LSI researchers and colleagues at Simon Fraser University combined 3D super-resolution microscopy and deep learning to gain important insights into how replication of the Zika virus reorganizes the endoplasmic reticulum (ER) that may help identify potential viral inhibitors.
The endoplasmic reticulum is a complex cellular organelle composed of tubules and sheets whose 30-100 nm width is below the diffraction limit of light. While diffraction limited confocal microscopy shows segregation of the ER into peripheral and central ER domains, better definition of ER structure by florescence microscopy has been a challenge, according to Dr. Ivan Robert Nabi, senior author of a new publication in Scientific Reports.
Super-resolution microscopy has better characterized the structure of peripheral ER tubules and sheets and identified peripheral convoluted networks of ER tubules, or matrices. Application of this technology to the denser, central ER is, however, more challenging. “In this publication, we used 3D STED super-resolution microscopy to define for the first time that a dense central tubular matrix is associated with Zika virus replication factories,” states Dr. Nabi, a Professor of cancer cell biology and director of LSI IMAGING, a super-resolution microscope facility located in the LSC.
“We further showed that a deep convolutional neural network can identify Zika-virus infected cells, based on ER reorganization. This represents the first application of deep learning to super-resolution microscopy images, and to the endoplasmic reticulum.”
The ability to reorganize the central ER by Zika virus is important, as the ability to detect viral infection through alteration of host cell properties is key to screening for viral inhibitors. “This has become even more imperative in light of the ongoing COVID-19 crisis,” says Nabi. As with flaviviruses such as Zika, SARS-CoV-2 coronavirus infection, which causes COVID-19, is also associated with ER reorganization.
“We are actively applying super-resolution microscopy and AI analysis to define ER morphological changes associated with SARS-COV-2, including defining ER contacts with other organelles,” adds Nabi. “Deep learning detection of Zika virus infected cells represents proof-of-principle of a sensitive detection approach to identify SARS-COV-2 infected cells, based on host cell reorganization of the ER.
“In light of the urgency of detecting inhibitors of SARS-CoV-2 infection, this publication represents not only a novel detection of Zika virus-induce central tubular matrix by super-resolution microscopy and deep learning, but also, a highly topical manuscript of relevance to important ongoing efforts to combat and develop novel therapeutics for treating COVID-19.”
Read the paper
Long, R.K.M., Moriarty, K.P., Cardoen, B. et al. Super resolution microscopy and deep learning identify Zika virus reorganization of the endoplasmic reticulum. Sci Rep 10, 20937 (2020). https://doi.org/10.1038/s41598-020-77170-3
Canadian Institutes for Health Research (CIHR; Nabi: PJT-148698; Jean: PJT-153434); Natural Sciences and Engineering Research Council of Canada (NSERC; IRN: RGPIN-2019-05179; GH: RGPIN-2015-06795, RGPIN-2020-06752; FJ: CRDPJ 531024-18); Infrastructure: Canadian Foundation of Innovation/BC Knowledge Development Fund; Strategic Investment Fund (Faculty of Medicine, UBC). COVID-19 research in the Jean, Nabi and Hamarneh labs is supported by CIHR COVID-19 May 2020 Rapid Research (VR3-172639) and NSERC Alliance COVID-19 grants.