Session Y11: Focus Session: Physics of Proteins IV

Beating the Heat: Fast Scanning Melts Beta Sheet Crystals

Beta-pleated-sheet crystals are among the most stable of protein secondary structures, and are responsible for the remarkable physical properties of many fibrous proteins, such as silk. Previous thinking was that beta-pleated-sheet crystals in the dry solid state would not melt upon input of heat energy alone. Indeed, at conventional heating rates ($\sim$1-50 $^{\circ}$C/min), silk exhibits its glass transition ($\sim$175 $^{\circ}$C), followed by cold crystallization, and then by immediate thermal degradation beginning at about 225 $^{\circ}$C. Here we demonstrate that beta-pleated-sheet crystals can melt directly from the solid state to become random coils, helices, and turns. We use fast scanning chip calorimetry at 2,000 K/s to avoid thermal degradation, and report the first reversible thermal melting of protein beta-pleated-sheet crystals, exemplified by silk fibroin. The similarity between thermal melting behavior of lamellar crystals of synthetic polymers and beta-pleated-sheet crystals is confirmed.   

Sterically allowed configuration space for amino acid dipeptides

Despite recent improvements in computational methods for protein design, we still lack a quantitative, predictive understanding of the intrinsic propensities for amino acids to be in particular backbone or side-chain conformations. This question has remained unsettled for years because of the discrepancies between different experimental approaches. To address it, I performed all-atom hard-sphere simulations of hydrophobic residues with stereo-chemical constraints and non-attractive steric interactions between non-bonded atoms for ALA, ILE, LEU and VAL dipeptide mimetics. For these hard-sphere MD simulations, I show that transitions between $\alpha$-helix and $\beta$-sheet structures only occur when the bond angle $\tau(N-C_{\alpha}-C)>110^{\circ}$, and the probability distribution of bond angles for structures in the `bridge' region of $\phi$-$\psi$ space is shifted to larger angles compared to that in other regions. In contrast, the relevant bond-angle distributions obtained from most molecular dynamics packages are broader and shifter to larger values. I encounter similar correlations between bond angles and side-chain dihedral angles. The success of these studies is an argument for re-incorporating local stereochemical constraints into computational protein design methodology.   

Generation of protein-like structures via simple rules imposed on a cubic lattice

In the current study, protein-like coarse-grained structures are generated by a simple set of rules on simple cubic lattice (SCL). The coarse-graining was based on individual amino acids. Detailed analysis of the average structure of 210 real proteins' radial distribution function (RDF) and number of neighbors as a function of cut-off distance suggest that SCL is an appropriate choice of lattice. Three simple rules were imposed to generate protein-like structures: finite size (presence of a molecular surface), random inclusion of voids, and a simple connectivity of remaining lattice sites. The set of on-lattice points (that mimic residues of a protein) satisfy many structural characteristics of real proteins. These on-lattice structures were subsequently relaxed either by random moves or by a combined Reverse Monte Carlo/Simulated Annealing (RMC/SA) algorithm that used the average RDF of proteins as its target function. The on-lattice and relaxed structures' characteristics were also analyzed via bond orientational order and graph theory. The results showed that although relaxation algorithms improved the structural characteristics of the generated structures, the improvement over the on-lattice structures are minimal. Based on various structural properties, our results indicate that the artificially generated structures closely resemble real proteins coarse-grained at the residue level.\textunderscore   

Monitoring Single-Molecule Protein Dynamics with a Carbon Nanotube Transistor

Nanoscale electronic devices like field-effect transistors have long promised to provide sensitive, label-free detection of biomolecules. Single-walled carbon nanotubes press this concept further by not just detecting molecules but also monitoring their dynamics in real time. Recent measurements have demonstrated this premise by monitoring the single-molecule processivity of three different enzymes: lysozyme [1], protein Kinase A [2], and the Klenow fragment of DNA polymerase I [3]. With all three enzymes, single molecules tethered to nanotube transistors were electronically monitored for 10 or more minutes, allowing us to directly observe a range of activity including rare transitions to chemically inactive and hyperactive conformations. The high bandwidth of the nanotube transistors further allow every individual chemical event to be clearly resolved, providing excellent statistics from tens of thousands of turnovers by a single enzyme. Initial success with three different enzymes indicates the generality and attractiveness of the nanotube devices as a new tool to complement other single-molecule techniques. Research on transduction mechanisms provides the design rules necessary to further generalize this architecture and apply it to other proteins [4]. The purposeful incorporation of just one amino acid is sufficient to fabricate effective, single molecule sensors from a wide range of enzymes or proteins.\\[4pt] [1] Y. Choi et. al., Science 335, 319 (2012); Y. Choi et. al., J. Am. Chem. Soc. 134, 2032 (2012).\\[0pt] [2] P. C. Sims et. al., JACS 135, 7861 (2013).\\[0pt] [3] T. J. Olsen et. al., JACS 135, 7855 (2013).\\[0pt] [4] Y. Choi et. al., Nano Lett. 13, 625 (2013).   

Novel computational methods to design protein-protein interactions

Despite the abundance of structural data, we still cannot accurately predict the structural and energetic changes resulting from mutations at protein interfaces. The inadequacy of current computational approaches to the analysis and design of protein-protein interactions has hampered the development of novel therapeutic and diagnostic agents. In this work, we apply a simple physical model that includes only a minimal set of geometrical constraints, excluded volume, and attractive van der Waals interactions to 1) rank the binding affinity of mutants of tetratricopeptide repeat proteins with their cognate peptides, 2) rank the energetics of binding of small designed proteins to the hydrophobic stem region of the influenza hemagglutinin protein, and 3) predict the stability of T4 lysozyme and staphylococcal nuclease mutants. This work will not only lead to a fundamental understanding of protein-protein interactions, but also to the development of efficient computational methods to rationally design protein interfaces with tunable specificity and affinity, and numerous applications in biomedicine.   

Structural basis underlying the metallic-like conductivity of microbial nanowires

Microbial nanowires are electrically conductive proteinaceous pili nanofilaments secreted by \textit{Geobacter sulfurreducens}. In contrast to current biochemical understanding that proteins are insulators, \textit{G. sulfurreducens} pili show organic metallic-like conductivity [1]. Pili also enable direct exchange of electrons among \textit{Geobacter} co-cultures [2]. Site-directed mutagenesis studies revealed that aromatic amino acids confer conductivity to pili [3]. In order to develop a structural understanding of the pili to probe the conduction mechanism at a molecular level, we employed three complementary structural methods -- X-ray microdiffraction using synchrotron radiation, rocking curve X-ray diffraction, and electron diffraction. Studies performed with all these three methods revealed a 3.2 {\AA} periodic spacing in wild-type \textit{G. sulfurreducens} pili, expected for metal-like conductivity and a lack of such spacing in genetically modified non-conductive pili. Notably, both the peak intensity and the conductivity increased 100-fold with lowering the pH from pH 10.5 to pH 2, demonstrating a structure-function correlation in pili. We also reconstructed the three dimensional tertiary structure of pili with homology modeling, which further suggested the 3.2 {\AA} spacing among aromatics associated with metal-like conductivity. \\[4pt] [1] \textit{Nature Nanotechnology}, 6, 573 (2011)\\[0pt] [2] \textit{Science}, 330, 1413 (2010)\\[0pt] [3] \textit{mBio} 4:e00105-13 (2013)   

Relaxation dynamics of proteins

We provide broadband dielectric spectra on aqueous lysozyme solutions of various concentrations and analyze the three dispersion regions commonly found. The beta-dispersion, occurring in the frequency range around 10 MHz and the gamma-dispersion arising around 20 GHz can be attributed to the rotation of the polar protein molecules in the aqueous medium and the reorientational motion of the free water molecules, respectively. The nature of the third relaxation (delta-relaxation) around 100 MHz, which is often ascribed to the motion of protein-bound water molecules, is not yet fully understood and the hydration-shell dynamics of biomolecules is an ongoing field of research [1-3]. Additional insight can be gained by analyzing the subzero temperature spectra, where the beta- and gamma-dispersions, which partly superimpose the delta-relaxation for temperatures above 273 K, disappear due to the freezing of the bulk water. In contrast, the water molecules in the protein hydration shell are known to remain in the liquid state well below the freezing point. This allows to investigate the delta-relaxation in an extended temperature range and to shed new light on the hydration-shell dynamics of biomolecules.\\[4pt] [1] W. Doster, S. Cusack, and W. Petry, Nature \textbf{337}, 754 (1989).\\[0pt] [2] M. Vogel, Phys. Rev. Lett. \textbf{101}, 225701 (2008).\\[0pt] [3] A. Benedetto, Biophys. Chem. \textbf{182}, 16 (2013).   

Protein dynamics, solvation, and quasielastic scattering

Quasielastic M\"ossbauer and neutron scattering (QES) have been used to measure protein dynamics for about 50 years. These low energy transfer spectra show two prominent features: a sharp elastic line and a broad quasielastic band. Current theory assumes that the elastic line and the quasielastic band are independent features of the spectrum, caused by motions in the sample. Current practice extracts information about dynamics from the spectra by assuming specific models with a few parameters that are determined by data fitting. We claim that this approach is flawed; it is based on questionable assumptions and has no predictive power. We propose a model where the elastic line and the broad band are one inhomogeneous spectrum of shifted, sharp natural-width lines. The model makes predictions of QES lineshapes and elastic fractions for M\"ossbauer and neutron scattering. Essential features of this description include: (i) QES lineshape and elastic fraction are sensitive to protein vibrations, and fluctuations slaved to the hydration shell and bulk solvent. (ii) Independently measured dielectric fluctuation spectra predict the QES lineshape.   

Why the observed mean square motional displacement depends on wave vector Q

The motional mean square displacement (MSD) of hydrogen (H) in proteins is extensively measured using neutron scattering techniques. The MSD increases rapidly with temperature near room temperatures and a large MSD is often associated with protein function. One shortcoming of these measurements is that the observed MSD depends on the wave vector (Q) of the neutron data used to obtain the MSD. This dependence is often attributed to use of the Gaussian approximation made to the scattering function in the analysis of the data. To test this we have simulated the protein lysozyme and calculated the intermediate scattering function (ISF), both the full ISF and the ISF in the Gaussian approximation. We find that the MSD extracted in the usual way was the same and still Q dependent in both cases. Also, direct calculation of the terms beyond the Gaussian approximation shows these terms are small. Rather, we find that the apparent Q dependence of the MSD arises from the ``dynamical diversity'' of the H in lysozyme. Specifically, if the ISF of an individual H in the protein is calculated and the MSD extracted in the usual way, then the MSD is independent of Q. This Q dependence arises from ignoring the dynamical diversity in the data analysis.   

Continuum model of non-conformational allosteric regulation

Allosteric regulation of proteins, in which the activity of one binding site on a protein is modified by the binding of a small ligand elsewhere on the protein, is traditionally understood as the result of conformational changes. It is now known that allostery is not always conformational: it may be attributed to an alteration of the thermal motion of the protein about an unchanged mean shape. We present a simple model in which the addition of a small ligand alters the thermal fluctuations about the equilibrium configuration of a continuum linear elastic caricature of a protein, and the attached ligand is treated as a small, localized shape perturbation. To determine the change in fluctuations, we develop a perturbation expansion for the change in the elastic fluctuation correlator due to the shape perturbation. We apply this scheme to a simple binding model, and calculate the change in binding energy due to the presence of a ligand.   

Minimal Mechanochemical Model for the Processivity of Myosin VI

Myosin VI is an ATPase responsible for force generation in cells. It dimerizes upon actin binding, and is proposed to walk along the actin filament. Single headed reaction mechanism of myosin VI is well understood but much of its walking mechanism remains unclear. We aim to construct a minimum model for the myosin VI walking mechanism and explore the minimal requirements for processivity. We constructed a kinetic model for the stepping mechanism of Myosin VI using minimum assumptions. The kinetics of the myosin VI dimer is modeled as a three state linear reaction network with reaction rates extracted from relevant experiments. The time limiting step in in-vitro experiments (low APT concentration) is the diffusion of detached head. In this process the myosin dimer is modeled as a tethered polymer with a flexible joint at the dimerization site. The relevance of this polymer model is checked with coarse-grained simulation. We found that the motor maintains processivity for a wide range of kinetic parameters, however long persistence length for the lever arm is crucial for processivity especially under resistive load.   

Comparison of Side-Chain Motion of Calbindin D-9k in Its Four Calcium Binding States by Molecular Dynamics Simulation

Calbindin D-9k,a small single domain protein found predominantly in tissues involved in the uptake and transport of calcium, consists of a single pair of a helix-loop-helix motif (called EF-hand) that binds calcium with the ligands provided by the loop residues and helical residues immediately adjacent to the loop. It exits in four calcium binding states: a doubly loaded state (a state with a calcium atom in each of its two binding sites), two singly loaded states (a state with calcium in its first binding site only and a state with calcium in its second binding site) and an apo-state (a state with no calcium atom). Experiments have shown that calcium binding occurs in a positive cooperative fashion. This fact is also supported by computational studies on dynamics of backbone of the protein.Studies of the methyl side chain dynamics of the doubly loaded state of the protein by molecular dynamics simulation further enhances the point. To further investigate by computation, the molecular dynamics simulation approach has been used to study the side chain dynamics of all four calcium binding states of the protein. In the study, the different kinds of force fields, especially the AMBER (a molecular dynamics simulation suit) force fields, and different kinds of water models are employed in the GPU environment.   

Lipid-Mediated Activation of G-Protein-Coupled Receptors in Membranes

The role of lipid-protein interactions in membrane function is an important question in the field of lipid membrane biophysics. Lipid effects on G-protein-coupled receptors (GPCRs) are revealed by UV-visible and FTIR spectroscopic studies of rhodopsin [1]. During rhodopsin light activation, the photoreactive 11-\textit{cis}-retinylidene chromophore is isomerized to all-\textit{trans} leading to an equilibrium between the inactive Meta-I and active Meta-II states. Modulation of the metarhodopsin equilibrium depends on the polar head groups and acyl chain composition of the membrane lipids. A flexible surface model (FSM) describes elastic coupling of the membrane bilayer to the conformational energetics of rhodopsin. According to the FSM, membrane lipids whose spontaneous curvature stabilizes the activated state within the membrane are involved in regulating protein function. The new biomembrane model explains the effects of bilayer thickness, nonlamellar-forming lipids, and osmotic stress on protein function. An ensemble-mediated activation mechanism is proposed for rhodopsin in a natural membrane lipid environment. Bulk water is involved in the activation of rhodopsin-like GPCRs in membranes [2]. Membrane proteins and membrane-bound peptides are affected by curvature forces due to elastic deformation of the bilayer, thus giving a new paradigm for membrane lipid-protein interactions in biophysics.\\[4pt] [1] M. F. Brown (2012) \textit{Biochemistry} \textbf{51}, 9782.\\[0pt] [2] A. V. Struts et al. (2011)~\textit{PNAS}$~$\textbf{108}, 8263.