You are on page 1of 15

Article

Cite This: Langmuir 2017, 33, 11496-11510 pubs.acs.org/Langmuir

Assessing the Structure and Stability of Transmembrane Oligomeric


Intermediates of an α‑Helical Toxin
Rajat Desikan,† Prabal K. Maiti,‡ and K. Ganapathy Ayappa*,†,§

Department of Chemical Engineering, ‡Centre for Condensed Matter Theory, Department of Physics, and §Centre for Biosystems
Science and Engineering, Indian Institute of Science, Bengaluru, India 560012
*
S Supporting Information
Downloaded via INDIAN INST OF TECHNOLOGY KANPUR on February 15, 2019 at 13:28:09 (UTC).
See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

ABSTRACT: Protein membrane interactions play an important role in our understanding of diverse phenomena ranging from
membrane-assisted protein aggregation to oligomerization and folding. Pore-forming toxins (PFTs) are the primary vehicle for
infection by several strains of bacteria. These proteins which are expressed in a water-soluble form (monomers) bind to the target
membrane and conformationally transform (protomers) and self-assemble to form a multimer transmembrane pore complex
through a process of oligomerization. On the basis of the structure of the transmembrane domains, PFTs are broadly classified
into β or α toxins. In contrast to β-PFTs, the paucity of available crystal structures coupled with the amphipathic nature of the
transmembrane domains has hindered our understanding of α-PFT pore formation. In this article, we use molecular dynamics
(MD) simulations to examine the process of pore formation of the bacterial α-PFT, cytolysin A from Escherichia coli (ClyA) in
lipid bilayer membranes. Using atomistic MD simulations ranging from 50 to 500 ns, we show that transmembrane oligomeric
intermediates or “arcs” form stable proteolipidic complexes consisting of protein arcs with toroidal lipids lining the free edges. By
creating initial conditions where the lipids are contained within the arcs, we study the dynamics of spontaneous lipid evacuation
and toroidal edge formation. This process occurs on the time scale of tens of nanoseconds, suggesting that once protomers
oligomerize, transmembrane arcs are rapidly stabilized to form functional water channels capable of leakage. Using umbrella
sampling with a coarse-grained molecular model, we obtain the free energy of insertion of a single protomer into the membrane.
A single inserted protomer has a stabilization free energy of −52.9 ± 1.2 kJ/mol and forms a stable transmembrane water channel
capable of leakage. Our simulations reveal that arcs are stable and viable intermediates that can occur during the pore-formation
pathway for ClyA.

■ INTRODUCTION
The eukaryotic cellular plasma membrane, composed mainly of
and infection propagation.5,6 The water-soluble PFT proteins
expressed by the organism (referred to as monomers)
amphipathic lipid molecules, serves as a semipermeable barrier spontaneously recognize and bind to the target plasma
membrane or to specific protein or lipid receptors.2,5 Upon
compartmentalizing the cytosolic cellular contents from the
binding to the membrane, monomers undergo a conforma-
extracellular environment and is vital for life. Pore-forming
tional change into a membrane-bound form (termed a
toxins (PFTs) are an ubiquitous class of lytic proteins that
protomer). The assembly-competent membrane-bound proto-
primarily undermine the structural integrity of the plasma
mer spontaneously undergoes membrane-assisted oligomeriza-
membrane and cause cell death by disrupting the osmotic and
ionic gradients across the cellular plasma membrane.1−5 The
pathogenesis of several bacterial infections in humans caused by Special Issue: Tribute to Keith Gubbins, Pioneer in the Theory of
Vibrio cholerae, Listeria monocytogenes, Bacillus anthracis, Liquids
Escherica coli, and Staphylococcus aureus is primarily PFT- Received: July 1, 2017
mediated; transmembrane pore formation causes unregulated Revised: September 19, 2017
transport across the cell membrane, thus leading to cell death Published: September 20, 2017

© 2017 American Chemical Society 11496 DOI: 10.1021/acs.langmuir.7b02277


Langmuir 2017, 33, 11496−11510
Langmuir Article

Figure 1. Schematics of the widely accepted prepore pathway and the growing pore pathway for PFTs, using ClyA as the model PFT, are illustrated
(top view of the membrane). The various membrane-bound pore-forming intermediates of the dodecameric ClyA pore (single protomers, trimers,
hexamers, and nonamers) are shown in red and are derived from the crystal structure of ClyA (PDB ID 2WCD), the membrane lipids are
represented as green circles, and water molecules in the pore lumen are represented as blue circles. In the prepore pathway, none of the
intermediates apart from the fully formed pore are capable of leakage.

tion with other protomers to form higher-order oligomeric pore transmembrane partially oligomerized pores or arcs that are
intermediates. The process of oligomerization is terminated lined by a lipid edge.1,11,12 The paucity of high-resolution
upon formation of the final multimeric transmembrane pore structures of membrane-inserted PFT pores and pore-forming
complex, whose diameters typically range from a few to tens of intermediates impedes our understanding of pore-forming
nanometers.2 PFTs are primarily classified as α- and β-PFTs, mechanisms. Although water-soluble monomeric structures of
depending on whether the transmembrane domains of the pore several PFTs have been resolved, the resolution of membrane-
assume amphiphilic α-helices or β-sheets as the dominant bound intermediates, such as the protomeric and intermediate
secondary structures.2 Unlike protein aggregation, which is oligomeric states, and of the final transmembrane pores is
driven primarily by misfolded states, the transmembrane pores challenging. In addition, tracking PFT oligomerization on
formed by the PFTs have a definite oligomeric structure membranes is a formidable challenge since pore-forming
implying the underlying stability and robustness associated with proteins lie well below the optical resolution limit, and the
this process of membrane-assisted self-assembly. In contrast to dynamics are typically complete on time scales of hundreds of
PFTs, integral host membrane proteins (especially those with microseconds.13 Ionic conductance and vesicle leakage experi-
α-helical transmembrane domains) are partitioned into the lipid ments provide indirect evidence of the underlying pore kinetics
membrane by the translocon protein complex, an active process and morphology.14,15 Hence, proposed mechanisms are either
that requires energy supply by GDP−GTP exchange and derived from indirect observations of pore assembly or
multiple cellular machineries to work in tandem.7,8 Therefore, reconstructed from the few available crystal structures of the
PFTs are unique from a membrane−protein perspective PFT monomers and the membrane-inserted pores.1,16
because the toxins are originally secreted in a water-soluble While pore structures and assembly mechanisms of β-PFTs
form but eventually sample the membrane environment to such as Staphylococcus aureus α-hemolysin,17 Staphylococcus
form multimeric integral transmembrane channels.1,2 aureus γ-hemolysin,18 and Vibrio cholerae cytolysin19 have been
Many mechanisms for understanding protomer oligomeriza- extensively studied, molecular mechanisms of α-PFT oligome-
tion on the target membrane have been proposed. According to rization and lipid reorganization during pore formation are
the widely accepted prepore paradigm (illustrated in Figure 1), poorly understood.1,2,4,14,20−23 Intrinsic participation of the
protomers oligomerize on an intact membrane into a fully membrane lipids along with the inserted transmembrane helices
formed prepore, and transmembrane pore formation occurs to form the pore complex is a recurring theme for α-PFTs.14,24
with the insertion of the prepore assembly into the plasma Escherichia coli cytolsin A (ClyA) is currently one of only two
membrane. This transition can be accompanied by a large α-helical PFTs with available crystal structures of both the
conformational change as in the case of cholesterol-dependent water-soluble monomer25 and the membrane-inserted pore
cytolysins such as pneumolysin.9 The prepore pathway has complex.16 Thus, ClyA can be conveniently used to decipher
been experimentally verified for many members of the class of the assembly mechanism for the class of α-helical PFTs.16,20,26
β-PFTs such as α-hemolysin and γ-hemolysin or the The crystal structure of the pore reveals the formation of a
cholesterol-dependent cytolysins such as perfringolysin O.10 dodecameric homo-oligomeric complex having a total length of
Alternately, in the growing pore paradigm (illustrated in Figure 13 nm, with the inner channel diameter varying from 7 nm at
1), membrane-inserted pore intermediates also known as arcs the extracellular end to 4 nm at the transmembrane or cytosolic
grow by oligomerization to form the pore complex. This end.16 The current understanding and challenges regarding the
pathway is supported for some PFTs by the observation of pore-forming pathway of ClyA are detailed below.
11497 DOI: 10.1021/acs.langmuir.7b02277
Langmuir 2017, 33, 11496−11510
Langmuir Article

Figure 2. Top and front views of starting configurations of the ClyA arcs (6-mer to 10-mer and the 12-mer) are illustrated here. The pink portion of
the arcs corresponds to the solvent-exposed domain, and cyan represents the transmembrane domain. The n-mer complexes were formed by simply
removing the required number of pore subunits from the 12-mer. During the course of all atomistic simulations, the arc structures do not deviate
significantly from the initial configurations shown here (discussed later).

The water-soluble ClyA monomer initially binds to the target ClyA pore as well as ClyA arcs at various levels of
membrane, which contains cholesterol that acts as a lipid oligomerization (6-mer to 10-mer). In both atomistic and
receptor for ClyA,5 and subsequent membrane-assisted coarse-grained simulations, we observe that the 1-mer is stable
oligomerization results in the formation of the transmembrane and creates a transmembrane water channel. Coarse-grained
pore complex. Surprisingly little is known about ClyA’s simulations using the MARTINI force field (with polarizable
oligomerization pathways, the molecular role for cholesterol, water44 and PME electrostatics) and umbrella sampling are
and the fate of lipids during pore formation.20,21 ClyA pore used to obtain the free energy of protomer−membrane binding.
intermediates (termed arcs) in electron micrographs have been These simulations show that a single membrane-inserted
observed upon inducing pore formation in detergent,27,28 in protomer is a thermodynamically favorable state. To address
planar lipid monolayer films,29 and in single-molecule experi- the issue of partially oligomerized intermediates, we performed
ments in detergent.30 Combined experimental and modeling to our knowledge for the first time extensive all-atom molecular
studies have shown that these ClyA arcs (1-mer to 11-mer) dynamics simulations of 6-mer to 10-mer arcs in both saturated
mostly oligomerize in a reversible and sequential manner on the DMPC (1,2-dimyristoyl-sn-glycero-3-phosphocholine) and un-
membrane,15,31 though nonsequential oligomerization has been saturated POPC (1-palmitoyl-2-oleoyl-sn-glycero-3-phospho-
shown in detergents.30,32 While ClyA was previously assumed choline) lipid membranes. In all simulations carried out over
to follow the prepore pathway16 where only the 12-mer causes 50−500 ns, the proteolipid arclike structures retain their
leakage, an experimental and modeling study (performed in our structural integrity, indicating that these purported structures
laboratory) of ClyA-induced calcein dye leakage from small are stable. We observe that lipids initially present in the arc
unilamellar vesicles15 has shown that 5-mer to 12-mer ClyA spontaneously evacuate the arc interior, leading to the
arcs cause leakage. An unlikely noncanonical pathway, where formation of a transmembrane water channel, and we discuss
intact ClyA prepores oligomerize in the absence of membranes the possible biological implications of our observations.
and directly insert into intact target membranes upon contact,
had previously been proposed,33 but this mechanism has been
challenged34 and is not considered in this study.
■ COMPUTATIONAL METHODS
All-Atom MD Simulations. The dodecameric crystal structure of
Several questions which require a molecular understanding of the ClyA pore channel (PDB ID 2WCD) has unresolved N-terminal
pore formation remain unanswered in this extended paradigm residues 1−7 and C-terminal residues 293−303, and these residues
of ClyA pore formation. Two of them are the following: (i) Is a were added to the crystal structure as described previously.35 Standard
membrane-inserted protomer a favorable state, and can it form residue protonation states according to the force field definitions have
a water channel? (ii) How do the membrane-inserted arcs clear been assumed for all protein molecules in this study. Coordinates of
the central lipids corresponding to the pore lumen to form the ClyA protomer as well as that of partial n-mer pore intermediates
stable water channels capable of leakage? Because experiments or arcs were obtained from the reconstructed crystal structure. The
implicit assumption here is that in the absence of further structural
are often resolution-limited on spatiotemporal scales, we information on these elusive pore intermediates it is reasonable to
attempt to answer these questions in this article with evidence assume that the protomers’ structure and assembly in the arcs are not
from fully atomistic and coarse-grained molecular dynamics drastically different from those of the protomers in the crystal structure
simulations. Fully atomistic MD simulations of PFTs enable us of the ClyA pore. As mentioned in the Introduction, multiple studies
to observe equilibrium configurations of these large mem- have observed ClyA arcs, and combined experimental and kinetic
brane−protein complexes at molecular resolution and have modeling studies have predicted their existence.15,27−32,34 Even if the
been used to study the structure and properties of the ClyA structures were slightly different, molecular dynamics simulations
pore,35−37 the mechanism of conformational transition from enable conformational sampling and thus often yield energetically
relaxed equilibrium structures at molecular resolution, which can be
the soluble ClyA monomer to the membrane-bound proto- utilized to deduce the mechanism and energetics of pore formation.
mer,38 and transport properties of various molecules through The 6-mer to 10-mer complexes and the 12-mer are illustrated in
the pore lumen of the β-PFT, α-hemolysin.39−43 Figure 2 and are obtained by deleting the requisite number of
In this article, we carry out MD simulations of the single protomers from the pore crystal structure. All of the MD simulations
transmembrane ClyA protomer (1-mer) and the full 12-mer presented in the main text of this article were performed using

11498 DOI: 10.1021/acs.langmuir.7b02277


Langmuir 2017, 33, 11496−11510
Langmuir Article

GROMACS software version 4.6.4 (www.gromacs.org),45 and some Table 1. Complete List of All of the Production Simulations
simulations in the Supporting Information were performed using Performeda
GROMACS software version 2016.3 (see the Supporting Information
for details). Visualization of the initial and final configurations obtained NaCl
from the simulations was performed with VMD version 1.9.3.46 simulation concentration
protein membrane ensemble time (ns) (M)
Additional analyses were carried out using self-written tools in
MATLAB version R2017a. ClyA protomer (1- DMPC NPT 500 0.15
The single ClyA protomer as well as the n-mer arcs were inserted mer)
into pre-equilibrated DMPC (1,2-dimyristoyl-sn-glycero-3-phospho- solvated ClyA NVT 300 0.15
protomer
choline) or POPC (1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine)
membranes by deleting the overlapping lipids between the protein and ClyA arc (6- DMPC NPT 224 charge-neutral
mer)
membrane. The lipid−protein complex was solvated using TIP3P47
ClyA arc (6- POPC NPT 50 charge-neutral
water molecules, with adequate solvent buffers on both the solvent- mer)
exposed side and the cytosolic side of the ClyA arcs. Each ClyA
ClyA arc (6-mer, POPC NPT 20 0.15
protomer carries a positive charge of 7, hence a sufficient number of CHARMM36, SI)
Na+ counterions (42, 49, 56, 63, 70 for the 6-mer, 7-mer, 8-mer, 9-mer, ClyA arc (7- DMPC NPT 50 charge-neutral
and 10-mer, respectively) were added to create charge-neutral initial mer)
configurations. The 12-mer simulations performed previously35 and ClyA arc (7- POPC NPT 50 charge-neutral
the 1-mer simulations were carried out at 0.15 M NaCl concentration. mer)
Upon solvent and ion addition, the solvent molecules in contact with ClyA arc (8- POPC NPT 50 charge-neutral
the lipid core in the bilayer−protein interstices were removed. The mer)
AMBER 99SB-ILDN force field48 with ϕ corrections,49 which was ClyA arc (9- DMPC NPT 50 charge-neutral
shown to perform well for proteins,50 was used to describe the intra- mer)
and intermolecular interactions of the protein and ions. The AMBER ClyA arc (9- POPC NPT 50 charge-neutral
force field-compatible Slipid parameters51 were used to model the mer)
DMPC and POPC lipids. ClyA arc (9-mer, POPC NPT 20 0.15
Electrostatic interactions were computed using the particle mesh CHARMM36, SI)
Ewald method52 with a 1.0 nm real space cutoff, and the van der ClyA arc (10- DMPC NPT 289 charge-neutral
mer)
Waal’s interactions were computed with a 1.0 nm cutoff. All bonds
were constrained using the LINCS algorithm.53 A leapfrog integrator ClyA arc (10- POPC NPT 50 charge-neutral
mer)
with a 2 fs integration time step was used along with Verlet buffered
ClyA pore DMPC NPT 100 0.15
lists (target energy drift of 0.005 kJ/mol/ns per atom). Neighbor lists (12-mer)
were updated every 10 steps; three-dimensional periodic boundary
ClyA protomer DMPC NPT 2400 charge-neutral
conditions were used. For the single solvated protomer simulations in (MARTINI, PMF)
the absence of a membrane, simulations were carried out with the ClyA protomer DMPC NPT 200 charge-neutral
above parameters but in the NVT ensemble without pressure coupling. (MARTINI, Figure
A synopsis of all simulations performed along with relevant detail is S2)
shown in Table 1. a
All of the simulations were carried out with explicit solvent and at 310
All systems were energy minimized for 10 000−100 000 steps using K. For charge-neutral simulations, only Na+ counterions are present to
the steepest-descent method (step size 0.01 nm). Initially, 200−500 ps ensure charge neutrality.
runs in the NPT ensemble with harmonic restraints on the protein
atoms were performed. Subsequently, 50- to 500-ns-long runs in the
NPT ensemble were performed for the ClyA protomer and all arcs in with spring constants of 100 kJ mol−1 nm−2. This is because windows
DMPC and POPC membranes as well the 12-mer pore in a DMPC 11 and 12 correspond to the water−bilayer interface and exhibit large
membrane (Table 1). We used the Nosé−Hoover thermostat54,55 or fluctuations in ζ for weak harmonic restraints. All simulations were
the velocity-rescale thermostat56 to control the system temperature at performed in the NPT ensemble, and the simulation parameters and
310 K by using a time constant of 0.5 or 0.1 ps, respectively. System choice of algorithms are similar to that used in the all-atom simulations
pressure corresponding to 1 bar was maintained in the plane of the (described above). Briefly, a leapfrog integrator with a 20 fs integration
membrane as well as in the direction normal to the membrane by time step was used, PME electrostatics were employed with an
using the semi-isotropic Parrinello−Rahman pressure-coupling effective dielectric constant of ϵr = 2.5 and a cutoff of 1.2 nm, the
scheme57 (isothermal compressibilities of κxy = κz = 4.5 × 10−5 velocity-rescale thermostat56 was used to set the system temperature at
bar−1, time constant of 10 ps). 310 K with a time constant of 0.1 ps, and the semi-isotropic
PMF of ClyA Protomer Membrane Binding from MARTINI Parrinello−Rahman pressure coupling scheme57 with a time constant
Simulations with Polarizable Water. The single ClyA protomer in of 12 ps was utilized to equilibrate the unit cell.
a DMPC membrane was set up using the MARTINI force field To ensure that the protein’s internal conformation is at equilibrium
(Elnedyn 2.2) with a polarizable water model44 and PME in order to at each window during umbrella sampling, we examine the RMSD of
capture the electrostatics accurately, especially at the complex the full protein and helix αA1 from the 100-ns-long unrestrained
protein−membrane−water interfaces. Upon equilibrating the ClyA equilibrium simulations of window 24 (described in Figure S2). We
protomer in the DMPC membrane for 100 ns, the steered MD feature find that the conformational change of the N-terminus is complete by
of GROMACS was used to create 24 initial configurations along the 20 ns (Figure S3), and a final snapshot after 100 ns is illustrated in
reaction coordinate ζ (defined as the distance between the centers of Figure S4. Hence, the first 20 ns of each window is censored, and the
mass of the membrane and protomer) such that each window was resulting PMF and associated binding energy are reported in Figure 5.
separated at intervals of Δζ = 0.2 nm. The phosphate atoms of the The effect of varying the censoring threshold on the PMF is reported
membrane were restrained in the Z coordinate using harmonic in Figure S5, where we find that the PMF is relatively insensitive to the
restraints of 100 000 kJ mol−1 nm−2. Solvent and unit-cell equilibration censoring threshold. The potential of mean force along the reaction
simulations for 2 ns and umbrella sampling simulations for 100 ns per coordinate was obtained by using the weighted histogram analysis
window were carried out with a weak restraining harmonic force at method (WHAM; implemented in the g_wham58 tool in GROMACS
that particular value of ζ. A spring constant of 10 kJ mol−1 nm−2 was version 4.6.4.) on histograms from the umbrella sampling trajectories,
used for all windows (see Figure S2 for the rationale behind choosing and the final PMF was generated by dividing histograms along the
this value), except windows 11 and 12 which were simulated again reaction coordinate into 200 bins. The error bars on the PMF were

11499 DOI: 10.1021/acs.langmuir.7b02277


Langmuir 2017, 33, 11496−11510
Langmuir Article

Figure 3. (A) Initial and final snapshots of a single ClyA protomer and the ClyA pore inserted into DMPC membranes (the N-terminal helix αA1 is
colored green, the C-terminal helix αG is colored blue, and the β-tongue consisting of residues 177−203 is colored magenta). (B) RMSD traces of
the various conformations of ClyA show structural differences. (C) Comparison of the equilibrium RMSFs of the single transmembrane ClyA
protomer (averaged over various blocks of time) with a protomer subunit in the transmembrane ClyA pore. (D) Axial tilts of helices αA1 and αB
with respect to the Z axis are plotted as a function of simulation time for the single protomer and are compared with the tilts averaged over all
protomer subunits in the ClyA pore. The axes for helices αA1 and αB are defined as vectors passing through the center of mass of residues 11 and 33
and residues 38 and 99 (shown as labeled yellow beads in the protomer on the right), respectively. (E) Temporal trend in the radius of gyration (Rg)
of the single protomer.

computed by generating 2000 bootstrapped trajectories from the time of λCOM(t), estimated by fitting a single exponential to the 6-mer
obtained histogram such that the new random trajectories had similar and 10-mer trajectories, was found to be τ6 = 1.13 ns and τ10 = 2.04 ns,
converged histograms and autocorrelation times. respectively. Subsequently, λCOM(t) from the 6-mer and 10-mer
Spring Constant and Force between the Edge Protomers trajectories was divided into blocks of length τ6 and τ10 after discarding
from Atomistic MD Simulations. If the distance between the edge the first 50 ns of both trajectories, and k and F were computed for
protomers in the arcs is normally distributed around a mean, then every block from eq 5. The mean k and F values were computed by
these edge protomers can be assumed to be connected by a virtual averaging over all of the blocks, and error bars were computed by 1-
harmonic spring with a spring constant k and a restoring force F. deletion block-jackknife resampling.59,60 From this method, the mean
Assuming a classical Hamiltonian for the spring in the canonical values of k and F for the 6-mer were calculated to be 39.68 ± 10.56
ensemble, the equation for the spring constant (in kJ mol−1 nm−2) at a pN/nm and 306.06 ± 81.42 pN, respectively, and for the 10-mer,
temperature of 310 K can be derived as shown below. 11.39 ± 3.78 pN/nm and 76.61 ± 25.03 pN, respectively.


The Hamiltonian of a one-dimensional harmonic oscillator is given
as RESULTS
2 2
p kr Membrane-Bound ClyA Protomer Retains Structural
/(p , r ) = +
2m 2 (1) Integrity and Stabilizes a Continuous Transmembrane
where r is defined as the normally distributed Euclidean distance Water Channel. A key factor in the pore-formation process of
between the center of mass of the two edge protomers. The variance in PFTs such as ClyA is the conformational transformation from a
the mean position can be calculated as water-soluble monomer to an assembly-competent protomer.
To the best of our knowledge, experimental data pertaining to
∬ r 2e(−β /(p , r)) dp dr ⎜⎛ ∬ r e(−β /(p , r)) dp dr ⎞⎟
2
the energetics of membrane binding of ClyA is absent in the
⟨r 2⟩ − ⟨r ⟩2 = −⎜ (−β /(p , r )) dp dr ⎟ literature. First, we determine if the single transmembrane
∬ e(−β /(p , r))dp dr ⎝ ∬e ⎠ protomer is structurally stable relative to a protomer bound in
(2) the fully oligomeric pore. To contrast the structural attributes
2 2 of ClyA in these two conformations (single transmembrane
∫ e(−βp /2m)
dp ∫ r 2e(−βkr /2)
dr
protomer and bound protomer in the pore complex), a 500-ns-
⟨r 2⟩ − ⟨r ⟩2 = 2 2
(−βp /2m) (−βkr /2) long all-atom MD simulation of a single transmembrane ClyA
∫e dp ∫ e dr
protomer and a 100-ns-long all-atom simulation of the
⎛ (−βp2 /2m) 2 ⎞ 2
∫e dp ∫ r e(−βkr /2) dr ⎟ transmembrane ClyA pore were performed. The initial and
−⎜ final simulation snapshots are illustrated in Figure 3A.
⎜ (−βp2 /2m) 2 ⎟
⎝∫e dp ∫ e(−βkr /2) dr ⎠ (3) The RMSD traces with respect to the initial simulation
Because these are Gaussian integrals, we get structures of the single membrane-inserted protomer and the
pore are contrasted (illustrated in Figure 3B). The single
2

⟨r 2⟩ − ⟨r ⟩2 =
(
κBT 1 − π ) membrane-inserted protomer has a RMSD with larger
k (4) fluctuations when compared to that of the pore. A comparison
of the per-residue root-mean-square fluctuations (RMSFs)
Hence, between the single transmembrane ClyA protomer in
2 2 comparison to protomer subunits in the dodecameric pore
k=
(
κBT 1 − π) and F = (
κBT 1 − π )⟨r⟩ (RMSF averaged over all 12 protomer subunits) is illustrated in
2 2
⟨r ⟩ − ⟨r ⟩ ⟨r ⟩ − ⟨r ⟩2
2
(5) Figure 3C. The fluctuations of all of the residues in the single
where magnitude of the force (F) is in pN molecule . Because we −1 protomer are consistently higher than the residues in the pore,
report only single-molecule results, we write the units of force simply and this is attributed to the absence of stabilizing protein−
as pN. From now onward, r(t) is referred to as λCOM(t). Results for the protein interactions for the single protomer. Changes in the
6-mer and the 10-mer are calculated as follows. The autocorrelation orientation of the single transmembrane protomer with respect
11500 DOI: 10.1021/acs.langmuir.7b02277
Langmuir 2017, 33, 11496−11510
Langmuir Article

to the membrane are illustrated by a comparison of the helix hydrophobic surface interacts with the hydrophobic lipid acyl
tilts of helices αA1 and αB, as shown in Figure 3D. It can be chains, whereas the hydrophilic surface stabilizes the sponta-
seen that the helix tilts of the single protomer exhibits drift as neously formed transmembrane water channel.
well as large fluctuations from the initial value, while a protomer Because the equilibration of lipids and solvent around the
in the pore is stable. This is consistent with previous Go-model protein is crucial, we carried out simulations with an alternate
simulations of the ClyA protomer in the presence of an explicit equilibration protocol and a completely different force field,
model membrane,38 where the protomer samples a large CHARMM36.61,62 The single protomer was inserted into a
number of helix tilt angels in the membrane-inserted state. The POPC membrane using the CHARMM-GUI membrane
temporal trend of the radius of gyration (Rg) of the single builder63−65 (http://www.charmm-gui.org/), ensuring that
protomer is shown in Figure 3E, and we observe stable trends lipids were completely packed around the protein, with
by the end of the 500-ns-long simulation. Overall, these results extensive unit cell and solvent equilibration simulations (details
confirm that the protomer in the pore is a more favorable state in Supporting Information). Using this initial configuration, 20
because of protein−protein interactions with neighboring ns simulations revealed the formation of the water channel as
protomers, but the single protomer does not lose its structural discussed above, indicating that the formation of the water
integrity in the absence of the other protomers. channel is a robust phenomenon (Figure S6).
The all-atom MD simulations of the single membrane- From this finding, it is conceivable that the oligomerization
inserted protomer (described above) also show for the first of the protomers in the noninserted state extends this
transmembrane domain into a Janus surface, with the alkyl
time to our knowledge that a stable and continuous membrane-
lipid tails stabilized by the hydrophobic outer surface and the
spanning water column is stabilized by the hydrophilic face of
aqueous channel in the pore lumen stabilized by the inner
the amphipathic transmembrane N-terminal helix (Figure 4A;
hydrophilic surface, as described in the growing pore
also see Figure S1). Transmembrane amphipathic helix αA1 is mechanism (Figure 1). Another α-PFT fragaceatoxin C also
cleanly separated into hydrophobic and hydrophilic surfaces, possesses an amphipathic N-terminal helix with a Janus surface
thus resembling a Janus surface (illustrated in Figure 4A). The topography which forms the transmembrane domain of the
membrane-inserted pore.24 These findings for ClyA indicate
that oligomerization and membrane leakage can be initiated by
a single membrane-inserted protomer.
We next assess the binding free energy of a single ClyA
protomer to a DMPC membrane by employing coarse-grained
simulations (MARTINI force field with polarizable water and
full electrostatics) with the umbrella sampling technique. We
point out that although it is desirable to obtain the PMF using
an all-atom description, complete umbrella sampling, given the
size of the protein and the corresponding simulation box, is
computationally prohibitive. Furthermore, MARTINI force
fields have been know to capture protein membrane
interactions quite accurately.66,67 The free-energy profile is
shown in Figure 5. It can be clearly seen that membrane
binding by ClyA is thermodynamically favorable, with a free-
energy gain of ΔE = −52.9 ± 1.2 kJ/mol, or ∼ −12.6 ± 0.3
kcal/mol, at ζ ≈ 6.0 nm for membrane-inserted conformations
relative to completely solvated conformations (configurations
illustrated in Figure 5). This is similar to a previous estimate of
−46 kJ/mol for ClyA.38
The membrane-inserted protomer (from the initial window)
of the coarse-grained simulations also stabilizes a trans-
membrane water channel as shown in Figure 4B. In addition
to this, the protomer is completely solvated in the final
windows of the umbrella sampling simulations and is seen to
undergo significant structural distortions in the N-terminus and
helix αA1. The final snapshot after 100 ns from umbrella
Figure 4. (A) Instantaneous snapshot at 300 ns of the transmembrane sampling simulations of configuration 24 is magnified in Figure
protomer from an all-atom MD simulation shows the transmembrane S4, and structural distortions such as unfolding of the N-
water channel. Final snapshots at 500 ns are shown in Figure S1. terminus and helix αA1 (colored orange) can be observed.
Transmembrane residues 1 to 36, which comprise the N-terminus and Fully Solvated Protomer Is Structurally Unstable
the amphipathic helix αA1, are depicted below with the hydrophobic Compared to a Membrane-Inserted Protomer. To
residues colored green, hydrophilic residues colored red, and proline analyze the structural distortions of the fully solvated protomer
36 colored purple. (B) Snapshots (100 ns each) of the transmembrane
in detail, fully atomistic simulations of the transmembrane vs
protomer from the first window of the MARTINI umbrella sampling
simulations (also illustrated in Figure 5) show that the single protomer solvated protomer conformations were compared, and the
forms a water channel in coarse-grained simulations as well. Note that initial and final snapshots of the protomer in both solvated and
the center-of-mass distances between the protomer and the membrane transmembrane configurations are shown in Figure 6A−F. The
are restrained by a weak harmonic force in MARTINI, but all-atom evident distortion in the N-terminus and helix αA1 (residues
simulations are not. 1−34) is due to the exposure of hydrophobic residues in this
11501 DOI: 10.1021/acs.langmuir.7b02277
Langmuir 2017, 33, 11496−11510
Langmuir Article

Figure 5. Free energy of the protomer binding to the membrane is


shown using the potential of mean force (PMF) along reaction
coordinate ζ (distance between the center of mass of the membrane
and the center of mass of the protomer). The PMF was obtained by
simulating 24 configurations of the transmembrane protomer, with
overlapping histograms at intervals of Δζ = 0.2 nm for 100 ns each
(Methods). These simulations were carried out using the MARTINI
coarse-grained force field with polarizable water and full electrostatics.
Snapshots (100 ns) from the 1st window, 13th window, and last
window are illustrated along with the value of ζ at 100 ns. The protein
residues corresponding to the N-terminus and helix αA1 and the rest
of the protein are represented by orange and red vdW spheres,
respectively. PO4 membrane beads are shown as green vdW spheres.

amphipathic domain to the aqueous environment (also see Figure 6. (A−F) Initial and final simulation snapshots of the fully
Figure S4). This is corroborated by the time traces of the solvated protomer compared to the transmembrane protomer show
backbone root-mean-square deviations (RMSD) of this domain that the structure of residues 1−34 comprising the N-terminus and
(shown in Figure 6G) which plateau at a higher value for the helix αA1 is significantly distorted when the protomer is away from the
solvated protomer as well as the decreased helicity of these membrane, and these residues are completely exposed to the solvent.
residues in the solvated protomer as compared to that of the The ClyA protomer is colored red except for transmembrane residues
membrane-inserted protomer. The loss of structural integrity in 1−34, which are colored orange, and phosphate beads in the
the putative transmembrane domain of the solvated protomer membrane colored green. The solvent is omitted for clarity. (G)
Final snapshots (all-atom) of the membrane-inserted protomer and
(illustrated in Figure 6G) suggests that the membrane-bound the completely solvated protomer without a membrane highlight the
protomer conformation with the N-terminus residues bound to structural distortions at the N-terminus. Corroboratively, the RMSD of
the membrane is structurally favorable. The protomer can be in residues 1−34 is significantly higher for the fully solvated protomer
a noninserted state with the amphipathic N-terminus and helix than for the transmembrane protomer, indicating the greater stability
αA1 bound to the upper leaflet, or the protomer may span the of the transmembrane state. Secondary structural analysis (% helicity)
membrane and create a membrane channel as shown in Figure shows the loss of helical structure for the stretch of residues 1−34 in
4. Similar evidence for membrane insertion of the N-terminal the solvated protomer but not for the membrane-inserted protomer.
helix prior to oligomerization has been shown for other α-
helical pore-forming proteins such as Equinatoxin II68 and snapshots of the initial and final configurations (50 ns) of the 6-
Bax.69 mer and 9-mer ClyA arcs in DMPC bilayers and 6-mer to 10-
Spontaneous Formation of Membrane-Permeabiliz- mer arcs in POPC bilayers are shown in Figure 7A,C. By 50 ns
ing Proteolipid ClyA Arcs. In the previous section (Figure for all arcs, the lipids initially present in the hydrophilic arc
4), we observed that a single membrane-inserted protomer interior were spontaneously displaced into the surrounding
stabilizes a transmembrane water channel. In this section, we membrane, with a transmembrane water channel replacing the
investigate the stability of larger transmembrane arcs to displace displaced lipids. The displacement sequence for the lipids over
central lipids and form stable water channels in the membrane. the 50-ns-long simulation for the 7-mer arc in the DMPC
To observe the spontaneous formation of water channels, the membrane is illustrated in Figure 7B. This indicates that the
transmembrane arcs were initially inserted into previously well- favorable configuration for transmembrane ClyA arcs is the
equilibrated DMPC lipid bilayers with the lipids initially formation of a transmembrane water channel that solvates the
present in and around ClyA arcs. No water defect was present inner hydrophilic surface of the transmembrane domain.
at the start of simulations, and the membrane was intact except Results from all-atom simulations of the 6-mer and 10-mer
for the inserted protein complexes. Representative simulation arcs in DMPC with longer run times (up to 289 ns) show that
11502 DOI: 10.1021/acs.langmuir.7b02277
Langmuir 2017, 33, 11496−11510
Langmuir Article

Figure 7. (A) Rapid lipid evacuation from the interior of the transmembrane ClyA pore intermediates (6-mer and 9-mer; water and ions removed for
clarity) into the plane of the surrounding DMPC membrane occurs within 50 ns. (B) A timeline of lipid evacuation is shown for the 7-mer. (C) Fully
atomistic simulations of ClyA arcs inserted into POPC (1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine) membranes show rapid lipid evacuation
within 50 ns for the 6-mer to the 10-mer (water and ions removed for clarity). Corresponding lipid survival probabilities for all n-mer arcs in DMPC
and POPC membranes are shown in Figure 8.

Figure 8. A line connecting residue 189 on chain A and residue 11 on chain G of a 7-mer (colored green) defines the boundary between the interior
and the exterior of the pore. The continuous survival probability (CSP) of DMPC and POPC lipids within the 6-mer, 7-mer, 9-mer, and 10-mer
ClyA arcs captures the probability that a lipid resides in the interior of an arc for a specific length of time.

the water channel and the proteolipid arc complexes on longer the lipids within the arc. The definition of the arc boundary is
time scales are stable (discussed subsequently). Using a similar assumed to be a line joining residues 11 and 189 in the edge
CHARMM-GUI protocol for lipid packing in the arc interior protomers of the arc as shown in Figure 8. For the ith lipid, the
(without water) as implemented for the single protomer, 20 ns instantaneous probability Pi(t) is either 1 if the lipid is inside
simulations with the CHARMM36 force field for the 7-mer arc
the arc or 0 otherwise. The lipid position is tracked by the
revealed rapid lipid evacuation and the formation of a
transmembrane water channel, similar to the results presented coordinates of its phosphorus atom. The survival probability for
above, reinforcing the observation of lipid evacuation and water N lipids at time t is given as
channel formation (Figure S7).
N to + t
Kinetics of Lipid Evacuation from the Atomistic MD
Simulations. Because lipids are displaced from the hydrophilic S(t ) = ∑ ∏ Pi(tk)
arc interior within a period of 50 ns, we computed the i=1 tk = to (6)
continuous survival probability (CSP) of the lipids originally
present in the arcs (illustrated in Figure 7) to quantify the where the angular brackets represent shifted time averages. The
kinetics of lipid evacuation. CSP defines the residence time of normalized CSP, ⟨S(t)/S(0)⟩, is calculated over 50 ns
11503 DOI: 10.1021/acs.langmuir.7b02277
Langmuir 2017, 33, 11496−11510
Langmuir Article

Figure 9. (A) Front view of a 9-mer (red) with the stable transmembrane water channel (blue) in a membrane (transparent green slab) illustrating
that the stable ClyA arcs are functional membrane-permeabilizing entities. A cross-section across the bilayer of the 9-mer arc in DMPC reveals the
toroidal orientation of the lipids at the edge of the bilayer facing the arc interior (lipid choline headgroups colored red and phosphate groups colored
green). (B) An analysis of the lipid tilts in the arc vicinity: the distribution of the angle subtended by the lipid vector with the z axis is shown. Two
peaks corresponding to the upper and lower lipid leaflets can be seen for lipids in the bulk membrane (blue). However, in the lipid shell immediately
surrounding the 9-mer (green), θ assumes intermediate values which correspond to toroidal lipids lying nearly perpendicular to the z axis. (C)
Toroidal lipids participate in a larger number of hydrogen bonds in both 7-mer and 9-mer arcs in comparison with lipids away from the proteolipidic
edge. (D) Water radial distribution functions (g(r)) around the phosphorus atoms of the toroidal and free lipids for 7-mer and 9-mer arcs are shown.

(sampling frequency of 10 ps, 5000 snapshots over each 50 ns mechanism, assembly first occurs on the membrane surface to
trajectory) and fit to the following biexponential function: form a prepore which subsequently inserts into the membrane.
The fate of the lipids during this stage is largely unknown.
S (t ) ⎛ t⎞ ⎛ t⎞ Recent AFM images70−72 show the presence of inserted arcs at
= A exp⎜ − ⎟ + B exp⎜ − ⎟ + k different stages of oligomerization for the cholesterol-depend-
S(0) ⎝ τf ⎠ ⎝ τs ⎠ (7)
ent toxins, listeriolysin O and suilysin, which are known to
In the above equation, constant B = 1 − A − k from initial follow a prepore pathway. In this scenario, partially assembled
conditions. Two distinct time constants are observed in the pores (or prearcs) on the membrane surface undergo a
CSP fits: τf describes the fast component of lipid evacuation, conformational change resulting in membrane insertion and the
and τs describes the slower component. Because the survival formation of a transmembrane pore. Our simulations shed light
probabilities have decayed to almost zero for most of the cases on a possible lipid displacement pathway to stabilize the
as shown in Figure 8, τs describes the residence time of the transmembrane arcs. Recent work in our laboratory15 suggests
lipids inside the arc. We attribute the fast component (<1 ns) that ClyA leakage on small unilamellar vesicles occurs
to the rapid response of the lipids whose hydrophobic acyl tails predominantly through membrane-inserted arcs formed by
are abruptly exposed to the inner hydrophilic arc interface. The the stochastic insertion of prearcs. Here again, rapid lipid
longer time constant (7.9−16.9 ns) represents the gradual displacement dynamics would lead to the formation of a stable
reorganization and formation of the toroidal lipid edge arc. However, it must be noted that while the lipid
(discussed in the subsequent section). Thus, we obtain 7.9− reorientational dynamics has been observed in our simulations
16.9 ns from the CSP of lipids inside the arcs as quantitative to be typically complete within tens of nanoseconds, the actual
estimates of the lipid survival time scales depending on the arc displacement of lipids from the arc interior on complex cellular
size and the lipid type (DMPC/POPC). We also observe that membranes with an underlying cytoskeleton may be signifi-
POPC has a relatively shorter survival time when compared to cantly slower.
DMPC, indicating that unsaturated lipids may undergo faster Molecular Analysis of the Toroidal Lipid Edge. An
expulsion from the arcs than fully saturated lipids such as important property of the lipid bilayer membranes is the
DMPC, despite the presence of longer acyl chains in POPC. propensity to form an edge defect, characterized by the line
The biological significance of the calculated lipid evacuation tension of the edge (Λ). This tension causes the reorientation
time scale is that because spontaneous channel formation upon of lipids at the edge of membranes in ribbon configurations into
inserting ClyA arcs occurs within a few tens of nanoseconds on configurations similar to the reorientation of lipids observed at
pure lipid bilayer membranes, this process is not predicted to the mouth of the arcs.73 The formation of the toroidal lipid
be a rate-limiting step in pore formation. We point out that edge shields the hydrophobic lipid acyl tails from solvent
pore formation and the ensuing lipid dynamics are dependent exposure, locally offsets the energy penalty due to missing
on the pore-formation pathway. In the case of the prepore protomers, and stabilizes the partially formed arcs and may
11504 DOI: 10.1021/acs.langmuir.7b02277
Langmuir 2017, 33, 11496−11510
Langmuir Article

relieve curvature stress.74,75 Lipid edges configured to form [Fn]toroidal lipid = ΛL(n); n = 1, ..., nmax (8)
stable water channels lined by proteolipid walls have also been
observed in electron micrographs,11,76,77 AFM images of stable where L(n) is the length of the lipid edge defined by the
partially oligomerized arcs of cholesterol-dependent PFTs in oligomeric status of the arc and Λ is the line tension associated
membranes,78 X-ray diffraction studies of Bax-derived pep- with the edge. The chord length, L(n), can be derived using
tides,79 and MD simulations of protegrin arcs in lipid bilayers.80 simplified geometric arguments as follows. If there are nmax
Several studies have hypothesized the presence of proteolipidic protomers in a fully formed pore, then each protomer subtends
oligomerization intermediates during pore formation.1,11 an angle of 2π . Therefore, the angle subtended by the arc in
nmax
However, the experimental observation of lipid reorientation
to form the toroidal edge during pore formation and its
structural characterization is yet to be reported.75
the center of the pore is θ(n) = ( )2π when
n
nmax
n
nmax
≤ 0.5 and

MD simulations of the proteolipid ClyA arcs show the


toroidal lipid edge in molecular detail, as illustrated in Figure
(
θ(n) = 1 −
n
nmax )2π when n
nmax
> 0.5, as illustrated in Figure
10. Assuming that the curvature of the arcs is similar to that of
9A for the 9-mer arc in the DMPC membrane. The toroidal
lipids at the water−lipid interface of the arcs where protomers
are missing form an incomplete toroidal edge with the lipids
orientated perpendicular to the membrane normal at the mouth
of the arcs. The polar lipid headgroups (shown in red and
green) are shown facing the continuous water channel enclosed
by the proteolipid complex (cyan). An analysis of lipid tilts in
the arc vicinity show distributions of the angle (θ) subtended
by the lipid vector with the membrane normal (Figure 9B). The
lipid vector is defined as a vector passing from the center of
mass of the last carbon atoms in the lipid alkyl tails to the
phosphate bead, and the membrane normal is oriented toward
the lower leaflet. Two distinct well-formed peaks corresponding
to the lower leaflet (0° < θ < 90°) and the upper leaflet (90° <
θ < 180°) can be seen in the second lipid shell (blue), which
represents the free lipids in the bulk membrane. However, in
the lipids which form the toroid edge of the arc (green), θ
assumes intermediate values around 90°, which correspond to
toroidal lipids lying nearly perpendicular to the z axis. The lipid
shells are chosen such that shell 1 (green) consists of lipids
situated in an angular cross section at a distance less than R
away from the arc center (where R is the radius of the arc), and Figure 10. Top-view schematics of two arc complexes where
n n
shell 2 (blue) consists of lipids situated in an angular cross n
≤ 0.5 and n > 0.5 are illustrated (protomers illustrated in
max max
section at a distance more than 2R away from the arc center. blue). The length of the lipid edge, L(n), the angle subtended by the n-
Additionally, the distribution corresponding to the upper leaflet mer arcs, θ(n), and the radius of the pore, Rpore, are also depicted. The
shows greater broadening compared to lipids in the lower energetic penalty for the formation of the lipid edge given by eq 9 is
leaflet, indicative of an induced leaflet asymmetry due to the plotted for the 1-mer to 12-mer arcs for Λ = 10 and 46.1 pN. Λ
depends on the length of the lipid acyl chains and whether the lipid is
presence of the arc. saturated or unsaturated, and the above range encompasses various
The toroidally reoriented lipids facing the water channel in lipid bilayers as reported previously.73
the arc interior participate in a larger number of hydrogen
bonds in comparison to free lipids (membrane lipids away from
the influence of the arc) in both 7-mer and 9-mer arcs as shown the fully formed pore, we can extend the argument that the
in Figure 9C. To find the water structuring around these protomers in a partially oligomerized arc also subtend the same
toroidal lipids, an analysis of the water radial distribution angle in the center of the pore. Because the ClyA pore has a
functions (g(r)) around the phosphorus atoms of the toroidal tapering transmembrane region, we use an average value of the
and free lipids in the membrane was carried out (Figure 9D). upper and lower pore radii (Ru and Rl, respectively) to obtain
The water g(r) for the free lipids is nearly identical for both 7- the edge length L(n). Figure 9A depicts the upper and lower
mer and 9-mer arcs. The water g(r) for toroidal lipids present in leaflets in the context of the arcs. The edge penalty for the
the lipid edge shows minor enhancements mostly in the second transmembrane n-mer arc complex can be expressed as
solvation shell as compared to free lipids in the membrane but ⎛ θ(n) ⎞
equilibrate to bulk densities with increasing r. This may be due [Fn]toroidal lipid = Λ(Ru + Rl)sin⎜ ⎟ ; n = 1, ..., nmax
to suboptimal packing of lipids in the toroidal edge. The higher ⎝ 2 ⎠
coordination number of water around the toroidal lipids may (9)
explain the greater hydrogen bonding observed in both 7-mer The typical value of Λ for the lipid membrane edge formation
and 9-mer arcs, as shown in Figure 9C. lies in the range of 10−46.1 pN,73 and for the DMPC lipids, the
Energetics of the Toroidal Lipid Edge. Because toroidal value is 19.2 ± 2.8 pN. Using this range for Λ and values of Ru
lipids are created at an energetic cost, we use a simple model to = 7 nm and Rl = 4 nm,16,37 we obtain the variation in
estimate the penalty incurred as a result of the presence of the [Fn]toroidal lipid as illustrated in Figure 10. The energetics are
toroidal lipid edge. For each arc, the free energy of the edge is proportional to L(n), reaching a maximum at θ = 90 or n = 6.
simply given as For the dodecameric ClyA pore where n = nmax = 12, the
11505 DOI: 10.1021/acs.langmuir.7b02277
Langmuir 2017, 33, 11496−11510
Langmuir Article

Figure 11. (A) RMSD time traces and corresponding histograms for the 6-mer and 10-mer arcs in DMPC membranes, simulated for 224 and 289 ns,
respectively. (B) Superposition of the initial (cyan) and final (red) snapshots of the 6-mer and 10-mer arcs shows that the arc geometry is
undistorted as the simulation proceeds. (C) Stable Rg (radius of gyration) time traces and histograms (inset) of the 6-mer and 10-mer arcs further
corroborate the notion of arc stability. (D) Time traces and distributions of the axial tilts of helices αA1 and αB (as illustrated in Figure 3D) averaged
over all of the protomer subunits in the 6-mer and 10-mer arcs are shown. (E) The time trace of the center-of-mass distance between the edge
protomers (λCOM(t)) is shown; the inset illustratates λCOM for the 6-mer.

penalty is zero when pore formation is complete. The estimate the simulated unit cell is illustrated for both the 6-mer and the
for forming the toroidal edge which lies in the range of 10−150 10-mer in Figure S8, and the trends are observed to be stable.
kJ/mol must be overcome for the n-mer arcs to be stable in the Time traces of the Euclidean center-of-mass distance
membrane as our MD simulations suggest. Thus, stabilization between the edge protomers (λCOM(t), illustrated schematically
of the arc must be driven by protomer−protomer and for a 6-mer in the inset of Figure 11E) are analyzed for both the
protomer−membrane interactions which offset the cost of 6-mer and 10-mer arcs (Figure 11E). Stable λCOM(t) trends can
forming the edge. Evaluating these energetic contributions to be used to evaluate the tendency of the arcs to close and form
derive a complete free-energy landscape for the oligomerization n-mer pores where n < nmax. Furthermore, the edge-protomer
process is outside the scope of this article, and we are currently center-of-mass distances can be used to compute the spring
pursuing these computations. constant (k) between the edge protomers (which gives a rough
Structural Stability of the 6-mer and 10-mer measure of the arc stiffness) and the equilibrium harmonic
Proteolipidic ClyA Arcs from Longer Atomistic MD force (F) between the arc edges (eq 5 of Methods).
Simulations. The fully atomistic 50-ns-long 6-mer and 10-mer The mean values of k and F for the 6-mer are 39.68 ± 10.56
ClyA arc simulations in DMPC membranes were extended to pN/nm and 306.06 ± 81.42 pN, respectively, which are ∼4
224 and 289 ns, respectively, in order to verify the structural
times greater than that of the 10-mer (11.39 ± 3.78 pN/nm
stability of the arcs at longer time scales (Figure 11). The
and 76.61 ± 25.03 pN, respectively). Intuitively, smaller
RMSDs of both arcs (Figure 11A) show small deviations from
oligomers have a higher spring constant due to lower arc-
initial structures, as confirmed by the superposition of initial
and final structures (Figure 11B) and the mostly constant backbone fluctuations. However, as oligomerization proceeds
radius of gyration (Rg, Figure 11C). These observations imply and additional protomers are added to the arcs, the flexibility of
that the 6-mer and 10-mer arcs are stable and reveal no n-mer arc-backbones increases as a result of the increasing arc
propensity to disassociate over simulation time scales of length and attractive forces between the edge protomers, which
hundreds of nanoseconds. RMSD and Rg histograms appear also become progressively proximal as oligomerization
to be normally distributed as shown in Figure 11A,C. The proceeds. This renders flexibility to the arcs and increases
orientation of the arcs measured by averaging helix tilts of edge-protomer fluctuations (Figure 11E), thus leading to a net
helices αA1 and αB over all protomer subunits in the arcs is decrease in arc stiffness as evidenced by lower spring constants
shown to be stable over the simulation time scales in Figure and forces for the 10-mer. However, the 10-mer arc does not
11D and similar to the observations of the 12-mer (Figure 3D). close and form a full pore on the simulated time scales, and this
Fluctuations in the orientation of the 6-mer are marginally is consistent with the notion that the 12-mer is the
higher than that of the 10-mer. The time trace of the length of predominant pore stoichiometry observed experimentally.16,32
11506 DOI: 10.1021/acs.langmuir.7b02277
Langmuir 2017, 33, 11496−11510
Langmuir Article

■ DISCUSSION AND CONCLUSION


We have carried out extensive molecular dynamics simulations
supports this view. Identifying the presence of partially
oligomerized intermediates akin to the arclike structures
in an effort to shed light on the state of the oligomeric investigated in this study supports the growing pore pathway.
intermediates that could form during pore formation for the α- A nuance to this pathway is that noninserted arcs can
PFT, ClyA. Fully atomistic MD simulations show that the oligomerize on the membrane surface (prepore pathway) and
transmembrane ClyA pore intermediates or arcs are structurally can stochastically insert to form transmembrane arcs.15
stable with toroidal lipids forming the part of the water channel Finally, we point out that the process of unregulated pore
where proteins are missing. Initial conditions with lipids present formation on a membrane surface accompanied by lipid
within the arc interior illustrate the dynamics of lipid evacuation displacement from the pore lumens could drive membrane
and reorientation to form toroidal lipid edges bordering the buckling and tension generation. The number of ClyA arcs and
water channel. While the lipid reorientational dynamics is pores required to lyse erythrocytes31 has been shown to be
typically complete within tens of nanoseconds, it must be noted ∼105 at the onset of lysis, thus leading to a maximum estimated
that lipid evacuation time scales on actual cellular membranes coverage of 7% of the erythrocyte cell surface. Note that in our
with an underlying cytoskeleton may be significantly slower. simulations the protein to lipid ratio is significantly higher.
Other studies on melittin81 and actinoporins14,24 have Because the actual scenario is an extremely dilute regime, the
suggested that lipid participation is essential in stabilizing the manner in which membrane mechanical properties are
transmembrane pore. Toroidal lipid topologies observed in our modulated by ClyA pores is still an open question. The
MD simulations, lining the free edges of the ClyA arcs, are situation is further complicated by the multicomponent cell
similar to the reoriented lipids observed during electro- membrane and the presence of an actin cytoskeleton,
poration82 and X-ray reconstructions of lipid toroids in embedded sugars, and integral membrane proteins.
transmembrane pores of α-PFTs.79 Our study suggests that In summary, the combined observations provide strong in
ClyA proteolipidic arcs (Figure 7) could potentially form the silico evidence at molecular resolution for simultaneous lipid
oligomerization intermediates in the pore-forming pathway of evacuation from the central pore lumen as oligomerization
ClyA, similar to the growing-pore paradigm proposed for PFTs proceeds and the ability of all membrane-inserted n-mers
(illustrated in Figure 1). We provide molecular details about including the single transmembrane protomer to spontaneously
the single transmembrane protomer conformation, about which form stable water channels capable of leakage. We speculate
no experimental details are currently known. Recent structure- that modulating ClyA arc distributions by operating in the
based models have been used to study the conformational appropriate concentration regimes in suitably packed mem-
change from a water-soluble monomer to a membrane-inserted branes may prove useful for selective membrane permeabiliza-
protomer, 38 where membrane-bound intermediates are tion and size-based analyte separations using ClyA nanopores
sampled during the transition into the membrane. Here we and arcs.89
show that the membrane-inserted protomer is a relatively stable
state with a favorable free energy of binding to the membrane
(estimated using umbrella sampling). Surprisingly, the existence

*
ASSOCIATED CONTENT
S Supporting Information
of a stable water channel seen in both the atomistic and
The Supporting Information is available free of charge on the
MARTINI simulations of the single membrane-inserted
ACS Publications website at DOI: 10.1021/acs.lang-
protomer shows its membrane-permeabilizing ability (Figure
muir.7b02277.
4).
We briefly comment on the existence of arcs within the All-atom simulations, spring constants in MARTINI
framework of two pore-forming paradigms. In the prepore sampling, protein confirmational changes, and simula-
model, protein oligomerization occurs at the membrane tions set up using CHARMM-GUI protocols (PDF)


interface until a prepore complex is assembled. Subsequently
concerted membrane penetration of the prepore completes the
formation of the transmembrane pore complex (Figure 1). In AUTHOR INFORMATION
this regard, β-PFTs have been experimentally studied and pore Corresponding Author
formation is found to occur with the formation of a prepore in *E-mail: ayappa@chemeng.iisc.ernet.in. Phone: +91 (80) 2293
several cases.9,17,78,83,84 In the growing pore model, as 2769. Fax: +91 (80) 2360 8121.
oligomerization proceeds, lipid evacuation and the formation
ORCID
of transmembrane water channels are assumed to occur
simultaneously. This phenomenon has not been observed or Rajat Desikan: 0000-0002-0785-8187
characterized at molecular resolution and remains an open Prabal K. Maiti: 0000-0002-9956-1136
problem. In addition, lipid evacuation during the formation of K. Ganapathy Ayappa: 0000-0001-7599-794X
the transmembrane α-helical barrel in α-PFTs such as ClyA Notes
remains unanswered20 because the oligomerization pathways The authors declare no competing financial interest.


and kinetics are not fully determined. Multiple models have
been proposed where arcs have been implicated to play a
critical role in PFT-mediated lysis pathways,1,75 specifically in ACKNOWLEDGMENTS
the case of ClyA.15,30 Semicircularized arcs have been shown to The authors thank the Department of Science and Technology
cause ion leakage in various membranes.4,12 The release of (DST), Government of India, for a grant under which this work
calcium and potassium ions by these arcs can instigate several was carried out. The authors thank Sandhya S. Visweswariah, J.
downstream cellular processes that can enhance the lytic ability K. Basu, Narendra M. Dixit, Rahul Roy, and Pranesh
of PFTs even at sublytic concentrations,4,85−88 and our study Padmanabhan for extensive discussions.
11507 DOI: 10.1021/acs.langmuir.7b02277
Langmuir 2017, 33, 11496−11510
Langmuir Article

■ REFERENCES
(1) Gilbert, R. J. C. Pore-forming toxins. Cell. Mol. Life Sci. 2002, 59,
ClyA, SheA): X-ray crystal structure of the toxin and observation of
membrane pores by electron microscopy. Cell 2000, 100, 265−276.
(26) Reitz, S.; Essen, L.-O. α-Helical Cytolysins: Molecular Tunnel-
832−844. Boring Machines in Action. ChemBioChem 2009, 10, 2305−2307.
(2) Parker, M. W.; Feil, S. C. Pore-forming protein toxins: from (27) Roderer, D.; Benke, S.; Müller, M.; Fäh-Rechsteiner, H.; Ban,
structure to function. Prog. Biophys. Mol. Biol. 2005, 88, 91−142.
N.; Schuler, B.; Glockshuber, R. Characterization of Variants of the
(3) Bayley, H. Toxin structure: Part of a hole? Curr. Biol. 1997, 7,
Pore-Forming Toxin ClyA from Escherichia coli Controlled by a
R763−R767.
Redox Switch. Biochemistry 2014, 53, 6357−6369.
(4) Bischofberger, M.; Gonzalez, M. R.; van der Goot, F. G.
(28) Eifler, N.; Vetsch, M.; Gregorini, M.; Ringler, P.; Chami, M.;
Membrane injury by pore-forming proteins. Curr. Opin. Cell Biol.
Philippsen, A.; Fritz, A.; Müller, S. A.; Glockshuber, R.; Engel, A.;
2009, 21, 589−595.
Grauschopf, U. Cytotoxin ClyA from Escherichia coli assembles to a 13-
(5) Peraro, M. D.; van der Goot, F. G. Pore-forming toxins: ancient,
meric pore independent of its redox-state. EMBO J. 2006, 25, 2652−
but never really out of fashion. Nat. Rev. Microbiol. 2016, 14, 77−92.
(6) Los, F. C. O.; Randis, T. M.; Aroian, R. V.; Ratner, A. J. Role of 2661.
(29) Wai, S. N.; Westermark, M.; Oscarsson, J.; Jass, J.; Maier, E.;
Pore-Forming Toxins in Bacterial Infectious Diseases. Microbiology and
Benz, R.; Uhlin, B. E. Characterization of Dominantly Negative
Molecular Biology Reviews 2013, 77, 173−207.
(7) van Geest, M.; Lolkema, J. S. Membrane Topology and Insertion Mutant ClyA Cytotoxin Proteins in Escherichia coli. J. Bacteriol. 2003,
of Membrane Proteins: Search for Topogenic Signals. Microbiology and 185, 5491−5499.
Molecular Biology Reviews 2000, 64, 13−33. (30) Benke, S.; Roderer, D.; Wunderlich, B.; Nettels, D.;
(8) White, S. H. Membrane Protein Insertion: The Biology-Physics Glockshuber, R.; Schuler, B. The assembly dynamics of the cytolytic
Nexus. J. Gen. Physiol. 2007, 129, 363−369. pore toxin ClyA. Nat. Commun. 2015, 6, 6198.
(9) van Pee, K.; Neuhaus, A.; D’Imprima, E.; Mills, D. J.; Kuhlbrandt, (31) Vaidyanathan, M. S.; Sathyanarayana, P.; Maiti, P. K.;
W.; Yildiz, O. CryoEM structures of membrane pore and prepore Visweswariah, S. S.; Ayappa, K. G. Lysis dynamics and membrane
complex reveal cytolytic mechanism of PneumolysineLife2017, 6, oligomerization pathways for Cytolysin A (ClyA) pore-forming toxin.
10.7554/eLife.23644. RSC Adv. 2014, 4, 4930−4942.
(10) Tweten, R. K. Cholesterol-Dependent Cytolysins, a Family of (32) Roderer, D.; Glockshuber, R. Assembly mechanism of the α-
Versatile Pore-Forming Toxins. Infect. Immun. 2005, 73, 6199−6209. pore-forming toxin cytolysin A from Escherichia coli. Philos. Trans. R.
(11) Bhakdi, S.; Tranum-Jensen, J.; Sziegoleit, A. Mechanism of Soc., B 2017, 372, 20160211.
membrane damage by streptolysin-O. Infect. Immun. 1985, 47, 52−60. (33) Fahie, M.; Romano, F. B.; Chisholm, C.; Heuck, A. P.; Zbinden,
(12) Gilbert, R. J. C.; Mikelj, M.; Dalla Serra, M.; Froelich, C. J.; M.; Chen, M. A non-classical assembly pathway of Escherichia coli
Anderluh, G. Effects of MACPF/CDC proteins on lipid membranes. pore-forming toxin cytolysin A. J. Biol. Chem. 2013, 288, 31042−
Cell. Mol. Life Sci. 2013, 70, 2083−2098. 31051.
(13) Thompson, J. R.; Cronin, B.; Bayley, H.; Wallace, M. Rapid (34) Roderer, D.; Benke, S.; Schuler, B.; Glockshuber, R. Soluble
Assembly of a Multimeric Membrane Protein Pore. Biophys. J. 2011, Oligomers of the Pore-forming Toxin Cytolysin A from Escherichia
101, 2679−2683. coli Are Off-pathway Products of Pore Assembly. J. Biol. Chem. 2016,
(14) Kristan, K. C.; Viero, G.; Serra, M. D.; Macek, P.; Anderluh, G. 291, 5652−5663.
Molecular mechanism of pore formation by actinoporins. Toxicon (35) Sathyanarayana, P.; Desikan, R.; Ayappa, K. G.; Visweswariah, S.
2009, 54, 1125−1134. S. The Solvent-Exposed C-Terminus of the Cytolysin A Pore-Forming
(15) Agrawal, A. Pore Forming Toxins: Unravelling Oligomerization Toxin Directs Pore Formation and Channel Function in Membranes.
Pathways and Related Kinetics. Ph.D. Thesis, Indian Institute of Biochemistry 2016, 55, 5952−5961.
Science, 2016. (36) Mandal, T.; Kanchi, S.; Ayappa, K. G.; Maiti, P. K. pH
(16) Mueller, M.; Grauschopf, U.; Maier, T.; Glockshuber, R.; Ban, controlled gating of toxic protein pores by dendrimers. Nanoscale
N. The structure of a cytolytic α-helical toxin pore reveals its assembly 2016, 8, 13045−13058.
mechanism. Nature 2009, 459, 726−730. (37) Desikan, R.; Patra, S. M.; Sarthak, K.; Maiti, P. K.; Ayappa, K. G.
(17) Song, L.; Hobaugh, M. R.; Shustak, C.; Cheley, S.; Bayley, H.; Comparison of coarse-grained (MARTINI) and atomistic molecular
Gouaux, J. E. Structure of Staphylococcal α-Hemolysin, a Heptameric dynamics simulations of α and β toxin nanopores in lipid membranes.
Transmembrane Pore. Science 1996, 274, 1859−1865. J. Chem. Sci. 2017, 129, 1017.
(18) Yamashita, K.; Kawai, Y.; Tanaka, Y.; Hirano, N.; Kaneko, J.; (38) Giri Rao, V. V. H.; Desikan, R.; Ayappa, K. G.; Gosavi, S.
Tomita, N.; Ohta, M.; Kamio, Y.; Yao, M.; Tanaka, I. Crystal structure Capturing the Membrane-Triggered Conformational Transition of an
of the octameric pore of staphylococcal γ-hemolysin reveals the β- α-Helical Pore-Forming Toxin. J. Phys. Chem. B 2016, 120, 12064−
barrel pore formation mechanism by two components. Proc. Natl. 12078.
Acad. Sci. U. S. A. 2011, 108, 17314−17319. (39) Aksimentiev, A.; Schulten, K. Imaging α-Hemolysin with
(19) De, S.; Olson, R. Crystal structure of the Vibrio cholerae Molecular Dynamics: Ionic Conductance, Osmotic Permeability, and
cytolysin heptamer reveals common features among disparate pore- the Electrostatic Potential Map. Biophys. J. 2005, 88, 3745−3761.
forming toxins. Proc. Natl. Acad. Sci. U. S. A. 2011, 108, 7385−7390. (40) Wong-ekkabut, J.; Karttunen, M. Assessment of Common
(20) Bayley, H. Membrane-protein structure: Piercing insights. Simulation Protocols for Simulations of Nanopores, Membrane
Nature 2009, 459, 651−652. Proteins, and Channels. J. Chem. Theory Comput. 2012, 8, 2905−2911.
(21) Bayley, H.; Jayasinghe, L.; Wallace, M. Prepore for a (41) Mathe, J.; Aksimentiev, A.; Nelson, D. R.; Schulten, K.; Meller,
breakthrough. Nat. Struct. Mol. Biol. 2005, 12, 385−386. A. Orientation discrimination of single-stranded DNA inside the α-
(22) Tilley, S. J.; Saibil, H. R. The mechanism of pore formation by hemolysin membrane channel. Proc. Natl. Acad. Sci. U. S. A. 2005, 102,
bacterial toxins. Curr. Opin. Struct. Biol. 2006, 16, 230−236. 12377−12382.
(23) Phillips, R.; Ursell, T.; Wiggins, P.; Sens, P. Emerging roles for (42) Di Marino, D.; Bonome, E. L.; Tramontano, A.; Chinappi, M.
lipids in shaping membrane-protein function. Nature 2009, 459, 379− All-Atom Molecular Dynamics Simulation of Protein Translocation
385. ̂
through an I±-Hemolysin Nanopore. J. Phys. Chem. Lett. 2015, 6,
(24) Tanaka, K.; Caaveiro, J. M.; Morante, K.; Gonzalez-Manas, J. 2963−2968.
M.; Tsumoto, K. Structural basis for self-assembly of a cytolytic pore (43) Stoddart, D.; Ayub, M.; Hofler, L.; Raychaudhuri, P.;
lined by protein and lipid. Nat. Commun. 2015, 6, 6337. Klingelhoefer, J. W.; Maglia, G.; Heron, A.; Bayley, H. Functional
(25) Wallace, A. J.; Stillman, T. J.; Atkins, A.; Jamieson, S. J.; truncated membrane pores. Proc. Natl. Acad. Sci. U. S. A. 2014, 111,
Bullough, P. A.; Green, J.; Artymiuk, P. J. E. coli hemolysin E (HlyE, 2425−2430.

11508 DOI: 10.1021/acs.langmuir.7b02277


Langmuir 2017, 33, 11496−11510
Langmuir Article

(44) Yesylevskyy, S. O.; Schäfer, L. V.; Sengupta, D.; Marrink, S. J. lations Using the CHARMM36 Additive Force Field. J. Chem. Theory
Polarizable Water Model for the Coarse-Grained MARTINI Force Comput. 2016, 12, 405−413.
Field. PLoS Comput. Biol. 2010, 6, e1000810. (66) de Jong, D. H.; Singh, G.; Bennett, W. F. D.; Arnarez, C.;
(45) Pronk, S.; Páll, S.; Schulz, R.; Larsson, P.; Bjelkmar, P.; Wassenaar, T. A.; Schäfer, L. V.; Periole, X.; Tieleman, D. P.; Marrink,
Apostolov, R.; Shirts, M. R.; Smith, J. C.; Kasson, P. M.; van der Spoel, S. J. Improved Parameters for the Martini Coarse-Grained Protein
D.; Hess, B.; Lindahl, E. GROMACS 4.5: a high-throughput and Force Field. J. Chem. Theory Comput. 2013, 9, 687−697.
highly parallel open source molecular simulation toolkit. Bioinformatics (67) Marrink, S. J.; Tieleman, D. P. Perspective on the Martini
2013, 29, 845−854. model. Chem. Soc. Rev. 2013, 42, 6801−6822.
(46) Humphrey, W.; Dalke, A.; Schulten, K. VMD: Visual molecular (68) Rojko, N.; Kristan, K. C.; Viero, G.; Zerovnik, E.; Macek, P.;
dynamics. J. Mol. Graphics 1996, 14, 33−38. Dalla Serra, M.; Anderluh, G. Membrane Damage by an α-Helical
(47) Jorgensen, W. L.; Chandrasekhar, J.; Madura, J. D.; Impey, R. Pore-forming Protein, Equinatoxin II, Proceeds through a Succession
W.; Klein, M. L. Comparison of simple potential functions for of Ordered Steps. J. Biol. Chem. 2013, 288, 23704−23715.
simulating liquid water. J. Chem. Phys. 1983, 79, 926−935. (69) Luo, L.; Yang, J.; Liu, D. Integration and Oligomerization of Bax
(48) Lindorff-Larsen, K.; Piana, S.; Palmo, K.; Maragakis, P.; Klepeis, Protein in Lipid Bilayers Characterized by Single Molecule
J. L.; Dror, R. O.; Shaw, D. E. Improved side-chain torsion potentials Fluorescence Study. J. Biol. Chem. 2014, 289, 31708−31718.
for the Amber ff99SB protein force field. Proteins: Struct., Funct., Genet. (70) Yilmaz, N.; Kobayashi, T. Visualization of Lipid Membrane
2010, 78, 1950−1958. Reorganization Induced by a Pore-Forming Toxin Using High-Speed
(49) Nerenberg, P. S.; Head-Gordon, T. Optimizing Protein-Solvent Atomic Force Microscopy. ACS Nano 2015, 9, 7960−7967.
Force Fields to Reproduce Intrinsic Conformational Preferences of (71) Mulvihill, E.; van Pee, K.; Mari, S. A.; Muller, D. J.; Yildiz, O.
Model Peptides. J. Chem. Theory Comput. 2011, 7, 1220−1230. Directly Observing the Lipid-Dependent Self-Assembly and Pore-
(50) Beauchamp, K. A.; Lin, Y.-S.; Das, R.; Pande, V. S. Are Protein Forming Mechanism of the Cytolytic Toxin Listeriolysin O. Nano Lett.
Force Fields Getting Better? A Systematic Benchmark on 524 Diverse 2015, 15, 6965−6973.
NMR Measurements. J. Chem. Theory Comput. 2012, 8, 1409−1414. (72) Leung, C.; et al. Stepwise visualization of membrane pore
(51) Jämbeck, J. P. M.; Lyubartsev, A. P. Derivation and Systematic formation by suilysin, a bacterial cholesterol-dependent cytolysin. eLife
Validation of a Refined All-Atom Force Field for Phosphatidylcholine 2014, 3, e04247.
Lipids. J. Phys. Chem. B 2012, 116, 3164−3179. (73) West, A.; Ma, K.; Chung, J. L.; Kindt, J. T. Simulation studies of
(52) Darden, T.; York, D.; Pedersen, L. Particle mesh Ewald: An structure and edge tension of lipid bilayer edges: effects of tail
Nlog(N) method for Ewald sums in large systems. J. Chem. Phys. 1993, structure and force-field. J. Phys. Chem. A 2013, 117, 7114−7123.
98, 10089−10092. (74) Garcia-Saez, A. J.; Chiantia, S.; Salgado, J.; Schwille, P. Pore
(53) Hess, B.; Bekker, H.; Berendsen, H. J. C.; Fraaije, J. G. E. M. formation by a Bax-derived peptide: effect on the line tension of the
LINCS: A linear constraint solver for molecular simulations. J. Comput. membrane probed by AFM. Biophys. J. 2007, 93, 103−112.
Chem. 1997, 18, 1463−1472. (75) Gilbert, R. J. C.; Serra, M. D.; Froelich, C. J.; Wallace, M. I.;
(54) Nosé, S. A unified formulation of the constant temperature Anderluh, G. Membrane pore formation at protein-lipid interfaces.
molecular dynamics methods. J. Chem. Phys. 1984, 81, 511−519. Trends Biochem. Sci. 2014, 39, 510−516.
(55) Hoover, W. G. Canonical dynamics: Equilibrium phase-space (76) Palmer, M.; Harris, R.; Freytag, C.; Kehoe, M.; Tranum-Jensen,
distributions. Phys. Rev. A: At., Mol., Opt. Phys. 1985, 31, 1695−1697. J.; Bhakdi, S. Assembly mechanism of the oligomeric streptolysin O
(56) Bussi, G.; Donadio, D.; Parrinello, M. Canonical sampling pore: the early membrane lesion is lined by a free edge of the lipid
through velocity rescaling. J. Chem. Phys. 2007, 126, 014101. membrane and is extended gradually during oligomerization. EMBO
(57) Parrinello, M.; Rahman, A. Polymorphic transitions in single Journal 1998, 17, 1598−1605.
crystals: A new molecular dynamics method. J. Appl. Phys. 1981, 52, (77) Vadia, S.; Arnett, E.; Haghighat, A.-C.; Wilson-Kubalek, E. M.;
7182−7190. Tweten, R. K.; Seveau, S. The Pore-Forming Toxin Listeriolysin O
(58) Hub, J. S.; de Groot, B. L.; van der Spoel, D. g_wham-A Free Mediates a Novel Entry Pathway of L. monocytogenes into Human
Weighted Histogram Analysis Implementation Including Robust Error Hepatocytes. PLoS Pathog. 2011, 7, e1002356.
and Autocorrelation Estimates. J. Chem. Theory Comput. 2010, 6, (78) Czajkowsky, D. M.; Hotze, E. M.; Shao, Z.; Tweten, R. K.
3713−3720. Vertical collapse of a cytolysin prepore moves its transmembrane β-
(59) Miller, R. G. The jackknife - a review. Biometrika 1974, 61, 1− hairpins to the membrane. EMBO J. 2004, 23, 3206−3215.
15. (79) Qian, S.; Wang, W.; Yang, L.; Huang, H. W. Structure of
(60) Efron, B.; Stein, C. The Jackknife Estimate of Variance. Annals of transmembrane pore induced by Bax-derived peptide: Evidence for
Statistics 1981, 9, 586−596. lipidic pores. Proc. Natl. Acad. Sci. U. S. A. 2008, 105, 17379−17383.
(61) Best, R. B.; Zhu, X.; Shim, J.; Lopes, P. E.; Mittal, J.; Feig, M.; (80) Prieto, L.; He, Y.; Lazaridis, T. Protein Arcs May Form Stable
Mackerell, A. D. Optimization of the additive CHARMM all-atom Pores in Lipid Membranes. Biophys. J. 2014, 106, 154−161.
protein force field targeting improved sampling of the backbone ϕ, ψ (81) Lee, M.-T.; Sun, T.-L.; Hung, W.-C.; Huang, H. W. Process of
and side-chain χ1 and χ2 dihedral angles. J. Chem. Theory Comput. inducing pores in membranes by melittin. Proc. Natl. Acad. Sci. U. S. A.
2012, 8, 3257−3273. 2013, 110, 14243−14248.
(62) Klauda, J. B.; Venable, R. M.; Freites, J. A.; O’Connor, J. W.; (82) Ho, M.-C.; Casciola, M.; Levine, Z. A.; Vernier, P. T. Molecular
Tobias, D. J.; Mondragon-Ramirez, C.; Vorobyov, I.; MacKerell, A. D.; Dynamics Simulations of Ion Conductance in Field-Stabilized
Pastor, R. W. Update of the CHARMM all-atom additive force field for Nanoscale Lipid Electropores. J. Phys. Chem. B 2013, 117, 11633−
lipids: validation on six lipid types. J. Phys. Chem. B 2010, 114, 7830− 11640.
7843. (83) Cheley, S.; Malghani, M. S.; Song, L.; Hobaugh, M.; Gouaux, J.
(63) Jo, S.; Kim, T.; Iyer, V. G.; Im, W. CHARMM-GUI: a web- E.; Yang, J.; Bayley, H. Spontaneous oligomerization of a staph-
based graphical user interface for CHARMM. J. Comput. Chem. 2008, ylococcal alpha-hemolysin conformationally constrained by removal of
29, 1859−1865. residues that form the transmembrane beta-barrel. Protein Eng., Des.
(64) Wu, E. L.; Cheng, X.; Jo, S.; Rui, H.; Song, K. C.; Davila- Sel. 1997, 10, 1433−1443.
Contreras, E. M.; Qi, Y.; Lee, J.; Monje-Galvan, V.; Venable, R. M.; (84) Yamashita, D.; Sugawara, T.; Takeshita, M.; Kaneko, J.; Kamio,
Klauda, J. B.; Im, W. CHARMM-GUI Membrane Builder toward Y.; Tanaka, I.; Tanaka, Y.; Yao, M. Molecular basis of transmembrane
realistic biological membrane simulations. J. Comput. Chem. 2014, 35, beta-barrel formation of staphylococcal pore-forming toxins. Nat.
1997−2004. Commun. 2014, 5, 4897.
(65) Lee, J.; et al. CHARMM-GUI Input Generator for NAMD, (85) Metkar, S. S.; Marchioretto, M.; Antonini, V.; Lunelli, L.; Wang,
GROMACS, AMBER, OpenMM, and CHARMM/OpenMM Simu- B.; Gilbert, R. J. C.; Anderluh, G.; Roth, R.; Pooga, M.; Pardo, J.;

11509 DOI: 10.1021/acs.langmuir.7b02277


Langmuir 2017, 33, 11496−11510
Langmuir Article

Heuser, J. E.; Serra, M. D.; Froelich, C. J. Perforin oligomers form arcs


in cellular membranes: a locus for intracellular delivery of granzymes.
Cell Death Differ. 2015, 22, 74−85.
(86) Metkar, S. S.; Wang, B.; Catalan, E.; Anderluh, G.; Gilbert, R. J.
C.; Pardo, J.; Froelich, C. J. Perforin Rapidly Induces Plasma
Membrane Phospholipid Flip-Flop. PLoS One 2011, 6, e24286.
(87) Kennedy, C. L.; Smith, D. J.; Lyras, D.; Chakravorty, A.; Rood, J.
I. Programmed Cellular Necrosis Mediated by the Pore-Forming α-
Toxin from Clostridium septicum. PLoS Pathog. 2009, 5, e1000516.
(88) Aroian, R.; van der Goot, F. Pore-forming toxins and cellular
non-immune defenses (CNIDs). Curr. Opin. Microbiol. 2007, 10, 57−
61.
(89) Soskine, M.; Biesemans, A.; Moeyaert, B.; Cheley, S.; Bayley, H.;
Maglia, G. An Engineered ClyA Nanopore Detects Folded Target
Proteins by Selective External Association and Pore Entry. Nano Lett.
2012, 12, 4895−4900.

11510 DOI: 10.1021/acs.langmuir.7b02277


Langmuir 2017, 33, 11496−11510

You might also like