Designed β-Boomerang Antiendotoxic and Antimicrobial Peptides

Lipopolysaccharide (LPS), an integral part of the outer membrane of Gram-negative bacteria, is involved in a variety of biological processes including inflammation, septic shock, and resistance to host-defense molecules. LPS also provides an environment for folding of outer membrane proteins. In this work, we describe the structure-activity correlation of a series of 12-residue peptides in LPS. NMR structures of the peptides derived in complex with LPS reveal boomerang-like β-strand conformations that are stabilized by intimate packing between the two aromatic residues located at the 4 and 9 positions. This structural feature renders these peptides with a high ability to neutralize endotoxicity, >80% at 10 nm concentration, of LPS. Replacements of these aromatic residues either with Ala or with Leu destabilizes the boomerang structure with the concomitant loss of antiendotoxic and antimicrobial activities. Furthermore, the aromatic packing stabilizing the β-boomerang structure in LPS is found to be maintained even in a truncated octapeptide, defining a structured LPS binding motif. The mode of action of the active designed peptides correlates well with their ability to perturb LPS micelle structures. Fourier transform infrared spectroscopy studies of the peptides delineate β-type conformations and immobilization of phosphate head groups of LPS. Trp fluorescence studies demonstrated selective interactions with LPS and the depth of insertion into the LPS bilayer. Our results demonstrate the requirement of LPS-specific structures of peptides for endotoxin neutralizations. In addition, we propose that structures of these peptides may be employed to design proteins for the outer membrane.

LPS 2 or endotoxin, a major component of the outer leaflet of the outer membrane of Gram-negative bacteria, is critically involved in health and diseases of humans (1,2). LPS is essential for bacterial survival through establishing an efficient permeability barrier against a variety of antimicrobial compounds including hydrophobic antibiotics, detergents, host-defense proteins, and antimicrobial peptides (3,4). Several studies have demonstrated that LPS catalyzes folding of outer membrane proteins as a chaperone (5)(6)(7).
LPS, a potent inducer of innate immune systems, hence called endotoxin, is primarily responsible for lethality in sepsis and septic shock syndromes associated with serious Gram-negative infections (8 -10). Circulating LPS in bloodstream is intercepted by the phagocytic cells of the innate immune system. Once induced by LPS, these phagocytes produce proinflammatory cytokines, e.g. tumor necrosis factor-␣, interleukin-6, and interleukin-1␤, through the activation of a Toll-like pattern recognition receptor (11,12). The release of cytokines in response to microbial invasion is a natural function of the innate immunity. However, an uncontrolled and overwhelming production of these cytokines may cause "endotoxic shock" or septic shock, typified by endothelial tissue damage, loss of vascular tone, coagulopathy, and multiple organ failure, often resulting in death (9,10). Sepsis is the major cause of mortality in the intensive care unit, accounting for 200,000 deaths every year in the United States alone (13). It was demonstrated that release of LPS from antibiotic-treated Gram-negative bacteria can indeed enhance sepsis (14). Therefore, an effective antibiotic should not only exert antibacterial activities but also have the ability to sequester LPS and ameliorate its toxicity. Therefore, an amalgamated property of LPS-neutralizing and antimicrobial activity would be highly desirable for antimicrobial agents. Polymyxin B is a prototypical antimicrobial and antiendotoxic antibiotic; however, its neurotoxicity and nephrotoxicity limit its application to topical use (15). The increasing emergence of bacterial strains that are resistant to conventional antibiotics has initiated vital structure/function studies of membrane-perturbing cationic antimicrobial peptides (16 -20). More recent studies have been conducted to understand interactions between antimicrobial peptides with LPS to gain insights into the mechanism of outer membrane perturbation, antibacterial activities, and LPS neutralization (21)(22)(23)(24)(25)(26). These studies have delineated the role of amino acid sequence properties, LPS-peptide interactions by biophysical methods, and global structural parameters, obtained by CD and FTIR.
Designing synthetic peptides and elucidation of three-dimensional structures in complex with LPS would be useful for the purpose of rational development of non-toxic antisepsis and antimicrobial therapeutics. Such studies will also be poten-tially instructive in establishing rules by which folded structures can be stabilized on the LPS surface. Extensive work in the field of peptide design primarily focuses on mimicking secondary structures and tertiary folds of proteins. Usually, short linear peptides are often structurally flexible; however, the functions of these peptides are highly dependent on their ability to adopt folded structures upon complex formation with their cognate receptors. In this regard, designed peptides that would yield high resolution structures in complex with LPS have not been well pursued. LPS, being a negatively charged amphiphilic molecule, interacts with naturally occurring peptides or protein fragments containing basic/polar and hydrophobic amino acids, although there are considerable variations in lengths, sequences, and amino acid compositions among these peptides (27,28).
Here, we have determined the three-dimensional structures of a series of 12-residue peptides in the context of LPS. To the best of our knowledge, these results show, for the first time, that atomic resolution structures of designed peptides obtained in LPS could be correlated with their antiendotoxic activities. Furthermore, the LPS-induced structures of active, inactive, and short peptide motif, presented here, may provide building blocks for the designing novel proteins for the outer membrane.

EXPERIMENTAL PROCEDURES
Reagents-LPS of Escherichia coli 0111:B4 and fluorescein isothiocyanate (FITC)-conjugated lipopolysaccharide from E. coli 055:B5 and spin-labeled lipids 5-doxyl-stearic acid (5-DSA) and 16-doxyl-stearic acid (16-DSA) were purchased from Sigma. Peptides were synthesized commercially by GL Biochem (Shanghai, China) and further purified by a reversephase HPLC, Waters TM , using a C 18 column (300 Å pore size, 5-m particle size) by a linear gradient of acetonitrile/water mixture. The molecular weight of the peptides was confirmed by mass spectrometry.
LPS Neutralization Assay-The ability of the designed peptides to neutralize or inhibit LPS was assessed using a quantitative chromogenic limulus amoebocyte lysate (LAL) with a QCL-1000 (Cambrex) kit. Endotoxin neutralization experiments were carried out following the protocols provided by the vendor and published else where (26,27). Stock solutions of peptides were prepared in pyrogen-free water provided with the kit. Peptides at concentrations of 0.01, 0.05, 0.1, 5, and 10 M were incubated with three different endotoxin units (EU) of LPS, namely 1, 3, and 8 EU/ml (1 EU ϳ0.1 ng of LPS), in a flat bottom nonpyrogenic 96-well tissue culture plate, at 37°C for 30 min to allow peptide binding to LPS (26). A total of 50 l of this mixture was then added to equal volume of LAL reagent, and the mixture was further incubated for 10 min followed by the addition of 100 l of chromogenic substrate (Ac-Ile-Ala-Arg-p-nitroaniline). The reaction was terminated by the addition of 25% acetic acid, and the yellow color that developed due to cleavage of the substrate was measured spectrophotometrically at 410 nm using a Benchmark plus microplate spectrophotometer (Bio-Rad). The reduction of A 410 as a function of peptide concentrations is directly proportional to the inhibition of LPS by the peptide (29,30). All assays were repeated twice, and average values are reported.
Determination of Minimum Inhibitory Concentration-Antimicrobial activities of the designed peptides were determined following a previously reported method (31). Briefly, bacterial cells used for this assay, e.g. E. coli DH5alpha, Bacillus subtilis, Pseudomonas aeruginosa ATCC 27853, and Staphylococcus aureus ATCC 25923, were cultured in Luria-Bertani (LB) media at 37°C overnight. Cells were centrifuged and washed with the assay buffer (10 mM sodium phosphate buffer, pH 7.4) and diluted to an A 600 of 0.2. About 50 l of these bacterial cell suspensions were incubated, in a sterile 96-well microtiter plate, with the same volume of peptides at various concentrations, ranging from 1 to 200 M, diluted from a stock solution of 0.3 mM (prepared in the same buffer) at 37°C for 2 h. The cell suspensions were then plated onto Mueller-Hinton agar plates and incubated overnight. The minimum inhibitory concentration was expressed as the lowest concentration of the peptide where there was no visible growth of the bacteria.
Fluorescence Studies-All of the fluorescence experiments were performed using a Cary Eclipse fluorescence spectrophotometer (Varian, Inc.). To study the interactions of peptides with FITC-conjugated LPS, 0.5 M FITC-LPS samples were excited at 480 nm, and change in the emission of FITC at 515 nm was monitored with various concentrations (0.01, 0.02, 0.1, 0.5, 1.0, 5.0, and 10 M) of peptides. Samples were prepared in 10 mM phosphate buffer, pH 6.0. For the intrinsic Trp fluorescence studies of peptides, 5 M of each peptide was titrated with varying concentrations of LPS or DPC in a 10 mM sodium phosphate buffer at pH 6.0. The intrinsic tryptophan fluorescence emission spectra of the peptides in their free or lipid-bound forms were acquired by exciting samples at 280 nm using band passes of 5 nm for both the excitation and the emission monochromators in a 0.1-cm path length quartz cuvette. Quenching of tryptophan fluorescence was examined following sequential additions of various concentrations (0.02-3 M) of acrylamide into solutions containing peptide (5 M) in its free and lipidbound forms. The results of the quenching reactions were analyzed according to the Stern-Volmer equation, F 0 /F ϭ 1 ϩ K SV [Q], where F 0 and F are the fluorescence intensities at the emission maxima in the absence and presence of quencher, respectively, K SV is the Stern-Volmer quenching constant, and [Q] is the molar quencher concentration.
Measurement of Depth of Insertion of the Peptides into LPS Vesicles-Quenching of Trp fluorescence by spin-labeled lipids (5-DSA and 16-DSA) was used to estimate the depth of insertion of the peptides into LPS bilayer or vesicle by parallax method (32). LPS bilayer was prepared by dissolving the appropriate amount of LPS in 2:1 chloroform/methanol solution. The organic solvent was evaporated to dryness under vacuum. The lipid film was hydrated with 10 mM phosphate buffer, pH 6.0, at 60°C and vortexed briefly. This mixture was frozen and thawed five times and extruded through a 0.1-m membrane with the extruder (Avanti Polar Lipids, Alabaster, AL). Various concentrations of spin-labeled lipids, 5-DSA or 16-DSA, were added, from a stock solution of 1 mM (prepared in methanol) into solutions containing 5 M peptides and 40 M LPS vesicle. The location of the tryptophan into LPS bilayer was determined by comparing the extent of quenching observed from shallow (5-DSA) and deep (16-DSA) quenchers following the equation (29) where Z 1F is the difference in depth between the shallow quencher and the tryptophan residue, and F 1 and F 2 are the difference between the tryptophan fluorescence intensities in the presence and absence of shallow and deep quenchers, respectively. Assuming the usual surface area of the lipid to be 70 Å, C is the quencher mole fraction in unit area. L 21 is the difference in depth between the two quenchers. Once Z 1F is calculated, the distance of tryptophan from the center of the bilayer was calculated from Z CF ϭ Z 1F ϩ L cl , where L cl is the distance from the center of the bilayer to the shallow quencher. Isothermal Titration Calorimetry (ITC)-ITC experiments were performed using a VP-ITC microcalorimeter (MicroCal Inc., Northampton, MA). All samples, dissolved in 10 mM phosphate buffer, pH 6.0, were degassed prior to use. LPS at a concentration of 0.05 mM was loaded into the sample cell (volume 1.4359 ml), and the reference cell was filled with the above mentioned buffer. Peptides, at a concentration of 1 mM, were placed into the injection. A typical titration involved 35 injections of 2.5-l aliquots of YI12 peptides into the sample cell, at an interval of 4 min, at 25°C. The reaction cell was stirred continuously at 300 rpm. Raw data were collected and integrated using the MicroCal Origin 5.0 software supplied with the instrument. A single site binding model was fitted to the data by non-linear least square analysis to yield the association constant (K a ) and enthalpy change (⌬H). ⌬G and ⌬S were calculated using the fundamental equations of thermodynamics: ⌬G ϭ ϪRT ln K a and ⌬S ϭ (⌬HϪ ⌬G)/T, respectively.
Dynamic Light Scattering-To obtain information on the ability of designed peptides to dissociate LPS aggregates, dynamic light-scattering measurements were carried out in an BI-9000AT with digital autocorrelator (Brookhaven Instruments Corp., Holtsville, NY). The peptide and buffer solutions were filtered through 0.45-m filters (Whatman Inc). Measurements were made for 1 M LPS (without any peptides), and upon incubation with 2 M peptides, the scattering data were collected at 90°. The data were analyzed through the standard CONTIN method using the dynamic light-scattering software supplied with the instrument.
NMR Experiments-All of the NMR spectra were recorded on a Bruker DRX 600 spectrometer, equipped with cryo-probe and pulse field gradients. Data acquisition and processing were performed with the Topspin software (Bruker) running on a Linux workstation. Sequence-specific resonance assignments of the peptides were achieved from two-dimensional total correlation spectroscopy (TOCSY) and nuclear Overhauser effect spectroscopy (NOESY) spectra acquired in aqueous solution containing 10% D 2 O at pH 4.8, 298 K. The peptide concentrations were 0.6 mM, and mixing times were 80 and 400 ms for TOCSY and NOESY, respectively. The interactions of the designed peptides with LPS were examined by recording series of one-dimensional proton NMR spectra whereby 0.6 mM peptides were titrated with various concentrations, 5, 10, and 16 M LPS. Tr-NOESY spectra were typically obtained either at 10 M or at 13 M LPS, which generated a large number of Tr-NOE cross-peaks. The two-dimensional Tr-NOESY spectra were recorded at three different mixing times: 100, 150, and 200 ms with 512 increments in t 1 and 2048 data points in t 2 . Tr-NOESY spectra were also obtained in D 2 O for unambiguous assignments of aromatic/aromatic or aromatic/aliphatic NOEs. The spectral width was normally 12 ppm in both dimensions. After 16 dummy scans, 72 scans were recorded per t 1 increment. NMR data analyses were carried out using the program SPARKY (T. D. Goddard and D. G. Kneller, University of California, San Francisco, CA). 31 P NMR spectra of LPS were recorded on a Bruker DRX 400 spectrometer at 298 K. Data acquisition and processing were performed with the Topspin software (Bruker) suite. The interactions of the designed peptides with LPS were examined by recording series of one-dimensional 31 P NMR spectra whereby 0.2 mM LPS in water (pH 4.5) was titrated with various concentrations (0.1, 0.2, and 0.4 mM) of designed peptides from a stock solution that was prepared in an unbuffered water (pH 4.5).
NMR-derived Structure Calculation-NMR structures were calculated using the DYANA program, version 1.5 (33). NOE intensities were qualitatively categorized as strong, medium, and weak based on cross-peak intensities in the Tr-NOESY spectra obtained at a mixing time of 150 ms. The NOESY crosspeaks were further translated to upper bound distance limits of 2.5, 3.5, and 5.0 Å, corresponding to strong, medium, and weak intensities, respectively. Only the dihedral angles were constrained between Ϫ30 and Ϫ180°to maintain a good stereochemistry of the calculated structures. Out of the 100 structures generated, the 20 lowest energy structures were used for more analysis.
Infrared Spectroscopy-FTIR spectra were recorded on a Nicolet Nexus 560 spectrometer (Thermo Fisher Scientific, Inc.) purged with N 2 and equipped with a mercury-cadmium-telluride (MCT/A) detector cooled with liquid nitrogen. Attenuated total reflection (ATR) spectra were measured with a 25-reflections ATR accessory from Graseby Specac (Kent, UK) and a wire grid polarizer (0.25 mM, Graseby Specac). Approximately 200 l of a D 2 O solution of LPS alone or in the presence of peptide in a 20:1 lipid/peptide molar ratio were applied onto a trapezoidal (50 ϫ 2 ϫ 20 mm) germanium internal reflection element. A dry, or D 2 O-saturated, N 2 stream flowing through the ATR compartment was used to remove bulk water (low hydration) or to fully hydrate the sample (high hydration), respectively. A total of 200 scans were collected at a resolution of 4 cm Ϫ1 , averaged, and processed with one-point zero filling and Happ-Genzel apodization.

RESULTS
Design of Peptides-In an earlier study, we have determined LPS-bound structure and antiendotoxic activity of a 12-residue synthetic peptide (34). This peptide, YVLWKRKRMIFI, was designed using a co-crystal structure of LPS/FhuA (1 QFG) an outer membrane protein. The outer membrane proteins of Gram-negative bacteria are rich in ␤-sheet, assuming a ␤-barrel topology (35). The folding and stability of these proteins are maintained by a specific environment of LPS (5-7). In complex with LPS, the designed peptide assumed a novel amphipathic structure whereby the cationic and hydrophobic residues were segregated into distinctly different regions (34). The C terminus of the peptide showed two consecutive ␤-turns, whereas the N terminus appeared to be extended. In LPS inhibition assays, the peptide showed relatively weak activity (IC 50 ϳ 10 M). Interestingly, in the LPS-bound state, residues Trp 4 and Met 9 of the designed peptide showed NOE contacts, indicating that they are within Յ5 Å apart. The Met residue was introduced into the primary sequence as an NMR chemical shift marker (34). Here, we have replaced Met 9 with aromatic residues, Trp, Phe, and Tyr, to enhance packing interactions with Trp 4 in complex with LPS. Such substitution may result in a defined folded structure with a large hydrophobic surface of the peptide in the context of LPS with a plausible enhancement in endotoxin neutralization and antimicrobial activities (Table 1). To underscore the role of presumable aromatic-aromatic interactions, peptides containing Ala 4 /Ala 9 and Leu 4 /Leu 9 were also prepared ( Table 1). The Leu residues were particularly introduced to determine correlation of the hydrophobicity and aromatic-aromatic packing to the folding and activities of these peptides. To understand the specific role of Trp in the structure/activity, another peptide containing Phe at positions 4 and 9 has also been made (Table 1). In addition, an octapeptide was prepared to elucidate the role(s) of the hydrophobic residues at the N and C termini (Table 1).
LPS Neutralization and Antimicrobial Activities of the Designed Peptides-To determine the ability of the peptides to inhibit or neutralize LPS, sensitive chromogenic LAL assays were conducted (see "Experimental Procedures"). This assay can detect endotoxin at very low concentrations down to ϳ1 pM. LAL assays were conducted at three different LPS concentrations, 1, 3, and 8 EU/ml, with six different concentrations of peptides. As can be seen, peptides containing aromatic residues at positions 4 and 9, i.e. YI12WF, YI12WW, YI12WY, and YI12FF (Table 1), demonstrated the inhibition of LPS-mediated activation of LAL enzyme (Fig. 1). Peptides YI12WF, YI12WW, and YI12WY inhibited Ն80% endotoxin even at a concentration of 10 nM at 1 EU/ml ( Fig. 1, top) and 3 EU/ml ( Fig. 1, middle). At these LPS concentrations, Ն95% inhibition is observed at 100 nM concentrations for YI12WF, YI12WY, and YI12WW peptides (Fig. 1, top and middle). The YI12FF peptide shows a weak inhibitory activity, only Յ40% at 10 M concentration at 1-and 3-EU/ml doses of LPS (Fig. 1, top and middle).
a A C 18 reverse phase semi-preparative column was used. The peptides were eluted using a 60-min linear gradient of acetonitrile (10%) and water (90%) containing 0.1% trifluoroacetic acid (v/v). b YI12FF peptide shows Ͻ40% neutralization of LPS at 10 M concentration. c ND: no detectable inhibition.
Even at a much higher concentration of LPS (8 EU/ml), YI12WF, YI12WY, and YI12WW peptides demonstrate ϳ80% neutralization of endotoxin at 10 M concentrations (Fig. 1, bottom). No detectable inhibition of LPS was found for the YI12LL or Y12AA peptides even at a higher concentration of Ն100 M (data not shown). The truncated 8-residue GG8WF peptide (Table 1) also lacks antiendotoxic activity (data not shown).
We have also examined antimicrobial activities of these peptides against two Gram-negative and two Gram-positive bacterial strains ( Table 1). The order of antimicrobial activities follows YI12WY Ͼ YI12WF Ͼ YI12WW (Table 1). Other peptides are found to be rather inactive against these bacteria, although YI12FF and YI12LL showed some antibacterial activity only against the E. coli strain (Table 1). Peptide hydrophobicity has been used as a parameter to explain the antibacterial and hemolytic activities of antimicrobial peptides (36). The hydrophobicity of peptides can easily be measured from their retention time in a C 18 column. A longer retention time is an indicative of a higher hydrophobicity. However, the lack of LPS neutralization properties of the YI12LL and YI12AA peptides or the diminished activity of the YI12FF peptide cannot be simply explained by the hydrophobicity of these peptides ( Table 1). The HPLC retention time of the active peptide YI12WY is similar to the largely inactive peptide YI12LL (Table 1). Furthermore, the YI12FF peptide, showing reduced LPS-neutralizing activity, also has a similar hydrophobicity to that of YI12WW (Table 1). One would find a positive correlation between LPS-neutralizing activities and hydrophobicity by comparing HPLC retention time for the inactive Y12AA peptide with the active peptides (Table 1). However, the lack or reduced activities of the YI12LL and YI12FF peptides clearly indicate the requirement of more specific interactions. These data suggest that rather than a global hydrophobicity, an explicit involvement of aromatic residues at positions 4 (Trp) and (Phe, Tyr, or Trp) 9 are important for LPS inhibitory activities of the designed peptide.
LPS Binding Affinity of the Peptides-ITC is used to determine binding interactions of the designed peptides with LPS ( Fig. 2). Binding parameters can be obtained for all but GG8WF peptide as a result of the unsaturable binding. Table 2 summarizes thermodynamic parameters of the interactions. The LPSpeptide interactions are characterized by an endothermic heat released or entropy-driven process, as indicated by upward trends of the ITC profiles (Fig. 2). The entropy-driven complex formation at 25°C between LPS and peptide had been reported earlier (37,38). In this work, the temperature was kept below the phase transition temperature (ϳ37°C) of LPS favoring endothermic binding reactions. As can be seen, all peptides interact with LPS at M affinities with K d ranging from as high as 5.9 M to 1 M (Table 2). These data indicate that mere LPS binding of peptides do not correlate with LPS neutralization or antimicrobial activities. Interestingly, the inactive peptides, YI12AA and YI12LL, show a somewhat high affinity binding to LPS as compared with the active peptides ( Table 2).
Effect of Designed Peptides on LPS Micelles-We have probed structural perturbations of LPS micelles upon interactions with designed peptides using fluorescence of FITC- conjugated LPS, dynamic light scattering (Fig. 3), and 31 P NMR of LPS (Fig. 4). As proposed earlier, LPS molecules form soluble aggregates, causing quenching of the fluorescence intensity of FITC (39). Binding of peptides or proteins with LPS may cause an enhancement of the FITC fluorescence as a result of the plausible dissociation of LPS aggregates (21,39,40). This observation has been correlated with the ability of the LPS-interacting peptides with their endotoxin neutralization activities (21,40). It has been shown that the aggregated forms of LPS or lipid A are biologically more potent than the monomeric forms (41).  (Fig. 3A). There was no increase in FITC fluorescence in the presence of inactive peptides, YI12LL, YI12AA, or GG8WF. Interestingly, additions of the YI12AA peptide in FITC-LPS solutions appeared to cause a quenching of FITC fluorescence (Fig. 3A), indicating a different structural change of LPS aggregates (see below). The disaggregation of LPS by active peptides, YI12WF, YI12WY, and YI12WW, is also seen from dynamic light-scattering studies (Fig. 3B). The LPS alone produces a polydisperse sample with a diameter centered at 7000 nm (Fig.  3B). There is a dramatic shift of the average size of LPS toward the lower values in the presence of YI12WF and YI12WW peptides (Fig. 3B). Similar results were also obtained for the active peptide YI12WY (data not shown). The YI12FF peptide, showing a reduced LPS neutralization, disaggregates LPS at a lower extent as compared with the highly active peptides (Fig. 3B). On the other hand, the inactive peptide, YI12AA, does not show any dissociation of LPS aggregates (Fig.  3B). Rather, in the presence of YI12AA, LPS becomes more polydisperse with populations having larger sizes (Fig. 3B). This observation may explain the quenching of FITC fluorescence detected in the case of the YI12AA peptide (Fig. 3A) because more aggregation of LPS will reduce the fluorescence intensity of FITC (Fig. 3A).
We have utilized 31 P NMR of LPS to study its interactions with the active peptide YI12WF and the inactive peptide YI12AA. LPS produces two well separated 31 P resonances at Ϫ2.00 and Ϫ0.72 ppm (Fig. 4). By comparison with 31 P NMR   (Fig. 4, left panel). The upfield shifted 31 P resonance (ϳϪ2.00 ppm) of the diphosphate groups experiences a progressive change in line width as a function of increasing concentrations of YI12WF peptide (Fig. 4). The resonance at Ϫ0.72 ppm for monophosphate groups also shows broadening effects, although at a lower extent, upon interactions with YI12WF peptide (Fig. 4). By contrast, additions of the YI12AA peptide do not yield any significant changes in the 31 P resonances of LPS (Fig. 4, right panel), indicating limited structural perturbation of LPS micelles by the inactive peptide. The dramatic changes in 31 P resonances of LPS caused by the YI12WF peptide indicate conformational changes and disordering of the head group regions of LPS (43). Such structural perturbations may essentially result in dissociation of larger LPS aggregates into small sizes. We preclude aggregation, as may be interpreted from the larger line width of 31 P resonances, of LPS micelles in complex with YI12WF peptide because two resonances are differently changed upon complex formation. The enhanced broadening of 31 P resonance of the diphosphate groups is an indication of conformational exchange(s) among different states of peptide-LPS complexes. In addition, it is also noteworthy that the diphosphate groups of LPS provide a facile interaction site for the cationic peptide as a result of its higher negative charge density. Collectively, these data suggest that LPS neutralization ability of the active peptides may occur from the plausible dissociation and perturbation of LPS micelles (see "Discussion").

Solution NMR and Tr-NOE Studies-
The solution conformations of all the designed peptides in their free states were examined using two-dimensional 1 H-1 H NOESY and TOCSY.
Sequence-specific resonance assignments for all the amino acids were achieved by the analyses of TOCSY and NOESY spectra (data not shown) (44). The NOESY spectra of the free peptide were predominantly characterized by intraresidue and sequential NOEs between the backbone protons and the side chain proton resonances (supplemental Fig. S1), indicating that peptides are highly mobile in their free forms and do not adopt any unique conformation(s).
Additions of low concentrations of LPS into the solutions containing peptides had caused concentration-dependent broadening of almost all the proton resonances without any significant change in chemical shifts, demonstrating a fast or intermediate exchange between free and LPS-bound states at the NMR time scale (supplemental Fig. S2) (see "Experimental Procedures"). Two-dimensional 1 H-1 H NOESY spectra of the active peptides, YI12WF, YI12WY, and YI12WW, were obtained in the presence of LPS at a peptide/LPS molar ratio of 35:1. A large number of NOE crosspeaks were observed as a result of Tr-NOE (45, 46) effects, implying well folded conformations of the LPS-bound states of the peptides (Fig. 5). LPS forms high molecular weight bilayers/micelles in solution at a very low concentration (14 g/ml) (47,48), enabling determination of LPS-bound structures of peptides by the Tr-NOE method (49 -53). Analyses of Tr-NOESY spectra reveal strong sequential ␣N (i, iϩ1) and weak HN/HN NOEs for the hydrophobic/aromatic residues at the N and C termini, whereas more medium range NOE contacts of the type HN/HN (i to iϩ2) and C␣H/HN (i to iϩ2) were observed for the central positively charged KRKR segment (supplemental Fig. S3). Most importantly, the NOE contacts between the Trp 4 ring protons with the aromatic residue at position 9, i.e. Phe 9 , Tyr 9 , or Trp 9 in YI12WF, YI12WY, and YI12WW, respectively, are unambiguously assigned (Fig. 5). For example, the well separated N ⑀ H proton, resonating at 10.18 ppm, of Trp 4 shows NOE contact with the ␤-protons (2.96 ppm) of Phe 9 residue (Fig. 5A, left  panel). There were NOE contacts between W4H4 and W4H5 to F9C ␤ Hs (Fig. 5A, right panel). Similarly, in the YI12WY peptide, long range NOE contacts were observed between the ring protons and indole N ⑀ H proton of Trp 4 with the C ␤ Hs of Tyr 9 (Fig. 5B). In the case of the YI12WW peptide, NOEs between the non-degenerate ␤-protons of Trp 4 and Trp 9 were detected (Fig. 5C, inset). In addition, NOEs involving W9H6 proton with Trp 9 C ␤ Hs (Fig. 5C) and N ⑀ H of Trp 4 with Trp 9 C6H proton were identified in complex with LPS (Fig. 5C, left panel). The YI12FF peptide also showed NOEs between residues Phe 4 and Phe 9 in complex with LPS (data not shown). Tr-NOESY spectra of the inactive peptide YI12AA showed different NOE contacts whereby long range NOEs were detected between A4C ␣ H/ Phe 11 ring protons and Tyr 1 ring protons with Ile 12 C ␦ H 3 group (supplemental Fig. S4).
Three-dimensional Structures of the Peptides Bound to LPS-Ensembles of high resolution structures of the active peptides, YI12WF, YI12WY, YI12WW, and the inactive YI12AA peptide, as a complex with LPS, were obtained by use of Tr-NOE-driven distance restraints (Table 3). The LPS-bound conformations of the designed peptides are well defined as determined by close superposition of the calculated structures and low root mean square deviation values ( Fig. 6 and Table 3). All three active peptides, YI12WF, YI12WY, and YI12WW, acquire an Designed ␤-Boomerang Antiendotoxic Peptides AUGUST 14, 2009 • VOLUME 284 • NUMBER 33

JOURNAL OF BIOLOGICAL CHEMISTRY 21997
amphipathic structure in complex with LPS. The positively charged residues at the center form a loop defining the cationic face of the molecule, whereas the N and C termini hydrophobic residues adopt extended or ␤-strand conformations, making the hydrophobic surface (Fig. 6, A-C). In complex with LPS, the hydrophobic face of the active peptides is defined by aromatic-aromatic packing between the critical residue Trp 4 , with aromatic residues Phe 9 , Tyr 9 , or Trp 9 in YI12WF, WY, and WW peptides, respectively (Fig. 6, A-C). In addition, in all three peptides, the N terminus residue Val 2 is packed against Trp 4 , and the C terminus Phe 11 makes facile contacts with Phe 9 , Tyr 9 , and Trp 9 , presumably strengthening the non-polar cluster (Fig. 6, A-C). The ␤-type structure with the aromatic-aromatic packing between residues 4 and 9 of the active peptides resembles a "boomerang" or ␤-boomerang structure. A helical boomerang structure was earlier described for the fusion domain of the influenza virus hemagglutinin protein in DPC micelles (54). The boomerang structure of the 20-residue fusion domain, defined by two helical segments, is dictated by the packing between two aromatic residues located at i and iϩ5 positions, i.e. Phe 9 and Trp 14 . These residues are placed at the base of an intervening hydrophilic loop (Fig. 7). The superposition of the LPS-bound structure of the YI12WF peptide with the DPC-bound fusion domain structure shows the boomerang architecture of the designed peptide (Fig. 7). Interestingly, replacement of aromatic residue with Ala had been shown to dramatically reduce the fusion activity of the domain with the concomitant disruption of the boomerang structure (55).
The LPS-bound NMR structure of the inactive peptide YI12AA has been determined ( Fig. 6D and Table  3). The YI12AA also folds into an amphipathic structure; in the context of LPS, however, the absence of the aromatic-aromatic packing at i to iϩ5 stabilizes a rather open struc-ture at the 4 AKRKRA 9 segment (Fig. 6D). The hydrophobic residues at the N and C termini are clustered into one side with aromatic residues Tyr 1 and Phe 9 facing each other (Fig. 6D).

An LPS Binding Structured Motif Defined by GG8WF Peptide-
Encouraged by the observation of a strong tendency of aromatic/ aromatic packing between residues 4 and 9, we determined the LPS-bound structure of the short 8-residue peptide GG8WF (Table 1). GG8WF also yields a large number of Tr-NOE crosspeaks while bound to LPS (Fig. 8A). The indole N ⑀ H of Trp 2 shows NOE contact with C ␤ Hs of Phe 7 (Fig. 8A, left panel). The C ␤ Hs of Phe 7 also shows NOEs with the C6H and C2H ring protons of Trp 2 (Fig. 8, right panel). Interestingly, more long range NOEs were detected between C␣H protons of residue Gly 8 with ring protons of residue Trp 2 (Fig. 8A). An ensemble of structures of GG8WF was obtained from Tr-NOE-driven distance constrains including 12 long range NOEs (Fig. 8B and Table 3). The well defined structure shows packing between Trp 2 /Phe 7 rings with the positively charged residues segregated out at the top (Fig. 8C). The phenyl ring of Phe 7 appeared to stack over the indole ring of Trp 2 (Fig. 8C). This structural scaffold derived for the GG8WF peptide in LPS may be termed as a "structured LPS binding motif." In principle, this independently folded sequence motif could be introduced into other peptides or proteins to develop endotoxin-neutralizing molecules or LPS targeting novel proteins (see "Discussion").
Localization of Peptides in LPS-Intrinsic Trp fluorescence was used to determine localization of the active peptides,   . Boomerang structure of the influenza hemagglutinin fusion peptide and the designed peptide YI12WF. Shown is the superposition of the structure of fusion domain of hemagglutinin determined in DPC micelles and the LPS-bound structure of YI12WF peptide. The 20-residue viral fusion domain adopts a boomerang-like structure, represented by the two segments of helices, held by a long range (i to iϩ5) packing between two aromatic residues Phe 9 and Trp 14 . Replacements of the aromatic residues with aliphatic ones showed a dramatic loss of fusogenic activity of the peptide. The LPS-bound conformation of the 12-residue designed peptide studied here shows a close topological similarity with the fusion domain structure. The image was produced using the program MOLMOL. Ramachandran plot for the mean structure % residues in the most favorable and additionally allowed region 90 90 90 100 100 % residues in the generously allowed region  (Table 4). By contrast, peptides show limited blue shift in the zwitterionic DPC micelles (Table 4), indicating that peptidelipid interactions are specific to the type of lipids. To assess the solvent exposure of the Trp, fluorescence-quenching studies were carried out using a neutral quencher acrylamide (Table 4). All peptides showed low Stern-Volmer quenching constants (K SV ) as compared with the free peptides and peptides in DPC micelles (Table 4), indicating that the Trp residue is well buried in LPS but largely exposed in the DPC micelles. Further, to determine insertion of the Trp residue into the LPS bilayer, fluorescence-quenching studies were carried out by two spin-labeled lipids, 5-DSA and 16-DSA. Trp fluorescence intensity was quenched for all four peptides either by the shallow quencher 5-DSA, containing spin label at position 5, or by the deep quencher 16-DSA, containing spin label at position 16. The extent of quenching was used to determine the depth of the insertion into LPS bilayers (see "Experimental Procedures"). The depths of penetration of the Trp residue for the four peptides in their LPS-bound states are shown in Table 4 and found to be located around ϳ7 Å from the center of LPS bilayer. These depth values indicate that the peptides are rather deeply inserted into LPS bilayer. It is noteworthy that the octapeptide GG8WF also demonstrated a similar depth of insertion (7.18 Å), emission maxima, and K SV values in LPS (Table 4), as compared with the 12-residue active peptides. These data indicate an independent ability of the GG8WF peptide to be localized into the same environment of LPS.

Secondary Structures and LPS Interactions by FTIR Studies-
The assignment of the bands in the amide I region was performed as reported (56). The spectra in the amide I region of the different peptides examined are shown in Fig. 9A. In all cases, except in the peptide GG8WF, there is a strong component at ϳ1630 cm Ϫ1 indicative of ␤-structure (57). When these structures are antiparallel, it is accompanied by a much less intense band at ϳ1690 cm Ϫ1 , but in this case, it may be obscured by the strong band at 1670 cm Ϫ1 corresponding to turns. By contrast, for the peptide GG8WF, the spectrum is consistent with turns (band at 1670 cm Ϫ1 ) and also some flexible structures (broad band centered at ϳ1640 cm Ϫ1 ). The effect of the peptides on the phosphate of LPS was examined by looking at the antisymmetric phosphate bands in the region 1200 -1260 cm Ϫ1 where the band is split in at least two vibrational bands with different hydration states, with the band at a higher wave number corresponding to representing lower hydration (57). In this case, two main bands were observed in LPS at both low and high levels of hydration, at 1237 and 1210 cm Ϫ1 (Fig. 9, B and C). In agreement with the above, when the sample was fully hydrated, the band at 1237 cm Ϫ1 (lower hydration) became attenuated (in Fig. 9, B and C, compare spectrum LPS, see arrow). After the addition of the peptides, another band appears at ϳ1200 cm Ϫ1 , as reported previously for the interaction of other peptides with LPS (22,57). At both low and high hydration, the bands from the antisymmetric phosphate of LPS disappear, indicating immobilization of the phosphate groups (22). Taken together, these results demonstrate that all of the 12-residue peptides, active or inactive, form ␤-type structures in LPS with a turn component. These results are in complete corroboration with the LPS-bound NMR structures of the peptides whereby the central positively charged residues form the turn or loop and the N and C termini hydrophobic residues are in ␤-strand conformations.

DISCUSSION
Due to the prime importance as an endotoxic agent, LPS is a target for the development of antisepsis drugs (27,59,60). LPS also serves as a potential barrier toward antimicrobials at the outer membrane of Gram-negative bacteria (3,4,61). In addition, LPS has been found to be highly necessary for the correct folding of the outer membrane proteins (5)(6)(7). Antimicrobial and endotoxic functionalities of LPS had recently generated considerable interest in investigating naturally occurring antimicrobial peptide-LPS interactions (18 -23, 62-66). Helical amphipathic cationic peptides, consisting of amino acids Leu and Lys, were designed to understand interactions with LPS (22,40). Highly ␤-sheet peptides (33-residue) were designed based on ␣-chemokines and human neutrophil bactericidal protein (BPI) for LPS neutralization and antimicrobial activities (67,68).
It is noteworthy that the design of YI12 peptides, as studied here, does not follow the usual amphipathic sequence pattern for helices or ␤-sheets as described earlier (22,40,67,68). Instead, we assume that the centrally located four positively charged residues will help in initiating a binding process via electrostatic interactions with LPS head groups; this may allow the hydrophobic sequences at the N and C termini to integrate into the acyl chains of LPS, stabilizing defined ␤-type structures (34). The structural organization of these peptides may essentially act as outer membrane protein mimics (see below). Here, we are able to demonstrate critical structural features that are essential for the LPS neutralization and antimicrobial activities of these peptides. We established that a long range aromaticaromatic packing (i to iϩ5) between residues located at positions 4 and 9 in the 12-residue sequences is highly important for the antiendotoxic and also antimicrobial activities of the designed sequences. In the LAL assay, the active peptides,  YI12WF, YI12WY, and YI12WW, showed very high LPS neutralization potency ( Fig. 1 and Table 1). A much lower LPS neutralization activity of the YI12FF peptide ( Fig. 1 and Table 1) indicates a specific role of the Trp residue at this position. Trp has been known to play an important function toward the membrane incorporation of antimicrobial peptides (69). The Tr-NOE-driven structures obtained for the active peptides, YI12WF, YI12WW, and YI12WY, in LPS showed aromatic packing interactions between Trp 4 and Phe 9 /Tyr 9 /Trp 9 (Fig. 6). The central positively charged residues assume loop-like structures (Fig. 6). The hydrophobic residues at the N and C termini are extended or ␤-strands and lay close to the packed aromaticaromatic cluster. IR studies also demonstrated the presence of turn and ␤-strand structures of the peptides in LPS (Fig. 9). The active and inactive peptides bind to LPS with similar affinity (Fig. 2 and Table 2) and acquire folded structure in complex with LPS (Fig. 6). It appears that the aromatic packing may render a specific orientation of the hydrophobic termini, resembling a boomerang. The replacement of the critical aromatic residues either by Ala or by Leu resulted in a non-packed structure, although amphipathic with impaired activities (Figs. 1 and 6). The electrostatic surface of the active peptides clearly shows a positively charged "head" and a well packed elongated non-polar "body" (Fig. 6). We have examined surface charges of a number of outer membrane proteins with known three-dimensional structures. Very interestingly, several outer membrane proteins contain structural regions that show similar surface topology to the designed peptides (Fig. 10). Fig. 10 illustrates a selected region from outer membrane proteins, FhuA (A), OmpA (B), and PhoE (C), showing positively charged patch, made of surface loops, at the top and hydrophobic ␤-sheets at the bottom. The positively charged loops are exposed to solvent, and hydrophobic ␤-sheets reside at the transmembrane. These structural motifs of the outer membrane proteins play critical roles for binding to LPS (70) and LPS-mediated folding from the unfolded states (6,7). Tryptophan fluorescence and depth measurements studies demonstrated that the non-polar ␤-strands of the designed peptides are indeed well inserted into the hydrophobic region of LPS micelles (Table 4). Therefore, LPS-bound structures of YI12 peptides determined here represent a mimic, at least partly, of the outer membrane proteins.
The probable mechanism by which active peptides are able to neutralize LPS appeared to correlate with their ability to disaggregate and/or perturb LPS micelles (Figs. 3 and 4). Dynamic light-scattering and FITC fluorescence studies clearly demonstrated that active peptides can dissociate LPS micelles. Dynamic light-scattering experiments revealed that the large aggregated forms of LPS converted into smaller sizes upon binding with active peptides (Fig. 3B). 31 P NMR resonances of LPS were highly affected in the presence of active peptide YI12WF (Fig. 4). However, the inactive peptide YI12AA did not cause any perturbation to 31 P resonance of LPS (Fig. 4), although ITC, NMR, and IR studies show that YI12AA peptide interacts and adopts a folded structure in LPS. Therefore, the perturbation of 31 P resonances of LPS micelles can be used as a specific probe for the LPS-neutralizing and antimicrobial peptides. Recently, we reported that 31 P resonances of LPS undergo significant changes while binding to a highly active antimicrobial peptide MSI-594 (71). Intrinsic Trp fluorescence emission spectra and quenching of fluorescence intensity by soluble quencher and spin-labeled lipid quenchers indicated that the peptides are rather deeply inserted into LPS micelles and also in the bilayer ( Table 4). The deletion of hydrophobic residues at the N and C termini of the most active LPS-neutralizing peptide YI12WF showed a dramatic diminution of LPS neutralization and antimicrobial activities (Table 1). Taken together, we propose a mechanistic model explaining the probable mode of action of the designed peptides (Fig. 11). Two aromatic residues, in particular Trp 4 and Phe/Tyr/Trp at the 9 th position appear to serve as a lock to secure the insertion of the peptide into LPS. The hydrophobic arms, at the N and C termini, are involved in the probable disorganization of LPS, leading to endotoxic neutralization activity of the peptides (Fig. 11). The mode of action of the designed peptides for antimicrobial activities may also be determined by the compact ␤-boomerang conformations in LPS micelles that may aid in permeabilization of the outer membrane of the bacteria (71).
Furthermore, our results show that the deleted analog, GG8WF, also assumes a well defined structure in complex with LPS, retaining the aromatic/aromatic packing (Fig. 8). The Trp residue of GG8WF is found to be in the non-polar acyl region of LPS (Table 4). The independently folded structure of this peptide in LPS may be considered as a structured LPS binding motif. LPS binding motif has been recognized in the x-ray structure of E. coli outer membrane protein FhuA, which was unexpectedly co-crystallized with one molecule of LPS (70). In that structure, four non-contiguous positively charged residues, Lys 306 , Lys 351 , Arg 382 , and Lys 482 , were found to be forming salt bridges and hydrogen bonds with phosphate groups of the LPS and were termed as an LPS binding motif (72). The LPS-bound structure of the G-WKRKRF-G peptide demonstrated a novel LPS binding motif that is determined by a stretch of contiguous residues. We surmise that this short sequence motif can be exploited as an LPS-anchoring sequence with the further incorporation of new synthetic functional ele-ments at its N and C termini. Such modifications can generate novel LPS-neutralizing molecules for plausible practical applications. Furthermore, the outer membrane of Gram-negative bacteria contains a number of important proteins. These proteins are transmembrane ␤-barrels having a variety of functions including ion transport, passive nutrient uptake, membranebound enzymes, and drug discharge (35). Many of these proteins, therefore, are targets for the development of antimicrobial compounds. The current LPS-induced structure-sequence correlation of the designed peptides, presented here, might aid in developing novel proteins that may fold into the outer membrane with ␤-sheet topology. Designing proteins for the outer membrane could be useful to disrupt transmembrane organization of the therapeutically important pathogenic proteins. Disruption of transmembrane helical interfaces by designed peptides has recently been demonstrated in type I membrane protein integrins (73). In conclusion, we have established, using a set of design peptides, vital structural determinants for LPS neutralizations by ␤-sheet peptides. The ability of the active and inactive peptides and the short peptide motif to fold into well defined ␤-type structures in LPS will be a starting point to build up ␤-sheet-rich native outer membrane proteins.