Spontaneous formation of detergent micelles around the outer membrane protein OmpX

Abstract

The structure and flexibility of the outer membrane protein X (OmpX) in a water-detergent solution and in pure water are investigated by molecular dynamics simulations on the 100-ns timescale and compared with NMR data. The simulations allow for an unbiased determination of the structure of detergent micelles and the protein-detergent mixed micelle. The short-chain lipid dihexanoylphosphatidylcholine, as a detergent, aggregates into pure micelles of approximately 18 molecules, or alternatively, it binds to the protein surface. The detergent binds in the form of a monolayer ring around the hydrophobic beta-barrel of OmpX rather than in a micellar-like oblate; approximately 40 dihexanoylphosphatidylcholine lipids are sufficient for an effective suppression of water from the surface of the beta-barrel region. The phospholipids bind also on the extracellular, protruding beta-sheet. Here, polar interactions between charged amino acids and phosphatidylcholine headgroups act as condensation seed for detergent micelle formation. The polar protein surface remains accessible to water molecules. In total, approximately 90-100 detergent molecules associate within the protein-detergent mixed micelle, in agreement with experimental estimates. The simulation results indicate that OmpX is not a water pore and support the proposed role of the protruding beta-sheet as a "fishing rod".

FIGURE 1

Root mean-square deviation of OmpX (C_α atoms) in water environment. The rmsd is given separately for the membrane-spanning β-barrel domain (thin black line), as well as for the protruding extracellular β-sheet (shaded line).

FIGURE 2

C_α atom rmsf of OmpX in a water environment. The residue-averaged rmsf is shown for four different time windows (0.5 ns, dashed gray; 1 ns, solid black; 5 ns. dashed black; and 10 ns, solid gray line) averaged over the simulation (A). (B) The ratio between the longer time windows and the shortest one (0.5 ns). (C) Comparison of fluctuations between the simulations of OmpX in a water environment (dotted line) and in a DHPC micelle (thin black line), with rmsf values calculated from the x-ray B-factors (PDB entries 1QJ8, dotted gray, and 1QJ9, solid gray line) and as obtained from the NMR ensemble (thick black line, PDB entry 1Q9F). The MD fluctuations are computed over the final 60 ns of the respective simulations (W2 and M1).

FIGURE 3

(A) Pore radius profile for OmpX in water. Shown is the average (thick black line, taken over 4000 snapshots of the final 80 ns of simulation W2) as well as 1 σ and 2 σ standard deviations (gray shaded regions) compared with the pore radius computed on the OmpX x-ray structure (dashed line). (B) Position of selected water molecules along the axis of the β-barrel of OmpX (W systems) as a function of simulation time. (C) Snapshot after 50 ns (system W2) with bound water molecules.

FIGURE 4

(A) Aggregation simulation system. Shown is the periodic simulation system with the OmpX (orange) solvated in a box of DHPC lipids at a concentration of 250 mM (system M1). Water molecules are not shown. (B) Crystal structure of OmpX (Vogt and Schulz, 1999) with notation of strands, loops, turns, and specific amino acids showing NOEs to the methyl groups of the polar lipid heads (Fernandez et al., 2002).

FIGURE 5

Number of lipids in contact with OmpX at various protein-lipid distances d ranging from 3.5 Å to 5.5 Å as a function of simulation time (A) and lifetime of contacts at d ≤ 5.5 Å (B) after smoothing the distances with a Gaussian fit with width σ = 800 ps (final 60 ns of simulation M1 considered). The given lipid-protein contact lifetime τ is calculated by fitting the lifetime distribution N(t) ∼ t/τ² exp – t/τ (solid line) to the lifetime of those lipids which left the protein surface within the simulation time span (gray). Lipids persisting at the end of the simulation are colored black.

FIGURE 6

Solvent-accessible surface. (A) Hydrophobic SAS as a function of M1 simulation time given separately for the OmpX (shaded line) and the DHPC lipids (dotted line). (B) Hydrophobic SAS as a function of residue number averaged over the first nanosecond of simulation time (black curve) and its difference (dotted line) to the last nanosecond of simulation M1 (shaded area). The region of the protruding β-sheet is highlighted by black bars. (C) Residual hydrophilic solvent accessible surface. Data shown in B and C are smoothed by using a Gaussian fit with a width of two amino acids.

FIGURE 7

Snapshot of DHPC-OmpX micelle after 80 ns (system M2) from three different perspectives (side views A and C, and view from the periplasmic side, E). Water molecules are not shown, for clarity. Eighty-two lipids are directly or indirectly aggregated on the protein surface. The polar heads of the DHPC molecules are colored magenta, the hydrophobic tails yellow. OmpX is depicted in surface representation with hydrophobic residues colored light green, aromatic residues orange, polar residues light blue, and positively and negatively charged residues blue and red, respectively. (B and D) Snapshots of the water atom distribution in a distance of 3.5 Å from the protein as seen from the views of panels A and C, respectively. Shown in F is a close-up view of the protruding β-sheet, including NOE distance constraints (green) and hydrogen bonds between the upper part of the β₄ and β₅ sheets. The respective NOE distances as computed from MD simulations W1 and M1 are shown in G.

FIGURE 8

Water distribution around the OmpX protein for three different time windows of simulation M1. Shown is the distance at which the cumulative radial distribution function of water oxygens around the backbone nitrogen atoms reaches a value of 1. The region of the protruding β-sheet is highlighted by dark shaded bars.

FIGURE 9

(A) Lipid cluster sizes and their color-coded frequency of occurrence (fo) as a function of simulation time for runs M1–M3. (B) Time-averaged distribution of cluster sizes. The weighted distributions are averaged over the final 60 ns (simulation M1) or the final 5 ns (M2 and M3), respectively. The gray dashed line indicates the aggregation number of DHPC molecules in pure lipid micelles as determined by small-angle neutron scattering experiments (Lin et al., 1986). I, II, and III are peaks of the protein-detergent mixed micelle.

FIGURE 10

Schematic picture of a DHPC micelle obtained from simulation results. The lengths of the axes correspond to the radii of gyration of the acyl chains and DHPC headgroups.

FIGURE 11

Contact map between backbone nitrogen atoms (indole nitrogens of Trp⁷⁶ and Trp¹⁴⁰; upper two rows) and both polar head methyl groups of DHPC (A) and the lipid tail CH_n groups (B) for simulation M1. Distances <5.5 Å are colored. The gray bars emphasize the loops and turns. Experimentally observed NOEs between amide backbone protons and polar head methyl groups of DHPC are marked by blue arrows (A). The lower graph shows the total number of the respective contacts below 5.5 Å as a function of simulation time. Additionally shown in A and B are pink bars with color intensities according to the measured paramagnetic relaxation enhancement ε using the paramagnetic relaxation probe Gd(DOTA) (A) or 5-DSA (Hilty et al., 2004) (B). White color corresponds to ε = 0, dark pink to ε = ε_max (Hilty et al., 2004).

FIGURE 12

Contact time (upper panels) between the backbone nitrogen atoms and the five different CH_n groups of the DHPC fatty acyl chains (A, light gray to black, with decreasing distance from the polar headgroup, in simulation M1), and between the backbone nitrogen atoms and the polar head methyl groups (B). Only contacts with distances <5.5 Å and with a lifetime of at least 300 ps are considered. The contact time is given as the fraction of the considered total simulation time (final 60 ns of simulation M1). The lower panels show the approximated NOE intensity by computing averages for the respective CH_n-amide backbone proton distances d_ij. The residues with experimentally observed NOEs between the DHPC choline head methyl groups and backbone amide protons (gray) or indole protons (black) have been marked by arrows.

FIGURE 13

Protein-detergent binding models. Detergent-binding in a prolate micelle (A) and in an oblate micelle (B) (le Maire et al., 2000).