Determination of three-dimensional protein structures from nuclear magnetic resonance data using fragments of known structures

Chemical shift-based methods in NMR structure determination

Chemical shifts are highly sensitive probes harnessed by NMR spectroscopists and structural biologists as conformational parameters to characterize a range of biological molecules. Traditionally, assignment of chemical shifts has been a labor-intensive process requiring numerous samples and a suite of multidimensional experiments. Over the past two decades, the development of complementary computational approaches has bolstered the analysis, interpretation and utilization of chemical shifts for elucidation of high resolution protein and nucleic acid structures. Here, we review the development and application of chemical shift-based methods for structure determination with a focus on ab initio fragment assembly, comparative modeling, oligomeric systems, and automated assignment methods. Throughout our discussion, we point out practical uses, as well as advantages and caveats, of using chemical shifts in structure modeling. We additionally highlight (i) hybrid methods that employ chemical shifts with other types of NMR restraints (residual dipolar couplings, paramagnetic relaxation enhancements and pseudocontact shifts) that allow for improved accuracy and resolution of generated 3D structures, (ii) the utilization of chemical shifts to model the structures of sparsely populated excited states, and (iii) modeling of sidechain conformations. Finally, we briefly discuss the advantages of contemporary methods that employ sparse NMR data recorded using site-specific isotope labeling schemes for chemical shift-driven structure determination of larger molecules. With this review, we aim to emphasize the accessibility and versatility of chemical shifts for structure determination of challenging biological systems, and to point out emerging areas of development that lead us towards the next generation of tools.

Improved chemical shift based fragment selection for CS-Rosetta using Rosetta3 fragment picker

A new fragment picker has been developed for CS-Rosetta that combines beneficial features of the original fragment picker, MFR, used with CS-Rosetta, and the fragment picker, NNMake, that was used for purely sequence based fragment selection in the context of ROSETTA de-novo structure prediction. Additionally, the new fragment picker has reduced sensitivity to outliers and other difficult to match data points rendering the protocol more robust and less likely to introduce bias towards wrong conformations in cases where data is bad, missing or inconclusive. The fragment picker protocol gives significant improvements on 6 of 23 CS-Rosetta targets. An independent benchmark on 39 protein targets, whose NMR data sets were published only after protocol optimization had been finished, also show significantly improved performance for the new fragment picker (van der Schot et al. in J Biomol NMR, 2013).

Recurrent structural motifs in non-homologous protein structures

We have extracted an extensive collection of recurrent structural motifs (RSMs), which consist of sequentially non-contiguous structural motifs (4–6 residues), each of which appears with very similar conformation in three or more mutually unrelated protein structures. We find that the proteins in our set are covered to a substantial extent by the recurrent non-contiguous structural motifs, especially the helix and strand regions. Computational alanine scanning calculations indicate that the average folding free energy changes upon alanine mutation for most types of non-alanine residues are higher for amino acids that are present in recurrent structural motifs than for amino acids that are not. The non-alanine amino acids that are most common in the recurrent structural motifs, i.e., phenylalanine, isoleucine, leucine, valine and tyrosine and the less abundant methionine and tryptophan, have the largest folding free energy changes. This indicates that the recurrent structural motifs, as we define them, describe recurrent structural patterns that are important for protein stability. In view of their properties, such structural motifs are potentially useful for inter-residue contact prediction and protein structure refinement.

Protein-Structure Prediction by Recombination of Fragments

The field of protein-structure prediction has been revolutionized by the application of "mix-and-match" methods both in template-based homology modeling and in template-free de novo folding. Consensus analysis and recombination of fragments copied from known protein structures is currently the only approach that allows the building of models that are closer to the native structure of the target protein than the structure of its closest homologue. It is also the most successful approach in cases in which the target protein exhibits a novel three-dimensional fold. This review summarizes the recent developments in both template-based and template-free protein structure modeling and compares the available methods for protein-structure prediction by recombination of fragments. A convergence between the "protein folding" and "protein evolution" schools of thought is postulated.

Molecular Fragment Replacement Approach to Protein Structure Determination by Chemical Shift and Dipolar Homology Database Mining

A novel approach is described for determining backbone structures of proteins that is based on finding fragments in the protein data bank (PDB). For each fragment in the target protein, usually chosen to be 7-10 residues in length, PDB fragments are selected that best fit to experimentally determined one-bond heteronuclear dipolar couplings and that show agreement between chemical shifts predicted for the PDB fragment and experimental values for the target fragment. These fragments are subsequently refined by simulated annealing to improve agreement with the experimental data. If the lowest-energy refined fragments form a unique structural cluster, this structure is accepted and side chains are added on the basis of a conformational database potential. The sequential backbone assembly process extends the chain by translating an accepted fragment onto it. For several small proteins, with extensive sets of dipolar couplings measured in two alignment media, a unique final structure is obtained that agrees well with structures previously solved by conventional methods. With less dipolar input data, large, oriented fragments of each protein are obtained, but their relative positioning requires either a small set of translationally restraining nuclear Overhauser enhancements (NOEs) or a protocol that optimizes burial of hydrophobic groups and pairing of beta-strands.

Modeling of globular proteins. A distance-based data search procedure for the construction of insertion/deletion regions and Pro----non-Pro mutations

A distance-based database search scheme is proposed for modeling Pro----in non-Pro and insertion/deletion regions of homologous globular proteins up to six residues in length. In the first step, geometric descriptors, the number of residues involved and target distances corresponding to the separation of C alpha atom positions adjacent to the "missing" segment, are chosen. In the second step, a database of high-resolution X-ray structures is scanned for segments with similar descriptors and selected segments are binned according to conformational type. In the third and fourth steps, the selected conformations are docked into the protein, and geometric and energetic criteria are used to determine their viability as segment models. The fifth step consists of an interaction scheme in which the geometric descriptors are redefined. This compensates for the use of a limited database and/or for the use of a poor original protein model adjacent to the missing segment. The procedure has been tested on Pro----non-Pro mutations in the homologous proteins penicillopepsin and endothiapepsin, and on the insertion/deletion regions of the homologs penicillopepsin and endothiapepsin, trypsin and gamma-chymotrypsin and hen and human lysozyme. The test cases represent a wide variety of secondary structural elements (helix, sheet, turn and coil) and insertion/deletion lengths (0 to 4 residues). It is shown that 79% of the test cases are accurately modeled (within 0.54 A root-mean-square (r.m.s.) deviation for main-chain atoms) using the proposed scheme. Failure of the scheme (main-chain atom r.m.s. deviations greater than 1.29 A) in 21% of the cases appears to be related to the presence of infrequently observed conformations or locally unique folds of the target proteins with respect to the database (18% of the test cases); the remaining 3% are unexplained. Geometric and energetic criteria are able to discriminate between trial conformations that correspond to the X-ray structures and those that are different in 97% of the conformations generated by the distance-weighted database search scheme. The scheme is shown to be relatively insensitive to uncertainty in the template co-ordinates, since the geometric descriptors were taken from the homologous protein (r.m.s. deviations in the position of descriptors range from 0.18 to 1.35 A for the accurately modeled test cases). It is demonstrated that the scheme can be used to correct local sequence misalignments.

1H-NMR stereospecific assignments by conformational data-base searches

A search procedure is described for making stereospecific assignments at prochiral centers in proteins on the basis of nuclear Overhauser enhancement and coupling constant data derived from nmr experiments. A data base comprising torsion angles, associated 1H-1H coupling constants and interproton distances is searched by a computer algorithm for sets of values that match the experimental data within specified error limits. Two different data bases are used. The first is a crystallographic data base derived from 34 well-refined crystal structures; the second is a systematic data base derived from conformations of a short peptide fragment with idealized geometry by systematically varying the phi, psi, and chi 1 torsion angles. Both approaches are tested for beta-methylene groups with model data obtained from 20 crystal structures. The results for the two methods are similar though not identical, so that a combination of the two methods appears to be useful. With an appropriate choice of error estimates, around 80% of the beta-methylene groups could be assigned in the test calculations. In addition, results with experimental nmr data indicate that a similar percentage of stereospecific assignments can be made in practical situations.

Comparative modeling methods: application to the family of the mammalian serine proteases

Comparative modeling methods are described that can be used to construct a three-dimensional model structure of a new protein from knowledge of its sequence and of the experimental structures and sequences of other members of its homology family. The methods are illustrated with the mammalian serine protease family, for which seven experimental structures have been reported in the literature, and the sequences for over 35 different protein members of the family are available. The strategy for modeling these proteins is presented, and criteria are developed for determining and assigning the reliability of the modeled structure. Criteria are described that are specially designed to help detect cases in which it is likely that the local structure diverges significantly from the usual conformation of the family.

Crystal structure of a protein-toxin alpha 1-purothionin at 2.5A and a comparison with predicted models

Alpha 1-Purothionin (alpha 1-P), a wheatgerm protein and lytic toxin, has a secondary and tertiary structure similar to that of crambin as revealed by CD and NMR studies. alpha 1-P crystallizes in the tetragonal space group 1422 with unit cell dimensions: a = b = 53.59 and c = 69.79 A. X-ray diffraction data have been measured to 2.5 A Bragg spacing. The crystal structure has been determined by molecular replacement methods, using an energy-minimized alpha 1-P model structure derived from crambin (Whitlow and Teeter: Journal of Biomolecular Structure and Dynamics 2:831-848, 1985, Journal of the American Chemical Society 108:7163-7172, 1986). The energy-minimized model gives a slightly cleaner rotation solution and better refinement against the x-ray data than do the crambin or unminimized alpha 1-P structures. The final crystallographic residual with the data in the 10-2.5 A resolution range is 0.216. The refined alpha 1-P structure has a backbone rms difference of 0.74 A from crambin and 0.55 A from the energy-minimized alpha 1-P model. A low resolution NMR model of alpha 1-P calculated from metric matrix distance geometry and restrained molecular dynamics differs from crambin's backbone by 2.3 A rms deviation (Clore et al.: EMBO Journal 5:2729-2735, 1986). Backbone dihedral angles for our predicted model differ from the refined alpha 1-P structure in only one region (at a turn where there is a deletion relative to crambin). The NMR model had differences in four regions.

Determination of three-dimensional structures of proteins and nucleic acids in solution by nuclear magnetic resonance spectroscopy

Nuclear magnetic resonance (NMR) spectroscopy has evolved over the last decade into a powerful method for determining three-dimensional structures of biological macromolecules in solution. Key advances have been the introduction of two-dimensional experiments, high-field superconducting magnets, and computational procedures for converting the NMR-derived interproton distances and torsion angles into three-dimensional structures. This article outlines the methodology employed, describes the major NMR experiments necessary for the spectral analysis of macromolecules, and discusses the computational approaches employed to date. The present state of the art is illustrated using a variety of examples, and future developments are indicated.

PKB: a program system and data base for analysis of protein structure

PKB is a computer program system that combines a data base of three-dimensional protein structures with a series of algorithms for pattern recognition, data analysis, and graphics. By typing relatively simple commands the user may search the data base for instances of a structural motif and analyze in detail the set of individual structures that are found. The application of PKB to the study of protein folding is illustrated in three examples. The first analysis compares the conformations observed for a short sequential motif, sequences similar to the cell-attachment signal Arg-Gly-Asp. The second compares sequences observed for a conformational motif, a 16-residue beta alpha beta unit. The third analysis considers a population of substructures containing ion-pair interactions, examining the relationship of frequency of occurrence to calculated electrostatic energy.

Determination of three-dimensional structures of proteins from interproton distance data by dynamical simulated annealing from a random array of atoms. Circumventing problems associated with folding

A new real space method, based on the principles of simulated annealing, is presented for determining protein structures on the basis of interproton distance restraints derived from NMR data. The method circumvents the folding problem associated with all real space methods described to date, by starting from a completely random array of atoms and introducing the force constants for the covalent, interproton distance and repulsive van der Waals terms in the target function appropriately. The system is simulated at high temperature by solving Newton's equations of motion. As the values of all force constants are very low during the early stages of the simulation, energy barriers between different folds of the protein can be overcome, and the global minimum of the target function is reliably located. Further, because the atoms are initially only weakly coupled, they can move essentially independently to satisfy the restraints. The method is illustrated using two examples of small proteins, namely crambin (46 residues) and potato carboxypeptidase inhibitor (39 residues).

Identification of predictive sequence motifs limited by protein structure data base size

Associations between short amino acid sequence patterns and protein secondary structure classes can be found by searching a data base of known protein structures. Analysis of these associations suggests that secondary structure of proteins can be determined locally by sequence motifs of high predictive value, but at present our ability to find these motifs is limited by the size of the available data bases.