Different binding modes of tropeines mediating inhibition and potentiation of α1 glycine receptors

Authors

  • Gábor Maksay,

    1. Department of Molecular Pharmacology, Institute of Biomolecular Chemistry, Chemical Research Center, Hungarian Academy of Sciences, Budapest, Hungary
    2. Department of Neurochemistry, Max-Planck-Institute for Brain Research, Frankfurt/Main, Germany
    Search for more papers by this author
  • Bodo Laube,

    1. Department of Neurochemistry, Max-Planck-Institute for Brain Research, Frankfurt/Main, Germany
    2. AG Zelluläre und molekulare Neurophysiologie, Technische Universität Darmstadt, Darmstadt, Germany
    Search for more papers by this author
  • Rudolf Schemm,

    1. Department of Theoretical and Computational Biophysics, Max-Planck-Institute for Biophysical Chemistry, Göttingen, Germany
    Search for more papers by this author
  • Joanna Grudzinska,

    1. Department of Neurochemistry, Max-Planck-Institute for Brain Research, Frankfurt/Main, Germany
    Search for more papers by this author
  • Malgorzata Drwal,

    1. Department of Neurochemistry, Max-Planck-Institute for Brain Research, Frankfurt/Main, Germany
    Search for more papers by this author
  • Heinrich Betz

    1. Department of Neurochemistry, Max-Planck-Institute for Brain Research, Frankfurt/Main, Germany
    Search for more papers by this author

  • The present address of Joanna Grudzinska is the School of Biomolecular and Biomedical Research, Conway Institute, University College Dublin, Belfield, Dublin 4, Ireland.

Address correspondence and reprint requests to Dr Gábor Maksay, Chemical Research Center, Hungarian Academy of Sciences, PO Box 17, H-1525 Budapest, Hungary. E-mail: maksay@chemres.hu

Abstract

Tropeines are bidirectional modulators of native and recombinant glycine receptors (GlyRs) and promising leads for the development of novel modulatory agents. Tropisetron potentiates and inhibits agonist-triggered GlyR currents at femto- to nanomolar and micromolar concentrations respectively. Here, the potentiating and inhibitory effects of another tropeine, 3α-(3′-methoxy-benzoyloxy)nortropane (MBN) were examined by voltage-clamp electrophysiology at wild type and mutant α1 GlyRs expressed in Xenopus laevis oocytes. Several substitutions around the agonist-binding cavity of the α1 subunit interface (N46C, F63A, N102A, R119K, R131A, E157C, K200A, Y202L and F207A) were found to reduce or eliminate MBN inhibition of glycine activation. In contrast, the binding site mutations Q67A, R119A and S129A which did not affect MBN inhibition abolished the potentiation of chloride currents elicited by low concentrations of the partial agonist taurine following pre-incubation with MBN. Thus, potentiation and inhibition involve distinct binding modes of MBN in the inter-subunit agonist-binding pocket of α1 GlyRs. Homology modelling and molecular dynamics simulations disclosed two distinct docking modes for MBN, which are consistent with the differential effects of individual binding site substitutions on MBN inhibition and potentiation respectively. Together these results suggest that distinct binding modes at adjacent binding sites located within the agonist-binding pocket of the GlyR mediate the bidirectional modulatory effects of tropeines.

Abbreviations used
5-HT3 receptor

type-3 5-hydroxytryptamine receptor

AChBP

acetylcholine binding protein

GABAA

A-type GABA

GlyR

glycine receptor

LBD

ligand-binding domain

MBN

3α-(3′-methoxy-benzoyloxy)nortropane

MD

molecular dynamics

Rg

radius of gyration

wt

wild type

Glycine is the major inhibitory neurotransmitter in spinal cord and brain stem (Aprison 1990). There, glycinergic synapses control motor and sensory pathways by activating strychnine-sensitive glycine receptors (GlyRs) which belong to the ‘Cys-loop’ family of ligand-gated ion channels (Laube et al. 2002; Lynch 2004). Five GlyR subunit genes (α1–4 and β) have been identified by molecular cloning. Upon heterologous expression, α subunits, but not β, form functional homopentameric chloride channels. In adult spinal cord, GlyRs are mainly heteropentamers composed of two α1 and three β subunits (Grudzinska et al. 2005), whereas at birth homomeric α2 GlyRs prevail (Becker et al. 1988). GlyRs containing α3 subunits participate in central inflammatory pain sensitization (Harvey et al. 2004).

Agonist-induced GlyR currents can be modulated by different allosteric agents or ‘modulators’ (reviewed in Laube et al. 2002). However, most of these modulators, such as anaesthetic alcohols, neurosteroids, Zn2+and tropeines, are not GlyR-selective but similarly affect the responses to A-type GABA (GABAA) receptors (Laube et al. 2002) or act at type-3 5-hydroxytryptamine (5-HT3) receptors (Chesnoy-Marchais 1996). Notably, several GlyR modulatory compounds such as tropeines, dihydropyridines and cannabinoids (Yang et al. 2008) produce bidirectional effects: high-affinity potentiation and low-affinity inhibition. For Zn2+, both potentiation and inhibition of GlyRs involve binding to the extracellular ligand-binding domain (LBD) which, like those of the other Cys-loop receptors, is homologous to molluscan acetylcholine binding proteins (AChBPs) (Brejc et al. 2001; Celie et al. 2004). Site-directed mutagenesis and homology modelling based on the structure of AChBP have allowed assigning residues of the α1 GlyR implicated in glycine and strychnine binding to the agonist-binding pocket located between subunit interfaces (Grudzinska et al. 2005). This binding pocket is positionally conserved in the Cys-loop receptor family and formed by distinct extracellular segments of the LBD [loops A, B and C on the (+), and loops D, E and F on the (−), sides of the binding interface; see Changeux and Taly 2008]. Notably, the GlyR residues mediating Zn2+ potentiation and inhibition also are located within this agonist-binding site (Nevin et al. 2003; Grudzinska et al. 2008). In contrast, inhaled anaesthetics are thought to bind in a cavity located between the transmembrane domains of the receptor (Beckstead et al. 2002).

Amongst the GlyR modulators investigated so far, tropeines have gained particular attention. Originally identified as 5HT3 receptor antagonists, these compounds were found to potentiate GlyR currents at low, non-saturating agonist concentrations (Chesnoy-Marchais 1996). Subsequently, amides and esters of 3α-hydroxy-tropane have been shown to display very high affinity to GlyRs (Maksay et al. 1998, 2004). Nortropeines like 3α-(3′-methoxy-benzoyloxy)nortropane (MBN, Fig. 1a), nor-O-zatosetron and nortropisetron displace [3H]strychnine binding to native rat striatal (Maksay et al. 2004) and recombinant human α1 (Maksay et al. 2008) GlyRs at nanomolar concentrations. Moreover, tropisetron (Fig. 1a) has been shown to potentiate α1 GlyR currents in the femtomolar concentration range (Yang et al. 2007). Thus, some tropeines are excellent in displaying both bidirectional modulatory properties and high-affinity binding.

Figure 1.

 3α-(3′-methoxy-benzoyloxy)nortropane (MBN) inhibition of wt and mutant α1 GlyRs. (a) Chemical structures of MBN and tropisetron. (b) Representative current traces reveal a reduced inhibition of the R131A and K200A GlyRs by 10 μM MBN. The scale bars correspond to 30 s and 2 μA. (c) Concentration dependence of MBN inhibition for wild type (wt) (), F63A (), R65A (•), R131A (□) and K200A (○) GlyRs. Data are mean ± SEM from 5 to 10 oocytes.

Different lines of evidence indicate that tropeines bind close to or within the agonist/antagonist binding sites of GlyRs. First, Yang et al. (2007) identified residue N102 located close to the subunit interfaces as a major determinant of tropisetron inhibition of the α1 GlyR. Second, binding of the structurally related 5HT3 receptor antagonist granisetron is thought to occur within the agonist-binding cleft of the 5HT3 receptor (Joshi et al. 2006), and the pharmacophores proposed for ‘setron-type’ 5-HT3 receptor antagonists and GlyR modulatory tropeines display considerable resemblance (Maksay et al. 2004). Here, we examined the effects of binding site mutations in the α1 GlyR on inhibition and potentiation by MBN. Our data are consistent with adjacent binding sites and distinct binding modes mediating tropeine potentiation and inhibition.

Materials and methods

Materials

Taurine, collagenase and urethane were obtained from Sigma Chemical Co. (St Louis, MO, USA). MBN was synthesized by Professor Péter Nemes (Szent István University, Budapest) as described (Maksay et al. 2004). The point mutations of the human α1 subunit have been described previously (Grudzinska et al. 2005). Linearized plasmid DNAs of the wild type (wt) and mutant GlyR α1 subunit were used for the in vitro synthesis of cRNA (mCAP mRNA Capping Kit; Stratagene, La Jolla, CA, USA) using T7 RNA polymerase as detailed previously (Kondratskaya et al. 2005). cRNA concentrations were determined by both measuring the optical density at 260 nm and comparing methylene blue staining intensities after gel electrophoresis.

Electrophysiology

Xenopus laevis oocytes were obtained, and voltage-clamp electrophysiology was performed as described previously (Maksay et al. 1999; Grudzinska et al. 2005). Animals were handled according to German law (licence no. V 54–19c20/15-F126/13; Regierungspräsidium Darmstadt). Agonist dose–response curves were recorded from cRNA injected oocytes under voltage-clamp conditions and fitted via GraphPad Prism 4 (GraphPad Software Inc., San Diego, CA, USA) as detailed previously (Grudzinska et al. 2005, 2008). Solution exchange was mediated by the BPS drug application system (Adams and List, Westbury, NY, USA) with a flow rate of 0.5 mL/s resulting in a half-exchange time of < 200 ms. Glycine and taurine were co-applied with MBN with or without 30 s of pre-treatment. Inhibitory concentrations of MBN and glycine concentrations corresponding to its half-maximal effect (EC50) were used to determine half-maximal inhibitory concentrations (IC50). Potentiation by MBN was examined with taurine at 10% of the maximal effect (EC10). Data represent mean ± SEM and were analysed using the KaleidaGraph program (Synergy Software, Reading, PA, USA). Statistical significance was determined by Student’s t-test and considered to be significant at < 0.01.

Receptor modelling, MBN docking and molecular dynamics simulations

Homology models of the LBDs of α1 GlyRs were generated by using the crystal structures of AChBP (Protein Data Bank code: 1I9B) and nicotinic acetylcholine receptor (2BG9) as templates. The sequences of Lymnaea stagnalis AChBP, nicotinic acetylcholine receptor and the LBD of human α1 GlyRs were aligned by Clustal-W and adjusted on the basis of multisequence alignments of Cys-loop receptors (Thompson et al. 1994). The program Composer included in the Sybyl 7.0 package (Blundell et al. 1988) was used to generate homology models of α1 GlyR-LBDs. AChBP (1I9B) was used as a template for the closed α1 GlyR-LBD model when co-crystallized HEPES molecules were replaced for MBN in the binding interfaces. The GlyR model based on nicotinic acetylcholine receptors (2BG9) was used as a reference structure. The resulting models were refined within Sybyl by alternating short (10–100 fs) molecular dynamics (MD) simulations and energy minimization steps in vacuo (Tripos force field, Gasteiger-Hückel charges). MBN was docked using Dock (Sybyl module) and FlexX (Tripos, BioSolveIT, GmbH, Sankt Augustin, Germany) respectively (Rarey et al. 1996). The protein-ligand complexes with lowest potential energies were then used as starting structures for MD simulations.

Pentameric α1 GlyR-LBDs with five docked MBN molecules were subjected to MD simulations by GROMACS (Berendsen et al. 1995) by using the following parameters: free MD with periodic boundary conditions, gmx force field, isobaric–isothermal ensemble, explicit solvent (spc water model), particle-mesh Ewald electrostatics, Berendsen temperature and pressure coupling, isotropic. The ligand topology of MBN was generated by using the Dundee PRODRG2 web server (Van Aalten et al. 1996). To obtain the most important protein-ligand interactions during a MD simulation, the respective parts of the trajectory exhibiting the lowest interaction energies of MBN with α1 GlyR LBDs were averaged and energy-minimized.

Results

To identify determinants of MBN inhibition and potentiation, we injected X. laevis oocytes with human wt or mutant α1 GlyR cRNAs. One to 2 days after injection, the oocytes responded to saturating glycine concentrations with chloride currents of 10–50 μA. GlyR currents could also be elicited by the partial agonist taurine (Schmieden et al. 1992). Taurine responses were used to investigate MBN potentiation, as (i) the efficacy of taurine can indicate the effect of mutations on coupling efficiency between agonist binding and activation; and (ii) allosteric modulation of native GlyRs in spinal cord membranes has been more efficacious for partial rather than full agonists (Bíró and Maksay 2004).

Effects of binding site mutations on MBN inhibition

The inhibitory potencies of MBN were examined at previously characterized mutant α1 GlyRs that carry point mutations within the LBD by recording glycine currents at the EC50 concentrations previously determined for the respective mutant receptors (Grudzinska et al. 2005). MBN alone did not elicit chloride currents. MBN was either pre-incubated for 30 s or coincubated with glycine. Pre-incubation with subnanomolar MBN (0.2 and 0.5 nM, up to 10 min) did not alter the responses of glycine and did not significantly affect the IC50 value of coincubated MBN. Therefore coincubation was applied subsequently. Glycine responses of the wt α1 GlyR were inhibited by MBN with an IC50 value of 1.5 μM (Fig. 1 and Table 1). The substitution F63A localized in loop D abolished inhibition by MBN while two other loop D mutations, R65A and Q67A, had no significant effect on inhibitory potency (Fig. 1 and Table 1). Mutations N46C, N102A (loop A), K200A, Y202L and F207A (all in loop C) eliminated the potency of MBN to inhibit glycine currents, whereas R131A (β strand/loop E) and E157C (loop B) increased its IC50 by about 20-fold (Table 1). Substitutions in loop E: S129A and R119A had no significant effect. Surprisingly, no inhibition was observed for R119K, a conservative mutation. In conclusion, most of the binding site mutations analysed either reduced or abolished inhibition by MBN.

Table 1.  Inhibitory potencies of and potentiation of wt and mutant α1 GlyRs by MBN
cRNA injectedInhibition by MBNa (IC50) (μM)Imax, tau/ Imax gly [FRACTION] × 100Potentiation by MBNb
2 nM (%)2 μM (%)
  1. Data are mean ± SEM from n (indicated in brackets) oocytes. Efficacies were determined with 4–6 oocytes.

  2. GlyRs, glycine receptors; MBN, 3α-(3′-methoxy-benzoyloxy)nortropane; IC50, half-maximal inhibitory concentrations.

  3. aDetermined at the respective EC50 values of glycine.

  4. bEnhancement by MBN of α1 GlyR currents elicited by the respective EC10 concentrations of taurine.

  5. cTaken from Grudzinska et al. (2005).

  6. dLow taurine activity. Y202A was not functional.

  7. *Significantly different (< 0.01) from wt α1 GlyRs (unpaired Student’s t-test).

  8. **Current enhancement by MBN was significant (< 0.001) over control (unpaired Student’s t-test).

wt1.5 ± 0.3 (8)25 ± 9c121 ± 2** (8)108 ± 3 (7)
N46C≫100* (4)50 ± 4*158 ± 11** (4)127 ± 7 (5)
F63A≫100* (6)20 ± 5c141 ± 11** (16)124 ± 8 (12)
R65A1.4 ± 0.3 (10)83 ± 19*,c121 ± 4** (8)194 ± 13* (6)
Q67A0.8 ± 0.2 (5)34 ± 5102 ± 4 (5)83 ± 6* (5)
N102A≫100* (3)26 ± 3176 ± 9** (4)246 ± 34* (4)
R119A2.5 ± 1.1 (6)7 ± 1*,c91 ± 4 (6)<100
R119K≫100* (5)37 ± 15142 ± 14 (6)423 ± 52* (4)
S129A3.4 ± 1.2 (5)51 ± 6110 ± 2 (5)54 ± 15* (4)
R131A30 ± 4* (5)65 ± 10*,c123 ± 3** (12)199 ± 21* (8)
E157C23 ± 12* (4)21 ± 6129 ± 8** (4)187 ± 29* (3)
K200A≫100* (5)17 ± 9182 ± 23** (6)298 ± 91* (3)
Y202Ld≫100* (6)23 ± 10114 ± 2** (6)148 ± 29 (4)
F207A≫100* (7)37 ± 1294 ± 3 (7)212 ± 35* (3)

Effect of binding site mutations on potentiation by MBN

As potentiation by MBN was studied in combination with activation by taurine, the relative efficacy of this partial agonist was compared to that of the full agonist glycine. As it has been shown previously for some of these binding site mutations (Grudzinska et al. 2005) they affect the efficacy of taurine. Efficacy varied between 83% for R65A and 7% for R119A α1 GlyRs (Table 1).

In another set of experiments, the chloride currents of the wt and mutant α1 GlyRs elicited by taurine at the respective EC10 concentration were recorded following pre-incubation of the oocytes with MBN. The representative current traces shown in Fig. 2(a) illustrate that nanomolar MBN produced small but significant potentiation of the wt α1 GlyR, while at 2 μM MBN potentiation was obscured by the inhibitory effects of the compound. In contrast, the R131A α1 GlyR showed highly significant potentiation at 2 μM MBN, which was much larger than the potentiation of the wt GlyR seen at 2 nM MBN. Apparently, reduced inhibition resulting from the R131A substitution (see Fig. 1c) permitted potentiation at micromolar MBN concentrations (see Fig. 2b). Table 1 shows that the point mutations N46C, F63A, R119K, R131A, E157C and K200A which all eliminate or reduce inhibition generated GlyRs that were significantly potentiated by MBN. With the inhibition-deficient mutants: N102A, R119K and K200A, MBN potentiation was even enhanced (Table 1). In contrast, mutations Q67A, R119A and S129A eliminated MBN potentiation (Table 1). Most of the mutations which did not affect potentiation by 2 nM MBN (Table 1) showed enhanced potentiation at micromolar MBN as demonstrated for the R131A GlyR in Fig. 2b.

Figure 2.

 Inhibition and potentiation of the wt and R131A GlyRs by MBN. (a) Representative current traces. Oocytes were pre-incubated with the indicated concentrations of MBN for 30 s followed by the application of taurine at the respective EC10 concentration. (b) Concentration dependence of MBN modulation of the wt and R131A GlyRs represented by black and grey columns respectively. Data are mean ± SEM from 5 to 10 oocytes. Note potentiation of the R131A receptor by MBN concentrations that inhibit wt α1 GlyRs. ***Significant enhancement (p < 0.01).

MD simulations suggest different binding modes for inhibition and potentiation

The effects of the binding site mutations described above indicate that MBN interacts with the α1 GlyR-LBD in the agonist-binding site formed between subunit interfaces. We therefore used a structural model of the pentameric extracellular domain of the α1 GlyR to dock MBN into the LBDs formed at the interfaces of wt and mutant α1 GlyR subunits. The docked structures were subsequently subjected to MD simulations. We paid primary attention to the rearrangements of loop C in concert with MBN. The MD simulations revealed that MBN might adopt two different positions in the agonist-binding cavity of the α1 GlyR. One position of MBN accompanied with permanent outward pushing of loop C to half-open conformation was associated with inhibition, while another position with inward motion (closure) of loop C was associated with potentiation. Notably, these MD simulations then led to MBN positions with the lowest energy.

3α-(3′-methoxy-benzoyloxy)nortropane was docked in the closed α1 GlyR-LBD model. The most stable docking position of MBN predicted from 20 ns MD simulation resulted in a ligand-LBD complex (Fig. 3, top left) with a binding energy of −258 kJ/mol. Here, MBN penetrated deeply in the interface cavity below the lower (cellular) side of loop C. It was in close contact with amino acid side chains whose mutations affected MBN inhibition, therefore we consider it ‘inhibitory’ binding. Accordingly, MBN could interact with residues N46, N102, F159 and, like strychnine (Grudzinska et al. 2005), with F63 and R131 (Fig. 3, top left). In this docking position, MBN is distant from residue R65, in agreement with the inability of mutation R65A to affect inhibition. The nortropane head is nested in a hydrophilic vestibule. Its secondary amino group is in hydrogen-bonding distance to the asparagines N102 in loop A and N46, while the carbonyloxy moiety of MBN seems to form strong hydrogen bonds with R131 in loop E. Notably, in this docking orientation the aromatic ring of MBN sits in a cleft between the aromatic rings of F63 and F159 intercalated as a wedge and suitable for π–π interactions. MD simulations suggest that the conformation of loop C is stabilized by interactions within the principal subunit: E157 with K200 and Y202 as well as π-π interactions between Y202 and F207. This docking position of MBN seems to prevent interface interactions and hinder the full closure of loop C that would elicit channel gating. This half-shut conformation lies in-between the agonist-bound (closed) and non-liganded (open) conformations. The superimposed structures shown in Fig. 3 (left bottom) also indicate the conformation-specific distances between the respective Cα atom positions of N203 at the tip of loop C.

Figure 3.

 Docking modes of MBN (ball and stick representation) in the LBD of α1 GlyRs. The central panel shows LBDs of two adjacent α1 GlyR subunits in green and blue with MBN docked in two positions correlated with inhibition and potentiation respectively. Top, left and right enlargements show the side chain interactions involved in MBN binding for the presumptive inhibitory and potentiating binding modes respectively. The potentiating docking mode of MBN is ‘above’ the inhibitory one by about 1 nm. It also differs by twofold rotations of about 180° around the y and z axes respectively. Bottom, left and right panels illustrate loop C positions (in green) for the two modes of MBN docking following MD simulations (top view). Loops C overlaid in white refer to the MD starting points. MD simulations of inhibition and potentiation started from closed and half-shut LBDs respectively. The bottom left panel demonstrates that MD simulation with MBN docked in the inhibitory orientation opens loop C half-way as compared with the non-liganded open (reference) loop C position overlaid in black.

To further explore the role of N102 in MBN binding and inhibition of α1 GlyRs, we also modelled the mutated N102A GlyR-LBD in order to determine whether mutation N102A not only impairs inhibition but also the docking stability of MBN. MD simulations were performed with MBN-docked pentameric α1 wt and N102A GlyR-LBDs under identical conditions. The compactness of a cylindrical pentamer can be characterized by its radius of gyration (Rg) of the backbone atoms. As the core structure of the pentamer was very stable, changes in Rg should mainly reflect conformational flexibility of loop C and the size of the binding cavity underneath it. With MBN docked below loop C and the NH moiety of MBN pointing to N102, the Rg value of the wt α1 GlyR increased from 2.90 to 2.92 nm, while the Rg of the N102A mutant declined substantially below 2.86 nm within 20 ns MD simulation (not shown). The Rgs were subsequently dissected into those of core pentamers (without loops C) and of loops C themselves. These MD simulations are illustrated in Fig. S1, where the time-dependent changes of the Rg values of loops C alone are plotted. The substantial difference of about 0.2 nm in Rg between loops C of wt α1 and N102A GlyRs reflects different loop C positions, with MBN being docked stably under the half-shut loop C of the wt receptor, whilst in N102A GlyRs the distorted interface cavity might collapse. We also measured the distance of the nortropane N atom of docked MBN from Cα atoms of N102 and of N102A following 20 ns MD simulation. The average distance of MBN was about 0.2 nm greater from N102A than from N102 (not shown). This indicates that MBN moved away from the mutated N102A parallel with the collapse of the binding cavity during MD simulation while it was tightly bound to N102. These findings support the suitability of MD simulations and underline the crucial role of N102 in MBN binding and inhibition of α1 GlyRs.

Our MD simulations also disclosed a second, even more stable docking position, whose ligand interactions can be reconciled with the mutations affecting MBN potentiation (Fig. 3, top right and centre). This position was obtained after docking MBN into a half-shut LBD structure, followed by MD simulation for 20 ns. In the course of MD simulation, MBN rotated clockwise by about 30° (seen from outside), relocated towards the centre of the LBD by about 0.4 nm and apposed to the inner surface of loop C. In this docking orientation with a calculated binding energy of −288 kJ/mol, the aromatic ring of MBN was buried in a hydrophobic cleft formed by residues F63, F159, Y202 and F207, while the nortropane NH group was close to residues Q67, R119 and S129 (Fig. 3, top right). If R119 is replaced for K, the Lys side chain which is more flexible than that of Arg might accommodate MBN at the potentiating site preferentially over the inhibitory site. This might explain the preferential potentiation of R119K α1 GlyRs.

In the ‘potentiating’ MBN-bound structure loop C adopted a closed conformation that appears to be stabilized by interactions with both MBN and side chains of the complementary (-)face of the binding site (data not shown). The resulting structure (Fig. 3, bottom right) strongly resembles the agonist-liganded LBD, and hence might facilitate channel opening.

Discussion

In this study, mutations within the agonist-binding site of the α1 GlyR are shown to affect potentiation and/or inhibition of the recombinant homomeric receptor by a representative tropeine, MBN. This result strongly supports the view that the biphasic modulatory effects of tropeines are mediated at the agonist-binding cavity localized at subunit interfaces, with low and high agonist and modulator occupancies resulting in potentiation and inhibition respectively.

Determinants of MBN inhibition and potentiation

Several GlyR α1 substitutions previously shown to increase the EC50 of glycine and other agonists were found here to affect MBN inhibition of glycine responses. Notably, these substitutions were localized in four of the six subdomains which form the agonist-binding site between adjacent subunits of the homomeric α1 receptor. Specifically, the mutations N102A in loop A; E157C in loop B; K200A, Y202L and F207A in loop C; R119K and R131A in β strand E respectively, all strongly reduced or eliminated MBN inhibition. The N102A substitution has been found to abolish inhibition of glycine responses by the closely related tropeine tropisetron (Yang et al. 2007). However, in that study other mutations at the ligand-binding site had no effect on tropisetron inhibition. Notably, substitutions of residues E157, Y202 and F207 have been shown to also impair the inhibition of α1 GlyR currents by Zn2+and Cu2+ (Grudzinska et al. 2008; Schumann et al. 2008). Thus, (partly) common binding site residues appear to be implicated in the inhibition of glycine responses by both tropeines and divalent metal ions.

Only few of the mutations that did not affect inhibition impaired potentiation by MBN. Specifically, substitutions Q67A, R119A and S129A eliminated potentiation by MBN. In contrast, several mutations which abolished or attenuated inhibition: N46C, F63A, N102A, R119K, R131A, E157C and K200A all maintained or enhanced GlyR potentiation by MBN. Some of these mutations which attenuated MBN inhibition (N102A, R131A, E157C and K200A) showed enhanced potentiation at micromolar MBN, consistent with impaired inhibition unmasking GlyR potentiation at higher modulator occupancy. Substitutions of residue F207 have also been found to affect inhibition and, importantly, potentiation by divalent metal ions (Grudzinska et al. 2008; Schumann et al. 2008). The side chain of F207 thus seems to constitute a crucial determinant of modulator action. Interestingly, R119 appears to function as a switch between inhibition and potentiation by MBN depending on its replacement. The R119A α1 GlyR was invariably inhibited by MBN while it was activated by taurine with very low efficacy which could not be potentiated by MBN. In contrast, the R119K α1 GlyR was not inhibited by MBN while its activation by taurine was potentiated by MBN to the greatest extent.

No linear correlation can be observed between the efficacy of taurine and potentiation by 2 nM MBN (r= 0.008) and 2 μM MBN (r= 0.0003). This suggests that the mutations did not affect MBN potentiation indirectly by modifying agonist binding.

Different docking positions of MBN predicted by MD simulation

The effects of our point mutations on MBN potentiation and inhibition indicate that both binding sites of tropeines are located at the subunit interface cavities of the α1 GlyR. The potentiating binding of MBN appears to overlap with glycine binding while inhibition requires MBN binding next to it as shown in Fig. 3. Our docking studies and MD simulations strongly suggest that MBN can adopt different positions and that these different docking modes are associated with distinct effects, i.e. potentiation and inhibition of GlyR currents. As the binding cavities of ionotropic Cys-loop and glutamate receptors are closed upon binding of agonists whereas antagonists prevent, or allow only partial closure (Furukawa and Gouaux 2003; Hansen et al. 2005), we propose that comparable differences in binding site conformations also underlie MBN potentiation and inhibition respectively. Moreover, the different binding energies and C loop orientations predicted for the two MBN binding modes can be attributed to the different efficacies of MBN: the higher binding energy of MBN docking with a closed loop C position can be associated with high-affinity potentiation, whilst the docking mode of lower binding energy with low-affinity inhibition.

It is informative to compare MBN binding with the binding of agonists and antagonists to GlyRs and other Cys-loop receptors. Glycine binding and activation of α1 GlyRs requires cation–π interaction of the amino group of glycine with residue F159 (Pless et al. 2008) and ionic interaction of its carboxylate group with R65 (Grudzinska et al. 2005). Inhibition of α1 GlyRs by tropisetron is abolished by mutations of N102, and it has been proposed that the tropane head interacts with residue N102 (Yang et al. 2007). Tropeine antagonism of 5-HT3A receptors is attenuated upon mutation of W89 (Yan et al. 1999), a residue homologous to F63 in α1 GlyRs. Granisetron binding to 5-HT3A receptors could be localized under both arms of loop C (Joshi et al. 2006). These findings support the existence of binding cavities of tropeines on GlyRs and 5-HT3 receptors in common with those of their agonists.

The conformational changes underlying GlyR potentiation are not fully understood. It is generally assumed that multiple subunit interface sites must be occupied in order to cause efficient channel opening of Cys-loop receptors such as of GABAA receptors via binding of GABA and benzodiazepines (Hanson and Czajkowski 2008). Potentiating binding of tropeines apparently occurs in the glycine-binding cavities, in the GlyRs interface. Closure of loops C around glycine bound at one interface, and around tropeines bound at another interface, may lead to allosteric facilitation of binding and channel opening, as known for GABA and benzodiazepines at GABAA receptors (Hanson and Czajkowski 2008). At high tropeine occupancy inhibition will predominate over potentiation. Elucidation of the distinct binding modes of modulators will hopefully lead to therapeutic agents with preferential potentiation over inhibition of Cys-loop receptors.

Acknowledgements

We thank Professor Péter Nemes (Szent István University, Budapest) for the synthesis of MBN and Tanja Schumann for experimental contributions. This study was supported by grants K-62203 (OTKA) and D-32 (Hungarian-German Intergovernmental S & T Cooperation Programme) to GM, Gemeinnützige Hertie-Stiftung to BL, Max-Planck-Gesellschaft, Deutsche Forschungsgemeinschaft (EXC 115) and Fonds der Chemischen Industrie to HB.

Ancillary