Deep learning applications in protein crystallography.

Deep learning techniques can recognize complex patterns in noisy, multidimensional data. In recent years, researchers have started to explore the potential of deep learning in the field of structural biology, including protein crystallography. This field has some significant challenges, in particular producing high-quality and well ordered protein crystals. Additionally, collecting diffraction data with high completeness and quality, and determining and refining protein structures can be problematic. Protein crystallographic data are often high-dimensional, noisy and incomplete. Deep learning algorithms can extract relevant features from these data and learn to recognize patterns, which can improve the success rate of crystallization and the quality of crystal structures. This paper reviews progress in this field.

TERSE/PROLIX (TRPX) - a new algorithm for fast and lossless compression and decompression of diffraction and cryo-EM data.

High-throughput data collection in crystallography poses significant challenges in handling massive amounts of data. Here, TERSE/PROLIX (or TRPX for short) is presented, a novel lossless compression algorithm specifically designed for diffraction data. The algorithm is compared with established lossless compression algorithms implemented in gzip, bzip2, CBF (crystallographic binary file), Zstandard(zstd), LZ4 and HDF5 with gzip, LZF and bitshuffle+LZ4 filters, in terms of compression efficiency and speed, using continuous-rotation electron diffraction data of an inorganic compound and raw cryo-EM data. The results show that TRPX significantly outperforms all these algorithms in terms of speed and compression rate. It was 60 times faster than bzip2 (which achieved a similar compression rate), and more than 3 times faster than LZ4, which was the runner-up in terms of speed, but had a much worse compression rate. TRPX files are byte-order independent and upon compilation the algorithm occupies very little memory. It can therefore be readily implemented in hardware. By providing a tailored solution for diffraction and raw cryo-EM data, TRPX facilitates more efficient data analysis and interpretation while mitigating storage and transmission concerns. The C++20 compression/decompression code, custom TIFF library and an ImageJ/Fiji Java plugin for reading TRPX files are open-sourced on GitHub under the permissive MIT license.

Machine learning for classifying narrow-beam electron diffraction data.

As an alternative approach to X-ray crystallography and single-particle cryo-electron microscopy, single-molecule electron diffraction has a better signal-to-noise ratio and the potential to increase the resolution of protein models. This technology requires collection of numerous diffraction patterns, which can lead to congestion of data collection pipelines. However, only a minority of the diffraction data are useful for structure determination because the chances of hitting a protein of interest with a narrow electron beam may be small. This necessitates novel concepts for quick and accurate data selection. For this purpose, a set of machine learning algorithms for diffraction data classification has been implemented and tested. The proposed pre-processing and analysis workflow efficiently distinguished between amorphous ice and carbon support, providing proof of the principle of machine learning based identification of positions of interest. While limited in its current context, this approach exploits inherent characteristics of narrow electron beam diffraction patterns and can be extended for protein data classification and feature extraction.

Cryptosporulation in Kurthia spp. forces a rethinking of asporogenesis in Firmicutes.

Endosporulation is a complex morphophysiological process resulting in a more resistant cellular structure that is produced within the mother cell and is called endospore. Endosporulation evolved in the common ancestor of Firmicutes, but it is lost in descendant lineages classified as asporogenic. While Kurthia spp. is considered to comprise only asporogenic species, we show here that strain 11kri321, which was isolated from an oligotrophic geothermal reservoir, produces phase-bright spore-like structures. Phylogenomics of strain 11kri321 and other Kurthia strains reveals little similarity to genetic determinants of sporulation known from endosporulating Bacilli. However, morphological hallmarks of endosporulation were observed in two of the four Kurthia strains tested, resulting in spore-like structures (cryptospores). In contrast to classic endospores, these cryptospores did not protect against heat or UV damage and successive sub-culturing led to the loss of the cryptosporulating phenotype. Our findings imply that a cryptosporulation phenotype may have been prevalent and subsequently lost by laboratory culturing in other Firmicutes currently considered as asporogenic. Cryptosporulation might thus represent an ancestral but unstable and adaptive developmental state in Firmicutes that is under selection under harsh environmental conditions.

Catalytic cycling of human mitochondrial Lon protease.

The mitochondrial Lon protease (LonP1) regulates mitochondrial health by removing redundant proteins from the mitochondrial matrix. We determined LonP1 in eight nucleotide-dependent conformational states by cryoelectron microscopy (cryo-EM). The flexible assembly of N-terminal domains had 3-fold symmetry, and its orientation depended on the conformational state. We show that a conserved structural motif around T803 with a high similarity to the trypsin catalytic triad is essential for proteolysis. We show that LonP1 is not regulated by redox potential, despite the presence of two conserved cysteines at disulfide-bonding distance in its unfoldase core. Our data indicate how sequential ATP hydrolysis controls substrate protein translocation in a 6-fold binding change mechanism. Substrate protein translocation, rather than ATP hydrolysis, is a rate-limiting step, suggesting that LonP1 is a Brownian ratchet with ATP hydrolysis preventing translocation reversal. 3-fold rocking motions of the flexible N-domain assembly may assist thermal unfolding of the substrate protein.

Cryo-electron Microscopy Imaging of Alzheimer's Amyloid-beta 42 Oligomer Displayed on a Functionally and Structurally Relevant Scaffold.

Amyloid-ß peptide (Aß) oligomers are pathogenic species of amyloid aggregates in Alzheimer's disease. Like certain protein toxins, Aß oligomers permeabilize cellular membranes, presumably through a pore formation mechanism. Owing to their structural and stoichiometric heterogeneity, the structure of these pores remains to be characterized. We studied a functional Aß42-pore equivalent, created by fusing Aß42 to the oligomerizing, soluble domain of the α-hemolysin (αHL) toxin. Our data reveal Aß42-αHL oligomers to share major structural, functional, and biological properties with wild-type Aß42-pores. Single-particle cryo-EM analysis of Aß42-αHL oligomers (with an overall 3.3 Šresolution) reveals the Aß42-pore region to be intrinsically flexible. The Aß42-αHL oligomers will allow many of the features of the wild-type amyloid oligomers to be studied that cannot be otherwise, and may be a highly specific antigen for the development of immuno-base diagnostics and therapies.

Ectopic positioning of the cell division plane is associated with single amino acid substitutions in the FtsZ-recruiting SsgB in Streptomyces.

In most bacteria, cell division begins with the polymerization of the GTPase FtsZ at mid-cell, which recruits the division machinery to initiate cell constriction. In the filamentous bacterium Streptomyces, cell division is positively controlled by SsgB, which recruits FtsZ to the future septum sites and promotes Z-ring formation. Here, we show that various amino acid (aa) substitutions in the highly conserved SsgB protein result in ectopically placed septa that sever spores diagonally or along the long axis, perpendicular to the division plane. Fluorescence microscopy revealed that between 3.3% and 9.8% of the spores of strains expressing SsgB E120 variants were severed ectopically. Biochemical analysis of SsgB variant E120G revealed that its interaction with FtsZ had been maintained. The crystal structure of Streptomyces coelicolor SsgB was resolved and the key residues were mapped on the structure. Notably, residue substitutions (V115G, G118V, E120G) that are associated with septum misplacement localize in the α2-α3 loop region that links the final helix and the rest of the protein. Structural analyses and molecular simulation revealed that these residues are essential for maintaining the proper angle of helix α3. Our data suggest that besides altering FtsZ, aa substitutions in the FtsZ-recruiting protein SsgB also lead to diagonally or longitudinally divided cells in Streptomyces.

Statistically correcting dynamical electron scattering improves the refinement of protein nanocrystals, including charge refinement of coordinated metals.

Electron diffraction allows protein structure determination when only nanosized crystals are available. Nevertheless, multiple elastic (or dynamical) scattering, which is prominent in electron diffraction, is a concern. Current methods for modeling dynamical scattering by multi-slice or Bloch wave approaches are not suitable for protein crystals because they are not designed to cope with large molecules. Here, dynamical scattering of nanocrystals of insulin, thermolysin and thaumatin was limited by collecting data from thin crystals. To accurately measure the weak diffraction signal from the few unit cells in the thin crystals, a low-noise hybrid pixel Timepix electron-counting detector was used. The remaining dynamical component was further reduced in refinement using a likelihood-based correction, which was introduced previously for analyzing electron diffraction data of small-molecule nanocrystals and was adapted here for protein crystals. The procedure is shown to notably improve the structural refinement, in one case allowing the location of solvent molecules. It also allowed refinement of the charge states of bound metal atoms, an important element in protein function, through B-factor analysis of the metal atoms and their ligands. These results clearly increase the value of macromolecular electron crystallography as a complementary structural biology technique.

Shape-Shifting Peptide Nanomaterials: Surface Asymmetry Enables pH-Dependent Formation and Interconversion of Collagen Tubes and Sheets.

The fabrication of dynamic, transformable biomaterials that respond to environmental cues represents a significant step forward in the development of synthetic materials that rival their highly functional, natural counterparts. Here, we describe the design and synthesis of crystalline supramolecular architectures from charge-complementary heteromeric pairs of collagen-mimetic peptides (CMPs). Under appropriate conditions, CMP pairs spontaneously assemble into either 1D ultraporous (pore diameter >100 nm) tubes or 2D bilayer nanosheets due to the structural asymmetry that arises from heteromeric self-association. Crystalline collagen tubes represent a heretofore unobserved morphology of this common biomaterial. In-depth structural characterization from a suite of biophysical methods, including TEM, AFM, high-resolution cryo-EM, and SAXS/WAXS measurements, reveals that the sheet and tube assemblies possess a similar underlying lattice structure. The experimental evidence suggests that the tubular structures are a consequence of the self-scrolling of incipient 2D layers of collagen triple helices and that the scrolling direction determines the formation of two distinct structural isoforms. Furthermore, we show that nanosheets and tubes can spontaneously interconvert through manipulation of the assembly pH and systematic adjustment of the CMP sequence. Altogether, we establish initial guidelines for the construction of dynamically responsive 1D and 2D assemblies that undergo a structurally programmed morphological transition.

Functional and Structural Insights into a Novel Promiscuous Ketoreductase of the Lugdunomycin Biosynthetic Pathway.

Angucyclines are a structurally diverse class of actinobacterial natural products defined by their varied polycyclic ring systems, which display a wide range of biological activities. We recently discovered lugdunomycin (1), a highly rearranged polyketide antibiotic derived from the angucycline backbone that is synthesized via several yet unexplained enzymatic reactions. Here, we show via in vivo, in vitro, and structural analysis that the promiscuous reductase LugOII catalyzes both a C6 and an unprecedented C1 ketoreduction. This then sets the stage for the subsequent C-ring cleavage that is key to the rearranged scaffolds of 1. The 1.1 Å structures of LugOII in complex with either ligand 8-O-Methylrabelomycin (4) or 8-O-Methyltetrangomycin (5) and of apoenzyme were resolved, which revealed a canonical Rossman fold and a remarkable conformational change during substrate capture and release. Mutational analysis uncovered key residues for substrate access, position, and catalysis as well as specific determinants that control its dual functionality. The insights obtained in this work hold promise for the discovery and engineering of other promiscuous reductases that may be harnessed for the generation of novel biocatalysts for chemoenzymatic applications.

Sub-pixel electron detection using a convolutional neural network.

Modern direct electron detectors (DEDs) provided a giant leap in the use of cryogenic electron microscopy (cryo-EM) to study the structures of macromolecules and complexes thereof. However, the currently available commercial DEDs, all based on the monolithic active pixel sensor, still require relative long exposure times and their best results have only been obtained at 300 keV. There is a need for pixelated electron counting detectors that can be operated at a broader range of energies, at higher throughput and higher dynamic range. Hybrid Pixel Detectors (HPDs) of the Medipix family were reported to be unsuitable for cryo-EM at energies above 80 keV as those electrons would affect too many pixels. Here we show that the Timepix3, part of the Medipix family, can be used for cryo-EM applications at higher energies. We tested Timepix3 detectors on a 200 keV FEI Tecnai Arctica microscope and a 300 keV FEI Tecnai G2 Polara microscope. A correction method was developed to correct for per-pixel differences in output. Timepix3 data were simulated for individual electron events using the package Geant4Medipix. Global statistical characteristics of the simulated detector response were in good agreement with experimental results. A convolutional neural network (CNN) was trained using the simulated data to predict the incident position of the electron within a pixel cluster. After training, the CNN predicted, on average, 0.50 pixel and 0.68 pixel from the incident electron position for 200 keV and 300 keV electrons respectively. The CNN improved the MTF of experimental data at half Nyquist from 0.39 to 0.70 at 200 keV, and from 0.06 to 0.65 at 300 keV respectively. We illustrate that the useful dose-lifetime of a protein can be measured within a 1 second exposure using Timepix3.

A lipocalin mediates unidirectional heme biomineralization in malaria parasites.

During blood-stage development, malaria parasites are challenged with the detoxification of enormous amounts of heme released during the proteolytic catabolism of erythrocytic hemoglobin. They tackle this problem by sequestering heme into bioinert crystals known as hemozoin. The mechanisms underlying this biomineralization process remain enigmatic. Here, we demonstrate that both rodent and human malaria parasite species secrete and internalize a lipocalin-like protein, PV5, to control heme crystallization. Transcriptional deregulation of PV5 in the rodent parasite Plasmodium berghei results in inordinate elongation of hemozoin crystals, while conditional PV5 inactivation in the human malaria agent Plasmodium falciparum causes excessive multidirectional crystal branching. Although hemoglobin processing remains unaffected, PV5-deficient parasites generate less hemozoin. Electron diffraction analysis indicates that despite the distinct changes in crystal morphology, neither the crystalline order nor unit cell of hemozoin are affected by impaired PV5 function. Deregulation of PV5 expression renders P. berghei hypersensitive to the antimalarial drugs artesunate, chloroquine, and atovaquone, resulting in accelerated parasite clearance following drug treatment in vivo. Together, our findings demonstrate the Plasmodium-tailored role of a lipocalin family member in hemozoin formation and underscore the heme biomineralization pathway as an attractive target for therapeutic exploitation.

Extracellular Nanovesicle Enhanced Gene Transfection Using Polyethyleneimine in HEK293T Cells and Zebrafish Embryos.

It is a hot topic to improve efficiency and decrease toxicity of gene transfection reagents. The extracellular nanovesicles (EVs) that are released by cells play an important role in intercellular communication and are naturally designed for genetic exchange between cells. Here, we show that the EVs have a large beneficial effect in polyethyleneimine (PEI)-mediated transfection of a GFP-encoding plasmid into HEK293T cells. An improvement of transfection efficiency of ~500% and a decrease in toxicity were observed in a specific concentration range of PEI. The EVs also greatly improved the transfection of the same plasmid into zebrafish embryos. To verify the generality of this gene transfection approach, we also tested the cell viability and gene transfection efficiency using two other plasmids (EpTEN and ELuc) and in another cell line (A549). The measured increase in transfection efficiency makes EV a promising candidate for enhancement of the quality of current PEI-based transfection technique.

Seeded Heteroepitaxial Growth of Crystallizable Collagen Triple Helices: Engineering Multifunctional Two-Dimensional Core-Shell Nanostructures.

Engineering free-standing 2D nanomaterials with compositional, spatial, and functional control across size regimes from the nano- to mesoscale represents a significant challenge. Herein, we demonstrate a straightforward strategy for the thermodynamically controlled fabrication of multicomponent sectored nanosheets in which each sector can be chemically and spatially addressed independently and orthogonally. Collagen triple helices, comprising collagen-mimetic peptides (CMPs), are employed as molecularly programmable crystallizable units. Modulating their thermodynamic stability affords the controlled synthesis of 2D core-shell nanostructures via thermally driven heteroepitaxial growth. Structural information, gathered from SAXS and cryo-TEM, reveals that the distinct peptide domains maintain their intrinsic lattice structure and illuminates various mechanisms employed by CMP triple helices to alleviate the elastic strain associated with the interfacial lattice mismatch. Finally, we demonstrate that different sectors of the sheet surface can be selectively functionalized using bioorthogonal conjugation chemistry. Altogether, we establish a robust platform for constructing multifunctional 2D nanoarchitectures in which one can systematically program their compositional, spatial, and functional properties, which is a significant step toward their deployment into functional nanoscale devices.

The wild-type flagellar filament of the Firmicute Kurthia at 2.8 Å resolution in vivo.

Bacteria swim and swarm by rotating the micrometers long, helical filaments of their flagella. They change direction by reversing their flagellar rotation, which switches the handedness of the filament's supercoil. So far, all studied functional filaments are composed of a mixture of L- and R-state flagellin monomers. Here we show in a study of the wild type Firmicute Kurthia sp., that curved, functional filaments can adopt a conformation in vivo that is closely related to a uniform, all-L-state. This sheds additional light on transitions of the flagellar supercoil and uniquely reveals the atomic structure of a wild-type flagellar filament in vivo, including six residues showing clearly densities of O-linked glycosylation.

3D Electron Diffraction: The Nanocrystallography Revolution.

Crystallography of nanocrystalline materials has witnessed a true revolution in the past 10 years, thanks to the introduction of protocols for 3D acquisition and analysis of electron diffraction data. This method provides single-crystal data of structure solution and refinement quality, allowing the atomic structure determination of those materials that remained hitherto unknown because of their limited crystallinity. Several experimental protocols exist, which share the common idea of sampling a sequence of diffraction patterns while the crystal is tilted around a noncrystallographic axis, namely, the goniometer axis of the transmission electron microscope sample stage. This Outlook reviews most important 3D electron diffraction applications for different kinds of samples and problematics, related with both materials and life sciences. Structure refinement including dynamical scattering is also briefly discussed.

2D Crystal Engineering of Nanosheets Assembled from Helical Peptide Building Blocks.

The successful integration of 2D nanomaterials into functional devices hinges on developing fabrication methods that afford hierarchical control across length scales of the entire assembly. We demonstrate structural control over a class of crystalline 2D nanosheets assembled from collagen triple helices. By lengthening the triple helix unit through sequential additions of Pro-Hyp-Gly triads, we achieved sub-angstrom tuning over the 2D lattice. These subtle changes influence the overall nanosheet size, which can be adjusted across the mesoscale size regime. The internal structure was observed by cryo-TEM with direct electron detection, which provides real-space high-resolution images, in which individual triple helices comprising the lattice can be clearly discerned. These results establish a general strategy for tuning the structural hierarchy of 2D nanomaterials that employ rigid, cylindrical structural units.

Supramolecular architectures of molecularly thin yet robust free-standing layers.

Stable, single-nanometer thin, and free-standing two-dimensional layers with controlled molecular architectures are desired for several applications ranging from (opto-)electronic devices to nanoparticle and single-biomolecule characterization. It is, however, challenging to construct these stable single molecular layers via self-assembly, as the cohesion of those systems is ensured only by in-plane bonds. We herein demonstrate that relatively weak noncovalent bonds of limited directionality such as dipole-dipole (-CN⋅⋅⋅NC-) interactions act in a synergistic fashion to stabilize crystalline monomolecular layers of tetrafunctional calixarenes. The monolayers produced, demonstrated to be free-standing, display a well-defined atomic structure on the single-nanometer scale and are robust under a wide range of conditions including photon and electron radiation. This work opens up new avenues for the fabrication of robust, single-component, and free-standing layers via bottom-up self-assembly.

Reducing dynamical electron scattering reveals hydrogen atoms.

Compared with X-rays, electron diffraction faces a crucial challenge: dynamical electron scattering compromises structure solution and its effects can only be modelled in specific cases. Dynamical scattering can be reduced experimentally by decreasing crystal size but not without a penalty, as it also reduces the overall diffracted intensity. In this article it is shown that nanometre-sized crystals from organic pharmaceuticals allow positional refinement of the hydrogen atoms, even whilst ignoring the effects of dynamical scattering during refinement. To boost the very weak diffraction data, a highly sensitive hybrid pixel detector was employed. A general likelihood-based computational approach was also introduced for further reducing the adverse effects of dynamic scattering, which significantly improved model accuracy, even for protein crystal data at substantially lower resolution.

Inelastic scattering and solvent scattering reduce dynamical diffraction in biological crystals.

Multi-slice simulations of electron diffraction by three-dimensional protein crystals have indicated that structure solution would be severely impeded by dynamical diffraction, especially when crystals are more than a few unit cells thick. In practice, however, dynamical diffraction turned out to be less of a problem than anticipated on the basis of these simulations. Here it is shown that two scattering phenomena, which are usually omitted from multi-slice simulations, reduce the dynamical effect: solvent scattering reduces the phase differences within the exit beam and inelastic scattering followed by elastic scattering results in diffusion of dynamical scattering out of Bragg peaks. Thus, these independent phenomena provide potential reasons for the apparent discrepancy between theory and practice in protein electron crystallography.