| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Department of Biochemistry and Biomedical Sciences, McMaster University, Hamilton, Ontario, Canada
Correspondence: Address reprint requests to Dr. Boris Zhorov, Dept. of Biochemistry and Biomedical Sciences, McMaster University, 1200 Main Street West, Hamilton, Ontario, L8N 3Z5 Canada. Tel.: 905-525-9140 ext. 22049; Fax: 905-522-9033; E-mail: zhorov{at}mcmaster.ca.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONSUPPLEMENTARY MATERIALACKNOWLEDGEMENTSREFERENCES |
|---|
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONSUPPLEMENTARY MATERIALACKNOWLEDGEMENTSREFERENCES |
|---|
). The channels are formed by four membrane-spanning domains arranged around the central pore. The pore-forming region of each domain includes the transmembrane outer helix S5, the membrane reentrant P-loop, and the transmembrane inner helix S6. The x-ray structures of several bacterial K+ channels in the closed and open states (Doyle et al., 1998; Jiang et al., 2002, 2003; Kuo et al., 2003) show that each P-loop folds up to form five segments:
The pore-forming 
1-subunit of Na+ channels has four nonidentical repeats within a single polypeptide chain. Each repeat incorporates the S1–S4 domain with the voltage sensor and the pore domain S5-P-S6. In K+ channels, the selectivity filter is formed by the backbone oxygens from the signature sequence TXGYG (Doyle et al., 1998). In the Na+ channel, highly conserved residues Asp, Glu, Lys, and Ala, in repeats I–IV, respectively form a ring called the DEKA locus. The side chains of the DEKA locus govern selective permeability of Na+ channels (Heinemann et al., 1992).
In all P-loop channels, residues governing selectivity are located near the C-termini of the pore helices. Since P-loop channels differ greatly in ion selectivity, it is expected that the structural details of their selectivity filters are very specific. In K+ channels, the conserved TXGYG motif forms a narrow tunnel, in which ions are coordinated by the main chain carbonyls. This three-dimensional structure is stabilized by inter- and intrasegment H-bonds (Doyle et al., 1998). Selectivity filters of other channels have been modeled. Models for glutamate-gated channels integrate data on the pore dimensions, activity of channel blocking drugs, and mutational analysis (Tikhonov et al., 1999, 2002). Models of the Na+ channel were proposed to explain experimental data on mutations and blockade by guanidinium toxins (Lipkind and Fozzard, 2000; Khan et al., 2002). Diverse models of the L-type Ca2+ channel were elaborated (Zhorov et al., 2001; Lipkind and Fozzard, 2001). All these models use x-ray structures of K+ channels as templates for general folding. However, the selectivity filters that control specific properties of K+, Na+, and Ca2+ channels should have different structures. For instance, in K+ channels, the permeating cations interact with the backbone carbonyl groups in the selectivity-filter region, whereas in other P-loop channels the permeating ions are believed to interact with the side chains of selectivity-filter residues. In the models of the Na+ channel (Lipkind and Fozzard, 2000) and glutamate-gated channels (Tikhonov et al., 2002) the pore helices were placed more distant from the pore axis than in K+ channels to make a wider channel. In the Na+ channel model (Lipkind and Fozzard, 2000), the wider channel provides access of tetrodotoxin (TTX) and saxitoxin (STX) to the selectivity-filter residues. Furthermore, the wider pore at the level of the selectivity filter could explain the fact that organic cations, which do not permeate via K+ channels, permeate via other P-loop channels (Burnashev et al., 1996; Hille, 1971, 1973; McCleskey and Almers, 1985).
There is a major problem with moving the pore helices from the pore axis. The pore helices form close contacts with the inner and outer helices. To preserve these contacts, the latter segments should be also moved. This relocation would affect folding of the entire channel. To avoid this, Zhorov et al. (2001) created a model of the L-type Ca2+ channel in which all the helices occupied the same positions as in KcsA. The wider pore at the selectivity-filter level was achieved by Monte Carlo minimizations of four hexapeptide segments (one from each repeat) around the selectivity-filter residues whose side chains were constrained to Ca2+ ions. The obtained structure did not have bad contacts, demonstrating that significant rearrangement of the selectivity-filter region is possible without disrupting the disposition of the pore helices.
The problem of the x-ray template flexibility is generally important for homology modeling of proteins. Obviously, the geometry of flexible loops, and the disposition of mobile domains and segments connected by hinges, can differ between homologous proteins. Can homologous proteins sharing the same fold differ significantly in the area of densely packed domains? This question is particularly important for studies of P-loop channels that employ high-resolution structures of bacterial K+ channels as templates. One of the goals of this work was to investigate whether experimentally available data on the Na+ channel selectivity-filter region can be explained without major modification of the x-ray template of the P-loop region. The selectivity filter of the Na+ channel is particularly suitable for such studies because a large body of experimental data is available. It is well known that Na+ selectivity is controlled by the DEKA locus. In Ca2+ channels, ion selectivity and permeation are controlled by the EEEE locus formed by four glutamates, which occupy positions homologous to the DEKA locus in Na+ channels (Kim et al., 1993; Yang et al., 1993). The replacement of the DEKA locus by four glutamates renders a Ca2+-selectivity to the channel (Heinemann et al., 1992; Favre et al., 1996).
Data on the Na+ channel blockade by guanidinium toxins provide unique possibilities to design and test molecular models. Tetrodotoxin (TTX) and saxitoxin (STX) block the voltage-gated Na+ channel in nanomolar concentrations (Narahashi et al., 1967). These toxins interact with residues in the DEKA locus and downstream from it (Noda et al., 1989; Terlau et al., 1991; Kirsch et al., 1994; Penzotti et al., 2001; Choudhary et al., 2003). The structure-activity relationships of TTX derivatives (Kao and Walker, 1982; Kao, 1986; Yotsu-Yamashita et al., 1999) revealed the importance of specific groups, thus providing additional information to be explained. TTX, STX, and their derivatives have a generally similar mechanism of action, but some residues of the Na+ channel interact differently with individual toxins (Satin et al., 1992; Penzotti et al., 1998).
Peptide toxins known as µ-conotoxins (Olivera et al., 1990) represent another class of ligands that act on the Na+ channel by the channel-blocking mechanism (Cruz et al., 1985; Ohizumi et al., 1986). The toxins have a critical arginine that could mimic the guanidinium group of TTX and STX (Yanagawa et al., 1987; Sato et al., 1991; Becker et al., 1992; Dudley et al., 1995; Chahine et al., 1995). However, mutations eliminating the sensitivity to TTX and STX are less critical for conotoxins (Chahine et al., 1998; Stephan et al., 1994; Li et al., 1997), indicating that residues beyond the TTX receptor are involved in the interaction with the large conotoxin peptide. The critical arginine residue of µ-conotoxin is likely to bind to the outer ring of negatively charged residues (Chang et al., 1998; Hui et al., 2002).
The above experimental data were integrated in models visualizing toxins bound to the selectivity filter and outer vestibule of the Na+ channel (Lipkind and Fozzard, 2000; Choudhary et al., 2003). The general pattern of interactions between specific functional groups of toxins and individual amino acids is well documented and presently does not need to be revised. However, the above models are descriptive rather than numerical. The ability of these models to sustain the above pattern of ligand-receptor contacts in molecular dynamics or Monte Carlo calculations has not been demonstrated. Independent testing of the models is not possible, because atomic coordinates are not available.
In this work we built a model of the Na+ channel P-loops region assuming its strong structural similarity with K+ channels by the disposition of the pore helices. We applied the Monte Carlo minimization protocol to search the energetically optimal position and orientation of the ligands in the selectivity-filter region. The optimal complexes found are in exact agreement with the general scheme of specific ligand-receptor interactions. The predicted binding energy of the TTX analogs quantitatively correlates with the experimental activity. The model is consistent with numerous experimental data that were not used at the model building stage, including mutational analysis, permeation of organic ions, and ion selectivity.
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONSUPPLEMENTARY MATERIALACKNOWLEDGEMENTSREFERENCES |
|---|
) was used as a template for homology modeling. The sequence alignment between MthK, Na+, and Ca2+ channels is shown in Table 1. This alignment was used in studies of Ca2+ channels (Zhorov et al., 2001; Yamaguchi et al., 2003) and Na+ channels (Yamagishi et al., 2001). Large extracellular linkers between the selectivity filter and the inner pore were not included in the model because the crystallographic structure of this part is not available. When viewed from the extracellular surface, the repeats I–IV were arranged in a clockwise fashion around the central pore (Dudley et al., 2000). The four repeats were not linked to each other.
|
). Electrostatic interactions were calculated with a distance-dependent dielectric (
= r). All ionizable groups in the protein, ligands, and peptide toxins were modeled in the ionized form. Nonbonded interactions were truncated at distances >8 Å. The cutoff was not applied to electrostatic interactions involving ionized groups; these interactions were computed at all distances. The hydration energy was calculated by the implicit-solvent method (Lazaridis and Karplus, 1999).
All-trans starting conformations were assigned for side chains of those residues that are different between MthK and the Na+ channel. The N- and C-termini of the P-loop fragments were constrained to positions seen in the MthK x-ray structure. Four residues in each repeat, including residues in the DEKA locus and three positions downstream, were treated as completely flexible. The -carbons of other residues were constrained to corresponding positions in the MthK template with the help of pins. A pin is a flat-bottom penalty function (Brooks et al., 1985) that allows penalty-free deviations of the respective atom up to 1 Å from the template and imposes an energy penalty for larger deviations. Flat-bottom constraints were also used to impose experimentally known ligand-receptor contacts.
The optimal conformations were searched by the Monte Carlo minimization (MCM) protocol (Li and Scheraga, 1987). Energy was minimized in the space of generalized coordinates using the ZMM program (Zhorov, 1981). MC-minimizations were terminated when the last 2000 energy minimizations did not improve the energy of the best minimum found. All MC-minimizations were performed in two stages. In the first stage, the energy was MC-minimized with pins and ligand-receptor constraints. After the constrained MCM trajectory converged, all constraints were removed and the model was refined by the unconstrained MCM trajectory. The difference between structures found in the constrained and unconstrained trajectories indicated whether the constrained search yielded a stable structure or a conformation with bad contacts. Other details of the MCM protocol implementation in the ZMM program are described elsewhere (Zhorov and Ananthanarayanan, 1996; Zhorov and Lin, 2000).
The ligands were built and their geometry was optimized using the ZMM program. All torsional and bond angles of ligands were treated as variables. The atomic charges of ligands were calculated by the AM1 method (Dewar et al., 1985) using the MOPAC program. Peptide toxins were built using structural information available from their nuclear magnetic resonance studies (Wakamatsu et al., 1992). The geometry of toxins was MC-minimized.
MC-minimized energy profiles of drugs pulled via the pore were computed as described elsewhere (Zhorov and Lin, 2000; Zhorov and Bregestovski, 2000). The plots of MC-minimized energy against the toxins' orientation in the selectivity filter were computed by constraining the dihedral angle between the plane comprising the pore axis and another plane passing through the long axis of the ligand. The latter was drawn from the central carbon atom in the guanidinium group to a carbon atom at the opposite side of the drug. The dihedral constraint freezes one of the six rigid-body degrees of freedom of the ligand. The constrained dihedral angle was increased with a 15° step. At each step the energy was MC-minimized with all other degrees of freedom being allowed to vary.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONSUPPLEMENTARY MATERIALACKNOWLEDGEMENTSREFERENCES |
|---|
The following interactions were used as constraints to build the TTX-based model:
The following interactions were used as constraints to build the STX-based model:
The experimental data that justify these constraints are summarized by Lipkind and Fozzard (2000) and Choudhary et al. (2003).
All bad contacts unavoidable in the MthK-based starting conformation were eliminated during MC-minimizations of both models. Importantly, no conflicts occurred between the ligand-receptor constraints and pin constraints in P-loops and residues following the selectivity-filter region. This indicates that the MthK-based model has enough space to accommodate TTX and STX without displacing the pore helices.
TTX and STX receptors have two rings with acidic residues. The guanidinium group of TTX and 7,8,9 guanidinium group of STX are known to bind to the DEKA locus. The opposite ends of the ligands have multiple donors of protons that bind to the outer ring of acidic residues containing Glu403, Glu758, and Asp1532. Obviously, various orientations of the toxins are possible, in which their multiple functional groups form H-bonds with the outer ring of the acidic residues. To explore whether the ligand orientation imposed by the experimental constraints is energetically optimal, we computed a MC-minimized energy profile for the ligand rotated around its long axis. At each point of the profile, one angle defining the ligand orientation was constrained, but all other degrees of freedom were allowed to vary in the MC-minimization (see Methods). In this stage, we did not impose any constraints biasing ligand-receptor interactions. The rotational profiles for both TTX and STX have a single well-shaped minimum (Fig. 1) at the orientations that correspond to the MC-minimized structures shaped around the ligands (see above). This result shows that energetically preferable orientations of TTX and STX do not depend on their initial placement.
|
|
-carbons in this region is as small as 0.29 Å. The obtained similarity between the independently built models is not trivial because toxins are different in structure. The similarity justifies our model-building approach. Since the TTX-based and STX-based models are practically similar, we describe below only the model shaped around TTX.
|
-carbons of Gly1238 and Gly1530, making knob-into-the-hole contacts. Some of these residues were reported to affect the ion permeation through Na+ channels (Tsushima et al., 1997). Tyr401 forms an H-bond with Glu403. As a result, the aromatic ring of Tyr401, a residue important for TTX binding, faces the pore and can interact with toxins. The arrangement of large residues around Gly1530 is shown in Fig. 2 B. In our model, Arg395 forms a salt bridge with Asp400, but does not impede interactions of Asp400 with TTX and STX. Specifically interacting partners were not found for Arg750.
At first sight, the fact that bulky guanidinium toxins bind to the Na+ channel selectivity filter suggests that its structure should significantly diverge from the narrow K+ channels. However, our models have rather small backbone deviations from MthK only in a limited region around the selectivity filter. These deviations provide enough space for the binding of toxins (Fig. 3). The narrowness of the K+ channel selectivity filter does not mean the lack of space in the P-loops region. Indeed, the pore-lining backbones are distant from pore helices due to H-bonds between large chains, including intrasubunit H-bonds Trp67
Asp80 and intersubunit H-bonds Trp68Tyr78 (KcsA numbering, see Table 1). Notably, these pairs of H-bonding residues are absent in the sequences of Na+ channels. This allows more compact packing of the selectivity-filter region against the pore helices, thus providing more space in the Na+ channel outer vestibule.
|
). The hydrophobic side of STX binds to the hydrophobic Met1240 side chain. The energy components of the ligand-receptor complexes are shown in Table 2. Solvation and van der Waals interactions provide significant contributions to the ligand-receptor energy. However, these interactions are similar for TTX and STX and weakly depend on their orientation. The reason is that the toxins have a conical shape, which is complimentary to the channel conical vestibule. Therefore, the contact surfaces between the toxins and receptor are similar, and similar amounts of water molecules are displaced upon the binding of both toxins. Electrostatic interactions provide the largest contribution to the ligand-receptor energy. This is not surprising, taking into account that the ligand and the receptor have oppositely charged groups. The charge complementarity with the receptor is the major determinant of the toxins' orientation. The largest contributions to the ligand-receptor energy are provided by aspartates and glutamates in the selectivity filter and carboxyl residues in the outer ring (Table 3). It should be noted that the absolute values of the electrostatic energy are obviously overestimated because solvent counterions were not taken into account.
|
|
). To test our model, we have selected four representative derivatives (Fig. 5 A) using the following criteria. First, we did not include compounds for which the lack of activity was obviously deducible from their structure. Second, the compounds selected have essentially different structures. Third, the activity of the selected compounds varies in a broad range, from nM to µM (Table 4). We did not make any adjustments of the model to dock the TTX analogs. For each analog, a rotational energy profile was calculated (Fig. 5 B). As in the case of TTX and STX, each profile has a well-defined minimum, which predicts unambiguously the optimal orientation of the ligand in the channel. As in the case of TTX and STX, the optimal orientation and the binding energy was mainly determined by the electrostatic complementarity of the ligands with acidic residues in the DEKA locus and in the outer ring. Notably, our calculations revealed a strong correlation between the experimentally observed activity and predicted binding energy (Table 4). Thus, our model allows quantitative analysis of the structure-activity relationships of guanidinium toxins.
|
|
, the most probable partner is Asp1532 (repeat IV), whereas Li et al. (2001) revealed a strong interaction of Lys16 with Asp1241 (repeat III).
To explore these possibilities, we MC-minimized the nuclear magnetic resonance structure of the toxin (Wakamatsu et al., 1992; PDB code: 1TCG) and manually docked it in the orientation suggested by Dudley et al. (2000), who defined the interaction patterns between µ-conotoxin specific residues and the channel repeats. Then this interaction pattern was used as a constraint in a two-stage MCM docking. The unconstrained search did not significantly change the binding modes of the toxin that were imposed in the constrained search. More importantly, the conotoxin-based model did not deviate significantly from the TTX/STX-based model. The RMSD of the main-chain atoms was 0.54 Å for the region upstream of the DEKA locus and 0.59 Å for the selectivity-filter region (between the DEKA and the outer ring of acidic residues). When 24 superficial residues in six positions downstream from the DEKA locus were taken into account, the RMSD between the TTX/STX- and conotoxin-based models increased to 1.67 Å. This is not surprising since the superficial residues do not significantly contribute to the TTX/STX binding but some of them (Asn404, Thr759, and Asp1241) were constrained to µ-conotoxin.
The binding of the conotoxin molecule is shown in Fig. 6. In the compact outer vestibule, the large guanidinium group of Arg13 can interact with Glu758 and simultaneously approach the DEKA locus. The side chain of Lys16 is projected toward the interface between repeats III and IV. The close disposition of the channel repeats allows the amino group of Lys16 to make a bridge between Asp1241 and Asp1532. Partitioned interaction energies for these two residues are shown in Table 5. It should be noted that both residues have more than one significantly interacting partner. The energies of specific interactions are in a qualitative agreement with experimental data (Chang et al., 1998; Dudley et al., 2000; Li et al., 2001).
|
|
). The probable reason is the strong electrostatic and H-bonding contacts between the DEKA side chains, which converge at the pore axis. However, strong transmembrane voltage results in a detectable permeability for large organic cations (Huang et al., 2000). The substitution of the DEKA lysine by alanine can break the inter-residue contacts and cause the channel to become less selective and permeable to cations as large as tetramethylammonium (TMA) and tetraethylammonium (Sun et al., 1997). However, the mutants do not behave as a simple molecular sieve because, paradoxically, the AAAA mutant of the DEKA locus is not permeable to TMA (Sun et al., 1997). The permeability to organic cations may be a critical test for our model, which is based on the K+ channel structure and therefore has a seemingly narrower pore lumen than previous models. To test whether the above experiments conflict with our model, we replaced the DEKA locus with AAAA and DEAA residues, constrained the TMA nitrogen to a plane normal to the pore axis, and computed 30 MCM trajectories by progressively shifting the constraining plane along the pore axis with a 1 Å step. The MC-minimized energy profiles (Fig. 7) have negative energies, clearly indicating the absence of critical sterical obstacles for TMA passing through the selectivity filter. The TMA energy in the AAAA mutant has two minima separated by a barrier. The first minimum corresponds to the electrostatic interaction with the external ring of acidic residues. The second minimum occurs close to the focus of four helical macrodipoles. The TMA permeation would be obstructed by the barrier, which is caused by unfavorable electrostatic interactions of TMA in the vicinity of the AAAA locus. The barrier disappears in the DEAA locus, whose flexible acidic residues attract TMA without imposing the steric hindrances. The outer ring of the acidic residues and the DEAA locus form a nucleophilic tunnel from the outer vestibule to the inner pore.
|
; Berneche and Roux, 2001; Shrivastava et al., 2002). Our model is not precise enough for such simulations, but it allows a qualitative analysis of structural determinants of selectivity. The Na+ channel selectivity-filter model (Lipkind and Fozzard, 2000) suggests that a water molecule is stabilized by H-bonds with the DEKA aspartate and lysine, whereas the DEKA lysine and glutamate form a salt bridge. An approaching Na+ ion would displace the water molecule and bind between the aspartate and glutamate, whereas lysine would move toward the DEKA alanine to provide room for Na+. To test the agreement between our model and this hypothesis, we placed a water molecule in the DEKA locus and imposed the above contacts by constraints. The two-stage MC-minimization demonstrated that the arrangement of the water molecule and DEKA residues is easily reproducible in our model and does not change after removing the constraints (Fig. 8). Numerous mutations in the DEKA locus are known to affect ion selectivity. The consistency of our model with the hypothesis of Lipkind and Fozzard (2000) implies similar explanations of how the mutations would affect the channel selectivity.
|
; Favre et al., 1996) deserves special mention, because these data are related to the problem of the three-dimensional similarity among K+, Na+, and Ca2+ channels. Structural models of the Ca2+ selectivity filter (Lipkind and Fozzard, 2001; Zhorov et al., 2001) propose that four glutamates in the EEEE locus bind two Ca2+. Lipkind and Fozzard (2001) proposed that the side chains of four glutamates converge to form a narrow ring; two Ca2+ ions bind at the top and bottom faces of the ring and permeate in a single-file mode. An alternative model suggests that at millimolar Ca2+ four glutamates split in two pairs, each chelating a single Ca2+ ion (Zhorov and Ananthanarayanan, 1996; Zhorov et al., 2001). To assess the ability of our model to reproduce both concepts, we created the EEEE mutant of the Na+ channel with two Ca2+ ions. Strong electrostatic attractions of two Ca2+ ions to four glutamates provided essential stability to both modes of Ca2+ chelation even without constraints. Therefore, other arguments and other methods of analysis should be employed to discriminate the two patterns of Ca2+ chelation. |
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONSUPPLEMENTARY MATERIALACKNOWLEDGEMENTSREFERENCES |
|---|
|
The sequence alignment between K+ and Na+ channels is of particular importance because any change in the alignment would dramatically affect the model. K+ channels are distant from Na+ and Ca2+ channels and the alignment of their sequences is not a trivial task. In this work we used the alignment (Table 1) previously proposed for the modeling of Ca2+ channels (Zhorov et al., 2001), which align unambiguously with Na+ channels. The same alignment was used in a Ca2+ channel model in which ligand-sensing residues in the outer, inner, and the pore helices were clustered to form a compact ligand-binding site (Yamaguchi et al., 2003). This alignment can also explain the common positions of ligand-sensing residues in the P-loops of the Shaker channel (Rauer and Grissmer, 1999) and Ca2+ channel (Yamaguchi et al., 2003). It deserves mentioning that the same alignment explains the involvement of certain P-loop residues in the binding of dihydropyridine ligands to Ca2+ channels (Yamaguchi et al., 2003) and guanidinium toxins to the Na+ channel.
The selectivity filter in the Na+ channel is essentially wider than in the K+ channel, whereas the pore helices in both channels have the same position. Although strange at first sight, this feature has a simple explanation. The narrow lumen in the K+ channel is lined by the backbones of residues, whose side chains are H-bonded to the side chains of other P-loop residues. In contrast, the Na+ channel selectivity filter is formed by the side chains of residues, whose backbones approach other P-loop residues. The compact packing of the selectivity-filter backbones against the pore helices provides a rather wide lumen in the Na+ channel model.
When TTX and STX bind to the Na+ channel, they approach the narrowest part but do not pass it. Therefore, these toxins cannot serve as probes to estimate the size of the lumen. The critical experimental data regarding the lumen size is the permeability of different organic cations (Sun et al., 1997), which do not pass through the wild-type Na+ channel, but pass through certain mutants. These data are especially important for our model, in which the P-loops are located closer to the pore axis than in other models. To address this question, we calculated MC-minimized energy profiles for TMA pulled through the Na+ channel and its mutants. These calculations did not reveal any critical steric obstacles along the permeation pathway.
Molecular pharmacology of Na+ channel
The energetically optimal complexes of STX and TTX with the Na+ channel provide a list of specific ligand-receptor contacts, some of which have not been previously resolved. Interestingly, practically all H-bonding groups of TTX and STX found H-bonding partners in the channel. This was achieved by the systematic search of the energetically optimal orientation of the ligands in the selectivity-filter region. Imposing of multiple contacts manually is hardly possible. Our results suggest that the multiple electrostatic and H-bonding interactions are responsible for the high activity of guanidinium toxins. The correlation between the predicted ligand-receptor energy of the TTX analogs and their experimental activity implies that the activity decreases with the number of electrostatic and H-bonding ligand-receptor contacts. The van der Waals and solvation contributions to the ligand-receptor energy do not depend significantly on the structure and orientation of the ligands, because different toxins occupy the same space and have approximately the same contact surface.
In this work we did not attempt to model the entire outer vestibule, since this is a standalone task. Therefore, the current model is not expected to predict the optimal orientation of conotoxins and their binding energy. Nevertheless, some aspects of interaction of certain groups of conotoxins, which are critical for their action at the selectivity-filter region, were analyzed. The results of MCM docking agree with multiple specific contacts found in experiments.
Our model is consistent with the cysteine scanning experiments on the Na+ channel pore region (Yamagishi et al., 1997; 2001) although our structural interpretation differs from that proposed in the original publications. In particular, the experiments show that the cysteine substitutions of Phe1236 and Thr1528 at positions immediately preceding the DEKA-locus residues in repeats III and IV, respectively (see Table 1), were accessible for the externally applied reagents (Yamagishi et al., 1997). At first glance, these data conflict with our model, in which Phe1236 and Thr1528 face the inner pore. Interestingly, however, the above residues face the tunnel formed by the inner helices of repeats III and IV and the pore helix in repeat III (Fig. 10). Residues that affect the action of externally applied Na+ and Ca2+ channel ligands also occur in the same tunnel. In the Ca2+ channel model with the pore helices arranged as in the current work, the tunnel is wide enough to accommodate relatively large dihydropyridine molecules (Yamaguchi et al., 2003). Thus, the comparison of the available mutational data with the KcsA x-ray structure suggests that there is a drug-access pathway to the inner pore in Na+ channels between domains III and IV.
|
). The selectivity-filter region in our model is tightly packed against the pore helices and almost completely screens the above residues from the lumen. Cysteine substitutions at these positions, on one hand, would not be readily accessible, but on the other hand, they would significantly affect the packing of the selectivity-filter region against the pore helices. This would result in the rearrangement of the side chains that interact with TTX. Since many specific contacts are critical for TTX binding, the rearrangement would reduce the TTX sensitivity of the mutants.
Limitations of our model
Our modeling results agree with a large number of experiments, but it is necessary to emphasize the obvious limitations of our approach. The quality of any model, which is not based on the high-resolution experimental structure, is affected by the imperfect force field and the limited power of the energy optimization methods. Therefore, the energetics alone cannot serve as the criterion of the model correctness. In these circumstances, it is crucial to define constraints that would harmonize the model with obvious experiments even at the expense of increasing the model energy. This limitation makes time-consuming calculations of free energy of ligand-receptor complexes, which aim to predict the absolute activity of ligands, impractical. However, the ligand-receptor energy calculated in this work correlates with the relative activity in a homologous series of ligands. The absolute value of this energy is overestimated because several important components are not included in the energy expression. Thus, the loss of entropy upon the ligand binding and displacement of cations from the acidic residues would compensate the large enthalpy. Experimental data indicate the importance of such effects (Li et al., 2003). The number and location of the counterions in the pore is unknown and the cost for their displacement is difficult to predict.
In the absence of the Na+ channel x-ray structure, we cannot estimate how correct the predicted torsional angles of the selectivity-filter region may be. But the general disposition of the functional groups in the TTX/STX binding site in our model seems reliable because it explains a wide range of experimental data. Our model cannot predict relative activities of peptide toxins because it does not include extracellular loops. We also did not address the quantitative aspects of selectivity, which would require extensive MD simulations. However, the proposed model could serve as a template for such studies. Coordinates of the model are available (see Supplementary Material).
|
SUPPLEMENTARY MATERIAL |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONSUPPLEMENTARY MATERIALACKNOWLEDGEMENTSREFERENCES |
|---|
|
ACKNOWLEDGEMENTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONSUPPLEMENTARY MATERIALACKNOWLEDGEMENTSREFERENCES |
|---|
This work was supported by a grant to B.S.Z. from the Canadian Institutes of Health Research. B.S.Z. is a recipient of the Canadian Institutes of Health Research Senior Investigator award.
|
FOOTNOTES |
|---|
Submitted on June 23, 2004; accepted for publication October 1, 2004.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONSUPPLEMENTARY MATERIALACKNOWLEDGEMENTSREFERENCES |
|---|
Backx, P. H., D. T. Yue, J. H. Lawrence, E. Marban, and G. F. Tomaselli. 1992. Molecular localization of an ion-binding site within the pore of mammalian sodium channels. Science. 257:248–251.
Becker, S., E. Prusak-Sochaczewski, G. Zamponi, A. G. Beck-Sickinger, R. D. Gordon, and R. J. French. 1992. Action of derivatives of µ-conotoxin GIIIA on sodium channels. Single amino acid substitutions in the toxin separately affect association and dissociation rates. Biochemistry. 31:8229–8238.[CrossRef][Medline]
Berneche, S., and B. Roux. 2001. Energetics of ion conduction through the K+ channel. Nature. 414:73–77.[CrossRef][Medline]
Brooks, C. L., B. M. Pettitt, and M. Karplus. 1985. Structural and energetic effects of truncating long-ranged interactions in ionic polar fluids. J. Chem. Phys. 83:5897–5908.[CrossRef]
Burnashev, N., A. Villarroel, and B. Sakmann. 1996. Dimensions and ion selectivity of recombinant AMPA and kainate receptor channels and their dependence on Q/R site residues. J. Physiol. 496:165–173.[Medline]
Chahine, M., L. Q. Chen, N. Fotouhi, R. Walsky, D. Fry, V. Santarelli, R. Horn, and R. G. Kallen. 1995. Characterizing the µ-conotoxin binding site on voltage-sensitive sodium channels with toxin analogs and channel mutations. Receptors Channels. 3:161–174.[Medline]
Chahine, M., J. Sirois, P. Marcotte, L. Chen, and R. G. Kallen. 1998. Extrapore residues of the S5–S6 loop of domain 2 of the voltage-gated skeletal muscle sodium channel (rSkM1) contribute to the µ-conotoxin GIIIA binding site. Biophys. J. 75:236–246.
Chang, N. S., R. J. French, G. M. Lipkind, H. A. Fozzard, and S. Dudley, Jr. 1998. Predominant interactions between µ-conotoxin Arg-13 and the skeletal muscle Na+ channel localized by mutant cycle analysis. Biochemistry. 37:4407–4419.[CrossRef][Medline]
Choudhary, G., M. Yotsu-Yamashita, L. Shang, T. Yasumoto, and S. C. Dudley, Jr. 2003. Interactions of the C-11 hydroxyl of tetrodotoxin with the sodium channel outer vestibule. Biophys. J. 84:287–294.
Cruz, L. J., W. R. Gray, B. M. Olivera, R. D. Zeikus, L. Kerr, D. Yoshikami, and E. Moczydlowski. 1985. Conus geographus toxins that discriminate between neuronal and muscle sodium channels. J. Biol. Chem. 260:9280–9288.
Dewar, M. J. S., E. G. Zoebisch, E. F. Healy, and J. J. P. Stewart. 1985. AM1: a new general purpose quantum mechanical model. J. Am. Chem. Soc. 107:3902–3909.[CrossRef]
Doyle, D. A., J. Morais-Cabral, R. A. Pfuetzner, A. Kuo, J. M. Gulbis, S. L. Cohen, B. T. Chait, and R. MacKinnon. 1998. The structure of the potassium channel: molecular basis of K+ conduction and selectivity. Science. 280:69–77.
Dudley, S. C., Jr., H. Todt, G. Lipkind, and H. A. Fozzard. 1995. A µ-conotoxin-insensitive Na+ channel mutant: possible localization of a binding site at the outer vestibule. Biophys. J. 69:1657–1665.
Dudley, S. C., Jr., N. Chang, J. Hall, G. Lipkind, H. A. Fozzard, and R. J. French. 2000. µ-Conotoxin GIIIA interactions with the voltage-gated Na+ channel predict a clockwise arrangement of the domains. J. Gen. Physiol. 116:679–690.
Favre, I., E. Moczydlowski, and L. Schild. 1996. On the structural basis for ionic selectivity among Na+, K+, and Ca2+ in the voltage-gated sodium channel. Biophys. J. 71:3110–3125.
Heinemann, S. H., H. Terlau, W. Stuhmer, K. Imoto, and S. Numa. 1992. Calcium channel characteristics conferred on the sodium channel by single mutations. Nature. 356:441–443.[CrossRef][Medline]
Hille, B. 1971. The permeability of the sodium channel to organic cations in myelinated nerve. J. Gen. Physiol. 58:599–619.
Hille, B. 1973. Potassium channels in myelinated nerve. Selective permeability to small cations. J. Gen. Physiol. 61:669–686.
Huang, C. J., I. Favre, and E. Moczydlowski. 2000. Permeation of large tetra-alkylammonium cations through mutant and wild-type voltage-gated sodium channels as revealed by relief of block at high voltage. J. Gen. Physiol. 115:435–454.
Hui, K., G. Lipkind, H. A. Fozzard, and R. J. French. 2002. Electrostatic and steric contributions to block off the skeletal muscle sodium channel by µ-conotoxin. J. Gen. Physiol. 119:45–54.
Jiang, Y., A. Lee, J. Chen, M. Cadene, B. T. Chait, and R. MacKinnon. 2002. The open pore conformation of potassium channels. Nature. 417:523–526.[CrossRef][Medline]
Jiang, Y., A. Lee, J. Chen, V. Ruta, M. Cadene, B. T. Chait, and R. MacKinnon. 2003. X-ray structure of a voltage-dependent K+ channel. Nature. 423.
Kao, C. Y., and S. E. Walker. 1982. Active groups of saxitoxin and tetrodotoxin as deduced from actions of saxitoxin analogues on frog muscle and squid axon. J. Physiol. 323:619–637.
Kao, C. Y. 1986. Structure-activity relations of tetrodotoxin, saxitoxin, and analogues. Ann. N. Y. Acad. Sci. 479:52–67.[Medline]
Khan, A., L. Romantseva, A. Lam, G. Lipkind, and H. A. Fozzard. 2002. Role of outer ring carboxylates of the rat skeletal muscle sodium channel pore in proton block. J. Physiol. 543:71–84.
Kim, M. S., T. Morii, L. X. Sun, K. Imoto, and Y. Mori. 1993. Structural determinants of ion selectivity in brain calcium channel. FEBS Lett. 318:145–148.[CrossRef][Medline]
Kirsch, G. E., M. Alam, and H. A. Hartmann. 1994. Differential effects of sulfhydryl reagents on saxitoxin and tetrodotoxin block of voltage-dependent Na channels. Biophys. J. 67:2305–2315.
Kuo, A., J. M. Gulbis, J. F. Antcliff, T. Raham, E. D. Lowe, J. Zimmer, J. Cuthbertson, F. M. Ashcroft, T. Ezaki, and D. A. Doyle. 2003. Crystal structure of the potassium channel KirBac1.1 in the closed state. Science. 300:1922–1926.
Lazaridis, T., and M. Karplus. 1999. Effective energy function for proteins in solution. Proteins. 35:133–152.[CrossRef][Medline]
Li, Z., and H. A. Scheraga. 1987. Monte Carlo-minimization approach to the multiple-minima problem in protein folding. Proc. Natl. Acad. Sci. USA. 84:6611–6615.
Li, R. A., R. G. Tsushima, R. G. Kallen, and P. H. Backx. 1997. Pore residues critical for µ-conotoxin binding to rat skeletal muscle Na+ channels revealed by cysteine mutagenesis. Biophys. J. 73:1874–1884.
Li, R. A., K. Hui, R. J. French, K. Sato, C. A. Henrikson, G. F. Tomaselli, and E. Marban. 2003. Dependence of µ-conotoxin block of sodium channels on ionic strength but not on the permeating [Na+]: implications for the distinctive mechanistic interactions between Na+ and K+ channel pore-blocking toxins and their molecular targets. J. Biol. Chem. 278:30912–30919.
Li, R. A., I. L. Ennis, R. J. French, S. C. Dudley, Jr., G. F. Tomaselli, and E. Marban. 2001. Clockwise domain arrangement of the sodium channel revealed by µ-conotoxin (GIIIA) docking orientation. J Biol Chem. 276:11072–11077.
Lipkind, G. M., and H. A. Fozzard. 2000. KcsA crystal structure as framework for a molecular model of the Na+ channel pore. Biochemistry. 39:8161–8170.[CrossRef][Medline]
Lipkind, G. M., and H. A. Fozzard. 2001. Modeling of the outer vestibule and selectivity filter of the L-type Ca2+ channel. Biochemistry. 40:6786–6794.[CrossRef]
