Untrained Neural Network Priors for Inverse Imaging Problems: A Survey

preprint version

Untrained Neural Network Priors for Inverse Imaging Problems: A Survey

preprint version

Imaging unlabeled proteins on DNA with super-resolution

Fluorescence microscopy is invaluable to a range of biomolecular analysis approaches. The required labeling of proteins of interest, however, can be challenging and potentially perturb biomolecular functionality as well as cause imaging artefacts and photo bleaching issues. Here, we introduce inverse (super-resolution) imaging of unlabeled proteins bound to DNA. In this new method, we use DNA-binding fluorophores that transiently label bare DNA but not protein-bound DNA. In addition to demonstrating diffraction-limited inverse imaging, we show that inverse Binding-Activated Localization Microscopy or ‘iBALM’ can resolve biomolecular features smaller than the diffraction limit. The current detection limit is estimated to lie at features between 5 and 15 nm in size. Although the current image-acquisition times preclude super-resolving fast dynamics, we show that diffraction-limited inverse imaging can reveal molecular mobility at ∼0.2 s temporal resolution and that the method works both with DNA-intercalating and non-intercalating dyes. Our experiments show that such inverse imaging approaches are valuable additions to the single-molecule toolkit that relieve potential limitations posed by labeling.