Subtomogram Averaging Research Articles

MDTOMO method for continuous conformational variability analysis in cryo electron subtomograms based on molecular dynamics simulations

Cryo electron tomography (cryo-ET) allows observing macromolecular complexes in their native environment. The common routine of subtomogram averaging (STA) allows obtaining the three-dimensional (3D) structure of abundant macromolecular complexes, and can be coupled with discrete classification to reveal conformational heterogeneity of the sample. However, the number of complexes extracted from cryo-ET data is usually small, which restricts the discrete-classification results to a small number of enough populated states and, thus, results in a largely incomplete conformational landscape. Alternative approaches are currently being investigated to explore the continuity of the conformational landscapes that in situ cryo-ET studies could provide. In this article, we present MDTOMO, a method for analyzing continuous conformational variability in cryo-ET subtomograms based on Molecular Dynamics (MD) simulations. MDTOMO allows obtaining an atomic-scale model of conformational variability and the corresponding free-energy landscape, from a given set of cryo-ET subtomograms. The article presents the performance of MDTOMO on a synthetic ABC exporter dataset and an in situ SARS-CoV-2 spike dataset. MDTOMO allows analyzing dynamic properties of molecular complexes to understand their biological functions, which could also be useful for structure-based drug discovery.

PickYOLO: Fast deep learning particle detector for annotation of cryo electron tomograms

Particle localization (picking) in digital tomograms is a laborious and time-intensive step in cryogenic electron tomography (cryoET) analysis often requiring considerable user involvement, thus becoming a bottleneck for automated cryoET subtomogram averaging (STA) pipelines. In this paper, we introduce a deep learning framework called PickYOLO to tackle this problem. PickYOLO is a super-fast, universal particle detector based on the deep-learning real-time object recognition system YOLO (You Only Look Once), and tested on single particles, filamentous structures, and membrane-embedded particles. After training with the centre coordinates of a few hundred representative particles, the network automatically detects additional particles with high yield and reliability at a rate of 0.24–3.75 s per tomogram. PickYOLO can automatically detect number of particles comparable to those manually selected by experienced microscopists. This makes PickYOLO a valuable tool to substantially reduce the time and manual effort needed to analyse cryoET data for STA, greatly aiding in high-resolution cryoET structure determination.

A method for restoring signals and revealing individual macromolecule states in cryo-ET, REST

Cryo-electron tomography (cryo-ET) is widely used to explore the 3D density of biomacromolecules. However, the heavy noise and missing wedge effect prevent directly visualizing and analyzing the 3D reconstructions. Here, we introduced REST, a deep learning strategy-based method to establish the relationship between low-quality and high-quality density and transfer the knowledge to restore signals in cryo-ET. Test results on the simulated and real cryo-ET datasets show that REST performs well in denoising and compensating the missing wedge information. The application in dynamic nucleosomes, presenting either in the form of individual particles or in the context of cryo-FIB nuclei section, indicates that REST has the capability to reveal different conformations of target macromolecules without subtomogram averaging. Moreover, REST noticeably improves the reliability of particle picking. These advantages enable REST to be a powerful tool for the straightforward interpretation of target macromolecules by visual inspection of the density and of a broad range of other applications in cryo-ET, such as segmentation, particle picking, and subtomogram averaging.

TomoTwin: generalized 3D localization of macromolecules in cryo-electron tomograms with structural data mining

Cryogenic-electron tomography enables the visualization of cellular environments in extreme detail, however, tools to analyze the full amount of information contained within these densely packed volumes are still needed. Detailed analysis of macromolecules through subtomogram averaging requires particles to first be localized within the tomogram volume, a task complicated by several factors including a low signal to noise ratio and crowding of the cellular space. Available methods for this task suffer either from being error prone or requiring manual annotation of training data. To assist in this crucial particle picking step, we present TomoTwin: an open source general picking model for cryogenic-electron tomograms based on deep metric learning. By embedding tomograms in an information-rich, high-dimensional space that separates macromolecules according to their three-dimensional structure, TomoTwin allows users to identify proteins in tomograms de novo without manually creating training data or retraining the network to locate new proteins.

In situ architecture of the ER-mitochondria encounter structure.

The endoplasmic reticulum and mitochondria are main hubs of eukaryotic membrane biogenesis that rely on lipid exchange via membrane contact sites1-3, but the underpinning mechanisms remain poorly understood. In yeast, tethering and lipid transfer between the two organelles is mediated by the endoplasmic reticulum-mitochondria encounter structure (ERMES), a four-subunit complex of unresolved stoichiometry and architecture4-6. Here we determined the molecular organization of ERMES within Saccharomyces cerevisiae cells using integrative structural biology by combining quantitative live imaging, cryo-correlative microscopy, subtomogram averaging and molecular modelling. We found that ERMES assembles into approximately 25 discrete bridge-like complexes distributed irregularly across a contact site. Each bridge consists of three synaptotagmin-like mitochondrial lipid binding protein domains oriented in a zig-zag arrangement. Our molecular model of ERMES reveals a pathway for lipids. These findings resolve the in situ supramolecular architecture of a major inter-organelle lipid transfer machinery and provide a basis for the mechanistic understanding of lipid fluxes in eukaryotic cells.

Cryo-electron tomography of intact cardiac muscle reveals myosin binding protein-C linking myosin and actin filaments.

Myosin binding protein C (MyBP-C) is an accessory protein of the thick filament in vertebrate cardiac muscle arranged over 9 stripes of intervals of 430Å in each half of the A-band in the region called the C-zone. Mutations in cardiac MyBP-C are a leading cause of hypertrophic cardiomyopathy the mechanism of which is unknown. It is a rod-shaped protein composed of 10 or 11 immunoglobulin- or fibronectin-like domains labelled C0 to C10 which binds to the thick filament via its C-terminal region. MyBP-C regulates contraction in a phosphorylation dependent fashion that may be through binding of its N-terminal domains with myosin or actin. Understanding the 3D organisation of MyBP-C in the sarcomere environment may provide new light on its function. We report here the fine structure of MyBP-C in relaxed rat cardiac muscle by cryo-electron tomography and subtomogram averaging of refrozen Tokuyasu cryosections. We find that on average MyBP-C connects via its distal end to actin across a disc perpendicular to the thick filament. The path of MyBP-C suggests that the central domains may interact with myosin heads. Surprisingly MyBP-C at Stripe 4 is different; it has weaker density than the other stripes which could result from a mainly axial or wavy path. Given that the same feature at Stripe 4 can also be found in several mammalian cardiac muscles and in some skeletal muscles, our finding may have broader implication and significance. In the D-zone, we show the first demonstration of myosin crowns arranged on a uniform 143Å repeat.

ATP‐induced conformational change of axonemal outer dynein arms revealed by cryo‐electron tomography

Axonemal outer dynein arm (ODA) motors generate force for ciliary beating. We analyzed three states of the ODA during the power stroke cycle using in situ cryo‐electron tomography, subtomogram averaging, and classification. These states of force generation depict the prepower stroke, postpower stroke, and intermediate state conformations. Comparison of these conformations to published in vitro atomic structures of cytoplasmic dynein, ODA, and the Shulin–ODA complex revealed differences in the orientation and position of the dynein head. Our analysis shows that in the absence of ATP, all dynein linkers interact with the AAA3/AAA4 domains, indicating that interactions with the adjacent microtubule doublet B‐tubule direct dynein orientation. For the prepower stroke conformation, there were changes in the tail that is anchored on the A‐tubule. We built models starting with available high‐resolution structures to generate a best‐fitting model structure for the in situ pre‐ and postpower stroke ODA conformations, thereby showing that ODA in a complex with Shulin adopts a similar conformation as the active prepower stroke ODA in the axoneme.

Structure of Anabaena flos-aquae gas vesicles revealed by cryo-ET

Gas vesicles (GVs) are gas-filled protein nanostructures employed by several species of bacteria and archaea as flotation devices to enable access to optimal light and nutrients. The unique physical properties of GVs have led to their use as genetically encodable contrast agents for ultrasound and MRI. Currently, however, the structure and assembly mechanism of GVs remain unknown. Here we employ cryoelectron tomography to reveal how the GV shell is formed by a helical filament of highly conserved GvpA subunits. This filament changes polarity at the center of the GV cylinder, a site that may act as an elongation center. Subtomogram averaging reveals a corrugated pattern of the shell arising from polymerization of GvpA into a β sheet. The accessory protein GvpC forms a helical cage around the GvpA shell, providing structural reinforcement. Together, our results help explain the remarkable mechanical properties of GVs and their ability to adopt different diameters and shapes.

Subtomogram averaging of protein structures in isolated HEK293 cell plasma membranes

The plasma membrane and its associated proteins together are key to the function and disfunction of many vital cell processes such as adhesion, sensing/signaling, endocytosis/exocytosis. To better understand their function and treat their disfunction with precision, we need high-resolution structures of membrane-associated molecular components in their native context. Cryo-electron tomography (cryoET) is a rapidly evolving structural determination technique capable of resolving proteins and protein complexes to ångström-level resolution. However, high-resolution reconstruction of membrane-associated proteins and organelles in situ has been met with obstacles including the necessity for ultrathin (<200 nm thick) sample preparation. Here, we isolate plasma membranes using a technique called “unroofing” immediately prior to sample vitrification. This produces 50-200-nm-thick vitreous samples containing the plasma membrane, the actin cortex, and the myriad of other protein complexes associated with the membrane. To verify this technique as a tool for membrane protein structure determination, we are performing a quantitative assessment of the preserved resolution using subtomogram averaging. We are currently working to utilize and modify existing subtomogram averaging tools to identify, segment, process, and refine objects of interest within reconstructed tomograms. We compare our reconstructions with known protein structures. This ongoing work is part of an effort aimed to develop a correlative pipeline for visualizing the proteome of the eukaryotic plasma membrane with efficiency and precision.

Structure of Anabaena flos-aquae gas vesicles revealed by cryo-ET.

Gas vesicles (GVs) are gas-filled protein nanostructures employed by several species of bacteria and archaea as flotation devices to enable access to optimal light and nutrients. The unique physical properties of GVs have led to their use as genetically-encodable contrast agents for ultrasound and MRI. Although bioengineering of GVs is rapidly progressing, it is hindered by our limited knowledge about their basic structure and assembly. Here we employ cryo-electron tomography to reveal how the GV shell is formed by a helical filament of highly conserved GvpA subunits. This filament changes polarity at the center of the GV cylinder—a site that may act as an elongation center. High-resolution subtomogram averaging reveals a corrugated pattern of the shell arising from polymerization of GvpA into a β-sheet. The accessory protein GvpC forms a helical cage around the GvpA shell, providing structural reinforcement. We extend our study with biochemistry and computational modeling to corroborate our model and explore its implications for GV engineering. Together, our results help explain the remarkable mechanical properties of GVs and their ability to adopt different diameters and shapes.

In situ cryo-electron tomography reveals the asymmetric architecture of mammalian sperm axonemes

The flagella of mammalian sperm display non-planar, asymmetric beating, in contrast to the planar, symmetric beating of flagella from sea urchin sperm and unicellular organisms. The molecular basis of this difference is unclear. Here, we perform in situ cryo-electron tomography of mouse and human sperm, providing the highest-resolution structural information to date. Our subtomogram averages reveal mammalian sperm-specific protein complexes within the microtubules, the radial spokes and nexin–dynein regulatory complexes. The locations and structures of these complexes suggest potential roles in enhancing the mechanical strength of mammalian sperm axonemes and regulating dynein-based axonemal bending. Intriguingly, we find that each of the nine outer microtubule doublets is decorated with a distinct combination of sperm-specific complexes. We propose that this asymmetric distribution of proteins differentially regulates the sliding of each microtubule doublet and may underlie the asymmetric beating of mammalian sperm.

Characterization of Microtubule Lattice Heterogeneity by Segmented Subtomogram Averaging.

Microtubule structure is commonly investigated using single-particle analysis (SPA) or subtomogram averaging (STA), whose main objectives are to gather high-resolution information on the αβ-tubulin heterodimer and on its interactions with neighboring molecules within the microtubule lattice. The maps derived from SPA approaches usually delineate a continuous organization of the αβ-tubulin heterodimer that alternate regularly head-to-tail along protofilaments, and that share homotypic lateral interactions between monomers (α-α, β-β), except at one unique region called the seam, made of heterotypic ones (α-β, β-α). However, this textbook description of the microtubule lattice has been challenged over the years by several studies that revealed the presence of multi-seams in microtubules assembled in vitro from purified tubulin. To analyze in deeper detail their intrinsic structural heterogeneity, we have developed a segmented subtomogram averaging (SSTA) strategy on microtubules decorated with kinesin motor-domains that bind every αβ-tubulin heterodimer. Individual protofilaments and microtubule centers are modeled, and sub-volumes are extracted at every kinesin motor domain position to obtain full subtomogram averages of the microtubules. The model is divided into shorter segments, and subtomogram averages of each segment are calculated using the main parameters of the full-length microtubule settings as a template. This approach reveals changes in the number and location of seams within individual microtubules assembled in vitro from purified tubulin and in Xenopus egg cytoplasmic extracts. Key features This protocol builds upon the method developed by J.M. Heumann to perform subtomogram averages of microtubules and extends it to divide them into shorter segments. Microtubules are decorated with kinesin motor-domains to determine the underlying organization of its constituent αβ-tubulin heterodimers. The SSTA approach allows analysis of the structural heterogeneity of individual microtubules and reveals multi-seams and changes in their number and location within their shaft. Graphical overview.

In Situ Structure Determination of Bacterial Surface Nanomachines Using Cryo-Electron Tomography.

Bacterial surface nanomachines are often refractory to structural determination in their intact form due to their extensive association with the cell envelope preventing them from being properly purified for traditional structural biology methods. Cryo-electron tomography (cryo-ET) is an emerging branch of cryo-electron microscopy that can visualize supramolecular complexes directly inside frozen-hydrated cells in 3D at nanometer resolution, therefore posing a unique capability to study the intact structures of bacterial surface nanomachines in situ and reveal their molecular association with other cellular components. Furthermore, the resolution of cryo-ET is continually improving alongside methodological advancement. Here, using the type IV pilus machine in Myxococcus xanthus as an example, we describe a step-by-step workflow for in situ structure determination including sample preparation and screening, microscope and camera tuning, tilt series acquisition, data processing and tomogram reconstruction, subtomogram averaging, and structural analysis.

Changes in seam number and location induce holes within microtubules assembled from porcine brain tubulin and in Xenopus egg cytoplasmic extracts.

Microtubules are tubes of about 25 nm in diameter that are critically involved in a variety of cellular functions, including motility, compartmentalization, and division. They are considered as pseudo-helical polymers whose constituent αβ-tubulin heterodimers share lateral homotypic interactions, except at one unique region called the seam. Here, we used a segmented sub-tomogram averaging strategy to reassess this paradigm and analyze the organization of the αβ-tubulin heterodimers in microtubules assembled from purified porcine brain tubulin in the presence of GTP and GMPCPP, and in Xenopus egg cytoplasmic extracts. We find that in almost all conditions, microtubules incorporate variable protofilament and/or tubulin subunit helical-start numbers, as well as variable numbers of seams. Strikingly, the seam number and location vary along individual microtubules, generating holes of one to a few subunits in size within their lattices. Together, our results reveal that the formation of mixed and discontinuous microtubule lattices is an intrinsic property of tubulin that requires the formation of unique lateral interactions without longitudinal ones. They further suggest that microtubule assembly is tightly regulated in a cytoplasmic environment.

Cryo-electron tomography reveals structural insights into the membrane remodeling mode of dynamin-like EHD filaments

Eps15-homology domain containing proteins (EHDs) are eukaryotic, dynamin-related ATPases involved in cellular membrane trafficking. They oligomerize on membranes into filaments that induce membrane tubulation. While EHD crystal structures in open and closed conformations were previously reported, little structural information is available for the membrane-bound oligomeric form. Consequently, mechanistic insights into the membrane remodeling mechanism have remained sparse. Here, by using cryo-electron tomography and subtomogram averaging, we determined structures of nucleotide-bound EHD4 filaments on membrane tubes of various diameters at an average resolution of 7.6 Å. Assembly of EHD4 is mediated via interfaces in the G-domain and the helical domain. The oligomerized EHD4 structure resembles the closed conformation, where the tips of the helical domains protrude into the membrane. The variation in filament geometry and tube radius suggests a spontaneous filament curvature of approximately 1/70 nm-1. Combining the available structural and functional data, we suggest a model for EHD-mediated membrane remodeling.

Structures of the eukaryotic ribosome and its translational states in situ

Ribosomes translate genetic information into primary structure. During translation, various cofactors transiently bind to the ribosome that undergoes prominent conformational and structural changes. Different translational states of ribosomes have been well characterized in vitro. However, to which extent the known translational states are representative of the native situation inside cells has thus far only been addressed in prokaryotes. Here, we apply cryo-electron tomography to cryo-FIB milled Dictyostelium discoideum cells combined with subtomogram averaging and classification. We obtain an in situ structure that is locally resolved up to 3 Angstrom, the distribution of eukaryotic ribosome translational states, and unique arrangement of rRNA expansion segments. Our work demonstrates the use of in situ structural biology techniques for identifying distinct ribosome states within the cellular environment.

ArtiaX: An electron tomography toolbox for the interactive handling of sub-tomograms in UCSF ChimeraX.

Cryo-electron tomography analysis involves the selection of macromolecular complexes to be used for subsequent sub-tomogram averaging and structure determination. Here, we describe a plugin developed for UCSF ChimeraX that allows for the display, selection, and editing of particles within tomograms. Positions and orientations of selected particles can be manually set, modified and inspected in real time, both on screen and in virtual reality, and exported to various file formats. The plugin allows for the parallel visualization of particles stored in several meta data lists, in the context of any three-dimensional image that can be opened with UCSF ChimeraX. The particles are rendered in user-defined colors or using colormaps, such that individual classes or groups of particles, cross-correlation coefficients, or other types of information can be highlighted to the user. The implemented functions are fast, reliable, and intuitive, exploring the broad range of features in UCSF ChimeraX. They allow for a fluent human-machine interaction, which enables an effective understanding of the sub-tomogram processing pipeline, even for non-specialist users.

Deep Learning-Based Segmentation of Cryo-Electron Tomograms

Cryo-electron tomography (cryo-ET) allows researchers to image cells in their native, hydrated state at the highest resolution currently possible. The technique has several limitations, however, that make analyzing the data it generates time-intensive and difficult. Hand segmenting a single tomogram can take from hours to days, but a microscope can easily generate 50 or more tomograms a day. Current deep learning segmentation programs for cryo-ET do exist, but are limited to segmenting one structure at a time. Here, multi-slice U-Net convolutional neural networks are trained and applied to automatically segment multiple structures simultaneously within cryo-tomograms. With proper preprocessing, these networks can be robustly inferred to many tomograms without the need for training individual networks for each tomogram. This workflow dramatically improves the speed with which cryo-electron tomograms can be analyzed by cutting segmentation time down to under 30 min in most cases. Further, segmentations can be used to improve the accuracy of filament tracing within a cellular context and to rapidly extract coordinates for subtomogram averaging.

Deep Learning-Based Segmentation of Cryo-Electron Tomograms.

Cryo-electron tomography (cryo-ET) allows researchers to image cells in their native, hydrated state at the highest resolution currently possible. The technique has several limitations, however, that make analyzing the data it generates time-intensive and difficult. Hand segmenting a single tomogram can take from hours to days, but a microscope can easily generate 50 or more tomograms a day. Current deep learning segmentation programs for cryo-ET do exist, but are limited to segmenting one structure at a time. Here, multi-slice U-Net convolutional neural networks are trained and applied to automatically segment multiple structures simultaneously within cryo-tomograms. With proper preprocessing, these networks can be robustly inferred to many tomograms without the need for training individual networks for each tomogram. This workflow dramatically improves the speed with which cryo-electron tomograms can be analyzed by cutting segmentation time down to under 30 min in most cases. Further, segmentations can be used to improve the accuracy of filament tracing within a cellular context and to rapidly extract coordinates for subtomogram averaging.

Multishot tomography for high-resolution in situ subtomogram averaging

Cryo-electron tomography (cryo-ET) and subtomogram averaging (STA) can resolve protein complexes at near atomic resolution, and when combined with focused ion beam (FIB) milling, macromolecules can be observed within their native context. Unlike single particle acquisition (SPA), cryo-ET can be slow, which may reduce overall project throughput. We here propose a fast, multi-position tomographic acquisition scheme based on beam-tilt corrected beam-shift imaging along the tilt axis, which yields sub-nanometer in situ STA averages.