, adopts a sterically constrained conformation (9). This structural change essentially mirrors agonist-induced loop closure/immobilization that is reported in AChBP and other members of the pLGICs (1,55,56). Here we probed the ␤4-␤5 loop that is positioned between the ␤9-␤10 and ␤6-␤7 loops on the principal face, close to the ␤8-␤9 loop on the complimentary face (Fig. 10, A and B). The loop is strategically poised to transmit structural changes at ␤9-␤10 (in the vicinity of the binding site) to the ECD-TMD interface. Mutational characterization in eukaryotic members points toward the importance of this region in both ligand binding and channel gating (57)(58)(59).
In the closed conformation, the pattern of mobility and NiEdda accessibility (alternate high and low values) clearly defines the boundaries of the ␤-strands (Fig. 10B, black bars) and the short unstructured region of the ␤4-␤5 loop. The NiEdda accessibility values within the ␤4-␤5 loop increases progressively toward the ␤5 strand. Upon desensitization, although the ␤4-␤5 loop undergoes marginal increase in mobility and O 2 accessibility, there is a dramatic decrease in NiEdda accessibility. There is also an overall decrease in dynamics of the ␤5 strand (Fig. 10C). This decrease in mobility may also contribute to a decrease in NiEdda accessibility observed at these positions. These changes suggest that ␤4-␤5 may undergo a small inward movement away from the subunit interface, accompanied by a decrease in the flexibility of the ␤5 strand. A more thorough investigation of the dynamics of the ␤-strands is warranted to understand the underlying changes in dynamics of the ECD during gating.

DISCUSSION
Despite several landmark breakthroughs in the structure determination of pLGICs, fundamental aspects of the allosteric mechanisms underlying activation, desensitization, and channel modulation remain unclear. Although the functional state of the LGIC structures is still debatable, a fundamental difference in the ECD-TMD interface in ELIC, GLIC, and GluCL in comparison to nAChR is the relative positioning of the interfacial loops and the extent of interaction between them (1,3,5,7,54,60). Several factors may contribute to the observed structural differences between these channels (17,61). The structural models of nAChR in the unliganded and liganded conformations are based on 4-and 6-Å resolution data, respectively (1,62). On the other hand, the GLIC, ELIC, and GluCL structures are solved at a much higher resolution (2.4 -3.3 Å) (3-5, 7, 54). However, the nAChR model from cryoelectron micrographs hugely benefits from the fact that these channels are in their native Torpedo membranes in contrast to the crystal structures of ELIC, GLIC, and GluCL, which are in a detergent-solubilized environment. There is also growing evidence that crystallization conditions may bias structures toward specific conformations (6,11,63). Gating in these channels is intrinsically modulated by membrane lipids (22,64,65), likely by an allosteric mechanism that alters the relay of communication through the interface between the ECD and the TMD. Furthermore, it is also likely that these channels from different phylogenetic origins may have intrinsically different mechanisms.
This study systematically investigates the conformational changes at the intra-and intersubunit interfaces during ligandmediated desensitization in GLIC in a membrane-embedded FIGURE 9. Conformational rearrangements at the ECD loops during desensitization. Differences in the ⌬H o Ϫ1 , ⌸O 2 , and ⌸NiEdda values for the closed and desensitized states mapped on the GLIC structure and color-coded, with red denoting an increase and blue representing a decrease in the environmental parameter. environment. We propose a working model (Fig. 11) that is consistent with our structural analysis. However, alternate mechanisms are also feasible. Our findings clearly demonstrate that, in the closed conformation, the ECD-TMD interface is much more tightly constrained and buried than implicated by the putative closed structure of ELIC (Fig. 11). Specifically, the M2-M3 loop is positioned close to the intrasubunit helices in close contact with the ␤1-␤2 and ␤6-␤7 loops. In addition, the evidence of dipolar coupling in residues up to Lys-248 (24Ј) also suggests a close proximity of the side chains to each other. The location of the M2-M3 linker close to the M2 pore helix is in remarkable agreement with studies in GABA A receptors, where the ␤K24ЈC is shown to spontaneously form disulfide bonds in the closed conformation (66). Because cysteine ␣-carbon atoms must be within a distance of 5-6 Å of each other to form a disulfide bond, it has been proposed that the linker was either closer than what is seen in nAChR (ϳ17 Å) and GLIC (ϳ23 Å) structures or may be associated with large movements during gating. This conformation of the M2-M3 linker is also consistent with the "locally closed" GLIC structure (67) and with the liganded-closed state of GLIC mutants (67,68). A considerable degree of conformational flexibility in the M2-M3 linker (between the 16Ј to 24Ј positions) is also evident in the NMR structure of the GlyR TMD domain (69). It is to be noted that there is a register discrepancy of four residues (one turn of helix) at the beginning of the M1-M2 linker in the nAChR model in comparison to GLIC and GluCl structures (54,61). As a consequence, the beginning of the GLIC M2-M3 linker overlays on the M2 helix in the nAChR model (54,61). The pre-M1 region and the C terminus of M4 are also closely associated with the interactions at the interface. Interestingly, this extensive contact between the loops in the closed state is distinctly divergent from the ELIC conformation and more closely resembles the nAChR model in the unliganded state.
Even though a canonical ligand-binding site is absent in GLIC, pH-dependent activation leads to immobilization of loop C, as reported previously (9). There is an increasing consensus that agonists affect the mobility of loop C at the binding site and elicit closure of the binding cavity via capping by the C-loop, a movement that seems to be the initial conformational change underlying channel activation (29,55,56). We propose that the transition to the desensitized conformation is accompanied by an outward movement of the M2-M3 linker disengaging from the intrasubunit interactions with the ␤1-␤2 and ␤6-␤7 loops to establish intersubunit contacts with the pre-M1 region of the adjacent subunit. Studies in GlyR have shown that the substituted Cys in the M2-M3 linker is more water exposed in the open state than when the channel is in the closed conformation (20). The outward movement of M2-M3 also releases the ␤6-␤7 linker to now move downward into the intrasubunit cavity, interacting with tightly associated membrane lipids. Consequently, there is an enhanced interaction between the principal and complementary faces of the subunits, particularly between the M2-M3 linker and the pre-M1 region and also between the ␤1-␤2 and the ␤8-␤9 loops. The weakening of the intrasubunit contacts also occurs with outward movement of pre-M1 and the C terminus of the M4 segment, leading to a widening of the intrasubunit cavity. This outward movement of M4 has been implicated in lipid-dependent coupling and desensitization in nAChR (22). Interestingly, a decrease in the dynamics of the ␤-strands in the vicinity of loop C was also observed. Overall, the desensitized conformation is character- ized by weaker tertiary contacts at the interface, and the proposed conformation of this region is consistent with the GLIC crystal structure.
In summary, our findings provide a comprehensive view of the dynamic changes at the subunit interfaces during gating and how they relate to the conformations in the currently available structural models. These studies now pave the way to understanding how allosteric modulators affect gating by altering the interaction between the two domains. FIGURE 11. Schematic of structural changes at the ECD-TMD underlying GLIC gating. An EPR-based model predicting the structural changes at the interfacial loops during the transition to the desensitized state. Arrows indicate the putative direction of individual loop motion during transition to the desensitized state. The pH-dependent immobilization of the ␤9-␤10 loop has been reported previously (9). Key conformational changes during desensitization involve weakening of the intrasubunit interactions involving the ECD loops (␤1-␤2 and ␤6-␤7) and the TMD regions (pre-M1 region, the M2-M3 linker, and the C-terminal end of M4). As a consequence, there are new interactions at the intersubunit interface between the pre-M1 region and the M2-M3 linker from the adjacent subunit and also between loops ␤1-␤2 and ␤8-␤9.