Back

A prioritized medium-throughput screen on human stem cell derived insulin secreting beta cells identifies FGF4, FGF5, FGF8F, FGF19 and FGF21 as protective factors

Chu, C. M. J.; Kamal, S.; Huang, H.; Bayly, C. L.; Chen, S.-y.; Omur, M. E.; Wasserman, W. W.; Lynn, F. C.; Johnson, J. D.

2025-02-16 molecular biology
10.1101/2025.02.12.637990 bioRxiv
Show abstract

Insulin is produced by pancreatic {beta} cells, whose dysfunction and death are hallmarks of diabetes. Stem cell-derived {beta}-like cells (SC{beta} cells) are a potential alternative to cadaveric islets for replacement therapy. However, SC{beta} cells face a multitude of stresses, including hypoxia, hyperglycemia, ER stress, inflammation, and autoimmunity that must be overcome for this treatment to cure diabetes. Pancreatic {beta} cells and SC{beta} cells reside in complex microtissues (islets or spheroids) and can therefore be potentially modulated by hundreds of autocrine and paracrine signals. Here, we leveraged scRNAseq, bulk RNAseq, and bulk proteomics data from late stage SC{beta} cells and human islets to map potential ligand-receptor pairs in these tissues and generate a prioritized list of ligands for high-content SC{beta} cell survival screening. We tracked cell number with Hoechst 33342, cell death using propidium iodide incorporation, and INS activity from INS-EGFP knock-in reporter human embryonic stem cells over several days using high throughput imaging. Members of the fibroblast growth factor (FGF) family significantly prevented cytokine induced cell death, with the top validated hits being FGF4, FGF5, FGF19, FGF21, and FGF8F. These results have the potential to improve SC{beta} cell survival with pathways that are readily targetable and translatable to the clinic.

Matching journals

The top 9 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.