Living cells are characterized by a great diversity of separate internal spaces, the boundaries of which are made of membranes forming convoluted surfaces of manifold shapes. Sculpting these shapes is achieved in many cases by proteins. A single protein is too small to bend the membrane into useful shapes, such as spheres or tubes, that measure 10-100 nm, or more, in diameter. Indeed, the proteins work in teams, but exactly how remained a mystery. Now a computational study elucidates the membrane-sculpting process for proteins called amphiphysin N-BAR domain. Simulations performed with NAMD had revealed a first glimpse earlier (see the Sep 2008 highlight). The new study showed that multiple N-BAR domains form lattices maintained through electrostatic interactions. Positively charged, banana-shaped surfaces of individual proteins bend the negatively charged membrane, while the lattice formation ensures a uniform bending force across a wide membrane surface. In a dramatic example of computational "microscopy" the 200 microsecond sculpting of a large flat membrane into a complete tube was observed. More here.

Computational modeling seeks to simulate biomolecules, particularly proteins. The dream of computational biologists is that their simulations realistically mirror the structure and dynamics of proteins, which act as molecular machines in living cells. Indeed, when researchers tried to simulate how a nascent protein folds into its known shape the chosen protein, called WW domain, did not fold properly (see the May 2008 Highlight). Until recently, simulations could follow protein movement for about 0.0000001 seconds and the computational mirror seemed to work well. Recently, however, simulations with the program NAMD began to follow proteins for almost 0.0001 seconds, a thousand times longer, and the mirror showed cracks. Thus, a question arose as to what went wrong and how the distortion could be repaired. It was unclear whether the simulations still did not last long enough, or whether the physical interactions in the protein were poorly described in the computer model that was used. As reported in a recent paper, the interactions show subtle errors, significant enough to throw off the energy balance in the folding protein. Fortunately, the results suggest ways to improve the computation of physical interactions to fold proteins more accurately, repairing the cracks in the mirror. More information is available at our protein folding website.

NCSA News Release: Biophysicists at the University of Pennsylvania used NAMD running on NCSA's Abe to clarify a mysterious interaction between cholesterol and neurotransmitter receptors. Research into how anesthesia works may eventually unlock not only that mystery but dozens of others as well. "Anesthetics have improved significantly over the last hundred years, but the mechanism of anesthesia is not understood at all," says Grace Brannigan, a researcher at the University of Pennsylvania's Center for Molecular Modeling (CMM).

To gain insight into how anesthetics work, a team consisting of Brannigan and fellow researcher Jérôme Hénin, University of Pennsylvania professors Michael Klein and Roderic Eckenhoff, and Richard Law of the Lawrence Livermore National Laboratory, is focusing on the nicotinic acetylcholine receptor (nAChR). This receptor, found in both brain and muscle cells, is a ligand-gated ion channel. The channel opens or closes in response to binding with a chemical messenger (ligand) such as a neurotransmitter, like acetylcholine. When the channel is open, ions can cross the membrane. Anesthetics are believed to close the channel, thus reducing sensations and possibly causing the memory loss associated with being under general anesthetic.

The researchers hope their work leads to improved drug design. "You could fine-tune the properties of the drug if you could understand how the mechanism works," says Brannigan. "For instance, by understanding how anesthetics work, you could design new anesthetics that could be more powerful yet maybe wouldn't have some of the side effects that current ones do."

Many living cells, so-called eukaryotic ones, organize their genetic materials in the cell's nucleus, enveloped by a double membrane with guarded access through pores that involve an amazing filter. Like an ordinary filter it permits passage of small particles (biomolecules), but not of large particles (e.g., proteins). However, certain large particles, proteins called transport receptors, can pass. The filter is made of long "finger" proteins anchored inside the pores. The transport receptors can intermittently widen the filter. But to observe how this is achieved is difficult since the finger proteins are highly disordered. As reported recently, simulations using NAMD suggest now a simple and elegant answer: the finger proteins bundle in groups of 2 - 6 and form a brush, filling with its bristles the nuclear pores. The bristles are bundles of finger proteins and have two key properties: (i) on their surface they are dotted with spots of amino acid pairs, phenylalanine and glycine, that are known to interact favorably with transport receptors (see the Aug 2007 highlight, the Feb 2007 highlight, and the Jan 2006 highlight); (ii) the bristles are also interconnected, namely where finger proteins change from one bundle to another bundle, which they do with some frequency. It appears then that the bristles of the nuclear pore filter form an energetically favorable environment for transport receptors and that the latter can tear a finger protein readily away from a bundle to form a wider space for passage. More information here.

Nanotechnology develops small devices with dimensions below 100 nm, one hundred times smaller than the diameter of a human hair. Nanodevices can be used for a wide range of applications, such as biomedical sensors (see the Jan 2005 highlight) or tools for studying DNA properties (see the Feb 2009 highlight and the Nov 2005 highlight). In building and controlling such small devices, researchers run into problems such as surface effects and significant thermal fluctuations. Furthermore, properties arising from the discrete nature of matter start to dominate at the nanoscale, producing phenomena not observed in larger devices. For instance, when immersed in electrolytic solution and under the influence of an electric field, nanopores act as diodes for ionic currents, conducting in one voltage polarity better than the other, a behavior which has been proposed as the basis for developing nanoelectronic devices. In a recent report, researchers have studied this so-called rectification behavior by means of molecular dynamics simulations using the program NAMD, the ionic rectification inside the nanopore being described in atomic detail. More information can be found here. See also our recent biotechnology review.

All cells making up the human body contain the same DNA in their nucleus, the DNA entailing about 30,000 genes and each gene containing instructions for a protein. Despite this sameness, the cells in different parts of our body are very different due to many factors, a key one being that the level of expression of genes into protein is highly regulated and differs strictly from cell to cell. One rather common regulation mechanism involves methylation of one of the four bases of DNA, cytosin. Researchers find that the long DNA in human cells show spots of methylated cytosins, the methylation being correlated with the expression level of the genes near the spots. In fact, medical researcher relate several cancers to improper methylation of DNA. Despite the common occurrence of regulation by methylation, researchers have little understanding how methylation, that changes an H (hydrogen atom) for a CH_3 (methyl group) here and there, i.e., just adds small bumps on a rather bulky DNA molecule, affect the physical properties of DNA such that expression levels are altered. It was found that there are proteins that can recognize the CH_3 groups, i.e., the bumps, on the DNA, but researchers have a hunch that methylation does affect DNA properties directly, i.e., without protein markers, but do not know which properties. In a collaboration between bioengineers measuring the passing of DNA through nanopores and computational biologists simulating this process with NAMD (see also the Nov 2005 highlight stretchable DNA) first hints emerge that methylation does in fact alter DNA's ability to stretch itself through a nanopore. As reported recently, pulling DNA electrostatically through nanopores is easier for methylated than for unmethylated DNA, as seen both in experiment and simulation. The findings promise insight into an important chapter in the field of genetic control. More on our methylated DNA website. See also our recent biotechnology review.

The ribosome is one of the largest molecular machines present in hundreds to thousands of copies in every cell, in charge of synthesizing every protein in the cell faithfully from genetic instruction. For this purpose the ribosome "reads" the sequence of bases on so-called messenger RNA, three bases at a time and depending on the base triple, the codon, elongates a nascent protein by one of 20 possible amino acids, avoiding to an impressive degree adding a wrong amino acid. So far one knew that the reading is done by transfer RNA molecules that have "foots" which match the possible codons and a "head" that brings along the associated amino acid. Each amino acid has its transfer RNA, the transfer RNAs checking if the next codon is "theirs," and if it is they add the proper amino acid to the nascent protein, elongating it. But how does the ribosome make the critical decision at the decoding center, namely if the transfer RNA "foot," the so-called anticodon, matches the codon? The answer is not known, but a key detail has now been discovered through a combination of electron microscopy and molecular dynamics simulation using NAMD, VMD, and a method called flexible fitting (MDFF, see the June 2008 highlight). It was known that a third molecular system is involved, called the elongation factor Tu (EF-Tu), which generates a key signal to the ribosome and transfer RNA through a chemical reaction. This reaction involves chemically attacking a substrate of EF-Tu, the molecule guanosine-triphosphate (GTP), with water, breaking a bond and turning GTP into guanosine-diphosphate (GDP). The puzzle was that EF-Tu is far away from the decoding center. The collaboration between experiment and simulation, reported here, revealed that "correct recognition" through anticodon-codon binding opens a gate in the EF-Tu that allows water access to the GTP inducing the signaling reaction. The finding promises to now establish how the decision at the decoding center is made and how an "open sesame" order is transmitted to EF-Tu. More on our ribosome website.

The most powerful compute engine in your laptop and home computer is what? The central processing unit, of course. Wrong, it is the graphics processing unit (GPU)! Since GPUs are specialized for graphics and games, popular uses of modern computers, their development benefitted from strong market forces. Modern GPUs reach Teraflop speed, passing entire computer clusters filling a room. Unfortunately, GPUs lie dormant when computers are employed for intensive biological computing like biomolecular simulations. GPUs are now getting a wake-up call from biomedical researchers who then enjoy many-fold speed-ups of their laptop computations (see the Oct 2007 highlight), vendors already now hawking GPU-powered "desktop supercomputers." The next challenge is to bring the GPU from the laptop to computer centers, speeding up the world's fastest supercomputers. Building on previous development experience of NAMD and VMD, a recent report describes NAMD running at impressive speed on a GPU cluster. The report suggests techniques, useful also for programmers of other applications to efficiently accelerate computer clusters through GPUs. The report is particularly timely with a new, large GPU-accelerated computer cluster coming on-line. The reported GPU-based speed-ups will permit biomolecular simulations largely unfeasible so far for studies of entire virus particles and cellular organelles. More information here.

If you have a leaky roof, you need to plug the hole quickly to avoid further damage. Similarly, cells do not tolerate open holes in their surrounding membranes for long. However, cell membranes do require channels for various molecules to get across, and proteins are no exception. The protein-conducting channel SecY, a membrane-bound protein itself, is the pathway used by other proteins to cross the membrane. When it's not in use, it needs to have a water-tight seal, which is provided by two elements in the channel, a bulky plug and a constrictive pore ring (see the April 2006 highlight). But do these two elements, plug and pore ring, function separately to provide two independent barriers or together, providing a single barrier? Molecular dynamics simulations on the channel and two mutants in which a portion of the plug was deleted have provided the answer, as reported in a recent publication: both components are necessary to prevent leaks in the channel. More information on the amazing channel can be found on our Protein Translocation website.

Biological cells are surrounded by a highly versatile, yet very feeble cellular membrane and need to balance differences between the cell's interior and exterior that otherwise would burst the membrane. For example, the osmotic pressures inside and outside the cell need to be closely balanced. Thousands of proteins in the membrane act as gatekeepers, opening pores that can also act as safety valves, helping to reduce the interior-exterior difference in pressure rapidly. One such protein, the mechanosensitive channel of small conductance MscS (see the Mar 2008 highlight, "Observation and Simulation Depict Cell's Safety Valve", the Feb 2007 highlight, "Observing and Modeling a Crucial Membrane Channel", the May 2006 highlight, "Electrical Safety Valve", and the Nov 2004 highlight, "Japanese Lantern Protein") opens in response to cellular membrane tension generated due to a drastic imbalance in osmotic pressure as it arises when a bacterial cell suddenly finds itself in fresh water, rather than a highly saline physiological medium. The MscS channel widens then to jettison molecules out of the cell and reduce tension on the cellular membrane quickly. In a recent report, researchers have combined experimental data from electron paramagnetic measurements and computer modeling to reveal in atomic detail how MscS opens and closes its channel. Combining measurement and modeling, the researchers established a highly resolving computational microscope, unmatched by existing microscopes (more on our MscS website).

University of Pennsylvania News Release: Biophysicists at the University of Pennsylvania have used 3,200 computer processors and long-established data on cholesterol’s role in the function of proteins to clarify the mysterious interaction between cholesterol and neurotransmitter receptors. The results provide a new model of behavior for the nicotinic acetylcholine receptor, a well studied protein involved in inflammation, Alzheimer's disease, Parkinson's disease, schizophrenia, epilepsy, the effect of general anesthetics and addiction to alcohol, nicotine and cocaine.

Moreover, the results apply to closely related receptors that bind serotonin and GABA, which are neurotransmitters directly involved in regulation of mood and sleep.

The findings have broad implications for, among other fields, pharmacology. Drug development in this arena has to take into account the structure and chemical makeup of this receptor, both of which researchers now say were incomplete. Drugs acting on the receptor have been thought to interact with the protein as though it were isolated.

Now, researchers believe that drugs binding to the receptor not only interact with amino acids — the building blocks of the protein receptor — but also cholesterol tucked away within the protein. The shift in thinking transforms the understanding of this receptor in many ways, from shape and structure to its interaction with its environment and its response to neurotransmitters. The new model should spark a reexamination of several decades of research on the receptor's structure and function.

Living cells organize many functions: molecular import and export, signaling, transcription of genes into proteins, movement, building, repair, and more. These functions are realized through a complex architecture of the cell interior reflected in a system of labyrinthine membranes forming manifold cellular organelles from tubes, vesicles and many other shapes. Accordingly, cells need to sculpt their membrane in never ceasing processes and have at their disposal a wide range of mechanisms. A key sculpting mechanism is furnished by proteins, so-called BAR domains, that apparently form lattice-like scaffolds adhering to membrane surfaces. Such scaffolds have been observed through electron microscopy and they are now also being described through molecular dynamics simulations using NAMD. As recently reported, the simulations, carried out at four different levels of resolution (from an atomic to a continuum level), revealed that different arrangements of BAR domains lead to different curvatures. The simulations help to explain why BAR domains working in teams, i.e., in lattice formation, sculpt intra-cellular membranes into different shapes, depending on the exact arrangement. An arrangement of BAR domains that is particularly efficient in bending membranes was identified. More information can be found on our BAR domain web page.

Chromatophores are the photosynthetic machineries of bacteria. Each chromatophore contains, embedded in a membrane, all the photosynthetic proteins needed to absorb sunlight and turn it into chemical fuel. Chromatophores come in different shapes: while some chromatophores are spherical, others are flat or tubular. It has puzzled scientists how all these different geometries arise, and a hypothesis has developed that it is the photosynthetic proteins that render the shape of chromatophore membrane. In a study reported recently, computational biologists using NAMD took an atomistic look at how the chromatophore proteins bend the membrane. Simulations showed that the most numerous photosynthetic proteins dome the membrane, building arched membrane patches that can then be assembled into a spherical chromatophore. These simulations demonstrated that photosynthetic proteins construct their individual membrane environment, and when many of such proteins come together in the bacterial membrane, they can build functional cellular units with unique geometries. For more details, please see our chromatophore website.

Ouch!! You cut your finger with a knife. Blood immediately starts coming out from the wound, and in a panic you search through your drawer for a band-aid. However, well before you can find the band-aid, and even before the "ouch" came out, just about the same time you sensed the pain, your body's self-healing mechanism was turned on. A multistep signaling cascade involving a dozen different proteins in the blood started immediately after the cut, calling for platelet cells to clog arround the wound and form a plug to stop the bleeding. How does our body sense a wound so fast? A possible answer is that the platelet cells carry on their surface a sensitive molecular flow sensor, a protein called GP1b, which senses erroneous blood flow caused by a cut in the blood vessel. It has been hypothesized that a small segment located on the alpha-subunit of GP1b, called the ß-switch, transforms from a random coil to a ß-hairpin in the presence of shear flow (see movies). The ß-switch, in its ß-hairpin form, is able to bind to a protein called von Willerbrand factor, which will then anchor the platelet to the damage site. To test this hypothesis, researchers resorted to computer simulation using the program NAMD and mimicked blood flow around GP1b. As reported recently, the researchers discovered that a long suspected part of GP1b indeed changes shape under flow conditions, the likely trigger of the body's self-healing system. For more information see our molecular flow sensor web site.

Living cells are brimming with activity, much of it due to their manifold molecular machines pulling cargo, importing and exporting molecules, digesting food molecules and transforming their energy, reading and copying genetic messages, or synthesizing proteins from these messages (the latter done by the ribosome). Static structures of the molecular machines have been resolved through crystallography: machines pressed into the confinement of crystals and frozen into inactivity reveal their atomic level geometry through this methodology. However, many machines, for example the ribosome, undergo large conformational transitions during their cyclic action, but active motions are hard to view in atomic detail. A way out is offered by electron microscopy which freeze-shocks machines into states characteristic for action cycle intermediates. Unfortunately, the method does not yield atomic resolution images, leaving the chemical detail needed for a comprehension of the mechanisms blurred. Computational methods can be used to bridge the resolution gap: atomic level structures of non-functional states of the machines captured in crystals are deformed to match electron microscopy images. Until recently, the method worked well only for very small machines. Now a team of electron microscopists and computational biologists using NAMD extended the method to common size machines and reported its successful application to the ribosome, providing astonishing detail about ribosome dynamics and function. For more details, see our MDFF website.

Proteins carry out most functions in living cells, from import of food substances to chemical synthesis to motion to signaling. Proteins are chains of amino acids like GLSDGEWQLVLNVWGKVEAD... where each letter stands for one of twenty amino acids that are the building blocks of proteins, i.e., G for glycine or L for leucine. In general, sequences of proteins native to cells fold into unique three-dimensional structures capable of executing the proteins' specific function. Living cells store the amino acid sequences of their many different proteins in the form of DNA sequences, safeguarding them in the cells genome. On demand, the DNA sequences are translated according to the famous genetic code into amino acid sequences. The amino acid chains of newly synthesized proteins have to fold into the proper structure, an essential process scrutinized by biologists for decades. The folding process often takes milliseconds or longer, but recently proteins were identified that fold within microseconds. This was still a time too long to be simulated through molecular dynamics which could reveal folding in atomic level detail. However, improvements of NAMD have now made simulations of 10 microseconds possible and in a recent report experimental and computational biologists describe a joint study of a protein segment, known as the WW domain, over this timescale. The great increase in simulation time revealed intricate details of WW domain folding, but also points to a need to further improve the computational model (force field) used to simulate proteins. See also our protein folding web site.

Adhesion between human cells organizes our body into its organs and parts. The adhesion comes about through an intricate system of molecules that perform their task in a highly selective manner such that the body's different types of cells will find the right cells and stick to them. This selectivity leads to tissue differentiation and the organization of organs as complicated as the brain. Cadherin proteins form a particular family of such adhesion molecules. Interestingly, they glue cells together only in the presence of calcium. Some members of the cadherin family of proteins are also involved in the transduction of sound and cadherin mutants are known to cause hereditary deafness (see the April 2005 highlight, "Hearing: Turning Sound into Voltage"). How cadherins selectively bind to each other and the role of calcium was not well understood, but now molecular dynamics simulations have offered magnificent insight into calcium's role as recently reported. The simulations took advantage of parallel supercomputers and NAMD's ability to harness their power. The simulations revealed that in the absence of calcium cadherins stick out of cell surfaces like ends of loose rope; in the presence of calcium cadherin molecules turn into stiff hooks that link cells together. The calcium-induced links can withstand the strong mechanical forces that arise between cells much larger than cadherin (more on our cadherin website).

The environment of cells can undergo drastic changes, for example from dry to wet, in which case cells shrivel or swell. However, they are protected from bursting by a system of safety valves in their cellular membranes that open and release cellular content. Some of the valves open already at low membrane tension, but only little, others open only at higher tension, but wide and without filtering outflow. The mechanosensitive channel of small conductance, MscS, is a low pressure safety valve in bacterial cells (see the Feb 2007 highlight, "Observing and Modeling a Crucial Membrane Channel", the May 2006 highlight, "Electrical Safety Valve", and the Nov 2004 highlight, "Japanese Lantern Protein"). MscS is able to rescue cells about to burst by releasing small solutes through a large and transient opening in the cell membrane, thereby relieving internal pressure. The only way to learn how MscS performs this vital task is by determining its atomic-level structure under native conditions. However, the only structure available for MscS was obtained for the purified and crystallized protein; inspection of the structure left doubt that it shows a functional protein, i.e., a closed safety valve. Now a team of experimentalists and modelers report the structure of MscS seen in its natural membrane environment. In their approach, simulations incorporate information from so-called paramagnetic resonance measurements experiments. This finding is yet another case where the combination of modeling and observation offers entirely new close-up views of living cells (more on our MscS website).

Bleeding through physical injury is stopped in animals through the formation of blood clots. Such clots, actually arising often in blood vessels without injury, can rupture due to the blood's shear forces and obstruct upstream smaller vessels leading to life threatening stroke, pulmonary embolism, and heart attack. Hence, a blood clot must be both mechanically stable to stop bleeding, yet elastic enough to avoid rupture. Fibrin, the main component of a blood clot, possesses the stated mechanical properties in healthy individuals, but in pathological circumstances needs to be managed through medication. Unfortunately, preventive treatment of blood clots is still a black art since the molecular basis of fibrin elasticity is unknown. Clinicians at the Mayo Clinic teamed up with computational biologists at the University of Illinois to investigate this elasticity, focusing on the protein fibrinogen, the building block of fibrin. The clinical researchers stretched individual fibrinogen molecules measuring the force needed to extend the molecules. They found a characteristic force - extension relationship and its dependence on blood pH and calcium concentration, but they could not interpret the finding chemically, a prerequisite for improving blood clot chemical management. The clinical researchers joined forces with computational biologists who could reproduce the measured force - extension relationship in steered molecular dynamics using NAMD. The simulations starting from known crystallographic structures of fibrinogen offered a full, i.e., atom resolution, chemical picture of fibrinogen elasticity. As reported recently by the clinical and computational researchers the insight gained opens new avenues for blood clot treatment. For example, it was found that pH and calcium concentrations alter the stiffness of blood clots, thereby opening pharmacological avenues for controlling the incidence of pathological blood clots. More on this investigation can be found here.

Biological cells protect their interior through their cellular membranes, yet rely on import of nutrients. They have evolved for this import fast conduction channels that include reliable checkpoints distinguishing desirable and undesirable compounds. A checkpoint puts up a veritable obstacle course that only the right compounds can pass quickly. Understanding the channel design is difficult due to lack of detailed experimental data on nutrient dynamics. Presently, the most detailed information comes from viewing channel dynamics computationally, starting from static crystallographic structures. A recent study investigated how glycerols, small nutrient molecules needed by some bacteria, pass through checkpoints realized through the membrane protein aquaporin (see also highlights Gas Molecules Commute into Cell - Mar 2007, Aquaporin and the Cambridge Five - Oct 2006, Cellular Faucets - Feb 2006). Aquaporin furnishes four parallel channels that were monitored computationally using NAMD and a novel algorithm that explores the channel energetics quickly enough to be methodologicaly feasible on today's computers. The results show how the physical characteristics of glycerol, for example the molecule's ability to form so-called hydrogen bonds, its electrical dipole moments, its diffusive mobility and intrinsic flexibility are probed along the channel, discriminating glycerol from other molecules. More on computational investigations of aquaporin here.

Ever since the 1953 discovery of DNA's double helix structure researchers wondered how the double strands are separated so that genetic information stored inside the helix can be delivered from generation to generation. A class of proteins known to achieve this separation are DNA helicases (see Sept 2006 highlight), molecular motors that operate at a fork where a double stranded DNA separates into two single stranded DNAs. Helicases translocate along one of the single stranded DNAs, pushing forcefully into the fork to split the double stranded DNA apart further. Helicases seem to work, though, both through force and through persuasion, exposing to the double stranded DNA a surface that is apparently conducive for strand separation. This property suggests itself on account of the fact that many of the amino acid side groups at the relevant surface are highly conserved among species or evolved from species to species through pairwise mutation. But what chemical strategy evolution had in mind in molding the surface was not realized. Recently, however, researchers seeking artificial means of splitting double stranded DNA apart might have found a key clue. They pulled double stranded DNA at one of its single strands by means of an atomic force microscope from DNA's native salt water environment into a so-called non-polar solvent. The force - distance curve measured suggested that the DNA actually split apart, but there existed no direct experimental means of viewing the splitting. The researchers employed instead molecular dynamics simulations, using NAMD, that indeed clearly revealed the splitting of the DNA strands at a water - oil (octane) interface as reported in a recent publication. The study suggests how helicases achieve the splitting of DNA strands, namely by altering the local environment of DNA from water-like (hydrophilic) to oil-like (hydrophobic). More information here.

Globins are oxygen-storing proteins, vital to life. In our blood, hemoglobins carry oxygen from our lungs to every cell in our body. In our muscles, myoglobins keep reserves of oxygen to make sure it is available when needed. In some plants, leghemoglobins capture oxygen molecules that would otherwise be harmful to the production of ammonium necessary for the plant's survival. All these globins possess an iron-containig "heme", that grabs on to oxygen for a short time, and share the same protein architecture, despite large variations in their sequences. Since the heme group is buried inside a globin, scientists wondered how oxygen makes its way inside the protein to reach it. Exploring the motion and energetics of globins using the program NAMD researchers learned to gather data that permitted them to visualize, utilizing the VMD software, all the pathways taken by oxygen migrating inside whale myoglobin (see the Aug 2006 highlight and related publication). However, when the researchers turned their attention to the rest of the globin family to compute their oxygen pathways, they found, on their computational spelunking trip, something surprising. Given the conserved architecture of all globins, they expected to see similar oxygen pathways throughout the globin family, but they saw the opposite! Aside from a conserved pocket right at the heme binding site, the distribution of oxygen pathways showed very little similarity from one globin to the next. This result is described in a recent report, which shows that oxygen-pathways are not conserved by evolution, and that their location is not determined by a protein's overall architecture, but rather by its local amino acids. The researchers also learned which amino acids are found more often than others lining oxygen pathways, recognizing that bulky side groups are not hindering, but favoring oxygen passage. More information can be found here.

Modern computers include a massively parallel graphics processing unit (GPU) designed to perform geometry and lighting calculations at blistering speeds. State of the art GPUs can perform 0.5 teraFLOPS with their hundred cores. The tremendous computational power of GPUs was untapped by scientific computations because it could only be accessed with difficulty until now. As reported in the Journal of Computational Chemistry, recent advances allowing GPUs to be used for general purpose computing have boosted the performance of a number of molecular modeling applications, including NAMD simulations and VMD electrostatic potential calculations. The accelerated versions of these applications run five to one hundred times faster than on the best CPU-based hardware, allowing a single desktop computer equipped with a GPU to provide processing power equivalent to an entire, large computing cluster. More information on GPU acceleration of molecular modeling applications is provided here.

Everyone knows oil and water don't mix. Proteins observe this rule, too, some choosing to stay in the watery cytoplasm and others choosing the oily membrane. But getting into the membrane is not easy, and most newly formed proteins require another protein, the membrane-bound translocon, to help them insert into the membrane. The translocon, surprisingly also serves as a conduit for proteins across the membrane, thus carrying out a unique dual function. The structure of the translocon showed evidence of a likely "lateral gate", i.e., an exit from the channel into the membrane. How the channel opened to the membrane though, and how it closed afterwards, were not clear from the structure alone. Now, molecular dynamics simulations performed with NAMD, covered in a recent publication, have permitted researchers to understand how the channel opens laterally, how it closes, and how the oily lipids are prevented from invading the water-filled pore. Furthermore, the novel simulation technique, residue-based coarse graining, allowed the researchers to simulate the lipid-channel interactions for up to one microsecond, clearly illustrating that the lipids do not want to mix with the channel interior. More information on these results can be found on the Protein Translocation website.

Muscle fibers, in contracting and extending, generate tremendous force that needs to be buffered to protect muscle from damage. This role falls to the protein titin, with about 27,000 amino acids the longest protein in human cells. Titin functions as a molecular rubber band, but unlike uniform rubber bands, titin is made from over 300 different protein domains strung into a chain. While experiments have found that the individual domains of titin feature remarkable resilience against mechanical stretching, little is known about the elasticity of the overall titin chain. Crystallographers teamed up with computational biologists to investigate this elasticity, focusing on two adjacent titin domains. Molecular dynamics simulations using NAMD suggest, as reported recently, that the overall elasticity of the titin chain stems in part from a zigzag, i.e., accordion-like, motion: as titin is contracted and extended, energy is stored and released in the angular tilt of adjacent domains. More on this investigation can be found here.

The cells of higher organisms store their genetic material, the genome, in the so-called nucleus where they organize transcription of DNA into messenger-RNA, the blueprint for proteins. The messenger-RNA leaves the cell to be decoded by ribosomes that synthesize the respective proteins. Transcription factors, also proteins, control in the nucleus which parts of the cells' genomes are transcribed. Naturally, the access to the nucleus as well as exit from it must be restricted to transcription factors and related biomolecules. This is achieved by the nuclear pores, wide channels lined with brushes of polymers. The polymers are disordered proteins and prevent passage for most cellular proteins, except for so-called transport factors which bus transcription factors, messenger RNA, and certain larger biomolecule into and out of the nucleus. How transport factors are permitted to pass the nuclear pores, despite many studies, has been largely unknown. Molecular dynamics simulations, based on relevant crystallographic structures, using NAMD provided a comprehensive picture on the passage mechanism as reported recently. The simulations, analyzed with VMD, revealed that transport factors are dotted rather regularly on their surface with spots that bind to the brushes of nuclear pore proteins. While any protein may accidentally exhibit such a binding spot or two, only transport factors offer a regular pattern of such spots on their surface that apparently is their passport permitting them movement into and out of the nucleus, i.e., helping them to glide through the pores' protein brushes. More on simulations of transport factors can be found here.

Because oxygen gas is very reactive, it is frequently employed by the cell as a reagent by proteins called enzymes, which build the organic compounds that the cell needs. One such enzyme belongs to the copper amine oxidase family. These proteins transform amine-containing compounds into molecules needed by the cell, by reacting the compounds with oxygen. Researchers have long been interested in finding out how the various reagents reach the buried copper active site before the final oxidation reaction can occur. While copper amine oxidases exhibit an obvious channel for capturing the amine compounds to be modified, it had been unclear until now how oxygen molecules make their way through the enzyme. With the help of computer simulations using NAMD, researchers have identified in a recent publication, the routes taken by oxygen inside various copper amine oxidases from different species. In order to accomplish this, they analyzed simulations of the motions of four copper amine oxidases, using the VMD analysis and visualization software, which can predict the probability of finding oxygen molecules anywhere inside the simulated proteins. This analysis found numerous oxygen conduction routes inside each copper amine oxidase, i.e., oxygen can enter the protein through many routes, as it would in a sponge. More information on finding O2 migration pathways in proteins can be found here.

Like all organisms, bacteria have to eat. However, bringing nutrients in from the outside world is not an easy task for many bacteria that are surrounded by an extra membrane. The second membrane, called the outer membrane, offers additional protection but at a cost: no energy can be generated or stored at the outer fringes of the cell. So, to import large, rare nutrients that cannot cross by diffusion alone, bacteria have evolved a unique transport system which couples the inner, energy-generating membrane to the passive outer membrane, known as the TonB-dependent transport system. TonB, an inner membrane-associated protein, transfers energy across the periplasm to a variety of outer-membrane transporters. These transporters have a large, beta-barrel structure which is blocked in the middle by a plug called the 'luminal domain'. How TonB transfers energy to the transporter and causes the luminal domain to come out is still a mystery though. Now with the help of computer simulations using NAMD and a recent crystal structure of TonB coupled to BtuB, the transporter responsible for vitamin B12 transport, researchers have shown that TonB can mechanically activate the transporter by pulling on the luminal domain, causing it to leave the barrel. Using steered molecular dynamics, it was found that TonB stayed firmly attached to the luminal domain of BtuB, even though the contact between the two is limited to just a handful of residues. Furthermore, this pulling initiated unfolding of the luminal domain, opening a transport pathway for the substrate. These results, the subject of a recent publication and also highlighted in Science, demonstrate how a mechanical coupling can bridge the gap between the two membranes, thus enabling outer membrane transport.

We all know sushi rolls, but just to be sure here is an easy definition: a wrapper encircles rice which holds a precious bit of fish. To make a sushi role is an art and the same holds true for molecular sushi that is made of two lipoproteins as wrapper, lipids as rice, and membrane proteins as filling. Sushi rolls are for eating. Molecular sushi roles are for holding membrane proteins in place for physical analysis; they actually come only in sliced form, one disc at a time. Due to their size, the discs are called nanodiscs. Since membrane proteins are notoriously difficult to study experimentally due to their need to be in a "native" membrane environment, nanodiscs are a great tool, furnishing a membrane environment that has been used to embed a variety of membrane proteins for biochemical assay, including cytochrome P450's, rhodopsin, bacterial chemoreceptors, blood clotting factors, and translocation proteins. Unfortunately, it is difficult to make either real or molecular sushi rolls (nanodiscs). In either case one needs to lay down the ingredients first. In the case of nanodiscs, one starts from the raw ingredients which are solubilized by the detergent cholate. Removing the detergent allows the nanodiscs to self-assemble. However, the assembly process is difficult to quantify experimentally, thus researchers rather studied the disassembly process, i.e., how detergent disassembles preformed nanodiscs. One can watch a sushi chef make rolls, but watching the disassembly and assembly of nanodiscs is harder due the the small size. Fortunately, a computer can image the process. In a recent publication, nanodisc disassembly through the addition of increasing concentrations of cholate was monitored through computer simulations using NAMD and verified through experimental small-angle X-ray scattering. The study showed how cholate molecules insert themselves at the interface between the lipids and lipoproteins towards complete disassembly. The simulations employed a new method called residue-based coarse-graining. For more information, see our webpage on nanodiscs.

Mechanical forces are everywhere in human life. Strong forces power machines and cars, our body's forces let us labor and move, soft forces are sensed through touch, even softer ones through hearing. Forces are also ubiquitous in the living cell, driving its molecular machines and motors as well as signaling ongoing action in its surroundings. Man made, force bearing machines rely on extremely strong materials not found in the cell. How can the cell bear substantial forces? Also, how do cells sense extremely weak forces as in hearing, surpassing most microphones? Single molecule measurements, reviewed in a recent issue of Science, begin to answer these questions offering information on biomolecules' mechanical responses and action. However, the information offered by these measurements is not enough to relate the biomolecular function to the biomolecular architecture. Biomolecules in cells can move in amazing ways, but we did not know why. As one contribution in the Science issue demonstrates, computational modeling comes to the rescue. It can simulate the measurements and, in doing so, can reveal the physical mechanisms underlying cellular mechanics at the atomic level. In as far as observed data are available, the simulations show impressive agreement with actual measurements. While initially only following experiments or, at best, guiding experiments, modeling has advanced now further and through simulated measurements discovered on its own entirely novel mechanical properties that were later verified by experimental measurements. Experimentalists reacted to the new competition and began to do simulations themselves. More here.

PSC News Release: University of Utah chemist Gregory Voth and grad student Phil Blood are using PSC’s Cray XT3 to tackle a basic question of endocytosis—the life-sustaining process by which cells absorb material from outside the cell by bending their membrane to form a “vesicle” and engulf it. All animal cells depend on endocytosis, which involves various steps, but begins with curvature of the membrane.

BAR domains are a family of banana-shaped proteins shown to bind to cellular membrane as it curves. Experiments suggest that BAR domains mold their concave surface to a section of membrane and induce a corresponding curvature. Voth and Blood undertook molecular dynamics simulations to look more closely. With the XT3 they’ve been able to run efficiently, using software called NAMD, with as many as 1,024 processors. “The XT3 has been amazing,” says Blood. “We haven’t found a hard limit on scaling up the number of processors.”

They used TeraGrid systems at SDSC, NCSA and University of Chicago/Argonne to construct a model and to explore how long a stretch of membrane they needed for curvature to occur. Their final simulations used the XT3 to include the protein with a 50-nanometer length of membrane—probably the longest patch of membrane ever simulated—for a total of 738,000 atoms. Their results, reported in Proceedings of the National Academy of Sciences (2006), show that the orientation of the BAR domain as it attaches to the membrane determines the degree of curvature.

SDSC News Release: SDSC and UC San Diego researchers are using NAMD to zero in on the causes of Parkinson's disease, Alzheimer's disease, rheumatoid arthritis and other diseases. The April 2007 FEBS Journal cover story offers—for the first time—a model for the complex process of aggregation of a protein known as alpha-synuclein, which in turn leads to harmful ring-like or pore-like structures in human membranes, the kind of damage found in Parkinson's and Alzheimer's patients. The researchers also found that the destructive properties of alpha-synuclein can be blocked by beta-synuclein—a finding that could lead to treatments for many debilitating diseases.

Lead author Igor Tsigelny, SDSC researcher and project scientist in chemistry and biochemistry at UCSD, said that the team's research helped confirm what researchers had suspected. “The present study—using molecular modeling and molecular dynamics simulations in combination with biochemical and ultrastructural analysis—shows that alpha-synuclein can lead to the formation of pore-like structures on membranes.” In contrast, he said, “beta-synuclein appears to block the propagation of alpha-synucleins into harmful structures.”

“This is one of the first studies to use supercomputers to model how alpha-synuclein complexes damage the cells, and how that could be blocked,” said Eliezer Masliah, professor of neurosciences and pathology at UC San Diego. “We believe that these ring- or pore-like structures might be deleterious to the cells, and we have a unique opportunity to better understand how alpha-synuclein is involved in the pathogenesis of Parkinson's disease, and how to reverse this process.”

Every morning, many people drive to work, while others may bike, take the bus or the metro. Similarly, various biomolecules in the human body also reach their destinations in diverse manners. For example, to cross the cellular membrane, small hydrophobic gas molecules diffuse through the lipid bilayer, while water molecules pass through specialized channel proteins named aquaporins (AQPs). Interestingly, just like one may get to work both by bus and by driving, it has been found recently that some gas molecules may have more than one way to cross the membrane, i.e., besides diffusion through lipids, oxygen and carbon dioxide may also pass through AQPs. However, the pathways that these gas molecules take remained elusive. Using molecular dynamics performed with NAMD, researchers have investigated the gas permeability of AQP1 in a recent study with two complementary methods (explicit gas diffusion simulation and implicit ligand sampling). The simulation results suggest that while the four monomeric pores of AQP1 function as water channels, the central pore of AQP1 may serve as a pathway for gas molecules to cross the membrane. More information on the simulations can be found on the aquaporin web page.

High-density lipoproteins, otherwise known as the "good cholesterol", are the body's way of naturally removing cholesterol in the blood stream. Since lipid and cholesterol molecules are not soluble in blood, lipoproteins are needed to collect and transport them. The proteins wrap themselves around the hydrophobic portions of lipids and cholesterol, effectively shielding them from the aqueous environment and allowing them to be transported through the bloodstream to the liver for degradation. High-density lipoproteins exhibit a variety of shapes and sizes and presently cannot be imaged through experimental observations. Computational methods, however, can provide detailed images of high-density lipoprotein particles, even showing how these particles form in the body. As recently reported (article 1, article 2), so-called coarse-grained molecular dynamics simulations using NAMD discovered how lipid molecules are corralled by lipoproteins to form disc-like high-density lipoprotein particles. The simulations show in remarkable detail the aggregation of proteins and lipids, starting from a random arrangement of lipids that are mopped-up by two lipoproteins, eventually forcing the lipids into a disc-shape surrounded on its circumference by belt-like lipoproteins. For more information, see our webpages on HDL & nanodiscs and coarse-grained modeling.

Bacteria employ membrane proteins as crucial safety valves that release water and small solutes under challenging osmotic conditions (see the May 2006 highlight, "Electrical Safety Valve" and the Nov 2004 highlight, "Japanese Lantern Protein"). There are valves for balancing small pressure differences between the inside and outside of bacterial cells, that open and close readily, but there are also ones for protection against large pressure differences as a safety measure of last resort. The valves for balancing small pressure differences, like the one shown in the figure, include a filter that presumably keeps the most valuable molecules inside the cell interior, though this is not understood yet in detail. To reveal the function of such channels a combination of X-ray crystallography, physiological measurements, and molecular dynamics simulations using NAMD has been employed. Crystallography, in a prior study, captured the channel in a half-way open state. Now a team of physiologists and modelers reported the details on valve opening and closing. The experiments, using a pipette small enough to measure currents from a single channel, MscS, along with the simulations revealed that the channel conducts both positive and negative ions when subjected to tension and voltage. The unprecedented comparison of experimental and computational results open a new era of quantitative cell biology that borrows successful research strategies from physics (more on our MscS website).

The nucleus is responsible for storing the genome of eukaryotic cells, isolating it from the cellular cytoplasm. Partitioning the genetic material is very important in protecting it from cellular processes or foreign molecules. However, the nucleus also needs to provide access for the rest of the cell to the information stored in the genome. Numerous nuclear pores in the nuclear envelope offer communication pathways between the nucleoplasm and cytoplasm. The pathways are restricted to so-called transport receptors, proteins that taxi molecules into and out of the nucleus. If a molecule wishes to enter or leave the nucleus, it associates with a transport receptor. The complex passes through the pore and then dissociates. The question is why transport receptors can pass the nuclear pores while other proteins cannot. The answer lies in the role of FG-repeat proteins lining the pores and filling much of their free volume. These proteins are disordered peptides, consisting of repeating phenylalanine-glycine (FG) residues separated by a sequence of hydrophilic linker residues. Only proteins that interact favorably with the FG-repeat regions can pass through, while other proteins are excluded. A recent report used molecular dynamics via NAMD to examine the way in which the transport factor NTF2 interacts with the FG-repeats. The study described binding spots for FG-repeat peptides on the surface of NTF2, confirming known binding spots discovered previously via experimental means, and suggesting the existence of further binding spots. The new binding spots may play a role in steering NTF2, upon import or export, along a particular path through the nuclear pore. See also a previous highlight from January 2006, "Gateway to the Nucleus", as well as our webpage on the nuclear pore complex.

The bacterial flagellum is a large biomolecular assembly used by many types of bacteria as a helical propeller for forward swimming and turning. The flagellum is remarkable in that its properties differ greatly depending on the direction in which it is rotated, allowing the bacterium to switch between swimming straight ("running") and turning ("tumbling"). The mechanics of the flagellum are of interest both to biologists and mechanical engineers. The molecular mechanisms of the transition in the flagellum between running and tumbling modes is unknown. Because of the flagellum's size (several micrometers in length) and composition (made up of 30,000 protein subunits) it presents a challenge to computational modeling. Researchers have now achieved an advance describing the flagellum in both its running and tumbling state. For this purpose, the researchers developed a computational model of the system that glosses over atomic level detail, but resolves the shapes of all proteins making up a bacterial flagellum, simulating a simplified version of the system using the program NAMD. The results, reported recently, showed that the flagellum's transition between swimming straight and tumbling is triggered by friction due to the water around the bacterium. More information on the flagellum project can be found here.

Viruses are the cause of many human diseases, from the common cold to AIDS, and medicine is continuously searching for better ways to battle viruses through vaccination or medication. Detailed knowledge of the life cycles of viruses should be useful in the treatment of viral diseases. A key focus of investigations is the virus capsid, a protein coat that protects the viral genome, but also triggers release of the genome and other viral factors upon contact with the body's cells. X-ray crystallography has resolved the average structures of many types of virus capsids, providing the basis for detailed investigations, for example by means of molecular dynamics methods, of capsid dynamical properties, e.g., in assembly and disassembly. Unfortunately, due to their large size most virus capsids are beyond the reach of molecular dynamics simulations, with one notable exception (see the March 2006 highlight "Simulating an Entire Life Form"). This earlier simulation allowed researchers to develop and test a method for coarse-grained molecular dynamics simulations that glosses over atomic detail and, thereby, permits microsecond descriptions of entire viral particles. As reported recently (see also journal cover) such simulations, employing the program NAMD, were applied to the empty capsids of several viruses. These simulations revealed a variety of behaviors, from rapid collapse to high stability, depending on the strength of interactions between the proteins from which capsids are built. The new method offers unprecedented views of capsid dynamics that may assist in battling viral diseases. More information on the simulations can be found on our virus web page.

Bionanotechnology involves a marriage of two different materials: inorganic solids, like silica, and biomolecules, like DNA. The new combinations of materials have to be mastered on the laboratory bench as well as in computer simulations. On the bench, devices are manufactured and tested, in the simulations, they are imaged and designed. So far inorganic solids and biomolecules were simulated successfully, but only separately. To join the materials requires as much effort in simulations as on the bench. Even just the interaction of inorganic solids (like silica) with physiological solutions (water and ions) demands challenging descriptions of silica surface properties. As reported recently, researchers have now succeeded in describing accurately the wetting (by water) of amorphous silica, an essential material for nanoelectronics, clearing a major hurdle to simulating bionanotechnological devices, for example, those suggested for rapid and economical sequencing of DNA (see also Nov 2005 and Oct 2004 highlights). More on silica-water interaction here.

Escherichia coli are bacteria living in the intestines of mammals as part of their healthy gut flora, but also causing disease outside of the gut. The bacteria import from their environment nutriments, for example molecules of lactose, a sugar. For this purpose Escherichia coli employs in its cell membrane a protein channel, lactose permease, that translocates the sugar outside-in. This is the bacterium's "sweet tooth". To establish the unidirectional sugar transport, the bacterium utilizes an electrical potential maintained in the form of a trans-membrane proton gradient (more protons on the outer cellular than on the inner cellular side of the membrane). Protons, very small ions, that enter the channel from the outside one at a time, open the outer channel entrance. This permits access of lactose that gets bound inside the channel. Release of the proton to the cell interior closes the outer channel entrance and opens the inner channel entrance, such that the bound lactose can enter the cell. Despite extensive and elegant biochemical studies, the physical mechanism that couples unidirectional proton and sugar translocation is not yet known in detail. A crystallographic structure of lactose permease permitted now investigations into this mechanism by means of molecular dynamics simulations using NAMD. The simulations, reported in a recent publication, showed one step of the proton - sugar translocation, namely how binding and unbinding of the proton activates a spring-like bond, a so-called salt bridge, that closes and opens the inner channel exit. More information on the lactose permease project can be found here.

Mammalian cells adhere to each other forming tissues. The adhesion is due to a network of proteins, so-called extracellular matrix proteins, "gluing" the cells together. The cell membranes are too soft to provide anchoring points for the extracellular matrix proteins; rather, the cells furnish on their outer surface specialized hooks for anchoring the extracellular matrix proteins. The hooks, in the form of surface proteins, are linked directly through the membranes to the intracellular cytoskeleton that stabilizes and shapes cells. Integrins are an important family of such surface proteins that form hooks specific for certain types of extracellular matrix proteins. The hooks are flexible, they can be open for contacts or closed, the switch being induced by signals from inside or outside the cell through interactions with other proteins. The interactions between integrins and extracellular matrix proteins are rather complex, as the proteins are composed of many subunits; fortunately, their overall structures are presently being solved through crystallography. In a recent report a major component of an integrin and an extracellular matrix protein have been investigated through molecular modeling using NAMD, including steered molecular dynamics. The study described in detail how the extracellular matrix protein induces a transition in integrin, potentially strengthening its adhesion property. See also previous highlights: the May 2006 "Killer's Entry Route", Dec 2004 "Snap Fastener on Biological Cells", Dec 2003 "Body's Glue", and Mar 2002 "Cells Sense Push and Pull". More on modeling of extracellular matrix proteins and integrins can be found here.

Most forms of life need to detect and respond to changes in their environment for survival and optimal growth. For this purpose organisms rely on receptors that are based on sensory proteins. In plants, several sensory proteins detect the ambient light for optimal exposure of their photosynthetic apparatus. One class of plant light sensors, the phototropins, influence photosynthesis and induce the transition between root and stem growth when seedlings emerge out of the ground. Induction is activated through several protein domains, two of which actually absorb light and for their sensitivity to light, oxygen, and voltage, are called LOV1 and LOV2 domains. Understanding the LOV domains' involvement in activation is important for studying the signaling mechanisms of other types of sensory proteins. Strangely, light absorbed by LOV domains is observed to lead to a distinct, but only very minute, structural change that does not explain how activation might come about. NAMD-based molecular dynamics simulations of the LOV domain have now revealed, as reported in a recent publication, that photoactivated LOV domains exhibit altered patterns of motion that can induce a signal for plant cells. More information may be found on our biological photoreceptors website.

Sometimes analogies go a long way, surprisingly long.  Aquaporins are ubiquitous water channels in living cells, known to be tetrameric, each unit contributing one pore.  This much is certain and this is where an analogy begins, namely with a British spy ring that passed information to the Soviet Union during world war II and into the 1950s.  The ring is often referred to as the Cambridge Four since the spies, when recruited, were undergrads at Cambridge Trinity College and there were four of them (cryptonyms Stanley, Homer, Hicks, and Johnson).  But a Fifth Man was long suspected, yet never formally identified. Here the analogy continues: aquaporin was suspected to sport a fifth pore, supposedly at its center, where its four subunits join (hence known as the tetrameric pore).  Strong, but not yet completely conclusive,  evidence has now been put forward in a recent report that the central pore, actually quite plainly visible to the eye when aquaporins are inspected by molecular graphics, e.g., with VMD, is an ion channel gated by a common cellular signaling molecule, cGMP. The evidence stems from a combined computational (molecular dynamics using NAMD) and experimental (verifying computationally suggested mutants) study.  More information on the five pores can be found on our aquaporin website, more on the Cambridge Five here. But the analogy goes further. Today it is suspected that the Cambridge Five actually had more than five members and the same holds for the pores of aquaporin.  An ongoing investigation has lead to evidence that the further pore members conduct gases, for example carbon dioxide.  Hopefully, we will know one day with certainty all members of the Cambridge Five and all pores of aquaporin.

The computer processor is the workhorse of biomolecular modeling, with NAMD the plow to which a single processor or a team of thousands may be hitched. The recent release of NAMD 2.6 has extended the drawbar to harness the power of several thousand processors: 2000 on a Cray XT3 and 8,000 on an IBM Blue Gene/L. This permits the efficient simulation of an entire ribosome, the cell's protein factory, comprising 3,000,000 atoms when solvated. But the features and increased performance of NAMD 2.6 are also available to the scientist with only a laptop, on which a domain of the muscle protein titin (10,000 atoms solvated) can be readily simulated. NAMD has also become more versatile, supporting more force fields (OPLS, CHARMM with CMAP cross terms), calculating free energies, and executing customizable replica exchange simulations. In addition, NAMD can now be called from the structure analysis program VMD to calculate, for example, interaction energies between protein domains. Like increased horsepower in transportation, increased simulation power opens new routes, routes to study entire systems of biopolymers like the ribosome, not just one piece.

DNA with its famous double helix structure stores the genetic information of all life forms known. In order that this information is read, the double helix needs to be first unwound and separated into single helices or strands. This is achieved by cellular motor proteins called helicases that operate on already separated DNA strands. The helicases specialize in unwinding and separating the DNA double helix by scooting along one of DNA's single strands against the point where the two strands merge into the double helix; pushing against this point unwinds and separates the double helix further. The helicases are driven by energy stored in molecules of ATP which bind to the protein and get released in their so-called hydrolyzed, lower energy, form. Based on atomic resolution structures, researchers have studied now one of the smallest helicases known, PcrA, from the electronic to the functional level carrying out quantum mechanical/molecular mechanical simulations (as described in a first publication), as well as a combination of classical molecular dynamics simulation, using NAMD, and stochastic modeling calculations (described in a second publication and a third publication). This resulted in an overall explanation of how ATP's hydrolysis powers helicase activity which has been reported in a fourth publication. The researchers discovered that PcrA moves with two "hands" along single stranded DNA; when ATP binds, one "hand" moves along the DNA; when ADP and Pi (the hydrolysis products of ATP) unbind, the other "hand" moves; through a molecular "trick" both "hands" move in the same direction. Amazingly, the hand movement arises mainly from an increase in random mobility of the hands. i.e., is not enforced. Physicists refer to the underlying mechanism as a ratchet mechanism that was indeed long suspected to drive molecular motors. Interestingly, the helicase motor is very closely related to a wide class of other biological motors, for example the FoF1-ATP synthase (see Mar 2004 and Nov 2004 highlights). For more information visit our helicase research website.

Biological cells, in particular neurons, maintain an inside-outside voltage gradient through active transport of ions (Na⁺, K⁺, Cl^-, and others) across their membranes. The flow of the ions down their gradients through membrane channels is highly selective for each ion. The high selectivity permits nerve cells to signal each other through voltage spikes, which are produced through transient changes of channel conductivities for Na⁺ ions (channels open and close in about a ms) and K⁺ ions (channels open and close in about 10 ms). Crucial for the generation of voltage spikes is the selective, yet quick, conduction of ions, but as one knows from personal experience at border crossings, high selectivity and quick crossing seem to be mutually exclusive. Yet biological ion channels reconcile selectivity and speed. Prior experimental work, primarily that of year 2003 Nobelist MacKinnon, as well as computational work suggested how potassium channels achieve selectivity and speed. But until recently no high resolution atomic structure of a potassium channel was known in the open form and the suggested mechanism could not be tested under natural conditions through atomic level simulations. Last year's solution of the structure of the potassium channel Kv1.2 in its open form made it finally possible to simulate, using NAMD, the conduction of ions through Kv1.2 driven by a voltage gradient. The results reported recently confirmed indeed the high selectivity - high speed mechanism suggested earlier, namely a billiard-type motion of two and three ions, the last ion kicking the first ion out. The simulations revealed for the first time, through movies, the overall permeation process, including the jumps of ions between energetically favorable binding sites and the sequence of multi-ion configurations involved in permeation. More on our potassium channel web site.

How far and how well molecular biologists can look into the living cell depends as much on microscopes and observations, as it depends on computers and their software. The premiere software for looking into the molecular world of the cell, VMD, has made a big leap forward in broadening the molecular horizon of life scientists through its new release, VMD 1.8.5. Researchers are offered now a fresh view through a modern unified bioinformatics environment, MultiSeq, combining sequence and structure analysis for proteins and amino acids. VMD, now literally more colorful, lets scientists quickly exchange VMD views through integration of BioCoRE, calculate APBS electrostatics maps, call on NAMD to calculate energies, build and mutate structures, determine easily force field parameters, and navigate through proteins with a flying camera. VMD 1.8.5, though only a minor version number different from the previous release, includes now many new structure building and analysis tools that make it easier for modelers to set up, run, and analyze computer simulations of biomolecules.

Many proteins interact with gas molecules such as oxygen to perform their functions. In most cases, the gas molecules must reach sites buried deep inside the proteins that bind the molecules, with no obvious way in. Understanding how, for example, oxygen enters the protein, and mapping out which pathways it takes has been a long-standing challenge. As reported recently, computational biologists, inspired by previous work on the hydrogenase enzyme (see the September 2005 highlight), have developed a method, called implicit ligand sampling, that maps the pathways taken by gas molecules inside proteins. The mapping is determined by monitoring fluctuations of the protein, surprisingly, in the absence of the gas molecules. The mapping method is available in the most recent version of the program VMD used for structure and sequence analysis of proteins. The researchers applied the method to myoglobin, an oxygen-storing protein present in muscle cells, and determined detailed three-dimensional maps of oxygen and carbon monoxide pathways inside the protein (for more information see our web page). While some details of these pathways were already known from experiment, the implicit ligand maps revealed a large number of new pathways and suggest that oxygen enters myoglobin using many different entrance doors.

The living state of biological cells manifests itself through mechanical motion on many length scales. Behind this motion are processes that generate and transform mechanical forces of various types. As with other cell functions, the machinery for cellular mechanics involves proteins. Their flexible structures can be deformed and restored, and are often essential for handling, transforming, and using mechanical force. For instance, proteins of muscle and the extracellular matrix exhibit salient elasticity upon stretching, mechanosensory proteins transduce weak mechanical stimuli into electrical signals, and so-called regulatory proteins force DNA into loops controlling, thereby, gene expression. In a recent review, the structure-function relationship of four protein complexes with well defined and representative mechanical functions has been described. The first protein system reviewed is titin, a protein that confers passive elasticity on muscle. The second system reviewed is the elastic extracellular matrix protein fibronectin and its cellular receptor integrin. The third protein system covered are the proteins cadherin and ankyrin involved in the transduction apparatus of mechanical senses and hearing. The last system surveyed is the lac repressor,  a protein which regulates gene expression by looping DNA. In each case, molecular dynamics simulations using NAMD provided insights into the physical mechanisms underlying the associated mechanical functions of living cells. (more on our mechanobiology web site).  

In a recent paper experiences designing and deploying NAMD-G, an infrastructure for executing biomolecular simulations using the parallel molecular dynamics code NAMD within the context of a Computational Grid, are described. The NAMD-G project, and associated paper, is the result of a collaboration between the Theoretical and Computational Biophysics Group (TCBG) and the National Center for Supercomputing Applications (NCSA). James Phillips, Senior Research Programmer for NAMD, and Jordi Cohen, a Ph.D. candidate in Physics at the University of Illinois at Urbana-Champaign, contributed on behalf of the TCBG, while Richard Kufrin, Senior Research Programmer, and Michelle Gower, Research Programmer, led development for NCSA.

NAMD (NAnoscale Molecular Dynamics) can simulate the movement of proteins with millions of atoms, making it the world's fastest parallel molecular dynamics program. The NAMD development team will continue to incorporate the latest parallel-computing advances into NAMD, which already runs efficiently on several thousand parallel processors. Pictured to the right is the current development team for NAMD. James Phillips, a PhD in Physics, is Senior Research Programmer for NAMD at the Theoretical and Computational Biophysics Group (TCBG). Dr. Laxmikant Kale, Professor of Computer Science at UIUC, is a Co-PI at TCBG and leader of the Parallel Programming Laboratory. Kale's computer science graduate students, Chee Wai Lee, Chao Mei, and David Kunzman, also contribute to various aspects of NAMD development.

Many proteins store gases like oxygen, carbon dioxide, and nitric oxide, or react with them. The gases are conducted into the protein through access routes that exist only in passing and as a result of a protein's fluctuations. Accordingly, access routes are difficult to establish, but researchers are now able to image gas access pathways inside proteins computationally. The new method has many implications for biotechnology and science (see our hydrogenase page and Sep 2005 highlight, "Hydrogen Fuel from Protein"). Imaging gas access systematically over whole protein families, e.g., the family of myoglobins, requires a large number of calculations that need to be run and monitored. The traditional means of doing so is very wasteful of the researchers' time. To solve this problem, NAMD-G, a grid-based automation engine for biomolecular simulations running the NAMD software, has been developed in a collaboration with the National Center for Supercomputing Applications (see recent paper). From the researchers' workstations, NAMD-G "farms out" a large number of calculations, in parallel, to supercomputers on the TeraGrid. NAMD-G monitors and manages multiple sequences of calculations at distant sites, and performs the necessary data transfers and backups on an as-needed basis. While the gas transport simulations provide a clear scientific driver for the development, NAMD-G is quite general and will aid any NAMD user with access to the TeraGrid. The result? Less time spent baby-sitting runs and more time for science.

For hard-working scientists, the task of maintaining a single desktop computer is an unwelcome distraction. But what if your work requires the power of ten or a hundred machines? Our recent series of workshops (Sep 2005, Nov 2005, Mar 2006, and Apr 2006) has given nearly one hundred participants hands-on experience installing and using low-cost Linux clusters. Students were taught to eliminate many sources of complexity, such as hard drives, and to automate what remained with cluster management software and a queueing system. Lectures on cluster design stressed the importance of knowing which applications would be run and choosing cost-effective hardware to meet those specific needs, as well as less-obvious aspects of cluster acquisition such as electrical power, cooling, and the purchasing process. After assembling and installing small four-node clusters, students ran both the molecular dynamics program NAMD and a more typical parallel application that they compiled from scratch. Most participants were motivated by concrete plans to build clusters for their own groups in the near future and felt better-equipped to do so following their experience.

Bacterial cells, like those of Escherichia coli, protect themselves against sudden inside-out pressure differences that arise osmotically from changes in a cell's environment and that could burst the cellular envelop. The protection is achieved through so-called mechanosensitive channels in the cell membrane.  One such channel, that dissipates like a safety valve pressure differences across the Escherichia coli cell membrane, is contributed by the protein MscS.  Upon tension in the cell membrane, that can also be applied systematically in the laboratory, the channel opens and permits molecules to pass, as best measured through an ion current leaking through the stretched membrane.  MscS is a channel with a balloon-like filter,  the function of the latter being still a mystery  (see Nov 2004 highlight, "Japanese lantern protein").  Now computational biologists using NAMD teamed up with device engineers using BioMoca to study MscS as reported recently.  The team monitored the mysterious MscS computationally over several microseconds, a record time for protein simulations.  MscS was found to permit water passage, but to also exhibit strong electrostatic forces that focus ions streaming through its filter balloon and channel.  This suggests MscS to be both a hydrostatic and an electrical safety valve.  Even though now better known, MscS' entire function remains shrouded in mystery (more on our MscS web site).

Bacillus anthracis, the cause of anthrax, is one of the most lethal bacteria. In addition to its ability to infect animal and human cells, the bacterium attacks also the cells of the host's immune system, the so-called macrophages. For this purpose the anthrax bacterium releases three types of proteins, or toxins, into the blood stream of the host: protective antigen, lethal factor, and edema factor, referred to as PA, LF, and EF, respectively. LF and EF team up with PA, which transports them into a host macrophage cell. Once inside the cell, LF converts ATP to cyclic AMP, while EF disables MAPKKs, a family of signaling proteins. These attacks disrupt various cellular signaling pathways of macrophages and some other cells, essentially shutting down the host's immune system and often leading to death of the host. To invade macrophages, the toxins take an intricate entry route that involves binding to a cell receptor, capillary morphogenesis protein 2 (CMG2), inducing the cell to internalize the toxins in a bubble like membrane (endosome), the bubble wall being then punctured by seven PAs forming a channel upon a chemical (acidifying) trigger from the host; the channel permits then their lethal cargo, LFs and EFs, to slip into the cell. How exactly the PAs punctured the endosome wall remained a mystery. In a recent report the entry route has been resolved now in greater detail through molecular dynamics simulations using NAMD. The report reveals how acidic conditions in the endosome trigger conformational changes of the PA complex necessary for pore formation, and provides structural insights into the role of unusual interactions between the PAs and its receptor CMG2. Visit also our anthrax toxin webpage.

Computational tools, like the molecular graphics program VMD and molecular dynamics program NAMD, move rapidly from theoretical to experimental biology. To train researchers in the proper use of computational tools, a series of hands-on workshops was organized in the US and Australia in 2003-2005 (see July 2005 highlight). This year the first European hands-on workshop started a new generation of training with three novel features. First, the workshop addressed mainly bench scientists in need of computational methods. Second, the workshop introduced a key expansion of VMD that turned a mainly structurally oriented visualization program into a structure and sequence analysis program. This is achieved through a multiple sequence analysis tool in VMD, called multiseq. Third, all training material has been extended to multiple platforms and participants could bring their own laptops for the training sessions. As in the previous series, participants enjoyed workshop lectures that introduced concepts and good uses of biocomputing software, but were most enthusiastic about practical tutorials that provided opportunities to learn by example and to apply newly mastered tools to their own research. The participants carried all lecture material and software home on a DVD; others can obtain the same material through our web site (workshop lectures, tutorials, case studies, VMD, NAMD).

Anyone who has attempted to fit a long piece of thread through a needle's eye realizes how difficult fitting something so small and flexible into such a small hole can be. Yet this action is carried out every second in every living cell. Flexible polypeptides, proteins, often have to cross a cellular membrane to get to their correct location, whether that location is an organelle within the cell or even outside of it. To accomplish this, they are pushed through a protein pore in an unfolded conformation much like a long string. The channel that accepts the string-like proteins, the protein translocon, allows only certain proteins to pass, while restricting access to molecules even much smaller than the macromolecular proteins. As reported in a recent publication, computer simulations using the molecular dynamics program NAMD helped to answer the question of how such a small channel could achieve this feat, demonstrating how the channel itself can be flexible yet resilient during a protein-crossing event and also elucidating in part how it can maintain such tight control over what is permitted to cross. For more information, see our Protein Translocation website.

Muscle fibers, through their so-called thick and thin filaments, contract and extend in doing their work. To render the fibers elastic and protect them from overstretching, the thick filaments are connected through a long and thin elastic protein, titin, to the base of the fibers. Titin, by far the longest protein in human cells, is a molecular bungee cord and, like such cord, must be affixed firmly to the base. How this is done was a mystery until crystallographers took the first atomic resolution image of the system: it turns out that two titins are spliced together at their ends like ropes. The splicing involves a third small protein, the titin-telethonin-titin system forming a U. The U apparently is thrown over a bollard-like cellular structure to hold the thick filaments much like boats are held by bollards and ropes at their mooring place. The crystallographers teamed up with computational biologists to investigate the mechanical strength of the titin - telethonin - titin cord by means of molecular dynamics simulations using NAMD. As reported recently, the cord has great mechanical strength due to an extended network of hydrogen bonds between beta-strands, common structural features in proteins, that in the present case form a sheet extending through all three proteins. This discovery explains how living cells can splice cellular proteins together through a system of hydrogen-bonded beta-strands that extend through several proteins. Interestingly, such beta-strands were seen previously in cases of diseases like Alzheimers where the feature leads, however, to pathological assembly of proteins. What needs to be understood now is how the telethonin glue is applied only to the right spots in the cell and how the cells prevent telethonin from splicing together the wrong proteins. For more information visit our titin-telethonin web page.

Viruses, the cause of many diseases, are the smallest natural organisms known. They are extremely primitive and parasitic such that biologists refer to them as "particles", rather than organisms. Viruses contain in a protein shell, the capsid, their own building plan, the genome, in the form of DNA or RNA. Viruses hijack a biological cell and make it produce from one virus many new ones. Viruses have evolved elaborate mechanisms to infect host cells, to to produce and assemble their own components, and to leave the host cell when it bursts from viral overcrowding. Because of their simplicity and small size, computational biologists selected a virus for their first attempt to reverse-engineer in a computer program, NAMD, an entire life form, choosing one of the tiniest viruses for this purpose, the satellite tobacco mosaic virus. As described in a recent report, the researchers simulated the virus in a small drop of salt water, altogether involving over a million atoms. This provided an unprecedented view into the dynamics of the virus for a very brief time, revealing nevertheless the key physical properties of the viral particle as well as providing crucial information on its assembly. It may take still a long time to simulate a dog wagging its tail in the computer, but a big first step has been taken to simulate living organisms. Naturally, this step will assist modern medicine (more on our satellite tobacco mosaic virus web page).

Lipoproteins are protein-lipid particles which circulate in the blood collecting cholesterol, fatty acids, and lipids. Low levels of one such lipoprotein particle, called high-density lipoprotein (HDL) or "good cholesterol", has been implicated in the increased risk of coronary heart disease. The ability of lipoproteins to transport lipid and cholesterol through the blood is amazing since these types of particles are not generally soluble in blood plasma. However, when HDLs assemble, proteins wrap themselves around the lipids and cholesterol, shielding the lipid tails from the aqueous environment. Native HDL exhibit a variety of shapes and sizes, for example forming a discoidal particle. Conventional high-resolution imaging techniques, such as NMR and X-ray crystallography, cannot resolve how lipid and cholesterol are being accommodated by HDL, but the assembly and geometry of HDL discs can be captured using computer simulations. Unfortunately, the long time scales required for HDL assembly was a major stumbling block. Now a new simulation method, coarse-grained modeling in conjunction with the molecular dynamics program NAMD, has permitted the simulation of HDL assembly as recently reported. The simulations show that lipids quickly aggregate into a bilayer from their initial spherical "micelle" shape and that the two proteins subsequently attach to either side forming a belt surrounding the lipid core. For more information see HDL & nanodisc and coarse-grained modeling.

Molecular modeling with NAMD (NAnoscale Molecular Dynamics) promises to become a key methodology for research and development in bionanotechnology. Molecular modeling provides nanoscale images at atomic and even electronic resolution, predicts the nanoscale interaction of yet unfamiliar combinations of biological and inorganic materials, and can evaluate strategies for redesigning biopolymers for nanotechnological uses. The methodology's value has been reviewed for three uses in bionanotechnology. The first involves the use of single-walled carbon nanotubes as biomedical sensors where a computationally efficient, yet accurate description of the influence of biomolecules on nanotube electronic properties and a description of nanotube - biomolecule interactions were developed; this development furnishes the ability to test nanotube electronic properties in realistic biological environments (see Dec 2005 highlight). The second case study involves the use of nanopores manufactured into electronic nanodevices based on silicon compounds for single molecule electrical recording, in particular, for DNA sequencing. Here, modeling combining classical molecular dynamics, material science, and device physics, describes the interaction of biopolymers, e.g., DNA, with silicon nitrate and silicon oxide pores, furnishes accurate dynamic images of pore translocation processes, and predicts signals (see Nov 2005 and Oct 2004 highlights). The third case involves the development of nanoscale lipid bilayers for the study of embedded membrane proteins and cholesterol. Molecular modeling tested scaffold proteins, redesigned lipoproteins found in mammalian plasma that hold the discoidal membranes in shape, and predicted the assembly as well as final structure of the nanodiscs (see Feb 2005 highlight). In entirely new technological areas like bionanotechnology qualitative concepts, pictures, and suggestions are sorely needed; the three exemplary applications document that molecular modeling can serve as a critical "imaging" method for bionanotechnology, even though it may still fall short on quantitative precision.

Your favorite flower pot would not survive a weekend in your office without watering, if it wasn't for a sophisticated cellular mechanism evolved in land plants to conserve water under drought conditions. Water exchange between cells and their environment is facilitated by a group of highly specialized membrane proteins called aquaporins. Although present in all life forms, plants are particularly dependent on their function. While in most species these channels function as always-open "cellular pipes" allowing water in and out of the cell, in plants they evolved into "cellular faucets" whose water permeability can be controlled by the cell. Nearly all plant aquaporins can be gated in response to drought or even flooding conditions, through basic biochemical signals, e.g., phosphorylation and change of pH. A recent Nature paper reporting a collaborative study between crystallographers who succeeded in solving the first structure of a plant aquaporin from spinach, and modelers provides the most detailed view of the mechanism of gating for a membrane channel. Molecular dynamics simulations of the channel performed by NAMD reveals a dual gating mechanism in which phosphorylation of certain protein residues unleashes a long cytoplasmic loop that physically blocks water access to the pore. More information on aquaporin research can be found here.

Eukaryotic cells envelop their genetic material in the cell nucleus whose boundary contains numerous pores. Only small molecules can pass through these nuclear pores unhindered. For all larger ones, passage is highly selective and controlled. The control involves import and export proteins (transport receptors) that load and release cargo on the proper side of the nucleus upon interaction with signaling proteins. Researchers are presently solving the structure of the nuclear pore and its transport receptors with increasing resolution, and the first atomic level investigation into the mechanism of nuclear pore selectivity has recently been reported [paper]. The study inspected the interaction between the transport receptor importin-β with key nuclear pore proteins that appear disordered near the center of the pore and contain characteristic phenylalanine-glycine sequence repeats. Molecular dynamics simulations using NAMD and analyzed using VMD revealed a key insight into the selectivity mechanism. The simulations showed that the key sequences of the repeat proteins interact strongly with certain spots on the surface of importin-β. The study confirmed spots that had previously been identified experimentally and, moreover, found numerous binding spots not yet seen in experiment. Further experiments and simulations promise an understanding of the selectivity of entry and exit from the nucleus, a key element of the cell's genetic control. For more information see our nuclear pore complex webpage.

The most celebrated molecule of living cells, DNA, owes its fame to its role as a carrier of genetic information. But DNA is also impressive through other amazing properties, for example its mechanical flexibility. At first sight, it might seem a dull question to ask what is the smallest pore DNA can be squeezed through, as the obvious answer is that the diameter of that pore should be slightly larger than the diameter of a DNA helix. However, recent studies (paper1, paper2) in asking the stated question discovered that double stranded DNA can permeate, without loosing its structural integrity, pores smaller in diameter than a DNA double helix. The discovery was initiated through molecular dynamics simulations, carried out using NAMD and VMD. The simulations demonstrated that if an electrical field, driving negatively charged DNA through a nanopore, exceeds some critical value, the force exerted on DNA stretches DNA to twice its equilibrium length, reducing thereby its diameter and allowing it to squeeze through narrow pores. The simulations predicted precise values of pore radii and associated critical fields. The predictions were validated experimentally by counting the number of DNA copies that passed at different electric fields through synthetic nanopores. Further details about this study can be found here.

It was 1995 when NAMD was introduced (Nelson et al.) as a parallel molecular dynamics code enabling interactive simulation by linking to the visualization code VMD. In 1999 a major improvement was accomplished in NAMD 2 (Kalé et al.), scaling to 200 processors at the time due to the efforts and software of the Parallel Programming Lab. NAMD has since matured, adding many features and scaling to thousands of processors, garnering accolades and users in the process. This progress is now collected in a NAMD review paper that presents, in a manner accessible to the novice researcher, the concepts and algorithms behind NAMD, features for steered and interactive MD and for free energy calculation, the elements of the NAMD design that enable parallel scaling, performance comparisons of a variety of platforms, and advice for productive use of NAMD on modern research projects. Case studies ranging from the typical to the elaborate demonstrate the capabilities and flexibility of NAMD. This new reference provides an excellent foundation for working through the extensive NAMD tutorials, either on your own or at a hands-on workshop.

For the sequence of DNA, the genetic instructions of cells, to be read, the double helix of DNA is split open, exposing single DNA strands to DNA binding proteins. Once bound to DNA, the proteins, in carrying out their functions, will crawl along the DNA strand in one of two directions, towards DNA's 3' or 5' end. A recent study of DNA discovered a surprising property of single DNA strands that seems to explain how DNA binding proteins recognize the right direction on DNA strands. By measuring the translocation of DNA through alpha-hemolysin, a membrane protein with a narrow pore, researches discovered that directed single stranded DNA moves much faster when entering the pore 3' end first, rather than 5' end first. The underlying mechanism of this directionality was discovered through molecular dynamics simulations using NAMD and VMD. The simulations revealed that, in a narrow pore, DNA bases tilt collectively towards the 5' end, transforming a wide space directionless DNA brush into a tight space DNA ratchet. The 360,000-atom MD simulation did not only reveal how the DNA bases align and move faster in the "smooth" direction, but did also predict how the directional DNA movement can be discerned by means of direction-sensitive ionic currents through the channel blocked by translocating DNA strands. More details about this study can be found here.

In an optimistic future, cars and appliances will be powered by renewable energy produced by burning hydrogen gas, with water being the only waste product. To supply this hydrogen gas, scientists are turning their attention to an enzyme called hydrogenase that is found in certain microorganisms, which produce hydrogen gas from sunlight and water. This enzyme, however, is sensitive to oxygen gas, which irreversibly deactivates its hydrogen-producing active site. Understanding how oxygen reaches the active site will provide insight into how hydrogenase's oxygen tolerance can be increased through protein engineering, and in turn make hydrogenase an economical source of hydrogen fuel. In a recent paper (also described in this webpage), the programs NAMD and VMD are used to analyze the gas diffusion process inside hydrogenase, and how it correlates with the protein's internal fluctuations, thereby creating a map of the oxygen pathways. The calculations revealed two distinct pathways for oxygen to reach the active site. Gases participate in physiological processes of many organisms and the new computational strategy developed promises to image gas diffusion pathways for many relevant proteins. In fact, the researchers are currently inspecting hundreds of proteins for their ability to internally transport gas molecules.

The import of nutriments over their cellular membranes is one of the main tasks of all living cells. Even though a major part of the cell's molecular machinery is devoted to this task, principles of selective membrane transport are not yet well understood, mainly due to the fact that the membrane proteins responsible are notoriously difficult to resolve in their structure, the latter a prerequisite for a full physical description of the function. Recently, cell biology got very lucky in having the structures of two closely related membrane proteins solved. Two highly homologous aquaporins from the same bacterium, Eschericha coli, have become structurally known: one that conducts only water, called AqpZ, and one that conducts water as well as the nutriment glycerol, called GlpF. The discoveries have permitted us through structure analysis with VMD and molecular modeling with NAMD to look over nature's shoulder in the evolutionary design of two similar import channels of different conductivity. As described in a recent paper and on our aquaporin site, in making a water channel also a glycerol channel, nature has turned to a very basic principle, namely adjusting the overall pore size of the channel from a very narrow channel, just wide enough for water, to one wide enough also for glycerol.

To guide cell biology research and explain observation through molecular structures and sequence data, life scientists resort increasingly to computational tools. Sequence and structure viewers (VMD) combined with molecular dynamics modeling software (NAMD) are primary methodologies that revolutionized modern biomedicine. The revolution happened so quickly, though, that traditional university training has not kept up with the pace of developments in computational biology. A series of computational biophysics workshops in Perth (Australia), Urbana, Boston, Lake Tahoe, Chicago, and San Francisco attempted to fill the gap through hands-on training. Theory sessions in the morning introduced the concepts and methods used in molecular modeling today; computer laboratories in the afternoon gave participants, students, postdocs, and faculty, opportunities to work through tutorials at their own pace on provided laptops, as well as work on their own research problems. The workshops funded by NIH, NSF, NCSA, UIUC, and UWA met the needs of novices and experts alike for instruction in a new generation of research methods. All workshop materials are available on the web.

In a biological cell, membrane channels act like miniature valves regulating the flow of ions and other solutes between intracellular compartments and across the cell's boundary. Assembled in complex circuits, they generate, transmit, and amplify signals orchestrating cell function. To investigate how membrane channels work, life scientists, using an extremely fine pipette, isolate a tiny patch of a cell membrane and, in so-called patch clamp measurements, determine electric currents in response to applied electric potentials. Dramatic increase in computational power and its efficient utilization by NAMD allows one today to reproduce such studies computationally, calculating the permeability of a membrane channel to ions and water directly from its atomic structure. In what is one of the largest molecular dynamics simulation to date, described in a recent paper as well as on our web site (here), one copy of the membrane channel alpha-hemolysin, submerged in a lipid membrane and water, was subjected to an external electric field that drove ions and water through the channel. The calculations produced also an image of the electrostatic potential across the channel (see figure).

When Escherichia coli bacteria enjoy lactose and related food molecules in their environment, the cells quickly furnish proteins needed for import and metabolic digestion of the food. A set of genes, called the lac operon, is transcribed into messenger RNA that directs the synthesis of these proteins. When lactose is not available, the protein synthesis would be wasteful and, indeed, is prevented by locking the lac operon. This is achieved by a protein called lac repressor that forces the segment of the lac operon needed to initiate transcription into a loop, but that can be unlocked by a lactose molecule binding to the protein as soon as the food becomes available again. A recent study of the lac repressor combines a 314,000-atom protein simulation using NAMD with a multiscale simulation technique coupling the protein to the DNA loop. The calculations reveal how the lac repressor stretches out two "hands" grabbing the genomic DNA and then keeps a tight grip on the DNA wrestling it into a loop. The discovery is described on our website as well as in a popular article.

The Materials and Processes Simulations (MAPS) platform from Scienomics is a user friendly environment for molecular modeling and simulations. Its plug-in based architecture enables scientists to use the best technology for a given problem. The MAPS platform runs on Linux, IRIX and Windows® XP operating systems. MAPS includes a series of tools enabling the construction of molecular systems, finite and periodic, 3D visualization and other utilities. The NAMD user can quickly create a molecular model, using standard sketching tools or MAPS' polymers builder, and set up calculations using the NAMD graphical user interface which gives access to many of NAMD advanced capabilities. Analysis tools and graphs available in MAPS enable an easy representation of NAMD results. MAPS' native client-server architecture allows to use the best computational resources available and run NAMD simulations on numerous operating systems. Finally, efficient interaction with office productivity software allows to produce quickly presentations and reports.

The ear is a sensitive and robust device, able to perceive the faint sound of flowing water and the thunderous blast of an air plane. Like a microphone, the ear transforms a complex, mechanical stimulus (sound), into an electrical signal, a voltage change in a nerve cell, that can be understood by our brain. This transformation is called "mechanotransduction" and is accomplished by a series of amazingly minute devices that each connect a soft spring to an ion channel, both located in specialized sensory cells, the hair cells of the inner ear. The springs, through their vibrations agitated by particular sound frequencies, control ion currents passing through the channels, thereby, modifying the hair cell internal electrical potential. This leads to neural signaling to the acoustic cortex of the brain. Recently reported molecular dynamics simulations using NAMD, some of the most extensive simulations accomplished to date both in size and duration, showed that the mechanical characteristics of hair cell signaling may be traced to a single protein, ankyrin, that acts as a helical spring. Imagine a soft spring that is extended several inches by the weight of a feather! Ankyrin is such a spring, but a billion times finer (see our ankyrin website).

Biological evolution left its many traces in the form of organisms as well as in "fine print" in the form of gene sequences and associated protein structures. From the "fine print" researchers can draw conclusions about the inner workings of living cells and derive opportunities to battle disease. Researchers enjoy easy access to sequence and to structure information, but so far mainly separately, i.e., either for sequence or for structure. VMD, our widely used structure viewing and analysis program, has already offered a glimpse of the viewed protein's sequence, but with its latest release has taken a key step further, assisting in viewing and aligning multiple structures and sequences with few mouse clicks. Users of VMD 1.8.3 find themselves routinely comparing their protein of interest with analogous ones getting VMD to color the protein by similarity in structure, in sequence, and showing conserved amino acids. VMD 1.8.3 surprises with numerous further features, including a new cartoon representation that follows the actual molecular structure closely and offers superb, publication quality images. VMD continues to work together with the molecular dynamics program NAMD, permitting viewing and analysis of huge trajectory files by supporting 64-bit processors.

Biological cells, the basic units of life, are organized assemblies of nanodevices. Nanobiotechnology can adapt Nature's solutions for its own purposes, using computational biology to redesign Nature's nanodevices. In the case of Nanodiscs, bioengineers thought to construct the smallest possible environment that mimics the native environment of membrane proteins. Researchers borrowed the amino acid sequence of a naturally occurring class of proteins, lipoproteins, which are involved in the transport of lipids and cholesterol. The lipoprotein found in humans, apolipoprotein A-1, was used to synthetically engineer "belts" that surround a discoidal lipid bilayer, shielding the hydrophobic lipid tail groups from water. As recently reported, molecular dynamics simulations using NAMD showed an atomic level image of the structure of such a nanodevice. The predicted discoidal shape, diameter, and thickness of Nanodiscs simulated were experimentally corroborated through so-called small angle X-ray scattering. For more details see the HDL & Nanodisc website.

An important means for generating genetic diversity to provide raw material for evolution and maintain genomic stability is sexual reproduction. At the molecular level, the genes of two individuals are mixed through a process called homologous recombination. This process is found also in many simple life forms, even bacteria. At the beginning of recombination, two DNA duplexes, e.g., from mother and father, are aligned next to one another as the result of homology search, i.e., like strands are brought together with like strands. The four single DNA strands, two in each duplex, cross reciprocally two of the strands between the duplexes. The result is a joint molecule that contains DNA crossovers, named Holliday junctions. The Holliday junction is highly polymorphic in moving along two DNA duplexes, exchanging their DNA. Researchers are now investigating the physical mechanism of Holliday junction migration. The polymorphic, dynamic character of this migration makes observations difficult and the researchers resorted to molecular dynamics simulations using NAMD. The results, reported recently, resolved the dynamics of maternal-paternal DNA exchange through Holliday junction transitions in unprecedented detail providing an atomic level view of sexual reproduction. Check a brief review on our website.

Energy for most of the earth's biosphere is gained when sun light absorbed drives electrons across a membrane through a protein called the photosynthetic reaction center (RC), leaving behind positive electron holes. The electrons join protons to become hydrogen atoms and move, bound pairwise to a quinone molecule, to another protein, the so-called bc1 complex. Here electrons and protons move together back over the membrane and become separated again, thereby establishing an electro-osmotic potential that fuels many cellular processes. However, the electrons need to return to the RC to fill the electron holes left behind. Nature employs for this purpose a kind of bioelectric extension cord in the form of a third protein, cytochrome c2, that shuttles the electrons back from the bc1 complex to the RC. A recent paper reports molecular dynamics simulations using NAMD that investigated how cytochrome c2 plugs into the RC. Landing on a broad face of the RC, interactions steer the protein such that its electron carrying heme group comes close to RC's chlorophylls with electrons missing, a chain of water molecules providing an electrical conduit. The study is yet another example of how simulations provide today complete views of the fundamental processes underlying life.

Biological cells must be capable of attaching themselves to their surroundings. For this purpose they utilize fibrillar proteins, such as fibronectins, that grasp cells through cell surface receptors integrins. The latter act as snap fasteners to the extra-cellular fibrils. The growth, movement, and survival of cells are all dependent on the ability of integrins to fasten cells upon intra-cellular signals or to signal inwards that something has become fastened on the cell surface. The major fastener on integrins are simple divalent ions like Mg⁺⁺ or Ca⁺⁺ that can adhere to specific molecules with amazing strength, even though the interaction at the cell surface is exposed to water. Computer simulations using NAMD, reported recently, revealed a dynamic picture of the interactions used by cells to link themselves to the extra-cellular matrix. They showed that it is actually a brave water molecule that is recruited by integrins as a protective shield for the interaction. The simulations provide for the first time a detailed view of how cell tissues are stabilized through surface ions against mechanical stress.

Many bacteria hang into their cellular membranes proteins that look like a Japanese lantern. These proteins have a flexible cylinder (pore) that crosses the membrane and opens and closes depending on membrane tension; from the cylinder hangs a balloon with seven small openings around its equator (see figure). The apparent function of the Japanese lantern protein, aptly called mechanosensitive channel of small conductance (MscS), is to protect the bacterial cell against osmotic stress: when a bacterium finds itself suddenly in an aqueous environment entering water can burst the cell. Before this happens the cell membrane experiences tension that opens the protein pore, permitting passage of water and ions, the efflux being controlled through the protein balloon. A recent study explores the dynamical properties of MscS, e.g., pore closure and opening as well as ion conduction, by means of molecular dynamics simulations using NAMD. Embedding the large protein into a lipid bilayer and water led to a simulation encompassing 220,000 atoms. Surprisingly, the protein balloon was found to control the arrangement of positive and negative ions through a peculiar pattern of charged, polar, and non-polar amino acids on its internal and external surfaces. This suggests that the Japanese lantern protein has a yet unknown second function in the bacterial cell.

Electrical devices on computer chips built from silicon compounds have reached the small length scales of the building blocks in biomolecules, namely, the amino acids in proteins and the bases in DNA. Using beams of electron microscopes, electrical engineers drill nanometer wide pores into silicon wafers that contain a central layer only a few atoms thick. The engineers surround these pores with transistors and electrodes that can detect charges moving in the nanopore. Electrical fields across such synthetic nanopores can thread charged molecules like DNA through, and electrical signals stemming from single molecules transiting the pores can be recorded. Since the size of the nanopores compares with the dimension of DNA bases, the signals should eventually become precise enough to distinguish DNA bases, such that nanopores can become recording heads reading off sequences of DNA. While such ultrafast recording of DNA sequences is still a distant goal, the manufactured nanopores have been used already for sizing short strands of DNA as reported recently (report1, report2). Molecular dynamics simulations with NAMD and molecular graphics with VMD played a crucial role in imaging the dynamic events (movies available here) involved in recording single molecules of DNA and for optimizing the design of nanopores towards efficient threading and accurate recording. The landmark collaboration between computational biologists and device engineers promises to further unlock the great potential of biomedical nanotechnology.

Proteins perform the many functions of biological cells. This ability arises from the particular three-dimensional structure into which proteins fold at physiological temperature. The quick and precise folding of proteins depends on their molecular environment, e.g., water or lipid membrane, and is being investigated in many laboratories today. A new study from a computational - experimental collaboration investigated the folding and resulting structure of a protein, ubiquitin, in ethylene-glycol, commonly known as antifreeze, mixed with water. In this mixture the folding could be monitored at very low temperature in "slow motion" and resolved in great detail. Computational modeling using NAMD and VMD suggested that adding antifreeze to water leaves ubiquitin folding unaffected and this was born out indeed by further observation. Antifreeze is offering now a wider window into the study of the amazing abilities of other proteins.

One of life's great achievements is the development and maintenance of multi-cellular organisms, from an embryo to adulthood. Multitudes of cells need to be sorted and resorted into tissues, organs, and body of living beings. One strategy towards this end is to endow cells with so-called adhesion proteins that connect a mechanical framework inside cells through the cell membrane with other cells. A key type of adhesion protein is cadherin (calcium-dependent adherent protein) that stretches through the cell surface five-tandem domains. The outermost domain can stick to a cadherin molecule from an adjacent cell. Crystallography provided the molecular structures of cadherin pairs and resolved in atomic detail the cadherin-cadherin contact between cells. This prompted a collaboration that aimed at probing the adhesion strength of cadherin pairs through steered molecular dynamics simulations stretching the pairs apart. Results of the simulations were reported in a recent publication that employed NAMD as well as VMD. As shown by crystallography, the cadherins each insert a tryptophan residue into the other protein. The link thereby established can be broken only through strong forces that induce a step-wise slippage of the residues first out of their binding pockets and then along the protein surface. This scenario suggests a mechanism for selectivity among cadherins, i.e., why among the various cadherins found on the surfaces of cells some adhere much better to each other than others.

Since Leuwenhook, microscopic images of living matter have been produced with radiation, from light to X-rays. With the advent of ever more reliable computational methodologies, molecular dynamics has reached the status of a trusted research instrument. This instrument is particularly powerful for imaging nanoscale, i.e., 10-100 nanometer size, systems. This month our group brought three computers on-line that can serve to image nanoscale systems, three 48-processor rack-mount Xeon clusters (pictured) running our MD program NAMD. One such cluster, a nanoscale system of 300,000 atoms imaged over a nanosecond at the most advanced simulation conditions possible today, requires four days of computing. The Clustermatic software from Los Alamos National Laboratory makes each cluster of 48 processors appear to biomedical researchers as a single machine and allows interactive simulations to temporarily displace long-running NAMD jobs. The clusters have been used already for a study of balance and hearing in the human inner ear. These senses are intrinsically mechanical, relying on hair cells to convert vibration to ion channel modulation. Ankyrin, a protein formed by repeats of a 33-amino-acid domain, is thought to act as a molecular spring in mechanotransduction channels. Explaining the mechanism of ankyrin elasticity requires large simulations of 340,000 atoms that apply repeatedly stretching forces to the protein and monitor its response, revealing a mechanical behavior ideally suited for its biological function. The clusters are presently also used to design artificial nanopores for sequencing of DNA.

Computational and experimental biologists investigate jointly the physical mechanisms underlying the function of the molecular machines in cells. Simulations used to encompass biomolecular systems of 10,000 atoms, but recently the size increased tenfold. For example, simulations of aquaporins, a water channel, published during 2001-2003 (1, 2, 3) involved about 100,000 atoms and had been cited in connection with last year's chemistry Nobel prize; simulations of cadherin, a cell adhesion protein, and of ATPase, a key metabolic protein complex, published 2004 included a similar number of atoms. Simulations of over 200,000 atoms for a protein, lac repressor, that regulates genes, and for a protein, MscS, that is a mechanically gated membrane channel will be published later this year. Simulations involving over 300,000 atoms, on a protein, ankyrin, acting as an elastic spring in hearing, have been completed. The increase in the size of simulated systems is prompted by a revolutionary advance in crystallography that resolves ever larger structures of biomolecules and simulations are made possible through the great increase of computer resources at the NSF centers (PSC, NCSA). This marks the beginning of a new era in which systems like virus capsids and the ribosome, entailing 1-3 million atoms, will be studied, too. NAMD (see Highlight Dec 2002) is ready for the challenge posed by simulations needing 250-500 processors today and 1000-10000 processors in the future, to keep up with the developments in the biology laboratories (more).

Far from playing only the role of bricks and mortar as a mere divider between the inside and outside of cells or between parts of the cell, lipid bilayers are an active, tightly regulated cellular component whose physical properties are critical for the proper function of the membrane proteins contained within them. Lipid bilayers are the preeminent domain of computational biology since despite their considerable stability and impenetrability they form disordered films that are best described through computer modeling, albeit tested by observation. One of the largest molecular modeling projects achieved so far has been recently reported that employed NAMD to investigate the mechanical properties of cellular membranes. The systems simulated were made of lipids and water, composed of about 40,000 atoms, and simulated for over 100 nanoseconds. The simulations revealed that membranes, in terms of their mechanical properties, are far from being homogenous films; rather, they exhibit a delicate multi-lamellar structure of layers that alternatively tend to shrink and expand the membrane, inducing strong forces on all proteins and molecules entering. The lamellar character of the cell's membranes plays a key role for cellular processes such as osmotic regulation and may explain even the action of anaesthetics.

What will car motors look like in a million years? It's hard to tell, but biological cells seem to have found an ideal engine that they use since the early days of evolution. A spoonful of their engines generates about as much total torque as the strongest car engine today. The engine is called FoF1-ATP synthase and synthesizes the molecule ATP by combining two generator-like motors, Fo-ATPase and F1-ATPase, coupled through a single axle, one motor (Fo) that converts the cell's electrical energy into rotation, another one (F1) that converts rotation into chemical synthesis (see November 2003 highlight). ATP synthase, found throughout the whole kingdom of life, can also work in reverse, turning ATP into electrical energy. Cells typically use the energy of sun light or of food to generate an electrical potential by pumping protons that carry a positive charge across their cell membrane. The energy stored drives the protons back through Fo-ATPase enforcing rotation of the axle; the rotation in turn induces ATP synthesis in F1-ATPase. In one of the largest computational and mathematical biomodeling projects undertaken to this day and reported recently, researchers build from available disjoint structural data a model of Fo-ATPase and demonstrated its function as a motor turning proton conduction into rotation of a cylindrical protein complex. By linking nanosecond molecular modeling to a mathematical model of the motor's key elements, they could follow Fo-ATPase function properly, even when the load arising from driving synthesis in F1-ATPase was added. FoF1-ATP synthase being one of the largest molecular machines in biological cells, modelers needed to employ for its study the most advanced tools, NAMD and massively parallel computers, along with a new approach that combined molecular dynamics and stochastic mathematics.

When their environment gets high in hydrochloric acid, bacteria must protect themselves from chloride ions leaking into the cells. Some bacteria rely on so-called ClC chloride channels to quickly evacuate these ions without letting other small particles pass through. A recent paper investigates the mechanism of chloride transport in a ClC channel from the bacterium E. coli. By means of computer simulations using NAMD, researchers visualized the pathway taken by chloride ions as they pass through the channel and identified how the channel's protein architecture optimizes ion conduction. The relevant architectural features have also been observed in aquaporin water channels (see November 2003 highlight) and the potassium ion channel, but are rarely seen in other proteins. The new discovery demonstrates how much computer simulations are contributing to the emerging picture of life's membrane channel design that is critical for functions of biological cells ranging from the maintenance of our body's hydration to electrical signaling in the brain.

Tissues of the human body are composed of specialized cells held together by a connective fabric of proteins, that form the knots of a net glueing cells together. Upon stretching tissues, the knots unravel, rendering the net larger, but mysteriously also firmer. A protein called fibronectin-III-1 plays a particularly important role in the latter respect. Atomic force microscopy revealed that under mechanical tension fibronectin-III-1 stretches to ten times its initial length; but is does so in two steps, the first stretching step leading to net strengthening. It had been discovered earlier that other fibronectins found between cells are made of two sheets packed like a sandwich, but the structure of fibronectin-III-1 remained elusive. In an experimental-computational collaboration reported recently, the structure has now been resolved that at first sight looked similar to the sandwich structure of the other fibronectins, but on closer inspection showed a weak and a strong sheet. Simulations using NAMD revealed that stretching of the protein unravels readily the weak sheet, and only therafter the strong sheet. It turns out that the strong sheet of fibronectin-III-1 by itself, known as anastellin, inhibits tumor growth. Stretching of fibronectin-III-1, as it occurs naturally in tissue, unravels apparently half of the protein to render it extremely adhesive, strengthening a protein net that prevents metastasis of cancer cells and also assists wound healing (press release, more).

This year's Nobel prize in chemistry goes to Peter Agre and Roderick MacKinnon for their groundbreaking work on membrane channels. We join all in congratulating our two colleagues who have advanced knowledge on membrane channels. Foremost, this advance was made possible through great achievements in experimental methodology by several distinguished researchers, but computational and theoretical investigations of the channels the two new Nobelists investigated, contributed significantly. Our group feels fortunate to have participated in the exciting development through modeling studies of the aquaporin channel (see previous highlights June and April 2003, June and May 2002, and November 2001). Our investigations were based on molecular dynamics simulations of channels in their native membrane environment and demonstrate the widening role of computing in modern life science: using the most advanced computers at the NSF centers in Pittsburgh and Urbana, we could establish the mechanisms of conduction and selectivity in aquaporins in a series of studies that are summarized in our aquaporin web page, and in three recent publications in Biophysical Journal that show how the molecular architecture of channels is optimally designed for their function (1, 2, 3).

Living cells sense their environment and respond to its changes through proteins integrated into their outer membrane. These proteins mediate a broad range of cellular activities, including active and passive transport of materials across the membrane as well as response to osmotic shock, which can strain the cell membrane to the point of catastrophic bursting. Cells protect themselves through so-called mechanosensitive channels that open before the membrane tension grows too large. Molecular dynamics simulations and advanced analysis using NAMD and VMD have revealed in a recent report how the joint mechanics of membrane and protein opens a mechanosensitive channel called MscL. The finding promises to revolutionize the modern view of membrane - protein interaction: the membrane, far from being a homogeneous elastic sheet, exhibits a dramatic variation of tension across its thickness that proves to be decisive for the opening of MscL. More here.

Modeling the molecular processes of biological cells is a craft and an art. Techniques like theoretical and computational skills can be learnt by training, but meaningful applications are achieved only with experience and sensitivity. A summer school in Theoretical and Computational Biophysics attempted to teach both the craft and art of modeling through learning by doing: nearly hundred participants from all over the world came for two weeks to the Beckman Institute in Illinois to stretch proteins, pull water through molecular channels, mine genomic data, build their own computer cluster, and study their favorate biomolecules. After lectures and discussions in the morning, afternoon and evening were devoted to learning by doing, assisted through 300 pages of tutorials, in computer laboratories humming with computational biology software, e.g., VMD, NAMD, and GAMESS, and linked to NCSA's fast pentium machines. The school lasted two weeks, but will go on much longer: all school materials remain available on the web; participants will use BioCoRE to stay in touch and continue the scholarship and friendship experienced in Illinois.

Biological cells have evolved highly optimized molecular devices made from nanometer size proteins that are the basis for all cellular processes. Engineers seek to conquer the domain of such nanodevices and naturally they look to nature for guidance. In turn, biologists look to nanoengineers for new tools to manipulate cells. A recent study attempts a third type of exchange at the cell - nanotechnology frontier, use of nanodevices that mimic proteins for better understanding of the functions of both: the engineered system can be much simpler than the protein and much easier to comprehend, whereas the protein is more evolved and can reveal the design principles for optimized function. The study links two renowned devices, nanotube and membrane channel. Using the molecular dynamics program NAMD, water flow through nanotubes is simulated and compared with earlier studies of aquaporin water channels in bacteria.

At present, the exchange benefits biology: the very simple nanotube reveals the principles of biologial water conduction, in particular, water flow without loss of electrical voltage. In the future, nanoengineering may benefit, for example, in designing filters for mechanical desalination of sea water.

The import of material into biological cells needs to be highly selective. Which physical interactions are the best capable of differentiating suitable from unsuitable imports? How can one keep import traffic fast despite strong selection? As one knows from border crossing, fast traffic and high security can be conflicting objectives. A recent investigation employed a new simulation method to answer the above questions for import of glycerol through a membrane channel. Linking the renown molecular graphics program VMD with the simulation program NAMD permits researchers to view "live" proteins, e.g., the membrane channel, and employing computer game technology, namely, a device that permits one to pull glycerol through the simulated channel, researchers can feel the mechanical resistance of the pulled glycerol, guide it through the channel, and directly observe the selection process involving glycerol's shape, capability for hydrogen bonding, and electric dipoles. The new methodology, interactive molecular dynamics, will revolutionize computational biology and is one of the first tools that take advantage of fast (TeraGrid) computer networks in linking local graphics workstations (running VMD) with a distant supercomputer (running NAMD). The method has received an enthusiastic editorial comment.

Infections are battled best by the human immune system when there exists a memory from a previous disease or vaccination. The first step in using this line of defense is recognition: Cells of the immune system capture antigens, e.g., microbes in the respiratory tract, then mature in the lymph system, and finally present on their surface pieces of the antigen to T-cells that may recognize the antigen and become activated. The recognition of the antigen by T-cells is dramatically enhanced through surface receptors, CD2 and CD58, on the T-cell and the antigen presenting cell. The receptors stick out from their cell, adhere to one another, and conjoin the T-cell and antigen presenting cell long enough to enable recognition and activation. The molecular basis of this adhesion has been probed in a recent collaborative study with UIUC chemical engineer D. Leckband. Starting from the available crystallographic structure of the CD2-CD58 complex the researchers carried out 90,000 and 100,000 atom simulations using NAMD and pulled the complex apart in steered molecular dynamics simulations. An analysis of the simulations with VMD revealed in atomic level detail how the human immune system is strengthened through elastic adhesion.

Scanning their environment for information such as food resources, signs of danger, and illumination is crucial for the well being of biological cells. Evolution has developed for this purpose a great variety of membrane proteins, so-called receptors, that receive physical cues from the external environment through encounters with molecules or absorption of photons and send respective signals into the cell. A common type of receptor sends its signal through interaction with intracellular G-proteins that convey the signal further; proteins of this type are called G protein coupled receptors (GPCRs). GPCRs exist in lower as well as higher life forms and, in fact, the human genome codes for over 1300 GPCRs that detect ions, organic odorants, amines, peptides, proteins, lipids, nucleotides and photons. As about half of modern drugs act on GPCRs, learning how they become activated once they receive their signal is highly relevant. In a recent study, advanced computer simulation techniques using NAMD and VMD have been employed to investigate the first key steps of activation of a GPCR, the visual receptor rhodopsin. more.

The parallel molecular dynamics program NAMD, and its sister visualization program VMD, have helped researchers at Illinois discern how muscles stretch, nerves sense pressure, and kidneys filter water. The latter project, for example, used simulations of 106,000 atoms to discover how aquaporins, which are ubiquitous in mammals, plants, and bacteria, allow water to pass while preventing the conduction of protons or ions (article in Science, more). Our years of work developing this software to apply the nation's fastest sup ercomputers to understand the tiny components of living cells were recognized at the SC2002 High Performance Networking and Computing conference with a Gordo n Bell Award for unprecedented parallel performance on a challenging computational problem (pdf of paper-497k). NAMD and VMD are distributed , free of charge, to thousands of scientists in industry and academia around the world, quickening the pace of drug discovery and other vital research to unravel biological processes.

What kind of function does the longest gene in the human genome code for? The answer is a bit mundane: a very long molecular spring that provides muscle with passive elasticity. Nature adjusts the protein, called titin, for many types of muscle, e.g., skeletal or cardiac muscle, as well as for other cellular functions. The molecular spring contains hundreds of elastic elements in series like beads on a string. One type of bead is the immunoglobulin domain, which can stretch to ten times its normal length. For a long time only one of the immunoglobulin domains was structurally known, permitting only a single peek into nature's design library. Recently, a second domain became structurally known and protein crystallographers and modelers joined forces to discover how nature designs its beads along titin, as described in a recent publication.

The biological control of inorganic crystal formation, morphology and assembly is of interest to biologists and biotechnologists studying hard tissue growth and regeneration, as well as to materials scientists using biomimetic approaches for control of inorganic material fabrication and assembly. A molecular-level understanding of the natural mechanisms involved in these processes can be derived from the use of metal surfaces to study surface recognition by proteins together with combinatorial genetics techniques for selection of suitable peptides.

In a recent study, the structure of a genetically engineered gold binding protein has been determined computationally, and the ability of the protein to recognize gold crystal surfaces has been explained. Molecular dynamics simulations were carried out with the program NAMD using the solvated protein at the gold surface to assess the dynamics of the binding process and the effects of surface topography on the specificity of protein binding.

Living cells rely on nutrients absorbed through their cell membranes, for example on glycerol that is key to the cells' metabolism. Proteins, so-called aquaporins, in the membranes form channels that act as sieves permitting passage of water, glycerol, and like molecules, but prevent other molecules of similar size from entry and clogging. For this purpose the channels interact strongly with molecules attempting to pass. In a recent publication, the energetics of the conduction process of glycerol for the aquaporin GlpF was measured in a molecular dynamics simulation, carried out with NAMD, that pulled glycerol through the channel, monitoring the forces needed to advance along the channel axis. An analysis that discounted the irreversible work done on glycerol, a difficult prerequisite, yielded the energy profile that glycerol experiences along the channel and that reflects how the protein decides which molecules are allowed to pass the sieve.

Human kidneys need to filter about a bathtub of water a day through cells that contain membrane channels made of proteins called aquaporins. Crystallographers from the University of California at San Francisco (R. Stroud and coworkers) that discovered the structure of one type of aquaporins, aquaglyceroporins, have teamed up with UIUC researchers to determine how these channels achieve their very high water throughput, yet prevent the cells' electrical potential from discharging by not permitting any flow of ions or conduction of protons. The team, combining 106,000 atom simulations using NAMD and crystallography, found that the channels achieve the impressive filtering function by conducting water single file and keeping the water molecules strictly oriented: water molecules enter the channel oxygen atom first and leave the channel oxygen atom last. Aquaporins are ubiquitous in mammals, plants, and bacteria and the finding, published recently in Science magazine, has implications for many biological functions as well as for human diseases, e.g., cataract of the eye, loss of hearing, or diabetes insipidus. (more, press release)

Biological cells process numerous types of information, for optimal control of their life cycles or to adapt to their environment, and recruit for this purpose signaling proteins. The best known among the latter are the G-proteins, involved in numerous diseases and related to many targets of drugs. G-proteins are closely related to motor proteins: G-proteins get switched on and off through the binding of GTP and its hydrolysis to GDP; motor proteins generate mechanical force through binding of ATP and its hydrolysis to ADP. A recent publication reports a 19,463 atom computer simulation using NAMD that reveals a "power stroke" in G-proteins likewise found in motor proteins. The stroke switches on and off G-proteins' ability to interact with other signaling proteins, with a power stroke that transforms the protein from an ordered into a disordered conformation.

Cells in animals adhere to dynamic, seemingly random assemblies with other cells that make up tissues like skin, organs, and brain. The cell's adhesion and motion is controlled by the extracellular matrix, with the protein fibronectin being a key component. The proteins have optimal mechanical elasticity and also signal to cell surface receptors, integrins, the tension exerted on them. How this is achieved is the subject of an ongoing collaboration with the research group of Viola Vogel of the Department of Bioengineering at the U. of Washington in Seattle (see also Oct 2001 highlight). The most recent publication from this effort reports a 97,884 atom steered molecular dynamics simulation using NAMD. It is revealed now that stretching two consecutive domains of fibronectin deforms two sites, the so-called RGD and synergy sites as well as their distance. This weakens binding to cell receptors and, as a result, integrins can function as gauges that signal the magnitude of exterior forces to a cell.

Adenosine triphosphate (ATP) is the fuel of life; every living cell must use ATP to carry out its functions, and the human body synthesizes its own weight in ATP every day. The ubiquitous molecular motor ATP synthase catalyzes the creation of ATP by precisely directing electrochemically generated torque. A detailed understanding of how this system functions can impact areas ranging from neurodegenerative disease research to nanotechnology development. Running at the Pittsburgh Supercomputing Center on LeM ieux, the most powerful computer system in the world for open research, the freely available simulation code NAMD can simulate a solvated all-atom model of ATP synthase with full electrostatics at 65% efficiency on 1000 processors. This achievement in scalability places NAMD an order of magnitude ahead of comparable packages for molecular dynamics simulation.

Deciphering the processes by which proteins recognize and bind to DNA is critical in our quest to understand cellular functions. To reach this goal, a collaboration with the group of Stephen Sligar, UIUC, explored the factors involved in protein-DNA recognition using hydrostatic pressure to perturb the binding of the BamHI endonuclease to cognate DNA. Our joint resulting publication outlines a new technique of high-pressure gel mobility shift analysis to test the effects of elevated hydrostatic pressure on the binding of BamHI (so-called restriction enzyme) to a specific DNA sequence. Upon application of a hydrostatic pressure of 500 bar, recognition between BamHI and the DNA sequence was weakened nearly 10-fold, suggesting an important role of water. An advanced 65,000 atom nanosecond molecular dynamics simulations with NAMD, at both ambient and elevated pressures, complemented the experiments and revealed how water-mediated interactions between BamHI and DNA control sequence recognition.

Aquaporins are channel proteins abundantly present in all life forms, for example, bacteria, plants, and in the kidneys, the eyes, and the brain of humans. These proteins conduct water and small molecules, but no ions, across the cell walls. Their defective forms are known to cause diseases, e.g., diabetes insipidus, or cataracts. The molecular modeling program, NAMD, along with large parallel computers at the Pittsburgh and Illinois supercomputing centers, permitted researchers now to model aquaporins in the natural environment of membrane and water in one of the largest molecular dynamics simulations ever (over 100,000 atoms). The simulations revealed in unprecedented detail how cells conduct water and glycerol, a molecule that serves cells' metabolism. The simulations provided a movie of the entire conduction process.

Our low-cost cluster of 32 Athlon PCs (see February 2001 highlight) has been in constant use by local users, providing a substantial and very cost-effective boost to our group's large-scale simulation capabilities. To satisfy demand, we have added two additional 32-processor clusters with higher performance at an even lower cost. On this platform, the freely available simulation code NAMD can complete a 1 nanosecond simulation of the 60,000 atom aquaporin-1 water channel with full electrostatics and constant pressure in a single week. We have given three tutorials, both filled to capacity, introducing participants to cluster hardware and software with the aid of a hands-on session assembling and installing four-processor clusters (see photo).

"How do you feel?" Biologists now have an answer that may surprise you. Our sense of touch relies upon the fact that cells in our fingertips can sense the pressure from a tabletop and transmit a signal to the brain. But until recently, the molecular mechanism for turning the stretching of a cell membrane into a cellular signal was unknown. An important step in understanding this process was the discovery of a protein known as a the mechanosensitive channel of large conductance, or MscL. Though this protein has been studied primarily in bacteria, homologues exist in all major kingdoms of life. Researchers in the Theoretical Biophysics Group have used molecular dynamics simulations to study, at the atomic level, how MscL opens in response to pressure changes. Models of MscL will give us new insight, not only into how we feel, but also how our hearts beat and how we keep our balance. Feel better now? ( more, publication )

"If I could just get my hands on that protein!" Single molecule manipulation techniques like atomic force microscopy have brought us closer to this frequently expressed wish. These techniques, however, do not "see" the atomic level detail needed to relate mechanism to protein architectures. True, computational methods do illuminate the elusive protein structures, but are limited to static structures, or trajectories yielded by weeks-long simulations. Now, with the advent of inexpensive, high-performance computing, interactive manipulation of molecular dynamics simulations has become a reality. Linking advanced molecular graphics with ongoing molecular dynamics simulations, and utilizing a haptic device to connect forces from a user's hand with forces in the simulation, researchers can interact with "live" proteins. The new methodology is described in a recent publication and the figure shown here demonstrates a Cl^- ion being pulled through a gramicidin A channel (see a 3.8mb Streaming Video).

Aspirin, the widely used pain killer, has revealed many beneficial effects such that it has attracted renewed attention. It has become known that aspirin acts as an inhibitor to prostaglandin synthase. Pharmacological researchers have succeeded to improve aspirin's effect by synthesizing analogue compounds, so-called superaspirins, that target the right type of prostaglandin synthase in the body. The continuing effort has been supported by basic research on the properties of prostaglandin synthases. Molecular dynamics simulations, carried out with our molecular dynamics program NAMD, have investigated how prostaglandin synthases select their substrates, arachidonic acid, through a binding channel that acts as a filter for compounds with the right stereochemical properties. The figure, taken from a recent publication, and made with our graphics program VMD, shows one monomeric subunit (in a cartoon/ribbon representation) of the ovine PGHS-1 homo dimer. To see both subunits click on the image. [More Information]

We have installed a low-cost cluster of 32 PCs with 1.1GHz Athlon processors, 256MB of RAM, and switched fast ethernet. On this new platform, the freely available simulation code NAMD runs 1 ns of our 92K atom ApoA1 PME benchmark in 8 days with 70% efficiency, the equivalent of a 100 processor Cray T3E. The new machine will be useful for simulations such as the stretching of the muscle protein titin. This work seeks to examine in atomic detail the dynamics and structure-function relationships of this 30,000 amino acid long filament in muscle contraction and elasticity. The cluster also provides a powerful engine for interactive simulations. The Linux-based Scyld Beowulf operating system (see Scyld's press release regarding NAMD) makes the entire cluster appear to computational biologists as a single machine.

Water can act as a conformational lubricant for protein folding. The giant muscle protein titin is a roughly 30,000 amino acid long filament which plays a number of important roles in contraction and elasticity. For example, upon stretching in muscle some of titin's protein domains can unfold one-by-one permitting titin to retain elastic properties in muscle over a very wide range of length. To examine in atomic detail the dynamics and structure-function relationships of this behavior, SMD simulations of force-induced titin domain unfolding were performed in close collaboration with atomic force microscopy observations. The simulations led to the discovery that water molecules play an essential role in breaking sets of hydrogen bonds that control the unfolding of titin's domains (see resulting publication).