Correction to “Conformation and Aggregation of LKα14 Peptide in Bulk Water and at the Air/Water Interface”

Abstract

We have found out that the peptide bonds in the LKα14 peptide were able to switch to the cis conformation (Figure 1) (Figure Presented) in the replica exchange molecular dynamics (REMD) simulation in bulk water. The results from bulk water REMD simulation were presented in Figures 3 and 4 of our original manuscript.1 These unphysical transitions took place at elevated temperatures in the REMD simulation, which spanned a range from 298 K up to 520 K. Since such unphysical transitions to the cis conformation alter the conformational phase space of the peptide, we repeated the corresponding REMD simulation by restraining the peptide dihedral angle (ω) with an additional potential of the following form Vdihr(ω) = {1/2κdihr (1ω - ω0| - Δ ω)2 for |ω - ω0| > Δ ω 0 for |ω - ω0| ≤ Δ ω where ω ∈ [0,2π], ω0 = π, Δ ω = 4π/18, and κdihr = 1000 kj/mol/rad2. This form has the advantage that the GROMOS 54A7 force field is unperturbed unless the dihedral angle diverges more than Δω from ω0, which does not take place at the target temperature of 298 K in standard molecular dynamics simulations. In order to benefit from GPU, for the current REMD simulation, we have used GROMACS 2018.4. The Verlet cutoff scheme is used with a cutoff value of 1.4 nm for both van der Waals and Coulomb interactions. We have used Parrinello-Rahman, with τp = 5, for pressure coupling. Forty replicas are used that cover a temperature range of 298-520 K, and the simulation was 1 μs long. All other simulation parameters are kept identical to the original manuscript. The DSSP2 analysis of the corrected REMD simulation is shown in Figure 2 for 5 selected replicas out of 40. In all replicas, the LKα14 peptide explores different conformations. The REMD simulation has an average replica exchange probability of 0.23. (Figure Presented). The DSSP analysis based on temperatures is also shown for five representative cases in Figure 3. In our original manuscript, at 298 K, α-helix- and β-hairpin-like structures were observed with almost equal probability. Current results show a conformational phase space which is heavily populated by the β-hairpin conformation. We used hierarchical clustering3 to extract the dominant conformations in bulk water by using the last 0.75 μs at 298 K. The distance root-mean-square deviation (DRMSD) of the conformations is used as the distance metric for clustering. The complete linkage method is used with a distance cutoff value of (Figure Presented) 0.433 nm. This yields a total of 14 clusters, where the top 3 clusters make up approximately 95% of all possible conformations, as shown in the histogram in Figure 4a. (Figure Presented). Snapshots for the cluster centers are also shown in Figure 4b. The β-hairpin conformation clearly dominates the conformational phase space. However, the molecule also retains the α-helix and Trp-cage-like conformations. Similar to the original manuscript, we have calculated the solvent accessible surface area (SASA) for the hydrophobic leucine side chains, the Coulomb energy for the lysine side chains, and the number of backbone-backbone hydrogen bonds for the top three clusters (Figure 5). The results based on these three metrics are identical to our original manuscript. Briefly, the β-hairpin structure has the advantage of minimizing (Figure Presented) the Coulomb repulsion due to lysine side chains; the α-helix displays the largest number of backbone-backbone hydrogen bonds; and, finally, the Trp-cage structure lowers the SASA by forming a hydrophobic core. The results from the renewed REMD simulation clearly demonstrate that the isolated LKα14 molecule in bulk water has a strong preference for the β-hairpin structure. At macroscopic (e.g., vacuum/water) or molecular (e.g., neighboring molecules) interfaces, however, the molecule solely adopts the α-helix conformation due to its secondary amphiphilic character. We have also checked the interface simulation results for the unphysical transitions to the cis conformation and verified that they are not influenced by this error. In this regard, the conformational transition to the α-helix structure when the molecule is in contact with macroscopic (e.g., vacuum/water) or microscopic (e.g., neighboring molecules hydrophobic patches) interfaces is even more remarkable. These results are in agreement with the conformation selection and population shift view.4,5 We sincerely apologize to the readers for any inconvenience caused by this error.