Structural Architecture of the Hexameric AMP Channel.

The structure of assembled DCD solved at 2.5-Å resolution exhibits a channel architecture comprising a hexameric bundle formed by elongated α-helices, which adopts overall dimensions of ∼8 × 4 nm (Fig. 1A). Channel formation involves trimerization of antiparallel peptide dimers resulting in a firmly enclosed channel structure. Each monomer presents two distinct interfaces to neighboring subunits (Fig. 1A). The extended interface, displaying a contact surface of 930 Ų, is mainly formed by salt bridges, and the second interface covers 520 Ų (IF1) and is primarily stabilized by Zn²⁺ ions that intercalate between the two helices (Fig. 1A and Fig. S1D). The zinc ions in DCD are coordinated by N- and C- terminal residues of dimerizing peptides (residues involved: Glu5, Glu9, His38′, and Asp42′) (Fig. 1 A and B). They are attached to the inner wall of the channel and substantially change the overall charge of the assembly (from −12e to neutral). In addition, they modify the local charge distribution especially at the entrance of the channel.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

Crystal structure and surface characteristics of the human dermcidin channel. (A) X-ray structure of the hexameric DCD channel shown in cartoon representation from the side and top (Middle and Right), and as surface representation (Left). The different orientations of the individual peptides relative to the membrane normal are marked in orange and dark blue, and termini are marked (NT, N terminus; CT, C terminus). Arrows combined with tilt angle and axes give the relative orientation. Residues involved in Zn binding are shown in stick representation and Zn ions are marked in gray. The symmetry axis of the channel is marked with C2 (for the side view) and C3 (for the top view). Two interfaces of different surface area are formed after trimerization and named IF1 and IF2. (B) Close-up into the Zn-binding site S1. Four residues (Glu5 and Asp11 from one peptide and Asp41 and His38 from the second) form each Zn-binding site. The distance between the Zn ions is marked by arrows. (C) Electrostatic surface representation of the channel with two monomers marked in ribbon representation. The channel comprises five alternating patches of elongated negative (red) and ring-like positive (blue) charge. (D) Side view of DCD (hydrophobic residues in magenta). (E) Ribbon model of DCD. The pore diameter is represented by spheres. The lateral entry points are marked with circles. (F) Hydrophilic residues on the trimeric interface (negatively charged residues in red, positively charged residues in blue, polar residues in green). Nonpolar residues are shown in white. (G) The hydrophilic channel interior. For clarity, the front dimer is omitted; colors as in F. (H) Channel radius along the pore axis.

DCD shows a unique distribution of charges, reflecting peptide oligomers that occur both in soluble and membrane-bound forms. The entire hexameric channel comprises 96 ionizable residues, which are all oriented toward the channel interior (Fig. 1 C, F, and G). This enormous charge density, which is not completely shielded to the exterior, is likely to contribute to the relatively high aqueous solubility of DCD. The inner space of the peptide channel has an apparent separation into five radially symmetric charge girdles I/II/III/II/I (Fig. 1C). Residues facing the acyl chain region of the membrane are exclusively hydrophobic (Ala, Val, Leu) (Fig. 1D and Fig. S1A). However, no aromatic residues are found that are commonly considered to be important for the lateral adjustment of proteins in membranes (20, 21). The channel diameter is not homogeneous and varies along the y axis with two rather narrow entry sites, followed by a widened interior with windowlike eyelets in the IF1 interface (Fig. 1 A, E, and H). The six lateral openings have a diameter of ∼1 nm and are surrounded by small amino acids as well as positively charged residues that may have an influence on the selection of ion entry. The distance between two adjacent eyelets is 2.5 (same plane) and 3 nm (opposite plane), respectively, roughly corresponding to the width of the membrane hydrophobic core.

Interaction of DCD with Lipid Bilayers.

To investigate the interaction of DCD with bilayers, we conducted ss-NMR spectroscopy experiments. The structural preferences of DCD were investigated after labeling of the peptide with ¹⁵N at the Gly22 position and ²H₃-Ala at position 25, its reconstitution into oriented lipid bilayers and investigation by proton-decoupled ¹⁵N ss-NMR spectroscopy. DCD labeled with ¹⁵N at Gly22 exhibits a ¹⁵N chemical shift of (62 ± 2) ppm in oriented 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoethanolamine/1-palmitoyl-2-oleoyl-sn-glycero-3-phospho(1'-rac-glycerol) (POPE/POPG, 3:1) membranes both in the absence or presence of a 10-fold excess of Zn²⁺ (Fig. 2 A and B). This indicates an in-planar alignment of the major population of the labeled peptide domain (22), in agreement with previous oriented circular dichroism data (16). The chemical shifts are within experimental error identical in POPE/POPG, 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine, and DPhPC/cholesterol 9:1 membranes and also independent of zinc ions (ZnCl₂ salt) and the sample preparation protocol. The ²H ss-NMR spectrum of the ²H₃-Ala25 labeled site exhibits an ²H quadrupolar splitting of (45 ± 3) kHz (Fig. 2C) and further restricts the tilt and pitch angle (23). A detailed topological analysis is shown by the red and black traces in Fig. S2 for the ¹⁵N chemical shift and the ²H quadrupolar splitting, respectively (see ref. 23). Both NMR parameters agree when their corresponding traces intersect leading to a set of tilt/pitch angular pairs in the proximity of (90°/90°), which would represent an ideal alignment of the amphipathic helix with the membrane interface (Fig. 2B). In agreement, the ²H spectra of the fatty acyl chains show pronounced membrane disordering by the presence of DCD (Fig. S3).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Electrophysiology and ss-NMR of DCD in the presence and absence of zinc ions. (A and B) Proton-decoupled ¹⁵N and (C) ²H ss-NMR spectra of 2 mol % [¹⁵N-Gly22,²H₃-Ala25]-DCD in oriented POPE/POPG without ZnCl₂ (A) and with fivefold molar excess of ZnCl₂ relative to the peptide concentration (B and C). ¹⁵NH₄Cl (40.0 ppm) and ²H₂O (0 Hz) were used as external references and the spectra processed with an exponential line broadening of 50 (A and B) and 500 Hz (C), respectively. The gray line in C shows a simulated spectrum arising from a 3° Gaussian distribution for the peptide alignment and a spectral line width of 2 kHz. (D) Current trace and point amplitude histogram representing DCD activity in 1 M NaCl, 5 mM Hepes, pH 7.1 recorded at a holding potential of +100 mV. Protein activity was only rarely observed under these conditions. (E and F) Current traces and point amplitude histograms of DCD-1L in 1 M NaCl, 3 mM ZnCl₂, 5 mM Hepes, pH 7.1 recorded at a holding potential of +100 mV; (E) 3.3 µM DCD-1L was added resulting in one defined conductance state. In (F) 7.9 µM DCD-1L was used and two distinct conductance levels each of about 80 pS were monitored, suggesting the insertion of two channels. (G) An event histogram of the conductance levels (n = 1,009) shows a mean conductance of G = (81 ± 14) pS. (H) Open lifetime analysis with τ = (4.4 ± 0.2) ms.

Functional Properties of the Channel Highlight the Importance of Zinc as Cofactor.

The solid-state NMR data represent probably the lowest energy state after extensive equilibration of the major population of the peptide and, in particular, in the absence of a membrane potential. In contrast, electrophysiological recordings monitor a population that may occur as a minor, but functionally relevant, species under the application of a transmembrane voltage. To obtain functional evidence for the activity of DCD channels in the membrane, we collected conductivity data in planar lipid bilayers. First, membranes were prepared in 1 M NaCl, 5 mM Hepes, pH 7.1 in the absence of Zn²⁺. After the addition of DCD in micromolar concentrations, two out of the six membranes showed weak activity of the peptide (Fig. 2D). Altogether, only 20 current steps were observed, from which a mean conductance of G = (31 ± 8) pS was evaluated.

By contrast, in the presence of Zn²⁺, the addition of DCD at concentrations of 850 nM or higher resulted in current fluctuations for every membrane preparation, which eventually led to rupture of the membrane (Fig. 2 E and F). These concentrations are similar to DCD concentrations used to efficiently remove bacteria in antimicrobial assays. Evaluation of 1,009 events revealed a mean conductance of G = (81 ± 14) pS (Fig. 2G) and the mean open lifetime was determined to be τ = (4.4 ± 0.2) ms (Fig. 2H). This shows that, in the presence of zinc ions, specific and individual channels with a defined conductance are formed in the membrane bilayer. Neither the membrane with DCD alone nor the addition of only Zn²⁺ gave rise to the occurrence of individual channels in the electrophysiology experiments. However, we found that the polarity of the electric field had a significant impact on the efficiency of DCD-Zn²⁺ insertion.

To further investigate the importance of the specific Zn²⁺ interactions and their impact on the active membrane-bound form of DCD, we mutated His38, one of the main interaction partners of DCD with Zn²⁺, to alanine and studied this mutant through electrophysiology. The single mutation was sufficient to abolish channel formation in membranes (Fig. S4). This finding is most compatible with the notion that the assembled structure, which is characterized by specific Zn²⁺-conferred peptide–peptide interactions, is the most prominent membrane-disruptive form. It is important to note that the current fluctuations seen for DCD-Zn²⁺ in electrophysiology are representative of a small number of channels efficiently conducting ions across membranes, whereas the ss-NMR measurements observe the major peptide fraction only.

Molecular Dynamics Simulations Reveal Unexpected Ion-Transfer Pathways.

To obtain a detailed mechanistic picture of DCD acting on membranes, we performed extensive atomistic MD simulations of the DCD channel in negatively charged, bacterial-like phosphatidyl ethanolamine/ phosphatidyl glycerol (POPE/ POPG, 3:1) bilayers (24), and carried out computational electrophysiology simulations to monitor its ion permeation properties (Fig. S5). The DCD assembly including Zn²⁺ exhibited a high level of structural stability in membranes (C-α root-mean-square deviation ∼2 Å; Fig. S6). In aqueous solution, the aggregate was found to be stable, but showed a markedly raised structural variability (Fig. S6B). Consistent with our experimental data, removal of the Zn²⁺ ions from the membrane-inserted assembly structure resulted in a substantial distortion of the oligomer, compromising its membrane-channel character (Fig. S6D). These findings agree well with previous NMR studies, in which both a membrane-mimetic solvent (TFE-²H₃) and Zn²⁺ were shown to induce oligomerization of DCD (16).

After initially positioning the complex normal to the bilayer plane, the assembly adopted tilted configurations relative to the membrane normal within 250 ns simulated time (γ∼30°; Fig. 3A). The inclination was found to compensate for the hydrophobic mismatch between the bilayer and the hydrophobic region on the outer surface of DCD, and its magnitude depended on the length of the surrounding lipids (Fig. S7A).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Computational electrophysiology and molecular dynamics simulations of the channel. (A) Evolution of DCD tilt in three 250-ns single-bilayer MD simulations in POPE/POPG(3:1). (B) Water distribution (red/white spheres) in and around the DCD channel in the bilayer (gray); view along the channel axis. (C) Pathway of ion permeation. One complete Cl⁻ permeation event is shown, with red spheres representing snapshots of Cl⁻ position. The brown arrows indicate the direction of permeation. (D) The I–V relationships from computational electrophysiology simulations. Each data point was obtained from a 100-ns MD trajectory by monitoring the number of permeation events (and thus the current) and the overall transmembrane potential during this time window. Blue and green circles are from simulations with 0.15 M and 1 M NaCl in solution, respectively. Dashed lines: linear fit of the conductance data, with the slope (and the asymptotic SE) showing overall conductance.

During the simulations, the interior of the oligomer rapidly filled with water and formed a permanent water channel across the membrane (Fig. 3 B and C). Water also partitioned into the three hydrophilic crevices at the trimer interfaces, such that a high osmotic water permeability coefficient of (327 ± 13) × 10⁻¹⁴ cm³⋅s⁻¹ was determined that exceeds that of aquaporin channels by up to 50-fold (25). To investigate the conductivity of DCD for ions, we used computational electrophysiology to model biologically realistic electrochemical gradients across membranes (26). At 1 M NaCl concentration, we obtained a DCD channel conductance of (108 ± 11) pS, in excellent agreement with our experimental data (Figs. 2G and 3D). In simulations at a salt concentration of 150 mM, a total single-channel conductance of (50 ± 7) pS was obtained (Fig. 3D). In all of our simulations, DCD showed pronounced anion selectivity (Fig. S7B).

The remarkable agreement between computational and experimental data strongly suggests that the hexameric crystal structure is identical to the functional state in the membranes used for electrophysiology. Any higher or lower oligomerization state of DCD would very likely lead to markedly different conductance values (SI Materials and Methods and Fig. S8). This interpretation is corroborated by the strong dependence of channel current on zinc, which we observed in both the electrophysiology experiments and MD simulations in membranes, and which corresponds with the abundance and function of the Zn²⁺ binding sites, linking the subunits in the crystal structure.

It is notable that DCD exhibited a unique ion-permeation pathway in the simulations, which offers an explanation for this unexpectedly high conductance (Movies S1 and S2), despite its limited channel cross-section. Through channel tilt, ions are capable of entering sideways into the pore across the eyelets that occur at the trimeric interfaces. This not only shortens the pathway across the channel, but importantly, exploits the increased ion concentration observed at the lipid head groups by enabling these ions to enter the channel directly, and to rapidly traverse the inner pore (Fig. 3C and Movies S1 and S2). Also within the channel, DCD shows an unusual anion traversal mechanism. Most anion transfer steps across the inner section of the pore consist of single ion “hopping” transitions. Near the channel termini, however, anions accumulate to form clusters of three or four ions, most clearly seen at the channel exit. Productive ion translocations exiting the channel usually involve multiion “knock-on” effects, through which individual anions are expelled from this cluster to the bulk solution (Movie S1).

The stabilization of DCD oligomers by a membrane mimetic, seen in earlier NMR studies (16), is also corroborated by molecular dynamics simulations, in which we tested the long-term stability of the DCD assembly in a membrane environment vs. aqueous solution. Fig. S6B shows that the assembled structure with Zn²⁺ shows a substantially raised stability in bilayers relative to solution, which indicates a preference of the hexameric assembly for the membrane environment.