Abstract
Neuronal α4β2 nicotinic acetylcholine receptors are attractive drug targets for psychiatric and neurodegenerative disorders and smoking cessation aids. Recently, a third agonist binding site between two α4 subunits in the (α4)3(β2)2 receptor subpopulation was discovered. In particular, three residues, H142, Q150, and T152, were demonstrated to be involved in the distinct pharmacology of the α4-α4 versus α4-β2 binding sites. To obtain insight into the three-dimensional structure of the α4-α4 binding site, a surrogate protein reproducing α4-α4 binding characteristics was constructed by introduction of three point mutations, R104H, L112Q, and M114T, into the binding pocket of Lymnaea stagnalis acetylcholine-binding protein (Ls-AChBP). Cocrystallization with two agonists possessing distinct pharmacologic profiles, NS3920 [1-(6-bromopyridin-3-yl)-1,4-diazepane] and NS3573 [1-(5-ethoxypyridin-3-yl)-1,4-diazepane], highlights the roles of the three residues in determining binding affinities and functional properties of ligands at the α4-α4 interface. Confirmed by mutational studies, our structures suggest a unique ligand-specific role of residue H142 on the α4 subunit. In the cocrystal structure of the mutated Ls-AChBP with the high-efficacy ligand NS3920, the corresponding histidine forms an intersubunit bridge that reinforces the ligand-mediated interactions between subunits. The structures further reveal that the binding site residues gain different and ligand-dependent interactions that could not be predicted based on wild-type Ls-AChBP structures in complex with the same agonists. The results show that an unprecedented correlation between binding in engineered AChBPs and functional receptors can be obtained and provide new opportunities for structure-based design of drugs targeting specific nicotinic acetylcholine receptor interfaces.
Introduction
The α4β2 subtype of the nicotinic acetylcholine receptor (nAChR) is the most abundant nAChR in the brain, and it is implicated in psychiatric and neurodegenerative disorders as well as in nicotine addiction, which makes it a highly interesting therapeutic target (Jensen et al., 2005; Taly et al., 2009). The receptor is known to assemble in two alternate stoichiometries, (α4)3(β2)2 and (α4)2(β2)3 (Fig. 1A). Interestingly, this has been shown to influence agonist sensitivities (Nelson et al., 2003; Moroni et al., 2006; Zwart et al., 2008; Carbone et al., 2009) and calcium ion permeability (Tapia et al., 2007) and cause distinctly different pharmacology (Nelson et al., 2003; Kuryatov et al., 2005; Moroni et al., 2006, 2008; Tapia et al., 2007). The two receptor stoichiometries were recently found to differ in the number of agonist binding sites. In addition to two identical sites in the interface between α4 and β2 subunits (α-β interfaces) (Fig. 1A) common to both isoforms, the receptor with 3α:2β stoichiometry contains a third unique agonist binding site in the interface between two α4 subunits (α-α) (Harpsøe et al., 2011). The presence of both α-β and α-α binding sites in the 3α:2β receptor results in biphasic concentration-response relationships (CRRs) for acetylcholine (ACh) and many other agonists. This is simply because they have different functional potencies at the two types of binding sites, and occupation of all three sites is necessary for full receptor activation.
Agonist-binding at α4β2 nAChRs. (A) Schematic representations showing subunit arrangement and interfaces of the (α4)2(β2)3 (left panel) and (α4)3(β2)2 (right panel) ligand-binding domains. The high-affinity (α-β) orthosteric agonist-binding sites are shown by black arrows, and the low-affinity orthosteric-binding site at the α-α interface is shown by a gray arrow. The (+) and (−) signs represent the principal and complementary sides of the interface, respectively. (B) Chemical structure of NS3920 in the right panel, and NS3573 in the left panel. (C) Concentration-response curves for NS3920 and NS3573 on (α4)2(β2)3 and (α4)3(β2)2 receptors expressed in Xenopus laevis oocytes. Thin hashed lines represent fitted curve extensions. Data are from Harpsøe et al., 2011 (NS3573) and Ahring et al., 2015 (NS3920).
Recently, we have shown that three residues, H142, Q150, and T152 on the complementary side of the α4 subunit (α4(−)) and V136, F144, and L146 on the corresponding side of β2 (β2(−)), comprise the difference between the core of α-α and α-β binding sites, respectively, and are accountable for differences in agonist sensitivities (Harpsøe et al., 2011) and agonist binding affinities between the two sites (Ahring et al., 2015). Further, we have shown that a modulator that selectively targets the α-α binding site may substitute for one of the agonist molecules to aid activation via an agonist-like mechanism (Olsen et al., 2014a). All in all, these unique features of the α-α site make it particularly desirable to study to aid in design of new interface-specific drug molecules.
To understand the structural details of the α-α binding site, high-resolution structures of receptors presenting detailed information on the specific ligand-receptor interactions are required. However, such structures are not available, and despite emerging insight from X-ray structures of full-length ligand gated ion channels (Hilf and Dutzler, 2008, 2009; Bocquet et al., 2009; Hibbs and Gouaux, 2011; Althoff et al., 2014; Hassaine et al., 2014; Miller and Aricescu, 2014; Sauguet et al., 2014), soluble Ach-binding proteins (AChBPs) remain the proteins of choice when it comes to high-resolution structural studies of nAChR ligand-receptor interactions (Brejc et al., 2001; Celie et al., 2004; Hansen et al., 2005; Rohde et al., 2012; Shahsavar et al., 2012). The Lymnaea stagnalis acetylcholine binding protein (Ls-AChBP) has been particularly useful as a model for the α4β2 nAChR and has provided insight into binding mode of numerous α4β2 ligands (Celie et al., 2005; Rohde et al., 2012; Shahsavar et al., 2012; Ussing et al., 2013; Olsen et al., 2014a,b). Ls-AChBP is superior to AChBPs from other species because it contains a set of five conserved aromatic residues that are present in all nAChR agonist-binding sites and that are known to play a crucial role in accommodating the positive charge of common agonists. However, despite this intact “aromatic box,” the Ls-AChBP agonist binding site still differs from that in α-β and α-α sites by three first shell residues, R104, L112, and M114, on the complementary subunit interface, which notably correspond to the three residues that also differ between α-β and α-α sites and determine their functional fingerprints.
In the present study, we have investigated structural determinants of the α-α binding site to learn what governs agonist binding and functional responses. We introduced the three α-α site “signature” residues in Ls-AChBP (Ls-AChBPHQT) and conducted radioligand-binding experiments. The derived ligand affinities were then correlated to α-α affinities obtained from a functional α4β2 receptor in which α-β sites were engineered to resemble α-α sites by mutation of the three signature residues in the β2 subunit (α4β2HQT) to ease detection of α-α binding (Ahring et al., 2015). Next, we determined crystal structures of complexes between Ls-AChBPHQT and the α4β2 agonists NS3920 [1-(6-bromopyridin-3-yl)-1,4-diazepane] and NS3573 [1-(5-ethoxypyridin-3-yl)-1,4-diazepane] (Fig. 1B), two compounds that have distinctly different pharmacologic profiles on (α4)3(β2)2 and (α4)2(β2)3 receptors. Based on the structures, and supported by mutational analysis, functional differences between the ligands could be attributed to their different interactions with residues in the α-α versus α-β binding sites (Fig. 1C). The results point toward a unique ligand-specific role of the α4 residue H142 and confirm that AChBPs engineered to resemble individual binding sites have a future role in interpretation of structure activity relationships and as structural tools in drug discovery.
Materials and Methods
NS3531 [1-(pyridin-3-yl)-1,4-diazepane], NS3573, NS3920, and (+/−)-epibatidine were prepared at NeuroSearch A/S (Ballerup, Denmark) (Peters et al., 1999; Audouze et al., 2006). ACh [2-acetyloxyethyl(trimethyl)azanium chloride], (−)-cytisine [(1S,9S)-3,11-diazatricyclo[7.3.1.03,8]trideca-5,7-dien-4-one], (−)-nicotine [(−)-1-methyl-2-(3-pyridyl)pyrrolidine (+)-bitartrate], and salts or other chemicals not specifically mentioned below were purchased from Sigma-Aldrich (Broenby, Denmark) and were of analytic grade. [3H]Epibatidine [250 µCi, (1R,2R,4S)- (+)-6-(6-chloro-3-pyridyl)-7-azabicyclo[2.2.1] heptane] was obtained from PerkinElmer (Skovlunde, Denmark). Restriction enzymes, Sf9 cells, and pFastBac vector were purchased from Invitrogen (Naerum, Denmark). Oligonucleotides as well as sequencing services were from Eurofins MWG Operon (Ebersberg, Germany).
Mutagenesis, Protein Expression, and Purification.
The wild-type (WT) Ls-AChBP cDNA was mutated using the overlap extension polymerase chain reaction method (Ho et al., 1989). Briefly, fragments were first primed on the WT sequence to obtain the R104H mutation construct. Then the primers containing both L112Q and M114T mutations were used on the R104H construct to obtain the R104H/L112Q/M114T construct. Polymerase chain reaction fragments were digested using EcoRI/BamHI and then purified using agarose gel electrophoresis. Digested inserts were ligated into a cut pFastBac vector. After transformation, each mutant plasmid was purified using a Qiagen miniprep kit (Qiagen GmbH, Hilden, Germany). The mutations were then verified by sequencing. WT and engineered recombinant Ls-AChBPs were expressed using the Bac-to-Bac baculovirus expression system in Sf9 insect cells and purified using ion exchange and gel filtration chromatography as described previously elsewhere (Brejc et al., 2001).
Cloning of the human α4 and β2 nAChRs and construction of concatenated receptors were described previously elsewhere (Timmermann et al., 2007). Point mutations to give α4H142V, α4Q150F, and α4T152L were introduced by site-directed mutagenesis, as previously described elsewhere (Harpsøe et al., 2011). The cRNA was prepared from linearized plasmid using the mMessage mMachine T7 Transcription kit (Invitrogen) according to the manufacturer’s protocol.
Radioligand Binding.
Binding experiments with Ls-AChBP and Ls-AChBPHQT were performed using the purified protein from Sf9 cells under conditions similar to previous descriptions (Harpsøe et al., 2011, 2013). Binding affinity of [3H]epibatidine (Kd) was determined in standard saturation assays where the amount of radioactivity bound after equilibration with protein was measured as a function of the concentration of [3H]epibatidine. Nine volumes of protein suspended in 50 mM Tris-HCl buffer (pH 7.4) were mixed with one volume of [3H]epibatidine (55.8 Ci/mmol, in 48% ethanol) in varying concentrations and incubated for 2 up to 4 hours at room temperature. Nonspecific binding was tested by incubation with an excess of (−)-nicotine (30 µM for WT Ls-AChBP and 500 µM for Ls-AChBPHQT). Binding was terminated by the addition of ice-cold Tris-HCl buffer, and the solution was poured directly onto Whatman GF/C fiberglass filters (Whatman/GE Healthcare, Little Chalfont, UK) presoaked for 30 minutes in 0.1% polyethyleneimine and washed twice with buffer.
Compound inhibition constants (IC50) were determined in [3H]epibatidine displacement experiments. Eight volumes protein suspension were mixed with one volume of compound solutions in 48% ethanol and one volume of a [3H]epibatidine solution in 48% ethanol. The final radioligand concentration was ∼1 nM. The specific concentration of the radioligand solution was determined by liquid scintillation counting for each experiment. Nonspecific binding was tested by incubation with (−)-nicotine as described earlier. After 1 hour of incubation at room temperature, the binding was terminated by filtration over GF/C fiberglass filters as described earlier.
Filters containing protein and bound radioligand were individually incubated for at least 4 hours with 3 ml of Ultima Gold (PerkinElmer). Radioactivity was then measured by conventional liquid scintillation counting on a Tri-Carb counter (PerkinElmer). For each experimental set, three measurements were performed to determine total and nonspecific binding. All measurements for controls and compound tests were performed as triplicate samples.
Dissociation constant (Kd) values of [3H]epibatidine for each saturation binding experiment was determined by fitting the data by nonlinear regression to the equation Y = (Sbmax * x)/(Kd + x), where Y equals the specific binding counts (Measured binding − Measured nonspecific binding), Sbmax is the maximal specific binding, and x is the concentration of radioligand. The average Kd values were calculated from pKd values.
For radioligand displacement experiments, compound IC50 values were determined by fitting the percentage of inhibition by nonlinear regression to the equation Y = 100/(1 + 10X − log(IC50)), where Y is the percentage inhibition and X is the logarithm of the compound concentration. The inhibition constant (Ki) values were calculated from IC50 values by the Cheng-Prusoff equation Ki = IC50/(1 + L/Kd), where L is the concentration of radioligand used in the assay and Kd is the affinity of the radioligand for the binding site under investigation. Average Ki values were calculated from pKi values. All fitting routines were performed using GraphPad Prism 5 (GraphPad Software, San Diego, CA).
Two-Electrode Voltage Clamp Electrophysiology.
Xenopus laevis oocytes were prepared as previously described elsewhere (Mirza et al., 2008). Briefly, to obtain isolated oocytes, lobes from ovaries of female adult Xenopus laevis were removed and defolliculated using collagenase. Oocytes were injected with a total of ∼25 ng of a mixture of cRNAs, encoding the nAChR subunits α4, α4H142V, α4Q150F, α4T152L, β2, or concatenated subunits thereof, and incubated for 2 to 7 days at 18°C in modified Barth’s solution (90 mM NaCl, 1.0 mM KCl, 0.66 mM NaNO3, 2.4 mM NaHCO3, 10 mM HEPES, 2.5 mM sodium pyruvate, 0.74 mM CaCl2, 0.82 mM MgCl2, 100 µg/ml gentamycin and pH adjusted to 7.5).
Electrophysiologic recordings were performed with oocytes placed in a custom-built recording chamber as described previously elsewhere (Mirza et al., 2008). Pipettes were backfilled with 2 M KCl, and open pipette resistances ranged 0.6 to 2.0 MΩ when submerged in OR2 (90 mM NaCl, 2.5 mM KCl, 2.5 mM CaCl2, 1.0 mM MgCl2, 5.0 mM HEPES, and pH adjusted to 7.5). Cells were voltage clamped at a holding potential ranging from −40 to −80 mV using an Axon Geneclamp 500B amplifier (Molecular Devices UK, Wokingham, UK). Compounds were dissolved in OR2 on the day of the experiment and solutions were applied to the oocytes with a flow rate of 2.0 ml/min via a glass capillary. Each application lasted ∼1 minute, and the application system ensured rapid solution exchange (in the order of few seconds). Amplified signals were low-pass filtered at 20 Hz and digitized at 200 Hz by an Axon Digidata 1322A (Molecular Devices UK). Traces were recorded using Clampex 9.2 and analyzed in Clampfit (Molecular Devices UK).
During analysis, traces were baseline subtracted and the responses to individual applications read as peak current amplitudes. Peak current amplitudes were next normalized to a 1 mM ACh-evoked response in the same oocyte and fitted by nonlinear regression to monophasic or biphasic Hill equations using GraphPad Prism. In the fitting routines, Hill slopes were constrained to 1, the efficacy at infinitely low compound concentrations set to 0, and the best fit, mono- versus biphasic, was determined by an F test.
Crystallization.
A solution of Ls-AChBPHQT in 20 mM Tris-HCl buffer (pH 8.0) and 20 mM NaCl was incubated with 10–25 mM of the ligand (in 20 mM Tris-HCl buffer, pH 8.0) before crystallization. Crystals of both complexes were obtained using the hanging drop vapor diffusion method at 20°C. Crystallization drops were made by mixing 1 µl of a 5 mg/ml protein:ligand solution with 1 µl of crystallization solution containing 0.1 M Tris-HCl buffer (pH 8.0), 1–3% v/v polyethylene glycol 400, and 1.8–2.3 M (NH4)2SO4. Crystals grew from 3 days to 2 weeks to a final length of 0.1–0.2 mm.
Crystallographic Data Collection, Refinement, Model Building, and Structural Analysis.
The mother liquor supplemented with 25% (v/v) glycerol was used as a cryoprotectant. The crystals were mounted in cryoloops and flash-cooled in liquid nitrogen after brief immersion in the cryoprotectant solution. X-ray data on the NS3920-Ls-AChBPHQT complex were collected on beamline I911-3 at the MAX Laboratory synchrotron (Lund, Sweden), and the data on the NS3573 complex were collected on ID23-2 at the European Synchrotron Radiation Facility synchrotron (Grenoble, France), using a MarMosaic 225 detector (Marresearch GmbH, Norderstedt, Germany) at a wavelength of 1 Å. Data were processed and scaled using XDS (Kabsch, 2010) and Scala (Leslie, 1992), respectively. We set aside 5% of the data during the scaling process as a test set for calculation of Rfree.
The structures were solved by the molecular replacement method using the program Phaser (McCoy et al., 2007). A pentamer of the Ls-AChBP WT complexes with corresponding ligands (PDB IDs 3u8m and 3u8k) was used as search models for the NS3920 and NS3573 complexes, respectively. The refinements were performed with Phenix (Afonine et al., 2012) and rebuilt interactively using Coot (Emsley and Cowtan, 2004; Emsley et al., 2010). Noncrystallographic symmetry was used during refinement of the NS3920-AChBPHQT complex with Phenix as there were eight pentamers in the asymmetric unit of the crystal. The translation-libration-screw refinement was also used in the last refinement cycles of NS3920-AChBPHQT structure. The B factors were refined as individual isotropic values. Data collection and refinement statistics are summarized in Table 1.
Data collection and refinement statistics
The quality of the final models was assessed by examination of the detailed stereochemistry using Molprobity (Chen et al., 2010) (Table 1). The structure and ligand analyses were performed using Coot and PyMOL (Schrödinger, New York, NY). The figures were generated using PyMOL.
Results
We first mutated the Ls-AChBP ligand binding site to make it resemble an α4β2 nAChR α-α interface (Ls-AChBPHQT) by introduction of three point mutations, R104H, L112Q, and M114T.
Radioligand Binding.
To evaluate the Ls-AChBPHQT construct as a model system for the α-α interface of α4β2 nAChRs assembled in the 3α:2β stoichiometry and to compare it to the WT Ls-AChBP, we initially investigated [3H]epibatidine radioligand binding to the two proteins. For both, [3H]epibatidine displayed saturable specific binding with linear Scatchard plots consistent with a single type of binding sites (higher order fits were not statistically better), albeit the concentration ranges of radioligand used were very different (Fig. 2, A and B). The average of determined epibatidine Kd values for the WT Ls-AChBP was 0.085 nM (pKd ± S.D. = 10.1 ± 0.31; n = 3), which is very close to the previously published value of 0.097 nM at a Ls-AChBP/5HT3A chimera construct (Rohde et al., 2012).
[3H]Epibatidine saturation binding at WT and engineered Ls-AChBPHQT. Protein purified from Sf9 cells was used in standard filtration binding assays. (A and B) For both Ls-AChBP and Ls-AChBPHQT, normalized specific binding plotted as a function of the [3H]epibatidine concentration was well approximated by a single site hyperbolic model using nonlinear regression. This indicates binding to a single saturable binding site in each case. Plotted data points are from n = 3 (Ls-AChBP) or n = 4 (Ls-AChBPHQT) independent experiments, with each data point representing the mean of a triplicate determination. Note that two of the four curves are overlapping for AChBPHQT. Individual Kd values derived from fitting are indicated in the graphs, and the average Kd values (calculated from pKd values) were 0.085 nM (pKd ± S.D. = 10.07 ± 0.31; n = 3) for Ls-AChBP and 16 nM (pKd ± S.D. = 7.80 ± 0.30; n = 4) for Ls-AChBPHQT. At the highest tested [3H]epibatidine concentrations, nonspecific binding (mean ± S.D.) equated 12 ± 9% and 25 ± 12% of total binding for Ls-AChBP and Ls-AChBPHQT, respectively. (C) Normalized specific binding plotted as a function of the logarithm of [3H]epibatidine concentrations were well approximated by the monophasic Hill equation using nonlinear regression. In the presented fits bottom was set to 0 and the Hill slope to 1. Data for WT α4β2 and α4β2HQT are from Ahring et al., 2015.
For Ls-AChBPHQT, the average Kd value was 16 nM (Table 2), which is approximately 200-fold lower compared with the WT Ls-AChBP but very similar to the Kd value of 5.3 nM reported for the functional α4β2HQT mutant receptor (Ahring et al., 2015). When normalized, binding data from all experiments were well approximated by the Hill equation, giving EC50 values similar to the average of Kd values from the individual experiments (Fig. 2C). Figure 2 shows that adequate concentration ranges of [3H]epibatidine have been applied on both WT and engineered Ls-AChBPHQT and that there is good agreement between the EC50 values derived for WT and engineered Ls-AChBPHQT and α4β2 and α4β2HQT, respectively. As is also evident from the figure, the curve for the Ls-AChBPHQT is right-shifted by more than two orders of magnitude relative to WT Ls-AChBP, while it quite accurately approximates the affinity measured in the α4β2HQT mutant receptor (Table 2).
Binding affinities of nAChR ligands at Ls-AChBPHQT and α4β2HQT
Ls-AChBPHQT protein was purified from Sf9 cells and used in standard filtration binding assays. The pKd value of [3H]epibatidine was obtained as described in the legend to Fig. 2. Binding affinities represented as pKi of the remaining compounds were determined as the potency for displacing a fixed concentration of [3H]epibatidine. pKi values are presented as mean ± S.D. for n = 3–4 determinations each conducted in triplicates. Kd (epibatidine) and Ki values are calculated from pKd and pKi values and presented as mean in nanomolar. All data for α4β2HQT are from Ahring et al., 2015.
Subsequently, a series of agonists were profiled by radioligand displacement studies at Ls-AChBPHQT and compared with data obtained on the α4β2HQT mutant receptor heterologously expressed in human embryonic kidney 293 (HEK293) cells (Ahring et al., 2015). The determined binding affinities for the agonists strictly follow the order obtained on the α4β2HQT receptor as clearly seen from a correlation plot (Fig. 3; Table 2). This underscores that the Ls-AChBPHQT construct is a viable model system that reproduces binding affinities obtained at its functional counterpart.
Correlation between affinity of nAChR agonists at the engineered Ls-AChBPHQT and α4β2HQT nAChR. Linear correlation of pKi values for a series of agonists at Ls-AChBPHQT versus α4β2HQT. Data are presented in Table 2. Dotted lines indicate 95% confidence intervals.
Crystal Structure Analysis.
Crystal structure analysis was undertaken on complexes between Ls-AChBPHQT and the compounds NS3920 and NS3573 (Fig. 1B), which were selected as representatives of ligands that show high and low affinity at the α4β2HQT nAChR, respectively (Table 2) and have distinctly different pharmacologic profiles as illustrated in Fig. 1C. Both compounds are prototypical nAChR agonists consisting of a pyridine ring connected to a homopiperazine ring that is protonated at physiologic conditions. The pyridine ring of NS3920 is substituted with bromine at the 6-position whereas NS3573 has an ethoxy substituent in the pyridine 5-position (Fig. 1B).
The crystal structures of Ls-AChBPHQT with NS3920 and NS3573 were determined at 2.7 and 2.8 Å resolution and refined to a final Rwork/Rfree of 0.19/0.24 and 0.19/0.23, respectively (for information on refinement statistics, see Table 1). The structures show the same homopentameric assembly as previously determined AChBP structures, in which individual monomers consist of an N terminal α-helix, two short α310 helices and a 10-stranded β-sandwich core (Fig. 4). The electron density maps clearly reveal the existence of a single binding orientation for both of the ligands (Fig. 5), whereas the loop F part of the structures (residues 154–160) could only be completely modeled in chain D of the NS3573 complex (Fig. 4), due to lack of clear electron density, signifying a greater flexibility of these parts of the protein.
Overall structure of Ls-AChBPHQT. The diagram shows homopentameric Ls-AChBPHQT on left and the ligand-binding pocket at the interface of two monomers formed by loops A, B, and C from the principal or (+) side (colored in red) and loops D, E, and F from the complementary (−) side (colored in green) on right.
Cocrystal structures of Ls-AChBPHQT with NS3920 and NS3573. The magnified view shows the ligand-binding site at the interface of two monomers in (A) NS3920 and (B) NS3573 complexes with Ls-AChBPHQT. The (+) and (−) signs represent the principal and complementary sides of the interface, respectively. The principal side is colored in deep teal and the complementary side in pink in the NS3920-Ls-AChBPHQT complex. For NS3573-Ls-AChBPHQT, the principal and complementary sides are colored in red and green, respectively. The interactions between the ligand and its surroundings are shown as dashed lines. The final 2mFo-DFc electron density maps at 1σ level carved around the ligands are also presented. (C) Illustration represents the loop C in the NS3920 (deep teal) and NS3573 (red) complexes with Ls-AChBPHQT. (D) The complementary monomer of Ls-AChBPHQT crystal structure in complex with NS3920 (pink) is superimposed on the complementary monomer of the Ls-AChBPHQT in complex with NS3573 (green). The principal monomers of NS3920 and NS3573 complexes with Ls-AChBPHQT are shown in deep teal and red, respectively. The hydrogen-bonding interactions are shown as dashed lines.
Comparing Ls-AChBPHQT complexes with NS3920 and NS3573 to the corresponding WT Ls-AChBP complexes (Rohde et al., 2012) shows that neither a significant Cα backbone change (root-mean-square deviation: 0.258 ± 0.007 Å between the NS3920 complexes and 0.246 ± 0.001 Å between NS3573 complexes) nor ligand repositioning to a different pose occurs (Fig. 5, A and B). Also, the size and shape of the binding site are similar between the WT Ls-AChBP and Ls-AChBPHQT complexes.
Ligand Interactions with Non-Mutated Residues in Ls-AChBPHQT and Ls-AChBP.
In the NS3920 structure with Ls-AChBPHQT, the core of the ligand (defined as the two conjugated ring systems without substituents) interacts essentially as observed in prototypical agonist-bound structures of AChBP (Celie et al., 2004). The protonated secondary nitrogen of the ligand lies within hydrogen-bonding distance of the backbone carbonyl oxygen of W143, and is also in close contact with the hydroxyl group of Y89 on the principal side of the interface (Fig. 5A). On the complementary side, NS3920 interacts with the protein main chain via a water-mediated hydrogen bond as previously reported in agonist-bound AChBP structures (Celie et al., 2004; Hansen et al., 2005; Rohde et al., 2012) (Fig. 5A). The anchoring of the ligand to the complementary subunit is further supported by a halogen bond interaction caused by the positive electrostatic potential on the tip of the bromine atom of the ligand and electrons in the π-system of the carbonyl oxygen of Q112 (Fig. 5A, Br-O distance of 3.1 ± 0.1 Å), similar to the interaction seen in the complex with WT Ls-AChBP (Rohde et al., 2012).
The Ls-AChBPHQT structure in complex with NS3573 reveals similar interactions between ligand and residues that are unaltered between Ls-AChBPHQT and Ls-AChBP, both on the principal and the complementary sides of the interface, as observed with the NS3920 (Fig. 5B).
Ligand-AChBPHQT Specific Interactions.
Stacking interaction is seen between H104 and the aromatic ring of NS3920 (3.5 ± 0.2 Å averaged over the shortest distance between H104 and the ligand). The ligand stabilizes H104 in a conformation where it forms a hydrogen bond across the interface to Y192 (loop C) on the principal side (Fig. 5A). In 4 out of 40 monomers of the asymmetric unit of the crystal (chains A, E, Q, and b), a water molecule is observed in optimal hydrogen-bonding distance to H104 (2.7 ± 0.1 Å), and two alternate conformations of the histidine (ring flip) have therefore been introduced at these sites. The side-chain amide group of Q112 is in direct contact with the positive edge of the aromatic ring of NS3920 (3.6 ± 0.2 Å), and T114 is within van der Waals interaction distance (3.5 ± 0.2 Å) of the aromatic ring of the ligand (Fig. 6A).
The ligand interactions with residues H/R104, Q/L112, and T/M114 in engineered/WT Ls-AChBP. (A) NS3920-Ls-AChBPHQT, (B) NS3573-Ls-AChBPHQT, (C) NS3920-WT Ls-AChBP, and (D) NS3573-WT Ls-AChBP. The (+) and (−) signs represent the principal and complementary sides of the interface, respectively. The atoms involved in van der Waals interaction are labeled. The sphere in the middle of the aromatic ring of NS3920 in panel A represents the centroid of this group. The average distances between the atoms of the protein and the ligands involved in van der Waals interactions are as follows: (A) imidazole ring of His104 and center of the aromatic ring of NS3920: 3.5 ± 0.2 Å; Oε1 of Q112 and C4 of the aromatic ring of NS3920: 3.5 ± 0.1 Å; Cβ of T114 and N1 of the aromatic ring of NS3920: 3.6 ± 0.2 Å; (B) Cβ and Cγ of His104 and O1 of the ethoxy substituent of NS3573: 3.5 ± 0.1 Å; Oε1 of Q112 and O1 of the ethoxy substituent of NS3573: 3.5 ± 0.2 Å; Cβ of T114 and N3 and C10 of the aromatic ring of NS3573: 3.6 ± 0.1 Å; (C) Cβ of R104 and Br of NS3920: 3.4 ± 0.1 Å; Cδ2 of Leu112 and C4 and C5 of NS3920: 3.4 ± 0.1 Å; Cε of M114 and C1, C2, and N2 of NS3920: 3.4 ± 0.1 Å; (D) Cβ of R104 and O1 of the ethoxy substituent of NS3573: 3.5 ± 0.1 Å; Cδ2 of Leu112 and C7, C12, and O1 of the ethoxy substituent of NS3573: 3.7 ± 0.1 Å; Cε of M114 and C4, N2, C9, and C10 of NS3573: 3.7 ± 0.1 Å.
In the NS3573 complex, the ethoxy substituent causes a different orientation of H104 and prevents it from interacting across the interface with Y192, causing the loop C to be more open compared with in the NS3920-Ls-AChBPHQT complex (7.8 ± 0.05 versus 7.1 ± 0.03 Å measured from the carbonyl oxygen of W143 to the sulfur atom of C187) (Fig. 5C). This conformation of H104 subsequently leads to a conformational change in its immediate neighbor, Q73. The residue Q73 forms hydrogen bonds across the interface to H146 and E149 on the principal side (Fig. 5D). However, as Q73 corresponds to a serine in α4 (S110), we have not found it relevant to pursue further. With regard to conserved residues, the side chain of H104 is within van der Waals distance of the ethoxy substituent of the ligand (3.5 ± 0.1 Å). Likewise, the residues Q112 and T114 are within van der Waals interaction distance of the ethoxy substituent and aromatic ring of the ligand, respectively (Fig. 6B).
Mutational Analysis.
To investigate the functional importance of the three α-α interface “signature” residues in WT α4β2 nAChRs, three point mutated α4 constructs were created. In these, the residues H142, Q150, and T152 on the complementary face of the α(−) subunit were individually mutated to the corresponding β(−) residues, V136, F144, and L146. Hence, each “signature” residue point mutant alters the α-α interface to become more α-β interface-like in one specific position. ACh, NS3920, and NS3573 were then evaluated by two-electrode voltage-clamp electrophysiology on the variant receptors expressed in Xenopus laevis oocytes in the 3α:2β stoichiometry. Initial tests with ACh confirmed that all mutants were functional.
Next, CRRs for NS3920 and NS3573 were obtained with each construct and compared with data for the WT 3α:2β receptor. For NS3920, the α4H142V mutation resulted in a dramatic right shift of the concentration response curve, rendering the compound approximately 100-fold less potent, whereas the two other mutations had little or no effect on potency (Fig. 7, A and B; Table 3). In contrast, for NS3573, the α4H142V mutation had negligible effect whereas both of the other two mutations, α4Q150F and α4T152L, rendered the compound approximately 10-fold more potent. Interestingly, efficacy levels of the compounds were also affected but by different residues. For NS3920 a superagonist response was observed with the α4Q150F mutant whereas for NS3573 the efficacy was more than doubled with both α4Q150F and α4T152L relative to WT α4β2 nAChR (Fig. 7B; Table 3).
Agonist-evoked responses at mutant α4β2 nAChRs using Xenopus laevis oocyte two-electrode voltage-clamp electrophysiology. To yield uniform populations of receptors in the 3α:2β stoichiometry, oocytes were injected with WT β2 subunits mixed with point mutated α4 subunits in a 4:1 cRNA ratio (red, purple, and green curves). Full CRRs were obtained on each oocyte. Peak current amplitudes were background subtracted and normalized to the amplitude evoked by 1 mM ACh (ACh_control) on the same oocyte. The calculated percentage activation was depicted as mean ± S.E. as a function of the concentration and fitted to a mono- or biphasic Hill equation using nonlinear regression with a fixed bottom of 0 and Hill slope(s) of 1 (A and B) NS3920 and NS3573 from n = 6–19 individual oocytes. Note the different scaling used for the ordinate axis. Thin hashed lines represent fitted curve extensions. Regression results are presented in Table 3. Data for (α4)2(β2)3 and (α4)3(β2)2 with NS3920 are from Ahring et al., 2015 and with NS3573 from Harpsøe et al., 2011.
Agonist-evoked responses at WT and mutant α4β2 nAChRs using Xenopus laevis oocyte two-electrode voltage-clamp electrophysiology
Data were obtained as described in the legend for Fig. 7. Background subtracted and normalized peak current amplitudes were fitted to a mono- or biphasic Hill equation using nonlinear regression with a fixed bottom of 0 and Hill slope(s) of 1 and are presented as EC50 in µM (pEC50 ± S.E.) and maximal efficacy Emax ± S.E. in percentage of ACh_control (1 mM ACh). Data for (α4)2(β2)3 and (α4)3(β2)2 are from Ahring et al., 2015.
Discussion
The binding and functional differences observed at 3α:2β and 2α:3β nAChRs are assigned to existence of a third ACh binding site at the α-α interface of α4β2 receptors with 3α:2β stoichiometry and more specifically to the three “signature” residues at the complementary side of this interface. To resolve the structural determinants leading to differential agonist characteristics at α-α and α-β binding sites, Ls-AChBP was engineered to resemble the α-α interface of 3α:2β nAChRs. The construct, Ls-AChBPHQT, reproduces binding affinities of the α4β2HQT receptor with high accuracy for a series of agonists with an affinity range spanning from nano- to micromolar (Fig. 3). The high degree of ligand affinity correlation between Ls-AChBPHQT and α4β2HQT receptors suggest that detailed structural information regarding ligand binding at the α-α interface can be obtained from cocrystal structures with Ls-AChBPHQT. Therefore, we used Ls-AChBPHQT to investigate interactions for two agonists, NS3920 and NS3573. From cocrystal structures and functional studies we show that residues specific to the α-α interface play differential roles in determining binding affinities of the ligands, of which the ligand-specific role of α4H142 is particularly pronounced.
NS3920 and NS3573 are both homopiperazine substituted pyridines and have comparable binding affinity and functional potency at 2α:3β receptors (Rohde et al., 2012). However, the pharmacologic properties of NS3920 and NS3573 differ significantly at 3α:2β receptors with respect to both binding affinities and functional electrophysiologic profiles. NS3920 has approximately 70-fold higher affinity compared with NS3573 at the α4β2HQT receptor (Table 2) and is virtually a full agonist at the WT 3α:2β receptor with a monophasic CRR, suggesting that it does not differentiate between the α-β and α-α sites to a notable extent (Fig. 1C). In contrast, NS3573 is a partial agonist with a biphasic CRR at 3α:2β receptors, displaying an almost 1000-fold potency difference between α-β and α-α sites (Fig. 1C) (Harpsøe et al., 2011; Ahring et al., 2015). NS3920 and NS3573 therefore represent two compounds that are particularly suitable for studying differences in interactions with α-α and α-β interfaces of 3α:2β receptors.
The most striking difference between the cocrystal structures of Ls-AChBPHQT with NS3920 and NS3573 is the different rotameric states of H104. In the Ls-AChBPHQT complex with NS3920, H104 stacks with the aromatic ring of the ligand (Fig. 6A), whereas in the complex with NS3573, H104 adopts an entirely different orientation, which is comparable to that of R104 in WT Ls-AChBP. In the observed binding mode for NS3920, H104 forms a hydrogen bond to Y192 on loop C. Effectively this is an intersubunit bridge that mediates contacts between principal and complementary subunits in this structure and reinforces the ligand-mediated contacts. At the same time, this interaction constrains loop C to be more closed relative to the structure in complex with NS3573 (Fig. 5C). In the NS3573 complex, H104 makes favorable van der Waals contacts to the ethoxy substituent and at the same time causes a displacement of Q112 to a position where it interacts with the ethoxy substituent of the ligand (3.5 ± 0.2 Å) (Fig. 6B). The consequences of the Q112 displacement in terms of binding affinity are unclear. The combination of no halogen bonding (Rohde et al., 2012), the lack of stacking interactions with H104 and no intersubunit bridge formation to Y192 in the NS3573 complex likely account for the decreased affinity of NS3573 relative to NS3920 at the Ls-AChBPHQT and α4β2HQT receptors.
The different positions of H104 in the two cocrystal structures is supported by the mutational data showing that mutation of α4H142 to a valine, which is the corresponding residue in the β2 subunit, has entirely different effects on the pharmacology of NS3920 and NS3573 at receptors expressed in the 3α:2β stoichiometry. For NS3920 the efficacy is decreased from 82 to 54% accompanied by a dramatic ∼50-fold loss of functional potency (Table 3). This suggests that for NS3920, H142 is important not only for binding but also contributes to gating of the receptor. For NS3573, where the histidine is pointing away from the binding site, mutation of H142 to valine showed only marginal effects on efficacy and potency and did not change the fundamental appearance of the CRR (Fig. 7B). We have previously suggested that intersubunit bridge formation rather than the closure of loop C is the important determinant of efficacy (Rohde et al., 2012; Harpsøe et al., 2013). The high efficacy of NS3920 along with the strong effect observed upon mutation of the histidine and the breakage of its hydrogen bond to Y192 further support this hypothesis.
Interestingly, mutation of the two remaining “signature” residues Q150 and T152 to their β2 counterparts, phenylalanine and leucine, respectively, has little effect on the functional potency of NS3920. For NS3573 on the other hand, the mutations each resulted in an ∼10-fold increase in potency toward that of the 2α:3β receptor (Fig. 7; Table 3). This signifies an important contribution of each of these residues in discriminating binding affinities between α-α and α-β binding sites. For NS3920, the dominant role of H142 along with the very small difference in binding affinity (approximately 10-fold) between α-α and α-β sites leave little room for improvement by mutation and it is therefore not surprising that no differences are observed by mutation of these residues.
The important observation from the mutational studies is that each residue plays a unique role in interaction with different ligands. For some residues, the effect appears mainly on binding affinity; for others, efficacies are also affected. For both compounds, the Q150F mutant increases efficacy to roughly twice that at the WT receptor, and NS3920 is essentially a superagonist. For NS3920 the T152L mutant lowers efficacy, whereas it increases efficacy of NS3573. The differential role of the signature residues is thus fully in line with the studies showing that the complementary face of the adjacent subunit plays a key role fine-tuning agonist function (Harpsøe et al., 2013; Marotta et al., 2014).
Interestingly, despite an overall conserved positioning of the ligands in the pocket, the interactions with the three signature residues is entirely different between Ls-AChBPHQT and its WT variant (compare Fig. 6, A and B to Fig. 6, B and D). The difference is reflected in a consistently higher affinity of the compounds in Table 2 at WT Ls-AChBP in agreement with an expected stronger interaction of lipophilic compounds with a hydrophobic binding site. In the Ls-AChBPHQT-binding site, the presence of H104, Q112, and Thr114 creates a more hydrophilic environment, which generally, unless specific interactions can be obtained, leads to lower affinity for central nervous system–type compounds.
In combination, WT Ls-AChBP and Ls-AChBPHQT structures may facilitate understanding of the stoichiometry selectivity of agonists such as sazetidine-A (Xiao et al., 2006), which is an agonist with a functional preference for 2α:3β receptors (Zwart et al., 2008). A recent study has suggested that sazetidine-A is excluded from binding at the α-α interface because of steric clashes with α4H142 (Mazzaferro et al., 2014). However, if this were the case, it could be expected that NS3573—which, like sazetidine-A, contains a large substituent in the pyridine 5-position—would be affected in a similar way. Our data indicate that the 5-substituent induced conformational change of His142 to an unfavorable rotameric state leads to a large drop in affinity and functional potency as well as efficacy of the ligand but not its exclusion from binding per se.
In the present study, we adopted a minimalistic approach and only mutated residues in immediate vicinity of the agonist-binding pocket in Ls-AChBP. A big advantage with this approach was that expression, purification, and crystallization conditions could be adapted from the conditions used with the WT protein without the need for time-consuming optimization of the process. The approach led to a construct that in terms of binding affinities correlates almost perfectly with binding affinities obtained in the functional α4β2HQT receptor for agonists (Fig. 3). The minimalistic mutational approach means that the binding pocket is engineered to recognize agonists. For antagonists that usually are larger structures and extend out of the agonist binding core, an equally good correlation cannot automatically be assumed. To accommodate such ligands a more comprehensive mutational approach could be adopted. This has been successfully done for the nAChR α7 in a chimeric construct with Ac-AChBP that was shown to accommodate both agonists and antagonists (Li et al., 2011; Nemecz and Taylor, 2011).
The interpretation of binding and function in a structural context requires a unique well defined response. For receptors with 3α:2β stoichiometry, the present and previous studies (Rohde et al., 2012; Ahring et al., 2015) have taught us that binding and function correlates for nAChRs if focusing on a response that can be mapped back to one individual type of binding sites. In these instances, binding and function can be interpreted in a structural context. This study shows that to investigate structure activity relationships superior correlation between an engineered Ls-AChBP and its functional counterpart can be obtained with a minimalistic mutation approach.
Targeting binding sites at non–α-β interfaces of nAChRs provides a unique opportunity to develop subtype or even stoichiometry selective therapeutics. To aid structure-based design of such ligands, engineered AChBPs may be used as structural models and can provide details about ligand receptor interactions that by far exceeds what can be obtained using WT AChBPs as models.
Acknowledgments
The authors thank the MAX Laboratory, Lund, Sweden, and the European Synchrotron Radiation Facility, Grenoble, France, for providing beamtime and support during data collection.
Authorship Contributions
Participated in research design: Shahsavar, Ahring, Olsen, Kastrup, Balle, Gajhede.
Conducted experiments: Shahsavar, Ahring, Olsen, Krintel.
Performed data analysis: Shahsavar, Ahring, Olsen, Kastrup, Balle, Gajhede.
Wrote or contributed to the writing of the manuscript: Shahsavar, Gajhede, Balle, Ahring, Kastrup.
Footnotes
- Received January 23, 2015.
- Accepted July 15, 2015.
This work was supported by the Drug Research Academy at the University of Copenhagen, Denmark; the Danish Research Agency for Science Technology and Innovation, DANSCATT [5072-00003B], NeuroSearch A/S, Denmark; the Australian National Health and Medical Research Council [APP1069417] and IMK Almene Fond, Denmark.
The atomic coordinates and structure factors (PDB IDs 4UM1 and 4UM3) have been deposited in the Protein Data Bank, Research Collaboratory for Structural Bioinformatics, Rutgers University, New Brunswick, NJ (http://www.rcsb.org/).
Abbreviations
- ACh
- acetylcholine
- AChBP
- acetylcholine-binding protein
- CRR
- concentration-response relationship
- Ls-AChBP
- Lymnaea stagnalis acetylcholine-binding protein
- nAChR
- nicotinic acetylcholine receptor
- NS3531
- 1-(pyridin-3-yl)-1,4-diazepane
- NS3573
- 1-(5-ethoxypyridin-3-yl)-1,4-diazepane
- NS3920
- 1-(6-bromopyridin-3-yl)-1,4-diazepane
- WT
- wild-type
- Copyright © 2015 by The American Society for Pharmacology and Experimental Therapeutics