Molecular dynamics simulation studies on Ca2+-induced conformational changes of annexin I
Department of Chemistry, East Carolina University, Greenville, NC 27858, USA
1 To whom correspondence should be addressed. E-mail: liyu{at}ecu.edu
 |
Abstract |
|---|
 Top Abstract IntroductionMethodsResultsDiscussionConclusionsReferences |
|---|
Cryo-electron microscopy (EM) and X-ray studies proposed different mechanisms for annexin-induced membrane aggregation. In this work, molecular dynamics (MD) simulation technique was utilized to gain an insight into the calcium-induced conformational changes on annexin I and their implication in membrane aggregation mechanism. MD simulations were performed on the Ca2+-free annexin I with the N-terminal domain buried inside the core (System 1), the Ca2+-bound annexin I without N-terminal domain (System 2) and the Ca2+-bound annexin I with the N-terminal domain exposed (System 3). Our results indicated that calcium binding increases the flexibility of annexin I core domain residues including the calcium coordinating residues. As a result, annexin I was activated to interact with the negatively charged membrane. The exposed N-terminal domain was very flexible and gradually lost the secondary structure during MD simulation, suggesting that the N-terminal may adopt a favorable conformation to bind a second membrane and also explaining the failure of attempts to crystallize the full-length annexin I in the presence of calcium ions. The measured dimensions of the averaged simulation structure of the Ca2+-bound annexin I with the N-terminal exposed (System 3) support the proposed membrane aggregation mechanism based on X-ray studies.
Keywords: annexin I/calcium-binding/conformational changes/MD simulation/membrane aggregation
|
Introduction |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionConclusionsReferences |
|---|
Annexins are a family of calcium binding proteins involved in many important cellular processes. Nearly all annexins share the property of binding to negatively charged phospholipids and cellular membranes in a calcium dependent manner. However, only a few members of annexin family with large N-terminal domain were known to be involved in membrane fusion events (Gerke and Moss, 2002
). Annexins are composed of two principal domains, the divergent N-terminal domain and the conserved C terminal domain or core domain. The core domain composed of four homologous repeats; each repeat consists of five 
-helices (A–E) forming an anti-parallel bundle. The domain arrangement of annexin I is shown in Fig. 1. The convex side of the core domain contains calcium binding sites and faces the membrane when an annexin is associated peripherally with phospholipids. The N-terminal domain of annexins which is unique in sequence and length for each member of the family is responsible for the specificity among the different members of annexin class.
|
Annexin I has been shown to cause membrane aggregation (Meers et al., 1992
, 1993; Wang et al., 1992, 1994; de la Fuente and Parra, 1995). The numerous biochemical studies have shown that N-terminal domain of annexin I is capable of membrane binding and important in annexin I-induced membrane aggregation but is not important for the binding of annexin I to negatively charged phospholipids (Wang et al., 1994; Bitto and Cho, 1999). The second membrane binding site is exposed only upon calcium-dependent membrane binding and is not specific for negatively charged phospholipids (de la Fuente and Ossa, 1997). Experimental studies have also indicated the impact of phosphorylation and removal of the N-terminal on the membrane aggregation (Bitto and Cho, 1999). These experimental results together led to the speculation that the N-terminal domain of annexin I is the secondary membrane binding site and is involved in membrane aggregation.
The Cryo-electron microscopy studies on the annexin-induced junctions formed between DOPG/DOPC liposome and several Annexins have been reported (Lambert et al., 1997). The six stripe junctions between DOPC/DOPG liposome and annexin I or annexin II were assigned as follows: the two outer stripes on each side are due to the membrane and the central two stripes are interpreted as the dimer of Annerin I or annexin II. On the basis of this assignment, the thickness of dimer of annexin I or annexin II is only 60 Å after accounting the width of the headgroup of the lipid. In the assignment made for the five-stripe junction between DOPC/DOPC and annexin I or annexin II monomer layer, the thickness of annexin I monomer is only 
25 Å. These assignments were in conflict with X-ray studies described below.
The complete annexin I structures in the absence of calcium or in the presence of calcium were solved by X-ray crystallography technique (Rosengarth et al., 2001, Rosengarth and Luecke, 2003). The striking difference between the structures lies in the repeat III of core domain and the N-terminal domain. In the absence of calcium, the helix formed by the first 12 residues of the N-terminal domain was buried into the repeat III of the core domain, which replaces the D-helix in repeat III and turns it into a flap over the N-terminal helix. In the presence of calcium, the exposed N-terminal domain is disordered and the D-helix of repeat III is back in place for calcium coordination. According to the X-ray study, an annexin double layer would be at least 70 Å in thickness and 90 Å in the case of full-length annexin I with a thickness of 45 Å per monomer in light of the novel N-terminal domain structure. Therefore, based on X-ray studies, an alternate interpretation of the six-stripe conjunction between DOPC/DOPG liposome and annexin I or annexin II is presented (Rosengarth et al., 2001). The two central stripes of the six-stripe conjunction would be due to annexin molecules randomly attached to one or the other bilayer via their calcium binding convex faces and consequently with their exposed N-terminal domains interacting with the opposing bilayers resulting in an average appearance of two layers. The thinner five-stripe sections observed for both annexin I and annexin II could then be interpreted as junctions where the majority of convex faces of annexin are attached to one of the bilayers resulting in a single asymmetric high-density feature between the phospholipid bilayers.
The computational studies were conducted predominantly on annexin V. Cregut et al. (1998) conducted MD simulation using AMBER on the calcium-free and calcium bound annexin V and the truncated calcium-bound annexin I lacking the N-terminal domain to elucidate the hinge bending motions in these protein systems. Several interesting observations were made. Hinge bending was found for all repeats of the core domain of annexin V and especially was greater for repeat I. Calcium binding increased the hinge bending motion. The results of these studies indicated a significant shape modification of annexin V in the presence of Ca+2. Annexin I and annexin V showed significant differences in dynamics and this difference was attributed to the core residues as the N-terminal domain was excluded in this study. The crystallographic studies indicated that the Ca+2 bind to domain III of annexin V cause a large conformational change. This was also indicated by the molecular dynamics (MD) simulations performed by (Sopkova-de Oliveira Santos et al., 2000) using the CHARMM program. The molecular switch for the large-scale conformational change in repeat III of annexin V upon calcium binding of the core domain was shown involving the aspartic acid residue D-226 (Sopkova-de Oliveira Santos et al., 2001), as indicated by the MD simulation studies (Sopkova-de Oliveira Santos et al., 2000).
In this work, we focus on annexin I-induced membrane aggregation. Although numerous experiments have been performed to understand the annexin I-induced membrane aggregation, the membrane aggregation mechanism is largely unknown. Whether the annexin I dimer or monomer is involved in the membrane aggregation is in dispute. Computational approaches including MD simulation techniques prove to be a very useful tool to understand the dynamics and functions of the complicated biomolecular systems at an atomic level, but their application to investigating the annexin I-induced membrane aggregation mechanism has not been reported. Previous computational studies on the annexins focused on annexin V, which does not cause membrane aggregation. In this work, MD simulations were performed on the three systems of annexin I, Ca2+-free annexin I with N-terminal domain buried inside core (System 1, Fig. 1a), Ca2+-bound annexin I without N-terminal domain (System 2, Fig. 1b) and Ca2+-bound annexin I with N-terminal domain exposed (System 3, Fig. 1c). This study gives insights into annexin I-induced membrane aggregation mechanism.
|
Methods |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionConclusionsReferences |
|---|
System preparation
Starting coordinates for Systems 1 and 2 were obtained from the X-ray structures of full-length annexin I in the absence of calcium (1HM6.pdb) (Rosengarth et al., 2001) and in the presence of calcium (1MCX.pdb) (Rosengarth and Luecke, 2003), respectively. The coordinates for the N-terminal residues are not available in the full-length annexin I in the presence of calcium since the N-terminal domain is disordered. System 1 include 343 residues with the first 42 residues forming the N-terminal domain. It represents the configuration of full-length annexin I with N-terminal buried inside the core domain. System 2 contains 312 residues in which 304 residues are of core domain and the 305–312 residues comprise the eight calcium ions bound to the core domain. Annexin I in this system 2 is devoid of ‘39’ N-terminal domain residues.
The System 3, with the N-terminal exposed, was built using the Insight II software and the X-ray structures of Systems 1 and 2. To form a complete annexin I with Ca2+ ion, the N-terminal from system I was added in an exposed position to the core domain of System 2. The addition was made at residue 40. System 3 has a total of 351 residues of which the first 42 are from the N-terminal domain. There are eight calcium ions in this system with residue numbers from 344–351. This system represents the full-length annexin I in the presence of calcium with the exposed N-terminal domain (Fig. 1c). All the three systems were neutralized and explicitly solvated using TIP3P waters (Jorgensen et al., 1983) in a rectangle box with a minimum distance of 14 Å from protein to the wall the box. The total number of atoms including protein, TIP3P water and the counter ions are 59 789, 57 424 and 79 029 for Systems 1–3, respectively.
All computations including the minimization and MD simulations for the three systems were performed on a 32-processor SGI origin 350. All calculations were performed using AMBER’s all atom force field (ff99) as implemented in AMBER 8 software (Case, 2005). The SANDER module of AMBER program was used for the computation. Simulations were carried out under periodic boundary conditions with a non-bonded cutoff of 10 Å to truncate the VDW non-bonded interactions. The particle-mesh Ewald (Darden et al., 1993; Essmann et al., 1995; Crowley et al., 1997; Sagui and Darden, 1999; Toukmaji et al., 2000) method was used to treat the long-range electrostatic contribution to the force field. The SHAKE algorithm (Ryckaert et al., 1977) was employed to constrain the hydrogen atoms and allow the usage of longer step size of 2 fs.
All the three systems were energy minimized before each MD simulation. A restrained minimization was first performed on solvent while keeping the protein fixed; then the entire system was minimized. In case of System 3, an additional minimization in vacuum was performed prior to the restrained minimization to relieve bad contacts in the exposed N-terminal domain. Also for System 3, the N-terminal domain was left unrestrained during minimizing solvent.
The minimized systems were warmed for 20 ps to a final temperature of 300 K. Constant pressure dynamics was performed on the warmed system for 100 ps. During this step, the density of the systems stabilized. Finally, constant volume dynamics (NVT) was performed on the three systems.
|
Results |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionConclusionsReferences |
|---|
The NVT was performed on the three systems for a time scale of 5426 ps for System 1, 6078.6 ps for System 2 and 4911.2 ps for System 3, respectively. The stabilities of the simulation systems were determined by the conservation of energy and maintenance of a constant average temperature. The backbone root-mean square coordinate deviations (RMSD) between the simulation coordinate snapshots (taken every 0.15 ps) and the minimized initial structure were calculated for the three systems, which were shown in Fig. 2. The plots in the left column of Fig. 2 are for the backbone RMSD of the entire system including N-terminal residues, and the plots in the right column are for the backbone RMSD of the core domain residues with the N-terminal residues deselected in the RMSD calculation.
|
As indicated in Fig. 2, for System 1 with the N-terminal domain buried in the repeat III of core domain, no significant difference in the RMSD values was observed with or without the N-terminal domain residues included. For System 3 with the N-terminal domain exposed outside the core domain, the significant difference in the RMSD was observed with N-terminal residues and without N-terminal residues. With the N-terminal domain included the RMSD of the System 3 reaches round 6.5 Å, and without the N-terminal domain included the RMSD of the System 3 is
2.0 Å. This indicated that in System 3, the N-terminal domain expressed much larger conformational changes from the initial structure than the core domain during the simulation; the N-terminal domain in System 3 has a much greater impact on the system than the N-terminal domain in System 1. To understand calcium-induced conformational changes on annexin I, the backbone root-mean square coordinate fluctuation (RMSF) of each residue in the system was calculated using ‘atomicflunct’ of ‘ptraj’ in the Amber program based on the equilibrated simulations, which are shown in Fig. 3a. Prior to the RMSF calculation using ‘atomicfluct’, the calculation with ‘rms’ of ‘ptraj’ was performed to remove the rot and tran motions from each step. The black plot is for the backbone RMSF of each residue of System 1, the red plot is for System 2 and the green plot is for System 3. As shown in Fig. 3a, the RMSF curves of the three systems showed similar patterns for the core domain residues 41–344. The calcium-free full-length annexin I (System 1) has the lowest RMSF value. The residues 247–252 have much higher RMSF in System 1 than in Systems 2 and 3. These residues 247–252 have the major conformational changes from an unfolded flap to a helical structure due to calcium-binding.
|
The N-terminal domain residues 1–40 in System 3 have much greater RMSF values than in System 1, indicating that the exposed N-terminal domain in System 3 is very flexible. The residue 41 in System 2 showed a high RMSF. This may be attributed to the increased flexibility of this particular residue as a result of truncation of N-terminal domain.
The calcium binding residues showed greater flexibility than the other residues in the core domain of annexin I. The previously reported eight calcium-binding sites (Rosengarth and Luecke, 2003) were indicated in Fig. 3. Clearly Ca2+-coordinating residues are within the peak RMSF regions. Some other residues including residues 171 and 196 are also involved in calcium binding.
On the basis of the simulation trajectories, the B factors were calculated for C atoms and compared with the B factors in the X-ray files for full-length calcium-free annexin I and calcium-bound annexin I without the truncated N-terminal. As shown in Fig. 3b, the qualitative character of the X-ray B factors (the dotted line) is reproduced well by the simulations. This demonstrated that the simulation preparations and the simulations performed in this work using force filed ff99 are reasonable.
The averaged simulation structure was generated for each system based on the simulation snapshots saved every 0.15 ps during the MD simulation. The RMSD between the averaged simulation structure and the minimized starting structure was calculated and is shown in Table I. System 3, the calcium-bound annexin I with the N-terminal domain exposed, showed the largest RMS deviation from the initial structure.
|
The three dimensions of the averaged simulation structures were measured for the three systems and the results are shown in Table I. To measure the three dimensions of the protein, the protein is oriented along the principle moments of inertia. The minimum and maximum coordinates of the protein in the x, y and z direction then give the three dimensions of the protein. Y dimension indicates the distance from the convex side of the core domain to the N-terminal domain for calcium-bound annexin I with the exposed N-terminal domain (System 3), and the distance from the convex side to the concave side of the core domain for calcium-free annexin I and the calcium-bound annexin I without the N-terminal. Our results indicated that the thickness (height) of the calcium-bound annexin I with the exposed N-terminal domain is about 62 Å. Therefore, the protein measuring 60 Å in dimension based on the EM-Cryo-assignment could not be the dimer of annexin I in the six-stripe conjunction formed by DOPC/DOPG liposome and annexin I. In other words, our results support the alternate proposals based on the X-ray studies which signify the role of N-terminal domain as a second membrane binding site.
N-terminal domain conformation
As shown by the backbone RMSD and RMSF plots, the exposed N-terminal domain is very flexible and has large structural changes from the starting structure, which can be observed by the simulation snapshots of System 3 shown in Fig. 1d–h. As shown by the structure snapshots of System 3, the exposed N-terminal domain gradually lost its secondary helical conformation during the simulation, which might be the reason for failure of attempts to crystallize the full-length annexin in the presence of calcium ions. The loss of helical conformation may signify the structural and dynamic instability of the N-terminal domain. The overlay (Fig 1i) of snapshots (Fig 1d–h) illustrated the high flexibility of the N-terminal domain.
Cross-correlation analysis proves to be a very useful tool in the investigation of protein structural correlation based on molecule dynamics simulation. In this work, the correlated motions of annexin I were analyzed using cross-correlation technique for calcium-bound annexin I with the exposed N-terminal domain. The cross-correlation map of backbone C of entire annexin I is shown in Fig. 4. As shown in Fig. 4, the residues 1–23 of the N-terminal domain are negatively correlated with the most part (helix A, B, C and the loops between) of repeat II, and with the most part (except for the helix E) of repeat IV. Repeat I, particularly the first 65 residues of the repeat I, is negatively correlated with repeat II and repeat IV. Repeat II is negatively correlated with repeat I more strongly than with repeat III, and is positively correlated with repeat IV. Repeat III is negatively correlated with repeat II and repeat IV more weakly than repeat I. Repeat IV is positively correlated with repeat II, and is negatively correlated with the repeat I more strongly than with repeat III. In summary, repeat II and repeat IV are positively correlated; both repeats II and IV are negatively correlated with repeat I more strongly than with repeat III; repeat I and repeat III are very weakly correlated. Moreover, the two calcium-binding sites of each repeat are positively correlated. Thus the eight calcium-binding sites of annexin I may be correlated. The implication of the correlated motions of annexin I in membrane aggregation will be investigated in future MD studies with the membrane included.
|
|
Discussion |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionConclusionsReferences |
|---|
Core domain
Annexins have a conserved core domain, which is comprised of four homology repeats. In the absence of calcium, the core domain of annexin I harbors the helix formed by the first 12 residues of the N-terminal domain within the hydrophobic pocket of the repeat III. Our study shows that System 1, the full-length annexin I in the absence of calcium, has the lower RMSF and is less flexible than Systems 2 and 3, the annexin I in the presence of calcium. This may explain why the calcium-free full-length annexin I is inactive and cannot interact with the membrane, as hypothesized by Rosengarth et al. (2001). Annexin I has the calcium-binding sites on the convex side of the core domain. Calcium-binding increases the flexibilities of core domain. Notably, calcium binding increases the flexibilities of the calcium-coordinating residues at the convex side of the core domain. This may explain why annexin I binds with the negatively charged membrane with the convex side of the core domain.
Calcium coordinating residues for annexin I were reported by Rosengarth and Luecke (2003). In the present study, the calcium coordinating residues were found to be more flexible than the remaining residues in the core domain, as shown in Fig. 3a. Calcium coordination is expected to reduce the atomic fluctuations of the calcium binding loops but, in contrary, the calcium-binding increases the flexibility of calcium coordinating residues. Similar phenomenon was found for annexin V by Cregut et al. (1998). The interdependence between the calcium binding sites could be the reason for this increased flexibility of calcium binding residues as observed by Trave et al., in K128E mutant of annexin I. In this work, the cross-correlation analysis on calcium-bound annexin I with the exposed N-terminal domain indicates the correlation of the calcium-binding sites. It is noticed that the RMSF of the calcium-binding residues around 210 is actually decreased due to calcium-binding. This may be consistent with the weak correlation between repeat III and the other structural repeats.
On the basis of the numerous experimental results, the N-terminal domain is speculated to be the secondary membrane binding site generated upon calcium-induced membrane interaction. This can be explained by the structure and dynamics of the N-terminal in different forms of annexin I. In calcium-free annexin I, the N-terminal domain is buried inside the core. Owing to the tight packing of N-terminal domain in the repeat III, the RMSF of the N-terminal domain is low. This makes it incapable of participating in membrane aggregation events. In the presence of calcium, according to Volker Gereke et al., the N-terminal domain is ejected from the core domain and made available to the cytoplasmic-binding partners. The present study has provided evidence for this statement. Significant higher RMSF values were found for the N-terminal domain residues (1–41) in calcium-bound annexin I (System 3), which indicated that the exposed N-terminal domain is very flexible and capable of binding to S100 protein, second annexin I and possibly a second membrane itself. The interaction of N-terminal domain peptide of annexin I with S100C was previously observed (Seemann et al., 1996).
In another study, it was proposed that the failure of attempts to crystallize full-length annexin I in the presence of Ca2+ was due to the presence of two -helices in the N-terminal domain which are free to move around the flexible linker formed by residues 27–41 increasing the solvent accessibility. Our study further strengthened this proposal. From the snapshots obtained throughout the simulation trajectories for the calcium bound annexin I with exposed N-terminal domain, it was noticed that the N-terminal domain gradually lost its secondary structure as the simulation progressed, suggesting that the presence of solvent may cause structural instability in N-terminal domain when it is outside the core domain. The mass spectroscopic analysis performed by Rosengarth and Luecke (2003), provided proof for the fact that the N-terminal domain was disordered in the presence of calcium. Our results were consistent with that of mass spectrometry.
Implication in membrane aggregation
Our results indicate that calcium-binding increases the flexibilities of the core domain, particularly increasing the flexibility of the calcium-coordinating loop residues at the convex side of the core. This makes it possible for annexin I to bind with the negatively charged membrane at the convex side of the core domain. The exposed N-terminal is the most flexible part of annexin I in the presence of calcium, gradually losing its secondary structure during the simulation, which makes it possible for the exposed N-terminal domain to adopt a favorable conformation to interact with a second membrane, thus inducing membrane aggregation. This speculation is consistent with the findings for the N-terminal domain of annexin II. The circular dichroism spectroscopy results showed that a peptide corresponding to the full-length sequence (residues 1–31) of human annexin II N-terminal tail domain adopts mainly a helical conformation in hydrophobic or membrane-mimetic environments, whereas a predominantly random structure is adopted in aqueous buffer (Hong et al., 2003).
Whether the annexin I dimer or the annexin I monomer is involved in the six-stripe junction formed by DOPC/DOPG liposome and annexin I is in dispute. According to Cryo-EM studies, the distance between the centers of mass of the outer bilayer leaflets was 80 Å for annexin I in a six-stripe structure which was interpreted as junction formed by annexin I dimer. The annexin I dimer thickness, after accounting for the width of the lipid headgroup, was 60 Å. Hence, the thickness of annexin I monomer was proposed to be 30 Å or less. However, according to X-ray studies, such an annexin dimer would be 70–90 Å in thickness. To explain this conflict, an alternate proposal was put forward by the X-ray studies, which suggested the formation of membrane aggregates in the presence of annexin I monomer, emphasizing the role played by N-terminal domain as a secondary lipid binding site. In the present study, the dimensions of full-length annexin I with N-terminal exposed were measured from the averaged simulation structure. The measured thickness of annexin I (62 Å) supports the alternate model proposed by X-ray studies.
|
Conclusions |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionConclusionsReferences |
|---|
The impact of the calcium binding on dynamics and structure of annexin I was studied with MD simulation technique. Calcium-binding increases the flexibility of annexin I. Notably, the calcium-coordinating residues at the convex side of the core were more flexible than the remaining residues of the core domain. In the calcium-bound annexin I with the exposed N-terminal, the N-terminal is the most flexible part of the system, indicating its potential role as the secondary membrane binding site in the annexin I-induced membrane aggregation. The measured dimensions of the averaged simulation structure of the full-length calcium-bound annexin I with the exposed N-terminal domain support the annexin I-induced membrane aggregation mechanism proposed based on X-ray studies. Future simulations will provide the details of annexin I interaction with membrane.
|
Footnotes |
|---|
Edited by Rebecca Wade
|
References |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionConclusionsReferences |
|---|
Bitto E., Cho W. Biochemistry (1999) 38:14094–14100.[CrossRef][Web of Science][Medline]
Case D.A. J. Comp. Chem. (2005) 26:1668–1688.[CrossRef]
Cregut D., Drin G., Liautard J.P., Chiche L. Protein Eng. (1998) 11:891–900.
Crowley M.F., Darden T.A., Cheatham T.E. III, Deerfield D.W. II. J. Supercomput. (1997) 11:255–278.[CrossRef]
Darden T., York D., Pedersen L. J. Chem. Phys. (1993) 98:10089–10092.[CrossRef]
de la Fuenta M., Ossa C.G. Biophys. J. (1997) 71:383–387.
de la Fuenta M., Parra V. Biochemistry (1995) 34:10393–10399.[CrossRef][Web of Science][Medline]
Essmann U., Perera L., Berkowitz M.L., Darden T., Lee H., Pedersen L.G. J. Chem. Phys. (1995) 103:8577–8593.[CrossRef]
Gerke V., Moss S.E. Physiol. Rev. (2002) 82:331–371.
Hong Y.H., Won H.-S., Ahn H.-C., Lee J. J. Biochem. (2003) 134:427–432.
Jorgensen W.L., Chandrasekhar J., Madura J.D., Impey R.W., Klein M.L. J. Chem. Phys. (1983) 79:926–935.[CrossRef]
Lambert O., Gerke V., Bader M.-F., Porte F., Brisson A. J. Mol. Biol. (1997) 272:42–55.[CrossRef][Web of Science][Medline]
Meers P., Mealy T., Pavlotsky N., Tauber A.I. Biochemistry (1992) 31:6372–6382.[CrossRef][Web of Science][Medline]
Meers P., Mealy T., Tauber A.I. Biochim. Biophys. Acta (1993) 1147:177–184.[Medline]
Rosengarth A., Luecke H. J. Mol. Biol. (2003) 326:1317–1325.[CrossRef][Web of Science][Medline]
Rosengarth A., Gerke V., Luecke H. J. Mol. Biol. (2001) 306:489–498.[CrossRef][Web of Science][Medline]
Ryckaert J.-P., Ciccotti G., Berendsen H.J.C. J. Comput. Phys. (1977) 23:327–341.[CrossRef]
Sagui C., Darden T.A. Simulation and Theory of Electrostatic Interactions in Solution—Pratt L.R., Hummer G., eds. (1999) Melville, NY: American Institute of Physics. 104–113.
Seemann J., Weber K., Gerke V. Biochem. J. (1996) 319:123–129.[Web of Science][Medline]
Sopkova-De Oliveira Santos J., Fisher S., Guilbert C., Lewit-Bentley A., Smith J.C. Biochemistry (2000) 39:14065–14074.[CrossRef][Web of Science][Medline]
Sopkova-De Oliveira Santos J., Vincent M., Tabaries S., Chevalier A., Kerboeuf D., Russo-Marie F., Lewit-Bentley A., Gallay J. FEBS Lett. (2001) 493:122–128.[CrossRef][Web of Science][Medline]
Toukmaji A., Sagui C., Board J., Darden T. J. Chem. Phys. (2000) 113:10913–10927.[CrossRef]
Wang W., Creutz C.E. Biochemistry (1992) 31:9934–9939.[CrossRef][Web of Science][Medline]
Wang W., Creutz C.E. Biochemistry (1994) 33:275–282.[CrossRef][Web of Science][Medline]
Received August 22, 2007; revised December 13, 2007; accepted December 20, 2007.
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||