EM Images Research Articles

Do Perineuronal Nets Stabilize the Engram of a Synaptic Circuit?

Perineuronal nets (PNNs), a specialized form of extra cellular matrix (ECM), surround numerous neurons in the CNS and allow synaptic connectivity through holes in its structure. We hypothesize that PNNs serve as gatekeepers that guard and protect synaptic territory and thus may stabilize an engram circuit. We present high-resolution and 3D EM images of PNN-engulfed neurons in mice brains, showing that synapses occupy the PNN holes and that invasion of other cellular components is rare. PNN constituents in mice brains are long-lived and can be eroded faster in an enriched environment, while synaptic proteins have a high turnover rate. Preventing PNN erosion by using pharmacological inhibition of PNN-modifying proteases or matrix metalloproteases 9 (MMP9) knockout mice allowed normal fear memory acquisition but diminished long-term memory stabilization, supporting the above hypothesis.

Artificial intelligence approaches to the volumetric quantification of glycogen granules in EM images of human tissue.

Skeletal muscle, the major processor of dietary glucose, stores it in myriad glycogen granules. Their numbers vary with cellular location and physiological and pathophysiological states. AI models were developed to derive granular glycogen content from electron-microscopic images of human muscle. Two UNet-type semantic segmentation models were built: "Locations" classified pixels as belonging to different regions in the cell; "Granules" identified pixels within granules. From their joint output, a pixel fraction pf was calculated for images from patients positive (MHS) or negative (MHN) to a test for malignant hyperthermia susceptibility. pf was used to derive vf, the volume fraction occupied by granules. The relationship vf (pf) was derived from a simulation of volumes ("baskets") containing virtual granules at realistic concentrations. The simulated granules had diameters matching the real ones, which were measured by adapting a utility devised for calcium sparks. Applying this relationship to the pf measured in images, vf was calculated for every region and patient, and from them a glycogen concentration. The intermyofibrillar spaces and the sarcomeric I band had the highest granular content. The measured glycogen concentration was low enough to allow for a substantial presence of non-granular glycogen. The MHS samples had an approximately threefold lower concentration (significant in a hierarchical test), consistent with earlier evidence of diminished glucose processing in MHS. The AI models and the approach to infer three-dimensional magnitudes from two-dimensional images should be adaptable to other tasks on a variety of images from patients and animal models and different disease conditions.

An aldehyde-crosslinking mitochondrial probe for STED imaging in fixed cells

Fluorescence labeling of chemically fixed specimens, especially immunolabeling, plays a vital role in super-resolution imaging as it offers a convenient way to visualize cellular structures like mitochondria or the distribution of biomolecules with high detail. Despite the development of various distinct probes that enable super-resolved stimulated emission depletion (STED) imaging of mitochondria in live cells, most of these membrane-potential-dependent fluorophores cannot be retained well in mitochondria after chemical fixation. This lack of suitable mitochondrial probes has limited STED imaging of mitochondria to live cell samples. In this study, we introduce a mitochondria-specific probe, PK Mito Orange FX (PKMO FX), which features a fixation-driven cross-linking motif and accumulates in the mitochondrial inner membrane. It exhibits high fluorescence retention after chemical fixation and efficient depletion at 775 nm, enabling nanoscopic imaging both before and after aldehyde fixation. We demonstrate the compatibility of this probe with conventional immunolabeling and other strategies commonly used for fluorescence labeling of fixed samples. Moreover, we show that PKMO FX facilitates correlative super-resolution light and electron microscopy, enabling the correlation of multicolor fluorescence images and transmission EM images via the characteristic mitochondrial pattern. Our probe further expands the mitochondrial toolkit for multimodal microscopy at nanometer resolutions.

Mimicking kidney flow shear efficiently induces aggregation of LECT2, a protein involved in renal amyloidosis

Aggregation of leukocyte cell-derived chemotaxin 2 (LECT2) causes ALECT2, a systemic amyloidosis that affects the kidney and liver. Previous studies established that LECT2 fibrillogenesis is accelerated by loss of its bound zinc ion and stirring/shaking. These forms of agitation create heterogeneous shear conditions, including air-liquid interfaces that denature proteins, that are not present in the body. Here, we determined the extent to which a more physiological form of mechanical stress—shear generated by fluid flow through a network of narrow channels—drives LECT2 fibrillogenesis. To mimic blood flow through the kidney, where LECT2 and other proteins form amyloid deposits, we developed a microfluidic device consisting of progressively branched channels narrowing from 5 mm to 20 μm in width. Shear was particularly pronounced at the branch points and in the smallest capillaries. Aggregation was induced within 24 h by shear levels that were in the physiological range and well below those required to unfold globular proteins such as LECT2. EM images suggested the resulting fibril ultrastructures were different when generated by laminar flow shear vs. shaking/stirring. Importantly, results from the microfluidic device showed the first evidence that the I40V mutation accelerated fibril formation and increased both size and density of the aggregates. These findings suggest that kidney-like flow shear, in combination with zinc loss, acts in combination with the I40V mutation to trigger LECT2 amyloidogenesis. These microfluidic devices may be of general use for uncovering mechanisms by which blood flow induces misfolding and amyloidosis of circulating proteins.

Electron Microscopy Images as Set of Fragments for Mitochondrial Segmentation

Automatic mitochondrial segmentation enjoys great popularity with the development of deep learning. However, the coarse prediction raised by the presence of regular 3D grids in previous methods regardless of 3D CNN or the vision transformers suggest a possibly sub-optimal feature arrangement. To mitigate this limitation, we attempt to interpret the 3D EM image stacks as a set of interrelated 3D fragments for a better solution. However, it is non-trivial to model the 3D fragments without introducing excessive computational overhead. In this paper, we design a coherent fragment vision transformer (FragViT) combined with affinity learning to manipulate features on 3D fragments yet explore mutual relationships to model fragment-wise context, enjoying locality prior without sacrificing global reception. The proposed FragViT includes a fragment encoder and a hierarchical fragment aggregation module. The fragment encoder is equipped with affinity heads to transform the tokens into fragments with homogeneous semantics, and the multi-layer self-attention is used to explicitly learn inter-fragment relations with long-range dependencies. The hierarchical fragment aggregation module is responsible for hierarchically aggregating fragment-wise prediction back to the final voxel-wise prediction in a progressive manner. Extensive experimental results on the challenging MitoEM, Lucchi, and AC3/AC4 benchmarks demonstrate the effectiveness of the proposed method.

Deep convolutional neural networks to restore single-shot electron microscopy images

Advanced electron microscopy techniques, including scanning electron microscopes (SEM), scanning transmission electron microscopes (STEM), and transmission electron microscopes (TEM), have revolutionized imaging capabilities. However, achieving high-quality experimental images remains a challenge due to various distortions stemming from the instrumentation and external factors. These distortions, introduced at different stages of imaging, hinder the extraction of reliable quantitative insights. In this paper, we will discuss the main sources of distortion in TEM and S(T)EM images, develop models to describe them, and propose a method to correct these distortions using a convolutional neural network. We validate the effectiveness of our method on a range of simulated and experimental images, demonstrating its ability to significantly enhance the signal-to-noise ratio. This improvement leads to a more reliable extraction of quantitative structural information from the images. In summary, our findings offer a robust framework to enhance the quality of electron microscopy images, which in turn supports progress in structural analysis and quantification in materials science and biology.

DeepCLEM: automated registration for correlative light and electron microscopy using deep learning

In correlative light and electron microscopy (CLEM), the fluorescent images must be registered to the EM images with high precision. Due to the different contrast of EM and fluorescence images, automated correlation-based alignment is not directly possible, and registration is often done by hand using a fluorescent stain, or semi-automatically with fiducial markers. We introduce “DeepCLEM”, a fully automated CLEM registration workflow. A convolutional neural network predicts the fluorescent signal from the EM images, which is then automatically registered to the experimentally measured chromatin signal from the sample using correlation-based alignment. The complete workflow is available as a Fiji plugin and could in principle be adapted for other imaging modalities as well as for 3D stacks.

Computational models of amorphous ice for accurate simulation of cryo-EM images of biological samples

Simulations of cryo-electron microscopy (cryo-EM) images of biological samples can be used to produce test datasets to support the development of instrumentation, methods, and software, as well as to assess data acquisition and analysis strategies. To be useful, these simulations need to be based on physically realistic models which include large volumes of amorphous ice. The gold standard model for EM image simulation is a physical atom-based ice model produced using molecular dynamics simulations. Although practical for small sample volumes; for simulation of cryo-EM data from large sample volumes, this can be too computationally expensive. We have evaluated a Gaussian Random Field (GRF) ice model which is shown to be more computationally efficient for large sample volumes. The simulated EM images are compared with the gold standard atom-based ice model approach and shown to be directly comparable. Comparison with experimentally acquired data shows the Gaussian random field ice model produces realistic simulations. The software required has been implemented in the Parakeet software package and the underlying atomic models are available online for use by the wider community.

#4694 AUTOMATIC SEGMENTATION OF GLOMERULAR SUBSTRUCTURES BY DEEP LEARNING

Abstract Background and Aims Electron microscopy (EM) complements light microscopy (LM) evaluation of the kidney biopsy. Foot process effacement, as assessed by EM, helps in diagnosing podocytopathies. However, human interpretation of EM images is time-intensive and often subjective. In this pilot, we investigate how deep learning techniques can help in adequate segmentation of the glomerular basement membranes (GBM) and podocytes in EM images. The ultimate goal would be to design an automatic tool for reliable and fast assessment of foot process effacement. Method Podocytes and glomerular basement membranes (GBM) of 10 patients, 5 with podocyte disease and 5 without glomerular changes (where biopsy showed tubulointerstitial pathology on LM) were annotated with a combination of manual annotation and thresholding for providing ground truth. After data preprocessing including splitting, flipping, rotating and tiling, a modified U-net architecture was applied (‘baseline model’). This baseline model was compared to a combination of U-net and a self-supervised contrastive learning approach with pretraining on 100 additional images (SimCLR framework, ‘fine-tuned model’). Segmentation performance was measured by IoU score. Results Segmentation of the glomerular basement membrane was best achieved by the baseline model and resulted in an IoU score of 0.711±0.089 compared to a IoU score of 0.675 ±0.093 in the fine-tuned model Segmentation of the podocytes was most successful in the fine-tuned model, with a IoU score of 0.609±0.118 compared to a IoU score of 0.591 ±0.118 in the baseline model Conclusion This study pioneers in segmenting glomerular substructures on EM images by means of a modified U-net architecture. The next step is training and validation in larger datasets. Data annotation remains a challenge. Inclusion of more images is expected to greatly improve the performance of the model. Guideline 1500-3600 characters

Guided-deconvolution for correlative light and electron microscopy.

Correlative light and electron microscopy is a powerful tool to study the internal structure of cells. It combines the mutual benefit of correlating light (LM) and electron (EM) microscopy information. The EM images only contain contrast information. Therefore, some of the detailed structures cannot be specified from these images alone, especially when different cell organelle are contacted. However, the classical approach of overlaying LM onto EM images to assign functional to structural information is hampered by the large discrepancy in structural detail visible in the LM images. This paper aims at investigating an optimized approach which we call EM-guided deconvolution. This applies to living cells structures before fixation as well as previously fixed sample. It attempts to automatically assign fluorescence-labeled structures to structural details visible in the EM image to bridge the gaps in both resolution and specificity between the two imaging modes. We tested our approach on simulations, correlative data of multi-color beads and previously published data of biological samples.

Sol-gel synthesis of nanostructured cobalt oxide in four different ionic liquids

In this paper we report on the sol-gel synthesis of cobalt oxide in four different ionic liquids with two different cations and anions. The employed ionic liquids are; 1-ethyl-3-methylimidazolium trifluoromethylsulfonate ([EMIm]TfO), 1-butyl-1-methylpyrrolidinium trifluoromethylsulfonate ([Py1,4]TfO), 1-ethyl-3-methylimidazolium bis(trifluoromethylsulfonyl)amide ([EMIm]TFSA), and 1-butyl-1-methylpyrrolidiniumbis(trifluoromethylsulfonyl)amide ([Py1,4]TFSA). The influence of the cation and anion of the ionic liquids on the morphology of the synthesized oxides was explored. The results showed that octahydron-shaped Co3O4 nanoparticles with different average sizes depending on the type of the employed ionic liquid. The imidazolium cation was found to play a significant role in reducing the particle size of the synthesized Co3O4 nanoparticles. The samples prepared in the imidazolium-based ionic liquids showed smaller particle sizes than those of the samples prepared in the pyrrolidinium-based ionic liquids. The shape of the particles of the synthesized Co3O4 was not affected by the type of the ionic liquids.Graphical abstractSEM images of sol-gel synthesized cobalt oxide obtained from the ionic liquids [EMIm]TFSA, [EMIm]TfO, [Py1,4]TFSA and [Py1,4]TfO, and calcined at 500 °C.

Disordered regions of respiratory supercomplexes offer new pathways for substrate channeling in crowded membranes.

Are protein supercomplexes biological redundancies or do they have a functional role? This question has been perplexing biophysicists and bioenergeticists for decades, especially on the significances in crowded environments. Now, cryo-EM has brought forth remarkable insights into the structures of supercomplexes, but functional relevance of their intrinsically disordered regions remains a mystery. First, using a combination of maximum entropy (MaxEnT)-guided molecular simulations with low-resolution cryo-EM data, we have resolved the disordered Qcr6 subunit of respiratory complex III (CIII). Second, porting this whole-CIII model into the crowded yeast CIII2CIV2 supercomplex, long-timescale Brownian Dynamics computations were performed. We found that the negative charge on the disordered region surprisingly cooperates with that of the anionic lipids in the mitochondrial membrane to attract a pool of complementary protein substrates in its vicinity. Third, we discovered how the transient fold of Qcr6 expedites directional diffusion of the substrates from CIII to CIV that simultaneously leverage the supercomplex architecture as well as the electrostatic environment of the membrane. A re-classification of the EM images based on our computational model have brought to light new density features that indeed describe the 2-dimensional dynamic of the substrate in crowded supercomplexes. Finally, using multisequence alignment via Alphafold2, mutations are designed to control the diffusivity of substrates by tuning the disorder of the protein surface, for future biochemical validation.

Improving the Neural Segmentation of Blurry Serial SEM Images by Blind Deblurring.

Serial scanning electron microscopy (sSEM) has recently been developed to reconstruct complex largescale neural connectomes, through learning-based instance segmentation. However, blurry images are inevitable amid prolonged automated data acquisition due to imprecision in autofocusing and autostigmation, which impose a great challenge to accurate segmentation of the massive sSEM image data. Recently, learning-based methods, such as adversarial learning and supervised learning, have been proven to be effective for blind EM image deblurring. However, in practice, these methods suffer from the limited training dataset and the underrepresentation of high-resolution decoded features. Here, we propose a semisupervised learning guided progressive decoding network (SGPN) to exploit unlabeled blurry images for training and progressively enrich high-resolution feature representation. The proposed method outperforms the latest deblurring models on real SEM images with much less ground truth input. The improvement of the PSNR and SSIM is 1.04 dB and 0.086, respectively. We then trained segmentation models with deblurred datasets and demonstrated significant improvement in segmentation accuracy. The A-rand (Bogovic et al. 2013) decreased by 0.119 and 0.026, respectively, for 2D and 3D segmentation.

New low-cost, flow-through carbon electrodes characterized in brackish water

We propose a simple and low-cost flow-through electrode for electrochemical cells used for instance in capacitive desalination. We have coated macro-porous carbon fiber papers with various loads of carbon microporous particles to combine both a high surface area and an open structure for good fluid dynamics. In this first study, we restrict our investigation to the charging/discharging behavior, the identification of side reactions, and the effect of geometry on the diffusion of ions. The electrochemical performance was first investigated by cyclic voltammetry and galvanic charge–discharge techniques. The specific capacitance increases by three orders of magnitude upon adding the carbon particles. Then, electrochemical impedance spectroscopy revealed the presence of charge transfer phenomena and modification in the mass transport by the diffusion process for the coated electrode.Graphical abstractSEM image of the surface morphology of the cross section of CP/CFP/CP structure of the flow-through electrode.

Novel Airborne EM Image Appraisal Tool for Imperfect Forward Modeling

Full 3D inversion of time-domain Airborne ElectroMagnetic (AEM) data requires specialists’ expertise and a tremendous amount of computational resources, not readily available to everyone. Consequently, quasi-2D/3D inversion methods are prevailing, using a much faster but approximate (1D) forward model. We propose an appraisal tool that indicates zones in the inversion model that are not in agreement with the multidimensional data and therefore, should not be interpreted quantitatively. The image appraisal relies on multidimensional forward modeling to compute a so-called normalized gradient. Large values in that gradient indicate model parameters that do not fit the true multidimensionality of the observed data well and should not be interpreted quantitatively. An alternative approach is proposed to account for imperfect forward modeling, such that the appraisal tool is computationally inexpensive. The method is demonstrated on an AEM survey in a salinization context, revealing possible problematic zones in the estimated fresh–saltwater interface.

Comparative Study for Patch-Level and Pixel-Level Segmentation of Deep Learning Methods on Transparent Images of Environmental Microorganisms: From Convolutional Neural Networks to Visual Transformers

Currently, the field of transparent image analysis has gradually become a hot topic. However, traditional analysis methods are accompanied by large amounts of carbon emissions, and consumes significant manpower and material resources. The continuous development of computer vision enables the use of computers to analyze images. However, the low contrast between the foreground and background of transparent images makes their segmentation difficult for computers. To address this problem, we first analyzed them with pixel patches, and then classified the patches as foreground and background. Finally, the segmentation of the transparent images was completed through the reconstruction of pixel patches. To understand the performance of different deep learning networks in transparent image segmentation, we conducted a series of comparative experiments using patch-level and pixel-level methods. In two sets of experiments, we compared the segmentation performance of four convolutional neural network (CNN) models and a visual transformer (ViT) model on the transparent environmental microorganism dataset fifth version. The results demonstrated that U-Net++ had the highest accuracy rate of 95.32% in the pixel-level segmentation experiment followed by ViT with an accuracy rate of 95.31%. However, ResNet50 had the highest accuracy rate of 90.00% and ViT had the lowest accuracy of 89.25% in the patch-level segmentation experiments. Hence, we concluded that ViT performed the lowest in patch-level segmentation experiments, but outperformed most CNNs in pixel-level segmentation. Further, we combined patch-level and pixel-level segmentation results to reduce the loss of segmentation details in the EM images. This conclusion was also verified by the environmental microorganism dataset sixth version dataset (EMDS-6).

DeepCLEM: automated registration for correlative light and electron microscopy using deep learning.

In correlative light and electron microscopy (CLEM), the fluorescent images must be registered to the EM images with high precision. Due to the different contrast of EM and fluorescence images, automated correlation-based alignment is not directly possible, and registration is often done by hand using a fluorescent stain, or semi-automatically with fiducial markers. We introduce "DeepCLEM", a fully automated CLEM registration workflow. A convolutional neural network predicts the fluorescent signal from the EM images, which is then automatically registered to the experimentally measured chromatin signal from the sample using correlation-based alignment. The complete workflow is available as a Fiji plugin and could in principle be adapted for other imaging modalities as well as for 3D stacks.

A scalable and modular automated pipeline for stitching of large electron microscopy datasets.

Serial-section electron microscopy (ssEM) is the method of choice for studying macroscopic biological samples at extremely high resolution in three dimensions. In the nervous system, nanometer-scale images are necessary to reconstruct dense neural wiring diagrams in the brain, so -called connectomes. The data that can comprise of up to 108 individual EM images must be assembled into a volume, requiring seamless 2D registration from physical section followed by 3D alignment of the stitched sections. The high throughput of ssEM necessitates 2D stitching to be done at the pace of imaging, which currently produces tens of terabytes per day. To achieve this, we present a modular volume assembly software pipeline ASAP (Assembly Stitching and Alignment Pipeline) that is scalable to datasets containing petabytes of data and parallelized to work in a distributed computational environment. The pipeline is built on top of the Render Trautman and Saalfeld (2019) services used in the volume assembly of the brain of adult Drosophila melanogaster (Zheng et al. 2018). It achieves high throughput by operating only on image meta-data and transformations. ASAP is modular, allowing for easy incorporation of new algorithms without significant changes in the workflow. The entire software pipeline includes a complete set of tools for stitching, automated quality control, 3D section alignment, and final rendering of the assembled volume to disk. ASAP has been deployed for continuous stitching of several large-scale datasets of the mouse visual cortex and human brain samples including one cubic millimeter of mouse visual cortex (Yin et al. 2020); Microns Consortium et al. (2021) at speeds that exceed imaging. The pipeline also has multi-channel processing capabilities and can be applied to fluorescence and multi-modal datasets like array tomography.

Myosin assembly of smooth muscle: from ribbons and side polarity to a row polar helical model

After decades of debate over the structure of smooth muscle myosin filaments, it is still unclear whether they are helical, as in all other muscle types, or square in shape. In both cases bipolar building units are proposed, but the deduced cross-bridge arrangements are fundamentally different. The opposite polarity of the adjusting longitudinal rows is proposed for the helical structure, while in the case of square filaments, or myosin ribbons, only their two faces are appositively polarized. Analysis of our unpublished archival data on light meromyosin (LMM) paracrystals and myosin rod assemblies as well as the filaments themselves indicated that the rods were assembled with a 6°–7° tilt angle from the rods’ longitudinal axis, in contrast to the lack of tilt in LMM, both exhibiting a 14.3 nm myosin periodicity. Optical diffraction analysis of EM images of the rod assemblies and those of intact myosin confirmed their helical architecture characterized by 28 nm residue translations, 172 nm repeats and 516 nm pitch. A detailed helical model of these filaments was elucidated with bipolar tetramer building units made of two polar trimers. The filaments elongate at their two ends in a head-to-head manner, enabling targeted cross-bridge polarity of the adjacent rows, in the form of a unique Boerdijk–Coxeter type helix, similar to that of collagen or desmin fibers, with the covalent links replaced by a head-to-head clasp.

From Flat to Bulb ‐ Novel Insights in Caveolae Membrane Curvature

Caveolae are 60‐100 nm sized bulb‐shaped plasma membrane invaginations found in various cell types including adipocytes, endothelial cells and myocytes. Their cellular functions comprise endocytosis and signaling processes, cellular lipid uptake, as well as sensing and regulation of membrane tension. Thereby, it was shown previously that caveolae can change their membrane shape from invaginated bulbs to flat domains and vice versa. However it is currently not clear which proteins are responsible for the membrane curvature changes. Here, we report the first detailed structural investigation of caveolae membrane shapes in vivo. To do so, Platinum replica EM images of the plasma membrane in mouse embryonic fibroblasts and endothelial cells were correlated to super‐resolution (STED) fluorescence intensities of proteins associated with caveolae domains. This correlative imaging approach (CLEM) allowed us to precisely localize caveolae related proteins to the underlying membrane curvature. The major structural caveolae proteins such as Caveolins and Cavins were found in all caveolae membrane domains independent of their curvature. Surprisingly, EHD2, a caveolar neck protein, was not only detected at bulb‐shaped invaginations but also at flat caveolae. Contrarily, Pacsin/Syndapin2 and EHBP1 were predominantly surrounding flat caveolae domains indicating that both proteins promote the neck formation. Because dynamin was previously associated with caveolae endocytosis and mobility we analyzed its spatial distribution to caveolae. However, our detailed CLEM analysis failed to determine dynamin to caveolae membrane domains in several cell types. Additionally, dynamin lacking cells did not show severe morphological changes of caveolae suggesting that dynamin is not involved in caveolae membrane curvature changes. Taken together, this study gives novel insights in caveolae structure and their formation from flat to invaginated membrane domains.