DiffraGAN: a conditional generative adversarial network for phasing single molecule diffraction data to atomic resolution.

Introduction: Proteins that adopt multiple conformations pose significant challenges in structural biology research and pharmaceutical development, as structure determination via single particle cryo-electron microscopy (cryo-EM) is often impeded by data heterogeneity. In this context, the enhanced signal-to-noise ratio of single molecule cryo-electron diffraction (simED) offers a promising alternative. However, a significant challenge in diffraction methods is the loss of phase information, which is crucial for accurate structure determination. Methods: Here, we present DiffraGAN, a conditional generative adversarial network (cGAN) that estimates the missing phases at high resolution from a combination of single particle high-resolution diffraction data and low-resolution image data. Results: For simulated datasets, DiffraGAN allows effectively determining protein structures at atomic resolution from diffraction patterns and noisy low-resolution images. Discussion: Our findings suggest that combining single particle cryo-electron diffraction with advanced generative modeling, as in DiffraGAN, could revolutionize the way protein structures are determined, offering an alternative and complementary approach to existing methods.

Ab initio structure determination of nanocrystals of organic pharmaceutical compounds by electron diffraction at room temperature using a Timepix quantum area direct electron detector. Corrigendum.

Corrections are made to Table 1 in the article by van Genderen et al. [Acta Cryst. (2016), A72, 236-242].

Protein structure determination by electron diffraction using a single three-dimensional nanocrystal.

Three-dimensional nanometre-sized crystals of macromolecules currently resist structure elucidation by single-crystal X-ray crystallography. Here, a single nanocrystal with a diffracting volume of only 0.14 µm3, i.e. no more than 6 × 105 unit cells, provided sufficient information to determine the structure of a rare dimeric polymorph of hen egg-white lysozyme by electron crystallography. This is at least an order of magnitude smaller than was previously possible. The molecular-replacement solution, based on a monomeric polyalanine model, provided sufficient phasing power to show side-chain density, and automated model building was used to reconstruct the side chains. Diffraction data were acquired using the rotation method with parallel beam diffraction on a Titan Krios transmission electron microscope equipped with a novel in-house-designed 1024 × 1024 pixel Timepix hybrid pixel detector for low-dose diffraction data collection. Favourable detector characteristics include the ability to accurately discriminate single high-energy electrons from X-rays and count them, fast readout to finely sample reciprocal space and a high dynamic range. This work, together with other recent milestones, suggests that electron crystallography can provide an attractive alternative in determining biological structures.

Lattice filter for processing image data of three-dimensional protein nanocrystals.

When 300 kV cryo-EM images at Scherzer focus are acquired from ∼ 100 nm thick three-dimensional protein nanocrystals using a Falcon 2 direct electron detector, Fourier transformation can reveal the crystalline lattice to surprisingly high resolutions, even though the images themselves seem to be devoid of any contrast. Here, it is reported how this lattice information can be enhanced by means of a wave finder in combination with Wiener-type maximum-likelihood filtering. This procedure paves the way towards full three-dimensional structure determination at high resolution for protein crystals.

Ab initio structure determination of nanocrystals of organic pharmaceutical compounds by electron diffraction at room temperature using a Timepix quantum area direct electron detector.

Until recently, structure determination by transmission electron microscopy of beam-sensitive three-dimensional nanocrystals required electron diffraction tomography data collection at liquid-nitrogen temperature, in order to reduce radiation damage. Here it is shown that the novel Timepix detector combines a high dynamic range with a very high signal-to-noise ratio and single-electron sensitivity, enabling ab initio phasing of beam-sensitive organic compounds. Low-dose electron diffraction data (∼ 0.013 e(-) Å(-2) s(-1)) were collected at room temperature with the rotation method. It was ascertained that the data were of sufficient quality for structure solution using direct methods using software developed for X-ray crystallography (XDS, SHELX) and for electron crystallography (ADT3D/PETS, SIR2014).

Medipix 2 detector applied to low energy electron microscopy.

Low energy electron microscopy (LEEM) and photo-emission electron microscopy (PEEM) traditionally use microchannel plates (MCPs), a phosphor screen and a CCD-camera to record images and diffraction patterns. In recent years, however, MCPs have become a limiting factor for these types of microscopy. Here, we report on a successful test series using a solid state hybrid pixel detector, Medipix 2, in LEEM and PEEM. Medipix 2 is a background-free detector with an infinite dynamic range, making it very promising for both real-space imaging and spectroscopy. We demonstrate a significant enhancement of both image contrast and resolution, as compared to MCPs. Since aging of the Medipix 2 detector is negligible for the electron energies used in LEEM/PEEM, we expect Medipix to become the detector of choice for a new generation of systems.

Cyclops: new modular software suite for cryo-EM.

Cyclops is a new computer program designed as a graphical front-end that allows easy control and interaction with tasks and programs for 3D reconstruction of biological complexes using cryo-electron microscopy. Cyclops' current plug-ins are designed for automated particle picking and include two new algorithms, automated carbon masking and quaternion based rotation space sampling, which are also presented here. Additional plug-ins are in the pipeline. Cyclops allows straightforward organization and visualization of all data and tasks and allows both interactive and batch-wise processing. Furthermore, it was designed for straightforward implementation in grid architectures. As a front-end to a collection of programs it provides a common interface to these programs, thus enhancing the usability of the suite and the productivity of the user.

The impact of single amino acid substitutions in CD3gamma on the CD3epsilongamma interaction and T-cell receptor-CD3 complex formation.

The human T-cell receptor-CD3 complex consists of at least eight polypeptide chains; CD3gamma- and delta-dimers associate with the disulfide linked alphabeta- and zetazeta-dimers to form a functional receptor complex. The exact structure of this complex is still unknown. We now have examined the interaction between CD3gamma and CD3 in human T-cells. For this purpose, we have generated site-directed mutants of CD3gamma that were introduced in human T-cells defective in CD3gamma expression. Cell-surface and intracellular expression of the introduced CD3gamma chains was determined, as was the association with CD3delta, CD3, and the T-cell receptor. Although the introduction of wild type CD3gamma and CD3gamma (78Y-F) fully restored T-cell receptor assembly and expression, the introduction of CD3gamma (82C-S), CD3gamma (85C-S), and CD3gamma (76Q-E) all resulted in an impaired association between CD3gamma and CD3 and a lack of cell-surface expressed CD3gamma. Finally, the introduction of CD3gamma (76Q-L) and CD3gamma (78Y-A) restored the expression of TCR-CD3deltagammazeta2 complexes, although the association between CD3gamma and CD3 was impaired. These results indicate that several amino acids in CD3gamma are essential for an optimal association between CD3gamma and CD3 and the assembly of a cell-surface expressed TCR-CD3deltagammazeta2 complex.

TYSON: robust searching, sorting, and selecting of single particles in electron micrographs.

We here present TYSON, a new program for automatic and semi-automatic particle selection from electron micrographs. TYSON employs a three-step strategy of searching, sorting and selecting single particles. In the first step, TYSON finds the positions of potential particles by one of three different methods: local averaging, template matching or local variance. The practical merits and drawbacks of these methods are discussed. In the second step, these potential particles are automatically sorted according to their probability of being true positives. Many criteria are provided for this sort. In the final -interactive- step, whole categories of poorly fitting false positives can be removed with a single mouse-click. We present results obtained using cryo-EM micrographs of both spherical virus particles and asymmetric particles. The procedures are fast and use of TYSON allowed, for example, some 20,000 particles to be selected in a single working day.

Apoptin's functional N- and C-termini independently bind DNA.

Apoptin induces apoptosis specifically in tumour cells, where Apoptin is enriched in the DNA-dense heterochromatin and nucleoli. In vitro, Apoptin interacts with dsDNA, forming large nucleoprotein superstructures likely to be relevant for apoptosis induction. Its N- and C-terminal domains also have cell-killing activity, although they are less potent than the full-length protein. Here, we report that both Apoptin's N- and C-terminal halves separately bound DNA, indicating multiple independent binding sites. The reduced cell killing activity of both truncation mutants was mirrored in vitro by a reduced affinity compared to full-length Apoptin. However, none of the truncation mutants cooperatively bound DNA or formed superstructures, which suggests that cooperative DNA binding by Apoptin is required for the formation of nucleoprotein superstructures. As Apoptin's N- and C-terminal fragments not only share apoptotic activity, but also affinity for DNA, we propose that both properties are functionally linked.

Matrix methods for solving protein substructures of chlorine and sulfur from anomalous data.

The weak signal obtained from the anomalous scattering (at lambda = 1.54 A) of naturally occurring elements such as sulfur, phosphorus and ordered solvent chloride ions is used to determine the atomic positions of these atoms. Two examples are discussed: the sulfur and chlorine substructure of tetragonal hen egg-white lysozyme and an oligonucleotide containing ten P atoms. The substructure of lysozyme was also solved from Cu K(alpha) radiation data collected on a standard rotating-anode generator. The results presented here are an illustration of the power of the matrix methods, which are to be implemented in next distribution of the direct methods package CRUNCH.

Structure elucidation of beta-mannanase: from the electron-density map to the DNA sequence.

The crystal structure of affinity-purified Thermomonospora fusca beta-mannanase has been solved despite the lack of the major part of the amino-acid sequence. A high-quality electron-density map allowed the identification of a stretch of eight amino acids close to the C-terminus which was used to design a degenerate downstream PCR primer. Together with a specific primer previously derived from the N-terminus, 95.7% of the mannanase gene sequence was obtained from genomic T. fusca DNA by PCR. The structure-derived sequence was then compared with the DNA-derived sequence and corrected when necessary. Applying the presented protocol, there was no need to manually build a model at an early stage of structure determination, an erroneous and tedious process, especially in the absence of the amino-acid sequence. Using the DNA sequence information and the current version of ARP/wARP, 281 residues, or 93% of the polypeptide chain (including side chains), were built and refined to an R factor of 16.5% without any manual intervention.

Crystal structure of the C-terminal SH2 domain of the p85alpha regulatory subunit of phosphoinositide 3-kinase: an SH2 domain mimicking its own substrate.

The binding properties of Src homology-2 (SH2) domains to phosphotyrosine (pY)-containing peptides have been studied in recent years with the elucidation of a large number of crystal and solution structures. Taken together, these structures suggest a general mode of binding of pY-containing peptides, explain the specificities of different SH2 domains, and may be used to design inhibitors of pY binding by SH2 domain-containing proteins. We now report the crystal structure to 1.8 A resolution of the C-terminal SH2 domain (C-SH2) of the P85alpha regulatory subunit of phosphoinositide 3-kinase (PI3 K). Surprisingly, the carboxylate group of Asp2 from a neighbouring molecule occupies the phosphotyrosine binding site and interacts with Arg18 (alphaA2) and Arg36 (betaB5), in a similar manner to the phosphotyrosine-protein interactions seen in structures of other SH2 domains complexed with pY peptides. It is the first example of a non-phosphate-containing, non-aromatic mimetic of phosphotyrosine binding to SH2 domains, and this could have implications for the design of substrate analogues and inhibitors. Overall, the crystal structure closely resembles the solution structure, but a number of loops which demonstrate mobility in solution are well defined by the crystal packing. C-SH2 has adopted a binding conformation reminiscent of the ligand bound N-terminal SH2 domain of PI3K, apparently induced by the substrate mimicking of a neighbouring molecule in the crystal.

The structure of bovine mitochondrial F1-ATPase: an example of rotary catalysis.

There is now compelling evidence in support of a rotary catalytic mechanism in F1-ATPase, and, by extension, in the intact ATP synthase. Although models have been proposed to explain how protein translocation in F0 results in rotation of the gamma-subunit relative to the alpha 3/beta 3 assembly in F1 [22], these are still speculative. It seems likely that a satisfactory explanation of this mechanism will ultimately depend on structural information on the intact ATP synthase.

New developments in phase refinement.

A longstanding problem in X-ray crystallography is that vital information regarding the crystal phases in missing from the experimental data that are gathered in the diffraction experiment. Prior knowledge needs to be introduced in order to resolve phase ambiguities whenever the diffraction data are not sufficient to unequivocally reconstruct the crystal phases through anomalous or isomorphous differences. Very recent developments include progress in the application of direct methods to small proteins and other compounds of a similar small size (Shake 'n' Bake, SHELXD, CRUNCH and SIR96), bias-free refinement through the gamma-correction (Solomon), improvements in the determination of phase probability distributions (SHARP) and automated atomic refinement (wARP).

Implications for function and therapy of a 2.9 A structure of binary-complexed antithrombin.

The crystal structure of a binary complex of human antithrombin with a peptide of the same sequence as its reactive loop (P14-P3) has been determined at 2.9 A. The peptide binds as the middle strand s4A in the A beta-sheet, homologously to that of the reactive loop in the latent and cleaved forms of antithrombin. Peptide binding results in the complete expulsion of the hinge region of the loop from the A beta-sheet although the conformation differs from that of heparin-activated antithrombin. The 36-fold increase in the rate of reaction of the binary complex with factor Xa indicates that full loop expulsion alone is not sufficient for complete heparin activation of antithrombin but that this is also dependent on the overall conformation of the molecule. Previous studies have demonstrated that reactive loop peptides can block or reverse the polymerisation of serpins associated with cirrhosis and thrombosis. The antithrombin binary complex structure defines the precise localisation of the blocking peptide in a serpin and provides the basis for rational drug design for mimetics that will prevent polymerisation in vivo and so ameliorate the associated disease.

Wild-type alpha 1-antitrypsin is in the canonical inhibitory conformation.

alpha 1-Antitrypsin is the archetypal member of the serine proteinase inhibitor or serpin superfamily. Members of the family show structural homology based on a dominant A beta-sheet and a mobile reactive centre loop. Our recent crystal structure of alpha 1-antitrypsin stabilized with a point mutation showed the loop to be in a canonical inhibitory conformation in the absence of significant insertion into the A beta-sheet. It could be argued that the stabilizing mutation may induce the reactive centre loop to adopt an artificial, and unrepresentative, conformation and the finding seems to be at variance with studies assessing rates of peptide insertion into the A beta-sheet and limited proteolysis of the reactive loop. Here we present a 2.9 A structure of recombinant wild-type alpha 1-antitrypsin with no stabilizing mutations. Again, the reactive loop is in a canonical conformation in the absence of significant insertion into the A beta-sheet. A stabilizing salt bridge between P5 glutamate and arginine residues 196, 223 and 281, already identified in the mutant, provides strong evidence that this conformation is not an artefact of crystallization but represents the conformation of the circulating inhibitor in vivo. Comparison with the structure of alpha 1-antitrypsin stabilized with the Phe51Leu mutation indicates that the increased thermal stability of the mutant results from enhanced packing of aromatic residues in the hydrophobic core of the molecule. The structure of wild-type alpha 1-antitrypsin reveals a hydrophobic pocket between s2A and helices D and E that is filled on reactive loop insertion and the formation of biologically relevant loop-sheet polymers. This pocket may provide a target for rational drug design to prevent the formation of polymers and the associated plasma deficiency, liver cirrhosis and emphysema.

Bias reduction in phase refinement by modified interference functions: introducing the gamma correction.

The chemical, physical and symmetry constraints of an electron-density map impose relationships between structure factors, and these relationships are exploited during refinement. However, constraints often allow an artificially high correlation between the model and the original structure factors, a flaw known as model or refinement bias. Elimination of the bias component of a constrained model, the component insensitive to constraints, enhances the power of phase-refinement techniques. The scale of the bias component, here denoted as gamma, is shown to be equal in magnitude to the origin vector of the interference function G that defines the relationships between the structure factors. The gamma correction leads to solvent flipping in the case of phase improvement by solvent flattening, and other types of constraint allow a similar treatment.