您当前的位置: 首页 > 网页快照
Dissecting subcellular architecture of whole cells - Nature Methods
Technological advances in volume electron microscopy allow imaging and analysis of whole cells at record resolution. You have full access to this article via your institution. Download PDF Download PDF Three-dimensional (3D) distribution of membrane-bound organelles and macromolecular assemblies within a cell account for its morphology and function. Our understanding of these details is limited by the resolution and speed afforded by imaging technologies. Transmission electron microscopy (TEM) and electron tomography (ET) visualize only a single slice or a relatively small volume of the cell, respectively, and although there has been some success has in stitching together these sections digitally, structural details are still compromised. Another challenge in analyzing subcellular organization is the error-prone process of manual annotations. Automated cell segmentation approaches are evolving quickly, but there is limited availability of ‘ground truth’–annotated training datasets for a wide variety of cells and tissues. 3D renderings of manually annotated training blocks and the classification of different organelles within the volume in an interphase HeLa cell. Image adapted from Heinrich et al. (2021), Springer Nature. Researchers at the Howard Hughes Medical Institute’s Janelia Research Campus have joined forces under the Cell Organelle Segmentation in Electron Microscopy (COSEM) Project to tackle the imaging, cellular segmentation and associated large-scale data analysis problems. A team of researchers led by C. Shan Xu and Harald F. Hess previously developed enhanced focused ion beam–scanning electron microscopy (FIB-SEM), which made it feasible to extend the advantages of FIB-SEM, such as excellent z -axis resolution, isotropic resolution and easy 3D data acquisition, to larger volumes by allowing long-term imaging for weeks, months or even years. Recently, thanks to advances in the precision and stability of FIB milling, as well as enhanced signal detection and faster SEM scanning, they have increased the volume that can be imaged with 4-nm voxels by two orders of magnitude. In one study, the researchers report a volume EM atlas with 4-nm voxels for ten biological specimens, including cancer cells, immune cells, mouse pancreatic islets and Drosophila melanogaster neural tissues. Each specimen required weeks of uninterrupted imaging, and comparison between 4-nm and 8-nm voxel sampling shows features that are visible only at the higher resolution, such as finer details inside the nucleus, Golgi cisternae and the close contacts of endoplasmic reticulum and mitochondria. “This new existence of whole-cell 3D EM data enables a comprehensive whole-cell analysis of anything that can be identified, counted or quantified. Biologist can now pose new kinds of questions and create characterizations from analysis of this data which cannot be answered from traditional EM section data,” says Hess. They make their data available via the OpenOrganelle repository. Stephan Saalfeld and Aubrey V. Weigel, along with the COSEM Project team, developed the OpenOrganelle. “We believe that a comprehensive ground-truth repository for cellular organelles and macromolecular structures and structures in as many cell types, tissues and preparation method variations, in combination with innovative methods to train classifiers on these data, will eventually lead to a push-button solution to analyze whole-cell FIB-SEM volumes fully automatically,” says Weigel. The researchers have manually annotated up to 35 cellular organelles and sub-organelle structures in 4-nm and 8-nm-per-voxel FIB-SEM volumes, spanning a variety of different cell types, and have trained deep-learning architectures to simultaneously segment these structures at different input resolutions. “We demonstrated that training on samples from different cell types and preparation protocols reliably improves performance on all samples, even when compared to training on data exclusively from the target sample,” explains Weigel. OpenOrganelle makes the FIB-SEM volumes, training data, reconstructions, open-source code and models available to the public, making important contributions towards the larger goals of the COSEM Project. “Our experiments show that this work is promising but not finished. While our networks perform excellently on datasets and structures that are well represented in the training data, performance degrades for rare structures and in samples whose properties are not well represented,” says Saalfeld. The researchers thus plan to continue expanding the list of imaged as well as annotated samples and tissues available on OpenOrganelle to further catalyze the development of better computational models. They are also experimenting with more targeted training objectives, new deep-learning architectures and different modes of annotation, as well as ways to optimally include unsupervised pre-training on raw data and active learning strategies that enable sparse annotation. “As the sample list and number of characterization tools grows, the open question will be if this can generate biological insight,” muses Hess. Research papers . Xu, C. S. et al. An open-access volume electron microscopy atlas of whole cells and tissues. Nature 599 , 147–151 (2021). CAS ? Article ? Google Scholar ? Heinrich, L. et al. Whole-cell organelle segmentation in volume electron microscopy. Nature 599 , 141–146 (2021). CAS ? Article ? Google Scholar ? Download references Author information . Affiliations . Nature Methods Arunima Singh Authors Arunima Singh View author publications You can also search for this author in PubMed ? Google Scholar Corresponding author . Correspondence to Arunima Singh . Rights and permissions . Reprints and Permissions About this article . Cite this article . Singh, A. Dissecting subcellular architecture of whole cells. Nat Methods (2021). https://doi.org/10.1038/s41592-021-01350-w Download citation Published : 03 December 2021 DOI : https://doi.org/10.1038/s41592-021-01350-w Share this article . Anyone you share the following link with will be able to read this content: Get shareable link Sorry, a shareable link is not currently available for this article. Copy to clipboard Provided by the Springer Nature SharedIt content-sharing initiative .
From:
监测目标主题     
(1)  
系统抽取主题     
(1)  
(1)  
(1)  
(1)  
(1)  
(1)  
(1)  
(1)  
(1)  
(1)  
(1)  
(1)  
(1)  
(1)  
(1)  
(1)  
(1)  
(1)  
(1)  
(1)  
(1)  
(1)  
(1)  
(1)  
(1)  
(1)  
(1)  
(1)  
(1)  
(1)  
(1)  
(1)  
(1)  
(1)  
(1)  
(1)  
(1)  
(1)  
(1)  
(1)  
(1)  
(1)  
(1)  
(1)  
(1)  
(1)