Folding pathway of the b1 domain of protein G explored by multiscale modeling

From coarse-grained to atomic-level characterization of protein dynamics: transition state for the folding of B domain of protein A

Atomic-level molecular dynamics simulations are widely used for the characterization of the structural dynamics of proteins; however, they are limited to shorter time scales than the duration of most of the relevant biological processes. Properly designed coarse-grained models that trade atomic resolution for efficient sampling allow access to much longer time-scales. In-depth understanding of the structural dynamics, however, must involve atomic details. In this study, we tested a method for the rapid reconstruction of all-atom models from α carbon atom positions in the application to convert a coarse-grained folding trajectory of a well described model system: the B domain of protein A. The results show that the method and the spatial resolution of the resulting coarse-grained models enable computationally inexpensive reconstruction of realistic all-atom models. Additionally, by means of structural clustering, we determined the most persistent ensembles of the key folding step, the transition state. Importantly, the analysis of the overall structural topologies suggests a dominant folding pathway. This, together with the all-atom characterization of the obtained ensembles, in the form of contact maps, matches the experimental results well.

Structural features that predict real-value fluctuations of globular proteins

It is crucial to consider dynamics for understanding the biological function of proteins. We used a large number of molecular dynamics (MD) trajectories of nonhomologous proteins as references and examined static structural features of proteins that are most relevant to fluctuations. We examined correlation of individual structural features with fluctuations and further investigated effective combinations of features for predicting the real value of residue fluctuations using the support vector regression (SVR). It was found that some structural features have higher correlation than crystallographic B-factors with fluctuations observed in MD trajectories. Moreover, SVR that uses combinations of static structural features showed accurate prediction of fluctuations with an average Pearson's correlation coefficient of 0.669 and a root mean square error of 1.04 Å. This correlation coefficient is higher than the one observed in predictions by the Gaussian network model (GNM). An advantage of the developed method over the GNMs is that the former predicts the real value of fluctuation. The results help improve our understanding of relationships between protein structure and fluctuation. Furthermore, the developed method provides a convienient practial way to predict fluctuations of proteins using easily computed static structural features of proteins.

GB1 is not a two-state folder: identification and characterization of an on-pathway intermediate

The folding pathway of the small α/β protein GB1 has been extensively studied during the past two decades using both theoretical and experimental approaches. These studies provided a consensus view that the protein folds in a two-state manner. Here, we reassessed the folding of GB1, both by experiments and simulations, and detected the presence of an on-pathway intermediate. This intermediate has eluded earlier experimental characterization and is distinct from the collapsed state previously identified using ultrarapid mixing. Failure to identify the presence of an intermediate affects some of the conclusions that have been drawn for GB1, a popular model for protein folding studies.

Folding simulations of the A and B domains of protein G

We study wild type and mutants of the A and B domain of protein G using all-atom Go-models. Our data substantiate the usefulness of such simulation for probing the folding mechanism of proteins and demonstrate that multifunnel versions of such models also allow probing of more complicated funnel landscapes. In our case, such models reproduce the experimentally observed distributions of the GA98 and GB98 mutants which differ only by one residue but fold into different structures. They also reveal details on the folding mechanism in these two proteins.

Durandal: fast exact clustering of protein decoys

In protein folding, clustering is commonly used as one way to identify the best decoy produced. Initializing the pairwise distance matrix for a large decoy set is computationally expensive. We have proposed a fast method that works even on large decoy sets. This method is implemented in a software called Durandal. Durandal has been shown to be consistently faster than other software performing fast exact clustering. In some cases, Durandal can even outperform the speed of an approximate method. Durandal uses the triangular inequality to accelerate exact clustering, without compromising the distance function. Recently, we have further enhanced the performance of Durandal by incorporating a Quaternion-based characteristic polynomial method that has increased the speed of Durandal between 13% and 27% compared with the previous version. Durandal source code is available under the GNU General Public License at http://www.riken.jp/zhangiru/software/durandal_released_qcp.tgz. Alternatively, a compiled version of Durandal is also distributed with the nightly builds of the Phenix (http://www.phenix-online.org/) crystallographic software suite (Adams et al., Acta Crystallogr Sect D 2010, 66, 213).

Simulation of chaperonin effect on protein folding: a shift from nucleation-condensation to framework mechanism

The iterative annealing mechanism (IAM) of chaperonin-assisted protein folding is explored in a framework of a well-established coarse-grained protein modeling tool, which enables the study of protein dynamics in a time-scale well beyond classical all-atom molecular mechanics. The chaperonin mechanism of action is simulated for two paradigm systems of protein folding, B domain of protein A (BdpA) and B1 domain of protein G (GB1), and compared to chaperonin-free simulations presented here for BdpA and recently published for GB1. The prediction of the BdpA transition state ensemble (TSE) is in perfect agreement with experimental findings. It is shown that periodic distortion of the polypeptide chains by hydrophobic chaperonin interactions can promote rapid folding and leads to a decrease in folding temperature. It is also demonstrated how chaperonin action prevents kinetically trapped conformations and modulates the observed folding mechanisms from nucleation-condensation to a more framework-like.

Coarse-grained force field: general folding theory

We review the coarse-grained UNited RESidue (UNRES) force field for the simulations of protein structure and dynamics, which is being developed in our laboratory over the last several years. UNRES is a physics-based force field, the prototype of which is defined as a potential of mean force of polypeptide chains in water, where all the degrees of freedom except the coordinates of α-carbon atoms and side-chain centers have been integrated out. We describe the initial implementation of UNRES to protein-structure prediction formulated as a search for the global minimum of the potential-energy function and its subsequent molecular dynamics and extensions of molecular-dynamics implementation, which enabled us to study protein-folding pathways and thermodynamics, as well as to reformulate the protein-structure prediction problem as a search for the conformational ensemble with the lowest free energy at temperatures below the folding-transition temperature. Applications of UNRES to study biological problems are also described.

Entropy-accelerated exact clustering of protein decoys

MOTIVATION

Clustering is commonly used to identify the best decoy among many generated in protein structure prediction when using energy alone is insufficient. Calculation of the pairwise distance matrix for a large decoy set is computationally expensive. Typically, only a reduced set of decoys using energy filtering is subjected to clustering analysis. A fast clustering method for a large decoy set would be beneficial to protein structure prediction and this still poses a challenge.

RESULTS

We propose a method using propagation of geometric constraints to accelerate exact clustering, without compromising the distance measure. Our method can be used with any metric distance. Metrics that are expensive to compute and have known cheap lower and upper bounds will benefit most from the method. We compared our method's accuracy against published results from the SPICKER clustering software on 40 large decoy sets from the I-TASSER protein folding engine. We also performed some additional speed comparisons on six targets from the 'semfold' decoy set. In our tests, our method chose a better decoy than the energy criterion in 25 out of 40 cases versus 20 for SPICKER. Our method also was shown to be consistently faster than another fast software performing exact clustering named Calibur. In some cases, our approach can even outperform the speed of an approximate method.

AVAILABILITY

Our C++ software is released under the GNU General Public License. It can be downloaded from http://www.riken.jp/zhangiru/software/durandal_released.tgz.

Characterizing protein energy landscape by self-learning multiscale simulations: application to a designed β-hairpin

Characterizing the energy landscape of proteins at atomic resolution is still a very challenging problem, since it simultaneously requires high accuracy in estimating specific interactions and high efficiency in conformational sampling. Here, for these two requirements to meet, we extended the self-learning multiscale simulation (SLMS) method developed recently and applied it to the designed β-hairpin CLN025. The SLMS integrates all-atom and coarse-grained (CG) models in an iterative way such that the conformational sampling is performed by the CG model, the AA energy is used to calibrate the energy landscape, and the CG model is improved by the calibrated energy landscape. We extended the SLMS in two aspects, use of the energy decomposition for self-learning of the CG potential and a two-bead/residue CG model. The results show that the self-learning greatly improved the CG potential, and with the derived CG potential, the β-hairpin CLN025 robustly folded to the native structure. The self-learning iteration progressively enhanced the context dependence in the CG potential and increased the energy gap between the native and the denatured states of the CG model, leading to a funnel-like energy landscape. By using the SLMS method, without prior knowledge of the native structure but with the help of the AA energy, we can obtain a tailor-made CG potential specific to the target protein. The method can be useful for de novo structure prediction as well.

Theoretical study of molecular mechanism of binding TRAP220 coactivator to Retinoid X Receptor alpha, activated by 9-cis retinoic acid

Study on molecular mechanism of conformational reorientation of RXR-alpha ligand binding domain is presented. We employed CABS--a reduced model of protein dynamics to model folding pathways of binding 9-cis retinoic acid to apo-RXR molecule and TRAP220 peptide fragment to the holo form. Based on obtained results we also propose a sequential model of RXR activation by 9-cis retinoic acid and TRAP220 coactivator. Methodology presented here may be used for investigation of binding pathways of other NR/hormone/cofactor sets.

Determination of side-chain-rotamer and side-chain and backbone virtual-bond-stretching potentials of mean force from AM1 energy surfaces of terminally-blocked amino-acid residues, for coarse-grained simulations of protein structure and folding. I. The method

In this and the accompanying article, we report the development of new physics-based side-chain-rotamer and virtual-bond-deformation potentials which now replace the respective statistical potentials used so far in our physics-based united-reside UNRES force field for large-scale simulations of protein structure and dynamics. In this article, we describe the methodology for determining the corresponding potentials of mean force (PMF's) from the energy surfaces of terminally-blocked amino-acid residues calculated with the AM1 quantum-mechanical semiempirical method. The approach is based on minimization of the AM1 energy for fixed values of the angles lambda for rotation of the peptide groups about the C(alpha)...C(alpha) virtual bonds, and for fixed values of the side-chain dihedral angles chi, which formed a multidimensional grid. A harmonic-approximation approach was developed to extrapolate from the energy at a given grid point to other points of the conformational space to compute the respective contributions to the PMF. To test the applicability of the harmonic approximation, the rotamer PMF's of alanine and valine obtained with this approach have been compared with those obtained by using a Metropolis Monte Carlo method. The PMF surfaces computed with the harmonic approximation are more rugged and have more pronounced minima than the MC-calculated surfaces but the harmonic-approximation- and MC-calculated PMF values are linearly correlated. The potentials derived with the harmonic approximation are, therefore, appropriate for UNRES for which the weights (scaling factors) of the energy terms are determined by force-field optimization for foldability.

Multiscale methods for protein folding simulations

Inherently hierarchic nature of proteins makes multiscale computational methods especially useful in the studies of folding and other functional dynamics. With the multiscale strategies, one can achieve improved accuracy and efficiency by coupling the atomistic and the coarse grained simulations. Depending on the problems studied, very different implementation protocols can be used to realize the multiscale idea. Here, we give detailed introductions to the currently used multiscale protocols, together with some recent applications to the protein folding simulations in our group. The advantages and weakness, as well as the application scopes of these multiscale protocols are discussed. The directions for the future developments are also proposed.

Mechanism of formation of the C-terminal beta-hairpin of the B3 domain of the immunoglobulin binding protein G from Streptococcus. III. Dynamics of long-range hydrophobic interactions

A 20-residue peptide, IG(42-61), derived from the C-terminal beta-hairpin of the B3 domain of the immunoglobulin binding protein G from Streptoccocus was studied using circular dichroism, nuclear magnetic resonance (NMR) spectroscopy at various temperatures and by differential scanning calorimetry (DSC). Unlike other related peptides studied so far, this peptide displays two heat capacity peaks in DSC measurements (at a scanning rate of 1.5 deg/min at a peptide concentration of 0.07 mM), which suggests a three-state folding/unfolding process. The results from DSC and NMR measurements suggest the formation of a dynamic network of hydrophobic interactions stabilizing the structure, which resembles a beta-hairpin shape over a wide range of temperatures (283-313 K). Our results show that IG (42-61) possesses a well-organized three-dimensional structure stabilized by long-range hydrophobic interactions (Tyr50 ... Phe57 and Trp48 ... Val59) at T = 283 K and (Trp48 ... Val59) at 305 and 313 K. The mechanism of beta-hairpin folding and unfolding, as well as the influence of peptide length on its conformational properties, are also discussed.

The energy profiles of atomic conformational transition intermediates of adenylate kinase

The elastic network interpolation (ENI) (Kim et al., Biophys J 2002;83:1620-1630) is a computationally efficient and physically realistic method to generate conformational transition intermediates between two forms of a given protein. However it can be asked whether these calculated conformations provide good representatives for these intermediates. In this study, we use ENI to generate conformational transition intermediates between the open form and the closed form of adenylate kinase (AK). Based on C(alpha)-only intermediates, we construct atomic intermediates by grafting all the atoms of known AK structures onto the C(alpha) atoms and then perform CHARMM energy minimization to remove steric conflicts and optimize these intermediate structures. We compare the energy profiles for all intermediates from both the CHARMM force-field and from knowledge-based energy functions. We find that the CHARMM energies can successfully capture the two energy minima representing the open AK and closed AK forms, while the energies computed from the knowledge-based energy functions can detect the local energy minimum representing the closed AK form and show some general features of the transition pathway with a somewhat similar energy profile as the CHARMM energies. The combinatorial extension structural alignment (Shindyalov et al., 1998;11:739-747) and the k-means clustering algorithm are then used to show that known PDB structures closely resemble computed intermediates along the transition pathway.

Modeling of loops in proteins: a multi-method approach

Background

Template-target sequence alignment and loop modeling are key components of protein comparative modeling. Short loops can be predicted with high accuracy using structural fragments from other, not necessairly homologous proteins, or by various minimization methods. For longer loops multiscale approaches employing coarse-grained de novo modeling techniques should be more effective.

Results

For a representative set of protein structures of various structural classes test predictions of loop regions have been performed using MODELLER, ROSETTA, and a CABS coarse-grained de novo modeling tool. Loops of various length, from 4 to 25 residues, were modeled assuming an ideal target-template alignment of the remaining portions of the protein. It has been shown that classical modeling with MODELLER is usually better for short loops, while coarse-grained de novo modeling is more effective for longer loops. Even very long missing fragments in protein structures could be effectively modeled. Resolution of such models is usually on the level 2-6 Å, which could be sufficient for guiding protein engineering. Further improvement of modeling accuracy could be achieved by the combination of different methods. In particular, we used 10 top ranked models from sets of 500 models generated by MODELLER as multiple templates for CABS modeling. On average, the resulting molecular models were better than the models from individual methods.

Conclusions

Accuracy of protein modeling, as demonstrated for the problem of loop modeling, could be improved by the combinations of different modeling techniques.

Implementation of molecular dynamics and its extensions with the coarse-grained UNRES force field on massively parallel systems; towards millisecond-scale simulations of protein structure, dynamics, and thermodynamics

We report the implementation of our united-residue UNRES force field for simulations of protein structure and dynamics with massively parallel architectures. In addition to coarse-grained parallelism already implemented in our previous work, in which each conformation was treated by a different task, we introduce a fine-grained level in which energy and gradient evaluation are split between several tasks. The Message Passing Interface (MPI) libraries have been utilized to construct the parallel code. The parallel performance of the code has been tested on a professional Beowulf cluster (Xeon Quad Core), a Cray XT3 supercomputer, and two IBM BlueGene/P supercomputers with canonical and replica-exchange molecular dynamics. With IBM BlueGene/P, about 50 % efficiency and 120-fold speed-up of the fine-grained part was achieved for a single trajectory of a 767-residue protein with use of 256 processors/trajectory. Because of averaging over the fast degrees of freedom, UNRES provides an effective 1000-fold speed-up compared to the experimental time scale and, therefore, enables us to effectively carry out millisecond-scale simulations of proteins with 500 and more amino-acid residues in days of wall-clock time.

Mechanism of formation of the C-terminal beta-hairpin of the B3 domain of the immunoglobulin-binding protein G from Streptococcus. IV. Implication for the mechanism of folding of the parent protein

A 34-residue alpha/beta peptide [IG(28-61)], derived from the C-terminal part of the B3 domain of the immunoglobulin binding protein G from Streptoccocus, was studied using CD and NMR spectroscopy at various temperatures and by differential scanning calorimetry. It was found that the C-terminal part (a 16-residue-long fragment) of this peptide, which corresponds to the sequence of the beta-hairpin in the native structure, forms structure similar to the beta-hairpin only at T = 313 K, and the structure is stabilized by non-native long-range hydrophobic interactions (Val47-Val59). On the other hand, the N-terminal part of IG(28-61), which corresponds to the middle alpha-helix in the native structure, is unstructured at low temperature (283 K) and forms an alpha-helix-like structure at 305 K, and only one helical turn is observed at 313 K. At all temperatures at which NMR experiments were performed (283, 305, and 313 K), we do not observe any long-range connectivities which would have supported packing between the C-terminal (beta-hairpin) and the N-terminal (alpha-helix) parts of the sequence. Such interactions are absent, in contrast to the folding pathway of the B domain of protein G, proposed recently by Kmiecik and Kolinski (Biophys J 2008, 94, 726-736), based on Monte-Carlo dynamics studies. Alternative folding mechanisms are proposed and discussed.

Exploring the parameter space of the coarse-grained UNRES force field by random search: selecting a transferable medium-resolution force field

We explored the energy-parameter space of our coarse-grained UNRES force field for large-scale ab initio simulations of protein folding, to obtain good initial approximations for hierarchical optimization of the force field with new virtual-bond-angle bending and side-chain-rotamer potentials which we recently introduced to replace the statistical potentials. 100 sets of energy-term weights were generated randomly, and good sets were selected by carrying out replica-exchange molecular dynamics simulations of two peptides with a minimal alpha-helical and a minimal beta-hairpin fold, respectively: the tryptophan cage (PDB code: 1L2Y) and tryptophan zipper (PDB code: 1LE1). Eight sets of parameters produced native-like structures of these two peptides. These eight sets were tested on two larger proteins: the engrailed homeodomain (PDB code: 1ENH) and FBP WW domain (PDB code: 1E0L); two sets were found to produce native-like conformations of these proteins. These two sets were tested further on a larger set of nine proteins with alpha or alpha + beta structure and found to locate native-like structures of most of them. These results demonstrate that, in addition to finding reasonable initial starting points for optimization, an extensive search of parameter space is a powerful method to produce a transferable force field.

The nucleation mechanism of protein folding: a survey of computer simulation studies

The nucleation mechanism of protein folding, originally proposed by Baldwin in the early 1970s, was firstly observed by Shakhnovich and co-workers two decades later in the context of Monte Carlo simulations of a simple lattice model. At about the same time the extensive use of φ-value analysis provided the first experimental evidence that the folding of Chymotrypsin-inhibitor 2, a small single-domain protein, which folds with two-state kinetics, is also driven by a nucleation mechanism. Since then, the nucleation mechanism is generally considered the most common form of folding mechanism amongst two-state proteins. However, recent experimental data has put forward the idea that this may not necessarily be so, since the accuracy of the experimentally determined φ values, which are used to identify the critical (i.e. nucleating) residues, is typically poor. Here, we provide a survey of in silico results on the nucleation mechanism, ranging from simple lattice Monte Carlo to more sophisticated off-lattice molecular dynamics simulations, and discuss them in light of experimental data.

Conformational studies of the alpha-helical 28-43 fragment of the B3 domain of the immunoglobulin binding protein G from Streptococcus

To determine whether the alpha-helix in the B3 immunoglobulin binding domain of protein G from group G Streptococcus has conformational stability as an isolated fragment, we carried out a CD and NMR study of the 16-residue peptide in solution corresponding to this alpha-helix. Based on two-dimensional H-NMR spectra recorded at three different temperatures (283, 305, and 313 K), it was found that this peptide is mostly unstructured in water at these temperatures. Weak signals corresponding to i,i+3 or i,i+4 interactions, which are characteristic of formation of turn-like structures, were observed in the ROE spectra at all temperatures. The absence of a stable three-dimensional structure of the investigated peptide supports an earlier study (Blanco and Serrano, Eur J Biochem 1995, 230, 634-649) of a possible mechanism for folding of other (B1 and B2) immunoglobulin binding domains of Protein G.

Predicting protein folding pathways at the mesoscopic level based on native interactions between secondary structure elements

Background

Since experimental determination of protein folding pathways remains difficult, computational techniques are often used to simulate protein folding. Most current techniques to predict protein folding pathways are computationally intensive and are suitable only for small proteins.

Results

By assuming that the native structure of a protein is known and representing each intermediate conformation as a collection of fully folded structures in which each of them contains a set of interacting secondary structure elements, we show that it is possible to significantly reduce the conformation space while still being able to predict the most energetically favorable folding pathway of large proteins with hundreds of residues at the mesoscopic level, including the pig muscle phosphoglycerate kinase with 416 residues. The model is detailed enough to distinguish between different folding pathways of structurally very similar proteins, including the streptococcal protein G and the peptostreptococcal protein L. The model is also able to recognize the differences between the folding pathways of protein G and its two structurally similar variants NuG1 and NuG2, which are even harder to distinguish. We show that this strategy can produce accurate predictions on many other proteins with experimentally determined intermediate folding states.

Conclusion

Our technique is efficient enough to predict folding pathways for both large and small proteins at the mesoscopic level. Such a strategy is often the only feasible choice for large proteins. A software program implementing this strategy (SSFold) is available at .

Computational techniques for efficient conformational sampling of proteins

In this review, we summarize the computational methods for sampling the conformational space of biomacromolecules. We discuss the methods applicable to find only lowest energy conformations (global minimization of the potential-energy function) and to generate canonical ensembles (canonical Monte Carlo method and canonical molecular dynamics method and their extensions). Special attention is devoted to the use of coarse-grained models that enable simulations to be enhanced by several orders of magnitude.