Implicit solvent models and the energy landscape for aggregation of the amyloidogenic KFFE peptide

This study compares the performance of four implicit solvent models in describing peptide aggregation. The solvent models are the effective energy function-1 (EEF1) and three generalized Born (GB) models: one following the original implementation of Still (GB1), the analytical continuum electrostatics (ACE) potential, and GB with simple switching (GBSW). For each solvent model the first step of aggregation, namely dimerization, is investigated for the KFFE peptide, which is one of the shortest peptides known to form amyloid fibrils in vitro. Using basin-hopping for global optimization and replica exchange molecular dynamics simulations, we conclude that of the four solvent models considered, the EEF1 potential provides the most reliable description for the formation of KFFE amyloid precursors. It produces results that are closest to the experimental findings of a partial beta-strand conformation for the KFFE peptide in solution along with the formation of fibrils exhibiting antiparallel beta-strand structure. The ACE and GB1 potentials also show a significant beta-propensity for the KFFE peptide but fail to produce stable KFFE dimers. The GBSW potential, on the other hand, supports a very stable antiparallel dimer structure, but in a turn rather than a beta conformation.