Three-Dimensional Cell and Tissue Reconstruction by Serial Block Face SEM

National Institute Of Biomedical Imaging And Bioengineering, Bethesda

Investigators

Linked publications & trials

Abstract

Our laboratory is equipped with two types of vEM techniques: (1) a serial block-face scanning electron microscope comprising a Zeiss SIGMA-VP SEM coupled with a Gatan 3View in situ ultramicrotomy system for determining the 3D ultrastructure of cells and tissues at a lateral (x,y) resolution of 5 to 10 nanometers and a z-resolution of 25 nm to 50 nm; and (2) a Zeiss Crossbeam 550 focused ion beam scanning electron microscope for determining the 3D ultrastructure of cells and tissues at a lateral (x,y) resolution of 3 to 5 nanometers and a z-resolution of 5 nm to 10 nm. Volume electron microscopy (vEM) enables biologists to visualize nanoscale 3D ultrastructure of entire eukaryotic cells and tissues processed by heavy atom staining and plastic embedding. The vEM technique with the highest resolution is focused ion-beam scanning electron microscopy (FIB-SEM), which provides nearly isotropic (~5–10 nm) spatial resolution at fluences up to 10,000 electrons per square nanometer. However, it is still not understood how such resolution is achievable, because serial block-face (SBF) SEM, which incorporates an in-situ ultramicrotome instead of a Ga+ FIB beam, results in radiation-induced collapse of similar specimen blocks at fluences of only ~20 electrons per square nanometer.. We have shown that FIB-SEM implants a thin concentrated layer of Ga ions, which greatly reduces electron beam-damage, reduces the depth from which backscattered electrons are detected, and prevents specimen charging and collapse. Furthermore, we have shown that the z-resolution (perpendicular to block-face) in FIB-SEM is substantially higher than predicted by Monte Carlo modeling of the backscattered signal when Ga implantation is not included. We have used a scanning transmission electron microscope and electron spectrometer in our laboratory to show that FIB-SEM implants an extremely high concentration of gallium into the top 20-25 nm of the specimen block creating an ultra-strong radiation-resistant glassy phase, which prevents damage and enhances the spatial resolution perpendicular to the block face. This phenomenon has not previously been reported in the context of biological FIB-SEM, which is rapidly becoming a widely adopted technique [1]. We have developed a new correlative volume microscopy technique, cellular omics with structural Integration (COSI), to address the limitation of traditional technologies in simultaneously obtaining gene expression profiles and super-resolution cellular structural information at the single-cell level [2]. The platform comprises three core functional modules: (1) a single-cell transcriptomics and super-resolution fluorescence microscopy integration module that enables simultaneous acquisition of gene expression profiles and super-resolution fluorescence images at the single-cell level; (2) an electron microscopy and super-resolution fluorescence microscopy integration module with deep learning resolution enhancement that further gives fluorescence image high resolution features; and (3) a comprehensive analysis module that integrates transcriptomic data with enhanced super-resolution morphological data. Application of this technology to primary liver sinusoidal endothelial cells successfully achieved efficient matching and analysis of ultrastructural information and gene transcription data at the single-cell level, revealing associations between specific genes and endothelial cell fenestration formation. Through correlation analysis and multivariate statistical methods, we identified specific gene sets associated with fenestration number and average area. Validation in published non-alcoholic steatohepatitis (NASH) and diabetic mouse models demonstrated that these gene sets can effectively assess disease status and drug intervention efficacy, with fenestration number-related gene sets showing significant reduction in NASH and time-dependent changes in response to diabetes treatments. These findings not only expand our understanding of the mechanisms underlying liver and kidney endothelial cell fenestration formation but also provide novel molecular markers and potential therapeutic targets for early diagnosis and treatment evaluation of metabolic diseases. As a fundamental research tool, COSI technology fills critical gaps in existing spatial omics and cellular biology research, particularly for studying cellular structures lacking specific markers, and demonstrates significant potential for clinical applications in chronic metabolic diseases. We have used SBF-SEM to investigate the organization and activation of platelets in the formation of blood clots in a mouse model. Cardiovascular diseases are a leading cause of mortality and morbidity worldwide. Aberrant thrombosis is a common feature of systemic conditions like diabetes and obesity, and chronic inflammatory diseases like atherosclerosis, cancer, and autoimmune diseases. Upon vascular injury, usually the coagulation system, platelets, and endothelium act in an orchestrated manner to prevent bleeding by forming a clot at the site of the injury. Abnormalities in this process lead to either excessive bleeding or uncontrolled thrombosis/insufficient anti-thrombotic activity, which translates into vessel occlusion and its sequelae. Such models involve endothelial damage and subsequent clot formation at the injured site, and provide a sensitive, quantitative assay to monitor vascular damage and clot formation in response to different degrees of vascular damage. Once optimized, this standard technique can be used to study the molecular mechanisms underlying thrombosis, as well as the ultrastructural changes in platelets in a growing thrombus). The complex and highly intertwined morphology of activated platelets within thrombi poses significant challenges for segmentation. We have therefore developed a robust dual-network pipeline for cell and organelle segmentation. This multi-network approach enables the detection of fine details near the membrane while simultaneously facilitating long-range smoothing in regions distal to the membrane, drastically improving the performance of the watershed clustering algorithm compared to single-network approaches. We have further enhanced segmentation performance by collecting and averaging neural network predictions along orthogonal axes, capturing 3D correlations using only 2D neural networks [3]. In this way, we have segmented and analyzed the 3D morphology of hundreds of platelets, and have obtained quantitative measurements showing that 3D volumes are consistent with hand-segmented results. In collaboration with Dr. Manu Platt and Dr. Hannah Song Lee in NIBIB’s Section on Mechanics and Tissue Remodeling Integrating Computational & Experimental Systems (MATRICES), we are collecting three-dimensional nanoscale images of carotid artery blood vessel walls in a mouse model of sickle cell disease that Dr. Platt has developed. The aim is to quantify structural changes caused by the sickle cell mutation hemoglobin relative to animals with normal hemoglobin, as well as to study the effects of other proteins that are knocked out. 1. Yang Z, Kim J, Baenen CM, Fulton SJ, Zhang G, Chen X, Aronova MA, Leapman RD. Biological volume EM with focused Ga ion beam depends on formation of radiation-resistant Ga-rich layer at block face. bioRxiv, doi: https://doi.org/10.1101/2024.09.16.613321. 2. Wei Z, Chen J, Leapman RD. Study on liver sinusoidal endothelial cell fenestrations based on cellular omics – structure integration technology and its application in metabolic diseases. bioRxiv, doi: https://doi.org/10.1101/2025.05.16.653525. 3. Fulton SJ, Baenen CM, Storrie B, Yang Z, Leapman RD, Aronova MA. Densely Populated Cell and Organelles Segmentation with U-Net Ensembles. bioRxiv, doi: https://doi.org/10.1101/2024.11.19.623228.