Generating Chromosome Geometries at the Single-Cell Level from Cryo-Electron Tomograms

To solve three-dimensional structures of biological macromolecules in situ, large numbers of particles often need to be picked from cryo-electron tomograms. However, adoption of automated particle-picking methods remains limited because of their technical limitations. To overcome the limitations, we develop DeepETPicker, a deep learning model for fast and accurate picking of particles from cryo-electron tomograms. Training of DeepETPicker requires only weak supervision with low numbers of simplified labels, reducing the burden of manual annotation. The simplified labels combined with the customized and lightweight model architecture of DeepETPicker and accelerated pooling enable substantial performance improvement. When tested on simulated and real tomograms, DeepETPicker outperforms the competing state-of-the-art methods by achieving the highest overall accuracy and speed, which translate into higher authenticity and coordinates accuracy of picked particles and higher resolutions of final reconstruction maps. DeepETPicker is provided in open source with a user-friendly interface to support cryo-electron tomography in situ.

BackgroundCryo-electron tomography is an important tool to study structures of macromolecular complexes in close to native states. A whole cell cryo electron tomogram contains structural information of all its macromolecular complexes. However, extracting this information remains challenging, and relies on sophisticated image processing, in particular for template-free particle extraction, classification and averaging. To develop these methods it is crucial to realistically simulate tomograms of crowded cellular environments, which can then serve as ground truth models for assessing and optimizing methods for detection of complexes in cell tomograms.ResultsWe present a framework to generate crowded mixtures of macromolecular complexes for realistically simulating cryo electron tomograms including noise and image distortions due to the missing-wedge effects. Simulated tomograms are then used for assessing the template-free Difference-of-Gaussian (DoG) particle-picking method to detect complexes of different shapes and sizes under various crowding and noise levels. We identified DoG parameter settings that maximize precision and recall for detecting particles over a wide range of sizes and shapes. We observed that medium sized DoG scaling factors showed the overall best performance. To further improve performance, we propose a combination strategy for integrating results from multiple parameter settings. With increasing macromolecular crowding levels, the precision of particle picking remained relatively high, while the recall was dramatically reduced, which limits the detection of sufficient copy numbers of complexes in a crowded environment. Over a wide range of increasing noise levels, the DoG particle picking performance remained stable, but dramatically reduced beyond a specific noise threshold.ConclusionsAutomatic and reference-free particle picking is an important first step in a visual proteomics analysis of cell tomograms. However, cell cytoplasm is highly crowded, which makes particle detection challenging. It is therefore important to test particle-picking methods in a realistic crowded setting. Here, we present a framework for simulating tomograms of cellular environments at high crowding levels and assess the DoG particle picking method. We determined optimal parameter settings to maximize the performance of the DoG particle-picking method.Electronic supplementary materialThe online version of this article (doi:10.1186/s12859-016-1283-3) contains supplementary material, which is available to authorized users.

Hierarchical detection and analysis of macromolecular complexes in cryo-electron tomograms using Pyto software.

The resolution of subtomogram averages calculated from cryo-electron tomograms (cryo-ET) of crowded cellular environments is often limited due to signal-loss in, and misalignment of the subtomograms. In contrast, single-particle cryo-electron microscopy (SP-cryo-EM) routinely reaches near-atomic resolution of isolated complexes. We have developed a method called “TomographY-Guided 3D REconstruction of Subcellular Structures” (TYGRESS) that is a hybrid of cryo-ET and SP-cryo-EM, and is able to achieve close-to-nanometer resolution of complexes inside crowded cellular environments. TYGRESS combines the advantages of SP-cryo-EM (images with good signal-to-noise ratio/contrast and minimal radiation damage) and subtomogram averaging (3D-alignment of macromolecules in a complex sample). Using TYGRESS, we determined the structure of the intact ciliary axoneme with up to 12 Å resolution. These results reveal many structural details that were not visible by cryo-ET. TYGRESS is generally applicable to cellular complexes that are amenable to subtomogram averaging, bringing us a step closer to (pseudo-)atomic models of cells.

Conical Fourier shell correlation applied to electron tomograms.

The creation and comparison of minimal genomes will help better define the most fundamental mechanisms supporting life. Mesoplasma florum is a near-minimal, fast-growing, nonpathogenic bacterium potentially amenable to genome reduction efforts. In a comparative genomic study of 13 M.florum strains, including 11 newly sequenced genomes, we have identified the core genome and open pangenome of this species. Our results show that all of the strains have approximately 80% of their gene content in common. Of the remaining 20%, 17% of the genes were found in multiple strains and 3% were unique to any given strain. On the basis of random transposon mutagenesis, we also estimated that ~290 out of 720 genes are essential for M.florum L1 in rich medium. We next evaluated different genome reduction scenarios for M.florum L1 by using gene conservation and essentiality data, as well as comparisons with the first working approximation of a minimal organism, Mycoplasma mycoides JCVI-syn3.0. Our results suggest that 409 of the 473 M.mycoides JCVI-syn3.0 genes have orthologs in M.florum L1. Conversely, 57 putatively essential M.florum L1 genes have no homolog in M.mycoides JCVI-syn3.0. This suggests differences in minimal genome compositions, even for these evolutionarily closely related bacteria. IMPORTANCE The last years have witnessed the development of whole-genome cloning and transplantation methods and the complete synthesis of entire chromosomes. Recently, the first minimal cell, Mycoplasma mycoides JCVI-syn3.0, was created. Despite these milestone achievements, several questions remain to be answered. For example, is the composition of minimal genomes virtually identical in phylogenetically related species? On the basis of comparative genomics and transposon mutagenesis, we investigated this question by using an alternative model, Mesoplasma florum, that is also amenable to genome reduction efforts. Our results suggest that the creation of additional minimal genomes could help reveal different gene compositions and strategies that can support life, even within closely related species.

Momentary configurations of long polymers at thermal equilibrium usually deviate from spherical symmetry and can be better described, on average, by a prolate ellipsoid. The asphericity and nature of asphericity (or prolateness) that describe these momentary ellipsoidal shapes of a polymer are determined by specific expressions involving the three principal moments of inertia calculated for configurations of the polymer. Earlier theoretical studies and numerical simulations have established that as the length of the polymer increases, the average shape for the statistical ensemble of random configurations asymptotically approaches a characteristic universal shape that depends on the solvent quality. It has been established, however, that these universal shapes differ for linear, circular, and branched chains. We investigate here the effect of knotting on the shape of cyclic polymers modeled as random isosegmental polygons. We observe that random polygons forming different knot types reach asymptotic shapes that are distinct from the ensemble average shape. For the same chain length, more complex knots are, on average, more spherical than less complex knots. 1

Computer methods for sampling statistical ensembles generate chains of configurations in which subsequent members differ only slightly. Statistical errors are determined by the number of independent configurations contained in the sample. A quantitative treatment of the persistence of correlation shows how the first two moments of the autocorrelation function of a variablef along the chain are connected with the expected variance of its mean. The variance of the potential energy in the canonical ensemble is shown be to larger than that in the microcanonical one by a factor which is the ratio of the system heat capacity to that of an ideal gas. A comparison has been made of the efficiency of Metropolis Monte Carlo (MC), the usual microcanonical molecular dynamic (MDM) and a modification of molecular dynamics for canonical ensemble sampling (MDC). The analysis is focused on three aspects: the mean square displacement of a representative point in configuration space, the persistence of correlation in the potential energyV and also in a function of interest in free-energy-difference calculations. In MDM simulations of crystalline solids, it was found thatV behaves as an «oscillatory» variable and that the variance of its mean is reduced by antithetic variations of its values.

Experiments show that with increasing temperature, microemulsion systems undergo Winsor transitions. The transitions occur from Winsor I (oil droplets in water media) to Winsor II (water droplets in oil media) via Winsor III (bicontinuous phase) with an increase in the temperature. In this paper, it has been shown, for the first time, how one can study the qualitative effects of temperature, head, tail, and oil chain lengths, on these transitions. Simple cubic lattice with excluded volume and periodic boundary conditions is used to mimic the box of the simulation as a bulk of solution. The simulations have been done using the standard traditional Metropolis algorithm in the canonical ensemble (N, V, T). Configurational bias Monte Carlo and reptation moves are used with an equal probability to relax the systems. A very simple interaction model, i.e., the repulsions of water (or heads of surfactants) with oil (or tails of surfactants), is used due to the main characteristic of oil-water mixtures or amphiphilic molecule that is the hydrophobicity. The interfacial tension between oil and water (gammaow) is related to the averaged total energy of the lattice. The model shows that the Winsor III has a minimum interfacial tension (gammaow) similar to experimental results. Changing the phase structure from Winsor III to Winsor I (or Winsor II), increases the interfacial tension which is in agreement with experiments. To relate interfacial tension with the interaction parameter, the simple theory of Bragg-Williams has been used. All of the results such as the effects of oil chain length, head and tail beads number are all similar to the experimental results. Using the Davies method for calculating hydrophilic-lypophilic Balance (HLB), similar to the experimental results, Winsor III phase is formed at HLB value nearly to 10.

A lattice-based method is presented for creating 3D models of entire bacterial nucleoids integrating ultrastructural information cryoelectron tomography, genomic and proteomic data, and experimental atomic structures of biomolecules and assemblies. The method is used to generate models of the minimal genome bacterium JCVI-Syn3A, producing a series of models that test hypotheses about transcription, condensation, and overall distribution of the genome. Lattice models are also used to generate atomic models of an entire JCVI-Syn3A cell.

Water-soluble polymers such as polyethylene glycol (PEG) are critical components of industrial processes ranging from drug delivery to water purification. However, the understanding of the microscopic structure of these polymers in water and of the thermodynamics of the mixtures is limited because available experimental techniques (such as SLS and SANS) give little information about conformations and provide even the radius of gyration only in the dilute limit (<~5 wt % PEG). Computer simulations employing Monte Carlo (MC) and molecular dynamics (MD) techniques can provide an atomistic molecular structure; however, such approaches have difficulties in predicting the equilibrium polymer configurations of high-molecular-weight polymers at normal densities and in obtaining entropies and free energies directly from the MD. Here, we develop the scaled effective solvent (SES) method to predict the equilibrium ensemble of polymer configurations, which we illustrate for the case of a 20 kDa PEG (455 monomers) at a 25 wt % PEG aqueous solution (3339 waters per PEG chain). We evaluate the free energy and entropy of the members of this ensemble including explicit water, validating that it leads to average sizes (R(g)) observed experimentally and that all members of the ensemble have favorable free energies. With the SES method validated to provide well-equilibrated polymer chains in water, it should be useful for predicting ensembles of polymer chains in polymer melts and in solvents.

A Trailing Paper Trail

In this paper we shall briefly review a few Markov Chain Monte Carlo methods for simulating closed systems described by canonical ensembles. We cover both Boltzmann and non-Boltzmann sampling techniques. The Metropolis algorithm is a typical example of Boltzmann Monte Carlo method. We discuss the time-symmetry of the Markov chain generated by Metropolis like algorithms that obey detailed balance. The non-Boltzmann Monte Carlo techniques reviewed include the multicanonical and Wang-Landau sampling. We list what we consider as milestones in the historical development of Monte Carlo methods in statistical physics.

Cell behaviors are reflections of intracellular tension dynamics and play important roles in many cellular processes. In this study, temporal variations in cell geometry and cell motion through cell cycle progression were quantitatively characterized via automated cell tracking for MCF-10A non-transformed breast cells, MCF-7 non-invasive breast cancer cells, and MDA-MB-231 highly metastatic breast cancer cells. A new cell segmentation method, which combines the threshold method and our modified edge based active contour method, was applied to optimize cell boundary detection for all cells in the field-of-view. An automated cell-tracking program was implemented to conduct live cell tracking over 40 hours for the three cell lines. The cell boundary and location information was measured and aligned with cell cycle progression with constructed cell lineage trees. Cell behaviors were studied in terms of cell geometry and cell motion. For cell geometry, cell area and cell axis ratio were investigated. For cell motion, instantaneous migration speed, cell motion type, as well as cell motion range were analyzed. We applied a cell-based approach that allows us to examine and compare temporal variations of cell behavior along with cell cycle progression at a single cell level. Cell body geometry along with distribution of peripheral protrusion structures appears to be associated with cell motion features. Migration speed together with motion type and motion ranges are required to distinguish the three cell-lines examined. We found that cells dividing or overlapping vertically are unique features of cell malignancy for both MCF-7 and MDA-MB-231 cells, whereas abrupt changes in cell body geometry and cell motion during mitosis are unique to highly metastatic MDA-MB-231 cells. Taken together, our live cell tracking system serves as an invaluable tool to identify cell behaviors that are unique to malignant and/or highly metastatic breast cancer cells.

Cryo-electron tomography is an imaging technique that provides 3D images (tomograms) in situ of cells with sub-nanometer resolution. Typically, the first step in the analysis is to classify the tomogram voxels into different structures, named semantic segmentation. However, the segmentation results are sets of voxels, hindering further quantitative analysis. In this paper, we define and implement algorithms to convert the semantic segmentation of the main structures in a cellular cryo-electron tomogram (membranes, filaments, cytosolic and membrane-bound macromolecules) into specific skeletons, preserving their topological and geometrical information. Additionally, we have defined a metric for comparing segmentations in cryo-ET coming from different methods more robust than the standard DICE. We also demonstrate how this approach can be used to trace cellular features by analyzing several in situ cellular cryo-electron tomograms.