PEDS Advance Access published online on June 26, 2007
Protein Engineering Design and Selection, doi:10.1093/protein/gzm029
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Molecular dynamics studies of AChBP with nicotine and carbamylcholine: the role of water in the binding pocket
Structural Bioinformatics and Computational Biochemistry, Department of Biochemistry, The University of Oxford, South Parks Road, Oxford OX1 3QU, UK
1 To whom correspondence should be addressed. E-mail: philip.biggin{at}bioch.ox.ac.uk
 |
Abstract |
|---|
 Top Abstract IntroductionMethodsResults and discussionConclusionsAcknowledgementsReferences |
|---|
The acetylcholine-binding protein (AChBP) is homologous to the ligand-binding domain of the nicotinic acetylcholine receptor (nAChR) and other members of the Cys-loop family of neurotransmitter receptors. The high-resolution X-ray structures of AChBP mean it has been used as a model from which to understand agonist and antagonist binding to nAChRs. We present here a molecular dynamics (MD) study of AChBP with nicotine and carbamylcholine bound. Our results suggest that the ligand imposes rigidity on the binding pocket residues. The simulations also suggest that the protein undergoes breathing motions with respect to the five-fold axis, a motion that has been postulated to be related to gating in the nAChR. We analyzed the behaviour of the water molecules in and around the binding site and found that they occupied five distinct sites within the binding pocket. Water occupied these sites in the absence of ligand, but the presence of ligand increased the probability that a water molecule would be found in these sites. Finally, we demonstrate how the positions of these waters might be used in the design of new ligands by comparing the positions of these sites with other recent structures.
Keywords: nicotinic acetylcholine receptor/computational/conformational change/cys-loop receptor/simulation
|
Introduction |
|---|
|
TopAbstractIntroductionMethodsResults and discussionConclusionsAcknowledgementsReferences |
|---|
The acetylcholine-binding protein (AChBP) is a ligand binding domain (LBD) homologue of the pentameric Cys-loop ligand gated ion channel (LGIC) family of membrane proteins (Brejc et al., 2001
; Smit et al., 2001, 2003; Sixma and Smit, 2003). The nicotinic acetylcholine receptor (nAChR), the serotonin receptor (5HT3), the GABAA,C receptors and the glycine receptor make up the Cys-loop family of LGICs (Ashcroft, 2000; Lester et al., 2004) and are characterized by a conserved pair of cysteines in one of the LBD loops (Karlin, 2002). Each subunit is composed of an N-terminal LBD, a transmembrane (TM) domain and an intracellular (IC) domain. Various combinations of subunits can form homomeric or heteromeric ion channels. The structures of AChBP from different species have been solved by X-ray crystallography with a variety of ligands (Brejc et al., 2001; Brejc et al., 2002; Celie et al., 2004, 2005a, b; Bourne et al., 2005; Hansen et al., 2005) that have allowed important comparisons to be made back to nAChR in particular.
The structure of AChBP (Fig. 1A) reveals that the individual subunits are comprised of one 
-helix, 10 stranded ß sheets and two 310 helices (Brejc et al., 2001; Sixma and Smit, 2003; Smit et al., 2003). The binding site is located between two adjacent subunits, a principle subunit (on the left in Fig. 1A) and a complementary subunit (on the right in Fig. 1A). Thus, there are up to five binding sites within the pentameric unit. In the nAChR, at least two of the subunits must be of subtype which forms the principle subunit of the binding site. In some cases, the receptor can be comprised of all subunits (for example receptors comprised of the 7 subunit), and thus will possess five binding sites.
|
The LBD of nAChR has high sequence identity with AChBP from Lymnaea stagnalis (Sixma and Smit, 2003
; Smit et al., 2003). The nAChR is found in the nervous system at neuronal–neuronal synapses and in the muscle at neuromuscular junctions (Nicke et al., 2004). The nAChR plays an important role in nicotine addiction and is implicated in various neurological and neuromuscular diseases. Mutations in the nAChR can result in myasthenic syndromes and frontal lobe epilepsy (De Fusco et al., 2000; Steinlein, 2001), and various cognitive malfunctions (Ashcroft, 2000). This channel is implicated in Parkinson's and Alzheimer's disease, but these mechanisms are not well understood (Hogg and Bertrand, 2004). The nAChR is a cationic channel which allows the passage of Na+, K+ and less commonly Ca2+ ions.
Agonist binding to nAChRs initiates conformational changes in the receptor that are transmitted to the TM domain which ultimately results in the opening of the ion channel to allow the conduction of ions (Karlin, 2002). The mechanism that leads from ligand binding to channel opening is incompletely understood (Cymes et al., 2002; Grosman, 2003). The AChBP structures have provided important clues as to how different ligands are recognized by this fold, but the precise extent and nature of the binding pocket and its behaviour is still poorly understood.
Molecular dynamics (MD) provides a useful way to examine the dynamic behaviour of proteins and indeed has been recently applied to AChBP (Gao et al., 2005) and also to models of the 7 nAChR (Henchman et al., 2003; Law et al., 2005), where the focus was on examining the large-scale motion of the protein thought to be related to gating in nAChRs. Steered molecular dynamics (SMD) has also been used to examine the motions (primarily of loop C) associated with ligand unbinding (Zhang et al., 2006) and more recently, normal-mode analysis (NMA) has been used to examine the opening motions of a model of the 7 nAChR leading to a proposed open-channel model (Cheng et al., 2006). Here though, we focus not on the large global motions but on the dynamics of residues and water in the binding pocket of the AChBP. Our results show that (i) a more stable structure of the binding pocket is maintained in the presence of a ligand; (ii) ligands exhibit varying degrees of mobility within the binding pocket—nicotine seems to prefer one mode of binding while carbamylcholine shows some flexibility of its chain; and (iii) waters exist in discrete pockets inside the binding pocket that maybe useful in the design of ligands. We illustrate this latter point with respect to the crystal structures of AChBP in complex with three toxins.
|
Methods |
|---|
|
TopAbstractIntroductionMethodsResults and discussionConclusionsAcknowledgementsReferences |
|---|
MD simulations of AChBP from L. stagnalis (Celie et al., 2004
) with and without nicotine and carbamylcholine were carried out for 10 ns each. The simulation with nicotine had five nicotine molecules bound, one to each binding site. The simulation with carbamylcholine had two molecules bound to the binding pockets located between subunits C and D and subunits D and E (the three remaining binding pockets were ligand-free; see Table I).
|
MD simulations were carried out with GROMACS v 3.1.4 (www.gromacs.org) (Berendsen et al., 1995
; Lindahl et al., 2001) using the GROMOS96 (van Gunsteren et al., 1996) force-field. Crystal structures of AChBP from L. stagnalis with nicotine (Fig. 1B) (PDB code: 1UW6) and AChBP with carbamylcholine (Fig. 1B) (PDB code: 1UV6
[PDB]
) (Celie et al., 2004) were used including their crystal waters (within 4 Å of the protein) for the liganded simulations. Apo structures were generated by removing the ligand. Topologies were generated for both nicotine and carbamylcholine using PRODRG (Schuettelkopf and van Aalten, 2004). The charges for nicotine and carbamylcholine were calculated with 6-31G* basis sets using Spartan (Hehre and Kluzinger, 1997). The pyrrolidine nitrogen (Fig. 1B) was assumed to be protonated (Graton et al., 2002; Pankow et al., 2003). Ligand topologies were tested by simulation in a box of approximately 900 simple point charge (SPC) (Hermans et al., 1984) water molecules for 1 ns.
Four protein simulations were carried out in total (see Table I) corresponding to both ligand-bound and apo conditions. Each system was energy-minimized until convergence using steepest descents algorithm. Energy minimization was carried out with position restraints, then MD with position restraints for 100 ps, followed by energy minimization again without position restraints, then equilibration for 1 ns and finally the production run of 10 ns. During the equilibration phase, the temperature and pressure was coupled with the Berendsen methods (Berendsen et al., 1984). During the production runs, the Parinello–Rahman (Parinello and Rahman, 1981) method was used for pressure coupling and the temperature was coupled using the Nosé-Hoover (Nose, 1984) method at 310 K. Electrostatics were calculated with the Particle Mesh Ewald (PME) method (Darden et al., 1993). The LINCS algorithm (Hess et al., 1997) was used to constrain bond lengths and a time step of 2 fs was used throughout. Protein-ligand interactions were visualised with LigPlot (Wallace et al., 1995) and VMD (Humphrey et al., 1996). Persistent water molecules were identified via an in-house program. The water density calculations were carried out using Gromacs 3.2.1 and an in house program g_ri3D (Oliver Beckstein) and images were generated with Chimera (Pettersen et al., 2004).
|
Results and discussion |
|---|
|
TopAbstractIntroductionMethodsResults and discussionConclusionsAcknowledgementsReferences |
|---|
Global motion
Although our focus was on the binding pocket and the behaviour of water molecules it is still useful to briefly examine the global motions in the four simulations. The overall dynamics of the receptor can be assessed by estimating the mean square fluctuations (MSFs) of the C atoms (Fig. 1C). This approach calculates the average fluctuation in a given window and can be used as an indicator of the level of sampling for a given conformational state (Faraldo-Gomez et al., 2004). The simulations without ligands showed more movement in the MSF plots than the simulations with ligands bound, perhaps indicating a degree of structural stability resulting from the presence of the ligand. However, in all simulations, the MSF values continue to increase which indicates that the conformational dynamics of the protein as a whole are far from being fully sampled.
Despite this, principal component analysis (PCA) of the large-scale motions from the simulations with ligands revealed a breathing motion of the receptor as has been recently observed for a model of the 7 nAChR (Henchman et al., 2005). The PCA data also revealed greater movement in two of the five subunits in both simulations with ligands bound, even in the case of the simulation where all five subunits were occupied with nicotine. These observations agree with previous MD studies of the 7 nAChR (Henchman et al., 2003) and the suggested gating mechanism of the Torpedo marmorata acetylcholine receptor (Unwin, 2002, 2005; Miyazawa et al., 2003). This asymmetrical motion is also observed to a smaller extent in the non-liganded (NCT-apo, CCE-apo) simulations.
As it is unclear how agonist binding leads to structural changes and channel gating in the homologous nACh receptor, we examined the dynamic behaviour of the binding pocket in the AChBP. In order to assess this, we calculated an RMSD of a subset of atoms known to form part of the binding pocket (specifically M114:N, W143:O, W143:HE1, T144:O and Y192:HH). The results are shown in Fig. 2, which shows that the RMSD for these atoms is higher in the case when the ligand is not present compared to their corresponding ligand-bound structures (note that the two apo simulations exhibit different RMSDs suggesting sampling is incomplete on this timescale). However, the relative (comparing liganded to its corresponding apo state) RMSDs in Fig. 2A and B indicate that there is some stabilizing effect of these atoms, although there are large periods of time when the RMSD of both apo and liganded simulations is similar suggesting that the apo states are still close to the initial structures. Figure 2 also shows that the stabilizing effect appears to be greater for carbamylcholine (Fig. 2B) where these protein atoms have a reduced RMSD compared to when nicotine is bound (Fig. 2A). Thus, simulations where the binding pocket was occupied by a ligand exhibited smaller fluctuations suggesting that the ligand brings about a structural rigidity to the binding pocket in agreement with the RMSF analysis in Fig. 1C.
|
Ligand flexibility
We examined the flexibility of the ligands within the binding site. The ligands for both the NCT and CCE simulations maintained their position within the binding pocket for the duration of the simulation. Although both ligands held their place and general orientation in the binding pocket, nicotine was much more rigid than carbamylcholine. Analysis of the dihedral angles (data not shown) showed a fairly rigid conformation for nicotine, consistent with only one binding mode that differs only by a slight tilting of the 6' ring. The dihedral data for carbamylcholine show more flexibility of the chain. The molecule maintains its orientation, but there are rotations at both ends of the ligand suggesting multiple conformations may be equally favourable within the pocket. Docking studies (Amiri et al., unpublished) also suggest that nicotine has one mode of binding while carbamylcholine may have more than one mode of binding.
We wanted to look at the role of water in the interactions of the ligand with the residues in the binding pocket. Other studies have alluded to the role of water in the binding pocket of another class of ligand-gated ion channel; namely the ionotropic glutamate receptors (Speranskiy and Kurnikova, 2004, 2005; Arinaminpathy et al., 2006; Kaye et al., 2006), and thus we wanted to further investigate its role in these simulations. Visual inspection of the trajectories revealed that in most subunits, water seems to play a significant role in the ligand's interaction with surrounding residues. Indeed, some of these bridging waters are seen throughout the entire simulation. Figure 3 shows both ligands interacting with surrounding residues via water molecules. In the nicotine-bound (NCT) and carbamylcholine (CCE)-bound simulations, a water molecule bridging the ligand and L102 or M114 on loop E is present for more than 92% of the time averaged for the subunits with ligands and sometimes the same water bridges both of these residues at the same time. Also commonly seen are water molecules bridging the ligand with W143 or T144 or both at the same time. Bridging waters also exist between the cysteines on loop C and the ligands. The cysteines on loop C from -subunits are thought to be functionally important as they are evolutionarily highly conserved. Recent MD data has shown that loop C comes in and covers the binding pocket in the presence of a ligand (Henchman et al., 2005) and thus the hydrogen bonding with the ligand could play an important role in determining the behaviour of loop C. Other bridging waters commonly exist between the ligand and Q55 and with Y164. Y164 has its side-chains facing upwards towards the ligand in the binding site which enables hydrogen bonding with the ligand. We quantified for what percentage of time these waters occupy different locations in the binding pocket (Table II). Those that occupied a position in the binding cleft for greater than 40% of the time we term ‘persistent’. Greater numbers of persistent waters were found in the simulations with ligand present (NCT and CCE). For the carbamylcholine simulation with only two ligands bound in two binding pockets, the greatest number of persistent waters was found in the liganded binding pockets throughout the simulation. For the binding pockets that do not possess a ligand, waters were observed, but were substantially more mobile and stayed for a much shorter amount of time than waters in the binding sites with ligands present.
|
|
In order to compare the distribution of water molecules in the binding site with bulk, we calculated the average water density throughout the simulations. Figure 4A shows the average water density plot for the simulation of one of the binding pockets of the AChBP with nicotine bound. Compared to bulk water, the average density of water molecules in the binding pocket appears to be discretised into distinct zones indicating that these areas are where water molecules are preferentially found over time. These density plots were compared with the positions of the persistant waters. Taking the visual inspection data, the average water density analysis and the persistency data enable us to describe five separate zones where water prefers to reside (Fig. 4B). These zones were populated frequently and for substantial stretches of time. Zone 1 lies within loop E between L102 and M114. This location is almost always occupied with at least one water molecule >90% of the time. Waters in this position have also been observed in some crystallography studies (Celie et al., 2004
). Water molecules can be seen ‘competing’ with each other for this location, and if one water molecule leaves the site, another will usually replace it very quickly, suggesting a strong preference for a water molecule in this area. Zone 2 is the area near the W143/T144 loop (ß8 sheet) where there is a bridging water connecting the ligand with either one or both residues in both cases of nicotine and carbamylcholine. Waters are commonly found in this position for subunits with ligands bound. Zone 3 is at the base of loop C near E193 and Y192. Y192 is heavily involved in hydrogen bonding to the ligand. Waters in this zone may play a role in keeping the integrity of loop C and restrict it from moving out freely in the presence of a ligand. Zone 4 describes the low ß5 sheet (loop A) of the binding pocket with residues Y89 and I92 that have been thought to be involved in ligand binding from previous experimental studies (Celie et al., 2005a, b). In several nanoseconds of the liganded simulations, this location is occupied with one or more water molecules, however this is seldom the case in apo simulations. Zone 5 is near the tip of loop C where there are two adjacent cysteines forming a disulfide bridge. These cysteines have been the focus of many studies and are highly conserved in the -subtype of nAChRs. It is believed that the cysteines interact with the ligands in the binding pocket of nAChR and can be seen to ‘close off’ the binding pocket in the presence of a ligand. There are several bridging waters in position between the cysteines and the ligands, and this interaction is often mediated by water molecules.
|
We examined the consequence on the structure if a water molecule in one of these zones moves out. Figure 5A shows the typical behaviour of waters into and out of a zone (Zone 1). Here we follow three water molecules for the duration of the simulation. Two waters are positioned in Zone 1 between residues L102 and M114 on loop E. A third water comes in and knocks one of the two waters out, and replaces it in the same exact position where it remains for most of the simulation. Upon leaving (just before the 9 ns mark), the ß-sheets of loop E collapse closer into each other and the nature of interactions with the ligand is altered (Fig. 5B).
|
Recently, several high resolution X-ray crystal structures of an AChBP in complex with agonists and antagonists have become available (Bourne et al., 2005
; Hansen et al., 2005; Ulens et al., 2006). Although obtained from a different species (Aplysia californica), the binding pocket is well conserved (Hansen et al., 2005) and so a direct comparison is useful. Comparative analysis of these structures demonstrated that loop C clusters into three distinct positions depending on the nature of the compound bound: (i) agonists bound (as exemplified by epibatidine or lobeline); (ii) the apo or alkaloid antagonist-bound structures (as exemplified by methyllycaconitine); or (iii) structures with peptide antagonists bound (ImI complex). We were interested to see how the positions of these compounds related to the positions of waters in the zones described above. We overlayed the structures of these compounds onto the water-density obtained from the NCT simulation (Fig. 6) to investigate this.
|
For the two agonists, epibatidine (a compound produced by the Ecuadorian frog, Epipedobates tricolor) and lobeline (an alkaloid from the Indian tobacco plant, Lobelia inflate), the position in the binding pocket with respect to the water zones is quite similar. The crystal structure of epibatidine bound to AChBP (PDB:2BYQ) shows that the nitrogen in the azobicycloheptane ring is close to Zone 1 (Fig. 6A). This toxin resembles the structure of nicotine and is bound in the same way inside the binding pocket with its 6' ring facing the high water density regions in Zone 1. Thus, we expect to see the same interactions with water molecules and surrounding residues. The other nitrogen atom lies near water densities in Zone 2, behind loop B. Lobeline is positioned in the pocket such that is quaternary nitrogen is in a similar position to epibatidine and nicotine. There is also potential interaction with water molecules in Zone 1 (Fig. 6B). This is confirmed by the presence of a water molecule in the crystal structure. Thus, the positions of these zones, particularly Zone 1 appears to most relevant for the agonist-bound conformation of the protein.
For methyllycaconitine, which is an alkaloid antagonist against nAChRs, there is no indication that the zones derived here play a role in stabilizing the complex (Fig. 6C). In the case of the -conotoxin ImI (PDB: 2BYP), the position of the toxin suggests that Zones 1 and 3 could play a role in stabilizing the complex (Fig. 6D). However, it is important to remember, for this structure, that the loop C is pulled out radially with respect to the central axis and with respect to the agonist-bound conformations from which the water zones were derived. A direct comparison of the zones derived here with the antagonist-bound structures should be treated with caution given the differences in the shape of the binding pocket, but we would suggest from this analysis that these zones become less important with respect to antagonist binding.
|
Conclusions |
|---|
|
TopAbstractIntroductionMethodsResults and discussionConclusionsAcknowledgementsReferences |
|---|
Our results indicate how the dynamics of the ligand in conjunction with the presence of explicit water molecules are an important consideration for our understanding of AChBP and the related Cys-loop receptors. The ligands are held fairly rigidly in the binding pockets. Furthermore, the presence of the ligands confers some stability on the binding site residues. The effect is aided by the presence of water molecules which form bridging hydrogen bonds with key residues. In the apo simulations, the binding site is flexible (higher RMSD values) with varying distances between neighbouring residues and more flexible loops. Loop C which is less mobile in liganded simulations, makes hydrogen bonds with ligands through water molecules. Loops A and B are held more rigid in the presence of water molecules which are more persistent in the presence of a ligand. These bonding patterns have structural implications for the binding site which are likely to be important in the homologous nAChR proteins. The location of zones in the binding sites where waters persist with nicotine and carbamylcholine bound suggest an extended pharmacophore that could be used in the design of new ligands. Furthermore, such a pharmacophore will be energetically stable within the binding pocket. This pharmacophore can be interpreted in two ways. The first is that the zones could offer a site where the water molecule is replaced by a chemical moiety. The second is where the zone is occupied by a water molecule and the ligand makes interactions with the protein via such a water (as exemplified by Zone 1, which appears to aid in the binding of nicotine, lobeline and epibatidine). Finally, these observations should be readily transferable to nAChRs in general and aid in new compounds targeted against them.
|
Footnotes |
|---|
Edited by Klaus Schulten
|
Acknowledgements |
|---|
|
TopAbstractIntroductionMethodsResults and discussionConclusionsAcknowledgementsReferences |
|---|
We thank the Wellcome Trust and Oxford Supercomputer Centre for supporting this work.
|
References |
|---|
|
TopAbstractIntroductionMethodsResults and discussionConclusionsAcknowledgementsReferences |
|---|
Arinaminpathy Y., Sansom M.S.P., Biggin P.C. Mol. Pharm. (2006) 69:11–18.
Ashcroft F.M. Ion Channels and Disease (2000) San Diego: Academic Press. 481.
Berendsen H.J.C., Postma J.P.M., van Gunsteren W.F., DiNola A., Haak J.R. J. Chem. Phys. (1984) 81:3684–3690.[CrossRef]
Berendsen H.J.C., van der Spoel D., van Drunen R. Comp. Phys. Commun. (1995) 95:43–56.
Bourne Y., Talley T.T., Hansen S.B., Taylor P., Marchot P. EMBO J. (2005) 24:1512–1522.[CrossRef][Web of Science][Medline]
Brejc K., van Dijk W.J., Klassen R.V., Schuurmans M., van Der Oost J., Smit A.B., Sixma T.K. Nature (2001) 411:269–276.[CrossRef][Medline]
Brejc K., van Dijk W.J., Smit A.B., Sixma T.K. Novartis Found Symp. (2002) 245:22–29.[Web of Science][Medline]
Celie P.H., et al. Nat. Struct. Mol. Biol. (2005a) 12:582–588.[CrossRef][Web of Science][Medline]
Celie P.H., Klassen R.V., van Rossum-Fikkert S.H., van Elk R., van Nierop P., Smit A.B., Sixma T.K. J. Biol. Chem. (2005b) 280:26457–26466.
Celie P.H., van Rossum-Fikkert S.H., van Dijk W.J., Brejc K., Smit A.B., Sixma T.K. Neuron (2004) 41:907–914.[CrossRef][Web of Science][Medline]
Cheng X., Lu B., Grant B., Law R.J., McCammon J.A. J. Mol. Biol. (2006) 355:310–324.[CrossRef][Web of Science][Medline]
Cymes G.D., Grosman C., Auerbach A. Biochemistry (2002) 41:5548–5555.[CrossRef][Medline]
Darden T., York D., Pedersen L. J. Chem. Phys. (1993) 98:10089–10092.[CrossRef]
De Fusco M., Becchetti A., Patrignani A., Annesi G., Gambardella A., Quattrone A., Ballabio A., Wanke E., Casari G. Nat. Genet. (2000) 26:275–276.[CrossRef][Web of Science][Medline]
Faraldo-Gomez J.D., Forrest L.R., Baaden M., Bond P.J., Domene C., Patargias G., Cuthbertson J., Sansom M.S.P. Proteins Struct. Funct. Genet. (2004) 57:783–791.[CrossRef][Web of Science][Medline]
Gao F., Bren N., Burghardt T.P., Hansen S.B., Henchman R.H., Taylor P., McCammon J.A., Sine S.M. J. Biol. Chem. (2005) 280:8443–8451.
Graton J., Berthelot M., Gal J., Girard S., Laurence C., Lebreton J., Le Questel J., Maria P., Naus P. J. Am. Chem. Soc. (2002) 124:10552–10562.[CrossRef][Web of Science][Medline]
Grosman C. Biochemistry (2003) 41:5548–5555.[CrossRef]
Hansen S.B., Sulzenbacher G., Huxford T., Marchot P., Taylor P., Bourne Y. EMBO J. (2005) 1–12.
Hehre W.J.Y., Kluzinger P. A guide to molecular mechanics and molecular orbital calculations in Spartan (1997) Irvine, CA: Wavefunction Inc.
Henchman R.H., Wang H.L., Sine S.M., Taylor P., McCammon J.A. Biophys. J. (2005) 88:2564–2576.[CrossRef][Web of Science][Medline]
Henchman R.H., Wang H.-L., Sine S.M., Taylor P., McCammon J.A. Biophys. J. (2003) 85:3007–3018.[Web of Science][Medline]
Hermans J., Berendsen H.J.C., van Gunsteren W.F., Postma J.P.M. Biopolymers (1984) 23:1513–1518.[CrossRef][Web of Science]
Hess B., Bekker J., Berendsen H.J.C., Fraaije J.G.E.M. J. Comp. Chem. (1997) 18:1463–1472.[CrossRef]
Hogg R.C., Bertrand D. Curr. Drug Targets CNS Neurol. Disord. (2004) 3:123–130.[CrossRef][Medline]
Humphrey W., Dalke A., Schulten K. J. Molec. Graph. (1996) 14:33–38.[CrossRef]
Karlin A. Nat. Rev. Neurosci. (2002) 102:102–114.
Kaye L.S., Sansom M.S.P., Biggin P.C. J. Biol. Chem. (2006) 281:12736–12742.
Law R.J., Henchman R.H., McCammon J.A. Proc. Natl Acad. Sci. USA (2005) 102:6813–6818.
Lester H.A., Dibas M.I., Dahan D.S., Leite J.F., Dougherty D.A. Trends Neurosci. (2004) 27:329–336.[CrossRef][Web of Science][Medline]
Lindahl E., Hess B., van der Spoel D. J. Mol. Model (2001) 7:306–317.
Miyazawa A., Fujiyoshi Y., Unwin N. Nature (2003) 424:949–955.
Nicke A. Eur. J. Biochem. (2004) 271:2293.[Web of Science][Medline]
Nose S. Mol. Phys. (1984) 52:255–268.[CrossRef][Web of Science]
Pankow J.F., Barsanti K.C., Peyton D.H. Chem. Res. Toxic. (2003) 16:23–27.[CrossRef][Web of Science][Medline]
Parinello M., Rahman A. J. Appl. Phys. (1981) 52:7182–7190.[CrossRef][Web of Science]
Pettersen E.F., Goddard T.D., Huang C.C., Couch G.S., Greenblatt D.M., Meng E.C., Ferrin T.E. J. Comp. Chem. (2004) 25:1605–1612.[CrossRef]
Schuettelkopf A.W., van Aalten D.M.F. Acta Cryst. D (2004) 60:1355–1363.[CrossRef][Medline]
Sixma T.K., Smit A.B. Annu. Rev. Biophys. Biomol. Struct. (2003) 32:311–334.[CrossRef][Web of Science][Medline]
Smit A.B., Brejc K., Syed N.I., Sixma T.K. Ann. N. Y. Acad. Sci. (2003) 998:81–92.[CrossRef][Web of Science][Medline]
Smit A.B., et al. Nature (2001) 411:261–268.[CrossRef][Medline]
Speranskiy K., Kurnikova M. J. Chem. Phys. (2004) 121:1516–1524.[CrossRef][Medline]
Speranskiy K., Kurnikova M. Biochemistry (2005) 44:11508–11517.[CrossRef][Medline]
Steinlein O.K. Am. J. Med. Genet. (2001) 106:139–145.[CrossRef][Web of Science][Medline]
Ulens C., Hogg R.C., Celie P.H., Bertrand D., Tsetlin V., Smit A.B., Sixma T.K. Proc. Natl Acad. Sci. USA (2006) 103:3615–3620.
Unwin N. Novartis Found Symp. (2002) 245:5–15.[Web of Science][Medline]
Unwin N. J. Mol. Biol. (2005) 346:967–989.[CrossRef][Web of Science][Medline]
van Gunsteren W.F., Krüger P., Billeter S.R., Mark A.E., Eising A.A., Scott W.R.P., Hüneberger P.H., Tironi I.G. Biomolecular simulation: the GROMOS96 manual and user guide. Groningen/Zürich: Biomos/Hochschulverlag AG an der ETH Zürich. (1996).
Wallace A.C., Laskowski R.A., Thornton J.M. Prot. Eng. (1995) 8:127–134.
Zhang D., Gullingsrud J., McCammon J.A. J. Am. Chem. Soc. (2006) 128:3019–3026.[CrossRef][Web of Science][Medline]
Received April 13, 2007; revised April 13, 2007; accepted April 26, 2007.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  L. M. Sharkey and C. Czajkowski Individually Monitoring Ligand-Induced Changes in the Structure of the GABAA Receptor at Benzodiazepine Binding Site and Non-Binding-Site Interfaces Mol. Pharmacol., July 1, 2008; 74(1): 203 - 212. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||