Combining genetic code expansion and bioorthogonal click labeling with deep learning and particle averaging for improved super-resolution imaging

• Applicants: Professor Dr. Philip Kollmannsberger; Professor Dr. Markus Sauer
• Subject Area: Biophysics ; Analytical Chemistry ; Bioinformatics and Theoretical Biology ; Organic Molecular Chemistry - Synthesis and Characterisation ; Structural Biology
• Term: from 2020 to 2024
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 440773101

• Final Report Year: 2025

Final Report Abstract

In recent years, super-resolution microscopy has been established as a powerful method for subdiffraction-resolution fluorescence imaging of cells and tissue. With further improvements in spatial resolution, e.g., in combination with expansion microscopy, both the size as well as the density of fluorescence markers have become key limiting factors for super-resolution microscopy. In this project, we addressed both challenges together using a combination of experimental and computational approaches. The focus of the experimental part of the project was on reducing fluorescent probe size with novel labelling strategies based on the use of unnatural amino acids, and combining them with optimized approaches of super-resolution microscopy. Here, we developed genetic code expansion (GCE) as a versatile approach to introduce unnatural amino acids site-specifically into proteins followed by click labelling with tetrazine dyes with minimal linkage error. This allowed us to stoichiometrically label tetrameric GluK2 receptors as well as trimeric PCNA proteins and demonstrate that sub-10 nm distances between the click sites can be spatially resolved by DNA-PAINT and expansion microscopy. The computational part of the project was about increasing the effective resolution at low labelling density in unexpanded and expanded samples using deep learning. In the first phase, we developed a novel algorithm based on Fourier Ring Correlation and Anisotropic Kernel Density Estimation to improve the reconstruction of microtubules. We then set out to improve the first stage of the reconstruction phase, the fitting of the raw SMLM movies, by combining deep learning with compressed sensing, and by extending the fitting context to the time domain. Finally, we developed novel particle averaging methods based on geometric deep learning. All three methods are openly available to the community and enable the mapping of subcellular structures from low-density expansion microscopy data with unprecedented resolution.

Publications

• Bioorthogonal labeling of transmembrane proteins with non-canonical amino acids unveils masked epitopes in live neurons. Nature Communications, 12(1).
  Bessa-Neto, Diogo; Beliu, Gerti; Kuhlemann, Alexander; Pecoraro, Valeria; Doose, Sören; Retailleau, Natacha; Chevrier, Nicolas; Perrais, David; Sauer, Markus & Choquet, Daniel
  
• Fourier Ring Correlation and Anisotropic Kernel Density Estimation Improve Deep Learning Based SMLM Reconstruction of Microtubules. Frontiers in Bioinformatics, 1.
  Berberich, Andreas; Kurz, Andreas; Reinhard, Sebastian; Paul, Torsten Johann; Burd, Paul Ray; Sauer, Markus & Kollmannsberger, Philip
  
• Photoblueing of organic dyes can cause artifacts in super-resolution microscopy. Nature Methods, 18(3), 253-257.
  Helmerich, Dominic A.; Beliu, Gerti; Matikonda, Siddharth S.; Schnermann, Martin J. & Sauer, Markus
  
• Super-resolving Microscopy in Neuroscience. Chemical Reviews, 121(19), 11971-12015.
  Werner, Christian; Sauer, Markus & Geis, Christian
  
• Photoswitching fingerprint analysis bypasses the 10-nm resolution barrier. Nature Methods, 19(8), 986-994.
  Helmerich, Dominic A.; Beliu, Gerti; Taban, Danush; Meub, Mara; Streit, Marcel; Kuhlemann, Alexander; Doose, Sören & Sauer, Markus
  
• ReCSAI: recursive compressed sensing artificial intelligence for confocal lifetime localization microscopy. BMC Bioinformatics, 23(1).
  Reinhard, Sebastian; Helmerich, Dominic A.; Boras, Dominik; Sauer, Markus & Kollmannsberger, Philip
  
• Coronaviruses Use ACE2 Monomers as Entry‐Receptors. Angewandte Chemie International Edition, 62(30).
  Eiring, Patrick; Klein, Teresa; Backes, Simone; Streit, Marcel; Jungblut, Marvin; Doose, Sören; Beliu, Gerti & Sauer, Markus
  
• PCNA as Protein‐Based Nanoruler for Sub‐10 nm Fluorescence Imaging. Advanced Materials, 36(7).
  Helmerich, Dominic A.; Budiarta, Made; Taban, Danush; Doose, Sören; Beliu, Gerti & Sauer, Markus
  
• Improving single molecule localisation microscopy reconstruction by extending the temporal context. openRxiv.
  Reinhard, Sebastian; Ebert, Vincent; Schrama, Jann; Sauer, Markus & Kollmannsberger, Philip