Analysis of ∼700 Human Tissue Samples Identifies microRNAs That Are Associated with, Predictive of, and Necessary for Insulin Gene Transcription

There is growing evidence that microRNAs, a class of non-coding RNA molecules, are associated with regulation of gene expression in general. Here we profiled microRNAs and key endocrine pancreatic gene transcripts in 698 human tissues, including ∼200 human islet preparations. Figure 1 shows the tissue types in our biobank and circos plots showing correlation of key pancreatic genes and microRNAs. Using novel approaches involving penalized regression and bootstrapping in our discovery set of 507 tissue samples, we identified microRNAs that are highly associated with (pro-)insulin expression. Then, using two different de-identified validation sets (N=91 and N=100) of human cells/tissue samples, we demonstrate that the levels of insulin-associated microRNAs can predict the level of insulin transcript with high accuracy. Finally, we show that forced expression of insulin gene does not change the levels of insulin-associated microRNAs, but forced expression of selected microRNAs promotes insulin gene expression. More generally, these data identify key microRNA genes associated with, predictive of and promoting for insulin gene transcription.