Nanodomain Formation of Ganglioside GM1 in ... - ACS Publications

2015, 31 (33), pp 9105–9114. DOI: 10.1021/acs.langmuir.5b01866. Publication Date (Web): August 6, 2015. Copyright © 2015 American Chemical Soci...
0 downloads 0 Views 489KB Size
Page 1 of 31

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Langmuir

Nanodomain Formation of Ganglioside GM1 in Lipid Membrane: Effects of Cholera Toxin-Mediated Cross-Linking Huijiao Sun, Licui Chen, Lianghui Gao,∗ and Weihai Fang Key Laboratory of Theoretical and Computational Photochemistry, Ministry of Education, College of Chemistry, Beijing Normal University, Beijing 100875, China E-mail: [email protected]

Abstract Cross-linking of specific lipid components by proteins mediates transmembrane signaling and material transport. In this work, we conducted coarse-grained simulation to investigate the interactions of binding units of chorela toxin (CTB) with mixed ganglioside GM1 and dipalmitoyl phosphatidylcholine (DPPC) lipid bilayer membrane. We determine that the binding of CTB pentamers cross-links GM1 molecules into protein-sized nanodomains that have distinct lipid order compared with the bulk. The toxin in the nanodomain partially penetrate into the membrane. The local disordering can also transmit across the membrane via lipid coupling. Comparison simulations on CTB binding to a membrane that is composed of various lipid components demonstrate that several factors are responsible for the nanodomain formation: (a) the negatively charged head group of a GM1 receptor is responsible for the multivalent binding; (b) the head groups being full of hydrogen-bonding donors and receptors stabilize the GM1 cluster itself and ensure the toxin binding with high affinity; and (c) significant size and ∗ To

whom correspondence should be addressed

1 ACS Paragon Plus Environment

Langmuir

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

order differences between the protein receptor lipids and bulk lipids are essential to promoting phase separation and signal transportation.

Introduction The cell membrane is composed of a large number of different lipid species and proteins that enclose the cell and mediate material transport and transmembrane signaling between the cytosol and extracellular environment. 1–3 To coordinate these functions, the membrane usually laterally segregates its constituents. In particular, changes in the membrane organization upon cross-linking of its components trigger a cell signaling response to exogenous factors. For example, many bacterial toxins, such as simian virus 40 (SV40), 4 shiga toxin, 5 and cholera toxin, 6 selectively bind to ganglioside glycolipids and gain entrance to target cells. Cholera toxin is frequently used as a cross-linker to investigate the protein-lipid interaction as well as the infection of the toxin itself. 6–13 This toxin belongs to the AB5 superfamily of toxins, which contains an active (A) unit and five identical binding (B) subunits. 7 The binding unit of the cholera toxin (CTB) binds specifically to monosialotetrahexosylganglioside (GM1) lipid via multivalent electrostatic attraction. Once bound, the toxin can translocate across the membrane, where the A unit undergoes proteolytic cleavage and gives rise to the enzymatically active A1 unit. Although much is known about the biochemistry of cholera toxin infection, the mechanism by which the toxin crosses the plasma membrane has attracted a vast amount of researchs but still remains elusive. Optical microscopy has shown that the cross-linking of GM1 via CTB binding can cause uniform model membranes 6 [that consist dioleylphosphatidylcholine (DOPC), sphingomyelin (Sph), and cholesterol in a certain molar ratio ] as well as plasma membranes 8,12 to phase-separate into large, coexistent liquid-ordered and liquid-disordered domains. Recently, by two-color z-scan fluorescence correlation spectroscopy, nanosized domain formation was observed in the early stage of the cross-linking-induced phase separation in model membrane at a lower sphingomyelin content. 13 This finding indicated that the local transient condensation of GM1 that is

2 ACS Paragon Plus Environment

Page 2 of 31

Page 3 of 31

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Langmuir

induced by the CTB cross-linking can compensate for a lack of Sph, and the CTB can also coalesce the existing nanodomains into large-scale phase separation. In those experiments, the microscale phase was enriched in cholesterol, which means that cholesterol could play important roles in the coalescence of CTB-cross-linked GM1. To investigate the interaction between CTB and GM1 and the resulting effects on the membrane, Miller et al used X-ray reflectivity and grazing incidence diffraction to study the structural changes of cholesterol-freed GM1/phospholipid monolayers induced by CTB binding. 9,10 In the region where the toxin was bound, less ordered domains were observed in a gel phase. The investigators also found that at biologically relevant surface pressures, CTB binding perturbed the lipid packing and resulted in an orientationally textured phase in both the monolayer and bilayer membranes. 11 Despite various experiments that focus on the CTB-induced phase transformation of membrane and toxin transport, the molecular mechanism by which the conformation and orientation of both the lipids and toxin change are still not well resolved because few techniques can track the structural details of the transient nanocluster with both high structural and temporal resolutions. Molecular simulation is a powerful alternative tool for providing structure and dynamics details that cannot be easily probed experimentally. However, because they are limited by the length and time scales that are accessible by all-atom Molecular Dynamics (AAMD) simulations, very few computational approaches on the action of CTB have been reported. 14,15 For example, in a recent work with AAMD simulation, 15 only one CTB, which was binding to a monolayer membrane composed of five GM1 molecules and a hundred phospholipids, was investigated. To handle the entire perspective of phenomena in complex materials, various coarse-grained (CG) methods, including coarse-grained Molecular Dynamics (CGMD) 16–20 and dissipative particle dynamics (DPD) simulations, 21–38 have been developed. The CG model groups a few atoms into one mesoscopic pseudoatom and thus can reach the biologically relevant time and length scales efficiently. 16 In this article, we perform DPD simulations to study the cross-linking effects of up to six CTB binding to bilayer membrane composed of dipalmitoyl phosphatidylcholine (DPPC) and GM1 lipids with sizes as large as 24 nm×24 nm. We find that the cross-linking of the bound re-

3 ACS Paragon Plus Environment

Langmuir

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ceptors significantly alter the structure of the membrane: Prior to the protein binding to the bilayer, at a certain molar percent, GM1 lipids form dimers and trimmers in the DPPC bilayer and induce an interdigitated phase. The binding of CTB pentamers cross-link GM1 molecules into raft-like nanodomains, which has a distinct lipid order from the bulk. The geometrical constraints between CTB and lipids promote the exclusion of DPPC lipids underneath the bound toxin and permits CTB to partially penetrate into the membrane. The signal of CTB binding on one leaflet transmits across the membrane by inducing a lipid order alteration in the opposing leaflet. Comparison simulations of CTB binding to membrane with different lipid components demonstrate that having a charged head group and a hydrogen-bonding capacity is essential for a lipid to be a protein receptor. Another factor that promotes the raft-like nanodomain formation is the size and orientation order incompatibility between the lipid components.

Simulation Methods In the coarse-grained DPD simulation, the elementary unit is a soft bead, in which each bead represents a fluid volume of several atoms. 21–23 In this work, water is explicitly modeled as a single bead(denoted by W), which represents 3 water molecules. DPPC and GM1 lipids are modeled as polymers that consist of hydrophilic and hydrophobic beads. The DPPC molecule has four hydrophilic head beads and eight hydrophobic beads. GM1 molecule have nineteen head beads and nine hydrophobic beads. Similar to lipids, an amino acid residue is represented by one skeleton bead plus one or more side-chain beads. 18 The atomic representation of DPPC, GM1, protein amino acids, and their corresponding CG models are given in Figures S1 and S2 in the Supporting Information. Similar to the Martini model, 17–19,39 the DPD beads are sorted into charged (Q), polar (P), nonpolar (N), and apolar (C) types. Each type is further divided into sublevels based on their hydrogen donor capacities (d), hydrogen acceptor capacities (a), and non-hydrogen-bond-forming capacities (0). The type of the backbone (B) bead of the amino acids depends on the secondary structure of the polypeptide: 18 In a coil or bend structure, B is a strongly polar P5 type. In a

4 ACS Paragon Plus Environment

Page 4 of 31

Page 5 of 31

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Langmuir

helical or β structure, the polarity is reduced due to the hydrogen bonding between the backbones, and therefore, the bead is a non-polar N0 -type for an α -helical structure and an Nda -type for a

β -structure. For two beads with separation distance ri j < r0 , the beads interact via short-ranged repulsive forces FiCj (ri j ) = ai j (1 − ri j /r0 )ˆri j ,

(1)

random forces p 2γi j kB T (1 − ri j /r0 )ζi j rˆ i j ,

(2)

FiDj (ri j ) = −γi j (1 − ri j /r0 )2 (ˆri j · vi j )ˆri j .

(3)

FiRj (ri j ) = and dissipative forces

Here, the vectors vi j ≡ vi − v j are the velocity differences between particles i and j. The parameq ters ai j (in unit of kB T /r0 ) are the repulsion strengths, γi j (in unit of kB T m0 /r02 ) are the friction coefficients, and ζi j are symmetrically and uniformly distributed random numbers. In DPD simulations, the repulsion parameter ai j are usually optimized to reproduce the compressibility of the system. 22,23 At ρ = 3, aWW is usually set to 78. The repulsion between beads of the same type is equal to aWW . The other force parameters ai j between beads of different types can be derived by linking the force parameters to χ -parameters in Flory-Huggins theory. Optimized DPD force parameters transferrable for both lipids and amino acids recently obtained by us 40 (also discussed in details in the Supporting Information) are applied here. All of the bonds interact via harmonic potentials 25 1 E2 (r) = K2 (r − L0 )2 2

(4)

with spring constant K2 and equilibrium length L0 . The bond bending stiffness is described by E3 (r) = K3 [1 − cos(θ − θ0 )]

5 ACS Paragon Plus Environment

(5)

Langmuir

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 6 of 31

with force constant K3 and equilibrium angle θ0 . Proper dihedral potentials E4 (r) = K4 [1 − cos(φ − φ0 )]

(6)

are used for dihedrals. Here, φ is the torsion angle between planes containing beads i, j, k and j, k, l, and K4 is the force constant. The equilibrium CG bond lengths, angles, and the respective force constants of lipids and amino acids are obtained by fitting the bond distributions derived from AAMD simulations 41 or the Protein Data Bank (PDB); these are given in in the Supporting Information. To mimic the hydrogen bonding that stabilizes the α -helix or β -structures of the polypeptides or proteins, dissociable Morse potentials are employed, 38 EM (r) = KM [1 − e−α (r−re ) ]2 .

(7)

Here, re is the equilibrium distance, α is the width of the potential well, and KM is the depth of the potential well. To model the helical structure, 1-3 Morse bonds between skeleton beads separated by two harmonic bonds and 1-5 bonds between skeleton beads separated by four harmonic bonds are introduced. We set re13 = 0.6r0 , re15 = 0.85r0 , KM13 = 6kB T , and KM15 = 9kB T to obtain the "vertical step" of the α -helix. To mimic the β -structure, the angle constraint in Eq. 5 with

θ0 = 180◦ is applied to each rigid stand. Beads that belong to two different stands are allowed to form Morse bonds with re = 0.6r0 and KM = 12kB T , to constrain the sheet structure. In contrast to models that use specific torsion potentials to enforce the secondary structure of the proteins, the dissociable H-bonds enable better modeling of changes in the secondary structure while preserving proper hydrodynamics. 38 Electrostatic interactions between charged beads are calculated by the method introduced by Groot. 24 The charges are distributed on the lattice, and the electrostatic potential is solved locally

6 ACS Paragon Plus Environment

Page 7 of 31

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Langmuir

on the grid using a real-space successive over-damped relaxation method

ψ (ri ) = ψ (ri ) + ζ [Γρ¯ e (ri ) + ∇ · (P(ri )∇ψ (ri ))].

(8)

Here, ζ = 0.15 is the analog of a friction factor, Γ = e2 /kB T ε r0 = 13.87 is a coupling constant at room temperature, ρ¯ e is the average local charge density, and P is the polarizability relative to pure water, which is 1 for polar and charged beads, 0.35 for non-polar beads, and 0.025 for apolar beads. In a constant pressure ensemble, the force and virial contributed by the self-energy term must be subtracted from the total force of each charged bead and the total virial, respectively, which were discussed in detail in our previous paper. 31 The initial configuration of a single phospholipid or ganglioside lipid is obtained from an MD simulation. To prepare bilayer membranes, 1568 (or 512) phospholipid molecules (DPPC or DOPC) are placed on square grids in the center of an Lx × Ly × Lz sized box with the head groups on the outside of the membrane and the alkyl chains inside the membrane (Fig. 1a). Then, 72 (or 32) GM1 molecules are evenly distributed in both of the leaflets of the membrane. In a biological system, GM1 may reside only in the outer leaflet of the membrane. In simulations, the set of symmetric distributions of GM1 is to achieve flat planar bilayers. The initial values of Lx = Ly are determined by the desired projected area per lipid a pr j , and Lz is set to 32r0 . Here, r0 corresponds to 0.65 nm. 23 Water beads and counterions (if needed) are distributed randomly in the space that is unoccupied by the membrane. The whole system has bead density ρ = 3/r03 . DPD simulations are performed with a constant surface tension, under normal pressure, and with a constant temperature, (Nγs P⊥ T ensemble), in combined with periodic boundary conditions. Surface tension is a macroscopic quantity that is defined as the average of the difference between the normal and tangential pressure multiplied by the dimension of the simulation box in the direction normal to the bilayer. Here, we employ the Langevin piston approach 42 to maintain a constant normal pressure P⊥ =68 kB T /r03 and zero surface tension by adjusting the length of the simulation box being perpendicular to the bilayer normal. Here, 68 kB T /r03 is the pressure for bulk water

7 ACS Paragon Plus Environment

Langmuir

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

at a density of ρ = 3/r03 . Simulations are performed using the velocity-Verlet algorithm with a time step of 0.02t0 , which corresponds to 2.84 ps in physical units. 23 Each trajectory is run for 1,400,000 time steps (approximately 4.0 µ s) for data collection. The initial CG structure of the CTB is obtained from the PDB (ID: 1XTC). 7 The five identical B subunits form a pentameric planar ring with a central pore. The polypeptide of each B subunit has 103 amino acid residues distributed among two α -helix and six β -strands. The β -strands form two sets of three-stranded anti-parallel sheets in a closed β -barrel. The short α -helix (residues 4 to 12) is anchored to the interior of the β -stands and caps the outer surface of the β -barrel. The long helix caps the β -barrel in the central pore with its hydrophilic face forming the boundary wall. The

β -strand (residues 26 to 30) of one subunit forms hydrogen bonding with the β -strand (residues 95 to 100) of its nearest subunit. The CTB pentamer is held together via hydrogen bonding and hydrophobic interactions. To investigate the binding of CTB to the membrane, one to six CTB units are placed at approximately 2 nm above the surface of a pre-relaxed bilayer with zero surface tension. The box is then refilled with water and counterions. Pre-simulations are run for 50,000 time steps with fixed CTB and lipid positions in a constant volume and constant temperature (NV T ) ensemble, in such a way that the solvent approaches equilibrium. Full simulations are then run for 2.84 µ s in the Nγs P⊥ T ensemble. For each system, five parallel samples are simulated to collect data. All of the DPD simulations are performed using a home-made code package.

Results and Discussion Mixed GM1/DPPC bilayer. A pure DPPC bilayer is in the gel phase at a reduced temperature T = 1 (corresponding to 298 K). The adding of 72 GM1 to a bilayer composed of 1568 DPPC lipids (GM1/DPPC ≈ 5 mol %) results in separated gel and interdigitated phases as shown in Figure 1. The formation of a striped pattern might be caused by the finite size effect in the periodic boundary condition. 43 When up to 162 GM1 (GM1/DPPC ≈ 10 mol %) are added to the bilayer, a pure interdigitated phase is obtained (Figure S5 in the Supporting Information). In the interdigitated

8 ACS Paragon Plus Environment

Page 8 of 31

Page 9 of 31

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Langmuir

phase, the alkyl tails of the GM1 lipids either tilt or penetrate into the opposite leaflet of the membrane, as shown in Fig. 1c. As a consequence, the mixed GM1/DPPC bilayer is less ordered than a pure DPPC bilayer. The bead density profile in Fig. 2a gives the distribution of the head groups, which indicates that the branched head groups of GM1 protrude out from the surface of the bilayer. The Gal2 group extends the most into the solvent; this group is approximately 1.5 nm away from the phosphate group of the DPPC, which is consistent with both experimental and simulation data. 19,44,45 Because GM1 and DPPC lipids have distinct structures and amphilicities, GM1 molecules tend to associate into dimers or trimers. The association can be quantitatively estimated by calculating the preferential partitioning, which is defined as the relative number of contacts between the GM1 molecules 39 Pgm1 =

Cgm1 /Ngm1 . Cgm1 /Ngm1 +Cd ppc /Nd ppc

(9)

Here, Cgm1 is the number of contacts between the GM1 molecules, and Cd ppc is the number of contacts between the GM1 and DPPC molecules. Ngm1 and Nd ppc are the total numbers of GM1 and DPPC lipids, respectively. Two molecules are considered to be in contact if the distance between their centers of mass is within 1.2 nm. The time-dependent preferential partitioning function of GM1 given in Fig. 2(b) shows that after 1 µ s, the system approaches equilibrium, with the number of contacts fluctuating around a constant number. This finding demonstrates that a GM1 lipid has a 50% probability of associating with another GM1. We note that the preferential partitioning obtained here exhibits strong fluctuations, which implies that the GM1 molecules associate and dissociate from time to time. This is in agreement with experiment that at low GM1 concentration (