To count proteins in cells, we developed a reference structure based on the nuclear pore complex, a channel to the nucleus. This not only allowed us to count precise copy numbers of other nuclear pore proteins and of proteins in the kinetochore, a protein machinery that separates the genetic material during cell division, the reference structure also became the gold standard for quality control in superresolution microscopy, used by many groups worldwide.
To increase the throughput of the notoriously slow SMLM we completely automated our microscopes so that they can acquire data around the clock without user intervention. To increase the resolution of SMLM, we developed a novel approach to 3D that uses a nearfield effect, by which a fluorophore can couple its emission directly into the substrate. In addition, we implemented a very complex microscope that uses interferometry to precisely localize fluorophores in 3D in entire cells. Finally, we integrated a new SMLM variant called MINFLUX in the lab. We established its use for fixed and live cell imaging and for tracking single proteins in cells. This allowed us to resolve for the first time the steps that a motor protein called kinesin takes while it transports molecules through a cell. Finally, we developed a new approach to MINFLUX that has the prospect to substantially simplify these still complex and expensive instruments.
We increased the number of fluorophores that can be imaged simultaneously in SMLM to four colors by new hardware and software technologies, that now allow us to precisely dissect the structure of complex protein machines. By developing technologies that directly correlate SMLM and EM, we hope to combine the best of both worlds and combine angstrom resolution and cellular context of EM with the molecular specificity of SMLM. To this end, we build an SMLM microscope that works under cryogenic temperatures and can image cells on an EM grid directly before the EM images are acquired.
We developed new algorithms and software to determine the positions of single fluorophores with a highest precision, even under challenging circumstances such as very high densities or multiple channels using new deep-learning approaches. Another important software development was an algorithm to extract specific structural parameters from the fluorophore positions, which can then be used to quantitatively answer biological questions. All our software developments are integrated into an open-source software platform.
All these tools now allowed us to see the process of endocytosis in a new light and to uncover unknown mechanisms. In the model organism yeast, we mapped out the positions of many endocytic proteins and discovered how the force is generated that pulls in the membrane. In human cells we investigated the mechanism how the protein coat around the growing vesicle is formed and could resolve a long-standing question the field.
