Study of the Interaction of Human Defensins with Cell Membrane

The interaction of eugenol with cell membrane models at the air–water interface is ... Rondinelli D. Herculano , Felippe J. Pavinatto , Luciano Case...
0 downloads 0 Views 473KB Size
Download PDF
11318

J. Phys. Chem. B 2007, 111, 11318-11329

Study of the Interaction of Human Defensins with Cell Membrane Models: Relationships between Structure and Biological Activity Marcos R. Lourenzoni,* Adriana M. Namba, Luciano Caseli, Le´ o Degre` ve, and Maria E. D. Zaniquelli Departamento de Quı´mica, Faculdade de Filosofia, Cieˆ ncias e Letras de Ribeira˜ o Preto, UniVersidade de Sa˜ o Paulo, 14040-901 Ribeira˜ o PretosS. P., Brazil ReceiVed: October 30, 2006; In Final Form: July 10, 2007

The HNP-1, HNP-2, and HNP-3 defensins are human antimicrobial peptides produced in response to microbial invasion. Their properties are distinct, with a more potent action for HNP-3. In this study, the relationship between their structural dissimilarities and their different microbial actions was evaluated by molecular dynamics simulation. Structural determinants related to their intra- and intermolecular interactions were defined for each HNP using a simplified membrane model consisting of a water/n-hexane interface. The hydrophobic portion of the HNPs promotes their diffusion to the interface with a concomitant, slight change in the structure induced by the intermolecular electrostatic interactions between the HPN molecules and the interface. As a consequence, different orientations are probably adopted by the HNPs at the interface, which may explain their different actions. The interaction of HNP-1 and HNP-2 with the surfaces was also studied using Langmuir monolayers as a biomimetic system. It was found that peptides adsorb rapidly at n-hexane/water interfaces as well as at phospholipid Langmuir monolayers but not at the air/liquid interface. This reveals that the presence of an organic phase is required for the exposure of the hydrophobic groups of the peptides. In addition, adsorption kinetics and surface pressure-area isotherms for Langmuir monolayers suggested that the lipidpeptide interaction is strongly influenced by the monolayer electrical charge and packing, depending also on the HPN structure. This study supports a model in which defensins, acting in a dimeric form, are able to disrupt membranes. The model also shows that the individual structures of the HNPs are responsible for their different actions on microbes.

Introduction Defensins are antimicrobial peptides produced by the innate immunity mechanism of higher vertebrates in response to microbial invasion.1,2 They contribute to host defense by acting rapidly to kill a broad range of potentially pathogenic microbes.3-6 Studies show that HNPs 1-3 suppress the activity and replication of the HIV-1 virus, thus having a powerful therapeutic value in AIDS treatment as well.7 Mammalian defensins are classified into the R, β, and θ categories8,9 on the basis of their size and disulfide bonding pattern. Human neutrophil peptides (HNPs) 1, 2, and 3 are the three most abundant forms of human R-defensins, constituting 30-50% of the protein contents of human azurophil granules and 5-7% of the neutrophil’s total cellular protein contents,10,11 and the primary sequences of HNPs are displayed in Figure 1.15 The three-dimensional structure of human defensin HNP-3 has been established by X-ray crystallography,12 and human HNP-1 structure has been studied by 2D-NMR.13,14 They are dimers consisting of two asymmetric monomers, each of them being formed by a hydrogen-bonded pair of antiparallel β-strands connected by a short turn, thus forming a β-hairpin.14 Although HNPs are structurally similar, they have distinct antimicrobial properties.1 Most HNPs are active against both Gram-negative and Gram-positive bacteria, fungi, viruses, and mycobacterium, while others are also active against protozoa and chlamydia. HNP-3 is less microbiocidal than HNPs 1 and * Corresponding author. E-mail: [email protected].

Figure 1. Alignment of the multiple amino acid sequences of the R-defensins (human HNP-1, HNP-2, and HNP-3, and rabbit NP-3a) following the work of Selsted and Harwig (ref 15). The disulfide bridges are represented by solid lines and the β-strand regions by arrows.

2 against most of the tested microbes.1 The mechanisms by which microorganisms are killed and/or inactivated by defensins are not yet fully understood. However, it is generally accepted that the killing of microorganisms is a consequence of microbial membrane disruption. A number of studies have contributed to evolving a mechanistic model of defensin microbiocidal activity, and they have revealed the general mode of action as being the permeabilization of the plasmatic membrane and the consequent leakage of cell contents, depending, therefore, on the structural amphiphilicity.11,16,17 In this sense, some studies have focused on the investigation of these defensins in biomembrane models. Wimley et al.18 have observed the permeabilization of large unilamellar vesicles (100 nm diameter) formed from 1-palmitoyl-2-oleoyl phosphatidylglycerol by the dimeric human neutrophil defensin HNP-2 in the absence of a membrane potential. Upon binding,

10.1021/jp067127g CCC: $37.00 © 2007 American Chemical Society Published on Web 09/05/2007

Human Defensins Interactions with Cell Membranes considered primarily as electrostatic, HNP-2 apparently breaches the bilayer by aggregation to form pores with a diameter of about 25 Å, formed by a hexamer of dimers. This is consistent with tryptophan fluorescence and circular dichroism measurements, which reveal that the secondary structure of human defensins does not change significantly upon interaction with bilayers.19 In summary, available evidence for the pore model suggest that the extracellular defensin initially binds electrostatically to anionic groups present on the bilayer surface.20 Insertion and reorientation across the bilayer may require, or is facilitated, by a transmembrane potential. The molecular dimensions of the dimer allow it to span the bilayer, and electrostatic interactions of the arginine side chains with phosphate head groups may stabilize the insertion.20 Formation of the pore may require lateral diffusion and assembly of six (or more) defensin dimers,20 and target cell death is thought to result from the efflux of cytoplasmic constituents through a pore.20 The validity of this model is currently under investigation. The available models for the HNPs-membrane interaction do not account for the role of the hydrophobic region of the membrane in the maintenance of the peptide structure inside the membrane. Therefore, we decided to use molecular dynamics simulations (MD) to identify what is the influence of the hydrophobic region of membranes in the maintenance of the quaternary structure of HNP dimers. In this way, we could identify the main factors providing the distinct antimicrobial activities of the dimeric forms of the human defensins HNP-1, HNP-2, and HNP-3 in a membrane interface model (MIM). In this work, MD studies of dimeric HNPs in an MIM formed by n-hexane and water were performed. This choice allowed us to study the role played by the interface formed by the polar and nonpolar phases in the absence of the complications arising from the slow relaxation of collective motions of lipid molecules that accompany conformational changes in the peptides as shown in the work of Gorfe et al.21 Many solutes, including peptides, exhibit qualitatively similar behavior at different interfaces between water and nonpolar media.22-25 These similarities are largely due to the importance of the balance between hydrophilic interactions with the aqueous environment and dispersive interactions between the hydrophobic portions of the defensin molecules and the hydrocarbon chains of the phospholipids.25 In this sense, the use of lipid Langmuir monolayers (LM) is a powerful tool to investigate peptide-membrane interactions.26 The MIM mimics the hydrophilic/hydrophobic membrane environment representing the unordered fatty acid chain matrix and the neighboring polar phase. The objective of the MD studies was to understand how the interactions between the inner part of the membrane and the HNP dimers are able to contribute to the maintenance of the quaternary structures of the HNPs, as well as gain some insight into their different activities. The hydrophobic portion of the bilayer membrane was mimicked by the n-hexane phase. This kind of MIM has been applied with success in the work of Freitas et al. in the study of the interaction of the Ebola fusion peptide in a membrane-mimetic environment.27 Moreover, in order to obtain more realistic information about MIM, LM experiments were performed with defensins HNP-1 and HNP-2 using both zwitterionic and negatively charged phospholipids. The main objective of this work is to compare the molecular data provided by MD studies, mainly for the HNP structure, with the results of LM as MIM. The results obtained by the different techniques add to our understanding of the action of human defensin on membranes.

J. Phys. Chem. B, Vol. 111, No. 38, 2007 11319 Experimental Methods Molecular Dynamics Simulation. Explicit solvent MD simulations were carried out using a GROMACS simulation package version 3.228 with a GROMOS96 (43A2) force field.29 Separate solvent boxes of water and n-hexane molecules were previously equilibrated and used to form a simulation box with an interface formed by two slabs, n-hexane and water. The simulation box, with dimensions in the directions x ) y ) 61.7 Å and z ) 88.0 Å, was divided into two slabs of 45.0 Å each, with 5232 water and 745 n-hexane molecules, thus constituting two phases. Chloride counterions were introduced in the aqueous phase (six, six, and four ions in the HNP-1, HNP2, and HNP-3 systems, respectively) in order to render each system electrostatically neutral. The Lennard-Jones and electrostatic interactions were considered until 14 Å, and the particle-mesh Ewald technique30 was used for the treatment of the electrostatic interactions since some residues are ionized at pH ) 7. The “leapfrog” algorithm31 was used to integrate the equations of motion with a time step of 2.0 fs over a total time of 40.0 ns. Periodic boundary conditions and minimum image convention were applied,32,33 and the neighbor list32 was updated every three steps. To maintain the system at a constant temperature of 300 K, a Berendsen thermostat was applied using a coupling time34 of 0.1 ps. The simulations were performed with a constant area in the xy plane, but variable in the other planes, where the pressure in the z direction was controlled and maintained in 1 bar using weak pressure coupling,34 with τp ) 1.0 ps. The protein bond lengths involving hydrogen atoms were constrained by means of the LINCS algorithm,35 and the SETTLE algorithm36 was used to constrain water molecules. The interface system was pre-equilibrated for 1.0 ns. Cell dimensions were adjusted to produce the correct experimental densities of 997 kg m-3 for SPC water model37 and 656 kg m-3 for n-hexane.38 The initial set of HNP-3 dimer atomic coordinates was obtained from X-ray crystallographic data (pdb code: 1DFN).12 The starting structures for the HNP-1 and HNP-2 dimers were obtained from mutation of the HNP-3 structure, since the structures of HNP-1 and HNP-2 are not deposited presently in the protein data bank.39 For HNP-1, the Asp2 residue side chain of HNP-3 structure was changed for the Ala side chain, and for HNP-2, the Asp2 residue of HNP-3 was eliminated. The HNPs were inserted in the middle of the aqueous phase with their hydrophobic residues oriented toward to n-hexane slab, forming an interface in the xy plane in the middle of the simulation box. Simulation Data Analysis Methods. Analyses were applied to the simulation data to investigate the influence of n-hexane on the structure of HNPs and their behavior at interfaces. Some properties such as the root-mean-square deviation (rmsd), intermolecular interaction potentials, gyration radii (GR) and hydrogen bonding (HB) were calculated and monitored during the simulation with the analyses program of the GROMACS package version 3.2.28 The rmsd between two structures reflects the conformational changes induced by the presence of HNP structures at interfaces. During the simulations, the atomic positions of the HNP structures were collected at every 2 ps and fitted with the reference structure obtained from X-ray diffraction data. Since the secondary and tertiary structures were maintained by three disulfide bonds in each monomer,13,19 this procedure was applied to identify the oscillation in the quaternary structures of HNPs. Intermolecular interaction potentials (IIP), defined as the sum of the electrostatic and Lennard-Jones interaction potentials between pairs of nonbonded atoms present in two different

11320 J. Phys. Chem. B, Vol. 111, No. 38, 2007 groups, were monitored during the simulation and then plotted in the graphic form. The pairs of different groups of atoms such as HNPs/n-hexane, HNPs/water, and HNPs monomers (Mo_A and Mo_B) were selected to calculate IIP. Since the interaction potentials can be different among HNPs, a comparative analysis can indicate the different behavior of the various HNPs at a hydrophobic/hydrophilic interface. Gyration radii was calculated for protein backbone atoms only, namely, CR carbon, amidic N and H, and the carbonyl group C and O atoms. GR is used to show the different packing of the HNP structures. Hydrogen bonding was determined by using a geometrical criterion where the distance (r) between the acceptor (A) and donors (D) atoms A‚‚‚H-D was within rAD < 3.5 Å and the angle (R) between the pairs of atoms A‚‚‚D-H was R < 60°. In order to calculate the electrostatic potential isocontours near the molecular surface, the PYMOL software package40 was used to run Poisson-Boltzmann equation (APBS).41 The last MD configuration of the HNPs was used to obtain the electrostatic potential isocontours, considering that the HNPs were inserted into a continuous pure solvent with a dielectric constant of 78.5. This procedure shows the possible differences in the electrostatic potential isocontours of the last MD configuration of the HNPs at the surface, detects the regions of the protein that are most probably in contact with the employed solvents (water or n-hexane), and shows their relationship with the maintenance of the structures of the HNPs. Two kinds of HNPs area were considered at the interface: (a) contact areas of the images projected on the xy plane, in which only HNP atoms immersed or in contact with the n-hexane slab are considered (AR-I) and (b) areas of the images projected on the xy plane of all the HNP atoms (AR-P). The areas were determined by means of a homemade program that computed the areas during the simulation. The AR-P value was obtained by adding the circular areas of each projected image of HNP atom in the xy plane, considering the van der Waals radii of the atoms. As for AR-I, the area values were computed for the HNP atoms that were in contact with n-hexane. A contact was scored when given pair of atoms (one from the HNP and one from the n-hexane carbon atom) were separated by the shortest distance possible as calculated from the sum of their van der Waals radii. Langmuir Monolayer Experiments. All experiments were performed with water purified by a Milli-Q system from Millipore (pH 5.6 and resistivity 18.2 MΩ cm, surface tension, γ0, 72.8 ( 0.08 mN‚m-1 at 23 °C). Defensins HNP-1 and HNP2, dipalmitoylphosphatidic acid (DPPA), and dimistoylphosphatidilcholine (DPPC) were acquired from Sigma; n-hexane and chloroform were purchased from Merck. The experiment temperature was 23 ( 1 °C. The LM experiments were performed with HNP-1 and HNP-2 because they are commercially available, while HNP-3 is not. Peptide Adsorption at Lipid-free Interfaces. Peptide adsorption at lipid-free interfaces was followed by surface pressure measurements with time (π(t) ) γ0 - γ(t)), by means of experiments using a pending drop of the peptide aqueous solution and the axisymmetric drop shape analysis (ADSA)42,43 (contact angle tensiometer from Dataphysics model OCA 20, Germany). Drop images were recorded at 3 s intervals, with the aid of a trigger system available in the software for the capture of the freshly formed surface created during the drop formation (final volume of about 40 µL). Surface tension, γ(t), was calculated from the extracted profile of the drops using the Laplace equation. A small glass box with water was placed

Lourenzoni et al. below the drop in the closed measurement compartment to avoid evaporation and consequent volume shifts during the experiment. In order to simulate a hydrophobic interface, a water/n-hexane interface was also investigated. Instead of air, the peptide aqueous solution drop was formed in a solution containing n-hexane, and the interfacial tension was determined as described for air/water interfaces. The surface pressure values were determined considering γ0 as the value measured for the pure n-hexane/water interfacial tension (50.2 mN/m). Adsorption on Lipid Monolayers. Adsorption experiments on lipid monolayers were also performed in a drop tensiometer from Dataphysics (model OCA 20, Germany). Aliquots of approximately 0.1 mmol L-1 of DPPA or DPPC dissolved in chloroform were spread on the surface of the drop by carefully touching it with a small drop of the lipid solution, formed in a 2 µL Hamilton syringe. After solvent evaporation (approximately 3 min), the drop size was changed (by gently sucking in or pushing out the drop liquid by means of an automated syringe) until a desired surface pressure was reached. Then, a small aliquot of the peptide was injected into the drop. The dilution factor was lower than 10, considering the drop volume and lipid aliquot. In the same way, drop images were recorded in 3 s intervals, by using the trigger system available in the software for the capture of the freshly formed surface created by the drop expansion. The time zero was considered immediately after the end of the drop expansion, and posterior changes in surface pressure were considered an effect of the peptide adsorption. By retrieving the recorded movie, the surface tension values were calculated as a function of the time. A control experiment was carried out by expansion of the drop containing the phospholipid but without injection of the peptide. No shift in surface tension was observed after complete drop formation in this case. As the lateral diffusion of the phospholipids is very fast if compared with the time resolution of our setup, spreading/adsorption kinetics were not followed. Surface Pressure-Area Isotherms. Contrary to former experiments44 using ADSA to obtain surface pressure-area curves (π-A isotherms), we did not use a coaxial tip but a normal needle. The phospholipid was spread in the same way as described in the adsorption experiments. Monolayer compression was accomplished by sucking the drop at a rate of 0.5 Å2 (lipid molecule)-1 s-1, with the help of an automated syringe. The π-A isotherms were obtained for pure lipid monolayers as well as for mixed peptide/lipid monolayers. In the latter case, the peptide was injected into the drop containing the lipid monolayer at zero surface pressure. After a new surface pressure stabilization, the drop was sucked in to obtain the desired interface compression. Isotherms for the surface of the drops formed by peptide solution only were also performed as a control, but no significant change in the surface pressure was detected. Results and Discussion Molecular Dynamics. Alignment of the multiple amino acid sequences of the R-defensins following the homology numbering of Selsted and Harwig15 is shown in Figure 1. The first amino acids of HNP-1 and HNP-3 are in the second position of the alignment, whereas for HNP-2 it is located in the third position. Six cysteine amino acids in the HNP sequences contribute to the stabilization of β-sheet secondary structures by three disulfide bonds. The crystal structure of HNP-3 is formed by a dimer in the shape of a basket, in which the bottom is hydrophobic and the top contains polar side chains as well as the ionized N- and C-termini of both monomers,13 as depicted in Figure 2. The ionized arginine residues enhance the am-

Human Defensins Interactions with Cell Membranes

J. Phys. Chem. B, Vol. 111, No. 38, 2007 11321

Figure 2. Two views of the R-carbon backbone of the HNP-3 dimer from the crystalline structure (ref 12) illustrating the basketlike shape with a flattened basket top containing the N- and C-termini; the hydrophobic region is located at the bottom of the image.

Figure 4. (a-d) Intermolecular interaction potential profiles between HNPs and n-hexane. The vertical lines delimit the instants when the HNPs reach the interface.

Figure 3. A1 and A2 angle profiles (a) HNP-3n and (b) HNP-3.

phiphilic character of the HNPs. These features enable the diffusion of the HNPs toward the interface. MD simulations initially were performed with HNP-3 in two different orientations, one of them being in the same orientation of HNP-1 and HNP-2, in which the hydrophobic part is toward to n-hexane phase. The other simulation with HNP-3 is denominated HNP-3n, and in this case, the top of basket that contains polar side chains is in the direction of the n-hexane phase. This procedure was adopted in order to test the effectiveness of the initial orientation of the other three HNPs in the MD to reduce the simulation time. This procedure was evaluated verifying the similarity between the final orientations of HNP-3 and HNP-3n close to interface. In order to follow the reorientation of protein in HNP-3n, the two angles formed between vectors depicted in Figure 2 were monitored. A1 is the angle formed between Vp and Vs, and A2 is between Vs and Vz. The Vz vector is perpendicular to the interface and aligned with the z dimension of the coordinate system, Vs is on the antiparallel β-strands in the bottom of basket (Figure 2), and Vp is perpendicular to Vs and is located in the middle of structure. Figure 3a shows the variation of A1 and A2 angles. The angle A1 remained close to 90°, and it sample that the HNP-3n structure remains unchanged, guaranteeing that the measured angle A2 has the information about reorientation only. At the beginning of simulation, the A2 angle is 180° because the top of basket is oriented and close to interface. In the first 1 ns, a radical change in the orientation (A2 ∼ 90°) is observed, but this angle increased in the next 4 ns reaching A2 ∼ 180°. After this instant (∼5 ns), a decrease of A2 (∼45°) is observed that remains until 22.5 ns, and after this moment the A2 angle

oscillates close to zero. This characterizes a complete rotation of the protein that orients the bottom of the basket to the n-hexane phase, which is the same orientation in the beginning of other HNPs simulations. The angles A2 of HNP-3n and HNP-3 (Figure 3, parts a and b) converge as shown by profiles that oscillate around similar angles. Figure 4a shows the IIP HNP-3n-n-hexane; the energy is nil until 5 ns, which is the same instant at which an abrupt decrease of A2 was observed, showing that the protein rotates in the aqueous phase, searching for a better orientation toward the interface. At 10 ns the effective contact between the protein and the n-hexane phase occurs. Such interactions take place in the time intervals required for the HNPs to reach the interface, which are indicated by vertical lines in Figure 4. The average properties calculated after these instants are listed in Table 1. The average energy is -156 ( 12 kcal‚mol-1 (Table 1), which is close to the HNP-3 simulation presented in Figure 4d and Table 1, -160 ( 12 kcal‚mol-1. The convergence of the orientations of HNP-3 in the two simulations can also be confirmed by comparing the values in the Table 1, which are similar. The average IIPs between Mo_A and Mo_B of HNP-3 and HNP-3n show a low divergence because the average IIP is calculated immediately after the contact between the HNP-3n and the n-hexane, where in this instant (∼10 ns) the HNP-3n has not reached the maximum interaction with n-hexane yet. Notice in Figures 3a and 4a that only after 22.5 ns the HNP-3n reaches an equilibrium position at the interface, being that two monomers have similar average IIP with n-hexane phase (profiles not shown). Consequently, the difference of average IIPs between monomers can be explained by the fact that the first contact of protein with n-hexane (∼10 ns) is realized by Mo_B in the case of HNP-3 and Mo_A in the case of HNP-3n as characterized by lower energy between two monomers in average IIP (Table 1). These results show that the initial orientations of HNPs in the three MD simulations are valid for the convergence of the orientations of HNP-3 and HNP-3n; however, the HNP-3n

11322 J. Phys. Chem. B, Vol. 111, No. 38, 2007

Lourenzoni et al.

TABLE 1: Some Average Properties after the Contact of HNPs with n-Hexane Phases

HNP-1 HNP-2 HNP-3 HNP-3n j

rmsda

EPb

EWc

EHd

EH Mo_Ae

EH Mo_Bf

AR-Ig

AR-Ph

GRi

2.3 ( 0.2 3.8 ( 0.2 2.3 ( 0.3 2.3 ( 0.3

-32 ( 7 -24 ( 8 -65 ( 9 -37 ( 9

-802 ( 75 -819 ( 46 -868 ( 45 -911 ( 58

-151 ( 21 -163 ( 13 -160 ( 12 -156 ( 12

-8 ( 21 -0 ( 10 -9 ( 10 -81 ( 8

-92 ( 8 -3 ( 12 -81 ( 8 -5 ( 10

324 ( 72 367 ( 70 272 ( 82 319 ( 67

698 ( 32 741 ( 21 733 ( 19 725 ( 23

10.3 ( 0.1 11.6 ( 0.2 10.3 ( 0.1 10.3 ( 0.1

a Average rmsd: (Å). b Average IIP between Mo_A and Mo_B (kcal‚mol-1). c Average IIP between HNPs/water (kcal‚mol-1). d Average IIP between HNPs/n-hexane (kcal‚mol-1). e Average IIP between Mo_A/n-hexane (kcal‚mol-1). f Average IIP between Mo_B/n-hexane (kcal‚mol-1). g Average AR-I of the HNP atoms in contact with n-hexane (Å2). h Average AR-P of the HNPs atoms (Å2). i Average GR (Å). j Data of the HNP-3n simulation.

requires a longer simulation time to reach a stable position at the interface than for the case of HNP-3. Therefore, the results of HNP-1, HNP-2, and HNP-3 only will be shown and discussed. In Figure 4, in the beginning of the IIP between HNPs/nhexane, the energy of HNP-1 reaches the minima at 0.7 and at 19.0 ns, the first minimum being observed in a shorter time than that obtained for the case of the other two HNPs, 2.5 and 3.0 ns. The energy minima that characterize the instant when the maximum interactions between HNP-1 and the n-hexane phase occur are observed after 19.0 ns only. The average IIP between HNPs/n-hexane at the interface are lower for HNP-2 and HNP-3 when compared to HNP-1, and they are around -163 and -160 kcal‚mol-1, respectively. The average IIP between the Mo_A of the HNP-1 and n-hexane is -58 kcal‚mol-1. It is contact with the interface before 19.0 ns since the average IIP for the other monomers at the interface are around -80 kcal‚mol-1, which suggests that the contact of HNP-1 occurs chiefly through Mo_B. The rmsd data presented in Table 1 for the HNP backbones show that deviation from the initial structure oscillates around 2.3 Å for HNP-1 and HNP-3 and 3.8 Å for HNP-2. These 2.3 Å deviations are not sufficiently significant to change the tertiary or quaternary structures, as already shown by MD simulation of the HNPs in aqueous solution45 and interface,46 thus indicating that the structures of these HNPs in this interface also are maintained. Under these conditions, the HNP-1 and HNP-3 dimers maintain the shape of a basket partially flattened on the top, according to the structures determined by crystallography and in aqueous solution by NMR and MD.13,45 In the case of HNP-2, the highest rmsd among HNPs (∼3.8 Å) indicates a more flattening on the top of the structure. The high rigidity of the monomers structures is explained by the presence of the three disulfide bonds, which had been previously characterized by nuclear magnetic resonance (NMR),13,14 circular dichroism (CD), and fluorescence spectroscopy measurements.19 The rmsd results for each monomer of HNP-1 and HNP-3, not shown in Table 1, are relatively low (∼2 Å), which demonstrates that the three disulfide bonds provide structural rigidity to the monomer. The rmsd for monomers of HNP-2 (∼2.5 and 3.0 Å) are higher than other two HNPs (∼2.0 Å), explaining the highest value of rmsd to the HNP-2 dimer. However, the average rmsd data for the dimers are more significant suggesting that the quaternary structures of the HNPs are maintained, while the structural behavior for each HNP is different since the same initial structure is used for the three HNPs in MD simulations (Table 1). The IIP values between Mo_A and Mo_B (IMo) shown in Table 1 indicate that the interfaces between the monomers are stabilized because of a minimum rearrangement of the structures of the HNPs. This can be characterized by the variation of energies between the beginning and the end of the simulations, that are lower for HNP-1 and HNP-2 (∼ -60 kcal‚mol-1) than

Figure 5. Electrostatic properties of HNPs at the water/n-hexane interface at the end of the simulations. Potential isocontours are shown at +1(kT/e) (blue) and -1(kT/e) (red) obtained with solute and solvent permittivities equal to 2 and 78.5, respectively. The different views of isocontours are identified by a, b, and c. (a) Lateral view of the systems in the zx plane. (b) The HNP basket shape into water slab. (c) The HNP basket shape into n-hexane slab.

for HNP-3 (∼ -89 kcal‚mol-1), indicating that HNP-3 reaches a higher structural stability than the other HNPs. HNP-2 has one residue less than HNP-1, the HNP-2 average IMo is higher than that of HNP-1, -124 > -132 kcal‚mol-1. In the case of HNP-3, IMo is the most attractive among the HNPs, around -165 kcal‚mol-1, because the Asp residue in the second position in the aligned sequence (Figure 1) is ionized. In the isocontours shown to HNP-3 in Figure 5, parts a and b, the Asp residues (in red) enhance the electrostatic interactions of the peptide with other ionized neighboring residues (in blue). It is necessary to consider that HNP-1 contains a nonpolar Ala in the second position, whereas HNP-2 does not contain any residue in this position, and HNP-3 has a polar Asp residue, as shown in Figure 1. The presence or not of residues in the second position in the primary sequence of the HNPs markedly changes the inter- and intramonomer interactions. These different interactions cause changes in some local residue conformations and in their neighborhoods as shown in Figure 6 for each of the HNPs; this will be argued in details in the next paragraphs. The comparison among HNPs structures evidence structural differences that can be followed by the hydrogen bonds net as showed in Figure 6 and quantified in the Table 2. These hydrogen bonds are important to maintenance of the dimeric structure of the HNPs at the interface. The different dimeric structures of the HNPs at the interface can characterize their different activities. The relationship between interaction and structure is also observed by the different isocontours and structures of HNPs in Figure 5, parts a and b. It has been suggested that the distribution patterns of hydrophobic and charged residues in human β-de-

Human Defensins Interactions with Cell Membranes

Figure 6. Snapshots of the HNPs structures at the end of the simulations. The water molecules are displayed in a space-fill mode, whereas the n-hexane molecules are shown in the light gray stick mode. The HBs involving the side chain and the N- and C-termini are represented by red dots, and the HB between backbone groups, carbonyl, and amine, >CdO‚‚‚H-NN-H‚‚‚OdC