Determining relevant features to recognize electron density patterns in x-ray protein crystallography

High-throughput computational methods in X-ray protein crystallography are indispensable to meet the goals of structural genomics. In particular, automated interpretation of electron density maps, especially those at mediocre resolution, can significantly speed up the protein structure determination process. TEXTAL(TM) is a software application that uses pattern recognition, case-based reasoning and nearest neighbor learning to produce reasonably refined molecular models, even with average quality data. In this work, we discuss a key issue to enable fast and accurate interpretation of typically noisy electron density data: what features should be used to characterize the density patterns, and how relevant are they? We discuss the challenges of constructing features in this domain, and describe SLIDER, an algorithm to determine the weights of these features. SLIDER searches a space of weights using ranking of matching patterns (relative to mismatching ones) as its evaluation function. Exhaustive search being intractable, SLIDER adopts a greedy approach that judiciously restricts the search space only to weight values that cause the ranking of good matches to change. We show that SLIDER contributes significantly in finding the similarity between density patterns, and discuss the sensitivity of feature relevance to the underlying similarity metric.     

Two-dimensional crystals: a powerful approach to assess structure, function and dynamics of membrane proteins

Electron crystallography and atomic force microscopy allow the study of two-dimensional membrane protein crystals. While electron crystallography provides atomic scale three-dimensional density maps, atomic force microscopy gives insight into the surface structure and dynamics at sub-nanometer resolution. Importantly, the membrane protein studied is in its native environment and its function can be assessed directly. The approach allows both the atomic structure of the membrane protein and the dynamics of its surface to be analyzed. In this way, the function-related conformational changes can be assessed, thus providing a detailed insight on the molecular mechanisms of essential biological processes.     

Automated macromolecular model building for X-ray crystallography using ARP/wARP version 7

ARP/wARP is a software suite to build macromolecular models in X-ray crystallography electron density maps. Structural genomics initiatives and the study of complex macromolecular assemblies and membrane proteins all rely on advanced methods for 3D structure determination. ARP/wARP meets these needs by providing the tools to obtain a macromolecular model automatically, with a reproducible computational procedure. ARP/wARP 7.0 tackles several tasks: iterative protein model building including a high-level decision-making control module; fast construction of the secondary structure of a protein; building flexible loops in alternate conformations; fully automated placement of ligands, including a choice of the best-fitting ligand from a 'cocktail'; and finding ordered water molecules. All protocols are easy to handle by a nonexpert user through a graphical user interface or a command line. The time required is typically a few minutes although iterative model building may take a few hours.     

A probabilistic approach to protein backbone tracing in electron density maps

One particularly time-consuming step in protein crystallography is interpreting the electron density map; that is, fitting a complete molecular model of the protein into a 3D image of the protein produced by the crystallographic process. In poor-quality electron density maps, the interpretation may require a significant amount of a crystallographer's time. Our work investigates automating the time-consuming initial backbone trace in poor-quality density maps. We describe ACMI (Automatic Crystallographic Map Interpreter), which uses a probabilistic model known as a Markov field to represent the protein. Residues of the protein are modeled as nodes in a graph, while edges model pairwise structural interactions. Modeling the protein in this manner allows the model to be flexible, considering an almost infinite number of possible conformations, while rejecting any that are physically impossible. Using an efficient algorithm for approximate inference--belief propagation--allows the most probable trace of the protein's backbone through the density map to be determined. We test ACMI on a set of ten protein density maps (at 2.5 to 4.0 A resolution), and compare our results to alternative approaches. At these resolutions, ACMI offers a more accurate backbone trace than current approaches.     

Constructing and validating initial Cα models from subnanometer resolution density maps with pathwalking

A significant number of macromolecular structures solved by electron cryo-microscopy and X-ray crystallography obtain resolutions of 3.5-6?, at which direct atomistic interpretation is difficult. To address this, we developed pathwalking, a semi-automated protocol to enumerate reasonable Cα models from near-atomic resolution density maps without a structural template or sequence-structure correspondence. Pathwalking uses an approach derived from the Traveling Salesman Problem to rapidly generate an ensemble of initial models for individual proteins, which can later be optimized to produce full atomic models. Pathwalking can also be used to validate and identify potential structural ambiguities in models generated from near-atomic resolution density maps. In this work, examples from the EMDB and PDB are used to assess the broad applicability and accuracy of our method. With the growing number of near-atomic resolution density maps from cryo-EM and X-ray crystallography, pathwalking can become an important tool in modeling protein structures.     

A database method for automated map interpretation in protein crystallography.

A significant portion of new protein structures contain folds that are related to those seen before. During the development of a computer program that can accurately position, in electron density maps, large protein domains with large structural deviations, it became apparent that the redundancy in protein folds could be used in a non trivial manner during a protein structure determination. As a result a computational procedure, Database Assisted Density Interpretation (DADI), was developed and tested to aid in the building of models in protein crystallography and to assist in interpreting electron density maps. The initial tests of the DADI procedure using a small database of protein domains are described. The philosophy is to first work with entire domains then with the secondary structure elements of these domains and finally with individual residues of the secondary structure elements via Monte Carlo, "chopping" and "clipping" procedures, respectively. The first test case was a traceable 3.2 A multiple isomorphous replacement with anomalous scattering (MIRAS) electron density map of a human topoisomerase I-DNA complex. The second test case uses poor electron density for the third domain of the diphtheria toxin repressor resulting from a molecular replacement solution with the first two domains. Despite the fact that a fairly small database was employed in these test cases, the DADI procedure was able to find a large portion of the protein backbone with very few errors. In the first case nearly 45% of the backbone and more than 80% of the secondary structure was placed automatically. In the second test case nearly 50% of the third domain was automatically detected. A particular encouraging result was that in both cases more than 75% of the beta sheet secondary structure was found automatically by the DADI procedure. Clearly, the procedures employed are promising avenues to exploit the current explosion of protein structures for the determination of future structures. Proteins 1999;36:526-541.     

Analytical trapping: extraction of time-independent structures from time-dependent crystallographic data

All chemical and biological reactions involve atomic motion, embodied in dynamic structural changes. Identifying these changes is the goal of time-resolved crystallography. The "raw" output of a time-resolved macromolecular crystallography experiment is the time-dependent set of difference electron density maps that span the desired time range and display the time-dependent changes in density (and underlying structure) as the reaction progresses. The goal is to interpret such data in terms of a small number of crystallographically refinable, time-independent structures, each associated with a reaction intermediate; to establish the pathways and rate coefficients by which the intermediates interconvert; and thus to establish a chemical kinetic mechanism. We review briefly the various strategies that may be used to achieve this goal and concentrate on two promising advances: singular value decomposition and cluster analysis. The strategies are illustrated by using data on the photocycle of the bacterial blue light photoreceptor, photoactive yellow protein.     

Sequence information within the lamB genes in required for proper routing of the bacteriophage lambda receptor protein to the outer membrane of Escherichia coli K-12

The structure of Satellite tobacco necrosis virus (STNV) has been determined to 3.0 Å resolution by X-ray crystallography. Electron density maps were obtained with phases based on one heavy-atom derivative and several cycles of phase refinement using the 60-fold non-crystallographic symmetry in the particle. A model for one protein subunit was built using a computer graphics display. The subunit is constructed mainly of a β-roll structure forming two β-sheets, each of four antiparallel strands. The N-termini of the subunits form bundles of three α-helices extending into the RNA region of the virus at the 3-fold axis. The topology of the polypeptide chain is the same as, and the conformation clearly similar to, that of the shell domains of the Tomato bushy stunt virus (TBSV) and Southern bean mosaic virus (SBMV) protein subunits. The subunit packing in the T = 1 STNV structure is, however, significantly different from the packing of these T = 3 viruses: parts of some of the structural elements facing the RNA in TBSV and SBMV are utilized for subunit-subunit contacts in STNV. No RNA structure is obvious in the present icosahedrally averaged electron density maps. The protein surface facing the RNA contains mainly hydrophilic residues, especially lysine and arginine.     

Combining efficient conformational sampling with a deformable elastic network model facilitates structure refinement at low resolution

Structural studies of large proteins and protein assemblies are a difficult and pressing challenge in molecular biology. Experiments often yield only low-resolution or sparse data that are not sufficient to fully determine atomistic structures. We have developed a general geometry-based algorithm that efficiently samples conformational space under constraints imposed by low-resolution density maps obtained from electron microscopy or X-ray crystallography experiments. A deformable elastic network (DEN) is used to restrain the sampling to prior knowledge of an approximate structure. The DEN restraints dramatically reduce over-fitting, especially at low resolution. Cross-validation is used to optimally weight the structural information and experimental data. Our algorithm is robust even for noise-added density maps and has a large radius of convergence for our test case. The DEN restraints can also be used to enhance reciprocal space simulated annealing refinement.     

Another piece of the ribosome: solution structure of S16 and its location in the 30S subunit

BACKGROUND: X-ray crystallography has recently yielded much-improved electron-density maps of the bacterial ribosome and its two subunits and many structural details of bacterial ribosome subunits are now being resolved. One approach to complement the structures and elucidate the details of rRNA and protein packing is to determine structures of individual protein components and model these into existing intermediate resolution electron density. RESULTS: We have determined the solution structure of the ribosomal protein S16 from Thermus thermophilus. S16 is a mixed alpha/beta protein with a novel folding scaffold based on a five-stranded antiparallel/parallel beta sheet. Three large loops, which are partially disordered, extend from the sheet and two alpha helices are packed against its concave surface. Calculations of surface electrostatic potentials show a large continuous area of positive electrostatic potential and smaller areas of negative potential. S16 was modeled into a 5.5 A electron-density map of the T. thermophilus 30S ribosomal subunit. CONCLUSIONS: The location and orientation of S16 in a narrow crevice formed by helix 21 and several other unassigned rRNA helices is consistent with electron density corresponding to the shape of S16, hydroxyl radical protection data, and the electrostatic surface potential of S16. Two protein neighbors to S16 are S4 and S20, which facilitate binding of S16 to the 30S subunit. Overall, this work exemplifies the benefits of combining high-resolution nuclear magnetic resonance (NMR) structures of individual components with low-resolution X-ray maps to elucidate structures of large complexes.     

Cluster analysis of time-dependent crystallographic data: Direct identification of time-independent structural intermediates

The initial output of a time-resolved macromolecular crystallography experiment is a time-dependent series of difference electron density maps that displays the time-dependent changes in underlying structure as a reaction progresses. The goal is to interpret such data in terms of a small number of crystallographically refinable, time-independent structures, each associated with a reaction intermediate; to establish the pathways and rate coefficients by which these intermediates interconvert; and thereby to elucidate a chemical kinetic mechanism. One strategy toward achieving this goal is to use cluster analysis, a statistical method that groups objects based on their similarity. If the difference electron density at a particular voxel in the time-dependent difference electron density (TDED) maps is sensitive to the presence of one and only one intermediate, then its temporal evolution will exactly parallel the concentration profile of that intermediate with time. The rationale is therefore to cluster voxels with respect to the shapes of their TDEDs, so that each group or cluster of voxels corresponds to one structural intermediate. Clusters of voxels whose TDEDs reflect the presence of two or more specific intermediates can also be identified. From such groupings one can then infer the number of intermediates, obtain their time-independent difference density characteristics, and refine the structure of each intermediate. We review the principles of cluster analysis and clustering algorithms in a crystallographic context, and describe the application of the method to simulated and experimental time-resolved crystallographic data for the photocycle of photoactive yellow protein.     

Review: electron crystallography: present excitement, a nod to the past, anticipating the future

From a modest beginning with negatively stained samples of the helical T4 bacteriophage tail, electron crystallography has emerged as a powerful tool in structural biology. High-resolution density maps, interpretable in terms of an atomic structure, can be obtained from specimens prepared as well-ordered, two-dimensional crystals, and the resolution achieved with helical specimens and icosahedral viruses is approaching the same goal. A hybrid approach to determining the molecular structure of complex biological assemblies is generating great interest, in which high-resolution structures that have been determined for individual protein components are fitted into a lower resolution envelope of the large complex. With this as background, how much more can be anticipated for the future? Considerable scope still remains to improve the quality of electron microscope images. Automation of data acquisition and data processing, together with the emergence of computational speeds of 10(12) floating point operations per second or higher, will make it possible to extend high-resolution structure determination into the realm of single-particle microscopy. As a result, computational alignment of single particles, i.e., the formation of "virtual crystals," can begin to replace biochemical crystallization. Since single-particle microscopy may remain limited to "large" structures of 200 to 300 kDa or more, however, smaller proteins will continue to be studied as helical assemblies or as two-dimensional crystals. The further development of electron crystallography is thus likely to turn increasingly to the use of single particles and small regions of ordered assemblies, emphasizing more and more the potential for faster, higher throughput.     

Neutron crystallographic studies of T4 lysozyme at cryogenic temperature

Bacteriophage T4 lysozyme (T4L) has been used as a paradigm for seminal biophysical studies on protein structure, dynamics, and stability. Approximately 700 mutants of this protein and their respective complexes have been characterized by X‐ray crystallography; however, despite the high resolution diffraction limits attained in several studies, no hydrogen atoms were reported being visualized in the electron density maps. To address this, a 2.2 Å‐resolution neutron data set was collected at 80 K from a crystal of perdeuterated T4L pseudo‐wild type. We describe a near complete atomic structure of T4L, which includes the positions of 1737 hydrogen atoms determined by neutron crystallography. The cryogenic neutron model reveals explicit detail of the hydrogen bonding interactions in the protein, in addition to the protonation states of several important residues.     

Photosystem II: the engine of life

Photosystem II (PS II) is a multisubunit membrane protein complex, which uses light energy to oxidize water and reduce plastoquinone. High-resolution electron cryomicroscopy and X-ray crystallography are revealing the structure of this important molecular machine. Both approaches have contributed to our understanding of the organization of the transmembrane helices of higher plant and cyanobacterial PS II and both indicate that PS II normally functions as a dimer. However the high-resolution electron density maps derived from X-ray crystallography currently at 3.7/3.8 A, have allowed assignments to be made to the redox active cofactors involved in the light-driven water-plastoquinone oxidoreductase activity and to the chlorophyll molecules that absorb and transfer energy to the reaction centre. In particular the X-ray work has identified density that can accommodate the four manganese atoms which catalyse the water-oxidation process. The Mn cluster is located at the lumenal surface of the DI protein and approximately 7 A from the redox active tyrosine residue (YZ) which acts an electron/proton transfer link to the primary oxidant P680.+. The lower resolution electron microscopy studies, however, are providing structural models of larger PS II supercomplexes that are ideal frameworks in which to incorporate the X-ray derived structures.     

Crystal and electron microscopy structures of sticholysin II actinoporin reveal insights into the mechanism of membrane pore formation

Sticholysin II (StnII) is a pore-forming protein (PFP) produced by the sea anemone Stichodactyla helianthus. We found out that StnII exists in a monomeric soluble state but forms tetramers in the presence of a lipidic interface. Both structures have been independently determined at 1.7 A and 18 A resolution, respectively, by using X-ray crystallography and electron microscopy of two-dimensional crystals. Besides, the structure of soluble StnII complexed with phosphocholine, determined at 2.4 A resolution, reveals a phospholipid headgroup binding site, which is located in a region with an unusually high abundance of aromatic residues. Fitting of the atomic model into the electron microscopy density envelope suggests that while the beta sandwich structure of the protein remains intact upon oligomerization, the N-terminal region and a flexible and highly basic loop undergo significant conformational changes. These results provide the structural basis for the membrane recognition step of actinoporins and unexpected insights into the oligomerization step.     

XtalView/Xfit--A versatile program for manipulating atomic coordinates and electron density

Xfit is a model-building and map viewing program in XtalView that is used by the structural biology community including researchers in the fields of crystallography, molecular modeling, and electron microscopy. Among its distinguishing features are built-in fast Fourier transforms that allow users flexibility in map calculations including the creation of OMIT maps and the updating of structure factors to reflect model changes from within the program. Written in C and using the freely available XView toolkit, it is highly portable to almost any X-windows based workstation including Intel-based LINUX systems. Its user interface is designed to aid in facile model-building and contains a semiautomated fitting system that allows the user to interactively and rapidly build chain de novo into an electron density map. The program is highly optimized to allow such features as interactive contour levels and map calculations to be completed within a few seconds. Features in the latest version including phase-combination, solvent-flattening, automated water addition, and small-probe dot contact surfaces, as well as basic design features, are discussed.     

Focusing-in on microtubules

A good approximation of the atomic structure of a microtubule has been derived from docking the high-resolution structure of tubulin, solved by electron crystallography, into lower resolution maps of whole microtubules. Some structural interactions with other molecules, including nucleotides, drugs, motor proteins and microtubule-associated proteins, can now be predicted.     

Knowledge-based real-space explorations for low-resolution structure determination

The accurate and effective interpretation of low-resolution data in X-ray crystallography is becoming increasingly important as structural initiatives turn toward large multiprotein complexes. Substantial challenges remain due to the poor information content and ambiguity in the interpretation of electron density maps at low resolution. Here, we describe a semiautomated procedure that employs a restraint-based conformational search algorithm, RAPPER, to produce a starting model for the structure determination of ligase interacting factor 1 in complex with a fragment of DNA ligase IV at low resolution. The combined use of experimental data and a priori knowledge of protein structure enabled us not only to generate an all-atom model but also to reaffirm the inferred sequence registry. This approach provides a means to extract quickly from experimental data useful information that would otherwise be discarded and to take into account the uncertainty in the interpretation--an overriding issue for low-resolution data.