Many antimicrobial peptides (AMPs) selectively target and form pores in microbial membranes. However, the mechanisms of membrane targeting, pore formation and function remain elusive. Here we report an experimentally guided unbiased simulation methodology that yields the mechanism of spontaneous pore assembly for the AMP maculatin at atomic resolution. Rather than a single pore, maculatin forms an ensemble of structurally diverse temporarily functional low-oligomeric pores, which mimic integral membrane protein channels in structure. These pores continuously form and dissociate in the membrane. Membrane permeabilization is dominated by hexa-, hepta- and octamers, which conduct water, ions and small dyes. Pores form by consecutive addition of individual helices to a transmembrane helix or helix bundle, in contrast to current poration models. The diversity of the pore architectures—formed by a single sequence—may be a key feature in preventing bacterial resistance and could explain why sequence–function relationships in AMPs remain elusive.
Since their discovery over 100 years ago, thousands of pore-forming antimicrobial peptides (AMPs) have been identified, revealing great variations in size, secondary structure and sequence composition1. Despite this wealth of data no common pore-forming motif has been discovered, and the molecular basis of antimicrobial activity remains poorly understood2,3. One of the key difficulties has been the transience of pores formed by AMPs, which has proved challenging for experimental structure determination. In the absence of experimental structures a range of theoretical models have been proposed that try to rationalize how soluble AMPs, which carry multiple charged residues, can insert into hydrophobic membranes to form pores and what these pores might look like2.
Here we report the molecular mechanisms of pore formation, as well as the structures, conductances and selectivities of pores formed by maculatin (GLFGVLAKVAAHVVPAIAEHF-NH2), a 21-residue peptide isolated from the skin of the green-eyed tree frog, Litoria genimaculata. Maculatin is a typical member of the large family of pore-forming amphiphilic AMPs, with broad-spectrum antibacterial activity at low micromolar concentrations. Previous studies have shown that maculatin forms long-lived pores, which cause all-or-none leakage comparable to other AMPs4,5,6,7.
Experimentally guided system set-up
On contact with phosphatidylcholine liposomes, maculatin peptides, which are unstructured in solution, fold into continuous α-helices that form pores in the liposome membrane. First, we used circular dichrosim (CD) spectroscopy to establish experimentally that the P15A mutation, which is active against Gram-positive bacteria4,8,9, stabilizes membrane-bound maculatin against thermal denaturation without affecting pore size or pore-forming capactiy for a range of different lipids and peptide-to-lipid ratios (Fig. 1). This thermally stabilized mutant allowed us to elevate the temperatures of unbiased long-timescale folding-partitioning molecular dynamics simulations to 90–150 °C, increasing sampling by 2–3 orders of magnitude, without significantly affecting the thermodynamics of the system10,11,12,13,14. Next, we experimentally determined the leakage of fluorescent dyes of increasing molecular weight from liposomes treated with maculatin (Fig. 1). This revealed that only the smallest dyes encapsulated (8-aminonaphthalene-1,3,6-trisulphonic acid/p-Xylene-bis-N-pyridinium bromide; ANTS/DPX), with molecular weights of ∼400 Da, are able to leak out of the liposome through pores formed by maculatin peptides, allowing us to estimate the approximate number of peptides forming a pore, by assuming simple channel geometries. Using this information, we built atomic detail systems of up to 16 maculatin peptides placed in phosphatidylcholine lipid bilayers of different acyl chain lengths (Supplementary Table 1).
Maculatin membrane insertion and oligomerization
All maculatin peptides initially reside in their stable surface-adsorbed state15, consistent with the −4.6±0.8 kcal mol−1 partitioning free energy determined for wild-type maculatin via CD titration (Fig. 2). Figure 3 shows that individual peptides subsequently insert transmembrane (TM) via a range of different mechanisms. This finding was unexpected, since TM insertion of AMPs like maculatin requires temporary burial and membrane translocation of charged side chains, which have theoretical membrane translocation barriers of ∼15–20 kcal mol−1 (ref. 16). In the dominant insertion mechanism this barrier is overcome by cooperative insertion, involving two peptides in a head-to-tail arrangement in combination with a water defect. TM-inserted maculatin can rapidly catalyse additional TM insertions through membrane defects induced by its charged and polar side chains. In contrast to some proposed models, AMP insertion via large surface aggregates was not observed17,18. At equilibrium the peptides, which were initially placed on one interface in simulations DM1–DM4, are symmetrically distributed along the bilayer normal, continually changing between marginally stable TM oligomeric assemblies and surface-bound states on both interfaces (Fig. 3). Subsequent pore-forming simulations were therefore started with an equal number of peptide on each interface.
Spontaneous formation of structurally diverse pores
Pore formation indicates that a significant number of peptides are TM-inserted. Overall, the surface-bound states slightly dominate the equilibrium with 53±7% occupation (averaged over all simulations ΔGS→TM=0.1 kcal mol−1), in agreement with the 50±15% interfacial population determined from oriented CD measurements in aligned palmitoyloleoyl-phosphocholine (POPC) bilayer stacks (Fig. 2). Figure 4 shows that 10–20 μs timescales capture numerous spontaneous pore-formation events. Remarkably, rather than a single, well-defined and stable pore structure, as would be expected for a channel protein, maculatin forms an ensemble of conformationally diverse channel-like pores, which continually assemble and disband in the membrane (Fig. 4). The formation of an equilibrium ensemble of structurally diverse channels by a single AMP has not been previously reported in the literature, but is consistent with experimental leakage data.
To assess the architectural diversity of the channel-like pores formed, we developed software to identify and cluster structurally similar oligomeric TM assemblies (see Methods). This revealed hundreds of different channel architectures for maculatin, of which usually <10 are significantly populated. Only small pores with 3–8 peptides were observed. Larger aggregates were found only rarely and consisted of smaller channels in lateral contact. The most populated cluster for each oligomer are shown in Figs 4 and 5. Interestingly, very similar oligomers were found for all simulations, with a strong preference for antiparallel peptide arrangements (90%). The recurring structural motif is an antiparallel dimer offset by 8–10 residues, with tight inter-helical packing of the C-terminal moieties of the peptides (Supplementary Fig. 1). Larger oligomers are symmetric combinations of this basic dimer motif.
Remarkably, single (P15A; E19Q) and double mutants (P15A, E19Q), which show small or little difference in experimental leakage and pore size (Fig. 1), revealed different ensemble weightings (Fig. 4), as did changes in membrane acyl chain length (Fig. 5). This shows that while maculatin forms proper channels in the membrane, rather than disordered pores or detergent-like holes, the channel equilibrium is influenced by the environment or minor mutations. Indeed, the heterogeneous nature of the channel-like structures precludes their classification into the generally assumed barrel stave or toridal models suggested for AMP pore structures2,19,20.
Unbiased dye-conductance simulations revealed that the octameric pores efficiently conduct ANTS/DPX fluorophores, which were the largest dyes observed to leak from vesicles treated by maculatin (Fig. 1). This confirms that these are indeed the largest pores that maculatin can form. The diversity of pores observed, and their variation on minor perturbation of the peptide sequence or membrane lipids explains why no clear pore-forming motif has been found in AMP sequences and why bacterial resistance against AMPs is remarkably low despite continuous exposure over millions of years.
Pore assembly processes
Pores form and dissipate many times over the course of the simulations, providing a converged description of the assembly mechanisms (Fig. 6). Current models of AMP activity propose detergent-like mechanisms of poration, where interfacial peptide aggregates form pores by concerted insertion. Contrary to these models, we found that the assembly of maculatin into pores is driven by the one-by-one, C-terminal-first addition of surface-bound peptides to an existing TM helix or oligomer. In this way, the translocation of the charged and polar side chains of the inserting peptide are catalysed by the polar face of the already TM-inserted peptide. Direct oligomerization of TM helices is observed only infrequently, and never for large oligomers (≥5). The pore-forming process appears similar to the spontaneous assembly of integral membrane proteins from fragments21,22,23, which is thought to mimic the assembly of multi-span membrane proteins after insertion of the correctly threaded TM helices by the translocon24.
Pore properties and relation to AMP activity
Each pore formed in the assembly simulations is only marginally stable and has its own characteristic architecture, oligomeric state, conductance properties, formation probability and lifetime in the bilayer. Since the typical lifetime of functional pores at this temperature (<1 μs) is much shorter than the average assembly time (multi-μs), only a small number of conduction events are typically captured before the pore disassembles again, too few to accurately determine ion selectivity and conductance properties. To quantify these properties we extracted the most populated pores with long lifetimes from the assembly simulations, equilibrated them in POPC bilayers and performed microsecond timescale spontaneous conductance simulations in the presence of applied voltages. Figure 7 shows the equilibrium ion and water conductance for each type of pore obtained from these simulations (Supplementary Table 2). This shows that conduction of water and ions is dominated by large oligomers with more than five peptides, but lower oligomers can occasionally also leak ions (Supplementary Fig. 2).
To quantify the pore lifetimes at room temperature, we determined the average unfolding times of each pore at highly elevated temperatures (167–207 °C). Unfolding times were found to follow Arrhenius kinetics (Fig. 8). Extrapolation to body temperature (37 °C) gives average pore lifetimes of 24 ms for the octamer, 22 ms for the heptamer and 58 ms for the hexamer.
In summary, experimentally guided simulations revealed that maculatin peptides form ensembles of structurally diverse channel-like pores in the membrane. The pores have a variety of different assembly mechanisms, conductances and lifetimes in the bilayer. Minor mutations or changes in lipid tail length result in different structures and ensembles, which explains AMP resiliance against bacterial resistance. Thus, AMPs prevail by fine-tuned interactions that balance pore stability, structural versatility and conductance efficiency.
In brief, solid-phase synthesized and purified maculatin peptides and mutants were validated using high-pressure liquid chromatography (HPLC) mass spectroscopy. Peptide secondary structure and thermostability was determined using synchrotron radiation CD (SRCD) spectroscopy of peptides after addition of large unilamellar vesicles (LUVs) at various peptide to lipid ratios and temperatures. Peptide-binding free energies were determined via vesicle titration, and the average membrane orientation of maculatin and mutant peptides was determined using SRCD spectroscopy of peptides in embedded into oriented lipid bilayer stacks. Peptide-induced pore formation in lipid bilayers was determined by measuring the leakage of dyes from LUVs using fluorescence microscopy. Atomic detail unbiased molecular dynamics simulations were performed using GROMACS (www.gromacs.org) and analysed using HIPPO (www.biowerkzeug.com), using the CHARMM force field.
Maculatin 1.1 and two mutants (P15A and P15A+E19Q) were synthesized and purified by GenScript using Fmoc chemistry, with 98% purity. Peptide purity and identity were confirmed by reverse phase HPLC and electrospray ionization mass spectrometry with the atomic mass of maculatin 1.1 wild-type, P15A and P15A+E19Q are 2145.55, 2119.51 and 2118.53 g mol−1, respectively. Peptides were dissolved in distilled water, and the concentration was determined using phenylalanine absorption at 257 nm, assuming an average molar extinction coefficient of ɛ=200 M−1 cm−1. POPC and other lipids were purchased from Avanti Polar Lipids (Alabaster, AL, USA). All other solvents and reagents were all purchased from Sigma-Aldrich.
A unit of 25 mg lipids (1 ml of 25 mg ml−1 lipids in chloroform) were dried under nitrogen gas in a glass vial, and the remaining chloroform was removed under vacuum overnight. Lipids were re-suspended in 10 mM sodium phosphate buffer (pH=7.0) with 100 mM potassium chloride. The lipids were then extruded 10 times through a 0.1 μm Nucleopore polycarbonate filter to give LUVs of 100 nm diameter. Vesicle size was validated by dynamic light scattering using a Zetasizer Nano ZS.
ANTS/DPX leakage assay
A unit of 5 mM ANTS (molecular weight=427.33 g mol−1; Life Technologies) and 12.5 mM DPX (molecular weight=422.16 g mol−1; Life Technologies) were encapsulated in 100 nm POPC LUVs. LUV size and homogeneity was verified by dynamic light scattering using a Zetasizer Nano ZS. Gel filtration chromatography of Sephadex G-100 (GE Healthcare Life Sciences Inc.) was used to remove external free ANTS/DPX from vesicle with entrapped contents. The ANTS/DPX-containing LUVs were diluted to 0.5 mM total lipid concentration. The peptide was added at specific peptide-to-lipid ratios. Leakage was quantified by the increase in ANTS fluorescence that occurs when the DPX quencher inside the liposome is diluted by leakage into the external buffer on peptide-induced permeabilization of the lipid bilayer. Leakage was measured after 3 h reaction time and measurements were repeated three times. Addition of Triton X-100 (0.05% vol.) to the LUVs, which fully permeabilizes the membrane, served as a positive control.
Macromolecule release assay
Several different-size dextrans were prepared and labelled with both TAMRA and biotin. Conjugated dextran was entrapped in POPC LUVs. External dextran was removed by incubation with immobilized streptavidin. Streptavidin labelled with an Alexa-488 fluorophore was added during the leakage experiment, with the peptide as previous described25. The sample was incubated for 3 h, before measuring the Alexa-488 fluorescence. A control without added peptide served as the 0% leakage signal, and addition of 0.05% vol. Triton X-100 was used to determine 100% leakage25.
Peptide binding to POPC and 1-palmitoyl-2-oleoylglycero-3-phosphoglycerol (POPG) bilayers was determined by vesicle titration. Peptide solutions (20 μM) in 10 mM phosphate buffer (pH=7.0) were titrated with 3 μl aliquots of 100 mM POPC/POPG LUVs in identical buffer. CD spectra were recorded using SRCD spectroscopy, using the CD beam-line on ASTRID2 (Aarhus University, Denmark). Spectra were recorded from 270 to 170 nm with a step size of 0.5 nm, bandwidth of 0.5 nm and a dwell time of 2 s. Each spectrum was averaged 3 over repeat scans. The averaged spectra were normalized to molar ellipticity per residue. The raw data were analysed using DichroWeb (http://dichroweb.cryst.bbk.ac.uk/)26,27.
Oriented CD spectroscopy
Peptides were dissolved in chloroform and added to POPC or POPG in chloroform at specific peptide to lipid ratios. Peptide/lipid mixtures were dried under a low flow of nitrogen gas and placed under high vacuum overnight to remove all organic solvent. Dried samples were dissolved in 20 μl of pure 2,2,2-trifluoroethanol, which was spread on a quartz glass slide to form oriented bilayer stacks. After drying the peptide–lipid film under high vacuum to remove all TFE, 2 μl distilled H2O was added to the glass plate, and the plate was placed in a chamber containing a saturated K2SO4 solution (120 g l−1 at 25 °C) and equilibrated overnight at 42 °C. CD spectra were recorded using the CD beam-line on ASTRID2 (Aarhus University). Spectra were recorded from 270 to 160 nm with a step size of 0.5 nm, bandwidth of 0.5 nm and a dwell time of 2 s, and averaged over 8 rotational angles, which rotated the sample around the beam axis by 360°. Each spectrum was averaged over 3 repeat scans. The averaged spectra were normalized to molar ellipticity per residue and the results were analysed following the protocols of Huang and Ulrich28,29.
Peptide thermostability assay
SRCD spectroscopy were carried out using the CD beam-line on ASTRID2 (ISA, Aarhus University). Peptides (20 μM) were embedded in POPC LUV using co-extrusion as described above and solvated in (peptide-to-lipid ratio=1:50) in 10 mM phosphate buffer (pH=7.0) were placed in a 1 mm path length quartz cuvette (Hellma Analytics). Spectra were recorded using 0.5 nm bandwidth, 0.5 nm resolution and 2 s dwell time. Temperature was controlled with a Peltier device, and varied between 20 and 90 °C in 10° steps. The actual temperature in the cell was determined using a thermometer and showed a linear correlation with TActual=0.766 TSetting+4.715 (R2=0.9999). The sample was equilibrated for 10 min at each temperature before collecting data, and measurements were repeated five times. The averaged spectra were smoothed using a Fourier filter and normalized to molar ellipticity per residue.
Binding free-energy assay
The free energy of folding-partitioning of peptides onto lipid bilayers was determined via vesicle titration. A unit of 200 μM peptide in 10 mM phosphate buffer (pH=7.0) was titrated by adding 200, 400, 4,000 and 10,000 μM POPC LUVs in phosphate buffer at 25 °C. The POPC LUVs had a diameter of 100 nm (determined using dynamic light scattering) and were made via extrusion. Folding-partitioning was determined from the CD signal change at 222 nm on LUV titration. This allows calculation of the binding free energy, as described by Ladokhin et al. using the formula ΔG=−RT ln KX, where KX is the partition coefficient. KX is calculated from the mean residue ellipticity at 222 nm, [θ]λ=222nm by assuming that the signal is composed of two states: folded peptide (F) and unfolded peptide (U), which gives KX=[F]/[U] (refs 30, 31).
Molecular dynamics simulations
All simulations were performed and analysed using GROMACS version 5.0.5 (www.gromacs.org)32 and HIPPO beta (www.biowerkzeug.com), using the CHARMM27 force field33, and TIP3P water34. CHARMM36 all-atom lipids parameters were used35. Electrostatic interactions were computed using particle-mesh-Ewald, and a cutoff of 10 Å was used for the van der Waals interactions. Bonds involving hydrogen atoms were restrained using LINCS36. Simulations were run with a 2 fs time-step, and neighbour lists were updated every five steps. All simulations were performed in the NPT ensemble (constant number of particles, constant pressure and temperature), with water, lipids and the protein coupled separately to a heat bath with T=80–120 °C and a time constant τT=0.1 ps using the velocity-rescaling method. Atmospheric pressure of 1 bar was maintained using the Parrinello−Rahman semi-isotropic pressure coupling method with compressibility κz=κxy=4.6 × 10–5 per bar and time constant τP=1 ps.
The high-temperature simulation protocol is robust: In particular, we have demonstrated that elevated temperature simulations of peptide folding-partitioning into lipid bilayers accurately reproduce the native-state structural equilibria and partitioning free energy, compared with both CD and translocon experiments, as well as low temperature simulations13,37,38,39,40,41. This approach exploits the fact that the cost of breaking a helical hydrogen bond in the bilayer is very high (∼5 kcal mol−1, which is much higher than kT), ensuring that membrane-inserted peptides stay fully helical even at highly elevated temperatures13,37,38.
The initial peptide set-up was the surface bound conformation. This stable interfacial α-helical minimum can be rapidly predicted39. Initial simulations were started with all peptides in the same membrane leaflet. These simulations revealed the long timescales needed for assembly of oligomers that require the equilibration of peptide mass in both leaflets. To speed up the pore-formation process, all subsequent simulations were started with both bilayer leaflets being initially populated. Typical bilayers consisted of 200 14:0 phosphatidylcholine:1,2-dimyristoyl-sn-glycero-3-phosphocholine, 16:0 phosphatidylcholine:1,2-dipalmitoyl-sn-glycero-3-phosphocholine or 18:0 phosphatidylcholine:1,2-distearoyl-sn-glycero-3-phosphocholine lipids, and ∼30 water molecules per lipids39. All simulations were performed on the PI supercomputer at Shanghai Jiao-Tong University and the MARCC supercomputing facility at Johns Hopkins University. Typical simulation length was between 20–50 μs for each trajectory, sufficient to converge the oligomer populations. For all quantities, an estimate of the s.e.m. was calculated by automatically block averaging over 10 blocks, and these are plotted as error bars.
Conduction simulations and pore lifetime simulations
Pore structures were taken from the assembly simulations and equilibrated for 100 ns in POPC bilayers and 500 mM Na/K/Cl solution, while holding the pore structure in place using positional restraints of the backbone carbon atoms. After equilibration of the membrane around the pore restraints were removed and spontaneous conduction was measured by counting the transport events of ions and water through the pore in the presence of applied voltages (Supplementary Table 2). Functional pore lifetimes were determined by repeating the simulations at increasing temperatures and recording the dissolution times of the pore (Supplementary Fig. 7).
The equilibration of the peptide mass from one leaflet across the membrane was calculated by performing a histogram of the peptide mass as a function of the membrane normal and simulation time. The histogram was smoothed over 10 consecutive trajectory frames.
To reveal the most populated pore assemblies during the simulations, a complete list of all oligomers was constructed for each trajectory frame. An oligomer of order n was considered any set of n peptides that are in mutual contact, defined as a heavy-atom (N, C and O) minimum distance of <3.5 Å. Population plots of the occupation percentage of oligomer n multiplied by its number of peptides n were then constructed. These reveal how much peptide mass was concentrated in which oligomeric state during the simulation time.
Permutational cluster analysis
The enormous amount of heterogeneous structures formed by maculatin 1.1 over the course of the simulations posed a significant challenge for automated analysis. Conventional cluster analysis could not be used, as the same specific oligomeric pore could consist of entirely different peptides in each instance. To address this, we have developed a new permutational cluster analysis tool as part of the HIPPO analysis programme. After the oligomer list was computed, in a second step, all oligomers of the same order n were conformationally clustered. A clustering algorithm was used, with a backbone root mean squared deviation (r.m.s.d.) similarity cutoff criterion of 4 Å. Since each oligomer could be made up of different peptides—or of the same peptides but in a different order—the clustering had to compare one oligomer with all n! permutations of peptide arrangements of the other oligomer. Permutations were generated using Heap’s algorithm. For large n, these computations could last several hours. The final r.m.s.d. value of the conformational similarity was considered the lowest r.m.s.d. value as obtained from the n! permutational comparisons.
The data that support the findings of this study are available from the corresponding author on reasonable request.
How to cite this article: Wang, Y. et al. Spontaneous formation of structurally diverse membrane channel architectures from a single antimicrobial peptide. Nat. Commun. 7, 13535 doi: 10.1038/ncomms13535 (2016).
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
We thank the Aarhus synchrotron CD beamline staff for support and beamtime (Grant #ISA-16-1038, to M.B.U.), and Johns Hopkins University for funding. This research was supported by a China 1000 Plan’s Program for Young Talents (#13Z127060001) to J.P.U. and by National Natural Science Foundation of China (NSFC) grants #91230202 and #11471213 to D.H. Simulation resources were supported by Center for High Performance Computing, Shanghai Jiao Tong University and the MARCC supercomputer facility at Johns Hopkins University.