Digital Reprogramming Decodes Epigenetic Barriers of Cell Fate Changes

Fates of differentiated cells in our body can be induced to change by nuclear reprogramming. In this way, cells valuable for therapeutic purposes and disease modeling can be produced. However, the efficiency of this process is low, partly due to the properties of somatic donor nuclei which stabilize their differentiated fate but also act as barriers reprogramming-associated cell fate changes. The identity of these reprogramming barriers is not fully understood. Here, we developed an artificial intelligence-based approach to model nuclear reprogramming and used it to identify the chromatin modification H3K27ac as a novel epigenetic barrier to reprogramming-induced cell fate changes. Using reprogramming by nuclear transfer (NT) to eggs of Xenopus laevis as a model system, we profiled chromatin modifications in different cell types alongside gene expression patterns before and after reprogramming. Our model integrated the data to accurately predict reprogramming outcomes on a transcriptome level. By leveraging model predictions, we find that genes resisting inactivation during reprogramming display specific chromatin modification barcodes, including the known reprogramming barrier H3K4me3 alongside a novel candidate barrier, H3K27ac. Accordingly, reducing H3K27ac levels using p300/CBP inhibitors before reprogramming led to an improved downregulation of genes linked to H3K27ac-modified enhancers after reprogramming. Importantly, these effects were accompanied by improved embryonic development of the resulting nuclear transfer embryos. In summary, our study developed Digital Reprogramming, an artificial intelligence approach capable of predicting resistance to cell-fate reprogramming, which led to the identification of H3K27ac as a critical barrier to reprogramming-associated cell-fate changes. Keywords: machine learning, predictive modeling, SCNT, nuclear reprogramming, H3K27ac, p300/CBP, epigenetic memory, cell fate stability.