Exploring a potential palonosetron allosteric binding site in the 5-HT3 receptor

Graphical abstract


Introduction
Cys-loop receptors constitute one of the largest families of ligand-gated ion channels responsible for the transmission of fast synaptic impulses in the CNS and PNS. The nicotinic acetylcholine, glycine, GABA A and 5-HT 3 receptors, all belong to this family, sharing homologous amino acid sequences and similar arrangements of subunits. 1,2 The 5-HT 3 receptor is formed by a pentameric assembly of identical (homomeric) or different (heteromeric) subunits surrounding a central ion-conducting pore. Each subunit has an extracellular domain, a transmembrane domain and an intracellular domain. The binding pocket is formed between two adjacent subunits, and is constituted of the + (or principle) face of one subunit and the À (or complementary) face of the adjacent subunit (e.g., A+AÀ when between two 5-HT3A receptor subunits). To date, five different 5-HT 3 receptor subunits have been identified. Receptors that contain only 5-HT3A receptor subunits (5-HT 3 A receptors) or 5-HT3A and 5-HT3B receptor subunits (5-HT 3 AB receptors) have been widely investigated, but receptors containing 5-HT3C, 5-HT3D and 5-HT3E receptor subunits have been less well characterized. [1][2][3][4][5][6][7][8] Radioligand binding and in situ hybridization studies have shown that 5-HT 3 receptors are present throughout the nervous system, and play a role in a variety of nervous and sensory functions, as well as in emesis, cognition and anxiety. 1,3,4 A number of 5-HT 3 receptor antagonists are therapeutic agents, including ondansetron, tropisetron, granisetron and palonosetron, with major uses in the management of chemotherapy-induced, radiationinduced and post-operative nausea and vomiting. 5 Palonosetron is the newest 5-HT 3 receptor antagonist licensed for clinical use and demonstrates improved control of acute (0-24 h) and delayed (24-120 h) emesis after moderate chemotherapy administration when compared to early generation antagonists such as granisetron (Kytril). Palonosetron also has a higher in vitro binding affinity for the 5-HT 3 receptor compared to other antagonists (pK i of 10.5 vs 8-9), and an extended in vivo plasma elimination half-life. 6,7 Palonosetron has unique structural characteristics when compared to other drugs in its class and data suggest that palonosetron interacts with 5-HT 3 receptors differently to granisetron or ondansetron, with early studies indicating allosteric effects and positive cooperativity. 12,13 These authors also provided data that suggests palonosetron triggers receptor internalisation, resulting in a long-lived inhibition of receptor function, although more recent data show that the dissociation of palonosetron is very slow, which could explain this effect. 8,9 Recent computational work modelled the extracellular domain of the human 5-HT 3 A and 5-HT 3 B subunits assembled into homoand heteromeric receptors to yield A+AÀ and B+AÀ interfaces. 10 Docking of serotonin, palonosetron and granisetron into these binding interfaces was used to probe the binding characteristics of palonosetron (Fig. 1), with the aim of providing a viable explanation for the higher efficacy of palonosetron, as well as investigating the allosteric binding and positive cooperativity. These studies suggested a second binding site for palonosetron at the A+AÀ interface that was not found for granisetron; this was located below the 5-HT binding site, closer to the membrane. Palonosetron was also proposed to bind to a B+AÀ interface. In silico studies such as these are increasingly being used to probe the mechanism of action of drugs, but because many of these use homology models, it is essential such data are confirmed using experimental studies. The present work uses HEK293 expressed mutant 5-HT 3 A and 5-HT 3 AB receptors to experimentally probe the residues identified as interacting with palonosetron in the previously published in silico study (Fig. 2, Table 1).

Materials
All cell culture reagents were obtained from Gibco BRL (Paisley, UK), except foetal calf serum which was from Labtech International (Ringmer, UK). [ 3 H]Granisetron (84.5 Ci/mmol) was from PerkinElmer (Boston, MA). FlexStation membrane potential dye was from Molecular Devices Ltd (Wokingham, UK). [ 3 H]Palonosetron (37.2 Ci/mmol) was custom synthesised for Helsinn Healthcare (Lugano, Switzerland), and both this and the unlabelled form of palonosetron were kindly gifted by Helsinn Healthcare (Lugano, Switzerland). All other reagents were of the highest obtainable grade.

Cell culture
Human embryonic kidney (HEK) 293 cells were cultured on 90 mm tissue culture plates at 37°C and 7% CO 2 in a humidified atmosphere. They were grown in DMEM:F12 (Dulbecco's Modified Eagle's Medium/Nutrient Mix F12 (1:1)) with GlutaMAX I™ containing 10% fetal calf serum and passaged when confluent. For radioligand binding studies, cells in 90 mm dishes were transfected at 60-70% confluency using polyethyleneimine and incubated for 3-4 days before use. For FlexStation assays, transfected cells were plated on 96-well plates and incubated 1-2 days before assay.

Site-directed mutagenesis
Mutagenesis reactions were performed using either the Kunkel method or the QuikChange kit (Agilent Technologies Inc., CA, USA) using human 5-HT3A or 5-HT3B receptor subunit cDNA (accession numbers: P46098 and O95264) in pcDNA3.1 (Invitrogen, Paisley, UK). Subunit numberings have been altered to the aligning residues in the mouse 5-HT3A subunit for ease of comparison with other studies.
Eight-point binding assays were performed to produce saturation binding curves. Membrane suspensions (50 ll) in 0.5 ml HEPES buffer were incubated at 4°C with 0.05-2 nM [ 3 H]-granisetron (for 1 h) or [ 3 H]-palonosetron (for 24 h). Unlabelled quipazine (0.1 mM) was used to determine non-specific binding. Membranes were harvested onto presoaked (0.3% polyethyleneimine) GF/B filters using a Brandel harvester, followed by two 2 ml washes of icecold HEPES buffer. Radioactive counts were measured using a Beckman LS6000SC liquid scintillation counter (Fullerton, California, USA). Data were analyzed by iterative curve fitting using GraphPad Prism v5 (San Diego, CA) according to the following equation: where B is the amount of bound radioligand, B max is the maximum specific binding at equilibrium, [L] is the concentration of radioligand, K d is the equilibrium dissociation constant of the radioligand, and n is the Hill coefficient.

Functional studies
Fluorometric studies were undertaken using a FlexStation Ò (Molecular Devices, Sunnyvale, CA, USA) as previously described. 11 Transfected cells grown on 96-well plates were washed using a standard buffer (10 mM HEPES, 115 mM NaCl, 1 mM KCl, 1 mM CaCl 2 , 1 mM MgCl 2 , 10 mM glucose, pH 7.4), loaded with FLIPR Ò Membrane Potential Dye (Molecular Devices) according to the manufacturer's instructions, and then incubated at 37°C for 45 min before being assayed.
where F is the fluorescence, F max is the maximum response, F min is the baseline fluorescence, [L] is the ligand concentration, n H is the Hill coefficient, and the EC 50 is the ligand concentration required to obtain 50% of maximal fluorescence.

Statistical analysis
Values are presented as mean ± SEM. Statistical analysis was performed using ANOVA in conjunction with a Dunnett's posthoc test; p <0.05 was taken as statistically significant. Mutant receptors (Y73A, Y73F, Y73S, F130A, F130Y, S163A, S163T, D165A and D165K) had K d values were not significantly different to wild type ( Table 2).

Mutant heteromeric receptors.
Most receptors containing mutations in the 5-HT 3 B receptor subunit (H73A, Y143A, and E170A) had K d values that were not significantly different to wild type. F130A containing heteromers showed a small ($fourfold) increase in K d compared to wild type receptors ( Table 2).

Mutant heteromeric receptors.
Most mutant receptors containing subunits with 5-HT 3 B subunit mutations (H73A, W90C, Y143A, and E170A) had K d values similar to wild type, but there was a small increase ($twofold) in K d for F130A containing heteromers.

Functional assays 3.1.3.1. Wild type receptors.
Our functional responses of WT 5-HT 3 AB receptors expressed in HEK cells and studied using membrane potential sensitive dye show a similar EC 50 and a decrease in Hill slope (EC 50 = 0.5 lM, n H = 1.3) when compared to wild type 5-HT 3 A receptors (EC 50 = 0.32 lM, n H = 2.3). These data are similar to previous reports. 14-16

Mutant homomeric receptors.
One of the mutant receptors, F130Y, was non-functional. As this mutant had binding characteristics similar to wild type, this indicates that the presence of Y at this position produces non-functional receptors. Y73A and Y73S mutant receptors had lower EC 50 values than wild type receptors. All other mutant receptors had EC 50 values that were not significantly different to wild type (Table 3).
Inhibition studies with palonosetron revealed its IC 50 was increased ($15-fold) at Y73A containing mutant receptors compared to wild type receptors. No other receptors were different to wild type receptors (Table 4).

Mutant heteromeric receptors.
None of the heteromeric receptors had 5-HT EC 50 values significantly different to wild type receptors, and none had differences in IC 50 values for palonosetron (Tables 3 and 4).

Probing the orthosteric binding site
To probe the orthosteric binding site we examined [ 3 H]palonosetron binding to receptors that had substitutions at two critical binding site residues in the 5-HT3A subunit; W183 in loop B, and W90 in loop D. The binding characteristics of receptors containing Cys substitutions at these locations revealed no saturable, specific binding with [ 3 H]palonosetron (Table 2). This is consistent with previously published data from [ 3 H]granisetron binding studies, and indicates that palonosetron binds in this pocket. Lack of specific [ 3 H]palonosetron binding could alternatively indicate that the mutant receptors were not expressed, but we consider this is unlikely as a previous work has demonstrated cell surface expression of 5-HT 3 receptors with mutations of W90 and W183. 17 To probe possible binding to A+BÀ, A+BÀ, B+AÀ or B+BÀ interfaces in 5-HT 3 AB receptors, we also substituted the equivalent residues in the 5-HT 3 B subunit (I183 and W90) and co-expressed them with wild-type 5-HT3A subunits. [ 3 H]palonosetron binding data revealed that binding affinities and functional responses were unaltered in these modified 5-HT 3 AB receptors.

Discussion
Palonosetron is a potent 5-HT 3 receptor antagonist that has improved properties for ameliorating the symptoms of chemotherapy-induced and post-operative nausea and vomiting when compared to earlier antagonists. A possible explanation for this high potency is the presence of an allosteric binding site, and such a site was suggested in a computational study by Moura Barbosa et al. 10 Here we used mutagenesis to probe the proposed interactions between palonosetron and residues in the second A+AÀ binding site and those in the 5-HT 3 B containing binding sites; the data confirm previous studies that show palonosetron binds to the orthosteric binding site located between two adjacent 5-HT 3 A subunits (with no contribution from 5-HT 3 B subunits), but do not support an alternative binding location. 6,7,18,13 There was some indication that Y73 had an interaction with palonosetron, although the effects of mutating this residue were small. The roles of each of the residues that we examined are discussed in more detail below.  4.40 ± 1.8 3 F130Y NF 9 S163A 1.57 ± 0.12 3 S163T 1.02 ± 0. 0.38 ± 0.05 9 E170A 0.36 ± 0.11 5 I183A 0.34 ± 0.08 4 Y73 is located just below the binding pocket and in the model presented by Moura Barbosa et al., 10 shows a side to edge interaction with palonosetron (Fig. 2). Our substitutions here caused a small increase in the K d for palonosetron binding and the IC 50 for inhibition of 5-HT-induced responses, suggesting that Y73 does not have a major interaction with palonosetron, but may influence the orthosteric site above, or may form part of a temporary binding location on route to the 'classic' binding pocket; a simulation study showing the trajectory of granisetron as it unbinds from the receptor indicates that ligands exits below the binding site close to Y73, 19 although altering the equivalent residue in mouse 5-HT 3 receptors has no effect on [ 3 H]granisetron binding. 20 Therefore, it is more likely that this residue is structurally important, as suggested in a recent computational work, where molecular dynamics simulations and computational alanine scanning mutagenesis revealed that Y73 is part of the 'hot centre' of the subunit-subunit interaction, an aromatic cluster located in the middle of the binding interface involved in the stabilization of the protein. 21,22 Further evidence for a structural role for this residue comes from our EC 50 values, which are lower for Y73A and Y73S mutant receptors.
F130 is located on loop A and has been extensively investigated in mouse 5-HT 3 A receptors, where it has been shown to have an important role in function: mutation to Ala or Trp decreases the 5-HT EC 50 yet increases the granisetron K d , while mutation to Tyr increases the EC 50 but has no effect on the K d . 23,24 These data suggest that this residue does not directly bind ligand, but that adding a hydroxyl is deleterious to receptor function; our results from human 5-HT 3 receptors reveal a similar but more pronounced effect as F130Y-containing receptors are expressed but are non-functional. The equivalent residue in the 5-HT 3 B subunit was the only mutated residue (F130A) that caused an increase in granisetron binding affinity. There is currently no evidence that granisetron binds to a 5-HT 3 B subunit-containing binding pocket, and we suggest that this mutation causes a subtle change in receptor structure that influences the ligand binding (A+AÀ) interface. Support for this hypothesis comes from our palonosetron binding data, which reveals a small increase in affinity in receptors containing F130Acontaining 5-HT 3 B receptor subunits. There was, however, no change in EC 50 suggesting no significant effect on agonist binding and/or channel gating. There was also no effect on palonosetron inhibition of function, suggesting the inhibitory effect of palonosetron does not involve binding to F130. S163 is located below the binding pocket and was predicted to be <5 Å from palonosetron in the allosteric site proposed by Moura Barbosa et al., 10 but not directly involved in its binding. Our data showed that mutation of S163 to either Ala or Thr resulted in no significant changes to binding or functional parameters, suggesting this residue does not play a significant role in palonosetron binding or receptor function. This is in agreement with a previous study that also concluded that there was no involvement of this Ser in channel activation or in coupling of ligand binding to channel gating as this residue was found to be a 'warm spot' only involved in subunit-subunit interactions stabilising the receptor. 21 D165 lies further from the orthosteric binding pocket than the residues discussed above but is a potentially novel palonosetron binding site identified in the Moura Barbosa et al. study, 10 and is also on the unbinding pathway for 5-HT and granisetron. 19 However our data revealed that mutation of D165 to either Ala or Lys resulted in no significant changes to binding or functional parameters, suggesting this residue does not play a role in palonosetron binding.
The docking study of Moura Barbosa et al. 10 was aimed at probing whether ligands could bind to an A+AÀ or a B+AÀ interface. These interfaces were chosen because the critical residues that differ between 5-HT 3 A and 5-HT 3 B receptor subunits are on the prin-cipal face, and thus A+AÀ would be equivalent to A+BÀ, and B+AÀ equivalent to B+BÀ, covering all possible subunit interface combinations that could occur in a 5-HT 3 AB receptor; furthermore computational studies have examined the role played by the extracellular moieties of the A and B subunits in the formation of functional or non functional receptors, which have been analyzed in terms of the hydrophobic and electrostatic properties of the different subunits and their complementarity. 25 At that time the stoichiometry of the receptor had been proposed to be BBABA, 25 but more recent data has shown that functional receptors require an A+AÀ interface and the heteromeric stoichiometry is 3A2B, with a likely subunit arrangement of AABAB 26,13,27 Our current data show the 5-HT 3 B subunit mutations have little or no effect on receptor binding or function, consistent with palonosetron binding only to an A+AÀ interface, although there may be some allosteric effects of 5-HT 3 B subunit mutations, as we have proposed for F130. A 5-HT 3 antagonist that does interact with both the A+AÀ and A+BÀ interfaces has recently been identified. 28 This compound (VUF10166) shows modified binding and functional effects when tested on A+BÀ receptor containing mutant 5-HT 3 B subunits, indicating an allosteric effect at an A+BÀ interface, although it also competed for granisetron at the A+A-binding site.

Conclusions
In conclusion our data support a binding location for palonosetron in the orthosteric (A+AÀ) binding site in both homomeric 5-HT 3 A and heteromeric 5-HT 3 AB receptors, but do not provide evidence for a second binding site for palonosetron at either the proposed allosteric site at the A+AÀ binding site or at the B+AÀ interface. The question remains as to whether there is another palonosetron binding site elsewhere in 5-HT 3 A or 5-HT 3 AB receptors. This is potentially possible but recent data suggest that at least some of the unusual effects of palonosetron can be explained by its unusual kinetics, and thus we consider that a second binding site is unlikely.