Nature Photonics, 28 September, 2026, DOI：https://doi.org/10.1038/s41566-026-02019-6
Lattice light sheet activation structured illumination super-resolution microscopy
Xue Dong, Quan Meng, Xiaoyu Yang, Haoyu Chen, Chang Qiao, Yuhuan Lin, Siwei Zhang, Xiaohan Geng, Linghui Luan, Tao Jiang, Wenfeng Fu, Amin Jiang, Wencong Xu, Jiabao Guo, Rongfei Wei & Dong Li
Abstract
Volumetric super-resolution (SR) live imaging at high speed has been a long-standing challenge, especially in thick fluorescent specimens. Here we develop lattice light sheet activation structured illumination microscopy (LA-SIM), which integrates the advantages of superior optical sectioning and low phototoxicity offered by light sheet microscopy with SR-SIM in both linear and nonlinear schemes, using reversibly photoswitchable fluorescent protein. We also implement rationalized deep-learning denoising in a self-supervised manner, enabling high-quality three-dimensional SR images at limited signal-to-noise ratios, while achieving spatial resolution improvement beyond twofold relative to the diffraction limit in both lateral and axial dimensions (up to 87?nm and 203?nm, respectively). We utilize the high-quality rationalized deep-learning LA-SIM data to train a large-scale transformer model, SRFormer-pMoE, which overcomes the inferior properties and limited palette of reversibly photoswitchable fluorescent proteins. We show that the LA-SIM system syncretized with the SRFormer-pMoE model enables multicolour volumetric SR imaging from cultured cells to multicellular embryos, shedding new light on bioprocesses by tracking subcellular dynamics at high spatiotemporal resolution for durations of hours.
Article link：
https://www.nature.com/articles/s41566-026-02019-6
附件下载：