Macromolecular Architecture Of The Synapse

National Institute Of Neurological Disorders And Stroke

Investigators

Linked publications & trials

Abstract

Progress Report Macromolecular Architecture of The Synapse A key priority for this project is developing methods to determine the molecular identity of structures within tomograms of the PSD. We developed a nanobody labeling method for high-resolution molecular identification in electron tomograms. A nanobody (~3 nm) conjugated to nanogold binds directly to the target protein with high specificity, allowing us to use small gold particle labels on synaptic proteins in neurons and identify them in dark-field STEM tomograms. We have analyzed tomograms labeled with nanobodies for PSD-95, CaMKII, and Homer1b in intact synapses, successfully identifying these proteins in spine synapses. Data for PSD-95 and CaMKII are compelling, and we are working to complete the analysis for Homer1b, which could reveal a new structural model involving Homer in the PSD. We have significantly refined a genetic labeling technique using APEX2 for EM tomography. This enzyme catalyzes the formation of electron-dense material, enabling detailed visualization of CaMKII in rat hippocampal neurons. Using dark-field STEM tomography, we identified APEX2-labeled CaMKII in dendritic spines, revealing its distribution within the spine, on the membrane, and in the PSD. This work has provided insights into CaMKII self-association, its structural role in the PSD, and its response to synaptic activity. For example, we observed increased electron-dense proteins in the PSD following synaptic stimulation, indicating heightened CaMKII activation and its association with NMDAR complexes. Even in basal states, CaMKII was detected in the core PSD structure, suggesting its dynamic role in organizing the PSD, likely in conjunction with NMDARs and AMPARs. We have analyzed over 70 dual-axis STEM tomograms for this study and are developing methods to combine immunogold labeling with APEX2 to further explore the relationship between CaMKII and glutamate receptors. Preliminary findings have been presented at 2023 SFN nanosymposia, and a publication is in preparation. Expanding on this approach, we developed a CRISPR knockin APEX2-tagged CaMKII construct to label endogenous CaMKII without overexpression. This enabled us to identify APEX2-labeled endogenous CaMKII in synapses within neuronal cultures, advancing our understanding of its native distribution and function. We have explored multiple approaches in cryo-EM tomography with notable progress. Our efforts include successful freezing and imaging of isolated PSDs, synaptosomes, and cultured neurons, resulting in the acquisition of over 150 cryo-ET series using advanced cryo-EM facilities. We processed these datasets on the NIH Biowulf supercomputer cluster, applying sophisticated software for motion correction, alignment, and 3D reconstruction. AI-based tools revealed lattice-like structures in isolated PSDs, akin to known NMDAR-type formations. In collaboration with the NIH cryo-EM facility, we obtained promising cryo-tomography data from synaptosomes, providing insights into molecular structures within brain synapses and PSDs. We are also integrating protein labeling techniques, such as immunogold and APEX2, into our cryo-EM studies, though additional microscope time is needed for further data acquisition. Moreover, we utilized cryo-focused ion beam milling to prepare thin lamellae for tomography, with initial results showing detailed synaptic structures. We aim to enhance this technique with cryo-correlated light and electron microscopy (cryo-CLEM) for more precise targeting in future tomography studies. Our overarching goal is to achieve or surpass the quality of results obtained with traditional HPF/FS techniques. Our transsynaptic assembly project investigates intracellular structures linked by cleft-spanning structures. Nearly all transcleft objects have some intracellular component. Assemblies form large domains or association domains, explaining the nanodomain phenomena. We classified and enumerated over three thousand structures, designing an algorithm to randomly display a structure, prompt the user for a description, and parse descriptions for common morphological elements. This information helped identify common structures associated with assemblies. This work was published last year. Further, using dark field STEM tomography of rodent synapses immunogold labeled for endogenous PSD-95, we found structural evidence that postsynaptic PSD-95 corresponds to presynaptic tethering, priming, and fusing vesicles. EM tomography images suggest a new structural model where transsynaptic assembly organization underlies synaptic transmission and plasticity. EM tomography also reveals various states of vesicle fusing and other trafficking events at presynaptic terminals, providing insights into vesicle exocytosis. This work was presented at the 2024 “Molecular and Cell Biology of the Neuromuscular System “, 2024 SFN meeting and 2025 Gordon conference on “Excitatory Synapses and Brain Function” is being prepared for publication. Unfortunately, technology limits our ability to analyze enough synapses to identify precise assembly structures. We are pairing this project with the automated segmentation project to increase the number of structures analyzed. Automated segmentation is crucial for the future of electron tomography. Our automatic segmentation optimization method (ASOM) segmented detailed structures of fragments isolated from sonicated and control PSDs imaged by cryo-EM. However, many structures within PSDs segmented by ASOM are still interconnected. To examine these structures in more detail, we improved ASOM by adding watershed segmentation, which separates connected structures automatically. This combination enabled the segmentation of hundreds of tightly packed granular structures in intact PSDs into individual modules. ASOM can automatically segment filaments connected to the postsynaptic membrane, revealing that PSD-95-like filaments can be automatically segmented. The improved ASOM also segments other distinct classes, such as those connected to the presynaptic membrane, postsynaptic membranes, and vesicle membranes. Further, the automatic segmentation of transsynaptic components is consistent with assemblies obtained by manual segmentation, expediting the segmentation process. We have implemented the Simultaneous Iterative Reconstruction Technique (SIRT), improved accuracy and reduced conventional ET and cryo-ET noise. Our SIRT method produces EM tomograms more efficiently than IMOD, with equal or better quality. Combining ASOM with skeletonization has shown that ASOM can automatically segment distinct transsynaptic structures as accurately as manual methods. More work is needed to match the detail of hand segmentation. While we continue to test other published or self-constructed machine learning-based software to automate the segmentation of 3D EM data, we are developing a software package that streamlines the entire tomography data pipeline, from image alignment to visualization. We will integrate machine learning and AI-based object classification algorithms to decrease the time needed to analyze a tomogram from several months to a few days. The Bioinformatics Core is helping us bring our segmentation software to web browsers, making it usable for anyone on any device—desktop, laptop, or tablet. This initiative increases the number of available tomographers per lab and provides them with advanced tools, accelerating analysis. This decreases the barrier for laboratories to move into structural neurobiology by electron tomography.