@article{ReinhardHelmerichBorasetal.2022,
  author    = {Reinhard, Sebastian and Helmerich, Dominic A. and Boras, Dominik and Sauer, Markus and Kollmannsberger, Philip},
  title     = {ReCSAI: recursive compressed sensing artificial intelligence for confocal lifetime localization microscopy},
  series = {BMC Bioinformatics},
  volume    = {23},
  journal   = {BMC Bioinformatics},
  number    = {1},
  doi       = {10.1186/s12859-022-05071-5},
  url       = {http://nbn-resolving.de/urn:nbn:de:bvb:20-opus-299768},
  year      = {2022},
  abstract  = {Background Localization-based super-resolution microscopy resolves macromolecular structures down to a few nanometers by computationally reconstructing fluorescent emitter coordinates from diffraction-limited spots. The most commonly used algorithms are based on fitting parametric models of the point spread function (PSF) to a measured photon distribution. These algorithms make assumptions about the symmetry of the PSF and thus, do not work well with irregular, non-linear PSFs that occur for example in confocal lifetime imaging, where a laser is scanned across the sample. An alternative method for reconstructing sparse emitter sets from noisy, diffraction-limited images is compressed sensing, but due to its high computational cost it has not yet been widely adopted. Deep neural network fitters have recently emerged as a new competitive method for localization microscopy. They can learn to fit arbitrary PSFs, but require extensive simulated training data and do not generalize well. A method to efficiently fit the irregular PSFs from confocal lifetime localization microscopy combining the advantages of deep learning and compressed sensing would greatly improve the acquisition speed and throughput of this method. Results Here we introduce ReCSAI, a compressed sensing neural network to reconstruct localizations for confocal dSTORM, together with a simulation tool to generate training data. We implemented and compared different artificial network architectures, aiming to combine the advantages of compressed sensing and deep learning. We found that a U-Net with a recursive structure inspired by iterative compressed sensing showed the best results on realistic simulated datasets with noise, as well as on real experimentally measured confocal lifetime scanning data. Adding a trainable wavelet denoising layer as prior step further improved the reconstruction quality. Conclusions Our deep learning approach can reach a similar reconstruction accuracy for confocal dSTORM as frame binning with traditional fitting without requiring the acquisition of multiple frames. In addition, our work offers generic insights on the reconstruction of sparse measurements from noisy experimental data by combining compressed sensing and deep learning. We provide the trained networks, the code for network training and inference as well as the simulation tool as python code and Jupyter notebooks for easy reproducibility.},
  language  = {en}
}
@article{BerberichKurzReinhardetal.2021,
  author    = {Berberich, Andreas and Kurz, Andreas and Reinhard, Sebastian and Paul, Torsten Johann and Burd, Paul Ray and Sauer, Markus and Kollmannsberger, Philip},
  title     = {Fourier Ring Correlation and anisotropic kernel density estimation improve deep learning based SMLM reconstruction of microtubules},
  series = {Frontiers in Bioinformatics},
  volume    = {1},
  journal   = {Frontiers in Bioinformatics},
  doi       = {10.3389/fbinf.2021.752788},
  url       = {http://nbn-resolving.de/urn:nbn:de:bvb:20-opus-261686},
  year      = {2021},
  abstract  = {Single-molecule super-resolution microscopy (SMLM) techniques like dSTORM can reveal biological structures down to the nanometer scale. The achievable resolution is not only defined by the localization precision of individual fluorescent molecules, but also by their density, which becomes a limiting factor e.g., in expansion microscopy. Artificial deep neural networks can learn to reconstruct dense super-resolved structures such as microtubules from a sparse, noisy set of data points. This approach requires a robust method to assess the quality of a predicted density image and to quantitatively compare it to a ground truth image. Such a quality measure needs to be differentiable to be applied as loss function in deep learning. We developed a new trainable quality measure based on Fourier Ring Correlation (FRC) and used it to train deep neural networks to map a small number of sampling points to an underlying density. Smooth ground truth images of microtubules were generated from localization coordinates using an anisotropic Gaussian kernel density estimator. We show that the FRC criterion ideally complements the existing state-of-the-art multiscale structural similarity index, since both are interpretable and there is no trade-off between them during optimization. The TensorFlow implementation of our FRC metric can easily be integrated into existing deep learning workflows.},
  language  = {en}
}
@article{KuhlemannBeliuJanzenetal.2021,
  author    = {Kuhlemann, Alexander and Beliu, Gerti and Janzen, Dieter and Petrini, Enrica Maria and Taban, Danush and Helmerich, Dominic A. and Doose, S{\"o}ren and Bruno, Martina and Barberis, Andrea and Villmann, Carmen and Sauer, Markus and Werner, Christian},
  title     = {Genetic Code Expansion and Click-Chemistry Labeling to Visualize GABA-A Receptors by Super-Resolution Microscopy},
  series = {Frontiers in Synaptic Neuroscience},
  volume    = {13},
  journal   = {Frontiers in Synaptic Neuroscience},
  issn      = {1663-3563},
  doi       = {10.3389/fnsyn.2021.727406},
  url       = {http://nbn-resolving.de/urn:nbn:de:bvb:20-opus-251035},
  year      = {2021},
  abstract  = {Fluorescence labeling of difficult to access protein sites, e.g., in confined compartments, requires small fluorescent labels that can be covalently tethered at well-defined positions with high efficiency. Here, we report site-specific labeling of the extracellular domain of γ-aminobutyric acid type A (GABA-A) receptor subunits by genetic code expansion (GCE) with unnatural amino acids (ncAA) combined with bioorthogonal click-chemistry labeling with tetrazine dyes in HEK-293-T cells and primary cultured neurons. After optimization of GABA-A receptor expression and labeling efficiency, most effective variants were selected for super-resolution microscopy and functionality testing by whole-cell patch clamp. Our results show that GCE with ncAA and bioorthogonal click labeling with small tetrazine dyes represents a versatile method for highly efficient site-specific fluorescence labeling of proteins in a crowded environment, e.g., extracellular protein domains in confined compartments such as the synaptic cleft.},
  language  = {en}
}
@article{PaulPauliEhmannetal.2015,
  author    = {Paul, Mila M. and Pauli, Martin and Ehmann, Nadine and Hallermann, Stefan and Sauer, Markus and Kittel, Robert J. and Heckmann, Manfred},
  title     = {Bruchpilot and Synaptotagmin collaborate to drive rapid glutamate release and active zone differentiation},
  series = {Frontiers in Cellular Neuroscience},
  volume    = {9},
  journal   = {Frontiers in Cellular Neuroscience},
  number    = {29},
  doi       = {10.3389/fncel.2015.00029},
  url       = {http://nbn-resolving.de/urn:nbn:de:bvb:20-opus-148988},
  year      = {2015},
  abstract  = {The active zone (AZ) protein Bruchpilot (Brp) is essential for rapid glutamate release at Drosophila melanogaster neuromuscular junctions (NMJs). Quantal time course and measurements of action potential-waveform suggest that presynaptic fusion mechanisms are altered in brp null mutants (brp\(^{69}\)). This could account for their increased evoked excitatory postsynaptic current (EPSC) delay and rise time (by about 1 ms). To test the mechanism of release protraction at brp\(^{69}\) AZs, we performed knock-down of Synaptotagmin-1 (Syt) via RNAi (syt\(^{KD}\)) in wildtype (wt), brp\(^{69}\) and rab3 null mutants (rab3\(^{rup}\)), where Brp is concentrated at a small number of AZs. At wt and rab3\(^{rup}\) synapses, syt\(^{KD}\) lowered EPSC amplitude while increasing rise time and delay, consistent with the role of Syt as a release sensor. In contrast, syt\(^{KD}\) did not alter EPSC amplitude at brp\(^{69}\) synapses, but shortened delay and rise time. In fact, following syt\(^{KD}\), these kinetic properties were strikingly similar in wt and brp\(^{69}\), which supports the notion that Syt protracts release at brp\(^{69}\) synapses. To gain insight into this surprising role of Syt at brp\(^{69}\) AZs, we analyzed the structural and functional differentiation of synaptic boutons at the NMJ. At tonic type Ib motor neurons, distal boutons contain more AZs, more Brp proteins per AZ and show elevated and accelerated glutamate release compared to proximal boutons. The functional differentiation between proximal and distal boutons is Brp-dependent and reduced after syt\(^{KD}\). Notably, syt\(^{KD}\) boutons are smaller, contain fewer Brp positive AZs and these are of similar number in proximal and distal boutons. In addition, super-resolution imaging via dSTORM revealed that syt\(^{KD}\) increases the number and alters the spatial distribution of Brp molecules at AZs, while the gradient of Brp proteins per AZ is diminished. In summary, these data demonstrate that normal structural and functional differentiation of Drosophila AZs requires concerted action of Brp and Syt.},
  language  = {en}
}