Non-Invasive, Label-free Image Approaches to Predict Multimodal Molecular Markers in Pluripotency Assessment

We detail an innovative non-invasive technique to predict gene and protein expression in pluripotent stem cells using advanced bright-field microscopy. The method employs machine learning algorithms to classify cells based on brightfield images, avoiding the need for traditional staining and manual annotation. Our approach uses DeepLearning technology to predict the gene expression (qPCR, RNA-seq) and protein expression (immunostaining, Flowcytometry) of cells from brightfield microscopy images. It provides a robust tool for non-destructive and continuous monitoring of the pluripotency status of stem cells, which will greatly advance regenerative medicine. It will be an approach that will contribute significantly to the manufacturing process of cellular products, especially where non-destructive and continuous monitoring is required. Overall design: Four different culture conditions were used for following iPSC maintenance and propagation. (1) Control: Cultured as in the general iPSC maintenance culture method. (2) Low nutrient: Cultured in inactivated medium (StemFit medium with heat treatment at 56°C for 30 min). (3) Differentiation: Cultured using differentiation medium (DMEM medium supplemented with 10% FBS, 1% MEM NEAA, 1% GlutaMax). (4) Physical stimulas: iPS cells were suspended by pipetting 20 times at passaging, and then cultured in a general way. Microscopic imaging was performed on days 1 and 4 under these four conditions, and various evaluations (qPCR, Flow Cytometry, immunocytochemistry, and RNA-seq) were performed on the cells after the imaging on day 5. Four conditions were placed in the same 6-well plate, one well per condition, and three plates were prepared in the same way (N=3).RNA-seq data were tied to each culture condition and named as follows. (1) control: YOK1, (2)Low nutrient: YOK2, (3)Differentiation: YOK3, (4)Physical stimulas: YOK4 YOK1-1 represents the first RNA-seq data of the YOK1 condition, for which data were acquired at N = 3.