INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION CONCLUSIONS REFERENCES

The inhibitory GlyR and GABA_A (GABAR) belong to the family of LGICs, which also comprise excitatory nAChR and 5-hydroxytryptamine (5-HT3) receptors, as well as the inhibitory glutamate receptors (Betz, 1990; Ortells and Lunt, 1995; Cleland, 1996). LGICs are pentameric proteins, each subunit having a large extracellular domain at the N-end, four transmembrane segments (M1-M4), and an intracellular (M3-M4) loop. Although the folding of LGIC proteins remains unknown, there is a consensus that five M2s, predominantly in the -helical conformation, contribute to the central pore (Changeux et al., 1992; Galzi and Changeux, 1995; Karlin and Akabas, 1995). M2s are, probably, kinked in their middle part (Unwin, 1995). The five-helical bundle of nAChR is believed to have a funnel-like shape with a narrow cytoplasmic end and several rings of homologous residues facing the pore (Changeux et al., 1992). The residues in the rings were demonstrated to govern ion conductance (Imoto et al., 1988), ion selectivity (Galzi et al., 1992), gating (Labarca et al., 1995), desensitization (Revah et al., 1991), and pharmacological properties (reviewed by Arias, 1998) of LGICs.

The central pore of LGICs is the only functional domain for which structure-function relationships are relatively well understood due to numerous electrophysiological, pharmacological, mutagenesis, and molecular modeling studies. Data on permeability of organic cations via nAChRs suggest the minimal profile of the open pore as a square of 6.5 × 6.5 Å (Dwyer et al., 1980) or a circle of 7.6 Å diameter (Nutter and Adams, 1995). Analogous approaches predict the minimal circular profiles of the open pore to have a diameter of 5.2-5.4 Å in GlyR (Bormann et al., 1987; Rundström et al., 1994), 5.6 Å in GABAR (Bormann et al., 1987), and 7.6 Å in the 5-HT3 receptor (Yang, 1990). The analysis of the conformation-activity relationships of noncompetitive pentamethylenebisammonium antagonists of nAChR (Zhorov et al., 1991; Brovtsyna et al., 1996) predicted the pore profiles at the levels of the two ammonium groups as rectangles of 6.1 × 8.3 Å and 5.5 × 6.4 Å. However, the low-resolution electron cryomicroscopy images of the nicotinic acetylcholine receptor in the open state imply a wider central pore (Unwin, 1995). The controversial data on the dimensions of the pore are reflected in recent structural models of the nAChR with M2s either tightly packed at the cytoplasmic half (Ortells and Lunt, 1996; Tikhonov and Zhorov, 1998) or distant from each other (Sankararamakrishnan et al., 1996; Adcock et al., 1998).

A knowledge of the pore architecture in LGICs is necessary for understanding functions of the pore-lining residues and mechanisms of the channel block by noncompetitive antagonists. A touchstone for structural models of the pore is their ability to explain structure-activity relationships of ligands. The pore region of nAChRs comprises binding sites for numerous noncompetitive antagonists (Arias, 1998). In contrast, little is known about the mechanisms of the block of ligand-gated Cl channels. CTB (see Fig. 1) inhibits GlyRs by a noncompetitive, channel-blocking mechanism (Rundström et al., 1994). PTX (Fig. 1) is the antagonist of GlyRs (Pribilla et al., 1992; Lynch et al., 1995), GABARs (Hosie et al., 1997; Zhang et al., 1995a), as well as of the inhibitory glutamate receptor (Cleland, 1996). All these receptors activate Cl-selective channels.

View larger version (13K): [in this window] [in a new window]   FIGURE 1   Molecular structures of blockers of ligand-gated Cl channels.

The interpretations of experimental results on PTX action are obscured by the fact that, unlike the case of the negatively charged CTB, which is expected to enter a Cl channel, the energy contributions that stabilize the electrically neutral PTX in the anionic pore remain unknown. Two main mechanisms of PTX action have been proposed. Pribilla et al. (1992) interpreted results of their mutagenesis, electrophysiological, and ligand-binding studies of GlyRs in favor of a noncompetitive mechanism of PTX block. Using a similar approach, Lynch et al. (1995) suggested that this compound is an allosterically acting competitive antagonist of GlyRs. Action of PTX on GABAR channels was also interpreted as the allosteric stabilization of the closed (desensitized) state of the receptor (Newland and Cull-Candy, 1992). However, mutations in the intracellular half of M2 significantly affect the sensitivity of GABARs to PTX (Wang et al., 1995; Zhang et al., 1995a; Curley et al., 1995), the facts compatible with the noncompetitive mechanism of the block. Coexistence of both competitive and noncompetitive mechanisms of the PTX inhibitory action in GABARs has been suggested (Yoon et al., 1993; Qian and Dowling, 1994; Wang et al., 1995).

The blocking activity of CTB and PTX is highly sensitive to the subunit composition of GlyRs and GABARs. Thus, micromolar concentrations of CTB block ₁ and ₁/ but not ₂ GlyRs (Rundström et al., 1994). PTX blocks homomeric ₁ GlyR and ₁ GABA_CR but weakly antagonizes heteromeric ₁/ GlyR and ₁/₂ GABAR (Pribilla et al., 1992; Zhang et al., 1995a). M2s of ₁ and ₂ GlyRs differ only by one residue (Gly^2' in ₁ GlyR and Ala^2' in ₂ GlyR, see Table 1). Mutation of ₁ Gly^2' to Ala makes this receptor insensitive to CTB, suggesting that the ligand binds at the cytoplasmic half of the pore (Rundström et al., 1994).

In the absence of an experimental atomic-scale resolution structure of LGICs, molecular modeling may help suggest mechanisms by which the pore-facing residues affect the functional properties of the channels. Tikhonov and Zhorov (1998) proposed a structural model of nAChR that accommodates various noncompetitive agonists. The model comprises the kinked -helical M2s that form three regions along the pore: the funnel-like extracellular part, the flexible kinked region comprising conserved Leu^9' residues, and a narrow cytoplasmic part where parallel helices may form close contacts with blockers.

In the present study we use this model as a template to built homology models of several GlyRs and GABARs. By using the MCM protocol we have calculated the energy profiles for CTB and PTX in the pore region of the receptors and predicted that these compounds have energetically optimal binding sites in the intracellular half of the pore. We further suggest that the low affinity of CTB and PTX for certain heteromeric receptors is caused by unfavorable contacts between these drugs and the pore-facing residues. These results support the narrow-pore model of LGICs and may help design new mutagenesis experiments and new blockers of Cl channels.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION CONCLUSIONS REFERENCES

Designating rings of the pore-facing residues

In nAChRs, the rings of pore-facing conserved residues are named after these residues, e.g., threonine ring (see Galzi and Changeux, 1995). However, in other LGICs, aligned positions do not necessarily have the same residues. For example, GlyRs have Gly and Pro residues in the positions aligned with the threonine ring of nAChRs (see Table 1). This causes a problem in naming the rings in the superfamily of LGICs. In an attempt to solve this problem, we designate the rings by the symbol "R" with a superscript referring to the position of the corresponding residue according to the numbering scheme used by Lester (1992). Thus, R^1', R^2', R^6', R^9', R^13', and R^16' denote, respectively, the rings aligned with the intermediate, threonine, serine, equatorial, valine, and outer leucine rings in nAChRs (Table 1 and Fig. 2).

View larger version (107K): [in this window] [in a new window]   FIGURE 2   A schematic view of the rings of the pore-facing residues in ligand-gated ion channels. Planes represent the pore levels near the rings. Relative dimensions of the rings and distances between the planes at the scheme do not exactly match the computational models. Amino acid sequence of the pore-forming segment M2 of 1 GlyR is shown as a column that should be read from the bottom to the top.

Structure of blockers

CTB is a permanently charged anion, the negative charge evidently promoting binding in the anion-selective pore. The blocker has a star-like shape with three phenyl rings extending from the central boron atom and a CN group capable of accepting H-bonds (Fig. 1). Repulsion between Ph rings hinders rotation around the B-Ph bonds, making CTB a relatively rigid molecule in which only limited conformational changes are possible.

A plant alkaloid, picrotoxin, consists of two compounds, PTX and PTN, which differ only by one group extending from a six-membered ring (Fig. 1). PTX is more potent than PTN in antagonizing effects of GABA (Jarboe et al., 1968; Curtis and Johnston, 1974) and in inhibiting GABA-gated Cl channels (Anthony et al., 1994; Shirai et al., 1995). Unlike the negatively charged CTB, PTX is a neutral molecule lacking ionizable groups. PTX has a rigid pentacyclic core with methyl, hydroxyl, and isopropenyl groups capable of rotating around the single bonds attaching them to the core. This compound has an egg-like shape with the isopropenyl group at the elongated end. The rounded end of PTX is trimmed by three ether oxygens and two carbonyl oxygens that may accept but not donate H-bonds. A hydroxyl group decorates the hydrophobic side surface of PTX.

General features of the models

In the present homology modeling study we have used as a template the predicted five-helix-bundle structure of nAChR (Tikhonov and Zhorov, 1998). We did not try to improve the spatial disposition of M2s, their backbone geometry and conformations of the flexible residues extending outside the pore. The borders of the M2s were specified as in the template. The sequences of the M2s are shown in Table 1. Heteromeric models were arranged in the (₁)₃/()₂ stoichiometry (Langosch et al., 1988). The overall topography of the template was preserved by restraining C atoms with the help of pins. A pin is a flat-bottom parabolic penalty function that increases with deviation of a C atom by >1 Å from the position specified in the template. A force constant of 10 kcal · mol¹ · Å¹ was used for the pins.

Software and force fields

All calculations have been carried out using the ZMM program described elsewhere (Zhorov, 1981; Zhorov and Ananthanarayanan, 1996). The program allows conformational searches in the space of arbitrarily specified internal coordinates such as torsion and bond angles, positions of free molecules (Cartesian coordinates of their root atoms), and the orientation of the molecules (Euler angles of the local systems of coordinates centered at the root atoms). Atom-atom interactions were calculated using the AMBER force field (Weiner et al., 1984) with a cutoff distance of 8 Å and a shifting function (Brooks et al., 1985). The energy components most sensitive to the chemical structure of a blocker, its conformation and position in the pore, are van der Waals interactions, hydrogen bonds, and electrostatic interactions involving polar groups. Therefore, we have ignored solvating effects and used Coulomb law for the calculations of the electrostatic energy. Since the charges at the ionizable residues should be compensated by counterions whose location is unknown, we considered all the ionized residues in their neutral forms, as was proposed by Momany et al. (1975). The standard atomic charges of amino acids with a distance-dependent dielectric parameter were used (Weiner et al., 1984). The partial charges at the atoms of blockers were calculated with the MOPAC software (Biosym Inc., San Diego, CA, USA). In most of the computational experiments, the CH, CH₂, and CH₃ groups of the receptor and CH groups of CTB were represented by united atoms specified in the AMBER force field. Since boron is not specified in the AMBER force field, its van der Waals interactions were calculated using parameters for a C_sp3 atom, while the length of the B-C bond was assigned 1.57 Å based on the respective covalent radii of 0.80 and 077 Å. Quantum-chemical calculations of CTB yielded charges of 0.13, 0.05, and 0.29 proton charge units, respectively, for B, C, and N atoms in the B-CN group, a charge of 0.63 units being distributed over three phenyl rings to yield a net CTB charge of 1 unit.

MCM protocol

For the search of optimal conformations the MCM protocol (Li and Scheraga, 1988) was used as described elsewhere (Zhorov and Ananthanarayanan, 1996). Trajectories were calculated at T = 600 K. A subsequent starting point in a trajectory was obtained by changing a randomly selected internal coordinate of the preceding point by a random increment. In the MCM calculations of ligand-free receptors, all internal coordinates were sampled. In the MCM calculations of the energy profiles, a list of internal coordinates to be sampled was formed and then updated periodically. It included generalized coordinates of the blocker and torsions of the receptor residues around the blocker (see below). From a given starting point, the energy was minimized until the norm of the energy gradient become <1 kcal · mol¹ · rad¹ or a limit of 200 calls to the procedure of the gradient calculation was exceeded. The resulting MEC was accepted in the trajectory if its energy E was less than that of the preceding point of the trajectory E_p or if a random number n  (0, 1) was <exp((EE_p)/RT). The obtained MEC was added to an internal stack (array) of MECs accumulated during the search if its energy did not exceed 7 kcal/mol above the lowest-energy structure in the stack and if it was geometrically distinct from any other MEC accumulated in the stack (Zhorov and Ananthanarayanan, 1993). Two MECs were considered distinct if they had a different backbone code (Zimmerman et al., 1977) or a side-chain torsion angle different by at least 10°. For the molecular systems under consideration, the internal stack had a capacity of 40 MECs. If and when the internal stack overflowed, its content was delivered to the external stack (file). After this procedure, redundant MECs and those with an energy above 7 kcal/mol from the lowest-energy MEC were removed from the internal and external stacks. A trajectory was terminated if 500 consecutive energy minimizations did not decrease the energy of the best MEC found nor added a new MEC to the internal stack. When calculating energy profiles, only the lowest-energy MEC found in each MCM trajectory was preserved for further analysis. These trajectories were usually terminated if 500 consecutive energy minimizations did not decrease the energy of the best MEC, but a large number of different MECs repeatedly overflowed the internal stack.

Positions 0', 11', 14', and 19' in M2s of GlyRs and GABARs comprise long-chain residues which are either at the borders of M2 or do not face the pore. Test MCM trajectories with all degrees of freedom sampled with equal probability converged slowly and yielded large external stacks of MECs with different orientations of the above residues. To speed up the convergence of standard trajectories, no MC sampling was performed for positions and orientations of M2s, all backbone torsions, and the side-chain torsions in positions 0', 11', 14', and 19'. However, all the generalized coordinates were varied in energy minimizations.

MC-minimized energy profiles

A special procedure was elaborated to pull a ligand through the pore. The pore axis is oriented along the z axis of the Cartesian coordinate system with ring R^16' close to the xy plane; other rings having negative z coordinates (Fig. 2). A position of a ligand along the pore is specified by z_R, the z coordinate of its root atom. The root atom of CTB is the tetrahedral carbon; the root atom of PTX is the central carbon shared by two five-membered rings and the six-member ring (Fig. 1). When z_R is fixed, the ligand is restrained at the given level of the pore. However, it may rotate around the root atom and move normally to the pore axis so that an MCM trajectory would yield an optimal position and orientation of the ligand at the given level of the pore. To build the MC-minimized energy profile of a ligand in the pore, a series of MCM trajectories with the z-constrained ligand were calculated and the lowest-energy structures were collected from each trajectory. Test MCM trajectories which were started from the same z_R, other internal coordinates being different, converged to similar but not identical structures. To minimize the dependence of an MC-minimized structure on the starting geometry, we have used the same starting geometry of the receptor and the same orientation of the ligand for the given series of MCM trajectories.

Each receptor model was initially optimized by a long MCM trajectory in the absence of a ligand with all side-chain torsions being sampled. To speed up calculations of energy profiles, we used the optimized structures of the receptors as the starting points and sampled only those variables that affect the ligand-receptor interactions by governing conformation, position, and orientation of the ligand and conformations of the residues close to the ligand. According to the standard ZMM protocol, each 50th energy minimization rebuilds an interaction list that includes pairs of atoms at less than the cutoff distance (8 Å in this study). Simultaneously, the list of variables to be sampled was rebuilt to include all degrees of freedom of the ligand and those side-chain torsions that govern positions of atoms within 8 Å from the ligand. The side-chain torsions were sampled using the biased probability MCM protocol (Abagyan and Totrov, 1994).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION CONCLUSIONS REFERENCES

Background of the models

In the absence of a high-resolution structure of LGICs, we have built homology models of GlyRs and GABARs using as a template the model of nAChR that explained structure-activity relationships of noncompetitive blockers (Tikhonov and Zhorov, 1998). The template consists of five -helical M2 segments kinked in their middle part. The helices are parallel to each other in positions 1' to 9', but diverge at position 9' to 19' to make a funnel-like pore. The C atoms at positions 1', 2', 6', 9', 13', and 16' face the pore. Residues in these positions were experimentally determined to line the pore in both cation-selective and anion-selective LGICs (Akabas et al., 1994; Xu and Akabas, 1996; footnotes § and ¶ in Table 1). This fact justifies using the model of nAChR as a template for homology models of chloride-selective LGICs. In the homology models, the C atoms are restrained to the template by pins, while side chains may move to adjust ligands.

Below we present the MC-minimized energy profiles for CTB and PTX in the homomeric and heteromeric Cl-selective GlyRs and GABAR. We further analyze interactions that stabilize the optimal ligand-receptor complexes and discuss structure-function relationships of the ligand and ion channels in view of the obtained results.

Locating optimal binding sites for blockers in the pore

In our models, the pore axis coincides with the z axis of the Cartesian system of coordinates, the plane xy being close to ring R^16' (see Fig. 2 and Table 3). To locate optimal binding sites for CTB and PTX, we have pulled the ligands through the pore from ring R^16' to ring R^1' with the step of 1 Å (from level 0 to level 29; see Fig. 2). At each level, an MCM trajectory was calculated with the z coordinate of the ligand's root atom being fixed and all other variables allowed to vary. The backbone topology of the pore was preserved by pinning C atoms. The following parameters of the lowest-energy MEC found in each trajectory were used to draw the energy profiles: the total energy E_t, the energy of ligand-receptor interactions E_lr, and electrostatic component of the energy of ligand-receptor interactions, E_lre.

With given starting geometry and parameters controlling an MC protocol, the energy E_t depends on the length of the trajectory. The top, middle, and bottom lines at Fig. 3 A show E_t obtained, respectively, after a single energy minimization, after 100 energy minimizations, and after the trajectory converged. (A usual reason for the convergence was repeated overflowing of the internal stack of MECs without decreasing E_t during 500 consecutive energy minimizations, see Methods.) An MCM trajectory takes 700-3300 minimizations (Fig. 3 B) while obtaining one MCM energy profile requires more than 30,000 minimizations. With the reference to the energy-minimized starting points (the top line in Fig. 3 A), the initial 100 minimizations decrease E_t by 30-40 kcal/mol while the converged trajectories decrease E_t by 60-80 kcal/mol. Since an MCM trajectory is not expected to reach the global minimum, the ragged shape of the profiles of E_t most probably reflect incomplete optimizations rather than essentially different ligand-receptor interactions at adjacent levels of the pore. Indeed, the profiles of E_lr are smoother than those of E_t (see Fig. 4, A-E) indicating that bad ligand-receptor contacts occurring after inserting the ligand at the given level of the pore relax faster than bad intra-receptor contacts. The latter occur as the pore-facing residues move away from the ligand at initial minimizations. These residues may be trapped at high-energy levels because the pinned C atoms lack the mobility necessary to bypass energy barriers and reach low-energy valleys. In contrast, bad ligand-receptor contacts relax faster because only the z coordinate of the root atom of the ligand is constrained.

View larger version (23K): [in this window] [in a new window]   FIGURE 3   (A) Energy profiles of Et for CTB in 1/ GlyR. The top, middle, and bottom thin lines show energies obtained after single minimization at each level, after 100 minimizations, and upon the convergence of the MCM trajectory (when the internal stack of MECs repeatedly overflowed without decreasing the lowest energy during 500 minimizations). Positions of rings of the pore-facing residues R1', R2', etc. along the pore are indicated by labeling levels accommodating C atoms of the corresponding residues (see Table 3). (B) The number of energy minimizations performed upon the convergence of the trajectories.

View larger version (38K): [in this window] [in a new window]   FIGURE 4   MC-minimized energy profiles of CTB in the 1/ (A), 1 (B), 2 (C), and 2/ (D) GlyRs. The profiles represent the total energy Et (dashed line), ligand-receptor energy Elr (bold line), and Elre, the electrostatic component of Elr (thin line). (E) Superposition of the Elr profiles shows the deepest minimum for 1 GlyR (squares) and the second deep minimum for 1/ GlyR (circles). The minimum for 2/ GlyR (triangles) is shifted toward ring R6', indicating that CTB does not interact properly with ring R2'. The energy minimum at the profile of CTB in for 2 GlyR (crosses) is both shallow and narrow as compared with other receptors. This is consistent with the low activity of CTB at 2 GlyR (see Table 2). Positions of rings of the pore-facing residues R1', R2',etc. along the pore are indicated by labeling levels accommodating C atoms of the corresponding residues (see Table 3).

CTB in GlyRs

As summarized in Table 2, CTB effectively blocks ₁ and ₁/ GlyR with IC₅₀ 1.3-2.8 µM. In contrast, CTB is inactive in ₂ GlyR whose M2 differs from that in ₁ GlyR by only one residue: Gly^2' in ₁ GlyR and Ala^2' in ₂ GlyR (Table 1). The mutation of Gly^2' in the ₁ GlyR to Ala yields receptors resistant to the CTB block (Rundström et al., 1994), clearly indicating that methyl groups of the five Ala^2' residues impede CTB binding. To understand structural reasons for these experimental data, we have calculated MC-minimized energy profiles of CTB in heteromeric (₁/ and ₂/) and homomeric (₁ and ₂) GlyRs.

CTB in ₁/ GlyR

The MC-minimized energy profile for CTB in ₁/ GlyR (Fig. 4 A) has deep and wide minima of E_lr and E_lre between levels 20 and 26. At level 24, a minimum of E_lr coincides with a deep minimum of E_t indicating that ligand-receptor interactions do not conflict with the intrareceptor interactions or the pins. At levels 21-24, the ligand fits between ₁ Gly^2' and  Pro^2' residues (Fig. 5). The profile of E_t has a barrier at level 16, close to ring R^6'. Fig. 6 shows three snapshots of CTB passing ring R^6'. At level 14, a phenyl group of CTB intrudes between Phe^6' and Thr^6' residues in adjacent M2s. A dramatic change of CTB orientation is seen at level 15 as two phenyl groups of CTB squeeze between side chains of Phe^6' and Thr^6' residues, while the third phenyl group of CTB remains at the extracellular side of ring R^6'. As CTB proceeds to level 16, side chains of Phe^6' residues move back toward the pore axis while a repulsion of CTB from three Thr^6' residues increases E_t (see Fig. 4 A). Thus, the MC-minimized energy profile predicts that a relatively large CTB may pass ring R^6' and reach ring R^2'. This explains why mutations of M2s at position 2' affect CTB binding (see Table 2).

View larger version (31K): [in this window] [in a new window]   FIGURE 5   Complexes of CTB with GlyRs. The ligand and side chains in the rings of pore-facing residues near the ligand are space-filled. Carbon, nitrogen, and boron atoms of CTB are colored yellow, blue, and light blue, respectively. One M2 segment is removed for a better view of the ligand. (A) CTB at levels 21 (left) and 24 (right) of the 1/ GlyR corresponding to two points in the deep and wide minimum of the profile Elr at Fig. 4 A. (B) CTB at level 19 of the 2 GlyR corresponding to the barrier of Et at Fig. 4 C. CTB is squeezed between rings R6' and R2', and lacks the room to maneuver and pass the barrier as readily as in the 1/ GlyR (see Fig. 6).

View larger version (39K): [in this window] [in a new window]   FIGURE 6   Extracellular and side views of CTB passing ring R6' in 1/ GlyR. Three rows, left to right, correspond, respectively, to levels 14-16. At side views, one M2 segment is removed for a better view of the ligand. See text for further explanation.

CTB in ₁ GlyR

The MC-minimized energy profiles of CTB in ₁ GlyR (Fig. 4 B) show minima of E_t, E_lr, and E_lre at level 23, close to Gly^2' residues. Five -branched side chains of Thr^6' residues in ₁ GlyR would be expected to form a higher barrier for CTB as compared to ₁/ GlyR, in which Phe^6' residues move away to give room for CTB (see Fig. 6). However, the wide ring of Gly^2' residues in the ₁ GlyR provides room for CTB to maneuver as it passes the Thr^6' ring (note a small barrier of E_t at level 16, Fig. 4 B). Thus, in accordance with the experimental data (Table 2) the MC-minimized energy profile predicts binding of CTB at the ring of Gly^2' residues in ₁ GlyR.

CTB in ₂ GlyR

The MC-minimized energy profiles of CTB in ₂ GlyR (Fig. 4 C) show minima of E_t at levels 4 and 23 and barriers at levels 10 and 13-22. The minimum at level 4 (present also at CTB profiles in other GlyRs) is due to the intercalation of the ligand between two diverging M2s. In real GlyRs, M1s that contribute to the synaptic end of the pore (Akabas and Karlin, 1995) would obstruct such intercalation. The minimum of E_t at level 23 is shallower and narrower than the corresponding minima for ₁ and ₁/ GlyRs (Fig. 4 E). This factor may contribute to the low activity of CTB in ₂ GlyR.

In general, the barriers of E_t in ₂ GlyR are essentially higher than in ₁ and ₁/ GlyRs. The barrier at level 10 is due to incomplete optimization: ₁ GlyR has the same residues but does not have a barrier of E_t at this level. However, the wide barrier at levels 13-22 has a physical reason: at level 19, CTB is squeezed between the rings of Thr^6' and Ala^2' residues (Fig. 5 B). The blocker lacks space to maneuver and cannot adjust its phenyl rings between Thr^6' residues as in ₁ GlyR. Since the permanently charged CTB cannot pass via the lipid bilayer to reach the binding site from inside the cell, the barrier in the pore may contribute to the low activity of CTB at the ₂ GlyR.

CTB in ₂/ GlyR

Heteromeric ₂/ GlyR with Ala^2'/Pro^2' and Thr^6'/Phe^6' residues is blocked by CTB rather effectively, with IC₅₀ of 7.5 µM (Table 2). This fact is surprising because CTB does not block ₂ GlyR, whose ring of Ala^2' residues is not expected to be wider than the ring of Ala^2'/Pro^2' residues in ₂/ GlyR. Moreover, substitution of Pro^2' by Ala^2' in the  GlyR subunit failed to abolish the inhibitory activity of CTB (Rundström et al., 1994) indicating that Pro^2' residues in ₂/ GlyR do not increase the affinity for the blocker. These observations also suggest that residues other than 2' may contribute to the inhibitory action of CTB. Indeed, Table 2 shows that rings R^2' and R^6' concertedly affect CTB activity: five Ala^2' and five Thr^6' residues impede the CTB binding, whereas the presence of Phe^6' residues promote CTB binding despite the five Ala^2' residues.

To understand these paradoxical facts, we have calculated the MC-minimized energy profiles of CTB in ₂/ GlyR (Fig. 4 D) and found the minimum of E_lr to be deeper and the barrier of E_t to be narrower than in ₂ GlyR (see superimposed profiles in Fig. 4 E). The results suggest that unlike the rings of Thr^6' and Ala^2' residues in ₂ GlyR, the rings of Phe^6'/Thr^6' and Ala^2'/Pro^2' residues in the ₂/ GlyR provide enough room for CTB to maneuver, helping the blocker to reach the low-energy binding site at ring R^2'. Thus, our calculations provide a possible explanation for the paradoxical relationships between the activity of CTB and the structure of M2s at positions 2' and 6'.

PTX in GlyRs

PTX effectively inhibits ionic currents in the homomeric human ₁ and ₂ GlyRs with IC₅₀ of 6-9 µM, but displays a low affinity for the heteromeric ₁/ GlyR (Table 2). Test calculations of PTX in GlyR with MCM trajectories, run under standard controlling parameters, yielded a ragged profile of E_t (not shown) because the bulky ligand induced many unfavorable intrareceptor contacts in the initial steps. Relaxation of these contacts required long MCM trajectories that smoothed the E_t profile but did not significantly change the E_lr profile because, as mentioned above, bad ligand-receptor contacts relax faster than bad intrareceptor contacts. Therefore, the MCM energy profiles for PTX are represented only by the ligand-receptor energy E_lr and by its electrostatic component E_lrt. Those degrees of flexibility that are not considered in our model (e.g., flexible bond angles) would smooth the profiles of E_t.

PTX in ₁ GlyR

The E_lr profile has minima at levels 12-14 and 19-24 and a barrier of ~8 kcal/mol at levels 15-18 (Fig. 7 A). At level 13, the long axis of PTX is normally oriented to the pore axis. The methyl group protruding from PTX side fits the hydrophobic ring formed by the methyl groups of Thr^6' residues, while carbonyl and ether oxygens of PTX accept H-bonds from three Thr^10' residues and an isopropenyl group approaches two other Thr^10' residues (not shown). A barrier of E_lr is observed at levels 15-18. At level 17, the long axis of PTX is parallel to the pore axis (Fig. 8 A), the H-bond between Thr^6' and PTX hydroxyl decreasing the barrier. The energetically preferable binding mode is observed at level 21, where the rounded end of PTX accepts H-bonds from Thr^6' residues (Fig. 8 B). Since our calculations do not take into account solvent effects, they underestimate the stabilizing energy of H-bonds between PTX and Thr^6' residues at the narrow level 21, where the ligand would displace most of the intrapore waters. Unlike the anionic CTB, the electroneutral PTX does not interact with the helical macrodipole, yielding the E_lre profile without a deep minimum at the cytoplasmic end. However, attraction of the negatively charged PTX oxygens to the positively charged C' atoms of the peptide backbone determine the negative values of E_lre all along the pore.

View larger version (23K): [in this window] [in a new window]   FIGURE 7   MC-minimized energy profiles of PTX in the 1 GlyR (A), 1/ GlyR (B), and in 1 GABAR (C) showing the ligand-receptor energy Elr (bold line), and Elre, the electrostatic component of Elr (thin line). (D) Superposed Elr profiles from the plots (A) and (B) show that the minimum of the PTX profile in 1 GlyR is wider and deeper than in 1/ GlyR (dotted line).

View larger version (20K): [in this window] [in a new window]   FIGURE 8   Complexes of PTX with GlyRs and 1 GABAR corresponding to the extreme points of the energy profiles in Fig. 7. PTX and side chains in the rings of the pore-facing residues near PTX are space-filled. Oxygen, carbon, and hydrogen atoms of PTX are colored, respectively, by the dark, normal, and light shades of yellow. (A and B) PTX at levels 17 and 21 in the 1 GlyR that correspond, respectively, to the maximum and minimum of the energy profile in Fig. 7 A. (C) PTX at level 21 of 1/ GlyR corresponding to the energy minimum of the profile in Fig. 7 B. (D) PTX at level 20 of 1 GABAR corresponding to the energy minimum of the profile in Fig. 7 C. The extracellular parts of M2s, which do not affect PTX binding between rings R2' and R6' are truncated after Thr10'.

PTX in ₁/ GlyR

The MC-minimized energy profile of E_lr for ₂/ GlyR has a minimum at levels 18-20 (Fig. 7 B). Superposition of the E_lr profiles for PTX (Fig. 7 D) shows that the minimum in ₁/ GlyR is narrower than in ₁ GlyR and is shifted by ~5 Å toward the R^6' ring. At level 21, PTX fits between rings R^6' and R^2'. Only three Thr^6' residues may donate H-bonds to the PTX oxygens because Phe^6' residues lack H-bond donors in the side chain (Fig. 8 C). This may be the major factor impeding PTX activity in the ₁/ GlyR.

Thus, our calculations predict that PTX may reach rings R^6' and R^2' from the extracellular side and accept several H-bonds from the ring of five Thr^6' residues present in the homomeric GlyR. Our model explains the low activity of PTX at the heteromeric receptors with  GlyR subunits by the lack of H-bond donors in the side chain of Phe^6' residue leading to a weaker interaction of PTX with the channel.

PTX in ₁ GABAR

In addition to its blocking action on GlyRs, PTX also antagonizes homomeric and heteromeric GABARs with high to moderate affinities (Table 2). Mutations in positions 6' and 2' dramatically affect PTX activity (see Table 2), suggesting that in GABARs the binding site for the blocker is located at the same levels of the pore as in GlyRs. To analyze the impact of the pore-facing residues on the binding of the blocker, we have calculated the energy profile of PTX in the ₁ GABAR. Because the united-atom model may underestimate a probable repulsion between five Pro^2's and PTX, we have built a more realistic all-atom model of ₁ GABAR.

The calculated MCM profile has a deep, narrow minimum at level 20 (Fig. 7 C). The MCM trajectory at this level yielded the lowest-energy MEC with the PTX elongated end at ring R^2' and the oxygens at the rounded end of PTX accepting H-bonds from Thr^6' residues (Fig. 8 D). Although the elongated end of PTX fits the ring of five Pro^2' residues, C atoms of Pro^2' impede the mobility of the blocker at this level. The observations that substitution of Pro^2' in ₁ GABA_CR by Ala increases activity of PTX (Table 2) is probably due to the larger R^2' ring providing more room for PTX mobility. Substitution of Thr^6' by Phe or Met essentially decreases the PTX activity (Table 2). This concurs with our above conclusion that Thr^6' residues in GlyRs stabilize PTX by donating several H-bonds.

Thus, our models qualitatively explain the relationships between the structure of rings R^6' and R^2' in GABARs and GlyRs, and their sensitivity to PTX. We conclude that the ring of Thr^6' residues that may donate several H-bonds to the rounded end of PTX is the key structural determinant of PTX activity at both GlyRs and GABARs.

Hydrated Cl in ₁ GlyR

Our calculations suggest that relatively large organic blockers may reach ring R^2' from the extracellular side. The pore between rings R^16' and R^2' should be wide enough to accommodate Cl with its inner hydration shell. Since the positively charged hydrogens of water molecules surround Cl, the oxygen atoms of the waters would accept H-bonds from the rings of polar residues in a manner that may resemble, to some extent, the interaction of the oxygens at the rounded end of PTX with the H-bond-donating rings. To highlight a possible analogy in the binding of PTX and hydrated Cl, we have calculated an MC-minimized energy profile of the cluster Cl/(H₂O)₈ in the ₁ GlyR. The water oxygens were constrained to Cl by the flat-bottom penalty functions with the upper distance limit of 6 Å. As in the case of computing energy profiles for CTB and PTX, the z coordinate of Cl was fixed at 30 levels of the pore and MCM trajectories were calculated to find optimal structures of the Cl hydration shell, optimal position of Cl at the given level of the pore, and optimal side-chain conformations of the receptor. This approach allows Cl to exchange waters from its surroundings for more attractive groups in the channel, to abandon waters in the pore constrictions, but rejoin the waters at the wide levels of the pore.

Fig. 9 A shows the MC-minimized energy profile of Cl/(H₂O)₈ in the ₁ GlyR. The lowest-energy structures found in each trajectory are characterized by the total energy (E_t), the energy of interactions of Cl with the receptor and waters (E_Cl), and E_{Cl_e}, the electrostatic component of E_Cl that provides the major contribution to E_Cl. E_t is high at levels 0-4 because Cl-bound waters do not establish favorable contacts with the channel. Eight Cl-OH₂ distances plotted against the Cl position in the pore (Fig. 9 B) show that five or six waters approach Cl at levels 0-6, while two to six waters remain in a close contact to Cl at other levels.

View larger version (45K): [in this window] [in a new window]   FIGURE 9   (A) MC-minimized energy profile of Cl/(H2O)8 in the pore of 1 GlyR: the total energy of the system (Et, dashed line), the energy of interactions of Cl with the receptor and waters (ECl, bold line), and ECl_e (thin line), the electrostatic component of ECl. (B) Distances between Cl and water oxygens in the cluster Cl/(H2O)8 against Cl position in the pore. Six, two, and four waters closely approach Cl at levels 5, 17, and 21, respectively, indicating that Cl hydration is affected by the channel micro-architecture. Each water molecule is kept within 6 Å of Cl by the flat-bottom constraint Cl-O with the upper limit of 6 Å.

The profiles of E_t and E_Cl have maxima at level 17, where most of the water molecules move away from Cl (Fig. 9 B) to interact with Thr^7' residues. The latter are specific for Cl-selective channels (Table 1) and may contribute to the anionic selectivity. The profile of E_Cl below ring R^6' (levels 20-24) has deep and wide minimum corresponding to the most preferable binding site for the hydrated Cl. The cluster Cl/(H₂O)₈ at level 22 is extended along the pore as four waters are retarded at Thr^6' residues to form a pattern resembling the rounded end of PTX (Fig. 10). The model of Cl hydration applied is too simple to account for the anion selectivity of GlyRs and GABARs. However, it shows that Thr^6' residues may stabilize the hydrated Cl inside the pore in the same manner as they stabilize PTX. This provides a possible explanation for the nature of Cl channel block by PTX.

View larger version (31K): [in this window] [in a new window]   FIGURE 10   MC-minimized complex of the 1 GlyR with Cl/(H2O)8 at level 23. Receptor, water molecules, and Cl are coded by light, medium, and dark shades, respectively.

Mechanosensitive receptor and LGICs

Recently, a crystallographic structure of a mechanosensitive ion channel (MscL) was published (Chang et al., 1998). It shows a five-helix bundle forming the central pore with the crossing angle between the pore helices similar to that observed in the KcsA K⁺ channel (Doyle et al., 1998). Ligand-gated and mechanosensitive ion channels have a different numbers of TM segments (20 and 10, respectively) and very different electrophysiological properties (Imoto et al., 1988; Newland and Cull-Candy, 1992; Fucile et al., 1999; Sukharev et al., 1994). Despite this fact, the five-helix bundle motif in the MscL was proposed as a template for the modeling of LGICs (Chang et al., 1998). Indeed, the aligned sequences of the pore-forming segments of MscL and ₁ GlyR have identical residues Gly^2', Thr^6' Thr^10', Thr^13' and Ser^15', as well as homologous residues in positions 1' 4', 8', and 9' (Table 1). However, the experimental data on the pore-lining residues in nAChR (Akabas et al., 1994) and GABA_AR (Xu and Akabas, 1996) are only partially consistent with the experimental structure of MscL. Val^1', Gly^2', Thr^6', and, probably, Val^9' would be reachable by the ligands from inside the pore of MscL, but Thr^13' and Ile^16' do not face the pore.

To compare our AChR-based models with MscL, we have built a preliminary homology model of ₁ GlyR using the MscL structure as a template. In the MscL-based homology model, Thr^6' residues may donate H-bonds to PTX and to the hydration shell of Cl in a manner similar to that found in AChR-based models of GlyRs and GABAR. However, ring R^6' in the MscL-based model is essentially larger than in the AChR-based models. The MCM docking of PTX in the MscL-based model of ₁ GlyR yielded a complex with only three H-bonds between PTX and Thr^6' residues (Fig. 11) and only three M2s being in a close contact with the ligand. As we discuss below, such a binding mode is energetically less preferable than the ligand binding in a close contact with five M2s.

View larger version (44K): [in this window] [in a new window]   FIGURE 11   Extracellular view of the preliminary homology model of the 1 GlyR built with the crystallographic structure of the mechanosensitive channel from Mycobacterium tuberculosis (Chang et al., 1998) as the template. PTX loosely binds at ring R6', accepting only three H-bonds from Thr6' residues. Receptor and PTX are coded by light and dark shades, respectively.

We further imposed five H-bonding constraints between Thr^6' residues and the rounded end of PTX, and allowed M2 helices to translate parallel to the xy plane toward the pore axis as rigid bodies while preserving the orientation of the helical axes. An MCM trajectory yielded a low-energy structure more consistent with the AChR-based models in terms of the dimensions of the pore (not shown). Thus, models of LGICs may be built by imposing the crossing angles between the pore helices seen in MscL and the dimensions of rings R^6' and R^2' obtained in the present study. Future systematic analysis should clarify whether such models would explain the pharmacological peculiarities of LGICs.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION CONCLUSIONS REFERENCES

The major aim of the present work is to analyze, using homology modeling and MCM calculations, the possible molecular mechanisms for the block of Cl channels by CTB and PTX, and to explain the dependence of their effect on the subunit composition of glycine and GABA receptors.

As described above, CTB and PTX have very different chemical structures. In particular, CTB is an anion expected to bind in the anion-selective pore while PTX is a neutral molecule, and the inhibitory mechanism of this plant alkaloid is still not clear. Experimental results suggest that PTX blocks GABAR- and GlyR-mediated currents in two different ways: 1) as a noncompetitive, channel-blocking antagonist; and 2) as an allosteric antagonist that does not occlude the pore lumen.

In favor of the allosteric mechanism are results of whole-cell and single-channel analysis using rat dissociated sympathetic neurons, suggesting that picrotoxin stabilizes an agonist-bound shut state (Newland and Cull-Candy, 1992). This view is also supported by observations on the homomeric ₁ GlyR (Lynch et al., 1995) and GABA_CR (Wang et al., 1994; Qian and Dowling, 1994; Zhang et al., 1995a) demonstrating the competitive component of PTX-induced inhibition. This component can be eliminated by mutations of a single amino acid in the M2 segment (Wang et al., 1995).

Several lines of evidence support the noncompetitive, channel-blocking mechanism of PTX action. First, this antagonist inhibits various anion-selective receptor-operated channels independently of the receptor type. Thus, in addition to blocking GABARs and GlyRs, PTX effectively blocks Cl-dependent currents activated by glutamate (reviewed by Cleland, 1996), acetylcholine (Yarowsky and Carpenter, 1978) or dopamine (Magoski and Bulloch, 1999). Moreover, this alkaloid does not modulate the binding of GABA to its receptor (Enna et al., 1977) and the onset rate of the PTX-induced inhibition of the homomeric GlyR does not depend on the presence of glycine (Lynch et al., 1995). These observations strongly suggest that the site of PTX action does not coincide with the neurotransmitter's recognition sites. Second, site-directed mutations in the pore-facing M2 segment dramatically affect the inhibitory activity of PTX. This was demonstrated on GlyR (Pribilla et al., 1992), homomeric human (Wang at al., 1994, 1995; Enz and Bormann, 1995) and rat (Zhang et al., 1995a) GABA_CRs, and , , or  subunits of GABA_AR (Gurley et al., 1995; see Table 2). Third, a synthetic four-helix bundle protein composed of M2 segments from GlyR formed anion-selective channels that were blocked by PTX (Reddy et al., 1993) although the properties of this block were different from those reported for authentic channels (Newland and Cull-Candy, 1992). Fourth, using the substituted-cysteine-accessibility method, Xu et al. (1995) convincingly demonstrated that PTX interacts with the pore-lining residues of the GABA_AR channel. The blocker protected the engineered ₁ Cys^2' residue from modification by sulfhydryl reagents. Moreover, in oocytes expressing the mutant with the engineered Cys^6' residue, a sulfhydryl reagent, methanethiosulfonate ethylammonium, decreased the PTX-induced block, demonstrating a competition with the blocker for the common binding site comprising Cys^6'. All these studies clearly indicate that the main target for PTX action is the Cl-selective pore region.

Although it is possible that receptors can have two distinct types of PTX binding sites (Davis and Ticku, 1981; Yoon et al., 1993), results of our study favor a "pore region" mechanism of PTX action. Assuming that different LGICs share a similar five-helix-bundle architecture of the pore region, we have included in this study homology models of several GlyRs and GABARs, using as a template the model of nAChR in the open state (Tikhonov and Zhorov, 1998). We further performed a systematic search for probable binding sites of two different blockers, CTB and PTX, in the models of Cl-selective channels.

Mutagenesis experiments suggest that interaction of CTB and PTX with GlyRs and GABARs depend on residues in positions 2' and 6' (see Table 2). A simple approach would be to dock the ligands at rings R^6' and R^2' that have been experimentally demonstrated to affect the ligand binding. However, to avoid a bias, we have performed systematic searches for the energetically most preferable binding sites for CTB and PTX. In each receptor model, a ligand was restrained at 30 levels of the pore, an MCM trajectory was calculated for each level, and parameters of the lowest-energy MECs found in the trajectories were presented as the MC-minimized energy profiles.

General properties of the energy profiles

The MC-minimized energy profiles show the total energy and its major components as a function of the ligand position in the pore. The extracellular half of the pore is wider than either CTB or PTX. At these levels, the ligands may interact with no more than two or three M2s simultaneously, yielding ligand-receptor energy E_lr of only ~10 kcal/mol (Figs. 4 and 7). The intracellular half of the pore is narrower. A priori, it was not clear whether CTB and PTX would fit there. The fact that all the E_lr profiles show large negative values in the intracellular half of the pore indicates that the blockers do fit there and interact with five M2s simultaneously. These interactions may drive the ligand into the narrow pore. Thus, the systematic search revealed that dimensions of both CTB and PTX match cross-sectional dimensions of the inner half of the pore. This is a computational argument in favor of a noncompetitive mechanism of PTX binding.

Because computing an MCM energy profile requires large computational resources, we have not attempted to statistically evaluate a variability of the energy profiles by simulating different runs of the same ligand via the same receptor. An indirect estimate of the variability comes from comparing central stretches of the energy profiles at Fig. 4 E that correspond to CTB passing levels 9-20 (M2 positions 10'-3'). At these positions, ₁, ₂, ₁/, and ₂/ GlyRs have identical or homologous residues (Table 1). At most levels in this area, the ligand-receptor energy of different profiles varies within 5 kcal/mol. In contrast, the ligand receptor energy at levels 5-10 and 21-26 (where M2s have different residues) varies up to 10 kcal/mol.

The minimal values of E_lr in the energy profiles (below 30 kcal/mol) are not compatible with the micromolar affinities observed for CTB and PTX in GlyRs and GABARs (Table 2). It should be noted that our calculations take into account only enthalpy contributions to the free energy of ligand-receptor interactions. Free energy calculations that require huge computational resources were not performed in this work (as well as in analogous modeling studies). Thus, the discrepancy between the calculated minimal values of E_lr and observed affinities of the blockers is explained by ignoring in our calculations destabilizing entropy contributions such as ligand and receptor dehydration and the lost of ligand and receptor degrees of freedom in the ligand-receptor complexes.

The profiles of electrostatic energy have deep minima for CTB but not for PTX at the cytoplasmic end of the pore. Since our models represent ionogenic residues in their neutral forms, the minima are due to the macrodipole effect of stabilizing an anion at the N-end of an -helix (Aqvist et al., 1991; Lockhart and Kim, 1992, 1993; Sitkoff et al., 1994). Macrodipoles from five M2s concertedly stabilize the anionic CTB at the cytoplasmic end of the pore. At first sight, the fact that PTX is not an anion may provide evidence against its binding in the anionic pore. However, the energy profiles for PTX also show a significant contribution of the negative electrostatic energy to the ligand-receptor interactions. Partitioning the ligand-receptor energy shows that interactions of the negatively charged PTX oxygens with the positively charged C' atoms in the peptide backbone are the major components of the electrostatic stabilization of PTX in the pore.

Another stabilizing factor revealed by our calculations are H-bonds between oxygen atoms at the rounded end of PTX and Thr^6' residues. The substitution of Thr^6' by Phe in any subunit of the rat ₁/₂/