Huafeng Xu

Scalable algorithms for molecular dynamics simulations on commodity clusters

Although molecular dynamics (MD) simulations of biomolecular systems often run for days to months, many events of great scientific interest and pharmaceutical relevance occur on long time scales that remain beyond reach. We present several new algorithms and implementation techniques that significantly accelerate parallel MD simulations compared with current state-of-the-art codes. These include a novel parallel decomposition method and message-passing techniques that reduce communication requirements, as well as novel communication primitives that further reduce communication time. We have also developed numerical techniques that maintain high accuracy while using single precision computation in order to exploit processor-level vector instructions. These methods are embodied in a newly developed MD code called Desmond that achieves unprecedented simulation throughput and parallel scalability on commodity clusters. Our results suggest that Desmonds parallel performance substantially surpasses that of any previously described code. For example, on a standard benchmark, Desmonds performance on a conventional Opteron cluster with 2K processors slightly exceeded the reported performance of IBMs Blue Gene/L machine with 32K processors running its Blue Matter MD code

Biomolecular Simulation: A Computational Microscope for Molecular Biology

Molecular dynamics simulations capture the behavior of biological macromolecules in full atomic detail, but their computational demands, combined with the challenge of appropriately modeling the relevant physics, have historically restricted their length and accuracy. Dramatic recent improvements in achievable simulation speed and the underlying physical models have enabled atomic-level simulations on timescales as long as milliseconds that capture key biochemical processes such as protein folding, drug binding, membrane transport, and the conformational changes critical to protein function. Such simulation may serve as a computational microscope, revealing biomolecular mechanisms at spatial and temporal scales that are difficult to observe experimentally. We describe the rapidly evolving state of the art for atomic-level biomolecular simulation, illustrate the types of biological discoveries that can now be made through simulation, and discuss challenges motivating continued innovation in this field.

Pathway and mechanism of drug binding to G-protein-coupled receptors

How drugs bind to their receptors—from initial association, through drug entry into the binding pocket, to adoption of the final bound conformation, or “pose”—has remained unknown, even for G-protein-coupled receptor modulators, which constitute one-third of all marketed drugs. We captured this pharmaceutically critical process in atomic detail using the first unbiased molecular dynamics simulations in which drug molecules spontaneously associate with G-protein-coupled receptors to achieve final poses matching those determined crystallographically. We found that several beta blockers and a beta agonist all traverse the same well-defined, dominant pathway as they bind to the β1- and β2-adrenergic receptors, initially making contact with a vestibule on each receptor’s extracellular surface. Surprisingly, association with this vestibule, at a distance of 15 Å from the binding pocket, often presents the largest energetic barrier to binding, despite the fact that subsequent entry into the binding pocket requires the receptor to deform and the drug to squeeze through a narrow passage. The early barrier appears to reflect the substantial dehydration that takes place as the drug associates with the vestibule. Our atomic-level description of the binding process suggests opportunities for allosteric modulation and provides a structural foundation for future optimization of drug–receptor binding and unbinding rates.

Activation mechanism of the β2-adrenergic receptor

A third of marketed drugs act by binding to a G-protein-coupled receptor (GPCR) and either triggering or preventing receptor activation. Although recent crystal structures have provided snapshots of both active and inactive functional states of GPCRs, these structures do not reveal the mechanism by which GPCRs transition between these states. Here we propose an activation mechanism for the β2-adrenergic receptor, a prototypical GPCR, based on atomic-level simulations in which an agonist-bound receptor transitions spontaneously from the active to the inactive crystallographically observed conformation. A loosely coupled allosteric network, comprising three regions that can each switch individually between multiple distinct conformations, links small perturbations at the extracellular drug-binding site to large conformational changes at the intracellular G-protein-binding site. Our simulations also exhibit an intermediate that may represent a receptor conformation to which a G protein binds during activation, and suggest that the first structural changes during receptor activation often take place on the intracellular side of the receptor, far from the drug-binding site. By capturing this fundamental signaling process in atomic detail, our results may provide a foundation for the design of drugs that control receptor signaling more precisely by stabilizing specific receptor conformations.

A conserved protonation-dependent switch controls drug binding in the Abl kinase

In many protein kinases, a characteristic conformational change (the “DFG flip”) connects catalytically active and inactive conformations. Many kinase inhibitors—including the cancer drug imatinib—selectively target a specific DFG conformation, but the function and mechanism of the flip remain unclear. Using long molecular dynamics simulations of the Abl kinase, we visualized the DFG flip in atomic-level detail and formulated an energetic model predicting that protonation of the DFG aspartate controls the flip. Consistent with our models predictions, we demonstrated experimentally that the kinetics of imatinib binding to Abl kinase have a pH dependence that disappears when the DFG aspartate is mutated. Our model suggests a possible explanation for the high degree of conservation of the DFG motif: that the flip, modulated by electrostatic changes inherent to the catalytic cycle, allows the kinase to access flexible conformations facilitating nucleotide binding and release.

Microsecond Molecular Dynamics Simulation Shows Effect of Slow Loop Dynamics on Backbone Amide Order Parameters of Proteins

A molecular-level understanding of the function of a protein requires knowledge of both its structural and dynamic properties. NMR spectroscopy allows the measurement of generalized order parameters that provide an atomistic description of picosecond and nanosecond fluctuations in protein structure. Molecular dynamics (MD) simulation provides a complementary approach to the study of protein dynamics on similar time scales. Comparisons between NMR spectroscopy and MD simulations can be used to interpret experimental results and to improve the quality of simulation-related force fields and integration methods. However, apparent systematic discrepancies between order parameters extracted from simulations and experiments are common, particularly for elements of noncanonical secondary structure. In this paper, results from a 1.2 micros explicit solvent MD simulation of the protein ubiquitin are compared with previously determined backbone order parameters derived from NMR relaxation experiments [Tjandra, N.; Feller, S. E.; Pastor, R. W.; Bax, A. J. Am. Chem. Soc. 1995, 117, 12562-12566]. The simulation reveals fluctuations in three loop regions that occur on time scales comparable to or longer than that of the overall rotational diffusion of ubiquitin and whose effects would not be apparent in experimentally derived order parameters. A coupled analysis of internal and overall motion yields simulated order parameters substantially closer to the experimentally determined values than is the case for a conventional analysis of internal motion alone. Improved agreement between simulation and experiment also is encouraging from the viewpoint of assessing the accuracy of long MD simulations.

Preconfiguration of the antigen-binding site during affinity maturation of a broadly neutralizing influenza virus antibody

Affinity maturation refines a naive B-cell response by selecting mutations in antibody variable domains that enhance antigen binding. We describe a B-cell lineage expressing broadly neutralizing influenza virus antibodies derived from a subject immunized with the 2007 trivalent vaccine. The lineage comprises three mature antibodies, the unmutated common ancestor, and a common intermediate. Their heavy-chain complementarity determining region inserts into the conserved receptor-binding pocket of influenza HA. We show by analysis of structures, binding kinetics and long time-scale molecular dynamics simulations that antibody evolution in this lineage has rigidified the initially flexible heavy-chain complementarity determining region by two nearly independent pathways and that this preconfiguration accounts for most of the affinity gain. The results advance our understanding of strategies for developing more broadly effective influenza vaccines.

Dynamic control of slow water transport by aquaporin 0: Implications for hydration and junction stability in the eye lens

Aquaporin 0 (AQP0), the most abundant membrane protein in mammalian lens fiber cells, not only serves as the primary water channel in this tissue but also appears to mediate the formation of thin junctions between fiber cells. AQP0 is remarkably less water permeable than other aquaporins, but the structural basis and biological significance of this low permeability remain uncertain, as does the permeability of the protein in a reported junctional form. To address these issues, we performed molecular dynamics (MD) simulations of water transport through membrane-embedded AQP0 in both its (octameric) junctional and (tetrameric) nonjunctional forms. From our simulations, we measured an osmotic permeability for the nonjunctional form that agrees with experiment and found that the distinct dynamics of the conserved, lumen-protruding side chains of Tyr-23 and Tyr-149 modulate water passage, accounting for the slow permeation. The junctional and nonjunctional forms conducted water equivalently, in contrast to a previous suggestion based on static crystal structures that water conduction is lost on junction formation. Our analysis suggests that the low water permeability of AQP0 may help maintain the mechanical stability of the junction. We hypothesize that the structural features leading to low permeability may have evolved in part to allow AQP0 to form junctions that both conduct water and contribute to the organizational structure of the fiber cell tissue and microcirculation within it, as required to maintain transparency of the lens.

A Simple Model of Multivalent Adhesion and Its Application to Influenza Infection

Adhesion between biological surfaces, which is typically the result of molecular binding between receptors on one surface and ligands on another, plays a fundamental role in biology and is key to the infection mechanisms of certain viruses, including influenza. The physiological outcome of adhesion depends on both the number of bound cells (or viruses, or other biological particles) and the properties of the adhesion interface that is formed, including the equilibrium number of receptor-ligand connections. Here, we introduce a quantitative model for biological adhesion by adapting thermodynamic models developed for the related problem of multivalent molecular binding. In our model, adhesion affinity is approximated by a simple, analytical expression involving the numbers of ligands and receptors at the interface. Our model contains only two fitting parameters and is simple to interpret. When applied to the adhesion between the hemagglutinin ligands on influenza viruses and the sialic acid receptors on biosensors or on host cells, our model generates adhesion affinities consistent with experimental measurements performed over a range of numbers of receptors, and provides a semiquantitative estimate of the affinity range of the hemagglutinin-sialic acid interaction necessary for the influenza virus to successfully infect host cells. The model also provides a quantitative explanation for the experimental finding that a mutant avian virus gained transmissibility in mammals despite the mutations conferring only a less than twofold increase in the affinity of its hemagglutinin for mammalian receptors: the model predicts an order-of-magnitude improvement in adhesion to mammalian cells. We also extend our model to describe the competitive inhibition of adhesion: the model predicts that hemagglutinin inhibitors of relatively modest affinity can dramatically reduce influenza virus adhesion to host cells, suggesting that such inhibitors, if discovered, may be viable therapeutic agents against influenza.

Key mutations stabilize antigen-binding conformation during affinity maturation of a broadly neutralizing influenza antibody lineage

Affinity maturation, the process in which somatic hypermutation and positive selection generate antibodies with increasing affinity for an antigen, is pivotal in acquired humoral immunity. We have studied the mechanism of affinity gain in a human B‐cell lineage in which two main maturation pathways, diverging from a common ancestor, lead to three mature antibodies that neutralize a broad range of H1 influenza viruses. Previous work showed that increased affinity in the mature antibodies derives primarily from stabilization of the CDR H3 loop in the antigen‐binding conformation. We have now used molecular dynamics simulations and existing crystal structures to identify potentially key maturation mutations, and we have characterized their effects on the CDR H3 loop and on antigen binding using further simulations and experimental affinity measurements, respectively. In the two maturation pathways, different contacts between light and heavy chains stabilize the CDR H3 loop. As few as two single‐site mutations in each pathway can confer substantial loop stability, but none of them confers experimentally detectable stability on its own. Our results support models of the germinal center reaction in which two or more mutations can occur without concomitant selection and show how divergent pathways have yielded functionally equivalent antibodies. Proteins 2014; 83:771–780.