Date Published: July 20, 2017
Publisher: Public Library of Science
Author(s): Gabor Nagy, Chris Oostenbrink, Jozef Hritz, Yaakov Koby Levy.
The 14-3-3 protein family performs regulatory functions in eukaryotic organisms by binding to a large number of phosphorylated protein partners. Whilst the binding mode of the phosphopeptides within the primary 14-3-3 binding site is well established based on the crystal structures of their complexes, little is known about the binding process itself. We present a computational study of the process by which phosphopeptides bind to the 14-3-3ζ protein. Applying a novel scheme combining Hamiltonian replica exchange molecular dynamics and distancefield restraints allowed us to map and compare the most likely phosphopeptide-binding pathways to the 14-3-3ζ protein. The most important structural changes to the protein and peptides involved in the binding process were identified. In order to bind phosphopeptides to the primary interaction site, the 14-3-3ζ adopted a newly found wide-opened conformation. Based on our findings we additionally propose a secondary interaction site on the inner surface of the 14-3-3ζ dimer, and a direct interference on the binding process by the flexible C-terminal tail. A minimalistic model was designed to allow for the efficient calculation of absolute binding affinities. Binding affinities calculated from the potential of mean force along the binding pathway are in line with the available experimental estimates for two of the studied systems.
14-3-3 proteins are important regulatory factors found in all kingdoms of life and are vital for the survival of higher organisms. In mammals, the 14-3-3 family consists of seven isoforms that can be found in large abundance within the brain. The human 14-3-3ζ isoform was selected for this project because of its high biological relevance. 14-3-3 proteins function mainly as dimers, which are composed of two 28-kDa monomers that are both capable of binding phosphorylated serine (pS) and threonine (pT) motifs in other proteins. Crystal structures of all seven mammalian homodimers are now available and show that each monomer is composed of nine α-helices, arranged in an antiparallel fashion. The helices form an amphipathic groove that mediates pS and pT target binding . Most 14-3-3 targets have two phosphoserine/threonine-containing motifs with a consensus sequence RSXpSXP (mode I) or RX[FY]XpSXP (mode II) , representing the optimal recognition sites for 14-3-3. Upon binding to these sites 14-3-3 proteins induce conformational changes in their target protein, (and ‘finish the job’) when phosphorylation alone may lack the power to drive the necessary allosteric changes for modulating the activity of an intracellular protein. Owing to their dimeric nature, 14-3-3 proteins are capable of distinguishing between non-, single- and double- phosphorylated binding protein partners; in this sense 14-3-3 proteins are considered to act as coupled binary devices . Recently, we have presented an approach based on experimental 31P NMR spectroscopy which revealed that a double-phosphorylated protein can be complexed with the 14-3-3ζ dimer in a much more dynamic fashion than was originally thought. In addition to the traditionally considered single partner with two phosphorylation sites occupying the individual binding cavities within the 14-3–3ζ dimer, two more major binding modes were confirmed . All that is currently known about the structural features of the phosphopeptide binding to 14-3-3 proteins originates from the available crystal structures of 14-3-3 proteins in apo and holo state. Comparison of the various 14-3-3 crystal structures also revealed the different width of the main peptide binding groove, thus suggesting a dynamic opening process .
We studied the binding of phosphorylated peptides to the 14-3-3ζ protein by constructing two different models of 14-3-3ζ, one containing a full length 14-3-3ζ dimer (dim) and one including only a truncated 14-3-3ζ monomer (tmon) without the flexible C-terminal stretch (tail, aa. 230–245). The proper simulation of a full length 14-3-3ζ dimer requires a large simulation box in order to avoid artificial periodic effects arising from the opening motion of 14-3-3ζ and the flexible C-terminal regions. Furthermore, the presence of the flexible C-terminal stretch also significantly slowed down the convergence of calculated free energies, as is demonstrated by the PMF calculations (see below). Four different phosphopeptide fragments are used as models for this study which were derived from the diphosphorylated PKC-ε and C-RAF kinase binding sites for 14-3-3ζ, and are referred to as the head and tail fragments of peptides 1 and 2, respectively (p1h, p1t, p2h, p2t), based on their location in the full protein sequence. These phosphopeptide fragments were chosen because their crystal structure bound to 14-3-3ζ was available and their binding affinities were previously measured [8–10].
We explored the phosphopeptide binding pathways of the 14-3-3ζ protein through molecular dynamics simulations of four phosphopeptide fragments derived from PKC-ε and C-RAF kinase. The pathways were explored by a novel Hamiltonian replica exchange molecular dynamics method with incorporated distancefield restraints (DF/HRE-MD). The eight DF/HRE-MD simulations (4 dimeric and 4 truncated monomeric complexes) combined corresponded to more than 6.7 μs of enhanced-sampling simulation time, allowing for the unbiased determination of the most probable binding/unbinding pathways, the corresponding structural changes of the phosphopeptides and 14-3-3ζ, as well as the PMF profiles along the binding pathway.
All molecular dynamics (MD) and Hamiltonian replica exchange MD (HRE-MD) simulations were performed using the GROMOS11 software package . The structure preparation and analysis of the simulation trajectories was based on the GROMOS++ analysis package . The PyMol molecular graphics system version 126.96.36.199  was used for visualization and calculation of the electrostatic surface potentials based on an adaptive Poisson-Boltzmann solver  as implemented in the PyMol software. Changes in the protein and peptide secondary structure were followed by the DISICL algorithm .