Binding Specificity of Biomolecular Interactions
Center for Bioinformatics, Saarland University, Saarbrücken (Germany)
Local Project ID:
HPC Platform used:
SuperMUC of LRZ
While some proteins of a biological cell are bound to cellular structures others diffuse freely. Especially in a crowded cellular environment, proteins constantly bump into other proteins which sometimes leads to biologically meaningful contact of the two proteins–the binding partners may either remain bound or a chemical reaction may take place. Performing atomistic molecular dynamics simulations on SuperMUC, bioinformaticists try to unravel the biophysical principles underlying such “specific” biomolecular interactions.
Biological cells are filled to about 30% with proteins. Except for those proteins that are bound to some cellular structures, proteins diffuse freely in the cell according to the well-known Brownian motion. In the crowded cellular environment, this means that proteins constantly bump into other proteins.
In some cases, such a collision actually leads to a biologically meaningful contact of the two proteins. Then, the binding partners may either remain bound or a chemical reaction may take place whereby, for example, an electron or a phosphate group is transferred from one protein to the other one before they unbind again. Such meaningful encounters are called “specific” interactions. In all other cases, the proteins only form short-lived “non-specific” contacts that are biologically not meaningful. Specific complexes often involve larger contact interfaces than non-specific complexes and are stabilized by favourable interactions such as hydrogen bonds or salt-bridges between the binding partners.
A group of researchers of the Center for Bioinformatics of the Saarland University, Saarbrücken, has a particular interest in unravelling the biophysical principles underlying such biomolecular interactions . In this project that was also supported by DFG normal grant HE 3875/11-1, they therefore studied the dissociation of three different protein-protein pairs. The selected protein pairs all form specific complexes that are stabilized by hydrophilic binding interfaces. The three protein complexes are illustrated on the left side of Fig. 1.
Results and Methods
The scientists employed atomistic molecular dynamics (MD) simulations that were performed with the GROMACS 4.5 software package  to characterize the free energy landscape governing the association and dissociation of the three protein-protein pairs. Precisely, they computed with the help of umbrella-potential restrained simulations the potential of mean force (PMF) between the two proteins along a one-dimensional reaction coordinate [3, 4].
Because the unbinding pathway up to distances of several nanometres had to be followed - where the proteins do not attract each other anymore – the simulation systems contained several hundreds of thousands of atoms. For each binding orientation, 21 windows at different separation distances of 10 – 40 ns length each were computed, consuming several million core hours of compute time on SuperMUC.
The right side of Fig. 1 illustrates the computed free energy profiles. Interestingly, the bioinformaticists found that all three protein complexes bind “downhill” into both the specific and non-specific complexes without passing over an activation energy barrier. This behaviour is comparable to some fast-folding proteins that fold “downhill” in a steep folding funnel.
The attractive interactions between both proteins extend to 1.2 to 1.5 nm of separation and then become flat. This behaviour was found for all three complexes and may be a general behaviour of specific hydrophilic protein complexes .
Studying such systems on a computer enables one to generate also starting conformations that cannot be studied easily in experiment. To contrast the binding into specific complex orientations with the binding into non-specific contact orientations mentioned before, the scientists performed unbiased MD simulations of the binding partners and monitored all contact orientations that differ from the specific complexes. Among all such contacts they selected the ones with largest binding interfaces and the longest-lived ones as starting positions for PMF calculations. In Fig. 1, these orientations are gold-coloured and silver-coloured for the three systems.
For these “non-specific” contacts, the bioinformaticists then computed the same association free energy profiles  as for the specific orientations before. As expected, the difference in free energy between the bound state and the unbound state turned out to be much smaller than for the specific complexes. Interestingly, also the range of the attractive interaction is shorter (0.7 – 1.0 nm) - except for the EIN-HPr pair - and the interface formed between the proteins is significantly smaller than for the native specific complexes.
On-going Research / Outlook
Due to the lengthy and large-scale MD simulations, only access to SuperMUC made this project feasible at all. No technical obstacles were faced in the course of this project which required careful fine-tuning of the simulation protocols in terms of how to generate “good” starting conformations for the 21 windows.
The group of researchers will continue working towards a mechanistic understanding of biomolecular interactions. Dr. Mazen Ahmad at the nearby Max Planck Institute for Informatics recently derived an intriguing theoretical model of how conformational changes contribute to binding thermodynamics , which the Saarbrücken based bioinformaticists plan to incorporate in their further research in order to study interconnected intramolecular and intermolecular processes.
Prof. Dr. Volkhard Helms (PI), Dr. Ozlem Ulucan.
References and Links:
 Ulucan O, Helms V (2014) J Chem Theor Comp, 10., 3512-3524.
 Ulucan O, Helms V (2015) J Phys Chem B, 119, 10524-10530.
 Ahmad M, Helms V, Lengauer T, Kalinina O (2015) J Chem Theor Comp 11, 2945-2957
Prof. Dr. Volkhard Helms
Center for Bioinformatics
Chair of Computational Biology
P.O. Box 15 11 50
e-mail: volkhard.helms [at] bioinformatik.uni-saarland.de