Volume Electron Microscopy Research Articles

High-fidelity Image Restoration of Large 3D Electron Microscopy Volume.

Volume electron microscopy (VEM) is an essential tool for studying biological structures. Due to the challenges of sample preparation and continuous volumetric imaging, image artifacts are almost inevitable. Such image artifacts complicate further processing both for automated computer vision methods and human experts. Unfortunately, the widely used contrast limited adaptive histogram equalization (CLAHE) can alter the essential relative contrast information about some biological structures. We developed an image-processing pipeline to remove the artifacts and enhance the images without CLAHE. We apply our method to VEM datasets of a Microwasp head. We demonstrate that our method restores the images with high fidelity while preserving the original relative contrast. This pipeline is adaptable to other VEM datasets.

NEURD offers automated proofreading and feature extraction for connectomics.

We are now in the era of millimeter-scale electron microscopy (EM) volumes collected at nanometer resolution. Dense reconstruction of cellular compartments in these EM volumes has been enabled by recent advances in Machine Learning (ML). Automated segmentation methods produce exceptionally accurate reconstructions of cells, but post-hoc proofreading is still required to generate large connectomes free of merge and split errors. The elaborate 3-D meshes of neurons in these volumes contain detailed morphological information at multiple scales, from the diameter, shape, and branching patterns of axons and dendrites, down to the fine-scale structure of dendritic spines. However, extracting these features can require substantial effort to piece together existing tools into custom workflows. Building on existing open-source software for mesh manipulation, here we present "NEURD", a software package that decomposes meshed neurons into compact and extensively-annotated graph representations. With these feature-rich graphs, we automate a variety of tasks such as state of the art automated proofreading of merge errors, cell classification, spine detection, axon-dendritic proximities, and other annotations. These features enable many downstream analyses of neural morphology and connectivity, making these massive and complex datasets more accessible to neuroscience researchers focused on a variety of scientific questions.

Volume electron microscopy reveals 3D synaptic nanoarchitecture in postmortem human prefrontal cortex.

Synaptic function is directly reflected in quantifiable ultrastructural features using electron microscopy (EM) approaches. This coupling of synaptic function and ultrastructure suggests that in vivo synaptic function can be inferred from EM analysis of ex vivo human brain tissue. To investigate this, we employed focused ion beam-scanning electron microscopy (FIB-SEM), a volume EM (VEM) approach, to generate ultrafine-resolution, three-dimensional (3D) micrographic datasets of postmortem human dorsolateral prefrontal cortex (DLPFC), a region with cytoarchitectonic characteristics distinct to human brain. Synaptic, sub-synaptic, and organelle measures were highly consistent with findings from experimental models that are free from antemortem or postmortem effects. Further, 3D neuropil reconstruction revealed a unique, ultrastructurally-complex, spiny dendritic shaft that exhibited features characteristic of heightened synaptic communication, integration, and plasticity. Altogether, our findings provide critical proof-of-concept data demonstrating that ex vivo VEM analysis is an effective approach to infer in vivo synaptic functioning in human brain.

Diffusion-based deep learning method for augmenting ultrastructural imaging and volume electron microscopy

Electron microscopy (EM) revolutionized the way to visualize cellular ultrastructure. Volume EM (vEM) has further broadened its three-dimensional nanoscale imaging capacity. However, intrinsic trade-offs between imaging speed and quality of EM restrict the attainable imaging area and volume. Isotropic imaging with vEM for large biological volumes remains unachievable. Here, we developed EMDiffuse, a suite of algorithms designed to enhance EM and vEM capabilities, leveraging the cutting-edge image generation diffusion model. EMDiffuse generates realistic predictions with high resolution ultrastructural details and exhibits robust transferability by taking only one pair of images of 3 megapixels to fine-tune in denoising and super-resolution tasks. EMDiffuse also demonstrated proficiency in the isotropic vEM reconstruction task, generating isotropic volume even in the absence of isotropic training data. We demonstrated the robustness of EMDiffuse by generating isotropic volumes from seven public datasets obtained from different vEM techniques and instruments. The generated isotropic volume enables accurate three-dimensional nanoscale ultrastructure analysis. EMDiffuse also features self-assessment functionalities on predictions’ reliability. We envision EMDiffuse to pave the way for investigations of the intricate subcellular nanoscale ultrastructure within large volumes of biological systems.

FAST-EM array tomography: a workflow for multibeam volume electron microscopy.

Elucidating the 3D nanoscale structure of tissues and cells is essential for understanding the complexity of biological processes. Electron microscopy (EM) offers the resolution needed for reliable interpretation, but the limited throughput of electron microscopes has hindered its ability to effectively image large volumes. We report a workflow for volume EM with FAST-EM, a novel multibeam scanning transmission electron microscope that speeds up acquisition by scanning the sample in parallel with 64 electron beams. FAST-EM makes use of optical detection to separate the signals of the individual beams. The acquisition and 3D reconstruction of ultrastructural data from multiple biological samples is demonstrated. The results show that the workflow is capable of producing large reconstructed volumes with high resolution and contrast to address biological research questions within feasible acquisition time frames.

Joint Learning Neuronal Skeleton and Brain Circuit Topology with Permutation Invariant Encoders for Neuron Classification

Determining the types of neurons within a nervous system plays a significant role in the analysis of brain connectomics and the investigation of neurological diseases. However, the efficiency of utilizing anatomical, physiological, or molecular characteristics of neurons is relatively low and costly. With the advancements in electron microscopy imaging and analysis techniques for brain tissue, we are able to obtain whole-brain connectome consisting neuronal high-resolution morphology and connectivity information. However, few models are built based on such data for automated neuron classification. In this paper, we propose NeuNet, a framework that combines morphological information of neurons obtained from skeleton and topological information between neurons obtained from neural circuit. Specifically, NeuNet consists of three components, namely Skeleton Encoder, Connectome Encoder, and Readout Layer. Skeleton Encoder integrates the local information of neurons in a bottom-up manner, with a one-dimensional convolution in neural skeleton's point data; Connectome Encoder uses a graph neural network to capture the topological information of neural circuit; finally, Readout Layer fuses the above two information and outputs classification results. We reprocess and release two new datasets for neuron classification task from volume electron microscopy(VEM) images of human brain cortex and Drosophila brain. Experiments on these two datasets demonstrated the effectiveness of our model with accuracies of 0.9169 and 0.9363, respectively. Code and data are available at: https://github.com/WHUminghui/NeuNet.

GAUSS-EM, guided accumulation of ultrathin serial sections with a static magnetic field for volume electron microscopy

Serial sectioning electron microscopy (EM) of millimeter-scale three-dimensional (3D) anatomical volumes requires the collection of thousands of ultrathin sections. Here, we report a high-throughput automated approach, GAUSS-EM (guided accumulation of ultrathin serial sections-EM), utilizing a static magnetic field to collect and densely pack thousands of sections onto individual silicon wafers. The method is capable of sectioning hundreds of microns of tissue per day at section thicknesses down to 35nm. Relative to other automated volume EM approaches, GAUSS-EM democratizes the ability to collect large 3D EM volumes because it is simple and inexpensive to implement. We present two exemplar EM volumes of a zebrafish eye and mouse olfactory bulb collected with the method.

Interview with FocalPlane Scientific Advisory Board member Lucy Collinson

ABSTRACT Lucy Collinson is Head of the Electron Microscopy (EM) Science Technology Platform (STP) at The Francis Crick Institute (the Crick), London, UK. Lucy completed her PhD in microbiology before switching to cell biology and microscopy for her postdoc with Colin Hopkins. Her first core facility position was as Head of EM at the Laboratory for Molecular Cell Biology (LMCB) at University College London before moving to the Cancer Research UK London Research Institute (LRI) and then the Crick. We caught up with Lucy to discuss her career, the volume EM revolution, FocalPlane and the upcoming JCS Special Issue on ‘Imaging Cell Dynamics’, which she'll be guest editing with Guillaume Jacquemet.

High-resolution volumetric imaging constrains compartmental models to explore synaptic integration and temporal processing by cochlear nucleus globular bushy cells.

Globular bushy cells (GBCs) of the cochlear nucleus play central roles in the temporal processing of sound. Despite investigation over many decades, fundamental questions remain about their dendrite structure, afferent innervation, and integration of synaptic inputs. Here, we use volume electron microscopy (EM) of the mouse cochlear nucleus to construct synaptic maps that precisely specify convergence ratios and synaptic weights for auditory nerve innervation and accurate surface areas of all postsynaptic compartments. Detailed biophysically based compartmental models can help develop hypotheses regarding how GBCs integrate inputs to yield their recorded responses to sound. We established a pipeline to export a precise reconstruction of auditory nerve axons and their endbulb terminals together with high-resolution dendrite, soma, and axon reconstructions into biophysically detailed compartmental models that could be activated by a standard cochlear transduction model. With these constraints, the models predict auditory nerve input profiles whereby all endbulbs onto a GBC are subthreshold (coincidence detection mode), or one or two inputs are suprathreshold (mixed mode). The models also predict the relative importance of dendrite geometry, soma size, and axon initial segment length in setting action potential threshold and generating heterogeneity in sound-evoked responses, and thereby propose mechanisms by which GBCs may homeostatically adjust their excitability. Volume EM also reveals new dendritic structures and dendrites that lack innervation. This framework defines a pathway from subcellular morphology to synaptic connectivity, and facilitates investigation into the roles of specific cellular features in sound encoding. We also clarify the need for new experimental measurements to provide missing cellular parameters, and predict responses to sound for further in vivo studies, thereby serving as a template for investigation of other neuron classes.

MorphoFeatures for unsupervised exploration of cell types, tissues, and organs in volume electron microscopy.

Electron microscopy (EM) provides a uniquely detailed view of cellular morphology, including organelles and fine subcellular ultrastructure. While the acquisition and (semi-)automatic segmentation of multicellular EM volumes are now becoming routine, large-scale analysis remains severely limited by the lack of generally applicable pipelines for automatic extraction of comprehensive morphological descriptors. Here, we present a novel unsupervised method for learning cellular morphology features directly from 3D EM data: a neural network delivers a representation of cells by shape and ultrastructure. Applied to the full volume of an entire three-segmented worm of the annelid Platynereis dumerilii, it yields a visually consistent grouping of cells supported by specific gene expression profiles. Integration of features across spatial neighbours can retrieve tissues and organs, revealing, for example, a detailed organisation of the animal foregut. We envision that the unbiased nature of the proposed morphological descriptors will enable rapid exploration of very different biological questions in large EM volumes, greatly increasing the impact of these invaluable, but costly resources.

Semi-Supervised Neuron Segmentation via Reinforced Consistency Learning.

Emerging deep learning-based methods have enabled great progress in automatic neuron segmentation from Electron Microscopy (EM) volumes. However, the success of existing methods is heavily reliant upon a large number of annotations that are often expensive and time-consuming to collect due to dense distributions and complex structures of neurons. If the required quantity of manual annotations for learning cannot be reached, these methods turn out to be fragile. To address this issue, in this article, we propose a two-stage, semi-supervised learning method for neuron segmentation to fully extract useful information from unlabeled data. First, we devise a proxy task to enable network pre-training by reconstructing original volumes from their perturbed counterparts. This pre-training strategy implicitly extracts meaningful information on neuron structures from unlabeled data to facilitate the next stage of learning. Second, we regularize the supervised learning process with the pixel-level prediction consistencies between unlabeled samples and their perturbed counterparts. This improves the generalizability of the learned model to adapt diverse data distributions in EM volumes, especially when the number of labels is limited. Extensive experiments on representative EM datasets demonstrate the superior performance of our reinforced consistency learning compared to supervised learning, i.e., up to 400% gain on the VOI metric with only a few available labels. This is on par with a model trained on ten times the amount of labeled data in a supervised manner. Code is available at https://github.com/weih527/SSNS-Net.

How innovations in methodology offer new prospects for volume electron microscopy

Detailed knowledge of biological structure has been key in understanding biology at several levels of organisation, from organs to cells and proteins. Volume electron microscopy (volume EM) provides high resolution 3D structural information about tissues on the nanometre scale. However, the throughput rate of conventional electron microscopes has limited the volume size and number of samples that can be imaged. Recent improvements in methodology are currently driving a revolution in volume EM, making possible the structural imaging of whole organs and small organisms. In turn, these recent developments in image acquisition have created or stressed bottlenecks in other parts of the pipeline, like sample preparation, image analysis and data management. While the progress in image analysis is stunning due to the advent of automatic segmentation and server‐based annotation tools, several challenges remain. Here we discuss recent trends in volume EM, emerging methods for increasing throughput and implications for sample preparation, image analysis and data management.

One for All, All for One: A Close Look at In-Resin Fluorescence Protocols for CLEM.

Sample preparation is the novel bottleneck for high throughput correlative light and electron microscopy (CLEM). Protocols suitable for both imaging methods must therefore balance the requirements of each technique. For fluorescence light microscopy, a structure of interest can be targeted using: 1) staining, which is often structure or tissue specific rather than protein specific, 2) dye-coupled proteins or antibodies, or 3) genetically encoded fluorescent proteins. Each of these three methods has its own advantages. For ultrastructural investigation by electron microscopy (EM) resin embedding remains a significant sample preparation approach, as it stabilizes the sample such that it withstands the vacuum conditions of the EM, and enables long-term storage. Traditionally, samples are treated with heavy metal salts prior to resin embedding, in order to increase imaging contrast for EM. This is particularly important for volume EM (vEM) techniques. Yet, commonly used contrasting agents (e.g., osmium tetroxide, uranyl acetate) tend to impair fluorescence. The discovery that fluorescence can be preserved in resin-embedded specimens after mild heavy metal staining was a game changer for CLEM. These so-called in-resin fluorescence protocols present a significant leap forward for CLEM approaches towards high precision localization of a fluorescent signal in (volume) EM data. Integrated microscopy approaches, combining LM and EM detection into a single instrument certainly require such an “all in one” sample preparation. Preserving, or adding, dedicated fluorescence prior to resin embedding requires a compromise, which often comes at the expense of EM imaging contrast and membrane visibility. Especially vEM can be strongly hampered by a lack of heavy metal contrasting. This review critically reflects upon the fundamental aspects of resin embedding with regard to 1) specimen fixation and the physics and chemistry underlying the preservation of protein structure with respect to fluorescence and antigenicity, 2) optimization of EM contrast for transmission or scanning EM, and 3) the choice of embedding resin. On this basis, various existing workflows employing in-resin fluorescence are described, highlighting their common features, discussing advantages and disadvantages of the respective approach, and finally concluding with promising future developments for in-resin CLEM.

Identifying long-range synaptic inputs using genetically encoded labels and volume electron microscopy

Enzymes that facilitate the local deposition of electron dense reaction products have been widely used as labels in electron microscopy (EM) for the identification of synaptic contacts in neural tissue. Peroxidases, in particular, can efficiently metabolize 3,3′-diaminobenzidine tetrahydrochloride hydrate (DAB) to produce precipitates with high contrast under EM following heavy metal staining, and can be genetically encoded to facilitate the labeling of specific cell-types or organelles. Nevertheless, the peroxidase/DAB method has so far not been reported to work in a multiplexed manner in combination with 3D volume EM techniques (e.g. Serial blockface electron microscopy, SBEM; Focused ion beam electron microscopy, FIBSEM) that are favored for the large-scale ultrastructural assessment of synaptic architecture However, a recently described peroxidase with enhanced enzymatic activity (dAPEX2) can efficienty deposit EM-visible DAB products in thick tissue without detergent treatment opening the possibility for the multiplex labeling of genetically defined cell-types in combination with volume EM methods. Here we demonstrate that multiplexed dAPEX2/DAB tagging is compatible with both FIBSEM and SBEM volume EM approaches and use them to map long-range genetically identified synaptic inputs from the anterior cingulate cortex to the periaqueductal gray in the mouse brain.

Deep learning based domain adaptation for mitochondria segmentation on EM volumes

Background and Objective: Accurate segmentation of electron microscopy (EM) volumes of the brain is essential to characterize neuronal structures at a cell or organelle level. While supervised deep learning methods have led to major breakthroughs in that direction during the past years, they usually require large amounts of annotated data to be trained, and perform poorly on other data acquired under similar experimental and imaging conditions. This is a problem known as domain adaptation, since models that learned from a sample distribution (or source domain) struggle to maintain their performance on samples extracted from a different distribution or target domain. In this work, we address the complex case of deep learning based domain adaptation for mitochondria segmentation across EM datasets from different tissues and species. Methods: We present three unsupervised domain adaptation strategies to improve mitochondria segmentation in the target domain based on (1) state-of-the-art style transfer between images of both domains; (2) self-supervised learning to pre-train a model using unlabeled source and target images, and then fine-tune it only with the source labels; and (3) multi-task neural network architectures trained end-to-end with both labeled and unlabeled images. Additionally, to ensure good generalization in our models, we propose a new training stopping criterion based on morphological priors obtained exclusively in the source domain. The code and its documentation are publicly available at https://github.com/danifranco/EM_domain_adaptation. Results: We carried out all possible cross-dataset experiments using three publicly available EM datasets. We evaluated our proposed strategies and those of others based on the mitochondria semantic labels predicted on the target datasets. Conclusions: The methods introduced here outperform the baseline methods and compare favorably to the state of the art. In the absence of validation labels, monitoring our proposed morphology-based metric is an intuitive and effective way to stop the training process and select in average optimal models.

Volume electron microscopy analysis reveals a new type of membrane junction required for mixing of parental genomes after fertilization

Imaging cells by light or fluorescence microscopy often provides great insights into mechanisms and dynamics of cellular processes. However, to gain information regarding the cellular structures that underlie these processes, nm‐scale resolution is often required. This type of resolution can be obtain using electron microscopy (EM). Traditionally, investigators imaged slices of chemically fixed cells or tissue in order to glimpse at relevant structures in 2‐dimensions. While powerful, this approach can potentially miss important information due to the somewhat random location of the planes analyzed, especially if a small number of images is a acquired. More recently, methodologies of volume EM have been developed that allow for sampling of a continuous volume at high resolution. Here we will describe how one such methodology, Focused Ion Beam‐ Scanning Electron Microscopy (FIB‐SEM), has been used to understand the process by which parental genomes mix after fertilization in C. elegans embryos. One‐cell embryos were collected during different stages of interphase and mitosis, fixed using high‐pressure freezing followed by freeze substitution, and stained to maximize membrane contrast. Images were acquired at 9 nm lateral (x,y) and axial (z) resolution, for volumes of approximately 10 X 10 X 25 µm3. These analyses revealed that for parental chromosomes to mix, the membranes of the two pronuclei must fused via a membrane structure that we termed “3‐way sheet junction”. The steps leading to the formation of 3‐way sheet junctions, and the more recent discovery of a novel membrane structure around centrosomes using these datasets, will be discussed.

The ultrastructural heterogeneity of lung surfactant revealed by serial section electron tomography: insights into the 3-D architecture of human tubular myelin.

Weibel's hypothetical three-dimensional (3-D) model in 1966 provided first ultrastructural details into tubular myelin (TM), a unique, complex surfactant subtype found in the hypophase of the alveolar lining layer. Although initial descriptions by electron microscopy (EM) were already published in the 1950s, a uniform morphological differentiation from other intra-alveolar surfactant subtypes is still missing and potential structure-function relationships remain enigmatic. Technical developments in volume EM methods now allow a more detailed reinvestigation, to address unanswered ultrastructural questions, we analyzed ultrathin sections of humanized SP-A1/SP-A2 coexpressing mouse and human lung samples by conventional transmission EM. We combined these two-dimensional (2-D) information with 3-D analysis of single- and dual-axis electron tomography of serial sections for high z-resolution (in a range of a few nanometers) and extended volumes of up to 1 µm total z-information, this study reveals that TM constitutes a heterogeneous surfactant organization mainly comprised of distorted parallel membrane planes with local intersections, which are distributed all over the TM substructure. These intersecting membrane planes form, among other various polygons, the well-known 2-D "lattice", respectively 3-D quadratic tubules, which in many analyzed spots of human alveoli appear to be less abundant than also observed nonconcentric 3-D lamellae, the additional application of serial section electron tomography to conventional transmission EM demonstrates a high heterogeneity of TM membrane networks, which indicates dynamic transformations between its substructures. Our method provides an ideal basis for further in and ex vivo structural analyses of surfactant under various conditions at nanometer scale.

Correlative Organelle Microscopy: Fluorescence Guided Volume Electron Microscopy of Intracellular Processes.

Intracellular processes depend on a strict spatial and temporal organization of proteins and organelles. Therefore, directly linking molecular to nanoscale ultrastructural information is crucial in understanding cellular physiology. Volume or three-dimensional (3D) correlative light and electron microscopy (volume-CLEM) holds unique potential to explore cellular physiology at high-resolution ultrastructural detail across cell volumes. However, the application of volume-CLEM is hampered by limitations in throughput and 3D correlation efficiency. In order to address these limitations, we describe a novel pipeline for volume-CLEM that provides high-precision (<100 nm) registration between 3D fluorescence microscopy (FM) and 3D electron microscopy (EM) datasets with significantly increased throughput. Using multi-modal fiducial nanoparticles that remain fluorescent in epoxy resins and a 3D confocal fluorescence microscope integrated into a Focused Ion Beam Scanning Electron Microscope (FIB.SEM), our approach uses FM to target extremely small volumes of even single organelles for imaging in volume EM and obviates the need for post-correlation of big 3D datasets. We extend our targeted volume-CLEM approach to include live-cell imaging, adding information on the motility of intracellular membranes selected for volume-CLEM. We demonstrate the power of our approach by targeted imaging of rare and transient contact sites between the endoplasmic reticulum (ER) and lysosomes within hours rather than days. Our data suggest that extensive ER-lysosome and mitochondria-lysosome interactions restrict lysosome motility, highlighting the unique capabilities of our integrated CLEM pipeline for linking molecular dynamic data to high-resolution ultrastructural detail in 3D.

An open-access volume electron microscopy atlas of whole cells and tissues.

Understanding cellular architecture is essential for understanding biology. Electron microscopy (EM) uniquely visualizes cellular structures with nanometre resolution. However, traditional methods, such as thin-section EM or EM tomography, have limitations in that they visualize only a single slice or a relatively small volume of the cell, respectively. Focused ion beam-scanning electron microscopy (FIB-SEM) has demonstrated the ability to image small volumes of cellular samples with 4-nm isotropic voxels1. Owing to advances in theprecision and stability of FIB milling, together with enhanced signal detection and faster SEM scanning, we have increased the volume that can be imaged with 4-nmvoxels by two orders of magnitude. Here we present a volume EM atlas at such resolution comprising ten three-dimensional datasets for whole cells and tissues, including cancer cells, immune cells, mouse pancreatic islets and Drosophila neural tissues. These open access data (via OpenOrganelle2) represent the foundation of a field of high-resolution whole-cell volume EM and subsequent analyses, and we invite researchers to explore this atlas andpose questions.

A Workflow for High-pressure Freezing and Freeze Substitution of the Caenorhabditis elegans Embryo for Ultrastructural Analysis by Conventional and Volume Electron Microscopy.

The free-living nematode Caenorhabditis elegans is a popular model system for studying developmental biology. Here we describe a detailed protocol to high-pressure freeze the C. elegans embryo (either ex vivo after dissection, or within the intact worm) followed by quick freeze substitution. Processed samples are suitable for ultrastructural analysis by conventional electron microscopy (EM) or newer volume EM (vEM) approaches such as Focused Ion Beam Scanning Electron Microscopy (FIB-SEM). The ultrastructure of cellular features such as the nuclear envelope, chromosomes, endoplasmic reticulum and mitochondria are well preserved after these experimental procedures and yield accurate 3D models for visualization and analysis ( Chang et al., 2020 ). This protocol was used in the 3D reconstruction of membranes and chromosomes after pronuclear meeting in the C. elegans zygote ( Rahman et al., 2020 ).