Abstract
When imaging the nucleus structure of a cell, the out-of-focus fluorescence acts as background and hinders the detection of weak signals. Light-sheet fluorescence microscopy (LSFM) is a wide-field imaging approach which has the best of both background removal and imaging speed. However, the commonly adopted orthogonal excitation/detection scheme is hard to be applied to single-cell imaging due to steric hindrance. For LSFMs capable of high spatiotemporal single-cell imaging, the complex instrument design and operation largely limit their throughput of data collection. Here, we propose an approach for high-throughput background-free fluorescence imaging of single cells facilitated by the Immersion Tilted Light Sheet Microscopy (ImTLSM). ImTLSM is based on a light-sheet projected off the optical axis of a water immersion objective. With the illumination objective and the detection objective placed opposingly, ImTLSM can rapidly patrol and optically section multiple individual cells while maintaining single-molecule detection sensitivity and resolution. Further, the simplicity and robustness of ImTLSM in operation and maintenance enables high-throughput image collection to establish background removal datasets for deep learning. Using a deep learning model to train the mapping from epi-illumination images to ImTLSM illumination images, namely PN-ImTLSM, we demonstrated cross-modality fluorescence imaging, transforming the epi-illumination image to approach the background removal performance obtained with ImTLSM. We demonstrated that PN-ImTLSM can be generalized to large-field homogeneous illumination imaging, thereby further improving the imaging throughput. In addition, compared to commonly used background removal methods, PN-ImTLSM showed much better performance for areas where the background intensity changes sharply in space, facilitating high-density single-molecule localization microscopy. In summary, PN-ImTLSM paves the way for background-free fluorescence imaging on ordinary inverted microscopes.