Super-resolution microscopy, or nanoscopy, enables the use of fluorescent-based molecular localization tools to study molecular structure at the nanoscale level in the intact cell, bridging the mesoscale gap to classical structural biology methodologies. Analysis of super-resolution data by artificial intelligence (AI), such as machine learning, offers tremendous potential for the discovery of new biology, that, by definition, is not known and lacks ground truth. Herein, we describe the application of weakly supervised paradigms to super-resolution microscopy and its potential to enable the accelerated exploration of the nanoscale architecture of subcellular macromolecules and organelles.
SuperResNET is an integrated machine learning‐based analysis software for visualizing and quantifying 3D point cloud data acquired by single‐molecule localization microscopy (SMLM). SuperResNET computational modules include correction for multiple blinking of single fluorophores, denoising, segmentation (clustering), feature extraction used for cluster group identification, modularity analysis, blob retrieval, and visualization in 2D and 3D. Here, a graphical user interface version of SuperResNET was applied to publicly available direct stochastic optical reconstruction microscopy (dSTORM) data of nucleoporin Nup96 and Nup107 labeled nuclear pores that present a highly organized octagon structure of eight corners. SuperResNET effectively segments nuclear pores and Nup96 corners based on differential proximity threshold analysis from 2D and 3D SMLM datasets. SuperResNET quantitatively analyzes features from segmented nuclear pores, including complete structures with eightfold symmetry, and from segmented corners. SuperResNET modularity analysis of segmented corners from 2D SMLM distinguishes two modules at 10.7 ± 0.1 nm distance, corresponding to two individual Nup96 molecules. SuperResNET is therefore a model‐free tool that can reconstruct network architecture and molecular distribution of subcellular structures without the bias of a specified prior model, attaining molecular resolution from dSTORM data. SuperResNET provides flexibility to report on structural diversity in situ within the cell, providing opportunities for biological discovery.
SuperResNET is a network analysis pipeline for the analysis of point cloud data generated by single-molecule localization microscopy (SMLM). Here, we applied SuperResNET network analysis of SMLM direct stochastic optical reconstruction microscopy (dSTORM) data to determine how the clathrin endocytosis inhibitors pitstop 2, dynasore and latrunculin A (LatA) alter the morphology of clathrin-coated pits. SuperResNET analysis of HeLa and Cos7 cells identified three classes of clathrin structures: small oligomers (class I), pits and vesicles (class II), and larger clusters corresponding to fused pits or clathrin plaques (class III). Pitstop 2 and dynasore treatment induced distinct homogeneous populations of class II structures in HeLa cells, suggesting that they arrest endocytosis at different stages. Inhibition of endocytosis was not via actin depolymerization, as the actin-depolymerizing agent LatA induced large, heterogeneous clathrin structures. Ternary analysis of SuperResNET shape features presented a distinct more planar profile for blobs from pitstop 2-treated cells, which aligned with clathrin pits identified with high-resolution minimal photon fluxes (MINFLUX) microscopy, whereas control structures resembled MINFLUX clathrin vesicles. SuperResNET analysis therefore showed that pitstop 2 arrests clathrin pit maturation at early stages of pit formation, representing an approach to detect the effect of small molecules on target structures in situ in the cell from SMLM datasets.
Single-molecule localization microscopy (SMLM) achieves nanoscale imaging of complex protein structures in the cell. However, the ability to capture structural variability across cell conditions (cell lines, gene expression, treatment) from 3D point cloud SMLM data remains limited. We present siMILe, a weakly-supervised multiple instance learning (MIL) machine learning method to close this gap in interpretable subcellular discovery. siMILe identifies condition-specific changes in protein assemblies by leveraging their shape and network features, without requiring structure-level supervision. siMILe improves structure classification by extending embedded instance selection (MILES) through adversarial erasing and a symmetric classifier. We validated siMILe by detecting caveolae from caveolin-1 (Cav1) labeled PC3 prostate cancer cells differentially expressing cavin-1. In PC3-CAVIN1 cells, cavin-1 closely associates with siMILe-identified caveolae, to a lesser extent with higher-order non-caveolar Cav1 scaffolds, but not small Cav1 oligomers corresponding to 8S complexes, supporting a role for progressive cavin-1 interaction in 8S complex oligomerization. We also validated siMILe on simulated SMLM data and in detecting inhibitor-induced structural variations within clathrin-coated pit data. These results highlight siMILe’s potential to identify differential molecular structures in distinct cell conditions. siMILe extends the SuperResNET SMLM software platform with the ability to detect interpretable structural differences across conditions.
Specificity of small molecules for their target molecule in the cell is critical to determine their effective use as biologics and therapeutics. Small molecule inhibitors of clathrin endocytosis, Pitstop 2, and the dynamin inhibitor Dynasore, have off-target effects and their specificity has been challenged. Here, we used SuperResNET to apply network analysis to 20 nm resolution dSTORM single-molecule localization microscopy (SMLM) to test whether Pitstop 2 and Dynasore alter the morphology of clathrin coated pits in intact cells. SuperResNET analysis of dSTORM data from HeLa and Cos7 cells identifies three classes of clathrin structures: small oligomers (Class I); pits and vesicles (Class II); and larger clusters corresponding to fused clathrin pits and clathrin plaques (Class III). SuperResNET analysis of high resolution MinFlux imaging identifies Class 1 oligomers as well as Class 2 structures including morphologically identifiable clathrin pits and vesicles. SuperResNET feature analysis of dSTORM data shows that Pitstop 2 and Dynasore induce the formation of distinct homogenous populations of clathrin structures in HeLa cells. Pitstop 2 blobs are smaller and more elongated than those induced by Dynasore, indicating that these two clathrin inhibitors arrest clathrin endocytosis at distinct stages. Pitstop 2 and Dynasore are not impacting clathrin structure via actin depolymerization as the actin depolymerizing agent latrunculin A (LatA) induced larger heterogeneous clathrin structures. Ternary analysis of SuperResNET shape features presents a distinct profile for Pitstop 2 Class II structures. The most representative Pitstop blobs align with and resemble MinFlux clathrin pits while control structures resemble Minflux clathrin vesicles. SuperResNET analysis of SMLM data is therefore a highly sensitive approach to detect the effect of small molecules on target molecule structure in situ in the cell.
Single‐molecule localization microscopy (SMLM) achieves nanoscale imaging of complex protein structures in the cell. However, the ability to capture structural variability across cell conditions (cell lines, gene expression, treatment) from 3D point cloud SMLM data remains limited. We present siMILe, a weakly supervised multiple instance learning machine learning method to close this gap in interpretable subcellular discovery. siMILe identifies condition‐specific changes in protein assemblies by leveraging their shape and network features, without requiring structure‐level supervision. siMILe improves structure classification by extending embedded instance selection through adversarial erasing and a symmetric classifier. We validated siMILe by detecting caveolae from caveolin‐1 (Cav1) labeled PC3 prostate cancer cells differentially expressing cavin‐1. In PC3‐CAVIN1 cells, cavin‐1 closely associates with siMILe‐identified caveolae, to a lesser extent with higher‐order noncaveolar Cav1 scaffolds, but not small Cav1 oligomers corresponding to 8S complexes, supporting a role for progressive cavin‐1 interaction in 8S complex oligomerization. We also validated siMILe on simulated SMLM data and in detecting inhibitor‐induced structural variations within clathrin‐coated pit data. These results highlight siMILe's potential to identify differential molecular structures in distinct cell conditions. siMILe extends the SuperResNET SMLM software platform with the ability to detect interpretable structural differences across conditions.