Protein-Peptide Interaction (together with Dr. Christian Freund at FMP Berlin): Biological function of adaptor domains controlled by the activity of peptidyl prolyl cis/trans isomerases
Versatile adaptor domains bind to proline-rich sequences involved in intracellular signalling events and guarantee the assembly of many signal promoting multi-protein complexes. The proline-rich ligands adopt a helical conformation upon binding to the adaptor domains and require the proline bonds to be in a trans conformation. Peptidyl prolyl cis/trans isomerases (PPIases) catalyse the interconversion of proline peptide bonds. Peptides will be designed experimentally that interact with the proline-rich cognition domains fynSH3 or GYF and also serve as substrates for the peptidyl prolyl cis/trans isomerases (PPIases) cyclophilin A or FKBP-12. A phage display system will be employed where randomized peptides are expressed as phage geneIII-fusion proteins. Peptide fusions that are selected after several rounds of panning will be sequenced and their binding properties will be investigated in vitro. The structure of the peptides in the bound and unbound state will be studied by heteronuclear NMR spectroscopy building on the published structures of peptide complexes of fynSH3 and of the GYF domain (Freund02). Furthermore, selected peptides will be introduced into living cells and their inhibitory potential will be evaluated. Peptides that bind to the adaptor domains in a PPIase dependent fashion will then be theoretically evaluated in more detail by docking techniques and by Molecular Dynamics and Monte Carlo simulations to identify their binding modes structurally and dynamically. By using NMR and molecular simulations, the conformational equilibiria of promising peptides will be studied in solution to investigate whether the peptides bind preformed to the adaptor domains or whether they adopt their poly-proline helix upon binding.
Protein-Protein Docking of electron transfer proteins (Helms)
Efficient association (and dissociation) is crucial for the association of electron transfer partners in biological redox chains. Examples are the membrane-bound processes of photosynthesis and respiration where cytochrome c acts as an electron carrier between reaction center and cytochrome bc1 complex (photosynthesis) or between cytochrome bc1 complex and cytochrome c oxidase (respiration). During recent years, crystal structures have become available for the isolated binding partners, and, in some cases, also of protein-protein complexes. Only small side chain rearrangments are observed during complexation which may be mechanistically favorable for such electron shuttle proteins. These are, therefore, ideal model systems for rigid body protein-protein docking. The FTDOCK package was first tested for docking complexes of cytochrome c with cytochrome c peroxidase using biochemical distance restraints (Flöck02) and the energy of the docked complexes was evaluated with the molecular mechanics package CHARMM after a short minization. When using a distance-dependent dielectric constant 5 ´ r , the crystal complex had the lowest energy. With this energy model, docked complexes were generated for the complexes of horse heart cytochrome c with cytochrome c oxidase from Paracoccus denitrificans and for cytochrome c552 with cytochrome c oxidase, both from Paracoccus denitrificans. Interestingly, the best complex of horse heart cytochrome c with cytochrome c oxidase was almost identical to a previously published structure for the complex of horse heart cytochrome c with cytochrome c oxidase from bovine that was generated using a different docking package (DOT). This shows the robustness of protein-protein docking techniques that can already be achieved currently in favorable cases.
In a second study, the complex was investigated between cytochrome c2 and reaction center from Rhodobacter sphaeroides (Chandran03). For this systems, crystal structures exist for the isolated proteins as well as for the complex. The intermolecular energies were again evaluated by CHARMM and the solvation energy was computed by solving the Poisson-Boltzmann equation with the UHBD package. Using 10 distance restraints, the crystal structure could reliably be found when the bound conformations of both proteins were used for the docking (ca. 2.0 Å RMSD from crystal complex). The spread between multiple solutions became larger (ca. 3.0 Å RMSD) when the free conformations were used. This is, of course, generally the case in docking studies and shows the importance of including optimization of side chain positions during protein-protein docking. In Saarbrücken, the Helms group will continue investigating complexes of cytochrome c variants to its redox partners by using the BALL package (see above).
The association kinetics of the cytochrome c552 with cytochrome c oxidase is being studied in atomistic detail by brownian dynamics simulations with the SDA package (Gabdoulline97). Although the association rates are computed a bit high for the proteins in solution compared to experimental data for solubilized proteins, the relative trends are well reproduced for two slower associating oxidase mutants. For the first time, the association is being studied for a protein embedded into a membrane. An atomistic membrane model was equilibrated around cytochrome c oxidase by molecular dynamics simulations and the electrostatic potential of the full system was computed as input for the brownian dynamics simulations.
Protein-Protein Docking of membrane protein complexes (Helms)
A very important area of protein-protein docking techniques is the prediction of the three-dimensional structure of supramolecular complexes within biomembranes because these will be very hard to study experimentally. Certainly, this also poses challenges to docking techniques because, in contrast to soluble proteins, two types of environment need to be considered, aqueous solution and the membrane interior. These challenges have currently been reviewed (Helms_EMBOReports02). A pilot study generated a molecular model of the SERCA Ca2+ATPase with its physiological inhibitor phospholamban, a 52-mer peptide that contains two alpha-helices in solution. Oriented in different starting configurations, phospholamban was pulled towards the ATPase in force field energy minimizations. The structure that is in agreement with most biochemical data had the most favorable energy (Hutter02). This model has recently been challenged by another docking model (Toyoshima03) which shows the enormous interest in structural models of such systems. A big problem for atomistic docking is currently the enormous size of the proteins involved. Here, residue pair potentials as the one being developed in the Helms group should allow a significant gain in computational efficiency at a marginal expense in accuracy. Often, it turns out that it may even be counter-productive to use atomistic models as long as the side chain positions are not known exactly.
Q-HOP MD (Prof. Helms)
Classical molecular dynamics simulations employ empirical molecular mechanics force fields where chemical bonds are modelled as harmonic springs. Therefore, chemical bonds defined in the molecular topology file at the start of a simulation remain bonded throughout the simulation thus prehibiting the modelling of any chemical reactions. Theoretical studies of chemical reactions are, however, routinely performed nowadays using (CPU intensive) quantum chemical techniques. In the field of enzymatic reactions, it is common to couple an inner quantum region including the "active site" where chemical reactions take place to an outer region modelled by molecular mechanics. This separation into two regions remains constant during a dynamics simulation. These techniques are, therefore, not suited to treat proton transport through entire membrane proteins that would require a travelling quantum region. A very efficient simulation scheme was developed by the Helms group over the past 4 years where instantaneous proton hopping events are included in MD simulations of arbitrary biomolecular or organic systems. The hopping rates are carefully parameterized against accurate quantum chemical calculations (Lill00, Lill01a, Lill01b, Lill01c, Herzog 06). The first application to a protein system was the proton shuttle in green fluorescent protein (Lill02).
The Q-HOP MD is now being implementated into the parallel quantum chemistry package NWChem. In addtion, we have optimized a separate set of atomic point charges on the donor and acceptor groups to mimic the environmental energy differences using the DFT/B3LYP theory. Base on these progress, we have generated 50-ns Q-HOP MD trajectories of solvated acetic acid. Without imposing any particular transfer mechanisms, we characterized the protonation equilibria from the observed spontaneous partitioning between titratable sites and solvent molecules. During the simulation, two different regimes of proton transfer were observed. Extended phases of frequent proton swapping between acetic acid and nearby water were separated by phases where the proton freely diffuses in the simulation box until it is captured again by acetic acid. The pKa of acetic acid was calculated around 3.0 based on the relative population of protonated and deprotonated states and the diffusion coefficient of excess proton was computed from the average mean squared displacement in the simulation. Both calculated values agree well with the experimental measurements.
Phosphoryl Transfer in Signal Transduction (Prof. Helms)
Phosphoryl transfer is one of the key reactions in cellular signaling and is mediated by kinases. The elucidation of the involved catalytic reaction mechanism in atomistic detail is thus desirable for the understanding of mutational effects as well as the design of kinase related drugs. Experimental determination of kinetic rates of the catalytic mechanism is difficult because of slow conformational changes of the protein accompaning ligand binding and product release release which often are the rate limiting steps. Theoretical methods allow an independent look at the actual reaction mechanism without having to account for the binding affinities of the substrates. Quantum chemical investigations by semiempirical molecular orbital calculations showed that the phosphoryl transfer involves the simultaneous shift of a proton onto the transferred phosphoryl group in active center models of cAMP-dependent kinase (Hutter99), UMP/CMP kinase (Hutter00), and in NDPK (Hutter02). Protein residues, however, were not found to participate in the reaction. Despite the chemical variety of the phosphoryl acceptors the actual enzymatic mechanism in all kinases is apparently identical and the associated energy profile is determined by the nature of the phosphoryl acceptor rather than the characteristics of the individual catalytic centers. The active sites of the specific kinases are tailored to accommodate the reactants via favorable electrostatic interactions and hydrogen bonding patterns, respectively.
Currently, a PhD student is characterizing the energetics for phosphoryl transfer between model compounds with the idea of deriving a simple parametrization of phosphate transfer reactions. This will then be used in molecular simulations of signal transduction cascades.
|
