Our study of both systems began by building ground-up models of function using a minimal scheme of system function. These models were then used to make quantitative hypotheses about rate-limiting processes in the two systems. For the ABCE1, we aimed to model an experimentally observed ten-fold difference in its rate of function. This ten-fold difference was highly unusual: it appeared to occur asymmetrically in one of two highly symmetric domains of the system. From our initial model, we were able to identify seven hypotheses that potentially described this asymmetry.
The ABCE1 contains structural motifs similar to the F-ATPase, but it exhibits an irregular behaviour. This highly symmetric two-domain molecule has an asymmetric behaviour: mutations that inhibit ATP hydrolysis in the first domain cut the overall hydrolysis rate in half, while mutations inhibiting hydrolysis in the second domain lead to an ten-fold increase in the overall rate. We aimed to determine how these two domains worked together and independently to produce the observed difference in ABCE1 hydrolysis rates.
Our first hypothesis was that the energy required for the system to transition from its inactive to its active state was altered by the mutations in question. The major difference was that mutated ABCE1 domains were unable to release ATP (and thus unable to bind a new ATP molecule). We calculated the energy required for the system to transition in four systems: ATP bound to both domains, ATP bound only in the first domain, ATP bound only in the second domain, or in the absence of ATP. We found that ABCE1 was only meta-stable if both binding pockets contained ATP. This suggested that even if the mutations themselves altered the probability of obtaining the closed active state, ATP binding in both pockets was a prerequisite to activation.
We therefore tested whether the mutations themselves stabilized or destabilized bound ATP and found no significant difference. However, we did observe that ATPs bound in the two domains had drastically different kinetics. ATP bound to domain 1 was more dynamic than ATP bound to domain 2—the ligand sampled three meta-stable states.
This difference led us to consider a non-intuitive solution to the difference in ATP hydrolysis. Based on our minimal models of ABCE1 function, we could envision a scenario where the domains themselves had intrinsically different rates and that mutations were able to “short-circuit” (that is, completely bypass) several intermediates. In the case of domain 2, this would lead to bypassing what would otherwise be the slowest step, while the mutation in domain 1 would create a symmetrically related short circuit but would still encounter the putative slowest step.
To test this hypothesis we computed the energetic contributions to mutating key regions responsible for ATP binding. We observed that these mutations only resulted in energetic differences when they altered the ATP’s dynamics.
With this knowledge in hand, we then identified key individual components of the system that were thought to be responsible for the asymmetric ATP dynamics. We performed a further set of simulations where these components were perturbed, and this perturbation eliminated the asymmetry.
Dissemination:
We worked to disseminate the results of our work through four conferences and an invited talk given at the University of Bristol. The conferences included the “Workshop on Computer Simulation and Theory of Macromolecules” at Hunfeld, Germany in both 2016 and 2017; the European Biophysical Societies Association held in Dresden, Germany; and the international Gordon Research Conference on Computational Chemistry held in Girona, Spain. Nicholas Leioatts also served as chair and organizer for the associated Gordon Research Seminar on Computational Chemistry, also held in Girona, Spain.
