Faculty of Chemistry, Department of Physical Chemistry, Gdansk University of Technology (Poland)
Local Project ID:
HPC Platform used:
SuperMUC of LRZ
The enzyme ATP synthase has been described by one of its foremost experts as “a splendid molecular machine”. It is capable of using energy stored in the form of protein gradients across the mitochondrial membrane to synthesise ATP – the ubiquitous universal energy carrier which provides energy for nearly all of the processes in the cell, including protein synthesis.
One of the most remarkable characteristics of ATP synthase is its efficiency. In terms of energy conversion, it beats anything that has been made by artificially by humans, with almost no energy being lost up to the limits of the second law of thermodynamics.
Jacek Czub of Gdansk University of Technology has been using HPC to try and understand the mechanisms behind this high level of efficiency. “We want to find out the design choices that evolution has taken to produce such an efficient enzyme,” says Czub. “ATP synthase is extremely well conserved evolutionarily – the version we see in humans is almost exactly the same as the version we see in primitive bacteria, so it is a molecule that is seen across most of the lifeforms that are present today on Earth.”
To be able to study the mechanisms of ATP synthase at an atomistic level, computer simulations are essential. X-ray crystallography and, more recently, cryo-electron microscopy, have provided a plethora of extremely valuable structural information about the enzyme at work, but these are limited to what are essentially still images – snapshots – of the mechanism, from which it is difficult to elucidate the entire dynamic cycle.
Czub and his colleagues have taken these still images from the field of structural biology as a starting point for their research, and are now using molecular dynamics simulations to try and fill in the gaps of the entire dynamic cycle of the enzyme, which involves a number conformational changes and movements of the active sites that are responsible for the synthetic chemical part of the mechanism. These changes are coupled to the motions of a rotary element that connects the ion transportation portion of the molecule to the chemical portion.
“The entire complex of the protein is huge, and along with solution that we simulate it in, consists of around one million atoms,” says Czub. “This is why we needed the HPC resources that we have been awarded by PRACE, in order to study the enzyme at the level of precision that we are interested in.”
The two functional parts of the enzyme – known as F1 and F0 – are, due to computational limitations, usually studied separately, but with the PRACE access received by Czub and his colleagues they have been able to model the entire protein as one.
Although it is at a stage where it is difficult to give any definite answers about the findings, Czub is optimistic about the project’s outlook. “We have produced somewhere between 10 and 20 terabytes of numerical data describing the dynamic evolution of all the atoms in the system, and we are now in the process of analysing that data in detail.
“We are now fairly confident that we should be able to solve one of the central puzzles of the mechanism of ATP synthase, which is in what way the ATP release – which is not spontaneous – is coupled to the binding of ADP, the substrate ATP is made of. This question of how the enzyme harnesses the energy of ADP binding to drive the release of the product has been unknown for many years, and so we will be very pleased if we can shed some light on it.”
The team is also hoping to explain the energy transmission mechanisms of the enzyme from the rotary element to its active sites. This stage of the processing of the data collected will no longer require the use of HPC, so it is now a case of carefully making sense of the findings.
In terms of the code used in the project, most of the work was done using the standard GROMACS code, which is well optimised for parallel computing. However, Czub has also been collaborating with colleagues from Germany and Sweden who have provided a computational tool for calculating electron densities.
Enhanced sampling has also played a large part in the project to allow the team to study how the enzyme operates at crucial stages of its mechanism. The technique – which uses the application of artificial external forcing to push the system in a certain way – allows the researchers to observe rare events in their simulations that would be unlikely to occur otherwise. The effects of the artificial forcing can then be subtracted to give a genuine picture of what happens, at a much lower cost of just using brute force and waiting for the rare events to occur. Czub is keen to highlight the positive experience of working with PRACE. “The whole project has gone very smoothly,” he says. “We were very pleased with the transparent nature of PRACE’s evaluation procedure, and we were also pleasantly surprised with the interesting modifications to our project that they suggested. This reassured us that we were being evaluated by real experts in the field, which is not always the case in this kind of situation. Although the architecture was fairly familiar to us, the people at the Leibniz Supercomputing Centre, where the SuperMUC system is hosted, were very helpful when we needed them, and we would definitely consider applying for resources with them in the future.”
Gdańsk University of Technology (Politechnika Gdańska)
Faculty of Chemistry
Department of Physical Chemistry
11/12 Gabriela Narutowicza Street
80-233 Gdańsk (Poland)
e-mail: jacek.czub [@] pg.edu.pl
NOTE: This is a reprint of the article published in PRACE Digest 2019, p.12-13. The simulation project was made possible by PRACE (Partnership for Advanced Computing in Europe) allocating a computing time grant on GCS HPC System SuperMUC of LRZ.
LRZ project ID: pn69fe