The binding characteristics and orientation of a novel radioligand with distinct properties at 5-HT3A and 5-HT3AB receptors

VUF10166 (2-chloro-3-(4-methyl piperazin-1-yl)quinoxaline) is a ligand that binds with high affinity to 5-HT3 receptors. Here we synthesise [3H]VUF10166 and characterise its binding properties at 5-HT3A and 5-HT3AB receptors. At 5-HT3A receptors [3H]VUF10166 displayed saturable binding with a Kd of 0.18 nM. Kinetic measurements gave monophasic association (6.25 × 107 M−1 min−1) and dissociation (0.01 min−1) rates that yielded a similar Kd value (0.16 nM). At 5-HT3AB receptors two association (6.15 × 10−7, 7.23 M−1 min−1) and dissociation (0.024, 0.162 min−1) rates were seen, yielding Kd values (0.38 nM and 22 nM) that were consistent with values obtained in saturation (Kd = 0.74 nM) and competition (Ki = 37 nM) binding experiments respectively. At both receptor types, specific binding was inhibited by classical 5-HT3 receptor-selective orthosteric ligands (5-HT, allosetron, d-tubocurarine, granisetron, mCPBG, MDL72222, quipazine), but not by non-competitive antagonists (bilobalide, ginkgolide B, picrotoxin) or competitive ligands of other Cys-loop receptors (ACh, bicuculline, glycine, gabazine). To explore VUF10166 ligand–receptor interactions we used in silico modelling and docking, and tested the predictions using site directed mutagenesis. The data suggest that VUF10166 adopts a similar orientation to 5-HT3 receptor agonists bound in AChBP (varenicline) and 5HTBP (5-HT) crystal structures.


Introduction
5-HT 3 receptors are transmembrane ligand-gated ion-channels that are responsible for fast synaptic neurotransmission in the central and peripheral nervous systems. They are composed of five subunits, each of which contains an extracellular, a transmembrane and an intracellular domain (Thompson et al., 2008a;Miller and Smart, 2012). In vivo 5-HT 3 receptor activation can result in nausea and vomiting, and for over three decades competitive antagonists of these receptors have been used to alleviate these symptoms arising from cancer therapy and general anaesthetics. There is also a limited use of antagonists for treating irritable bowel syndrome and pre-clinical interest in the use of partial agonists for the same disorder (Thompson and Lummis, 2007;Walstab et al., 2010;Thompson, 2013).
There are currently five 5-HT 3 receptor subunits (5-HT3Ae5-HT3E), with further complexity arising from splice variants and species differences (Walstab et al., 2010). 5-HT3A subunits can form homomeric receptors, but the subunits 5-HT3Be5-HT3E must combine with 5-HT3A subunits to function. The functional properties of these receptor subtypes have been reported by several groups, but to date only the pharmacologies of 5-HT 3 A and 5-HT 3 AB receptors have been studied in detail (Holbrook et al., 2009;Walstab et al., 2010;. Until recently only pore-blocking antagonists were known to have different properties at 5-HT 3 A and 5-HT 3 AB receptors, and these differences could be attributed to the varying pore-lining amino acids of the 5-HT3A and 5-HT3B subunits . However, the utility of these compounds is limited as they tend to be of low affinity (mM range) and also target other receptor types. More recently there have been descriptions of two compounds with other sites of action that discriminate between 5-HT 3 A and 5-HT 3 AB receptor subtypes. One of these, topotecan, primarily an anticancer drug, was found to inhibit 5-HT 3 A and potentiate 5-HT 3 AB receptors, although this compound also has a relatively low (mM) potency (Nakamura et al., 2013). The second compound is VUF10166 (2-chloro-3-(4methylpiperazin-1-yl)quinoxaline), which is highly potent, with an affinity at 5-HT 3 A receptors (pK i~1 0) that is~100-fold greater than at 5-HT 3 AB receptors . We previously showed that VUF10166 binds to the orthosteric binding site of both 5-HT 3 A and 5-HT 3 AB receptors (formed at the interface of two 5-HT3A subunits, AþAÀ) and that a second, allosteric, binding site (AþBÀ) in the 5-HT 3 AB receptor was responsible for causing ligands at the AþAÀ binding site to dissociate more rapidly.
Here we perform a detailed characterisation of VUF10166 binding to 5-HT 3 A and 5-HT 3 AB receptors with a radiolabelled version of this compound and use mutagenesis to explore the residues that interact with VUF10166 at the AþAÀ binding site. H]VUF10166 was fit with a mono-exponential function to yield k obs . (c) Linear regression was used to fit k obs against the radioligand concentration, yielding the k on (slope) and k off (intercept at y ¼ 0) values in Table 1. (d) Dissociation of [ 3 H]VUF10166 was best fit with a single exponential (k off ¼ 0.011 ± 0.001 min À1 , n ¼ 4). (e) For [ 3 H] granisetron, association was also best fit with mono-exponential functions that were used to plot k obs against the concentration to yield the k on and k off values in Table 1 were added for 1 h at room temperature. The reaction was quenched with 500 ml semi-prep HPLC eluent and subjected to semi-preparative HPLC purification, using a Reprosphere C18-DE 5 mM, 50*8 mm column as stationary phase (Dr. Maisch, Ammerbuch-Entringen, Germany) and acetonitrile/water 75/25 (v/v) with 0.1% diisopropylethylamine as eluent at a flow of 3 ml min À1 , with UV monitoring at 254 nm (Jasco UV-1575, Jasco, de Meern, Netherlands). 30 s fractions were collected, 5 ml of each added to 5 ml scintillation fluid, and counted for 1 min in a beta well counter (Rackbeta 1219 LSC, LKB-Wallac, Netherlands). Fractions containing 2chloro-3-(4-[ 3 H]methylpiperazin-1-yl)quinoxaline were diluted with 45 ml sterile water and passed over a preconditioned Waters tC18 plus Sep-Pak, washed with 20 ml of water, and the product obtained by elution with 1.5 ml ethanol; 35 MBq (83% radiochemical yield) of [ 3 H]VUF10166 was obtained. The specific activity of the product was 3.13 TBq/mmol (84.5 Ci/mmol) and the radiochemical purity was >98%, as determined by HPLC with a Platinum C18 100a, 5 mM 250*4.6 mm column (Grace Alltech, Breda, Netherlands) as stationary phase and acetonitrile/10 mM ammoniumdihydrogen phosphate buffer pH 2.5 50/50 (v/v) as eluent at a flow of 1 ml min À1 , with UV monitoring at 254 nm (Jasco UV-1575) and radioactivity monitoring (Lablogic b-RAM model 4, Metorix, Goedereede, Netherlands).

Site-directed mutagenesis
Mutagenesis was performed using the QuikChange method (Agilent Technologies Inc., California, USA) on human 5-HT3A cDNA (accession number: P46098) cloned into pcDNA3.1 (Invitrogen, Paisley, UK). Cysteine residues were substituted for amino acids throughout each of the binding loops AeE (Fig. 1). To facilitate comparisons with previous work, we use the numbering of the equivalent residues in the mouse 5-HT3A subunit; for human numbering 5 should be subtracted from each residue number.

Radioligand binding
Transfected HEK 293 cells were washed twice with phosphate buffered saline (PBS) at room temperature, scraped into 1 ml of ice-cold HEPES buffer (10 mM, pH 7.4), homogenised and frozen. After thawing, they were washed with HEPES buffer, resuspended, and 50 mg of cell suspension incubated in 0.5 ml HEPES buffer and the relevant concentration of radioligand at 0 C. Non-specific binding was determined using 2 mM quipazine. Equilibrium reactions were incubated for at least 3 h for [ 3 H] granisetron (63.5 Ci/mmol, PerkinElmer, Boston, Massachusetts, USA) and 48 h for [ 3 H]VUF10166. Incubations were terminated by vacuum filtration onto GF/B filters pre-soaked in 0.3% polyethyleneimine, followed by three rapid washes with 3.5 ml ice cold buffer. Radioactivity was determined by scintillation in Ecoscint A (National Diagnostics, Atlanta, Georgia) using a Beckman LS6000SC (Fullerton, California, USA). Each method was performed on at least three independent cell samples on at least three separate days.

Saturation binding
To construct saturation binding curves a range of [ 3 H]granisetron (0.25e2 nM) or [ 3 H]VUF10166 (0.04e50 nM) concentrations were used according to the conditions described above. Final counts were monitored to ensure that binding never exceeded 10% of the added concentrations of radioligands.

Kinetic measurements
To determine the association rate (k on ), the observed association rate (k obs ) was measured for a range of radioligand concentrations. The experiment was started (t ¼ 0) by the addition of radioligand to 500 ml cell suspension in HEPES buffer and harvested at varying time points to construct association curves. Dissociation was measured by allowing each radioligand to reach equilibrium according to the times described above and then adding a final concentration of 2 mM quipazine (~K d Â 10 6 ) to each tube for varying time periods.

Data analysis
All data were analysed using GraphPad Prism 4.03. Individual saturation binding experiments were fitted to Equ (1), and the values averaged to obtain mean ± sem: where B max is maximum binding at equilibrium, K d is the equilibrium dissociation constant and [L] is the free concentration of radioligand. Individual competition binding experiments were analysed by iterative curve fitting using the following equation and the values averaged to obtain the mean ± sem: where B min is the non-specific binding, B max is the maximum specific binding, [L] is the concentration of competing ligand and IC 50 is the concentration of competing ligand that blocks half of the specific bound radioligand.
A simple bimolecular binding scheme for receptor and ligand can be represented as: where L is the free ligand concentration, R receptor concentration, LR bound receptor concentration, and k on and k off microscopic association and dissociation rate constants. In a simple scheme such as this, the equilibrium dissociation constant (K d ) is equal to the ratio of dissociation to association rate constants, such that: Dissociation data were fitted to either a single or double exponential decay to yield k off . Association data were fitted to a single exponential association to calculate k obs . If k obs is plotted against the radioligand concentration, according to a simple model, the slope of this plot equals the association constant (k on ) and the y-intercept of this line (at x ¼ 0) is the dissociation constant (k off ). k on can also be calculated as described by Hill (Hill, 1909), where k off is predetermined from radioligand dissociation rate experiments.

Homology modelling
The protein sequence of the human 5-HT3A subunit (accession number; P46098) was aligned with a tropisetron bound AChBP template (PDB ID; 2WNC) using FUGUE. Using Modeller 9.9, five homology models were generated using default parameters and the best model selected using Ramachandran plot analysis. For the ligand, the protonated form of VUF10166 was constructed in Chem3D Ultra 7.0 (CambridgeSoft, Cambridge, UK). The binding site was defined as being within 5 Å of the a-carbon of W183, a residue critical in the binding of other 5-HT 3 competitive ligands. VUF10166 was docked into this site using the GOLD docking program (version 3.0, The Cambridge Crystallographic Data Centre, Cambridge, UK) with the GOLDScore function and default settings. Ten docking poses were generated for each of the five homology models and the poses visualised with PyMol v1.3.

[ 3 H]VUF10166 binding at 5-HT 3 A receptors
[ 3 H]VUF10166 showed high affinity saturable binding at 5-HT 3 A receptors with low levels (<5%) of non-specific binding. The K d value was similar to the K i value from competition of unlabelled Table 1 Binding parameters for VUF10166 and BRL43694.

VUF10166 kinetic parameters at 5-HT 3 A receptors
Association curves for [ 3 H]VUF10166 were best fit with a single exponential function (Fig. 1b), and the resultant rates (k obs ) plotted against ligand concentration to yield k on and k off (Fig. 1c, Table 1). The value for k on was similar to values determined directly from k obs values using Equ (5) (8.24 Â 10 7 M min À1 ). Dissociation of [ 3 H] VUF10166 in the presence of excess cold quipazine was also monophasic (Fig. 1d), with k off values that were similar to those determined from plots of k obs against ligand concentration (Table 1). K d values calculated from these kinetic measurements (Equ (4)) were similar to those derived from the saturation and competition binding (Table 1). These results indicate [ 3 H] VUF101666 binding can be best described by a simple bi-molecular binding scheme.

Specificity of binding
A range of competitive and non-competitive ligands of 5-HT 3 and related Cys-loop receptors were tested for their ability to compete with [ 3 H]VUF10166 binding (

Granisetron binding at 5-HT 3 A receptors
To compare [ 3 H]VUF10166 with a well-established 5-HT 3 receptor competitive ligand, experiments were also conducted using [ 3 H]granisetron. As expected, [ 3 H]granisetron showed high affinity binding at 5-HT 3 A receptors (Table 1). Competition binding with a range of known 5-HT 3 receptor agonists and antagonists gave K i values similar to those determined using competition with [ 3 H] VUF10166 (Table 3) and to those published elsewhere (Brady et al., 2001). Similar to [ 3 H]VUF10166, nicotine and strychnine competed with [ 3 H]granisetron.
[ 3 H]granisetron association rates were best fit with a monophasic curve. k obs increased with free ligand concentration and a straight line was fitted (Fig. 1e) to yield the k on and k off values in Table 1. K d values calculated from these kinetic measurements (Equ (4)) were in agreement with affinities calculated from our saturation binding studies (Table 1). Dissociation was also monophasic and the rate agreed well with that from our k obs versus concentration plots described above (Fig. 1f, Table 1).
These observations show that using a well-established radiolabelled 5-HT 3 receptor antagonist ([ 3 H]granisetron) we are able to accurately reproduce the binding characteristics reported elsewhere and, similar to [ 3 H]VUF10166, they are consistent with a simple bi-molecular binding scheme.

Homology modelling & docking
To gain insights into the residues that potentially interact with VUF10166 at the orthosteric site (AþAÀ interface), five 5-HT 3 A receptor homology models were generated and in silico docking of VUF10166 performed on each one (Fig. 2). A total of 50 docked poses were generated and for each of these the amino acids within 5 Å of VUF10166 were identified (Table 4). 26% of residues were common to all models, comparable to a previous docking study with granisetron, where 31% of residues were common to all of the predicted binding orientations (Thompson et al., 2005). A selection of these residues were chosen for mutagenesis based upon the following criteria, 1) side chains accessible to the ligand, 2) residues known to interact with other 5-HT 3 ligands or, 3) residues present in a limited number of docked poses to provide support for specific orientations. Of the 39 amino acids identified, 23 were mutated to cysteine (Fig. 3); cysteine substitution of these residues was chosen as all of the Cys mutants have been previously shown to express on the cell-surface, and the residue positions have been similarly used for the study of our radioligand standard, [ 3 H]granisetron (Thompson et al., 2005(Thompson et al., , 2011.

VUF10166 binding at 5-HT 3 AB receptors
VUF10166 was previously shown to discriminate between 5-HT 3 receptors subtypes  and so binding properties of the new radioligand were also tested at 5-HT 3 AB receptors. [ 3 H]VUF10166 showed high affinity binding at 5-  Table 4. HT 3 AB receptors, but unlike at 5-HT 3 A receptors, it was complex and could not be fit with a single site model (Fig. 5a). Dissociation of [ 3 H]VUF10166 at these receptors was best fit with a double exponential curve, which contained both a fast and a slow component; the latter was not significantly different (p < 0.05) to the single rate measured at 5-HT 3 A receptors (Fig. 5b, Table 1). Association curves were monophasic (Fig. 5c), but when k obs was plotted against radioligand concentration, the data were also best approximated by a two site fit (Fig. 5d, Table 1). At concentrations of [ 3 H]VUF10166 < 3 nM the k off and k on values were similar to 5-HT 3 A receptors; below 3 nM, average k on values determined from k obs (Equ (5)) were also similar to 5-HT 3 A receptors (8.77 Â 10 7 M min À1 ). At concentrations >3 nM, k off and k on had slower rates that yielded a K d (22.4 nM; Equ (4)) close to the value from competition binding (36.7 nM; Table 1). Competition binding with a range of ligands was performed using 0.6 nM [ 3 H] VUF10166 and K i values were similar to values at 5-HT 3 A receptors (Table 2). Table 4 Residues within 5 Å of docked VUF10166 in 5 different homology models of the 5-HT 3 A receptor binding site. These results show that [ 3 H]VUF10166 has different binding properties at 5-HT 3 A and 5-HT 3 AB receptors. In the latter effects are complex and some only become apparent at higher concentrations of [ 3 H]VUF10166.

Granisetron binding at 5-HT 3 AB receptors
Unlike [ 3 H]VUF10166, [ 3 H]granisetron saturation binding at 5-HT 3 AB receptors yielded K d values that were the same as those at 5-HT 3 A receptors, as reported elsewhere (Table 1) (Brady et al., 2001). Association (Fig. 5e), dissociation (Fig. 5f) and K i values from competition binding (Table 3) were also the same as those at 5-HT 3 A receptors.
These results show that the binding properties of [ 3 H]granisetron are the same at 5-HT 3 A and 5-HT 3 AB receptors unlike those of [ 3 H]VUF10166.

Discussion
[ 3 H]VUF10166 binds specifically and with high affinity to 5-HT 3 A and 5-HT 3 AB receptors, with evidence of a second, lower affinity, binding site in 5-HT 3 AB receptors. The effects of this second site are apparent at concentrations of [ 3 H]VUF10166 > 3 nM, and are consistent with previous work that identified an additional allosteric binding site for unlabelled VUF10166 at the AþBÀ interface . Docking of this competitive ligand into the orthosteric (AþAÀ) binding site, combined with data from mutagenesis, suggest that VUF10166 is oriented with its quinoxaline rings close to W183 and its basic nitrogen extended towards loop E. Individual residues, many of which have been previously shown to be important in studies of other 5-HT 3 receptor ligands (including d-tubocurarine, granisetron, lerisetron, meta-chlorophenylbiguanide and tropisetron) are also important for VUF10166 binding (Hope et al., 1999;Mochizuki et al., 1999;  Venkataraman et al., 2002a;Price and Lummis, 2004). The residues are discussed in more detail below.

The role of loop A residues
VUF10166 binding was abolished by Cys substitution of E129, slightly modified by L126C (~4 fold change in K d ) and not altered by N128C. E129 was previously identified as an important 5-HT 3 receptor binding residue and may form a hydrogen bond with bound ligand, which is consistent with our data (Price et al., 2008). However, data from 5HTBP (a modified AChBP with high affinity binding for 5-HT 3 receptor ligands) suggest that E129 may hydrogen bond with the side chain of T179 (Kesters et al., 2013), and therefore might have primarily a structural role. L126 may also have a structural role but is less important as the effects of altering this residue were small, while N128 has been shown to play a role in gating but not binding (Price et al., 2008;Kesters et al., 2013).

The role of loop B residues
Loop B has been previously identified as both a critical structural component of the binding pocket, and it contributes to ligand binding. W183 is especially important as a constituent of the 'aromatic box' that exists in all Cys-loop receptor binding sites (Beene et al., 2002;Thompson et al., 2008b;Duffy et al., 2012). Other residues (T179, H185, D189) are known to stabilise the binding site structure via hydrogen bonds (Thompson et al., 2008b;Kesters et al., 2013). It is therefore not surprising that all our loop B mutations altered or abolished [ 3 H]VUF10166 binding and we suggest that T181 and W183 interact with VUF10166 while T179, H185 and D189 have a structural role.

The role of loop C residues
F226 and Y234 are also constituents of the aromatic box and mutations here alter or eliminate VUF10166 binding. F226A has no effect on granisetron binding affinity, indicating this residue is more important for VUF10166 binding (Thompson et al., 2005). In 5HTBP Y234 (Y193) interacts with 5-HT and also contributes to a conserved water network that stabilises the granisetron-bound structure (Kesters et al., 2013); a conserved water network is also seen at this location in many AChBP crystal structures and may be important in many Cys-loop receptors. E236C also abolished VUF10166 binding, consistent with studies where substitutions affect binding of both GR65630 and granisetron, as well as altering the maximal current and EC 50 of 5-HT responses (Schreiter et al., 2003;Nyce et al., 2010). However E236 mutations may adversely affect the correct assembly of the binding site rather than interfering with specific ligand interactions as Nyce et al. (2010) and Schreiter et al. (2003) showed that some E236 mutant receptors are trapped within the cell. As this hypothesis is supported by the lack of interactions in the 5HTBP structure, we consider it unlikely that E236 contributes to VUF10166 binding (Kesters et al., 2013).

The role of loop D residues
W90 is another aromatic box residue that contributes to binding. In 5HTBP the equivalent residue (W53) is involved in van der Waals interactions with granisetron and W90 may have a similar role in binding VUF10166 (Spier and Lummis, 2000;Price and Lummis, 2004;Thompson et al., 2005;Yan and White, 2005). Substitutions at W90 decrease the affinity of other potent 5-HT 3 receptor-specific ligands such as curare, lerisetron and 5-HT (Yan et al., 1999;Venkataraman et al., 2002a) R92 interacts with granisetron in 5HTBP (R55), and the effects of its substitution on the affinity of VUF10166, ondansetron, granisetron and MDL72222, suggest an interaction with all of these ligands (Thompson et al., 2005;Yan and White, 2005).

The role of loop E residues
All of the mutations in loop E (Y141, Y143, R145, Q151, Y153) caused significant changes to [ 3 H]VUF10166 binding. In the 5HTBP crystal structure granisetron does not extend towards loop E, but instead lies horizontally between loops B and D, similar to the orientations of the closely related ligands tropisetron (2WNC) and cocaine (2PGZ) in AChBP. In contrast, in 5HTBP 5-HT hydrogen bonds with the backbone carbonyls of I104 (Y141 in 5-HT 3 ) and I116 (Y153), and has hydrophobic interactions with M114 (Q151), explaining why 5-HT activation is strongly affected by mutations at these locations, but effects on granisetron are less apparent (Venkataraman et al., 2002b;Price and Lummis, 2004;Thompson et al., 2011;Kesters et al., 2013). Here the affinity of VUF10166 was decreased 10-fold by Y153C and abolished by Y143C, indicating that bound VUF10166 extends towards, and may interact with, loop E residues. As VUF10166 is also a low efficacy partial agonist at mM concentrations, and must therefore induce the same structural changes as 5-HT, it is likely it adopts an orientation that at least partially mimics that of 5-HT.

The orientation of VUF10166 in the ligand binding pocket
Our results show that VUF10166 binding is affected by many of the residues previously identified as important for binding 5-HT 3 receptor antagonists, while mutation of R58, I71, K112 and S114, which are close to VUF10166 in models 1, 2 and 5, did not alter its affinity, suggesting that these models are less probable. Also in model 1 the predicted ligand orientations do not extend towards Loop E and yet residues here were important for VUF10166 binding. Similarly R145 is within 5 Å of VUF10166 in model 2, but our mutagenesis data show that altering this residue has little effect on binding affinity. Model 3 seems unlikely as these poses are positioned closer to the complimentary face of the binding site, and do not significantly interact with key principal face residues such as Fig. 5. Radioligand binding at 5-HT 3 AB receptors. (a) Binding at 5-HT 3 AB receptors could not be well fit with a standard one site model; deviation occurs at a radioligand concentration of~3 nM (arrow). Inset competition binding of unlabelled VUF10166 with [ 3 H]granisetron. (b) Dissociation was best fit with a double exponential at 5-HT 3 AB receptors (0.010 ± 0.003 min À1 and 0.227 ± 0.056 min À1 , n ¼ 8). (c) Association was mono-exponential, but a plot of k obs against radioligand concentration. (d) revealed two components, showing that it was rate-limited at higher concentrations. (e) The association of [ 3 H]granisetron was best fit with a mono-exponential function, but unlike [ 3 H]VUF10166, the fit of k obs against the radioligand concentration was linear at across all concentrations, yielding the values for k on and k off in Table 1. (f) Consistent with this plot, dissociation of [ 3 H] granisetron was also best described by a single exponential function (k off ¼ (0.012 ± 0.002 min À1 , n ¼ 5)) that was not significantly different to 5-HT 3 A receptors (p > 0.05, Student's t-test).
T181, W183 and Y234. Models 4 and 5 have quite similar docked poses with only F226 distinguishing them; F226C mutant receptors had a 7-fold lower affinity than wild type receptors suggesting that this residue is close enough to interfere with VUF10166 binding, which would best fit with model 4.
In previous work we presented a structure-activity study (SAR) of VUF10166 analogues  and the active analogues from these studies would fit well into model 4 in two distinct orientations (Fig. 6). These data showed substitutions of the chlorine atom in VUF10166 (Fig. 6a, region 2) are poorly tolerated, suggesting an important interaction at this location; in both poses in Fig. 6 the chlorine atom is closely located to R92 and W90. In contrast, substitutions in regions 1 and 3 are fairly well tolerated, providing that they are not too large; neither of the poses in Fig. 6 are sterically restricted around these regions of VUF10166. The poses also explain the importance of the charged Nmethylpiperazine nitrogen atom, as there are possible cationep interactions with W183 and Y234 in one pose, with these residues contributing to pep stacking of the quinoxaline ring in the other.
We therefore suggest that the docked poses in model 4 are most consistent with the mutagenesis data described here and our previously published SAR. It is difficult to predict whether the N- Fig. 6. Chemical structure of VUF10166 and its binding mode. (a) Three regions of the ligand are identified and are described in the text. Its protonation site, which is also its tritiation site, is indicated. (bec) The volume occupied by the two main docked pose clusters in model 4. In (c) cationep interactions are possible with W183 (5.06 Å away) and Y234 (4.46 Å). VUF10166 is shown as a stick and wire mesh representation (white), with the residues mutated in this study colour coded similar to Fig. 4. (dee) Cartoons showing our interpretation of the binding to heteromeric receptors. Below 3 nM, VUF10166 binds to a single population of binding sites at the AþAÀ interface of both 5-HT 3 A and 5-HT 3 AB receptors; consequently, at these concentrations both receptors share common values for k on and k off . At concentrations of VUF10166 > 3 nM, binding also occurs at a second AþBÀ binding site and allosterically influences the adjacent AþAÀ site; therefore, additional rates are apparent and saturation binding is confounded by rates associated with multiple binding sites and allosteric interactions. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) methylpiperazine ring or the quinoxaline ring is positioned toward loop E, but the orientation in Fig. 6c is most reminiscent of varenicline co-crystallised into AChBP (PDBID ¼ 4AFG & 4AFT) and 5-HT in 5HTBP (2YMD), both of which are agonists at 5-HT 3 receptors (Billen et al., 2012;Rucktooa et al., 2012). This similarity in orientation may explain why VUF10166 also displays partial agonist activity . However, it should be stressed that we must exercise caution when making these predictions as the physiological relevance of these structures have not yet been fully ascertained, for example three ligand molecules have been observed in a single AChBP binding site, something we would not have predicted (Brams et al., 2011;Stornaiuolo et al., 2013).

Conclusion
Our results show that VUF10166 interacts with several of the core binding site residues found at the AþAÀ interface and, combined with homology modelling and ligand docking, we propose it adopts an orientation similar to that of other 5-HT 3 receptor agonists in AChBP and 5HTBP crystal structures. At 5-HT 3 receptors our kinetic measurements are consistent with a single AþAÀ binding site, but at 5-HT 3 AB an additional fast component is seen. This is consistent with the lower affinity of VUF10166 for the 5-HT 3 AB receptor and is likely to result from an allosteric effect that is evident when the concentration of VUF10166 exceeds 3 nM (as summarised in Fig. 6).