Structural Determinants of Peptide Nanopore Formation

We have evolved the nanopore-forming macrolittin peptides from the bee venom peptide melittin using successive generations of synthetic molecular evolution. Despite their sequence similarity to the broadly membrane permeabilizing cytolytic melittin, the macrolittins have potent membrane selectivity. They form nanopores in synthetic bilayers made from 1-palmitoyl, 2-oleoyl-phosphatidylcholine (POPC) at extremely low peptide concentrations and yet have essentially no cytolytic activity against any cell membrane, even at high concentration. Here, we explore the structural determinants of macrolittin nanopore stability in POPC bilayers using atomistic molecular dynamics simulations and experiments on macrolittins and single-site variants. Simulations of macrolittin nanopores in POPC bilayers show that they are stabilized by an extensive, cooperative hydrogen bond network comprised of the many charged and polar side chains interacting with each other via bridges of water molecules and lipid headgroups. Lipid molecules with unusual conformations participate in the H-bond network and are an integral part of the nanopore structure. To explore the role of this H-bond network on membrane selectivity, we swapped three critical polar residues with the nonpolar residues found in melittin. All variants have potency, membrane selectivity, and cytotoxicity that were intermediate between a cytotoxic melittin variant called MelP5 and the macrolittins. Simulations showed that the variants had less organized H-bond networks of waters and lipids with unusual structures. The membrane-spanning, cooperative H-bond network is a critical determinant of macrolittin nanopore stability and membrane selectivity. The results described here will help guide the future design and optimization of peptide nanopore-based applications.


Figure S1
. Time courses of the number of direct and water-mediated H-bonding of M70 and M159 during the production runs of the simulations.We present the total number of direct H-bonds between peptide sidechains (self and non-self), and water-mediated bridges between sidechains with up to three H-bonded water molecules per bridge.All H-bonds sampled at least once, regardless of the average H-bond occupancies, are included in these time series.The origin of time is the start of the production runs without any constraints.Black and blue profiles represent time courses computing from the main and repeat simulations, respectively.(A, B) Time series computed for macrolittin M70 (panel A) and M159 (panel B).Ellipticity were converted to mean residues ellipticity.Maximum helicity for this family of peptides corresponds to about -20 mdeg dmol -1 cm -1 .As shown in Fig, S4, the macrolittins and their variants bind well to these lipid vesicles with 85-99% bound under typical conditions.The only exception is MelP5 I17Q, which does not bind measurably to POPC+30% cholesterol.These data show that the macrolittins are their variants are α-helical when bound to these various bilayers.The only exception is MelP5 I17Q which is not helical in the presence of POPC+30% cholesterol.This agrees with the conclusion that MelP5 I17Q does not bind to cholesterol-containing bilayers.Most importantly, these data show that the M159 E4A and E8V variants bind strongly (Fig. S4) and have helical secondary structure in POPC bilayers.Their decrease in nanopore forming activity is due to a decrease in nanopore stabilizing interactions.For clarity, we include only H-bonds with a minimum occupancy of 30%.

Figure S16
Results of the repeat simulation of macrolittin M159 E8V.These images display the average number of water molecules in the H-bonded bridges (top) and the average % occupancy of each direct or water bridged H-bond (bottom).Values were computed from the last ~200ns of the repeat simulation of macrolittin M159 E8V.
For clarity, we include only H-bonds with a minimum occupancy of 30%.

Figure S17
. Sliding window hydropathy plot for the classical single span membrane protein glycophorin A is shown in black.The TM helix of GPA has positive hydrophobicity.In red, the TM helix of glycophorin A is replaced by the sequence of macrolittin M70, with the glutamates protonated, equivalent to pH << 4. In blue, the TM helix of glycophorin A is replaced by the sequence of macrolittin M70, with the glutamates deprotonated, equivalent to neutral pH.In both cases the sequence of M70 is not hydrophobic enough to be identified as a transmembrane helix.

Figure S2 .
Figure S2.Results of repeat simulation of macrolittin M70.(A) Cut-away views showing water molecules interacting in the M70 pore.Peptides are shown as yellow ribbons, and water molecules are shown as van der Waals spheres with oxygen and H atoms colored red and white, respectively.The W19 sidechain of each peptide is shown as van der Waals spheres; for clarity, H atoms of the W19 sidechains are not shown.(B) Cut-away view illustrating lipid phosphate groups.(C, D) Illustration of lipid molecules closely associated with the M70 pore.In panel C, the pore is viewed from the membrane surface, while panel D shows a lateral view of the pore.Note that, for clarity, lipids of only one membrane leaflet are shown; see panel B for the phosphate groups of both membrane leaflets.(E, F) H-bond graph computed from the last ~200ns of the repeat simulation of macrolittin M70.For clarity, we include only H-bonds with a minimum occupancy of 30%.Detailed information on the average occupancy and average number of water molecules for each edge of the graph is presented in FigureS4.(F) Summary of the number of H-bonds counted based on the H-bond graph presented in panel E.

Figure S9 .
Figure S9.Circular dichroism spectra of macrolittins and variants.A. 25 µM of each peptide in buffer was characterized at room temperature using a JASCO 810 CD spectrometer.C-D.Circular dichroism spectra were collected in samples containing 25 µM peptide plus 1 mM of lipid vesicles made from POPC (B), POPC+30% cholesterol (C) and diC20:1PC (D).Ellipticity were converted to mean residues ellipticity.Maximum helicity for this family of peptides corresponds to about -20 mdeg dmol -1 cm -1 .As shown in Fig,S4, the macrolittins and their variants bind well to these lipid vesicles with 85-99% bound under typical conditions.The only exception is MelP5 I17Q, which does not bind measurably to POPC+30% cholesterol.These data show that the macrolittins are their variants are α-helical when bound to these various bilayers.The only exception is MelP5 I17Q which is not helical in the presence of POPC+30% cholesterol.This agrees with the conclusion that MelP5 I17Q does not bind to cholesterol-containing bilayers.Most importantly, these data show that the M159 E4A and E8V variants bind strongly (Fig.S4) and have helical secondary structure in POPC bilayers.Their decrease in nanopore forming activity is due to a decrease in nanopore stabilizing interactions.

Figure S10 .Figure S11 .Figure S12 .Figure S13 .
Figure S10.Time courses of the number of direct and water-mediated H-bonding of M159 E4A and M159 E8V during the production runs.We present the total number of direct H-bonds between peptide sidechains (self and non-self), and water-mediated bridges between sidechains with up to three H-bonded water molecules per bridge.All H-bonds sampled at least once, regardless of the average H-bond occupancies, are included in these time series.The origin of time is the start of the production runs without any constraints.Black and blue profiles represent time courses computing from the main and repeat simulations, respectively.(A, B) Time series computed for macrolittin M159 E4A (panel A) and M159 E8V (panel B).

Total: 156 A B C D M159 Repeat Simulation E F Figure S3. Results
of repeat simulation of macrolittin M159.(A)Cut-away views showing water molecules interacting in the M159 pore.Peptides are shown as yellow ribbons, and water molecules are shown as van der Waals spheres with oxygen and H atoms colored red and white, respectively.The W19 sidechain of each peptide is shown as van der Waals spheres; for clarity, H atoms of the W19 sidechains are not shown.(B)Cut-awayview illustrating lipid phosphate groups.(C,D)Illustration of lipid molecules closely associated with the M159 pore.In panel C, the pore is viewed from the membrane surface, while panel D shows a lateral view of the pore.Note that, for clarity, lipids of only one membrane leaflet are shown; see panel B for the phosphate groups of both membrane leaflets.(E,F)H-bond graph computed from the last ~200ns of the repeat simulation of macrolittin M159.For clarity, we include only H-bonds with a minimum occupancy of 30%.Detailed information on the average occupancy and average number of water molecules for each edge of the graph is presented in FigureS7.(F) Summary of the number of H-bonds counted based on the H-bond graph presented in panel E. Results of the main simulation of macrolittin M70.These images display the average number of water molecules in the H-bonded bridges (top) and the average % occupancy of each direct or water bridged H-bond (bottom).Values were computed from the last ~200ns of the main simulation of macrolittin M70.For clarity, we include only H-bonds with a minimum occupancy of 30%.Results of the repeat simulation of macrolittin M70.These images display the average number of water molecules in the H-bonded bridges (top) and the average % occupancy of each direct or water bridged Hbond (bottom).Values were computed from the last ~200ns of the repeat simulation of macrolittin M70.For clarity, we include only H-bonds with a minimum occupancy of 30%.Results of the main simulation of macrolittin M159.These images display the average number of water molecules in the H-bonded bridges (top) and the average % occupancy of each direct or water bridged Hbond (bottom).Values were computed from the last ~200ns of the main simulation of macrolittin M159.For clarity, we include only H-bonds with a minimum occupancy of 30%.Results of the repeat simulation of macrolittin M159.These images display the average number of water molecules in the H-bonded bridges (top) and the average % occupancy of each direct or water bridged Hbond (bottom).Values were computed from the last ~200ns of the repeat simulation of macrolittin M159.For clarity, we include only H-bonds with a minimum occupancy of 30%.The measured Kx values are shown in this panel on the left axis.On the right axis, we calculate the fraction of peptide bound to 1 mM lipid vesicles, as used in leakage, fusion and circular dichroism experiments.The macrolittins and their variants bind well to these lipid vesicles with 85-99% bound under typical conditions.The only exception is MelP5 I17Q, which does not bind measurably to POPC+30% cholesterol.
Figure S8.Membrane binding of macrolittins and variants.A-C. 10 µM peptides were titrated with lipid vesicles, and the tryptophan fluorescence spectra were measured.The fluorescence intensity at 333 nm was measured for each lipid concentration and was divided by the intensity measured in the absence of lipid.The intensity increase indicates membrane binding, and the more fraction partition coefficient Kx can be determined from these curves by Eq. 4 main text.Binding was measured to vesicles of 100% POPC (A); 70% POPC and 30% Cholesterol (B); 100% diC20:1-PC (C).D.
These images display the average number of water molecules in the H-bonded bridges (top) and the average % occupancy of each direct or water bridged H-bond (bottom).Values were computed from the last ~200ns of the main simulation of macrolittin M159 E4A.For clarity, we include only H-bonds with a minimum occupancy of 30%.Results of the repeat simulation of macrolittin M159 E4A.These images display the average number of water molecules in the H-bonded bridges (top) and the average % occupancy of each direct or water bridged H-bond (bottom).Values were computed from the last ~200ns of the repeat simulation of macrolittin M159 E4A.For clarity, we include only H-bonds with a minimum occupancy of 30%.Results of the main simulation of macrolittin M159 E8V.These images display the average number of water molecules in the H-bonded bridges (top) and the average % occupancy of each direct or water bridged H-bond (bottom).Values were computed from the last ~200ns of the main simulation of macrolittin M159 E8V.