Structure and Interactions of A Host Defense Antimicrobial Peptide Thanatin in Lipopolysaccharide Micelles Reveal Mechanism of Bacterial Cell Agglutination

Host defense cationic Antimicrobial Peptides (AMPs) can kill microorganisms including bacteria, viruses and fungi using various modes of action. The negatively charged bacterial membranes serve as a key target for many AMPs. Bacterial cell death by membrane permeabilization has been well perceived. A number of cationic AMPs kill bacteria by cell agglutination which is a distinctly different mode of action compared to membrane pore formation. However, mechanism of cell agglutinating AMPs is poorly understood. The outer membrane lipopolysaccharide (LPS) or the cell-wall peptidoglycans are targeted by AMPs as a key step in agglutination process. Here, we report the first atomic-resolution structure of thanatin, a cell agglutinating AMP, in complex with LPS micelle by solution NMR. The structure of thanatin in complex with LPS, revealed four stranded antiparallel β-sheet in a ‘head-tail’ dimeric topology. By contrast, thanatin in free solution assumed an antiparallel β-hairpin conformation. Dimeric structure of thanatin displayed higher hydrophobicity and cationicity with sites of LPS interactions. MD simulations and biophysical interactions analyses provided mode of LPS recognition and perturbation of LPS micelle structures. Mechanistic insights of bacterial cell agglutination obtained in this study can be utilized to develop antibiotics of alternative mode of action.

LPS-thanatin interactions were observed to be exothermic in nature as suggested by downward position of titration peaks and integrated heat values (Fig. 1d). ITC derived LPS-thanatin binding parameters are provided in Table 1. As seen, thanatin binds to LPS with K d ~ 1.55 μM and binding is essentially favoured by negative enthalpy (ΔH) change (Table 1). Note, a somewhat higher K d value of thanatin-LPS interactions was estimated from the dansyl fluorescence changes compared to ITC. The difference may arise from the type of interactions determined by the two different methods. Dansyl fluorescence probes only localized changes whereas ITC measures more global interactions. Next, we examined ability of thanatin to displace BODIPY-cadaverine (BC), a fluorescence probe from LPS. BC binds to the lipid A domain of LPS through ionic interactions between protonable amine groups of the probe and anionic phosphates of lipid A 19,20 . BC probe displacement assays demonstrated that thanatin binds efficiently to LPS, with an estimated ED 50 of 0.87 ± 0.15 μM (Fig. 1e). BC displacement was also carried out for polymyxin B, an archetype LPS binding peptide that displayed a somewhat higher ED 50 of 3.7 ± 0.30 μM (Fig. 1e). Surface charge neutralization of E. coli was examined by measuring zeta potential as a function of concentrations of thanatin (Fig. 1f). The anionic LPS-outer membrane predominantly contributes to a negative zeta potential of E. coli cells [21][22][23] . As expected, addition of thanatin to E. coli cell solutions, at various doses, caused a dramatic increase in zeta potential toward a neutral surface potential (Fig. 1f). Even at 5 μM of thanatin, zeta potential changed from −50 mV to −5 mV and the surface charge effectively became neutral (zeta potential ~0 mV) at a peptide concentration of 15 μM (Fig. 1f). These data suggests that thanatin binds to E. coli cells utilizing ionic interactions among cationic residues and anionic membranes. The aforementioned data, BC probe displacement ITC and zeta potential, thus demonstrated that thanatin-LPS interactions could be largely dominated by ionic interactions. Although, dans-thanatin fluorescence data suggested that thanatin also inserts into the hydrophobic milieu of LPS.

Structure of thanatin in free solution by NMR. 3-D structure of thanatin in free solution has been
previously reported in an aqueous solution at a low temperature of 5 °C by using NMR spectoscopy 24 . In order to realize thanatin structure and interactions in LPS, we have analysed NMR spectra, two-dimensional 1 H-1 H TOCSY and NOESY, and determined solution structure of thanatin in aqueous solution at 25 °C. An ensemble of structures, in free solution, of thanatin was determined based on 140 NOE driven distance constraints and dihedral angle constraints (Fig. 2a). As can be seen, thanatin assumes an extended conformation for N-terminal residues G1-P7 followed by a beta hairpin like structure containing two stranded antiparallel β-strands, residues I8-N12 and residues K17-M21, connected by a loop comprising residues R13-G16 (Fig. 2b). The β-hairpin structure is stabilized by a single disulfide bond involving residues C11 and C18 (Fig. 2b). The 3-D structure of thanatin determined here at 25 °C is akin to the structure previously determined at 5 °C 24 .
NMR analyses of thanatin in LPS micelle. Transferred nuclear Overhauser effect spectroscopy (tr-NOESY) 25,26 can be utilized to determine atomic-resolution structures of AMPs and LPS neutralizing peptides in LPS micelle [27][28][29][30][31][32][33] . Owing to the low critical micelle concentration (<1 μM), LPS readily forms high molecular weight micelles in solution 34,35 . AMPs undergoing chemical exchange at the time scale of fast or intermediate exchange between free and LPS bound states may yield tr-NOEs permitting determination of bound conformations [27][28][29][30][31][32][33] . Additions of LPS, from a stock solution, into a solution containing thanatin, showed broadening of resonances in one-dimensional NMR spectra ( Supplementary Fig. 1). Such an NMR line broadening effect of a ligand upon addition of large macromolecules can be interpreted as either fast or intermediate chemical exchange 25,26 . Fast or intermediate time scale of chemical exchange between thanatin and LPS micelles has been further supported by the observation of tr-NOEs. Identification of amino acid spin systems or sequence specific resonance assignment of thanatin in LPS micelles were carried out by combined analyses of TOCSY and NOESY spectra ( Supplementary Figs 2 and 3). Chemical shift of amino acids of thanatin has been provided in Supplementary  Table S1. Two-dimensional tr-NOESY spectra of thanatin, in presence of LPS micelle, yielded more number of NOE connectivities in comparison to thanatin in free solution involving sidechain/sidechain and backbone/ sidechain resonances ( Supplementary Fig. 4). Moreover, there were several new long range NOEs detected for residues, V6, I8, Y10, R14, K17 of thanatin in LPS micelles ( Supplementary Fig. 5). Notably, residue Y10 demonstrated as many as ~40 NOEs in complex with LPS micelle (Supplementary Fig. 5). Interestingly, a number of unique long-range NOEs were detected among residues e.g. V6/C11, I8/N12, I8/R13, I8/R14 for thanatin in complex with LPS (Supplementary Table S2, Fig. 2c). These NOEs demonstrated a close proximity among residues at the extended N-terminus and residues at the β-turn of the β-sheet structure of thanatin. Figure 3a shows superposition of backbone atoms (Cα, N and C') of twenty low energy structures of dimeric thanatin in complex with LPS. NOE driven distance constrains for residues V6/C11, I8/N12, I8/R13, I8/R14 (Supplementary Table S2) were found to be incompatible to a monomeric β-sheet structure and were utilized as inter-monomeric or dimeric NOEs for structure determination. The ensemble of dimeric structures of thanatin was determined utilizing 386 NOE driven distance constrains including 78 long range and 10 inter subunit NOEs (Supplementary Table S3 stranded β-sheet topology, whereby the residues at the long N-terminal region (residues 1-13) of the two monomers are engaged in a non-covalent interface in an antiparallel fashion. The two subunits of the dimer sustained close packing interactions between sidechains of aromatic and hydrophobic residues, Y10/Y10′, M21/M21′ and Q19/Q19′ (Fig. 3b). Further, a number of cationic residues, namely R13, R14 at the β-turn region and K3 and K4 at the N-terminus, in the dimeric structure of thanatin are situated in spatially proximal orientation (Fig. 3b). Figure 3c shows the superposition of structures of free thanatin and in complex with LPS (for one of the subunit). As seen, there are significant differences between the LPS bound structure and free structure of thanatin both in backbone and sidechain dispositions. The RMSD values for backbone atoms and all heavy atoms were estimated to be 5.1 Å and 6.2 Å, respectively. Strikingly in complex with LPS micelle, the cationic N-terminal segment of thanatin comprising residues G1-P7, acquires a define conformation and folds back toward the β-hairpin structure. In particular, residues K 3 KPV 6 adopt a β-turn conformation centring residues K4 and P5 at i + 1 and i + 2 positions, respectively ( Supplementary Fig. 6). The electrostatic potential surfaces of dimeric and monomeric thanatin are shown in Fig. 3. The dimeric structure of thanatin in LPS demonstrates large cationic patches at the distal ends and an extended non-polar surface at the centre of the four stranded β-sheet. Therefore, in comparison to the monomeric structure, the dimeric structure of thanatin possesses higher cationicity (12.8 vs 7) and exposed hydrophobic surface (2203 Å 2 vs 1603 Å 2 ) amenable for LPS interactions.

Interactions of thanatin with LPS by Saturation Transfer Difference (STD) NMR. STD-NMR
method can map contact residues of ligands in complex with receptor molecules at an atomic resolution within a fast chemical exchange regime [36][37][38] . STD-NMR has been utilized to determine contact residues of potent pore forming AMPs in LPS micelle [39][40][41][42] . One-dimensional 1 H STD and two-dimension 1 H-1 H STD-TOCSY experiments were carried out to identify residues of thanatin in contact with LPS micelle (Supplementary Fig. 7). Due to better resolution, two-dimension STD-TOCSY spectra were analyzed identifying residues of thanatin in contact with LPS micelle (Fig. 4a). The STD-TOCSY spectrum of thanatin showed cross-peaks arising from CαHs to sidechains and among sidechain/sidechain resonances for a number of residue (Fig. 4a). As reported in earlier works, resonances demonstrating cross-peaks in the STD-TOCSY spectra are supposed to be receiving a high degree of saturation from the binding partners due to their close proximity [36][37][38] . STD-TOCSY correlations appear to be more intense among the sidechains resonances compared to that of backbone CαH to sidechain resonances, indicating closest proximity of sidechains to LPS micelle (Fig. 4a). In particular, conspicuous STD-TOCSY correlations could be detected for the basic residues including K3, K4, R13, R14, K17 and R20, indicating their proximity with LPS micelle. Residues P5, V6, P7 also delineated STD effect. Note, these N-terminal residues K3, K4, P5 and V6 have experienced large conformational change in complex with LPS micelle. By contrast, among the polar residues, (S2, N12, C11, T15, C18, and Q19), residues N12, T15 and Q19 showed STD-TOCSY correlations. The single aromatic residue Y10 displayed strong STD effect involving ring protons and CαH/CβH protons, indicating its association with LPS micelle (Fig. 4a (inset), Supplementary Fig. 7). Note, although residue Y10 is engaged at the dimeric interface of thanatin, calculation of solvent accessible surface area showed that the sidechain of residue Y10 is not completely buried and remains amenable for interactions with LPS micelles. The non-polar sidechains of residues I8, I9 and M21 remain in a close contact with LPS micelle as demonstrated by the strong STD-TOCSY correlations. Taken together, STD analyses suggest that most of the residues, including cationic, aromatic and non-polar, of thanatin are essentially in contact with LPS micelle.
Effect of thanatin on LPS micellar structure. We have examined structural changes of LPS micelles in complex with thanatin using FITC labelled LPS or FITC-LPS and 31 P NMR analyses. Emission intensity of FITC-LPS is quenched due to FRET among the fluorophore in an aggregated LPS micelle structure [43][44][45] . Dissociation of LPS micelle upon binding to proteins or pore forming AMPs often results in an enhancement of FITC emission intensity 44,45 . Note, in FITC-LPS conjugate, FITC is covalently linked with the sugar residues located at the solvent permeable surface of LPS 46,47 . Therefore, FITC fluorescence change would probe the LPS surface interactions with peptides. Figure 4b shows fluorescence emission spectra of FITC-LPS at various concentrations of thanatin, ranging from 0 to 100 μM. As seen, emission intensity of FITC-LPS demarcated a marked quenching with increased concentrations of thanatin (Fig. 4b). The diminution of emission intensity of FITC-LPS in complex with thanatin, indicated a higher FRET among the fluorophore FITC. In other words, FITC-LPS fluorescence data demonstrate that LPS micelles undergo a higher order association upon binding with thanatin. Perturbation of LPS micelle structure by thanatin was further probed from 31 P NMR experiments. 31 P NMR spectra of LPS were acquired in free solution and at three different molar ratios, 4:1, 2:1 and 1:1 of LPS:thanatin (Fig. 4c). The free LPS micelles produces two well resolved NMR signal at ~−1.00 ppm and −2 ppm, respectively 39,44 . Both the phosphate groups demonstrated extensive line broadening when in complex with thanatin, indicating involvement of the phosphate head groups in binding with thanatin (Fig. 4c). As evident, at higher concentration of thanatin, at 1:1 molar ratio, 31 P NMR signals of LPS appeared to be extremely broadened. Such resonance broadening effect may potentially arise from slower tumbling motion of aggregated LPS micelles in complex with thanatin. Taken together, the aforementioned experiments, FITC-LPS fluorescence and 31 P NMR indicated that the LPS micelles undergo structural rearrangements or an association to larger LPS aggregates in complex with thanatin.

Modelling of LPS-dimeric thanatin complex by Molecular Dynamics (MD) simulations. In order
to gain further insights into LPS-thanatin interactions, MD simulations were carried out for 3 repeats of 400 ns each in LPS/DPPE bilayer and dimeric thanatin (see Materials and Methods). The LPS/DPPE bilayer may be considered as a mimic of outer membrane of Gram negative bacteria where the outer-leaflet of the outer membrane is predominantly occupied by LPS molecules. Figure 5 summarises MD simulations of dimeric thanatin and LPS/ DPPE bilayer. Figure 5a shows that interactions between dimeric thanatin and LPS lipids during MD simulations revealed a rapid, within 40-50 ns, encountering and stabilization of the LPS-thanatin complex. Both the subunits of the dimeric thanatin have been found to form a network of interactions with multiple LPS molecules, primarily with the head regions, through multiple electrostatics, polar and van der Waals contacts (Fig. 5b,c and d). The dimeric thanatin latches onto the LPS surface following two different binding modes. In binding mode I, residues at the N terminal tail from one subunit and cationic residues R13 and R14 from the middle loop of another subunit appeared to establish interactions with LPS molecules (Fig. 5b and c). As seen, residues R13 and R14 form multiple hydrogen bonds with hydroxyl groups of LPS (Fig. 5c). Moreover, residues, P5′, V6′, P7′ and I8′ interact with the sugar rings of LPS through non-polar contacts (Fig. 5b). In the binding mode II, both the N terminal tails of the dimer are engaged in interactions with LPS surface (Fig. 5d, Supplementary Fig. 8). The basic residues K3′ and K4′ of the N terminal tail in the dimeric thanatin establish electrostatic interactions with LPS phosphate groups (Fig. 5d).
Ala scanning mutations of cationic and hydrophobic residues of thanatin. In order to correlate LPS-dimer thanatin structures and activity, critical residues Y10, M21, R13 and R14 of thanatin, involved in potential LPS recognition, were mutated and antimicrobial activities and LPS interactions were determined. We hypothesized that these analogs with double replacements may generate a larger difference in LPS binding and antimicrobial activity compared to single mutants. As seen, replacements of aromatic/hydrophobic packing residues Y10 and M21 to Ala had substantially decreased antimicrobial activity as indicated by higher MIC values of the analog peptide Y10M21AA. On the other hand, the R13R14AA analog was found to be largely inactive ( Table 2). Figure 6a shows dose dependent changes in zeta potential of E. coli cells by peptides Y10M21AA and R13R14AA. Although, additions of mutant peptides caused surface charge neutralization toward lower negative values, however, complete surface charge neutralization was not observed even at the highest concentration of peptides (30 μM). Therefore, unlike native thanatin, mutant analogs are less efficient in complementing surface charge of bacterial cells. We further estimated LPS binding of the mutant analogs by BC probe displacement studies. Figure 6b shows BC probe displacement of the mutant analogs Y10M21AA and R13R14AA. The ED 50 for Y10M21AA and R13R14AA peptides were estimated to be 1.22 μM and 3.20 μM, respectively. Note, the native thanatin demonstrated an ED 50 of 0.87 μM in probe displacement (Fig. 1d). These data indicate that the mutant analogs bind to LPS with lower affinity compared to the native thanatin. In particular, LPS binding affinity appears to be more diminished for R13R14AA analog. Perturbations of LPS micelle structure by analog peptides were further investigated. 31 P NMR studies of LPS demonstrated sustained intensity of 31 P signals after additions of Y10M21AA and R13R14AA analogs (Fig. 6c). Further, FITC-LPS fluorescence spectra were obtained for the analog peptides. As seen, additions of analog peptides into FITC-LPS solutions had yielded limited changes in Thanatin R13R14AA GSKKPVPIIYCNAATGKCQRM >100 >100 >100 >100 >100 >100 >100 >100 FITC fluorescence intensity, compared to the native thanatin (Fig. 6d). The aforementioned investigations clearly established that the analog peptide R13R14AA substantially lacked bactericidal activity whereas antibacterial activity of Y10M12AA peptide has also been reduced ( Table 2). These peptides were inefficient in bacterial surface charge neutralization. Furthermore, these analog peptides possess lower binding to LPS and were unable to generate structural changes to the LPS structures.

Discussion
The outer member LPS plays important roles in survival of Gram negative bacteria creating a permeability barrier against antibiotics and bactericidal agents [48][49][50] . Endotoxin function of LPS is also known to cause fatal septic shock syndromes in humans and other animals [51][52][53] . At present, there are no drugs of therapeutic modality against sepsis or septic shock 54,55 . LPS being a strong inflammatory molecule is also associated with other diseases 56,57 . Therefore, atomic-resolution structures and interactions analyses of AMPs in complex with LPS would not only yield mechanistic insights for bacterial outer membrane disruption, but structural scaffolds can be utilized for the generation of antibiotics and endotoxin neutralization drugs [58][59][60] . Towards this, 3-D structures of several pore forming AMPs in complex with LPS have been reported in past and recent years [39][40][41][42] . Opposed to pore formation which requires AMPs to disrupt inner membrane, bacterial cell agglutinating AMPs target outer membrane components 10,11,17,19 . Bacterial cell killing by agglutination has been described for a number of antimicrobial peptides, proteins and recently for amyloid Aβ peptides [61][62][63] . However, the mechanism of cell agglutination is yet to be elucidated. One of the key steps in cell agglutination of Gram negative bacteria is the binding of AMPs with the outer membrane LPS. In order to glean knowledge in cell agglutination process, we have determined the first 3-D structure of a host defence cell agglutinating AMP, thanatin, in complex with LPS micelle and investigated its interactions. Biophysical and optical spectroscopic studies demonstrated characteristic interactions and estimated thanatin-LPS binding parameters. Dansyl fluorescence emission of dans-thanatin indicated potential inclusion of thanatin into non polar environment of LPS micelles (Fig. 1b and c). On the other hand, exothermic binding interactions suggested that LPS-thanatin complex may be stabilized by ionic and/or hydrogen bonding (Fig. 1d, Table 1). Whereas, negative entropy change appears to suggest more ordering of the LPS-thanatin  (Table 1). Further, thanatin mediated efficient release of BC probe from LPS (Fig. 1e) and E. coli surface charge neutralization (Fig. 1f) indicated predominant ionic interactions with the lipid A head groups and outer membrane LPS, respectively. STD-NMR analyses demonstrated that thanatin has been intimately associated with LPS micelle utilizing many of the cationic and non-polar and aromatic residues (Fig. 4a). Further, binding of thanatin into LPS micelle yielded changes of LPS micellar structure, towards lager aggregates, as observed from 31 P NMR and FITC-LPS fluorescence studies (Fig. 4b and c). Such interactions and complex formation between LPS-thanatin would contribute to the outer membrane binding and cell agglutination process. In order to better correlate, atomic resolution structures of thanatin were determined in free solution and in complex with LPS micelle. Free thanatin assumed a β-hairpin structure for the central sequence (residues I8-M21) along with an N-terminal extended region ( Fig. 2a and b). By contrast, thanatin in complex with LPS micelles demonstrated a higher order structural assembly of a dimeric four stranded β sheet (Fig. 3). Notably, 3-D structures of pore forming β-sheet antimicrobial peptides in complex with LPS have been determined [64][65][66] . These structures revealed monomeric topology of β-sheet AMPs which may be necessary for outer membrane permeabilization followed by traversing to the bacterial plasma membrane As seen, the dimeric structure of thanatin has resulted in close proximity of the N-terminal β-strand of one subunit and N-terminal β-strand of another subunit in antiparallel manner (Fig. 3b). The antiparallel topology of thanatin also negates formation of larger aggregated structures in LPS micelle. The dimeric interface of thanatin sustained by aromatic residues Y10 (Y10′) and hydrophobic residues M21 (M21′) packing interactions across the two β-strands (Fig. 3b). These hydrophobic residues and potentially their mutual packing interactions appear to be important for the antibacterial activity and LPS interactions. As seen, mutated analog peptide Y10M21AA demonstrated diminished antibacterial activity, limited ability of surface charge neutralization and perturbation of LPS structure (Fig. 6). Furthermore, in the dimeric structure of thanatin, the extended N-terminus assumed more defined conformations and folded back along the β-sheet (Fig. 3c, Supplementary Figure 6). These structural changes have yielded two apposite large cationic surfaces in the dimeric structure of thanatin (Fig. 3e). Note, analyses of fluorescence and ITC data using dimer based peptide/LPS interactions were not successful, since the stoichiometric association between LPS micelle and thanatin is not well defined. Regardless, NMR studies further demonstrate that tr-NOESY is a sensitive method for detection of intermolecular NOEs of AMP ligands in complex with LPS micelle. NMR ( 31 P, STD), FITC and dansyl fluorescence and MD simulations studies indicated that thanatin assumed a surface localization with LPS micelle or bilayer interacting with the sugar and phosphates at the head group of LPS. MD simulations studies in LPS-DPPE bilayer demonstrated dimeric thanatin can interact with multiple LPS molecules using more than one binding mode or dynamical in nature. Thanatin interactions were predominantly observed with the sugar and phosphate head groups of LPS molecules. The basic residues e.g. K3, K4, R13, R14 and K17, belonging to the cationic patches were found to be primarily involved in LPS recognition ( Fig. 5c and  d). Antibacterial activity and biophysical studies of the analog peptide R13R14AA indicated critical involvement of residues R13 and R14 (Fig. 6). The atomic-resolution structure of thanatin solved in complex with LPS micelles thus provides mechanistic insights towards bacterial cell agglutination. The antiparallel topology of the dimeric structure of thanatin would potentially allow binding of two or more LPS molecules at the distal ends. The ability of the dimeric thanatin structure to interact with multiple LPS molecules may permit bacterial cells proximity. Further, surface charge neutralization and perturbation of LPS-outer membrane integrity may perhaps facilitate cell-cell associations and agglutination. Taken together, the atomic resolution structures and mode of interactions of thanatin presented in this work can be potentially used for further development of novel cell agglutinating antimicrobial agents.

Dansyl Cadaverine experiment.
To study the binding affinity of the peptides to LPS, peptides were dansylated at the N-terminus and the fluorescence measurements were carried out with Cary Eclipse fluorescence spectrophotometer (Varian inc). 50 μM dansylated peptides were excited at 330 nm wavelength and their emission fluorescence were monitored from 450-650 nm followed by the addition of increasing concentration of LPS. ΔF max was calculated as (F − F o )/(F max − F o ) and was plotted against concentration of LPS, where, F is the fluorescence intensity at maxima for any LPS concentration. F o is the basal fluorescence intensity at maxima. F max is the fluorescence intensity at maxima for maximum LPS concentration. To calculate the dissociation constant, the plot was fitted using Hill's equation.
Bodipy Cadaverine Displacement Assay. Relative affinities of thanatin and its analogs to LPS were examined in terms of their ability to displace BODIPY TR cadaverine (BC) from LPS-BC mixture. BC has a high affinity to LPS and its displacement from LPS by the peptides results in dequenching of its fluorescence. 5 μM BC and 50 μg/ml LPS mixture was prepared in 50 mM Tris buffer, pH 7.4. In a 96-well plate, 50 μl of peptides were prepared at final concentrations ranging from 500 μM to 0.01 μM by serially diluting in 50 mM Tris buffer, pH 7.4. Another 50 μl of the aforementioned LPS-BC mixture was added to the wells and fluorescence was measured by exciting samples at 580 nm and emission was recorded from 604-650 nm. BC occupancy factor was calculated as: BC Occupancy Factor = (F max − F)/(F max − F min ) Where, F = Fluorescence intensity at maxima at a particular mediator concentration, F max = Fluorescence intensity at maxima at maximum mediator concentration, F min = Fluorescence intensity at maxima at minimum mediator concentration. Occupancy factor was plotted NMR Spectroscopy. All the NMR spectra were recorded on a Bruker DRX 600 spectrometer, equipped with a cryo-probe and pulse field gradients. Data acquisition and processing were performed with topspin software running on Linux workstation. 0.4 mM peptides in aqueous solution containing 10% D 2 O at pH 5 were used to acquire two dimensional TOCSY (total correlation spectroscopy) and NOESY (nuclear overhauser effect spectroscopy) spectra at 298 K. Mixing times of 80 ms for TOCSY and 300 ms for NOESY were used. DSS (2,2-dimethyl-2-silapentane 5-sulfonate sodium salt) was used as an internal reference for chemical shift. NOESY experiments were performed with 400 increments in t 1 and 64 transients. WATERGATE procedure was used for water signal suppression. A total of 2 K data points were used in t2 and NMR data was analysed in SPARKY (T.D. Goddard and D.G. Kneller, University of California, San Francisco, CA, USA) program. To record a two-dimensional transferred NOESY spectrum, 30 μM of LPS was added to 0.4 mM peptides in aqueous solution contacting 10% D 2 O at pH 5. The experiment was performed with the same parameters as the NOESY experiment except that the mixing time was 75 ms. In order to study the interaction of peptides with LPS, one dimensional 31 P NMR spectra of 400 μM LPS in water, pH 6 was recorded on a Bruker DRX 400 spectrometer. Following this, a series of one-dimensional 31  further analysis. PROCHECK 67 was used to produce Ramachandran plot, which was in turn used to validate the structures calculated.

Molecular Dynamics (MD) Simulation.
We performed all atom molecular dynamics simulations based on a periodic boundary condition (PBC) box with 120 1,2-Dipalmitoyl-sn-glycero-3-phosphoethanolamine (DPPE) (in lower leaflet), 48 LPS (in upper leaflet) molecules and the thanatin dimer, in a 84 × 84 × 120 cubic box with around 11000 TIP3P water molecules and 192 Ca 2+ ions, as well as Cl − ions. This LPS/DPPE hybrid bilayer lipid initial model was adopted and modified from the long equilibrium lipid model by Kirschner et al. 68 The schematic diagram of LPS model used in this study has been depicted in Supplementary Fig. 9. The Glycam based force field was used for lipid system. As for two thanatin dimer, NMR model was utilized for simulations, using Amber 99SB-ILDN force field. We adopted the Amber99SB force field parameters for TIP3P explicit water model and ions in all the simulations. Once the simulation boxes were constructed, the systems went through a 1000-steps energy minimization to eliminate possible close contacts among atoms, followed by 100 ps equilibrium in NVT ensemble with restraints added to the heavy atoms of DPPE and LPS molecules. 10 ns equilibrium was employed in NPT ensemble simulations with similar restraints akin to NVT equilibration. Thanatin dimer was randomly inserted into the simulation box replacing water molecules. At 310 K, 100 ps NVT and 10 ns NPT equilibriums, heavy atoms of the thanatin dimer were also imposed by position restraints of 1000 kJ/(mol. nm2). Finally, three repeat runs for 400 ns each, with different initial velocities, were carried out under temperature 310 K. All the simulations were performed with Gromacs package together with Plumed package. The time step for all simulations was 2 fs, while the coordinate data were stored every 2 ps. In NVT and NPT ensembles, temperature coupling was realized using velocity rescaling thermostat scheme updating velocities every 0.1 ps, whereas in NPT ensemble simulations, semi-isotropic pressure coupling method Parrinello-Rahman was utilized to maintain the lipid bilayer structure. SHAKE algorithm and LINCS algorithm were applied, to restrict the bonds between hydrogen-heavy atoms and the heavy-heavy atoms covalent bonds, respectively, to an equilibrium value. Furthermore, PME scheme enabled the long range electrostatic interactions. For both long range electrostatic and van der Waals interactions, the distance cutoff was 1.2 nm. During simulations, dimeric structure of thanatin was maintained by applying eight distance restraints using a quadratic bias potential (force constant 1000) according to Gromacs manual 5.0. It may be noted that structure of dimeric thanatin in water is not favourable due to repulsion among cationic sidechains. The dimer structure is only stabilized in complex with LPS. The minimum distance and center of mass distance between LPS and thanatin dimer were calculated using Gromacs g_mindist and g_dist respectively. The coordination number (CN) of LPS/thanatin was defined as ij ij ij whereas, N LPS and N than are the total number of atoms in LPS and thanatin dimer respectively, and dij is the distance between atom i in LPS and atom j in thanatin dimer, a distance cutoff = 0.4 nm was used. The thanatin-LPS residue contact probability matrix ( × NRes NRes LPS t han ) was calculated using a distance cutoff = 0.4 nm by counting and normalizing through all the frames in the trajectories, whereas NRes LPS = 15 and NRes than = 42 were the number of residues in the LPS and the thanatin dimer. For each frame of thanatin-LPS complex, 0 or 1 was assigned to each bin according to equation (3) with cutoff = 0.4 nm to have a contact matrix for all residue-residue pairs, and all the matrices summed up and normalized by dividing by the total number of frames.
Availability of materials and data. All data generated or analysed during this study are included in this published article (and its Supplementary Information files).