NMR Structure and Dynamics of a Designed Water-soluble Transmembrane Domain of Nicotinic Acetylcholine Receptor

Abstract

The nicotinic acetylcholine receptor (nAChR) is an important therapeutic target for a wide range of pathophysiological conditions, for which rational drug designs often require receptor structures at atomic resolution. Recent proof-of-concept studies demonstrated a water-solubilization approach to structure determination of membrane proteins by NMR (Slovic et al., PNAS, 101: 1828-1833, 2004; Ma et al., PNAS, 105: 16537-42, 2008). We report here the computational design and experimental characterization of WSA, a water-soluble protein with ∼83% sequence identity to the transmembrane (TM) domain of the nAChR α1 subunit. Although the design was based on a low-resolution structural template, the resulting high-resolution NMR structure agrees remarkably well with the recent crystal structure of the TM domains of the bacterial Gloeobacter violaceus pentameric ligand-gated ion channel (GLIC), demonstrating the robustness and general applicability of the approach. NMR T₂ dispersion measurements showed that the TM2 domain of the designed protein was dynamic, undergoing conformational exchange on the NMR timescale. Photoaffinity labeling with isoflurane and propofol photolabels identified a common binding site in the immediate proximity of the anesthetic binding site found in the crystal structure of the anesthetic-GLIC complex. Our results illustrate the usefulness of high-resolution NMR analyses of water-solubilized channel proteins for the discovery of potential drug binding sites.

1. Introduction

Nicotinic acetylcholine receptors (AChRs) belong to a superfamily of neurotransmitter-gated ion channels, which also include glycine (Gly), γ-amino-butyric acid type A (GABA_A), and serotonin (5-HT₃) receptors. These receptors, composed of five homologous subunits in a pentameric assembly, mediate fast synaptic transmissions. All subunits are comprised of a large extracellular domain, four transmembrane (TM) domains, and a long intracellular loop linking the third and fourth TM domains [1, 2]. Mutagenesis and functional investigations have shown that the binding of at least two agonists at the interfaces of the extracellular domains triggers the channel opening, allowing ions to pass through a TM pore surrounded by the TM2 helices from the five subunits.

The most studied member of this receptor superfamily is the highly abundant nAChR found in the Torpedo electric organs. The Torpedo nAChR, consisting of two α1 subunits and one each of β1, γ, and δ subunits, is homologous to the mammalian muscle-type nAChR responsible for rapid signaling at the neuromuscular junction. Although no atomic resolution structure is yet available for nAChRs, a 4-Å-resolution structure model has been determined by pioneering cryo-electron microscopy (cryo-EM) work on the Torpedo nAChR [3, 4]. More recently, high-resolution structural insights were gained from several crystal or NMR structures of related proteins, including an isolated extracellular domain of the nAChR α1 subunit bound to α-bungarotoxin [5], isolated four TM domains of the nAChR β2 subunit [6], the snail acetylcholine binding protein [7] which resembles the extracellular domain of the nAChR, and two prokaryotic pentameric ligand-gated ion channels from the bacterium Erwinia chrysanthemi (ELIC) [8] and the cyanobacterium Gloeobacter violaceus (GLIC) [9, 10]. The overall architecture of the extracellular domains is similar among the currently available structures. The TM domains, however, exhibit significant variation in high-resolution detail. The discrepancy among the TM domain structures may result from the difference in the membrane mimetic environments used for the structure determination, or from the difference in the ion-conducting states of the TM segments. It is worth noting that the 4-Å-resolution cryo-EM structure of the muscle-type nAChR shows four straight long helices in the TM domains. In contrast, the GLIC and ELIC crystal structures and the nAChR β2 NMR structure all exhibit relatively short and curved helical structures. The cryo-EM nAChR structure is presumably in the closed-channel state, whereas GLIC is either in a desensitized state or in a non-physiological artificial state due to the unexpected presence of detergent molecules within the aqueous pore. The ELIC crystal structure is thought to be in a collapsed state. Potential interactions with cholesterol, which are present in the eukaryotic membrane but absent in the prokaryotic membrane, can add another layer of complexity in TM domain structure determination.

The uncertainty in the TM domain structures reflects the technical difficulties associated with the structure determination of TM channel proteins. Crystallization trials involve the use of nonnative detergents that may or may not support the functional fold of the TM domains. The success in the expression, purification, and crystallization of prokaryotic channels such as ELIC and GLIC does not readily translate into practice for mammalian channels, which are difficult to overexpress in suitable quantities for structural studies. Because of their large hydrophobic surface in the TM segments, and the need for posttranslational modifications, mammalian channel proteins are also prone to aggregation, rendering standard solution-phase purification methods ineffective. Most biophysical studies involve dispersing membrane proteins in aqueous media, usually using detergents, lipids, auxiliary proteins, and other membrane-mimetic mixtures. Obtaining conditions to achieve monodispersity for X-ray crystallography or NMR studies remains a sensitive and exceedingly time consuming process.

An alternate, systematic approach is to redesign membrane proteins to remove the exposed hydrophobic exterior. In essence, TM proteins are transformed into soluble ones while retaining important structural and functional features. Using computational approaches, water-soluble variants of integral membrane proteins have been designed to facilitate structural studies. Recently, a computationally designed water-soluble variant of the bacterial potassium ion channel KcsA has been accomplished [11, 12] and its high-resolution structure has been elucidated by NMR [13]. The NMR structure of one of the three designed sequences of water-solubilized KcsA (WSK3) is in striking agreement with the crystal structure of the wild-type protein [13, 14]. These findings emphasize the promise of developing water-soluble variants of membrane proteins suitable for biophysical studies and potentially for structure-based drug discovery.

In this study, we present a similar approach for the TM domains of the nAChR α1 subunit. A water-soluble variant of the TM domains was computationally designed, resulting in the replacement of 23 hydrophobic residues with hydrophilic ones at the lipid-exposed surface. The resulting protein, called water-soluble acetylcholine receptor channel (WSA), was experimentally expressed and purified, and its structures determined using high-resolution NMR. Because the nAChRs are potential molecular targets for general anesthetics [15-19], we also determined the binding of two representative anesthetics, azi-isoflurane and azi-propofol, to the designed protein using photoaffinity labeling. We found that the designed protein retained an overall four-helix bundle topology and overlapped remarkably well with the crystal structure of the GLIC TM domains. A common binding site was identified for azi-isoflurane and azi-propofol in the general vicinity of a propofol binding site found in the crystal structure of anesthetic-bound GLIC. These results provide additional experimental evidence for the general applicability of the water solubilization approach to high-resolution structural investigation of membrane proteins.

2 Methods

2.1 Computational sequence design for water solubilization

The design principle involves the identification and redesign of exposed hydrophobic residues in the TM domains in order to transform a membrane protein into a completely or partially soluble protein while maintaining the structure- and function-related properties. The redesign of the nAChR α1 subunit was based on the 4-Å-resolution cryo-EM structure model as a template (PDB accession code: 1OED; Figure 1a) [3, 4]. The exposed residues were selected as those with more than 40% solvent exposure (Figure 1b) according to GETAREA1.1 [20]. Residues targeted for mutation were both hydrophobic (AFILMVW) and expected to reside in the TM region [3]. Hydrophobic residues at the interfaces between helices within the subunit were retained as in the wild type protein.

Figure 1.

Sequence redesign. (a) Ribbon representation of the pentameric structural template (PDB code: 2BG9) of nicotinic acetylcholine receptor (nAChR) used for the water solubilization sequence redesign. (b) The transmembrane domain of the α1 subunit in van der Waals representation before (left) and after (right) the computational redesign. Residues are colored by amino acid types: Green: hydrophilic (GNQSTY); white: hydrophobic (ACFILMPVW); blue: basic (HKR); and red: acidic (DE). (c) Cartoon representation of helix configuration in water-solubilized acetylcholine receptor (WSA). A polyglycine linker was added between TM1 (helical N-terminus) and TM4 (helical C-terminus) so that the TM domains can be expressed as a single chain. (d) Sequences of the transmembrane domains of nAChR α1 subunit, two calculated water-solubilization designs, and the bacterial pentameric ion channel GLIC are aligned for comparison. Water-solubilization mutations are marked in red. The most probable mutations are in italic boldface. Mutations changed by the inclusion of sequence diversity constraints are highlighted in green. Putative pore-lining residue positions are highlighted in blue. The transmembrane domain helices in the corresponding structures are underlined for comparison. Conformation heterogeneity in WSA is underlined with dashed lines.

The site-specific probabilities of the amino acids at the potential mutation sites were calculated as described previously [11, 21-25]. For each of the sites, all 20 amino acids were permitted while the remaining residues were constrained as wild type. Side chain conformations were restricted to those from a library of frequently observed conformations (rotamer states) [26] with at most 10 rotamer states per amino acid. Two energy functions, averaged over the sequence of the entire protein, were constrained to characterize the ensemble of soluble variants. The first energy term accounts for the inter-atomic interactions using a molecular mechanics force field (the AMBER force field) [27]. The second term quantifies the hydrophobic potential (environmental or solvation energy) that was constrained to an averaged value expected for a soluble protein with a size similar to the TM domains of the nAChR α1 subunit [11].

The computation for site-specific probabilities of amino acids at each mutation position was carried out in two stages. The first stage used the most probable amino acids at each of the targeted positions. At the second stage, only the sites where one amino acid was highly favored (probability of an amino acid exceeding 0.8) were constrained to that most probable amino acid. At the rest of the positions, an additional constraint was imposed to increase the diversity of different amino acids that appear in the final sequence. Amino acid diversity is desirable to avoid spectral over-crowding in the NMR spectra, thereby facilitating spectral assignment and structure determination. The sequence diversity constraint was implemented as an inverse participation ratio:

$$\frac{1}{m_{\mathit{\text{eff}}}}=\sum _{\alpha =1}^m{f}^2(\alpha )$$ (1)

$$f(\alpha )=\frac{1}{n}\sum _{i=1}^n{w}_i(\alpha )$$ (2)

where m is the number of possible amino acids, e.g., m = 20 if all the naturally occurring amino acids are permitted. The function f(α) is the frequency with which amino acid α appears among the sequences and is defined as the probability of amino acid α summed over the n targeted positions. The effective number of amino acids that appear at the targeted sites, m_eff, can be defined using the calculated amino acid probabilities; m_eff is bounded 1 ≤ m_eff ≤ m and it represents an average over the ensemble of sequences. For a homopolymer where only one amino acid is used to construct the sequence, m_eff = 1; if all 20 amino acids appear with equal frequency, m_eff = 20. Thus m_eff may be viewed as the effective number of amino acids used to construct the sequences.

The current design focuses on the TM domain without the extracellular domain and the intracellular loop. The removal of the intracellular loop separates TM4 from TM1-TM3. In order to express the designed protein as a single chain, a polyglycine linker was designed using the loop builder in MODELLER [28]. To minimize any structural biases and effects of an artificial linker on TM3 and TM4, and to take advantage of the additional length from the residues in the wild-type sequence leading to the TM1 helix and ending the TM4 helix (the C-terminus), we placed the artificial polyglycine linker between the C-terminus of helix TM4 and the N-terminus of helix TM1 (Figure 1c). This arrangement preserves the orientations of individual TM domains without restricting the relative spatial positions of TM3 and TM4, as in the topology of the original cryo-EM template.

2.2 Protein expression and purification

Hexahistidine-tagged WSA was expressed using BL21(DE3) competent Escherichia coli cells. Cells were first grown at 37°C in lysogeny broth (LB) media supplemented with 100 μg/ml ampicillin to an OD₆₀₀ of approximately 0.6. Cells were then spun-down and re-suspended into a quarter of the starting volume in ¹³C- and ¹⁵N-rich M9 media containing 100 μg/ml ampicillin and uniformly ¹³C-labeled glucose and ¹⁵NH₄SO₂ as the sole sources of carbon and nitrogen, respectively. After an hour of recovery at 30°C, cells were induced with 1 mM IPTG and grown overnight. Cells were then collected by centrifugation at 4°C and re-suspended in lysis buffer (50 mM Tris-HCl pH 7.5, 15% glycerol and 1 mM NaN₃). The suspended cells were incubated with lysozyme and DNase for ∼1 hour at room temperature, followed by sonication. The lysate was centrifuged at 18000 rpm (∼3880 g) for 20 min. The inclusion bodies were solubilized in 8 M urea in 20 mM Tris-HCl buffer (pH 8.0). After centrifugation (18000 rpm for 20 min), the supernatant was incubated with 2 mL Ni-NTA resin (Sigma) for 1 hour at room temperature. The resin was then washed with 40 mL of the washing buffer (8 M urea, 20 mM Tris-HCl, pH 8.0) and eluted with 6 mL of 250 mM imidazole, 8 M urea, 20 mM Tris-HCl, at pH 8.0. A size exclusion column was optionally used to improve the sample purity. The final protein yield was approximately 7 mg/L from M9 media.

WSA was then refolded in water at pH 8.0 by removing urea through dialysis. The protein was water soluble at high pHs and could be concentrated to at least 2 mg/mL. To prevent soluble aggregation when the pH was adjusted from 8 to 5.8, which was preferred for NMR measurements of the exchangeable backbone amide protons, 2% of 1-palmitoyl-2-hydroxy-sn-glycero-3-[phospho-rac-(1-glycerol)] (LPPG) (Avanti Lipids) was added to NMR samples to ensure monodispersity of the solubilized WSA.

2.3 Circular dichroism (CD)

CD measurements were performed using a Jasco Model J-810 spectrophotometer and analyzed using the K2D algorithm on the DICHROWEB server.

2.4 NMR spectroscopy and structure determination

Uniformly ¹⁵N- and ¹³C-labeled WSA was prepared in 2% LPPG, 10% D₂O, and 20 mM sodium phosphate buffer at pH 5.8. NMR data were collected at 40°C using Bruker Avance 600, 700, 800, or 900 MHz spectrometer controlled by the Topspin 1.3 software. The following pulse sequences from the Topspin sequence library were used for backbone and side chain spectral assignment: HNCOCA, HNCA, HNCACB, CBCACONH, HCCH-TOCSY, and 3D ¹⁵N- and ¹³C-edited NOESY. Mixing times for the 3D NOESY spectra were 120 ms and 150 ms for ¹⁵N-edited NOESY, and 150 ms for ¹³C-edited NOESY. Most of the 3D experiments were acquired with a complex time domain size of 1024 (direct observation dimension) × 48-80 (edit dimension) × 128-256 (indirect observation dimension).

The longitudinal and transverse relaxation rate constants, R₁ and R₂, and the steady-state ¹⁵N-{¹H} NOE were measured at 14.1 T using 4, 2.5, and 4 s recycle delays, respectively. R1 was recorded with 9 inversion-recovery times: 10, 50, 150, 300, 600, 1000, 1600, 2200, 3000 ms. R₂ was recorded using 7 spin-echo delays: 16, 32, 48, 64, 96, 128, and 176 ms. The steady-state ¹⁵N-{¹H} NOE was measured with and without ¹H saturation in an interleaved fashion. The R₂ dispersion measurements were carried out at 14.1 T and 40°C using the relaxation-compensated, constant-time, Carr-Purcell-Meiboom-Gill (CPMG) sequence with the following 8 CPMG field strengths: 25, 50, 75, 100, 150, 250, 400, and 650 Hz [29, 30]. The reference spectrum was acquired without the CPMG period. A relaxation delay of 2.5 s was used.

The NMRPipe [31] and Topspin programs were used to process the NMR spectra. The Sparky program [32] was used for resonance assignments. Structure calculation was performed using CYANA 2.1 [33] based on the distance constraints generated from ¹⁵N- and ¹³C-edited NOESY data, the dihedral angle constraints predicted by Talos [34] based on the chemical shifts, and the hydrogen bonding restraints derived from the temperature effect on the proton chemical shifts [35]. A total of 100 structures were calculated and the 30 structures with the lowest target function were refined using Cyana 3.0. The top 15 structures were further analyzed using Procheck-NMR [36, 37] and rendered using VMD [38] and Pymol [39].

The ¹⁵N relaxation parameters and ¹⁵N-{¹H} NOE were analyzed quantitatively using DYNAMICS with the extended model-free approach [40, 41]. A ¹⁵N-¹H internuclear bond distance of 1.02 Å and a ¹⁵N chemical shift anisotropy of −170 ppm were used. The global tumbling correlation time (τ_m) was determined by an extensive search for the τ_m value that gives the minimum χ² value and the least number of non-fitted residues in the Modelfree data fitting.

2.5 Photoaffinity labeling and anesthetic binding analysis

WSA expressed and purified as above was diluted into phosphate-buffered saline, pH 7, at 1 mg/ml, with either azi-isoflurane (1 mM final) or azi-propofol-m (azi-Pm, 100 μM final) added. These solutions were exposed to 300 nm or 350 nm light, respectively, at ∼1 cm distance for 15 min in 300-μL, 1-mm pathlength quartz cuvettes. Samples were then washed with 3 kDa centrifuge filters, and submitted for nano-LC/MS analysis. Briefly, samples were trypsinized and injected into a 10-cm C18 capillary column run at 200 nl/min for 60 minutes with gradient elution. MS-detected peptides were searched for adducts of the appropriate mass (196 Da for azi-isoflurane; 217 Da for azi-Pm) and then further fragment patterns were searched using Sequest to determine the adduct attachment site. MS work was performed at the University of Pennsylvania Proteomics Core Facility.

3. Results

3.1 WSA sequence design

Using the cryo-EM structure model of the nAChR α1 subunit [3, 4] as a template and 40% solvent accessibility surface as the criterion for exposed residues, we identified 29 hydrophobic residues within the TM domains as potential redesign targets. Of these, six residues at the interfaces between the helices were considered and kept unchanged: I219, M243, V261, F414, I420 and V425. The remaining 23 positions were targeted for computational redesign. They are I220, L223, L224, F227, L245, I247, L251, V255, F256, L258, V259, M282, I283, I286, I290, V293, V294, L411, M415, L416, I419, V423, and F426.

At the first stage of site-specific probability computation, the most probable amino acids at each of the 23 positions were calculated. The resulting sequence is shown in Figure 1d as Cal_1. This sequence was found to be cytotoxic to Escherichia coli, and the resulting expression yield was low. In the second stage of redesign, only the positions with amino acid probability > 0.8 were constrained to the most probable amino acids. The following six mutations satisfied this condition: L224K, F227E, L245E, V259K, I290K and I419K. At the remaining 17 positions, sequence diversity constraints were used according to Eq. (1) and Eq. (2). With 23 variable residues, the site-specific probabilities yield m_eff = 4.3 in the absence of any constraints on diversity. With the mutations at the above six positions fixed, the constraint was applied so as to achieve m_eff = 5.0. The most probable amino acid was selected at each of the 17 variable positions, yielding sequence Cal_2 in Figure 1d. This sequence has a theoretical isoelectric point (pI) of = 6.39 [42]. Also displayed in Figure 1d is an aligned sequence for the GLIC TM domains. The sequences are 83% identical between Cal_2 and the original α1 subunit of nAChR, and 11% between Cal_2 and GLIC. The residue numbering used in this study and the corresponding sequence numbering in the nAChR α1 subunit and in GLIC are shown in Figure 1d.

The lengths of the polyglycine linker (4 to 8 Glys) were investigated, and 50 independent calculations for each length of loop were performed using MODELLER [28]. The 5-glycine segment had the lowest effective energy (best score in MODELLER). The quality of the choice of loop length was also confirmed by energy minimization of the loop structures with NAMD [43] using the CHARMM force field [44].

3.2 Secondary structure of WSA by CD and NMR

CD measurements of WSA in water with 2% LPPG show an α-helical content of ∼50% (Figure 2a), significantly lower than the helical content of the cryo-EM template (∼83-87%) used in the sequence redesign, but in reasonable agreement with the estimation from FTIR and CD measurements of the authentic full-length nAChR [45] and GlyR [46, 47]. NMR chemical shift index (CSI), which provides residue-specific measures of the secondary structure, clearly shows four helical segments in WSA (Figure 2b), with ∼70% residues having a C_α chemical shift consistent with an α-helix secondary structure. This result is in excellent agreement with the helical content of the TM domains of GLIC. It is worth noting that residues S104-T105-S106-S107 (S266-S269 in the original sequence) between the TM2 and TM3 domains showed two sets of peaks (Figure 2c and Supplemental Figure S1): one has α-helical CSI, whereas the other set is non-helical. The fact that two peaks were observed on the NMR timescale for each residue indicated co-existence of two conformations in slow conformational exchange in this segment.

Figure 2.

Secondary structure determination. (a) Far-UV circular dichroism (CD) spectrum of WSA in water with 2% LPPG, showing an α-helical content of ∼50%. (b) The NMR chemical shift index (CSI), determined based on the backbone C_α resonance frequencies relative to the random coil values, is plotted as a function of residue numbers. Four helical domains are clearly defined. (c) Conformation heterogeneity in the segment between TM2 and TM3 is visible in the resonance peak doubling for residues S104-T105-S106-S107. SCI suggests slow exchange between helical and non-helical conformations.

3.3 High-resolution NMR structure of WSA

The advantage of water solubilization of TM channel proteins is clearly displayed in the NMR spectra of WSA (Figure 3a). Not only can WSA be expressed and purified in a large quantity for structural studies, the NMR spectra also showed that the water solubilization of this TM protein greatly improved monodispersity of the protein sample, allowing high-resolution NMR spectra to be acquired with homogenous peak intensities. Well-resolved 3D NMR spectra permitted spectral assignment of nearly all backbone resonances and many side chain resonances. With the exception of the segment STSS mentioned above, the majority of residues in the WSA sequence showed one set of NMR peaks, indicating a structural homogeneity. A total of 1165 NOE cross peaks were assigned, of which 42 were long-range inter-helical NOEs. These inter-helical NOEs defined the tertiary fold of WSA. The structures of WSA with the STSS segment in the major and minor conformations were determined separately using standard solution-state NMR methods. The 15 lowest target-function structures for each conformation are shown in Figure 3b (PDB ID: 2LKG and PDB ID: 2LKH for the major and minor conformations, respectively), and the structure with minimal root-mean-square deviation (RMSD) from the rest in the same conformation is highlighted in the ribbon representation. The NOE connectivity and hydrogen bonding constraints are depicted in supplemental Figure S2. The statistics of the structure calculation and results of the structure analyses are given in Table 1 and Table 2 for the major and minor conformations, respectively. The backbone RMSD over the four helical domains is 0.69 ± 0.13 Å and 0.93 ± 0.19 Å for the major and minor conformations, respectively. None of the structures have a distance violation > 0.5 Å or an angle violation > 5°, and 99.2% of the residues are in the most favored and additional allowed regions of the Ramachandran plot and no residues are in the disallowed region.

Figure 3.

Structure determination of WSA by high-resolution NMR. (a) Representative ¹H-¹⁵N HSQC NMR spectrum of WSA acquired at 40°C on an 800-MHz spectrometer, showing well resolved resonances of the water-solubilized membrane protein. (b) Bundles of the 15 lowest target function structures of major (left) and minor (right) conformations. The structures with the least RMSD from the rest of the bundle are depicted in the ribbon representation, with the TM4-TM1-TM2-TM3 domains colored in a red-to-blue color scale. The STSS segment is highlighted in magenta, and Loop-23 in the Cryo-EM template, which is a part of the TM3 helix in the WSA structure, is highlighted in cyan. (c) Superposition of the WSA NMR structure (blue) onto the TM domains of the crystal structure of GLIC (left, white) and the NMR structure of nAChR β2 subunit (right, white), which was solved in hexafluoroisopropanol [6].

Table 1. Statistics of the NMR structures of WSA in major conformation.

NMR structure	Statistics
Number of distance restraints	1165
Intraresidue (|i − j| = 0)	426
Short range (|i − j| = 1)	423
Medium range (1 < |i − j| ≤ 4)	274
Long-range, inter-helical (|i − j| ≥ 5)	42
Number of dihedral angle restraints	211
Number of hydrogen bond restraints	64
Number of upper limit restraints violations > 0.5 Å	0
Number of dihedral angle restraints violations > 5°	0
Backbone RMSD (Residues 11-40, 50-69, 80-95, 110-132)	0.69 ± 0.13 Å
Heavy atom RMSD (Residues 11-40, 50-69, 80-95, 110-130)	1.06 ± 0.12 Å
Ramachandran plot
Residues in most favored regions	93.2 %
Residues in additionally allowed regions	6.0 %
Residues in generously allowed regions	0.9 %
Residues in disallowed regions	0.0 %

Table 2. Statistics of the NMR structures of WSA in minor conformation.

NMR structure	Statistics
Number of distance restraints	1143
Intraresidue (|i − j| = 0)	414
Short range (|i − j| = 1)	416
Medium range (1 < |i − j| ≤ 4)	271
Long-range, inter-helical (|i − j| ≥ 5)	42
Number of dihedral angle restraints	201
Number of hydrogen bond restraints	64
Number of upper limit restraints violations > 0.5 Å	0
Number of dihedral angle restraints violations > 5°	0
Backbone RMSD (Residues 11-40, 50-69, 80-95, 110-132)	0.93 ± 0.19 Å
Heavy atom RMSD (Residues 11-40, 50-69, 80-95, 110-132)	1.28 ± 0.18 Å
Ramachandran plot
Residues in most favored regions	92.0 %
Residues in additionally allowed regions	7.2 %
Residues in generously allowed regions	0.8 %
Residues in disallowed regions	0.0 %

3.4 Backbone dynamics of WSA

Backbone ¹⁵N R₁ and R₂ relaxation parameters, as well as the steady-state heteronuclear NOEs, were measured for WSA at 14.1 T (600 MHz for ¹H resonances). R₂/R₁ ratios were calculated for all residues. Figure 4a summarizes these backbone dynamics data. The majority of the R₁ values are between 0.8 and 1.2 s^-1, and the R₂ values range from 7 to 18 s^-1. The hetNOE values are >0.6 for most of the residues in the helical regions, but significantly lower in the poly-glycine linker and loops between TM1 and TM2 (Loop-12) and between TM2 and TM3 (Loop-23). The average R₂/R₁ ratio for WSA is ∼15, suggesting that the NMR data describes a monomeric form of WSA as designed.

Figure 4.

Dynamics analysis of WSA. (a) Backbone amide ¹⁵N relaxation of WSA was measured at pH 5.5 and 40°C using a 600-MHz spectrometer. Longitudinal and transverse relaxation rate constants, R₁ and R₂, respectively, and ¹⁵N-¹H steady-state nuclear Overhauser effect (NOE) are plotted as a function of residue numbers. Error bars are the standard errors of fitting to the exponential decay function for R₁ and R₂ and the calculated uncertainties for NOE. (b) Quantitative model-free analysis of WSA dynamics using DYNAMICS program. S², generalized order parameters; t_e, the correlation time of local motion; R_ex, exchange contribution to R₂ relaxation; S_f², generalized order parameters for effective fast local motion. (c) Representative R₂ dispersions are plotted as a function of v_CPMG in unit of Hz, showing dynamics on the μs-ms timescale (TM2, blue; TM3, green; and TM4, red).

Quantitative analyses with the extended model-free approach were performed using the program DYNAMICS [41]. The overall correlation time used in the calculation was 12 ns, which was determined by minimizing the reduced χ² value. Results are also depicted in Figure 4b. The analyses yielded the generalized order parameters (S²), the local correlation times (τ_e), the exchange term (R_ex), the generalized order parameter for the fast effective local motion (S_f²), and the models used to fit the data for each residue. S² is above 0.7 and S_f² is 1 in the helical regions, which are typical for globular proteins. For the loop regions, most residues have τ_e ∼1-3 ns and S_f² is smaller than 1. Several residues displayed non-zero R_ex contributions to R₂, indicating that these residues might undergo conformation exchange. Particularly, the TM2 region showed more non-zero R_ex terms than other TM domains. This likely resulted from the computational redesign to promote monomeric structure, where the restriction from other TM2 domains that would otherwise be present in a pentameric structure is removed. CPMG R₂ dispersion measurements, which can detect slow motion on the μs-ms timescale, confirmed the presence of conformational exchanges in the TM2 domain and in Loop-12 and Loop-23 (Figure 4c). In particular, residues T105 and S106, which are a part of the STSS segment and show two sets of NMR peaks, have non-zero R_ex. The adjacent segment V109-P110-L111-I112 (271VPLI274), which forms Loop-23 in the Torpedo cryo-EM structure, also shows R₂ dispersion (V109) and non-zero R_ex (L111), suggesting conformational flexibility in this region.

3.5 Anesthetic binding to WSA

Using photoaffinity labeling, we investigated potential anesthetic binding to WSA. Peptides accounting for 48% of the sequence were detected for the azi-isoflurane labeled sample, and a single adduction site was found at V-31. Similarly, peptides accounting for 67% of the WSA sequence were detected in the aziPm labeled sample, and the same residue (V-31) was found to be the only modified site.

4. Discussion

4.1 Structural characteristics of WSA

We have solved the structure of a water-solubilized analogue of the TM domains of the nAChR α1 subunit using solution-state NMR spectroscopy. The protein was designed to be monomeric in solution, permitting secondary and tertiary structures to be determined at atomic resolution. For clarity, we focus our discussion on the WSA structures with the STSS segment in its major conformation. The structure shows the expected four-helix bundle fold, but with four unequal helix lengths. TM1 and TM2 are both six helix-turns long, shorter than TM3 (9 helix turns) and TM4 (8 helix turns). In comparison, the cryo-EM structure of Torpedo nAChR, which was used as the template in the computational sequence redesign, has four straight helices of roughly equal lengths (8 helix turns for TM1, TM2, and TM3, and 10 helix turns for TM4). The residues belonging to the helices as determined in the cryo-EM structure and the WSA NMR structure, along with those in the crystal structure of GLIC, are underlined in the sequences in Figure 1 for comparison. A superposition of WSA structure onto the monomer structure of GLIC is depicted in Figure 3c.

Although WSA has ∼83% sequence identity to the wild-type nAChR and only 11% identity to the GLIC TM domains, the high-resolution NMR structure of WSA agrees more with that of GLIC than with the cryo-EM template. This is rather unexpected. Several contributing factors can be considered. First, the experimental conditions under which the structures were determined were very different. The cryo-EM structure of nAChR was determined in highly ordered tubular crystal arrays rapidly frozen in liquid-nitrogen-cooled ethane. WSA NMR structure was determined at room temperature in a significantly more dynamic aqueous solution of lyso-lipids; 2% LPPG was needed to maintain monodispersity of the NMR sample when pH was lowered to 5.8. Similarly, GLIC was crystallized at a non-freezing temperature in ∼3% n-Undecyl-β-D-maltoside detergent. Moreover, nAChR normally resides in eukaryotic membranes that contain cholesterol, and cholesterol has been proposed to form a slowly exchangeable structural component of this family of ligand-gated ion channels [48]. The absence of cholesterol in WSA may therefore make it more analogous to the GLIC structure. Hence, the environment might play a more critical role than the amino acid sequences in affecting the high-resolution details of the structures. The debatable question remains as to which biophysical conditions are more relevant for membrane protein structure determination. Before high-resolution structures can be determined under physiologically relevant conditions, all currently available methods, including crystallography, can conceivably introduce some degree of artifact. Second, the high-resolution difference between WSA and cryo-EM nAChR structures may result from different equilibriums of conformational states. The cryo-EM structure is believed to mainly capture the closed state of the channel [3], whereas GLIC crystal is thought to be in either an open or desensitized state [9, 10]. Because the WSA structure was determined as a monomer, its conformational state can only be inferred. The fact that the WSA structure agrees so well with the GLIC structure suggests that the WSA monomer may reflect the conformation of the open or desensitized state of the nAChR TM domains. Third, there is also a possibility that the WSA structure is distorted due to the absence of extracellular and intracellular domains or due to water solubilization itself. This scenario is very unlikely, however, given the agreement between WSA and GLIC structures. GLIC was determined as an intact channel protein. Taken together, our results suggest that water-solubilization of membrane proteins is equally relevant for structure determination compared to more conventional lipid or detergent solubilization approaches. The ease in protein expression and purification and, more importantly, the greatly improved spectral quality of the resulting proteins after water solubilization, make water solubilization an attractive alternative. We believe that the NMR structures of water-solubilized proteins, including WSK3 published previously [13] and WSA determined in this study, represent the first approximation of the native structures. These structures provide the high-resolution templates for further refinement by other biophysical means.

4.2 Dynamic characteristics of WSA

It is conceivable that water solubilization of membrane proteins can change the motional characteristics of these proteins. For WSA, several residues in the TM2 domain were mutated from hydrophobic to charged amino acids for the purpose of promoting monomeric form in order to increase NMR spectral resolution. There are seven mutations in the TM2 helix, four of which, L89K (L9′K), V93E (V13′E), L96K (L16′K), and V97K (V17′K) (see Figure 1d for relative numbering of TM2), are located at the putative pore-lining sites of the nAChR channel. The charged residues likely prevent the formation of oligomerization by TM2. Without association with other TM2 helices, the TM2 domain becomes particularly dynamic in WSA. Both backbone dynamics measurements and R₂ dispersion measurements showed either non-zero R_ex contributions to the R₂ relaxation (residues 8′, 10′, 15′, 17′, 18′, and 19′, see Figure 4) or R₂ dispersion on the μs timescale (residues −2′, 2′, 5′, 9′, 17′, and 20′). The slow motion captured in R₂ dispersion suggests possible conformation exchange. In comparison, TM1, TM3, and TM4 domains showed significantly fewer numbers of residues with R_ex term and R₂ dispersion (Figure 4). Although the increase in dynamics of TM2 in WSA is undoubtedly the result of water solubilization, the fact that other TM domains do not experience the same degree of increase suggests that the motional characteristics of TM2, controlled by Loop-12 and Loop-23, might be intrinsically different from the rest of the protein. Indeed, it has been suggested that nAChR TM domains can be separated into inner helices, comprised of TM2 domains, and outer helices, comprised of TM1, TM3, and TM4 helices [3]. Channel gating is thought to involve the movement of the inner helices relative to the outer helices.

The dynamics of the two loops, Loop-12 and Loop-23, deserves further examination. As shown in Figure 1d, the WSA loop regions coincide more with those in the GLIC structure than in the nAChR cryo-EM structure. Loop-12 consists of five residues L73-S77 (LPTDS) in WSA, which differs from the very tight, three-residue turn (SGE) in the cryo-EM structure. The loop residues are also shifted. Loop-12 in WSA is a part of the TM1 helix in the nAChR cryo-EM structure, whereas Loop-12 in the cryo-EM structure forms the beginning of the TM2 helix in WSA. The NMR dynamics measurements show that the segment from Y72 to K80, encompassing Loop-12 in both the WSA (LPTDS) and cryo-EM (SGE) structures, is highly flexible. The relaxation data in this segment can only be fitted with model 4 or model 5 with ∼1-3 ns local motions (τ_e) and reduced S² and S_f² (see Figure 4b). Although the Loop-12 conformation as shown in the NMR structures is the preferred conformation based on the NMR data, the shifted Loop-12 conformation as displayed in the Cryo-EM structure is also possible. We found that E79 (E–1′) did not meet the basic Lipari-Szabo assumption in the model-free analysis and could only be fitted by model 5. The adjacent residue, G78 (G–2′), shows R₂ dispersion on the μs timescale. Nevertheless, we believe that the Loop-12 conformation as shown in the NMR structures, with P74 in the loop instead of in the TM1 helix, is the dominant loop population in the conformational equilibrium.

Similarly, Loop-23 between the TM2 and TM3 domain is also highly dynamic. In the intact receptor, this loop is believed to interact with the Cys-loop and β12 loop in the extracellular domain to mediate ligand binding to channel gating [17, 18, 49, 50]. In the WSA structure, the TM2 helix ends at E20′, whereas in the Cryo-EM structure, the TM2 helix extends to the extracellular space and ends at S27′ or A28′. None of the high-resolution structures solved so far for Cys-loop receptors, including the WSA structure solved in this study, the NMR structure of the nAChR β2 subunit TM domains [6], the high resolution NMR structures of glycine receptor TM2-TM3 domains solved in DMPC/DHPC bicelles [51], the crystal structure of GLIC [10], and the crystal structure ELIC [8], agree with the cryo-EM structure in this region. Our dynamics data suggest that several conformations can co-exist in Loop-23. Even with water solubilization, the dynamic characteristics seem to have been preserved. High-resolution NMR spectra of pentameric glycine receptor TM domains in lyso-lipid micelles also exhibit peak doubling in this segment (data not shown), suggesting that slow conformational exchange is an intrinsic dynamic feature in the region between TM2 and TM3 helices. The conformation heterogeneity of Loop-23 may be required for the TM2 movement during channel gating.

4.3 Anesthetic binding to WSA

To further confirm the similarity of WSA to GLIC, we performed binding studies. Recently, X-ray structures of GLIC were solved with two different general anesthetics—propofol and a haloether, desflurane—bound [52]. Like GLIC, nAChR is also sensitive to general anesthetics, so confirmation of anesthetic binding will provide a functional measure of the fidelity of the water solubilization paradigm. We found that both a propofol photolabel and a haloether photolabel bound to the same residue (V31) in the WSA structure (Figure 5). Also highlighted in this figure are residues that border the propofol-binding site in the GLIC-anesthetic co-crystal structure. The agreement with respect to the Z-direction along the bundle is remarkable. Although the side chain of V31 is oriented away from the intrasubunit GLIC anesthetic site, we have prior evidence suggesting that backbone carbonyl atoms are the preferred photoadduction sites [53], especially for residues with apolar hydrocarbon side chains. The V31 carbonyl oxygen is oriented such that it is easily accessed from the intrasubunit cavity. Thus, it would appear that the anesthetic site in GLIC and WSA is similarly positioned within the bundle. This result further validates the water-solubilization approach: mutation of surface residues to improve water solubility does not greatly alter the overall fold and internal packing of the protein, preserving potential drug binding sites for structure-based drug screening. However, it is important to note that neither the propofol nor the isoflurane binding site has been definitely located in nAChR, thus we cannot yet conclude with confidence that the WSA site well represents the anesthetic binding site in the native nAChR.

Figure 5.

Comparison of anesthetic binding site in WSA and in GLIC. The WSA structure (purple) is fitted to the TM domain of the crystal structure of GLIC (yellow). The anesthetic binding site in WSA was determined by photoaffinity labeling. Both Azi-isoflurane and Azi-propofol bind to residue V31 located in TM4 (cyan). The dominant propofol binding site in GLIC crystal is highlighted in orange. Notice that the general location of the binding sites is essentially the same, suggesting that the drug-binding site is preserved in the water-solubilized protein.

4.4 WSA-detergent interaction site

Although WSA was soluble in water at a high pH, lowering the pH to pass the theoretical pI for NMR experiments can lead to soluble aggregations. To maintain WSA monodispersity in the NMR sample, 2% of the lyso-lipid LPPG was used. Previous experimental and computational studies have suggested that the nicotinic acetylcholine receptor requires the proper compositions of lipids in order to function. The location of lyso-lipid molecules in WSA can thus provide structural information about protein-lipid interaction. In the ¹⁵N-edited NOESY spectra, we found several inter-molecular cross peaks between WSA and LPPG. Absence of cross-peaks does not indicate absence of interactions, but presence of intermolecular NOESY peaks signifies strong and stable interactions. We found that several residues at the TM1 and TM4 interface have cross peaks with LPPG. Specifically, residues 48, 50, 64, and 67 of TM1 and residues 24, 33, 34, 35, 39, 40, 41, and 43 of TM4 border an interfacial cavity that can accept a lipid molecule. The same cavity can be found in the nAChR cryo-EM structure and in ELIC and GLIC X-ray structures. As shown in Figure S3 in the supplement, the superposition of WSA and GLIC reveals that the strong WSA-LPPG interaction site coincides nearly perfectly with the location where surface lipid molecules were found in the GLIC crystal. This finding indicates that despite water solubilization, the helix interfaces are preserved to maintain possible interactions as in the wild-type protein.

In summary, the monomer structure of a water-soluble analog of the Torpedo nAChR α1 subunit was resolved to the backbone RMSD of 0.69 ± 0.13 Å. The overall structure is similar to the crystal structure of GLIC, demonstrating the robustness of the water solubilization approach. The WSA structure preserved the anticipated intra-helical drug binding site and lipid interaction site in the intact form of homologous receptors.

Highlights.

>We designed a water-soluble analogue to the acetylcholine receptor α1 subunit. >We determined the high-resolution structure of the water-soluble analogue by NMR. >The analogue preserved structural features observed in homologuous TM domains. >Photoaffinity labeling identified a relevant anesthetic binding site. >We affirm the water-solubilization method for high-resolution studies of TM proteins.