Simulating the dynamics of proteins to understand protein functions

Yuji Sugita
Team Leader
Biomolecular Dynamics Simulation Research Team
Associate Chief Scientist
Theoretical Biochemistry Laboratory
RIKEN Advanced Science Institute
Proteins are found everywhere in living bodies, and are vital and essential for life. Scientists since the latter half of the twentieth century have been conducting research into structural biology to clarify the three-dimensional structure and to understand the functions of proteins based on their structures. Research by the many scientists involved in the field has brought about a significant advance in our understanding of life phenomena. “However, the knowledge obtained solely by observing the structures of proteins is insufficient for a full understanding of the mechanism of how proteins function in the body,” says Yuji Sugita, associate chief scientist and leader of the Biomolecular Dynamics Simulation Research Team in the Theoretical Biochemistry Laboratory at the RIKEN Advanced Science Institute. With the aim of understanding protein functions at the atomic or molecular level, the laboratory has simulated the behavior of proteins using various theoretical calculation methods, specifically focusing on molecular dynamics. The laboratory is also working on the development of a new calculation method in close collaboration with structural biologists from within RIKEN and from other institutions.

Structural biology and simulation

Sugita is inundated with applications for research collaboration from many structural biologists within and outside RIKEN. “We use computers to simulate structural changes in biomacromolecules such as proteins and cellular membranes so as to clarify their functions. These people say, ‘I have just solved the three-dimensional structure of a new protein, so why don’t we conduct a joint simulation?’” he says. Proteins, responsible for many of life’s mechanisms, consist of a number of amino acids and fold into the specific three dimensional structures. The three-dimensional structure of a protein is closely related to its function, and so protein functions can be investigated by clarifying the three-dimensional structure. This is the essence of structural biology.

It was in 1958 that the three-dimensional structure of a protein was first elucidated. John Kendrew, a British biochemist, and others created crystals of myoglobin, which stores and carries oxygen in the blood. They exposed the crystals to x-rays to clarify their three-dimensional structure at the atomic level, for which they were awarded the 1962 Nobel Prize in Chemistry. This technique, x-ray crystallographic analysis, is still a mainstream technique in three-dimensional structural analysis. Nuclear magnetic resonance (NMR) is also widely used for three-dimensional structural analyses of solutions.

As Sugita points out, “So far, more than 60,000 proteins have been clarified in terms of their three-dimensional structure, contributing significantly to our understanding of protein functions. However, the artificially arranged three-dimensional structure of protein crystals is not an exact match to the structure within the living organism. Proteins in a living organism either exist in solution in the cytoplasm or are embedded in biological membranes such as the cellular membrane and the endoplasmic reticulum membrane. Furthermore, a protein can change its structure dynamically by itself when it functions. However, when we want to observe these changes in greater detail, there is a limit to observation based only on x-ray analysis and NMR. Thus, much attention has been paid to computer-based molecular simulation techniques as a new approach that allows observation of the dynamics of proteins at the atomic level.”

How the three-dimensional structure of proteins is computed

To simulate the structural changes in proteins, atomic-level information on the three-dimensional structure of the target protein is required. The resolution of the information should be equivalent to that obtained by x-ray analysis or NMR. As the information on the three-dimensional structure does not provide sufficient information with regard to hydrogen atoms, the positions of the hydrogen atoms around the target protein are first predicted theoretically before creating the full atomic model. When the target is a water-soluble protein, the water molecules are properly arranged around the protein in the model. When the target is a membrane protein, it is embedded in a lipid bilayer membrane, and water molecules are properly arranged around the target protein. Finally, the forces between the atoms are calculated by solving the classical Newton’s equation to analyze and observe the dynamics of proteins, that is, how the positions of the atoms change with time. This procedure is known as molecular dynamics simulation.

The molecular dynamics simulation of proteins started in 1977. “The first report was of results calculated for a small protein called BPTI, a chain of 58 amino acids, in a vacuum for a period of one picosecond. Accurate representation of the motion of atoms, however, requires calculation at intervals of one femtosecond. The amount of calculation required increases from being proportional to the number of atoms or the square of the number of atoms. Computers at that time did not have sufficient computing power for such computations, and so the interaction between water molecules had to be neglected, and the calculation time was limited to very short time periods.”

Current-day computing power makes it easy to simulate a target protein and surrounding water molecules for periods of one microsecond, and a BPTI protein in water can even be simulated for a period of up to one millisecond. Molecular dynamics simulations are now being applied to more complex, large-scale models such as membrane proteins embedded in biological membranes and DNA-protein complexes.

Sugita notes that there are three factors contributing to the rapid development in the molecular dynamics simulation of proteins. The first factor is the dramatic improvement in computing power, which at present is least a million times what it was when molecular dynamics simulation first began. The second factor is the amount of information accumulated on three-dimensional structures with the development of structural biology, and the third factor is the development of new calculation methods. “The molecular dynamics simulation of proteins has developed rapidly, supported by the multiplied computing power, three-dimensional structure information, and various calculation methods. We have also made a contribution through the development of calculation methods.”

Development of an epoch-making replica-exchange molecular dynamics method

Sugita successfully developed a new calculation technique called the ‘replica-exchange molecular dynamics method’ when he worked for the Institute for Molecular Science. “The three-dimensional structure of a protein is most stable when it is at the lowest energy level. However, the energy level of a protein is constantly changing, and its energy distribution is like a rough landscape with many valleys and cliffs. Thus, it has many metastable states. We may be trapped in a valley during calculation and lost in the middle of this endless energy distribution. We could get out of the valley if we used an ultra-high performance computer for an indefinite period of time, but that is not realistic. Under these circumstances, the replica-exchange molecular dynamics method, which I and Prof. Yuko Okamoto at the Institute for Molecular Science (at present, Nagoya University) jointly developed, is one of the most effective calculation methods to solve this problem.”

The replica-exchange molecular dynamics method is based on a technique developed in theoretical solid-state physics. The technique was then modified and applied to molecular dynamics calculations. In the replica-exchange molecular dynamics method, multiple replicas of a target protein are prepared. These individual replicas are then simulated simultaneously at different temperatures. These temperatures are exchanged during the calculations, and the operation is repeated. “We can obtain energetically stable structures when calculations are performed at low temperatures, but it is difficult to get out of the valleys. In contract, we can easily get out of the valleys when calculations are performed at high temperatures, but we tend to obtain unstable structures. So we exchange the temperatures so as to take advantage of the two calculation methods and to drive various structures. We use these results to restructure the structural changes of the protein at a constant temperature and to achieve a long simulation period. A normal molecular dynamics calculation would require several hundred to several thousand times the computing time needed for the replica-exchange molecular dynamics method.”

The replica-exchange molecular dynamics method is computationally very efficient because individual replicas are calculated in parallel. This method has been included in major molecular simulation software packages and is used widely around the world. The original article has been cited more than 600 times, and that number is still increasing. “Thanks to the replica-exchange molecular dynamics method, we are now able to simulate the folding of proteins in water, which had been considered impossible. The development of calculation methods for simulation can be compared to the development of equipment for experimental science. As advanced research results are created by a group with new equipment, so new simulations are created by a group that is actively working on the development of new calculation methods.”

Moving proteins as if they were alive

“The advantage of research into simulation is high versatility. We can apply the calculation methods to various life phenomena after they have been examined in detail,” says Sugita. In his laboratory, researchers are working on various simulations for phenomena such as structural changes in membrane proteins, structural prediction of amyloid proteins (considered to be a major cause of Alzheimer's disease), folding and degenerative processes of proteins in water, and the behavior of lipid molecules in biological membranes. “I have been amazed by the ingenious functions of many biomacromolecules. In particular, I was deeply impressed by the complex molecular mechanism by which the calcium ion pump changes its structure to enable active transport of Ca2+. This is the research project I have been continuing in collaboration with Prof. Chikashi Toyoshima of the Institute of Molecular and Cellular Biosciences at the University of Tokyo since I once worked for the institute.”

The Ca2+ pump is a membrane protein embedded in the endoplasmic reticulum membrane in muscle cells that transports Ca2+ from the cytoplasm into the endoplasmic reticulum. The application of x-ray crystallographic analysis to membrane proteins is very difficult because they do not crystallize easily. Toyoshima and others, however, successfully determined multiple three-dimensional structures of the Ca2+ pump in different states. Sugita is using computer simulation to connect and arrange all of these ‘snapshots’ in order. “I can really see the true meaning of ‘understanding functions through three-dimensional structures’ when I discuss the structures of the Ca2+ pump with Prof. Toyoshima.”

However, they are only half way toward meeting the challenge of understanding the functions of the Ca2+ pump through its three-dimensional structure. The driving force that the Ca2+ pump requires for ion transport is the chemical energy produced when adenosine triphosphate (ATP) is hydrolyzed into adenosine diphosphate (ADP). However, the molecular dynamics simulation based on classical mechanics are unable to deal with chemical reactions. Thus, the effects of the chemical reactions cannot be taken into account. “We need to combine multiple theoretical calculation methods, including quantum-chemistry calculations.” Setting this challenge as one of his goals, Sugita started the RIKEN Theoretical Biochemistry Laboratory in 2007.

One of the laboratory’s recent research targets is related to the simulation of a membrane protein called translocon, which has the dual function of transporting proteins across a biological membrane and embedding other membrane proteins in biological membranes. The three-dimensional structure of translocon has been clarified to have a closed form before a protein is transported, but Osamu Nureki of the University of Tokyo and others discovered another three-dimensional structure for translocon, one involving the binding of an antibody molecule. They wanted to prove that this three-dimensional structure is the stage following the closed form, and asked Sugita for his cooperation. “Dr Takaharu Mori, a contract researcher in our laboratory, and his team successfully conducted a simulation for a period of 100 nanoseconds. We found that the new structure changed into the closed form when the antibody molecule was removed. The article including pictures generated by our simulations was published in Nature in 2008.” The team is now moving forward with joint research toward elucidating the molecular mechanism of how translocon transports proteins.

Sugita’s laboratory is also making progress on the simulation of several membrane proteins with as yet undisclosed three-dimensional structures. These studies have been made possible by close collaboration between the theoretical biochemistry laboratory and structural biologists.

The field of protein simulation is very competitive, but Sugita has confidence in his laboratory. “In our laboratory, we work on both the development of calculation methods and computer simulation. We also have a close relationship with structural biologists. Furthermore, RIKEN’s computer environment is wonderful. We have no parallel in the world in our ability to make progress with our research because of these conditions.” In addition to their own computers, they can use MDGRAPE-3, a special-purpose computer system for molecular dynamics simulations incorporated into the RICC, the RIKEN Supercomputer System. The RICC supercomputer is very powerful, and MDGRAPE-3 is the largest system in Japan for molecular dynamics simulations of proteins.

Suyong Re, another contract researcher, and his team, who work on simulations based on quantum chemistry, are beginning to produce visible results. However, they need a faster computer because the amount of calculation required in quantum chemistry is proportional to the fourth power of the number of atoms. “We will get molecular dynamics and quantum chemistry simulations into full swing using the Next-Generation Supercomputer that RIKEN is now developing. We want to move large and complex proteins such as membrane proteins in the simulation as if they were alive. We have decided on a time length of one millisecond as a target because that time length allows us to observe the entire function cycle of the protein. Next, we would like to attempt the simulation of life phenomena involving multiple proteins,” says Sugita.

Inspiration for theoretical biochemistry

In 1987, shortly before Sugita entered the Faculty of Science at Kyoto University, Susumu Tonegawa, director of the RIKEN Brain Science Institute, won the Nobel Prize for Physiology or Medicine. Sugita seemed to be interested in biology, but he could not give up his interest in physics. So he enrolled in Nobuhiro Go’s laboratory, which was conducting research applying theory and calculation in the field of biology. “Go’s laboratory belonged to the Department of Chemistry. At first, I was not interested in chemistry because I thought it consisted of purely memorization study. However, as I studied it further, I found it more and more interesting. In a broader sense, chemistry is the study of various phenomena driven by the interaction between atoms and molecules. We use the information obtained through theoretical calculations and experiments to understand the functions and structures of proteins at the atomic or molecular level. I once asked Dr Ryoji Noyori, president of RIKEN, whether our research activities were in the area of chemistry. I was very pleased when he told me that our research activities were exactly in the area of chemistry,” says Sugita. The Theoretical Biochemistry Laboratory symbolizes Sugita’s research interests. “I would like to focus on understanding living organisms from the perspective of chemistry, based on molecules and atoms.”

About the Researcher

Yuji Sugita

Yuji Sugita was born in Niigata, Japan, in 1969. He graduated from the Faculty of Science, Kyoto University, in 1993, and obtained his PhD in 1998 from the same university. After half a year postdoctoral training at RIKEN, he worked as a research associate at the Institute for Molecular Science in Japan. In 2002, he moved to the University of Tokyo as a lecturer. From 2007, he has served as associate chief scientist at the RIKEN ASI, directing his own research group. His research focuses on computer simulation of biomolecules and the development of new simulation methodologies.