Cryo-electron tomography (cryo-ET) and subtomogram averaging have emerged as powerful techniques for investigating cellular structures and their spatial organization. However, the exact localization of proteins in the crowded and noisy environment of cellular tomograms is challenging. This review provides a comprehensive overview of existing deep learning-based particle-picking procedures, which were proposed to overcome these challenges. We evaluate both annotation-based and annotation-free methods, highlighting their respective strengths, weaknesses, and ideal use cases. Furthermore, we assess these methodologies based on various criteria, such as the effort required to generate the necessary input data, inference runtime, and filament support. Additionally, we consider practical factors such as the availability of documentation and tutorials to guide researchers in selecting the most appropriate approach for their needs.

Disease-causing bacteria secrete numerous toxins to invade and subjugate their hosts. Unlike many smaller toxins, the secretion machinery of most large toxins remains enigmatic. By combining genomic editing, proteomic profiling and cryo-electron tomography of the insect pathogen Yersinia entomophaga , we demonstrate that a specialized subset of these cells produces a complex toxin cocktail, including the nearly ribosome-sized Tc toxin YenTc, which is subsequently exported by controlled cell lysis using a transcriptionally coupled, pH-dependent type 10 secretion system (T10SS). Our results dissect the Tc toxin export process by a T10SS, identifying that T10SSs operate via a previously unknown lytic mode of action and establishing them as crucial players in the size-insensitive release of cytoplasmically folded toxins. With T10SSs directly embedded in Tc toxin operons of major pathogens, we anticipate that our findings may model an important aspect of pathogenesis in bacteria with substantial impact on agriculture and healthcare.

The thick filament is a key component of sarcomeres, the basic units of striated muscle 1 . Alterations in thick filament proteins are associated with familial hypertrophic cardiomyopathy and other heart and muscle diseases 2 . Despite the central importance of the thick filament, its molecular organization remains unclear. Here we present the molecular architecture of native cardiac sarcomeres in the relaxed state, determined by cryo-electron tomography. Our reconstruction of the thick filament reveals the three-dimensional organization of myosin, titin and myosin-binding protein C (MyBP-C). The arrangement of myosin molecules is dependent on their position along the filament, suggesting specialized capacities in terms of strain susceptibility and force generation. Three pairs of titin-α and titin-β chains run axially along the filament, intertwining with myosin tails and probably orchestrating the length-dependent activation of the sarcomere. Notably, whereas the three titin-α chains run along the entire length of the thick filament, titin-β chains do not. The structure also demonstrates that MyBP-C bridges thin and thick filaments, with its carboxy-terminal region binding to the myosin tails and directly stabilizing the OFF state of the myosin heads in an unforeseen manner. These results provide a foundation for future research investigating muscle disorders involving sarcomeric components.

The thick filament is a key component of sarcomeres, the basic force-generating and load-bearing unit of striated muscle. Mutations in thick filament proteins are associated with familial hypertrophic cardiomyopathy and other heart and muscle diseases. Despite this central importance for sarcomere force generation, it remains unclear how thick filaments are structurally organized and how its components interact with each other and with thin filaments to enable highly regulated muscle contraction. Here, we present the molecular architecture of native cardiac sarcomeres in the relaxed state, determined by electron cryo-tomography. Our reconstruction of the thick filament reveals the three-dimensional organization of myosin heads and tails, myosin-binding protein C (MyBP-C) and titin, elucidating the structural basis for their interaction during muscle contraction. The arrangement of myosin heads is variable depending on their position along the filament, suggesting that they have different capacities in terms of strain susceptibility and activation. Myosin tails exhibit a distinct arrangement and pattern of interactions. These are likely orchestrated by three alpha and three beta titin chains that are arranged like a spring, suggesting the existence of specialized roles of thick filament segments in length-dependent activation and contraction. Surprisingly, while the three titin alpha chains run along the entire length of the thick filament, titin beta does not. The structure also demonstrates that the C-terminal region of MyBP-C binds myosin tails and unexpectedly also directly interacts with the myosin heads, suggesting a previously undescribed direct role in the preservation of the myosin OFF state. Furthermore, we visualize how MyBP-C forms links between thin and thick filaments. These findings establish a robust groundwork for forthcoming research endeavors aiming to explore muscle disorders that involve sarcomeric structural components.

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.

Disease-causing bacteria use a variety of secreted toxins to invade and subjugate their hosts. While the machinery responsible for secretion of many smaller toxins has already been established, it remains enigmatic for larger ones such as Tc toxins from human and insect pathogens, which approach the size of a prokaryotic ribosome. In the present study, we combine targeted genomic editing, proteomic profiling and cryo-electron tomography of the insect pathogen Yersinia entomophaga to reveal that a specialized subset of bacterial cells produces the Tc toxin YenTc as part of a complex toxin cocktail released into the environment by controlled cell lysis using a transcriptionally-coupled, pH-dependent type 10 secretion system (T10SS). Our results dissect the process of Tc toxin export by a T10SS in hitherto unprecedented detail, identifying that T10SSs operate via a previously unknown lytic mode of action, and establishing them as crucial players in the size-insensitive release of cytoplasmically folded toxins. With T10SSs directly embedded in Tc toxin operons of major human pathogens such as Yersinia pestis and Salmonella enterica , we anticipate our findings to model an important aspect of pathogenesis in bacteria with a significant impact on global human health.

In skeletal muscle, nebulin stabilizes and regulates the length of thin filaments, but the underlying mechanism remains nebulous. In this work, we used cryo–electron tomography and subtomogram averaging to reveal structures of native nebulin bound to thin filaments within intact sarcomeres. This in situ reconstruction provided high-resolution details of the interaction between nebulin and actin, demonstrating the stabilizing role of nebulin. Myosin bound to the thin filaments exhibited different conformations of the neck domain, highlighting its inherent structural variability in muscle. Unexpectedly, nebulin did not interact with myosin or tropomyosin, but it did interact with a troponin T linker through two potential binding motifs on nebulin, explaining its regulatory role. Our structures support the role of nebulin as a thin filament “molecular ruler” and provide a molecular basis for studying nemaline myopathies.

This is a demo dataset for TomoTwin The ZIP file contains all necessary files for this demo. A readme.txt give instructions how to run it. More information can be found in the official documentation of TomoTwin. TomoTwin is open source and the code can be found on GitHub.

A widely used approach to analyze single particles in electron microscopy data is 2D classification. This process is very computationally expensive, especially when large data sets are analyzed. In this paper we present GPU ISAC, a newly developed, GPU-accelerated version of the established Iterative Stable Alignment and Clustering (ISAC) algorithm for 2D images and generating class averages. While the previously existing implementation of ISAC relied on a computer cluster, GPU ISAC enables users to produce high quality 2D class averages from large-scale data sets on a single desktop machine equipped with affordable, consumer-grade GPUs such as Nvidia GeForce GTX 1080 TI cards. With only two such cards GPU ISAC matches the performance of twelve high end cluster nodes and, using high performance GPUs, is able to produce class averages from a million particles in between six to thirteen hours, depending on data set quality and box size. We also show GPU ISAC to scale linearly in all input dimensions, and thereby capable of scaling well with the increasing data load demand of future data sets. Further user experience improvements integrate GPU ISAC seamlessly into the existing SPHIRE GUI, as well as the TranSPHIRE on-the-fly processing pipeline. It is open source and can be downloaded at https://gitlab.gwdg.de/mpi-dortmund/sphire/cuISAC/.

Automated single-particle picking in electron cryo-microscopy data has seen important advances in the past couple of years and now enables computer-assisted particle selection even for challenging datasets. These advances have implications for streamlined and automated image processing, with potential benefits for improving the resolution of resulting structures.

Nebulin is a major structural protein of skeletal sarcomeres and is essential for proper assembly and contraction of skeletal muscle 1 . It stabilises and regulates the length of thin filaments, 2 but the structural mechanism remains nebulous. Using electron cryotomography and sub-tomogram averaging, we present the first structure of native nebulin bound to thin filaments within the A-band and I-band of intact sarcomeres. This in-situ reconstruction reveals unprecedented detail of interaction at pseudo-atomic resolution between nebulin and actin, providing the basis for understanding the structural and regulatory roles of nebulin. The position of nebulin on the thin filament indicates that there is no contact to tropomyosin or myosin, but an unexpected interaction with a troponin-T linker, possibly through two binding motifs on nebulin. In addition, our structure of myosin bound to the thin filaments reveals different conformations of the neck domain, both within the same sarcomere and when compared to purified structures, highlighting an inherent structural variability in muscle. We provide a complete description of cross-bridge formation on fully native, nebulin-containing thin filaments at near-atomic scale. Our structures establish the molecular basis for the role of nebulin as a thin filament “molecular ruler” and the impact of nemaline myopathies mutations that will aid future development of therapies.

Sarcomeres are force-generating and load-bearing devices of muscles. A precise molecular picture of how sarcomeres are built underpins understanding their role in health and disease. Here, we determine the molecular architecture of native vertebrate skeletal sarcomeres by electron cryo-tomography. Our reconstruction reveals molecular details of the three-dimensional organization and interaction of actin and myosin in the A-band, I-band, and Z-disc and demonstrates that α-actinin cross-links antiparallel actin filaments by forming doublets with 6-nm spacing. Structures of myosin, tropomyosin, and actin at ~10 Å further reveal two conformations of the "double-head" myosin, where the flexible orientation of the lever arm and light chains enable myosin not only to interact with the same actin filament, but also to split between two actin filaments. Our results provide unexpected insights into the fundamental organization of vertebrate skeletal muscle and serve as a strong foundation for future investigations of muscle diseases.

Structure determination of filamentous molecular complexes involves the selection of filaments from cryo-EM micrographs. The automatic selection of helical specimens is particularly difficult, and thus many challenging samples with issues such as contamination or aggregation are still manually picked. Here, two approaches for selecting filamentous complexes are presented: one uses a trained deep neural network to identify the filaments and is integrated in SPHIRE-crYOLO , while the other, called SPHIRE-STRIPER , is based on a classical line-detection approach. The advantage of the crYOLO -based procedure is that it performs accurately on very challenging data sets and selects filaments with high accuracy. Although STRIPER is less precise, the user benefits from less intervention, since in contrast to crYOLO , STRIPER does not require training. The performance of both procedures on Tobacco mosaic virus and filamentous F-actin data sets is described to demonstrate the robustness of each method.

Particle selection is a crucial step when processing electron cryo microscopy data. Several automated particle picking procedures were developed in the past but most struggle with non-ideal data sets. In our recent Communications Biology article, we presented crYOLO, a deep learning based particle picking program. It enables fast, automated particle picking at human levels of accuracy with low effort. A general model allows the use of crYOLO for selecting particles in previously unseen data sets without further training. Here we describe how crYOLO has evolved since its initial release. We have introduced filament picking, a new denoising technique, and a new graphical user interface. Moreover, we outline its usage in automated processing pipelines, which is an important advancement on the horizon of the field.

Sarcomeres are the force-generating and load-bearing devices of muscles. A precise molecular understanding of how the entire sarcomere is built is required to understand its role in health, disease and ageing. Here, we determine the in situ molecular architecture of vertebrate skeletal sarcomeres through electron cryo-tomography of cryo-focused ion beam-milled native myofibrils. The reconstructions reveal the three-dimensional organisation and interaction of actin and myosin filaments in the A-band, I-band and Z-disc and demonstrate how α -actinin cross-links antiparallel actin filaments to form a mesh-like structure in the Z-disc at an unprecedented level of molecular detail. A prominent feature is a so-far undescribed doublet of α-actinin cross-links with ∼ 6 nm spacing. Sub-volume averaging shows the interaction between myosin, tropomyosin and actin in molecular detail at ∼ 10 Å resolution and reveals two coexisting conformations of actin-bound heads. The flexible orientation of the lever arm and the essential and regulatory light chains allow the two heads of the “double-headed” myosin not only to interact with the same actin filament but also to split between two actin filaments. Our results provide new insights into the conformational plasticity and fundamental organisation of vertebrate skeletal muscle and serve as a strong foundation for future in situ investigations of muscle diseases. ### Competing Interest Statement The authors have declared no competing interest.

Single particle cryo-EM requires full automation to allow high-throughput structure determination. Although software packages exist where parts of the cryo-EM pipeline are automated, a complete solution that offers reliable on-the-fly processing, resulting in high-resolution structures, does not exist. Here we present TranSPHIRE: A software package for fully-automated processing of cryo-EM datasets during data acquisition. TranSPHIRE transfers data from the microscope, automatically applies the common pre-processing steps, picks particles, performs 2D clustering, and 3D refinement parallel to image recording. Importantly, TranSPHIRE introduces a machine learning-based feedback loop to re-train its picking model to adapt to any given data set live during processing. This elegant approach enables TranSPHIRE to process data more effectively, producing high-quality particle stacks. TranSPHIRE collects and displays all metrics and microscope settings to allow users to quickly evaluate data during acquisition. TranSPHIRE can run on a single work station and also includes the automated processing of filaments.

Selecting particles from digital micrographs is an essential step in single-particle electron cryomicroscopy (cryo-EM). As manual selection of complete datasets—typically comprising thousands of particles—is a tedious and time-consuming process, numerous automatic particle pickers have been developed. However, non-ideal datasets pose a challenge to particle picking. Here we present the particle picking software crYOLO which is based on the deep-learning object detection system You Only Look Once (YOLO). After training the network with 200–2500 particles per dataset it automatically recognizes particles with high recall and precision while reaching a speed of up to five micrographs per second. Further, we present a general crYOLO network able to pick from previously unseen datasets, allowing for completely automated on-the-fly cryo-EM data preprocessing during data acquisition. crYOLO is available as a standalone program under http://sphire.mpg.de/ and is distributed as part of the image processing workflow in SPHIRE.

The era of research on (engineered) nanomaterials (NM) has been thriving for more than a decade and has delivered many beneficial applications, but also raised concerns about potential negative impacts on human health and ecosystems. The precautionary principle, hence, calls for a regulation of certain types of NM, which in consequence requires their unambiguous identification. Most of the currently available definitions of NM rely on an evaluation of the size of the constituent particles and therefore methods have to be developed to measure this parameter. Transmission electron microscopy (TEM) is one of the most promising techniques, as its resolving power well covers the nanosize range. However, limited automation of TEM analyses and possible user bias are major drawbacks of the technique and currently put severe constraints on its broader applications in nanometrology. Therefore, the goal of this study was to develop a software code, referred to as AutoEM, to automatically acquire TEM images, measure particle sizes and extract the respective particle size distributions (PSD) of (nano)-materials. The AutoEM software also incorporates methods for elemental analyses of individual particles using electron energy loss and energy dispersive X-ray spectroscopy (EELS/EDX) allowing the extraction of element-specific PSDs. Additionally automated acquisition of energy-filtered images (EFTEM) is implemented in the AutoEM software, which can be used, e.g., to derive thickness maps and, thus, to evaluate the thickness of individual (plate-like) particles.

An approach for the size measurement of particulate (nano)materials by transmission electron microscopy was evaluated. The approach combines standard operating procedures for specimen preparation, imaging, and image analysis, and it was evaluated on a series of certified reference materials and representative test materials with varying physical properties, including particle size, shape, and agglomeration state. The measurement of the median value of the minimal external particle diameter distribution was intra-laboratory validated. The validation study included an assessment of the limit of detection, working range, selectivity, precision, trueness, robustness, and ruggedness. An uncertainty that was associated to intermediate precision in the range of 1–7% and an expanded measurement uncertainty in the range of 7–20% were obtained, depending on the material and image analysis mode. No bias was observed when assessing the trueness of the approach on the certified reference materials ERM-FD100 and ERM-FD304. The image analysis method was validated in an inter-laboratory study by 19 laboratories, which resulted in a within-laboratory precision in the range of 2–8% and a between-laboratory precision of between 2% and 14%. The automation and standardization of the proposed approach significantly improves labour and cost efficiency for the accurate and precise size measurement of the particulate materials. The approach is shown to be implementable in many other electron microscopy laboratories.

The function of actin is coupled to the nucleotide bound to its active site. ATP hydrolysis is activated during polymerization; a delay between hydrolysis and inorganic phosphate (P i ) release results in a gradient of ATP, ADP–P i and ADP along actin filaments (F-actin). Actin-binding proteins can recognize F-actin’s nucleotide state, using it as a local ‘age’ tag. The underlying mechanism is complex and poorly understood. Here we report six high-resolution cryo-EM structures of F-actin from rabbit skeletal muscle in different nucleotide states. The structures reveal that actin polymerization repositions the proposed catalytic base, His161, closer to the γ-phosphate. Nucleotide hydrolysis and P i release modulate the conformational ensemble at the periphery of the filament, thus resulting in open and closed states, which can be sensed by coronin-1B. The drug-like toxin jasplakinolide locks F-actin in an open state. Our results demonstrate in detail how ATP hydrolysis links to F-actin’s conformational dynamics and protein interaction.