Energy landscapes in protein science

The following article by David Bradley

is a summary report commissioned by the Royal Society to accompany its recent meeting on the subject of protein energy landscapes. For information on RS meetings please visit their website - http://www.royalsoc.ac.uk

The shifting landscapes of life science

Navigating energy landscapes

The Royal Society brought together experimentalists and theoreticians from a wide range of disciplines to map a bird's eye view of energy landscapes. Their disparate vantage points act as field guides to a range of problems including the medically important folding and misfolding of proteins and the behaviour of molecular clusters including silica, the workhouse of materials science and electronics. Navigating energy landscapes could even explain how water can act as a glass and the real reason icebergs float.

The scientists discussed many different aspects of energy landscapes. These ranged from those emerging from the ultrafast studies of Caltech's Ahmed Zewail and his colleagues. That team is laying bare the chemical reaction. Dortmund University's Roland Winter, on the other hand, explained how studies of fatty, lipid molecules are helping us to understand how organisms survive extreme cold and pressure at the bottom of the oceans. Cambridge's Chris Dobson provided new insights into why Alzheimer's disease, type II diabetes, and other diseases are becoming major scourges of the modern world.

Energy landscapes help scientists visualise how the geometry of molecules relates to their energy. A clear map of the landscape will ultimately lead to a better understanding of a multitude of physical and chemical phenomena.

Frame by frame chemistry - in space and time

The landscapes of nature can describe the energy peaks and valleys of the reactions of the smallest molecule to the most complex protein, explained Ahmed Zewail of the California Institute of Technology. But, scientists face a problem: how to reduce the number of coordinates they need to map this landscape to describe a phenomenon in a comprehensible reduced space. Once we can do that we might navigate phenomena as diverse as freezing water and the folding or misfolding of a protein.

Scientists have learned how to measure the coordinates of chemical reactions, their trajectories, on the timescale of atomic movements but recent developments are now letting them see the changing structures of molecules as they chemically react.

Zewail's team is probing chemical reactions using the emerging technique of ultrafast electron diffraction to reveal how structures evolve in time and space. The challenge remains to understand the multitude of changes that take place in any given chemical reaction and to find the molecular shapes and structures at key points in a reaction.

Pulses of ultraviolet light that last a few billionths of a millionth of a second trigger a chemical reaction, while a second beam stimulates a photocathode to spit out electrons, which are focused by electrical plates on to the reacting molecule. This diffracts them around the short-lived chemical species produced by the first ultraviolet pulse. The researchers record the diffraction pattern and, as Zewail explained, this reveals direct information about the structure of the intermediates.

With this technique, scientists can now record "frame-by-frame" a chemical reaction in space and time showing how the intermediates between reactants and products evolve with time. Zewail's team has achieved unprecedented atomic detail in observing ice melting and the interaction of water with proteins on the femtosecond timescale.

Insightful protein simulations

Living cells are a collection of molecular machines, Harvard University's Martin Karplus, explained. These machines are mostly proteins and keep cells running, allow new cells to develop functions and assist in cell replication.

Proteins are complex polymer chains composed of many amino acids that fold into functional three-dimensional structures. Because proteins are so complicated, experimental results must work synergistically with theoretical studies to obtain a full understanding of folding and function. One problem facing theoreticians who study protein function is that most conformational transitions take milliseconds to occur but, explained Karplus, the techniques used to study molecules, such as molecular dynamics, are normally limited to much faster processes.

Karplus outlined how he and his team have studied two very different cellular machines that work by undergoing such structural transitions. The first, GroEL helps other proteins to fold into their own functional shape. A typical cell is a concentrated solution of many different molecules, explained Karplus, so that folding an amino acid chain properly without interference from other molecules requires help from the "chaperone" GroEL.

Karplus' team have devised a new theoretical approach to study the workings of GroEL that essentially speeds up the conformational transition to make it possible to use molecular dynamics calculations to study how it works. Their findings backed by experimental data, show that GroEL effectively forms a tiny reaction chamber in the cell into which an unfolded protein can fit, be unfolded by GroEL if it is misfolded, and then fold to its functional form without interference from other molecules in the cell.

The second protein, F1-ATPase, synthesizes the cell's high-energy molecule ATP (adenosine triphosphate), which makes possible biochemical reactions required by the cell. For example, one biochemical reaction that involves ATP is the binding together of the muscle proteins actin and myosin, that flex our muscles. Nanotechnologists are also exploiting the motor-like activity of F1-ATPase, Karplus pointed, in efforts to build a molecular-scale motor.

Realising the potential

According to David Wales of Cambridge University the observed structure, the dynamics and the thermodynamics of any molecular system, whether a small molecule like water or a complicated entity such as a protein, depend on the underlying potential energy surface (PES) of the system. This surface, a unique landscape in itself for every system, helps scientists bridge the divides between various phenomena, such as protein folding, the formation of glasses, and quantum effects such as tunnelling that occur in small clusters of molecules.

There are, explained Wales, many experimental challenges facing scientists hoping to understand the PES, which modern techniques are addressing. Experimental techniques such as ultrafast electron diffraction developed by Ahmed Zewail at Caltech and far-infra red vibration-rotation tunnelling spectroscopy invented by Berkeley's Richard Saykally can both provide new insights into the PES and so the molecular system it underlies.

Recent results from Wales' team on what he terms the "discrete path sampling approach" have begun to show how the energy landscape perspective can provide approximate dynamical information about molecular systems on an experimental timescale. For instance, for small molecules, ranging from fullerenes to borohydrides and carboranes, it is often possible to map out a complete reaction graph containing every possible arrangement of the molecules involved and the transition states that link them. By studying these graphs, researchers can glean new knowledge about how important molecules might react.

The same approach is useful in studying reactions that make and break small clusters of molecules. Wales and his colleagues have used the PES to help them interpret the tunnelling spectra obtained by Saykally and so provide a clearer picture of that ubiquitous yet infinitely puzzling material - water.

Studies of the landscapes of proteins too will help scientists understand how these molecules fold and why occasionally they misfold in certain diseases. Equally, a quantitative analysis of the PES might explain the puzzling behaviour of glassy materials from the familiar amorphous solids based on silica that we use in windows to modern materials such as metallic glasses and amorphous engineering plastics.

From the lowest low to the highest high

The free energy surface of a complex system can often characterized by the presence of deep minima separated by large barriers - the valleys and peaks of the energy landscape. The minima correspond to distinct states of the system, explained Michele Parrinello of the Swiss Federal Institute of Technology (ETH). Transitions among these valleys reflect important changes in the system. Such transitions might be the conversion of a reactant into a product - a chemical reaction; the melting of ice - a phase transition; or a change in molecular shape and structure - a conformational modification. But, in real life, understanding energy surfaces is not quite so simple.

The exponential dependence of the rate of change in a particular process on the barrier height often frustrates direct computer simulation of such processes. Too deep a valley or too high a peak and the simulation will not surmount the energy barriers and the results will reveal information only about the energy minimum where they start; unless one has many years to run the simulation. Parrinello and his colleagues have found a way around this problem.

Parrinello explained how he and his team have developed a new form of statistical analysis for molecular systems known as "coarse-grained non-Markovian dynamics", or more succinctly "metadynamics". Metadynamics can overcome the large energy barriers in a modest amount of computer time by modelling the energy landscape using a new set of coordinates that does not disturb the landscape's overall layout but fills in the valleys so reducing the height of the peaks. Over time, the computer recognises the places it has already been in the landscape and so does not revisit, or recalculate, those features.

Parrinello's novel approach could tackle a wide variety of problems, from understanding such seemingly simple processes as the dissolving of common salt in water to the melting of ice. The fact that the approach works well in describing these processes, explained Parrinello, bodes well for its application to other more complex processes.

The mass probing of proteins

Mass spectrometry is widely employed in protein research for identifying individual proteins after separation. Less well known is its utility in protein transitions studies, said Cambridge University's Carol Robinson. The technique maintains intact protein complexes and so complements existing structural biology methods, reaching regions of the energy landscape inaccessible to other techniques.

Robinson outlined how mass spectrometry helps scientists probe the inner workings of protein-RNA complexes involved in protein folding and translation, for instance. The first example she described revealed how mass spectrometry probes the molecular composition and arrangement of other molecules in the "TRAP" protein complex. This complex binds to RNA and regulates the synthesis of the essential amino acid tryptophan in bacterial cells.

By using a small amount of a very dilute solution of the protein complex the researchers can produce a very fine spray of droplets that can be cooled to prevent the complex disintegrating in the mass spectrometer. Using a technique known as nano-electrospray mass spectrometry together with extensive cooling and a modified mass spectrometer the researchers can record a spectrum for the intact protein complex.

Spectra recorded with other biomolecules added to the system reveals detailed information about how TRAP behaves in solution, difficult with crystallography. Robinson's studies have thus revealed that the complex is not the simple ring of eleven sub- units seen in its previously reported crystal structure, but under her conditions, a stack of 12-mer rings.

Robinson's second example of the power of mass spectrometry described one of the most important species found in the cell - ribosomes, the site of protein synthesis. Ribosomes, explained Robinson, are an enormous target for mass spectrometry as they are composed of three very large RNA molecules and over fifty different proteins. Understanding the evolution and function of ribosomes might ultimately lead to new treatments for certain protein disorders.

Again, using the nano-electrospray technique, she and others can examine the proteins released because of ribosome activation. Because the particles remain intact in the mass spectrometer it is possible to reveal structural features of ribosomes and changes in the protein RNA architecture that occur when different biomoleules bind to ribosomes. Robinson and her colleagues have found that they could identify significant difference in the ribosome structure in the presence of cofactors not observed in the mass spectra of naked ribosomes.

Chilly nanodroplets freeze biomolecular action

Roger Miller of North Carolina University at Chapel Hill is studying nucleic acids in what he described as "almost the gas phase" by using tiny, nanoscopic droplets of superfluid liquid helium. Giacinto Scoles of Princeton University pioneered this nanodroplet technique. It offers scientists the chance to wander the potential energy surface of biomolecules, perhaps even proteins, some distance away from the realm available to spectroscopy.

Spectroscopy is certainly a powerful tool for studying potential energy landscapes, said Miller, but spectra are often broad because spectral lines smear out as warm molecules vibrate and change form. The nanodroplet approach side steps this problem. By rapidly cooling a biomolecule in a helium mist four degrees above absolute zero, spectroscopic action is frozen. The different forms of a particularly interesting molecule are locked in and can be studied without the broadening effects.

Miller's team has investigated nucleic acid bases (NAB) - the building blocks of DNA - using this approach. Infrared laser spectroscopy of NABs, such as cytosine, in helium nanodroplets reveal nuances of the molecules that are inconsistent with theoretical studies, which explained Miller, implies that the theory might be wrong.

As an example of how the technique can improve our understanding of NABs, Miller considered cytosine. Despite the huge number of studies of this molecule, researchers cannot agree about which of its possible chemical arrangements occurs in particular chemical situations. Miller asked whether the nanodroplet technique could help overcome the tendency to shoehorn experimental results into the theoretical picture. The researchers have made conclusive assignments on cytosine that were impossible previously and revealed that even the best previous theoretical studies are not consistent with the real structures seen for this NAB.

The method, said Miller, can now provide a benchmark for studying this type of biomolecule might improve the power of the theoretical studies. The results of metal-containing biomolecules, he added, could have major implications for studying DNA as a nanotechnology material, for instance.

Liquids behaving badly

According to Austen Angell of Arizona State University, most liquids are well behaved when they are cooled. They either crystallize or, for more complex molecules, smoothly change into glassy solids at a "glass-transition temperature". He explained that this vitrification is represented as the descent of a "system point" on an "energy landscape". Some materials such as liquid silicon and water, however, behave "badly", with their smooth descent interrupted by a sudden jump to a new liquid state with a different density and different behaviour.

At densities in between, liquids cannot exist and the transition from one liquid to the other happens only with nucleation and growth. A process that transforms one liquid into another at the same temperature must be a highly cooperative process. Angell argued that for liquids that make the jump, simple models could explain the process by invoking clusters of defects.

For water, added Angell, research results suggest that this "bad" behaviour is only encountered at quite high pressures - above 1000 atmospheres (the pressure at the bottom of the Marianas Trench). At normal pressure, water converts very rapidly to an almost "ideal" glass.

A similar situation exists for proteins, which must "fold" into their working three-dimensional structures. Folding happens suddenly at a special temperature for each protein. The sudden change resembles the smooth transition of water, if it is "well-behaved", or like the liquid-liquid transition in silicon if it behaves badly.

The correct description of protein folding remains controversial, Angell said. Researchers do not know if there is a no-man's land for individual protein molecules across which they must pass by nucleation, or whether proteins slip smoothly into the folded state. Angell suggested his group might have a way to tell, using a "magic" solvent that allows them to quench proteins, in their unfolded states, to low temperatures and then watch their behaviour during slow re-heating. It is too early yet to decide what different proteins really do, he said, but probably both scenarios have their special members.

The landscapes of protein folding and misfolding

It is understood, said Cambridge University's Chris Dobson, that proteins are complicated objects. They represent a highly select group of biological molecules with very special characteristics and properties compared with random sequences of amino acids - namely, that specific sequences can fold into unique structures. The folding of natural proteins underpins the ability of biological systems to carry out an enormous range of functions with an astonishing degree of specificity in chemical processes.

A unique combination of experimental results and theory has led to recent dramatic progress in scientists' understanding of protein folding. Of particular importance, explained Dobson, is our increasing ability to understand at least in outline the nature of the energy landscapes for proteins. He and his colleagues have recently developed a novel approach to improving our understanding of the properties of proteins still further. It involves the use of a wide range of experimental data to constrain computer simulations of protein folding to those regions of the energy landscape that are consistent with experimental measurements. In other words, they avoid the highly time-consuming process of investigating those hypothetical folding processes that lead to unrealistic protein structures.

Using this procedure, Dobson's team has been able to define many of the key regions on the energy surfaces for the folding of a range of representative proteins. Understanding protein folding is more than an academic challenge, however, as misfolding can give rise to very serious problems in the cell and so cause disease. Dobson and his colleagues have focused on misfolding diseases that are associated with the conversion of normally soluble proteins into insoluble aggregates. These insoluble clumps of protein can form fibrous deposits, or amyloid plaques, in the liver, spleen and brain and there are about twenty known disorders of this type, including Alzheimer's disease, the spongiform encephalopathies and type II diabetes.

Dobson's research has shown that the formation of such amyloid fibrils is a generic property of polypeptide chains rather than an aberrant property of a few rogue proteins. So, when old age sets in, for instance, and the molecular repair mechanisms that normally prevent the appearance of aggregated proteins fail, our proteins begin to form amyloid fibrils that can accumulate in vital organs and cause disease. Dobson added that his team is mapping energy landscapes comprehensively. Their aim is to understand these phenomena at the molecular level, which might lead to new therapeutic approaches to misfolding diseases.

Protein folding funnels

Peter Wolynes of the University of California San Diego hopes to discover what it is scientists do wrong when trying to predict the way an amino acid chain will fold into the active structure of the protein. It is, he said, possible to model known protein structures, but it is not possible with current computer programs to take a sequence of amino acids and use it to predict the structure of the folded protein reliably in all cases. Energy landscapes, however, could provide new insights into improving our predictive models.

In the energy landscape, a balance exists between the width of the valleys, or funnels - the entropy, or disorder - and the depth, - the enthalpy, or internal energy. These two properties almost exactly balance so folded and unfolded proteins can exist together in the test-tube, explained Wolynes. Yet, the subtle differences between the two gives rise to barrier between the unfolded and the folded state. This barrier, added Wolynes, is analogous to the barrier to crystallisation that means nucleation is needed to push a liquid into crystallising fully.

The way proteins interact too is governed by the balance between enthalpy and entropy and so can help us predict how aggregates of proteins will behave if we know enough about the structure of the individual protein. Learning about protein aggregates is important in protein misfolding diseases, such as the spongiform encephalopathies like BSE, scrapie and CJD. In these diseases many identically misfolded proteins clump together.

One factor that has emerged from recent studies is that water plays a special role in protein folding, added Wolynes. Proteins use water to help them fold, so it comes as little surprise that if scientists include data on the interactions between water and proteins then they can markedly improve the computer predictions of protein structure. As Wolynes pointed out, previous research assumed that water is excluded from the folds of a newly folding protein. But, this overlooks the role of water in the stages before contacts between amino acids along the protein chain form. The new insight provided by including water in models of protein folding, especially for larger proteins has, Wolynes explained, greatly improved their predictions.

Water-borne landscapes

John Finney of University College London described experimental results on configurational energy landscapes determined in aqueous solutions. In his work with Daniel Bowron of the ISIS Facility at the Rutherford Appleton Laboratory in Chilton, Didcot, Oxfordshire, Finney is examining structures in liquids, and how they change with changes in external conditions such as temperature, pressure, concentration, and added co-solutes.

As Finney explained, solvents are, perhaps obviously, the key to the structures and interactions of molecules in solution. Solvents modulate the interactions between solute molecules; if these are large, flexible molecules such as proteins or nucleic acids, then this includes their conformation in solution. If the solvent conditions change, then structural changes and transitions in molecules can occur. For instance, soap-like molecules undergo assembly processes to form microscopic bubbles known as micelles. On the other hand a natural chain of amino acids will fold into an active protein in the right solvent conditions.

Finney pointed out that experimentalists are now quite capable of navigating the energy landscapes of even rather complex liquid systems, something that was previously the realm only of the theoretician. By exploring the configurational energy landscapes that underlie processes such as protein folding, experimentalists can provide new insights into biology at the molecular level. By examples, Finney focused on an aqueous solution of an amphiphile - a molecule that is water loving at one end and water-repelling at the other. Amphiphiles undergo structural transitions induced by changes in temperature, concentration, and salts added to the system. They can be examined in a straightforward at the molecular level and some critical regions of their configurational landscape identified.

The potentials, or energies, of the average forces that describe how solvents modulate the interactions between molecules in solution can now be experimentally accessed, Finney explained. Such systems can serve as models of the more complex systems of biological models such as proteins, and could provide important indicators on how these might be studied in more detail experimentally. An example described in some detail was the salting out process - the addition of a salt to the aqueous amphiphile causes it to aggregate. Finney's experimental results showed that the beginnings of this process were determined by a chloride ion in their example, linking the polar ends of two amphiphile molecules. Experimental studies also revealed how the potential of mean force between the two amphiphiles changes on salt addition. This work might also help rationalize how the salting out process depends on the particular ions used, a problem unsolved since Hofmeister raised it in 1896.

The changing state of water

Richard Saykally of University of California, Berkeley uses ultra-high resolution gas-phase spectroscopy of ultra-cold clusters, what some researchers might refer to as ultra-slow spectroscopy. In particular he described how he and his colleagues are studying one of the most commonly misunderstood molecules - water. Water exists as boomerang-shaped units composed of an oxygen atom flanked by two hydrogen atoms. It is this configuration that endows it with some of its unique properties, explaining them fully will only be possible with spectroscopic studies that probe the way it forms clusters and the vibrations and oscillations of these clusters.

Water molecules can link to each other by way of transient bonds known as hydrogen bonds. The slight negative charge on the oxygen atom can hook up to the slight positive residue on a hydrogen atom on another water molecule. Through hydrogen bonds, which form and break very rapidly, water can form short-lived clusters and networks that underlie much of its behaviour. Understanding these processes will ultimately allow researchers to calculate accurately the properties of liquid water, such as its heat capacity, density, dielectric constant, and compressibility over a range of temperatures and pressures. This is something that hundreds of theoretical models of water have yet to achieve.

Having previously revealed the likely mechanism of hydrogen bond formation and breaking, Saykally's current work focuses on a deeper understanding of water. He explained that his team hopes to define a universal potential for water - a watery energy landscape. This energy landscape, or forcefield, will use vibrational and rotational spectroscopy of water clusters to refine the model.

The problem that Saykally's group face is that because there are six different ways in which even the simplest water cluster - the dimer - can move and vibrate, its energy surface is very complex with the result that so too are the spectra. However, by refining the model iteratively by comparison with their water cluster spectra Saykally's team is getting close to a "universal" model of water that might eventually allow them to accurately predict the complex behaviour of liquid water and ice.

A glassy water

The most common form of water in the universe is not a liquid but a disordered solid known as glassy water. Nevertheless, researchers are familiar with the anomalous behaviour of more common forms of water, such as its expansion on cooling from about 4 Celsius to below its normal freezing point. To understand such properties of water and how it interconverts between its different forms, scientists need to understand its static and its dynamic properties.

Gene Stanley of Boston University explained that the local geometry of a water molecule is tetrahedral - the two hydrogen atoms form two corners and the two pairs of non-bonding electrons on the oxygen atom form the other two corners, with the oxygen atom sitting at the centre. It is this local tetrahedral geometry that gives rise to water's unusual thermodynamics, its static properties. On the other hand, its dynamic properties are inextricably linked to its potential energy landscape.

Depending on how glassy water forms, whether by the vapour being deposited, from the liquid, or through ice forming under high pressures, there are many different amorphous solids of quite different structure that can exist. The densities of these various types of water vary by as much as 40% between the lowest density glassy water (low density amorphous, LDA) and the highest (high-density amorphous, HDA). Recent experimental, numerical and theoretical studies undertaken by Stanley's team attempt to explain how and why these forms exist.

LDA forms by extremely rapid cooling of a liquid, while HDA glassy water can form by compressing either LDA or crystalline ice. But, there are still gaps in our understanding of exactly how this occurs. An important clue lies in the discovery of a second "distinct" high-density structural state - very high-density amorphous solid (VHDA). This form of water is about 7-8% denser than HDA. According to Stanley, computer simulations have helped reveal the relationship between HDA, VHDA and liquid water and hint that VHDA is the stable version of HDA and forms by the annealing of HDA regions on heating. If such findings bear up to his team's experimental scrutiny, then such findings could hold the key to understanding the strange behaviour of liquid water.

Going with the liquid silica flow

Silica, silicon dioxide, is what Peter Poole of St. Francis Xavier University, in Canada refers to as a strong liquid - it remains very viscous and is easily manipulated, when heated enough to flow like a liquid. We tend not to think of the stuff of sand and glass as a liquid at all. However, glassblowers are familiar with raising silica above its glass transition temperature. The tetrahedral arrangement seen in molecules of silica resembles that seen in liquid water, at least from a broad physical point of view rather than in the detailed chemistry. Perhaps, Poole explained, this could provide the key to understanding its behaviour and so controlling its technological applications.

Poole presented the results of a comprehensive study carried out by his team on the "BKS" potential. This is a widely used model of silica due to van Beest, Kramer and van Santen that can simulate the molecular dynamics of the material. For the liquid state, his team has evaluated over a wide range of pressures and temperatures various properties of silica including the equation of state, the absolute free energy, the diffusion coefficient, and the average energy of structures inherent in its potential energy landscape. They have also quantified the local molecular order of the liquid state.

Another important result of their studies was the elucidation of the boundaries between stable and metastable forms of this liquid and three crystalline phases, beta-quartz, coesite and stishovite. The liquid phase, he explained, was found to show anomalies typical of tetrahedral liquids, again alluding to water, such as a density maximum, and the potential for a liquid- liquid phase transition at low temperature. He also revealed that silica shows a crossover in the energy landscape that strongly influences the nature of the glass transition.

So comprehensive is the characterization of the BKS model system for silica that Poole is confident that the product is entirely self-consistent and can relate the anomalous behaviour of the liquid state with characteristic features of its energy landscape. The results, he suggested, could provide researchers with a guide to identifying novel technological materials with similar properties.