proteins STRUCTURE O FUNCTION O BIOINFORMATICS
The N-terminal of annexin A1 as a secondary membrane binding site: A molecular dynamics study Matthew P. Donohue, Libero J. Bartolotti, and Yumin Li* Department of Chemistry, East Carolina University, Greenville, North Carolina 27858
ABSTRACT Annexin A1 has been shown to cause membrane aggregation and fusion, yet the mechanism of these activities is not clearly understood. In this work, molecular dynamics simulations were performed on monomeric annexin A1 positioned between two negatively charged monolayers using AMBER’s all atom force field to gain insight into the mechanism of fusion. Each phospolipid monolayer was made up of 180 DOPC molecules and 45 DOPG molecules to achieve a 4:1 ratio. The space between the two monolayers was explicitly solvated using TIP3P waters in a rectilinear box. The constructed setup contained up to 0.14 million atoms. Application of periodic boundary conditions to the simulation setup gave the desired effect of two continuous membrane bilayers. Nonbonded interactions were calculated between the N-terminal residues and the bottom layer of phospholipids, which displayed a strong attraction of K26 and K29 to the lipid head-groups. The sidechains of these two residues were observed to orient themselves in close proximity (~3.5 A˚) with the polar head-groups of the phospholipids. Proteins 2014; 82:2936–2942. C 2014 Wiley Periodicals, Inc. V
Key words: interaction energy; membrane aggregation; domain flexibility; side-chain reaction; molecular dynamics.
INTRODUCTION Interactions between proteins and membrane surfaces involve complex molecular mechanisms, which are critical in allowing the cell to function properly, yet remain to be fully elucidated in cell biology. Anchoring of many peripheral proteins requires electrostatic or ionic interactions with membrane lipids. Annexins constitute a family of proteins that have the ability to bind onto negatively charged phospholipid bilayers in a reversible and calcium dependent manner through the unique architecture of their calcium binding sites. In addition to membrane binding, annexins with relatively larger N-terminal domains (between 30 and 100 residues) have been identified to cause membrane aggregation and fusion events.1 Annexins are structurally divided into a conserved core domain, which has the shape of a slightly curved disc, and a divergent N-terminal that is unique for a given member of the family. The core domain comprises four (in annexin A6 eight) homologous repeats (labeled I–IV) of about 75 amino acid residues that fold into five alpha-helices (A–E) and form an anti-parallel bundle (Fig. 1). High-resolution crystal structures have identified the calcium binding sites to be located on the convex
2936
PROTEINS
face of the protein. As shown in Figure 1, the calcium ions are depicted as light-blue spheres. The bound calcium ions serve as a hypothetical “bridge” between the protein and membrane by simultaneously coordinating ligands from acidic side chains of the protein and from phosphoryl moieties of the lipids.2 The N-terminal is variable in sequence and length for given members of the family, and is thought to regulate the specific physiological functions of each annexin.1 Annexin A1, a 37 kDa protein previously known as lipocortin, has been identified to cause membrane aggregation.3–7 The N-terminal domain of annexin A1 contains 41 residues. Biochemical studies involving the deletion and phosphorylation of the N-terminal have demonstrated that the N-terminal is important for membrane aggregation, but is of little significance for binding to negatively charged phospholipids.8 Further, a chimera comprising the core domain of annexin V (which does not promote membrane fusion activity and has a shorter *Correspondence to: Yumin Li, Department of Chemistry, East Carolina University, Greenville, North Carolina 27858. E-mail: [email protected] Received 24 January 2014; Revised 19 May 2014; Accepted 28 May 2014 Published online 10 June 2014 in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/prot.24623
C 2014 WILEY PERIODICALS, INC. V
N-Terminal of Annexin A1 as a Secondary Binding Site
Figure 1 Annexin A1 is positioned between two negatively charged monolayers. Annexin A1 is color coded as red (repeat I), green (repeat II), blue (repeat III), yellow (repeat IV), and black (the N-terminal). The location of K26 and K29 within the protein is indicated with an arrow. Calcium ions are shown as light-blue spheres. A 90 rotation provides an axial view of the protein.
N-terminal with only 19 residues) and the N-terminal of annexin A1 was found to cause membrane aggregation.9 The results of these studies led to the supposition that the N-terminal of annexin A1 harbors the second membrane-binding site of the protein. Cryo-electron microscopy studies reported two types of junctions formed connecting DOPG/DOPC liposomes and annexin A1.10 The 160 A˚ thick six stripe junctions were more frequently observed than the 125 A˚ thick five stripe junctions. The six striped junctions were interpreted as follows: the two outer stripes on each side were attributed to the membrane inner and outer leaflets, while the two inner stripes were due to an annexin A1 or A2 dimer. The five stripe junctions attributed the central stripe as an annexin A1 or A2 monomer. The
distances between the centers-of-mass of the outer bilayer leaflets were shown to be 80 A˚ (six-stripe junctions), which after accounting for the width of the lipid head-groups corresponds to a protein thickness of 60 A˚. In the case of the five-stripe junctions, a thickness of 25 A˚ was assigned to the annexin A1 monomer. These results were in conflict with X-ray crystallography techniques, and whether membrane aggregation is caused by the annexin A1 monomer or dimer remains unclear. The structures of porcine annexin A1 both in the presence and absence of calcium have been solved using Xray crystallography techniques.11,12 The results identified a drastic change in structure between the conformations of the N-terminal and repeat III of the core domain. In the absence of calcium, the amphipathic alpha helix formed by residues 2–12 of the N-terminal domain is buried inside repeat III of the protein core, replacing the D-helix, which forms a flap over the N-terminal. Upon calcium binding, the D-flap folds into the proper conformation in repeat III, forming a Ca21 binding site and ejecting the N-terminal. Based on crystal structures, an annexin A1 double layer between two lipid bilayers would be at least 70 A˚ thick and about 90 A˚ (accounting for the N-terminal) in the case of full length annexin A1. Explanation of the six stripe junctions by Rosengarth et al. proposed that the two central stripes were due to annexin A1 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 bilayer, resulting in an average appearance of two layers.11 The thinner fivestripe junctions could then be interpreted as sections where the majority of the convex faces of annexins are attached to one of the bilayers, resulting in a single asymmetric high-density feature between the phospholipid bilayers. Previous computational studies on annexins conducted by Cregut et al. were designed to study the hinge bending motions of calcium-bound annnexin V and A1 and apo-annexin V.13 MD and ED (essential dynamics) methods using AMBER were used to elucidate conformational and dynamic differences between the protein systems. Several interesting observations were made in the analysis of this study. Hinge-bending motions were shown to occur in each repeat of annexin V from the inter-module groove, and repeat I showed the most mobility. Calcium binding was shown to amplify the hinge bending motions and to stabilize more open conformations. Significant differences were observed in the conformations and dynamics between annexins A1 and V. The differences were attributed to the core residues as the N-termini were excluded from the study. Recent MD simulations were also performed on three protein systems to elucidate conformational changes during annexin A1 induced membrane aggregation.14 Ca21 free annexin A1 with the N-terminal buried inside the core, Ca21 PROTEINS
2937
M.P. Donohue et al.
bound annexin A1 lacking the N-terminal, and Ca21 bound annexin A1 with the N-terminal in an exposed position outside the core comprised the three systems under study. The results indicated that the calcium coordinating residues on the convex face of the protein showed relatively higher fluctuation values than noncoordinating residues. RMSD analysis indicated that the Nterminal is the most flexible region of the protein when in an exposed position, implicating its possible role as a second membrane binding site. The measured dimensions of the annexin system with the exposed N-terminal supported the mechanism of membrane aggregation proposed by X-ray studies. In this study, we investigate the role of the N-terminal as a possible second lipid binding site to induce membrane aggregation using molecular dynamics simulations. The mechanism of how annexin A1 causes membrane aggregation and fusion is still not fully understood and is clearly very complicated. Molecular dynamics is a proven and useful tool for studying biological phenomena at an atomic level, but its application to an annexin-membrane system has not yet been reported to date. The constructed setup contained up to 0.14 million atoms. This work focused on the analysis of MD simulations comprising an annexin A1 molecule with an exposed N-terminal positioned between two phospholipids monolayers. Periodic boundary conditions applied to the systems gave the desired effect of two continuous membrane bilayers. This work provides insight into annexin-membrane interactions. COMPUTATIONAL METHODS Annexin construction
There are no crystal structures available for full-length, calcium-bound annexin A1 with its N-terminal in an exposed position outside of the core domain. Therefore, our first task was to construct this protein under study. The calcium bound core domain starting coordinates were obtained from its X-ray structure (1MCX.pdb). The 41 residue N-terminal coordinates were taken from the X-ray structure of annexin A1 in the absence of calcium (1HM6.pdb), and fused to the core domain in an exposed position using Insight II software. This annexin construction was part of the work of previous molecular dynamics experiments, and therefore was readily available at the start of this work.14 The constructed protein has a total of 351 residues, eight of which are calcium ions. System construction
DOPG, an anionic phospholipid, and DOPC, a zwitterionic phospholipid, were also constructed using the Insight II software,15 and optimized using DMol3.16 The restrained electrostatic potential method17 was used to generate charges for the atoms in the DOPC/DOPG
2938
PROTEINS
molecules. The electrostatic potential was obtained using Gaussian 0318 at the HF/6-31G* level of theory, the same procedure that was used to generate charges for the Amber force field. The program Antechamber implemented in Amber suite was then used to construct a parameter library for each lipid. A total of 225 phospholipids were used per monolayer (180 DOPC:45 DOPG), creating a 15 3 15 array of phospholipids. The phospholipid layer was created in PyMOL,19 placing the individual lipids 8 A˚ngstr€ oms apart. The DOPG lipids were arranged within the layer to give maximum distance between these negatively charged lipids. The ratio of 4:1 DOPC to DOPG was chosen to model after the phospholipid composition used in the cryo-EM experiment.10 Three separate pdb files were made; the top layer of phospholipids, the annexin protein, and the bottom layer of phospholipids. All three were loaded into Insight II at the origin (0, 0, 0) in Cartesian space. Then, each pdb file (top, protein, or bottom) was translated along the y-axis until the annexin protein was positioned between the two constructed monolayers at the desired distance. The annexin/phospholipid system was then loaded as a pdb file into the xLEaP program of AMBER, after which Na1 counter ions were added to neutralize the negative charges of the system. Extra Na1 and Cl2 ions were then added to achieve the 0.1M experimental concentrations. Finally, the space between the two monolayers was explicitly solvated using TIP3P waters in a rectilinear box.20 Figure 1 depicts the fully constructed annexin-membrane setup. The simulation setup contains the calcium-bound annexin A1 monomer with an exposed N-terminal positioned between two phospholipid monolayers made up of 180 DOPC molecules and 45 DOPG molecules. MD simulation
All computations were performed on a 128-processor SGI Altix 4700. All calculations were carried out using AMBER’s force field (ff03) in the AMBER 9 software package. The SANDER module carried out the energy minimizations and molecular dynamics computations. Periodic boundary conditions were applied to the system with a nonbonded cutoff of 12 A˚ to truncate Van der Waals interactions. The particle-mesh Ewald method21,22 was used to treat long range electrostatic interactions, with a cubic B-spline interpolation. The SHAKE algorithm23 was used to constrain all bonds involving hydrogen atoms with a geometric tolerance of 0.00001 A˚. The system was energy minimized twice before the MD simulation. A restrained minimization was first performed on the solvent, the counter ions, and the phospholipid layers while keeping the protein fixed. Next, the entire system was minimized. The minimized system was then warmed for 20 ps to 300 K from an initial temperature of 10 K and the temperature was controlled using Langevin dynamics.
N-Terminal of Annexin A1 as a Secondary Binding Site
Table I
Dimensions (in A˚) and Time Scales for MD Simulations
Initial box dimensions () Total number of atoms Restrained minimization Minimization Equilibration (ps) NPT run (ps) NVT run (ps)
Figure 2 Area per molecule on the constructed lipid layer and density of the simulation cell used during a 200 ps NPT simulation.
Two molecular dynamics simulations were performed on the setup after the warming step; one simulation used NVT conditions (constant particle number, volume, and temperature) and the other used NPT conditions (constant particle number, pressure, and temperature). The choice of producing trajectories with both NVT and NPT ensembles allowed for comparison of the methodological aspects of the simulations. In the case of the NVT simulations, a short NPT run using anisotropic pressure scaling was first performed for 200 ps, which permitted the dimensions of the simulation cell to fluctuate; however, the angles between the walls of the cell were fixed at 90 . As shown in Figure 2, the projected lateral area per molecule on the constructed bilayer was 57.3 A˚2 after 200 ps using the NPT ensemble, which is in good agreement with the value of 59.3 A˚2 reported for DOPC bilayers at 66% relative humidity.24 Moreover, the density of the simulation cell equilibrates within this time scale to 1.02 g/cm3. Stabilization of these parameters, both lipid surface area and simulation cell density, were taken as indicators that the bilayer was equilibrated and ready for longer time scale simulations. A time step of 2 fs was used for all molecular dynamics simulations. Time scales used for the various steps in the MD simulations are reported in the following Table I. The simulations in the work are very computationally expensive. It took one week to generate 4 ns using 16 processors of Altix 4700 at East Carolina University.
NVT
NPT
122.75 3 122.00 3 112.94 142,364 1000 steps 2500 steps 20 200 26502.2
122.75 3 122.00 3 112.94 142,364 1000 steps 2500 steps 20 38204.4 N/A
generated for using NPT dynamics. The stability of each simulation was determined by the conservation of energy and the maintenance of a constant average temperature. The b-factor of the Ca of each of the residue of annexin A1 was calculated after the system was equilibrated (Fig. 3). This was accomplished using the “atomicfluct” command of the “ptraj” module in the AMBER 9.0 suite of programs. Prior to using the “atomicfluct” command in the b-factor calculation, “rms” command of “ptraj” was performed on the core domain to remove the rotational and translational motions of each step. The 41 residues of the N-terminal have much higher b-factor values than the rest of the protein, reaffirming that this region is indeed very flexible. The calculated b-factors of the Ca’s over the course of the simulation were compared with the b-factor taken from X-ray crystallography data (Fig. 3). As shown, the calculated b-factors from the simulations align very well with the b-factor values taken from X-ray data. This helps to validate that the simulation setup and the force field ff99 used during the simulation are rational choices giving fairly consistent data with experimental results.
RESULTS AND DISCUSSION Domain flexibility
About 26 ns simulation trajectory data were generated for annexin A1 positioned between negative charged monolayers using NVT dynamics and 38 ns were
Figure 3 The calculated b-factors for alpha carbons of NVT dynamics (red), NPT dynamics (green), and the X-Ray b-factors (black). [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]
PROTEINS
2939
M.P. Donohue et al.
Figure 4 The interaction of each residue of N-terminal with the bottom layer of phospholipids for both NPT and NVT ensembles. Serine 5 and lysines 26 and 29 are indicated in text.
Anja Rosengarth et al. proposed that calcium binding to the convex face of the core domain of annexin A1 triggers a series of events in which the N-terminal is ultimately ejected from repeat III of the core domain and the D helix folds back into the proper helical conformation.12 In this proposed active conformation, the N-terminal would be free to move around via the flexible linker formed by residues 27–41. Electron density studies indicated that hydrophobic residues of the N-terminal (M-3, V-4, and F-7) would favorably be packed into a hydrophobic pocket formed by residues F-221, L-225, F237, and V-268 of repeat III. This idea is supported by our analysis of the b-factor values, which are highest in the core domain at repeat III, giving further evidence to the structural role played by this region of the protein during membrane binding. Furthermore, several calcium coordinating residues displayed relatively high magnitudes of fluctuation compared with noncoordinating residues as shown on Figure 3. Anja Rosengarth et al. reported the calcium coordinating residues of annexin A1.12 The results of a study conducted by D. Cregut et al.,13 in which molecular dynamics simulations were performed on annexin V, also found that calcium coordinating residues displayed an increased flexibility. Previous simulations on annexin A1 by Shesham et al. also had the similar findings.14 The secondary membrane binding site The N-terminal and membrane interaction
Nonbonded interactions between the N-terminal residues of the protein and the bottom monolayer phospholipids were analyzed. This helped to identify which residues had a greater impact in the overall conformational changes the protein underwent. This was
2940
PROTEINS
accomplished using MM-GBSA module in the AMBER suite of programs. Van der Waals interactions were truncated at 12 A˚ and were found to have little influence in the nonbonded interactions. The calculations indicated that the primary non-bonded interactions were contributed by electrostatic forces. The electrostatic contribution to the solvation free energy, DGelst, is calculated using the Generalized Born method in the pbsa program of AMBER. MM-GBSA calculations were performed on both simulations, the results are shown in Figure 4. The net value of DGelst between the N-terminal and the bottom layer of phospholipids was calculated to be 248.07 kcal/mol for the entire trajectory using NVT dynamics and 214.66 kcal/mol using NPT dynamics. Moreover, a recent finding by Dorovkov et al. demonstrated that phosphorylation of S-5 prevented peripheral a-helical association of the first 18 residues of the N-terminal of annexin A1 with anionic and zwitterionic membrane mimetics and phospholipid vesicles.25 They hypothesized an electrostatic repulsion between phosphorylated S-5 and negatively charged micelles would prevent the Nterminal from binding to membranes. The results of our electrostatic calculations suggest that if S-5 did indeed possess a net negative charge, then, based on the magnitude of the energy between charged residues and the membrane, the overall sign of the free energy between the N-terminal and the membrane would shift from negative to positive, thereby weakening the potential for binding. According to Eduard Bitto and Wonhwa Cho, K-26 and K-29 play an essential role in the membrane aggregation activity of annexin A1.8 Bitto and Cho systematically assessed the contribution of the amino-terminal to membrane aggregation by first truncating the entire Nterminal and then measuring the effect of incremental addition of amino terminal residues on vesicle aggregation activity. They reported that annexin A1 D1–41 showed no detectable aggregating activity under normal assay conditions, that D1–29 lowered the activity of the core, whereas D1–24 fully restored the wild-type activity.8 The analysis of MM-PBSA data from simulation trajectories confirms their statement. The K-29 sidechain reorientation
Throughout both simulations, those two lysine residues displayed the strongest electrostatic attractions to the phospholipid layer. The positively charged sidechains of these residues, especially K-29, were observed to orient themselves in close proximity (3.5 A˚) with the polar head-groups of the phospholipids over the trajectory of the NVT simulation, as shown in Figure 5. K-26 and K-29 are absolutely conserved to all annexins I from different species, so this mechanism would apply to all annexins I. In this study, during the NVT simulation, K-26 and K-29 both orient their positively charged side chains
N-Terminal of Annexin A1 as a Secondary Binding Site
region of the N-terminal, and as indicated on the b-factor plot, this region has the highest flexibility of any region in the protein. Moreover, Lizarbe et al. reported that proteolysis of the N-terminal domain negatively regulates annexininduced vesicle aggregation.27 Based on the crystal structure, the concave side of the core domain of annexin A1 lacks any lysine or arginine residues located in loops regions. Removal of the N-terminal would eliminate the only structural motif in the protein containing flexible, positively charged side-chains. Therefore, it is their flexibility and their charge that permit K-26 and K-29 to anchor to an anionic membrane. Based on the simulations it appears that the secondary membrane binding site of annexin A1 is mediated by membrane-annexin interactions.
CONCLUSION
Figure 5 Snapshots taken from the trajectory using NVT dynamics showing the orientation of the side-chains of K-26 and K29. It was observed the K29 rotated from its initial position to be in closer proximity with the anionic phospholipid layer. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]
down toward the negatively charged phospholipid layer, possibly acting as an electrostatic anchor so that the protein may adopt a more favorable conformation in its environment. Additionally, the results from combined circular dichroism and neutron diffraction studies by Hu et al. established that the calcium-independent interaction of annexin A1 is indeed mediated by the N-terminal domain, and that the N-terminal adopts a peripheral mode of binding to the membrane surface and that some sidechains intercalate with phospholipids.26 We observe a reorientation of the sidechains of K-26 and K29 so that the e-amino groups come into close contact with the phospholipid head-groups, and the reorientation appears to be electrostatically driven. K-26 displayed the greatest electrostatic attraction to the bottom phospholipid layer out of all N-terminal residues throughout both simulations. Lysines 26 and 29 are reported to be the electrostatic anchor because they are located in the unstructured coil
Based on the results of the present study, the following mechanism of membrane aggregation is proposed for annexin A1. The convex face of the protein serves as the initial membrane binding site. The N-terminal has been speculated to function as the second membrane binding site in annexin A1. Our molecular dynamics simulations, consisting of two separate simulations of isobaric and isochoric ensembles at 300 K, generated strong electrostatic attractions of both K-26 and K-29 with the constructed membrane bilayer. The positively charged sidechains of these residues were observed to orient themselves in close proximity (3.5 A˚) with the polar headgroups of the phospholipids. REFERENCES 1. Gerke V, Moss SE. Annexins: From Structure to Function. Physiol Rev 2002;82:331–371. 2. Swairjo MA, Concha NO, Kaetzel MA, Dedman JR, Seaton BA. Ca21-bridging mechanism and phospholipids head group recognition in the membrane-binding protein annexin V. Nat Struct Biol 1995; 2:968–974. 3. Meers P, Mealy T, Pavlotsky N, Tauber AI. Annexin I-mediated vesicular aggregation: mechanism and role in human neutrophils. Biochemistry 1992;31:6372–6382. 4. Meers P, Mealy T, Tauber AI. Annexin I interactions with human neutrophil specific granules: fusogenicity and coaggregation with plasma membrane vesicles. Biochim Biophys Acta 1993;1147:177–184. 5. Wang W, Creutz CE. Regulation of the chromaffin granule aggregating activity of annexin I by phosphorylation. Biochemistry 1992;31: 9934–9939. 6. Wang W, Creutz CE. Role of the amino-terminal domain in regulating interactions of annexin I with membranes: effects of aminoterminal truncation and mutagenesis of the phosphorylation sites. Biochemistry 1993;33:275–282. 7. de la Fuenta M, Parra V. Vesicle aggregation by annexin I: role of secondary membrane binding site. Biochemistry 1995;34:10393–10399. 8. Bitto E, Cho W. Structural determinant of the vesicle aggregation activity of annexin I. Biochemistry 1999;38:14094–14100. 9. Andree HA, Willems G, Hauptmann R, Maurer-Fogy I, Stuart MC, Hermens WT, Frederik PM, Reutelingsperger CP. Aggregation of phospholipid vesicles by a chimeric protein with the N-terminus of annexin I and the core of annexin V. Biochemistry 1993;32:4634–4640.
PROTEINS
2941
M.P. Donohue et al.
10. Lambert O, Gerke V, Bader MF, Porte F, Brisson A. Structural analysis of junctions formed between lipid membranes and several annexins by cryo-electron microscopy. J Mol Biol 1997;272: 42–45. 11. Rosengarth A, Gerke V, Luecke H. X-ray structure of full-length annexin 1 and implications for membrane aggregation. J Mol Biol 2001;306:489–498. 12. Rosengarth A, Luecke H. A calcium-driven conformational switch of the N-terminal and core domains of annexin A1. J Mol Biol 2003;326:1317–1325. 13. Cregut D, Drin G, Liautard JP, Chiche L. Hinge-bending motions in annexins: molecular dynamics and essential dynamics of apoannexin V and of calcium bound annexin V and I. Protein Eng 1998;11:891–900. 14. Shesham RD, Bartolotti LJ, Li Y. Molecular dynamics simulation studies on Ca21-induced conformational changes of annexin I. Protein Eng 2008;21:115–120. 15. INSIGHT II. San Diego, California: Accelrys Inc. 16. Delley, B. DMol 3. San Diego, California: Accelrys Inc.; Delley, B. J Chem Phys 1990;92:508. 17. Bayly CI, Cieplak P, Cornell W, Kollman PA. A well-behaved electrostatic potential based method using charge restraints for deriving atomic charges: the RESP model. J Phys Chem 1993;97: 10269–10280. 18. Frisch MJ, Trucks GW, Schlegel HB, Scuseria GE, Robb MA, Cheeseman JR, Montgomery JA, Jr., Vreven T, Kudin KN, et al. Gaussian 03, revision E.01. Wallingford, CT: Gaussian, Inc.; 2004.
2942
PROTEINS
19. DeLano WL. The PyMOL Molecular Graphics System. San Carlos, CA: DeLano Scientific LLC. http://www.pymol.org. 20. Jorgensen WL, Chandrasekhar J, Madura JD, Impey RW, Klein ML. Comparison of simple potential functions for simulating liquid water. J Chem Phys 1993;79:926–935. 21. Darden T, York D, Pedersen L. Particle mesh Ewald: an N-log(N) method for Ewald sums in large systems. J Chem Phys 1993;98: 10089–10092. 22. Essmanmn U, Perera L, Berkowitz MX, Darden T, Lee H, Pedersen LG. A smooth particle mesh Ewald method. J Chem Phys 1995;103: 8577–8593. 23. Ryckaert JP, Ciccotti G, Berendsen HJC. Numerical integration of the cartesian equations of motion of a system with constraints: molecular dynamics of n-alkanes. J Comput Phys 1977;23:327–341. 24. T€ uchsen E, Jensen MØ, Westh P. Solvent accessible surface area (ASA) of simulated phospholipid membranes. Chem Phys Lipids 2003;123:107–116. 25. Dorovkov MV, Kostyukova AS, Ryazanov AG. Phosphorylation of annexin A1 by TRPM7 kinase: a switch regulating the induction of an a-helix. Biochemistry 2011;50:2187–2193. 26. Hu N, Bradshaw J, Lauter H, Buckingham J, Solito E, Hofmann A. Membrane-induced folding and structure of membrane-bound annexin A1 N-terminal peptides: implications for annexin-induced membrane aggregation. Biophys J 2008;94:1773–1781. 27. Lizarbe MA, Barrasa JI, Olmo N, Gavilanes F, Turnay J. Annexinphospholipid interactions functional implications. Int J Mol Sci 2013;14:2562–2683.
The N-terminal of annexin A1 as a secondary membrane binding site: A molecular dynamics study Matthew P. Donohue, Libero J. Bartolotti, and Yumin Li* Department of Chemistry, East Carolina University, Greenville, North Carolina 27858
ABSTRACT Annexin A1 has been shown to cause membrane aggregation and fusion, yet the mechanism of these activities is not clearly understood. In this work, molecular dynamics simulations were performed on monomeric annexin A1 positioned between two negatively charged monolayers using AMBER’s all atom force field to gain insight into the mechanism of fusion. Each phospolipid monolayer was made up of 180 DOPC molecules and 45 DOPG molecules to achieve a 4:1 ratio. The space between the two monolayers was explicitly solvated using TIP3P waters in a rectilinear box. The constructed setup contained up to 0.14 million atoms. Application of periodic boundary conditions to the simulation setup gave the desired effect of two continuous membrane bilayers. Nonbonded interactions were calculated between the N-terminal residues and the bottom layer of phospholipids, which displayed a strong attraction of K26 and K29 to the lipid head-groups. The sidechains of these two residues were observed to orient themselves in close proximity (~3.5 A˚) with the polar head-groups of the phospholipids. Proteins 2014; 82:2936–2942. C 2014 Wiley Periodicals, Inc. V
Key words: interaction energy; membrane aggregation; domain flexibility; side-chain reaction; molecular dynamics.
INTRODUCTION Interactions between proteins and membrane surfaces involve complex molecular mechanisms, which are critical in allowing the cell to function properly, yet remain to be fully elucidated in cell biology. Anchoring of many peripheral proteins requires electrostatic or ionic interactions with membrane lipids. Annexins constitute a family of proteins that have the ability to bind onto negatively charged phospholipid bilayers in a reversible and calcium dependent manner through the unique architecture of their calcium binding sites. In addition to membrane binding, annexins with relatively larger N-terminal domains (between 30 and 100 residues) have been identified to cause membrane aggregation and fusion events.1 Annexins are structurally divided into a conserved core domain, which has the shape of a slightly curved disc, and a divergent N-terminal that is unique for a given member of the family. The core domain comprises four (in annexin A6 eight) homologous repeats (labeled I–IV) of about 75 amino acid residues that fold into five alpha-helices (A–E) and form an anti-parallel bundle (Fig. 1). High-resolution crystal structures have identified the calcium binding sites to be located on the convex
2936
PROTEINS
face of the protein. As shown in Figure 1, the calcium ions are depicted as light-blue spheres. The bound calcium ions serve as a hypothetical “bridge” between the protein and membrane by simultaneously coordinating ligands from acidic side chains of the protein and from phosphoryl moieties of the lipids.2 The N-terminal is variable in sequence and length for given members of the family, and is thought to regulate the specific physiological functions of each annexin.1 Annexin A1, a 37 kDa protein previously known as lipocortin, has been identified to cause membrane aggregation.3–7 The N-terminal domain of annexin A1 contains 41 residues. Biochemical studies involving the deletion and phosphorylation of the N-terminal have demonstrated that the N-terminal is important for membrane aggregation, but is of little significance for binding to negatively charged phospholipids.8 Further, a chimera comprising the core domain of annexin V (which does not promote membrane fusion activity and has a shorter *Correspondence to: Yumin Li, Department of Chemistry, East Carolina University, Greenville, North Carolina 27858. E-mail: [email protected] Received 24 January 2014; Revised 19 May 2014; Accepted 28 May 2014 Published online 10 June 2014 in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/prot.24623
C 2014 WILEY PERIODICALS, INC. V
N-Terminal of Annexin A1 as a Secondary Binding Site
Figure 1 Annexin A1 is positioned between two negatively charged monolayers. Annexin A1 is color coded as red (repeat I), green (repeat II), blue (repeat III), yellow (repeat IV), and black (the N-terminal). The location of K26 and K29 within the protein is indicated with an arrow. Calcium ions are shown as light-blue spheres. A 90 rotation provides an axial view of the protein.
N-terminal with only 19 residues) and the N-terminal of annexin A1 was found to cause membrane aggregation.9 The results of these studies led to the supposition that the N-terminal of annexin A1 harbors the second membrane-binding site of the protein. Cryo-electron microscopy studies reported two types of junctions formed connecting DOPG/DOPC liposomes and annexin A1.10 The 160 A˚ thick six stripe junctions were more frequently observed than the 125 A˚ thick five stripe junctions. The six striped junctions were interpreted as follows: the two outer stripes on each side were attributed to the membrane inner and outer leaflets, while the two inner stripes were due to an annexin A1 or A2 dimer. The five stripe junctions attributed the central stripe as an annexin A1 or A2 monomer. The
distances between the centers-of-mass of the outer bilayer leaflets were shown to be 80 A˚ (six-stripe junctions), which after accounting for the width of the lipid head-groups corresponds to a protein thickness of 60 A˚. In the case of the five-stripe junctions, a thickness of 25 A˚ was assigned to the annexin A1 monomer. These results were in conflict with X-ray crystallography techniques, and whether membrane aggregation is caused by the annexin A1 monomer or dimer remains unclear. The structures of porcine annexin A1 both in the presence and absence of calcium have been solved using Xray crystallography techniques.11,12 The results identified a drastic change in structure between the conformations of the N-terminal and repeat III of the core domain. In the absence of calcium, the amphipathic alpha helix formed by residues 2–12 of the N-terminal domain is buried inside repeat III of the protein core, replacing the D-helix, which forms a flap over the N-terminal. Upon calcium binding, the D-flap folds into the proper conformation in repeat III, forming a Ca21 binding site and ejecting the N-terminal. Based on crystal structures, an annexin A1 double layer between two lipid bilayers would be at least 70 A˚ thick and about 90 A˚ (accounting for the N-terminal) in the case of full length annexin A1. Explanation of the six stripe junctions by Rosengarth et al. proposed that the two central stripes were due to annexin A1 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 bilayer, resulting in an average appearance of two layers.11 The thinner fivestripe junctions could then be interpreted as sections where the majority of the convex faces of annexins are attached to one of the bilayers, resulting in a single asymmetric high-density feature between the phospholipid bilayers. Previous computational studies on annexins conducted by Cregut et al. were designed to study the hinge bending motions of calcium-bound annnexin V and A1 and apo-annexin V.13 MD and ED (essential dynamics) methods using AMBER were used to elucidate conformational and dynamic differences between the protein systems. Several interesting observations were made in the analysis of this study. Hinge-bending motions were shown to occur in each repeat of annexin V from the inter-module groove, and repeat I showed the most mobility. Calcium binding was shown to amplify the hinge bending motions and to stabilize more open conformations. Significant differences were observed in the conformations and dynamics between annexins A1 and V. The differences were attributed to the core residues as the N-termini were excluded from the study. Recent MD simulations were also performed on three protein systems to elucidate conformational changes during annexin A1 induced membrane aggregation.14 Ca21 free annexin A1 with the N-terminal buried inside the core, Ca21 PROTEINS
2937
M.P. Donohue et al.
bound annexin A1 lacking the N-terminal, and Ca21 bound annexin A1 with the N-terminal in an exposed position outside the core comprised the three systems under study. The results indicated that the calcium coordinating residues on the convex face of the protein showed relatively higher fluctuation values than noncoordinating residues. RMSD analysis indicated that the Nterminal is the most flexible region of the protein when in an exposed position, implicating its possible role as a second membrane binding site. The measured dimensions of the annexin system with the exposed N-terminal supported the mechanism of membrane aggregation proposed by X-ray studies. In this study, we investigate the role of the N-terminal as a possible second lipid binding site to induce membrane aggregation using molecular dynamics simulations. The mechanism of how annexin A1 causes membrane aggregation and fusion is still not fully understood and is clearly very complicated. Molecular dynamics is a proven and useful tool for studying biological phenomena at an atomic level, but its application to an annexin-membrane system has not yet been reported to date. The constructed setup contained up to 0.14 million atoms. This work focused on the analysis of MD simulations comprising an annexin A1 molecule with an exposed N-terminal positioned between two phospholipids monolayers. Periodic boundary conditions applied to the systems gave the desired effect of two continuous membrane bilayers. This work provides insight into annexin-membrane interactions. COMPUTATIONAL METHODS Annexin construction
There are no crystal structures available for full-length, calcium-bound annexin A1 with its N-terminal in an exposed position outside of the core domain. Therefore, our first task was to construct this protein under study. The calcium bound core domain starting coordinates were obtained from its X-ray structure (1MCX.pdb). The 41 residue N-terminal coordinates were taken from the X-ray structure of annexin A1 in the absence of calcium (1HM6.pdb), and fused to the core domain in an exposed position using Insight II software. This annexin construction was part of the work of previous molecular dynamics experiments, and therefore was readily available at the start of this work.14 The constructed protein has a total of 351 residues, eight of which are calcium ions. System construction
DOPG, an anionic phospholipid, and DOPC, a zwitterionic phospholipid, were also constructed using the Insight II software,15 and optimized using DMol3.16 The restrained electrostatic potential method17 was used to generate charges for the atoms in the DOPC/DOPG
2938
PROTEINS
molecules. The electrostatic potential was obtained using Gaussian 0318 at the HF/6-31G* level of theory, the same procedure that was used to generate charges for the Amber force field. The program Antechamber implemented in Amber suite was then used to construct a parameter library for each lipid. A total of 225 phospholipids were used per monolayer (180 DOPC:45 DOPG), creating a 15 3 15 array of phospholipids. The phospholipid layer was created in PyMOL,19 placing the individual lipids 8 A˚ngstr€ oms apart. The DOPG lipids were arranged within the layer to give maximum distance between these negatively charged lipids. The ratio of 4:1 DOPC to DOPG was chosen to model after the phospholipid composition used in the cryo-EM experiment.10 Three separate pdb files were made; the top layer of phospholipids, the annexin protein, and the bottom layer of phospholipids. All three were loaded into Insight II at the origin (0, 0, 0) in Cartesian space. Then, each pdb file (top, protein, or bottom) was translated along the y-axis until the annexin protein was positioned between the two constructed monolayers at the desired distance. The annexin/phospholipid system was then loaded as a pdb file into the xLEaP program of AMBER, after which Na1 counter ions were added to neutralize the negative charges of the system. Extra Na1 and Cl2 ions were then added to achieve the 0.1M experimental concentrations. Finally, the space between the two monolayers was explicitly solvated using TIP3P waters in a rectilinear box.20 Figure 1 depicts the fully constructed annexin-membrane setup. The simulation setup contains the calcium-bound annexin A1 monomer with an exposed N-terminal positioned between two phospholipid monolayers made up of 180 DOPC molecules and 45 DOPG molecules. MD simulation
All computations were performed on a 128-processor SGI Altix 4700. All calculations were carried out using AMBER’s force field (ff03) in the AMBER 9 software package. The SANDER module carried out the energy minimizations and molecular dynamics computations. Periodic boundary conditions were applied to the system with a nonbonded cutoff of 12 A˚ to truncate Van der Waals interactions. The particle-mesh Ewald method21,22 was used to treat long range electrostatic interactions, with a cubic B-spline interpolation. The SHAKE algorithm23 was used to constrain all bonds involving hydrogen atoms with a geometric tolerance of 0.00001 A˚. The system was energy minimized twice before the MD simulation. A restrained minimization was first performed on the solvent, the counter ions, and the phospholipid layers while keeping the protein fixed. Next, the entire system was minimized. The minimized system was then warmed for 20 ps to 300 K from an initial temperature of 10 K and the temperature was controlled using Langevin dynamics.
N-Terminal of Annexin A1 as a Secondary Binding Site
Table I
Dimensions (in A˚) and Time Scales for MD Simulations
Initial box dimensions () Total number of atoms Restrained minimization Minimization Equilibration (ps) NPT run (ps) NVT run (ps)
Figure 2 Area per molecule on the constructed lipid layer and density of the simulation cell used during a 200 ps NPT simulation.
Two molecular dynamics simulations were performed on the setup after the warming step; one simulation used NVT conditions (constant particle number, volume, and temperature) and the other used NPT conditions (constant particle number, pressure, and temperature). The choice of producing trajectories with both NVT and NPT ensembles allowed for comparison of the methodological aspects of the simulations. In the case of the NVT simulations, a short NPT run using anisotropic pressure scaling was first performed for 200 ps, which permitted the dimensions of the simulation cell to fluctuate; however, the angles between the walls of the cell were fixed at 90 . As shown in Figure 2, the projected lateral area per molecule on the constructed bilayer was 57.3 A˚2 after 200 ps using the NPT ensemble, which is in good agreement with the value of 59.3 A˚2 reported for DOPC bilayers at 66% relative humidity.24 Moreover, the density of the simulation cell equilibrates within this time scale to 1.02 g/cm3. Stabilization of these parameters, both lipid surface area and simulation cell density, were taken as indicators that the bilayer was equilibrated and ready for longer time scale simulations. A time step of 2 fs was used for all molecular dynamics simulations. Time scales used for the various steps in the MD simulations are reported in the following Table I. The simulations in the work are very computationally expensive. It took one week to generate 4 ns using 16 processors of Altix 4700 at East Carolina University.
NVT
NPT
122.75 3 122.00 3 112.94 142,364 1000 steps 2500 steps 20 200 26502.2
122.75 3 122.00 3 112.94 142,364 1000 steps 2500 steps 20 38204.4 N/A
generated for using NPT dynamics. The stability of each simulation was determined by the conservation of energy and the maintenance of a constant average temperature. The b-factor of the Ca of each of the residue of annexin A1 was calculated after the system was equilibrated (Fig. 3). This was accomplished using the “atomicfluct” command of the “ptraj” module in the AMBER 9.0 suite of programs. Prior to using the “atomicfluct” command in the b-factor calculation, “rms” command of “ptraj” was performed on the core domain to remove the rotational and translational motions of each step. The 41 residues of the N-terminal have much higher b-factor values than the rest of the protein, reaffirming that this region is indeed very flexible. The calculated b-factors of the Ca’s over the course of the simulation were compared with the b-factor taken from X-ray crystallography data (Fig. 3). As shown, the calculated b-factors from the simulations align very well with the b-factor values taken from X-ray data. This helps to validate that the simulation setup and the force field ff99 used during the simulation are rational choices giving fairly consistent data with experimental results.
RESULTS AND DISCUSSION Domain flexibility
About 26 ns simulation trajectory data were generated for annexin A1 positioned between negative charged monolayers using NVT dynamics and 38 ns were
Figure 3 The calculated b-factors for alpha carbons of NVT dynamics (red), NPT dynamics (green), and the X-Ray b-factors (black). [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]
PROTEINS
2939
M.P. Donohue et al.
Figure 4 The interaction of each residue of N-terminal with the bottom layer of phospholipids for both NPT and NVT ensembles. Serine 5 and lysines 26 and 29 are indicated in text.
Anja Rosengarth et al. proposed that calcium binding to the convex face of the core domain of annexin A1 triggers a series of events in which the N-terminal is ultimately ejected from repeat III of the core domain and the D helix folds back into the proper helical conformation.12 In this proposed active conformation, the N-terminal would be free to move around via the flexible linker formed by residues 27–41. Electron density studies indicated that hydrophobic residues of the N-terminal (M-3, V-4, and F-7) would favorably be packed into a hydrophobic pocket formed by residues F-221, L-225, F237, and V-268 of repeat III. This idea is supported by our analysis of the b-factor values, which are highest in the core domain at repeat III, giving further evidence to the structural role played by this region of the protein during membrane binding. Furthermore, several calcium coordinating residues displayed relatively high magnitudes of fluctuation compared with noncoordinating residues as shown on Figure 3. Anja Rosengarth et al. reported the calcium coordinating residues of annexin A1.12 The results of a study conducted by D. Cregut et al.,13 in which molecular dynamics simulations were performed on annexin V, also found that calcium coordinating residues displayed an increased flexibility. Previous simulations on annexin A1 by Shesham et al. also had the similar findings.14 The secondary membrane binding site The N-terminal and membrane interaction
Nonbonded interactions between the N-terminal residues of the protein and the bottom monolayer phospholipids were analyzed. This helped to identify which residues had a greater impact in the overall conformational changes the protein underwent. This was
2940
PROTEINS
accomplished using MM-GBSA module in the AMBER suite of programs. Van der Waals interactions were truncated at 12 A˚ and were found to have little influence in the nonbonded interactions. The calculations indicated that the primary non-bonded interactions were contributed by electrostatic forces. The electrostatic contribution to the solvation free energy, DGelst, is calculated using the Generalized Born method in the pbsa program of AMBER. MM-GBSA calculations were performed on both simulations, the results are shown in Figure 4. The net value of DGelst between the N-terminal and the bottom layer of phospholipids was calculated to be 248.07 kcal/mol for the entire trajectory using NVT dynamics and 214.66 kcal/mol using NPT dynamics. Moreover, a recent finding by Dorovkov et al. demonstrated that phosphorylation of S-5 prevented peripheral a-helical association of the first 18 residues of the N-terminal of annexin A1 with anionic and zwitterionic membrane mimetics and phospholipid vesicles.25 They hypothesized an electrostatic repulsion between phosphorylated S-5 and negatively charged micelles would prevent the Nterminal from binding to membranes. The results of our electrostatic calculations suggest that if S-5 did indeed possess a net negative charge, then, based on the magnitude of the energy between charged residues and the membrane, the overall sign of the free energy between the N-terminal and the membrane would shift from negative to positive, thereby weakening the potential for binding. According to Eduard Bitto and Wonhwa Cho, K-26 and K-29 play an essential role in the membrane aggregation activity of annexin A1.8 Bitto and Cho systematically assessed the contribution of the amino-terminal to membrane aggregation by first truncating the entire Nterminal and then measuring the effect of incremental addition of amino terminal residues on vesicle aggregation activity. They reported that annexin A1 D1–41 showed no detectable aggregating activity under normal assay conditions, that D1–29 lowered the activity of the core, whereas D1–24 fully restored the wild-type activity.8 The analysis of MM-PBSA data from simulation trajectories confirms their statement. The K-29 sidechain reorientation
Throughout both simulations, those two lysine residues displayed the strongest electrostatic attractions to the phospholipid layer. The positively charged sidechains of these residues, especially K-29, were observed to orient themselves in close proximity (3.5 A˚) with the polar head-groups of the phospholipids over the trajectory of the NVT simulation, as shown in Figure 5. K-26 and K-29 are absolutely conserved to all annexins I from different species, so this mechanism would apply to all annexins I. In this study, during the NVT simulation, K-26 and K-29 both orient their positively charged side chains
N-Terminal of Annexin A1 as a Secondary Binding Site
region of the N-terminal, and as indicated on the b-factor plot, this region has the highest flexibility of any region in the protein. Moreover, Lizarbe et al. reported that proteolysis of the N-terminal domain negatively regulates annexininduced vesicle aggregation.27 Based on the crystal structure, the concave side of the core domain of annexin A1 lacks any lysine or arginine residues located in loops regions. Removal of the N-terminal would eliminate the only structural motif in the protein containing flexible, positively charged side-chains. Therefore, it is their flexibility and their charge that permit K-26 and K-29 to anchor to an anionic membrane. Based on the simulations it appears that the secondary membrane binding site of annexin A1 is mediated by membrane-annexin interactions.
CONCLUSION
Figure 5 Snapshots taken from the trajectory using NVT dynamics showing the orientation of the side-chains of K-26 and K29. It was observed the K29 rotated from its initial position to be in closer proximity with the anionic phospholipid layer. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]
down toward the negatively charged phospholipid layer, possibly acting as an electrostatic anchor so that the protein may adopt a more favorable conformation in its environment. Additionally, the results from combined circular dichroism and neutron diffraction studies by Hu et al. established that the calcium-independent interaction of annexin A1 is indeed mediated by the N-terminal domain, and that the N-terminal adopts a peripheral mode of binding to the membrane surface and that some sidechains intercalate with phospholipids.26 We observe a reorientation of the sidechains of K-26 and K29 so that the e-amino groups come into close contact with the phospholipid head-groups, and the reorientation appears to be electrostatically driven. K-26 displayed the greatest electrostatic attraction to the bottom phospholipid layer out of all N-terminal residues throughout both simulations. Lysines 26 and 29 are reported to be the electrostatic anchor because they are located in the unstructured coil
Based on the results of the present study, the following mechanism of membrane aggregation is proposed for annexin A1. The convex face of the protein serves as the initial membrane binding site. The N-terminal has been speculated to function as the second membrane binding site in annexin A1. Our molecular dynamics simulations, consisting of two separate simulations of isobaric and isochoric ensembles at 300 K, generated strong electrostatic attractions of both K-26 and K-29 with the constructed membrane bilayer. The positively charged sidechains of these residues were observed to orient themselves in close proximity (3.5 A˚) with the polar headgroups of the phospholipids. REFERENCES 1. Gerke V, Moss SE. Annexins: From Structure to Function. Physiol Rev 2002;82:331–371. 2. Swairjo MA, Concha NO, Kaetzel MA, Dedman JR, Seaton BA. Ca21-bridging mechanism and phospholipids head group recognition in the membrane-binding protein annexin V. Nat Struct Biol 1995; 2:968–974. 3. Meers P, Mealy T, Pavlotsky N, Tauber AI. Annexin I-mediated vesicular aggregation: mechanism and role in human neutrophils. Biochemistry 1992;31:6372–6382. 4. Meers P, Mealy T, Tauber AI. Annexin I interactions with human neutrophil specific granules: fusogenicity and coaggregation with plasma membrane vesicles. Biochim Biophys Acta 1993;1147:177–184. 5. Wang W, Creutz CE. Regulation of the chromaffin granule aggregating activity of annexin I by phosphorylation. Biochemistry 1992;31: 9934–9939. 6. Wang W, Creutz CE. Role of the amino-terminal domain in regulating interactions of annexin I with membranes: effects of aminoterminal truncation and mutagenesis of the phosphorylation sites. Biochemistry 1993;33:275–282. 7. de la Fuenta M, Parra V. Vesicle aggregation by annexin I: role of secondary membrane binding site. Biochemistry 1995;34:10393–10399. 8. Bitto E, Cho W. Structural determinant of the vesicle aggregation activity of annexin I. Biochemistry 1999;38:14094–14100. 9. Andree HA, Willems G, Hauptmann R, Maurer-Fogy I, Stuart MC, Hermens WT, Frederik PM, Reutelingsperger CP. Aggregation of phospholipid vesicles by a chimeric protein with the N-terminus of annexin I and the core of annexin V. Biochemistry 1993;32:4634–4640.
PROTEINS
2941
M.P. Donohue et al.
10. Lambert O, Gerke V, Bader MF, Porte F, Brisson A. Structural analysis of junctions formed between lipid membranes and several annexins by cryo-electron microscopy. J Mol Biol 1997;272: 42–45. 11. Rosengarth A, Gerke V, Luecke H. X-ray structure of full-length annexin 1 and implications for membrane aggregation. J Mol Biol 2001;306:489–498. 12. Rosengarth A, Luecke H. A calcium-driven conformational switch of the N-terminal and core domains of annexin A1. J Mol Biol 2003;326:1317–1325. 13. Cregut D, Drin G, Liautard JP, Chiche L. Hinge-bending motions in annexins: molecular dynamics and essential dynamics of apoannexin V and of calcium bound annexin V and I. Protein Eng 1998;11:891–900. 14. Shesham RD, Bartolotti LJ, Li Y. Molecular dynamics simulation studies on Ca21-induced conformational changes of annexin I. Protein Eng 2008;21:115–120. 15. INSIGHT II. San Diego, California: Accelrys Inc. 16. Delley, B. DMol 3. San Diego, California: Accelrys Inc.; Delley, B. J Chem Phys 1990;92:508. 17. Bayly CI, Cieplak P, Cornell W, Kollman PA. A well-behaved electrostatic potential based method using charge restraints for deriving atomic charges: the RESP model. J Phys Chem 1993;97: 10269–10280. 18. Frisch MJ, Trucks GW, Schlegel HB, Scuseria GE, Robb MA, Cheeseman JR, Montgomery JA, Jr., Vreven T, Kudin KN, et al. Gaussian 03, revision E.01. Wallingford, CT: Gaussian, Inc.; 2004.
2942
PROTEINS
19. DeLano WL. The PyMOL Molecular Graphics System. San Carlos, CA: DeLano Scientific LLC. http://www.pymol.org. 20. Jorgensen WL, Chandrasekhar J, Madura JD, Impey RW, Klein ML. Comparison of simple potential functions for simulating liquid water. J Chem Phys 1993;79:926–935. 21. Darden T, York D, Pedersen L. Particle mesh Ewald: an N-log(N) method for Ewald sums in large systems. J Chem Phys 1993;98: 10089–10092. 22. Essmanmn U, Perera L, Berkowitz MX, Darden T, Lee H, Pedersen LG. A smooth particle mesh Ewald method. J Chem Phys 1995;103: 8577–8593. 23. Ryckaert JP, Ciccotti G, Berendsen HJC. Numerical integration of the cartesian equations of motion of a system with constraints: molecular dynamics of n-alkanes. J Comput Phys 1977;23:327–341. 24. T€ uchsen E, Jensen MØ, Westh P. Solvent accessible surface area (ASA) of simulated phospholipid membranes. Chem Phys Lipids 2003;123:107–116. 25. Dorovkov MV, Kostyukova AS, Ryazanov AG. Phosphorylation of annexin A1 by TRPM7 kinase: a switch regulating the induction of an a-helix. Biochemistry 2011;50:2187–2193. 26. Hu N, Bradshaw J, Lauter H, Buckingham J, Solito E, Hofmann A. Membrane-induced folding and structure of membrane-bound annexin A1 N-terminal peptides: implications for annexin-induced membrane aggregation. Biophys J 2008;94:1773–1781. 27. Lizarbe MA, Barrasa JI, Olmo N, Gavilanes F, Turnay J. Annexinphospholipid interactions functional implications. Int J Mol Sci 2013;14:2562–2683.
