Predictive modeling of long non-coding RNA chromatin (dis-)association

Long non-coding RNAs (lncRNAs) are involved in gene expression regulation in cis and trans. Although enriched in the chromatin cell fraction, to what degree this defines their broad range of functions remains unclear. In addition, the factors that contribute to lncRNA chromatin tethering, as well as the molecular basis of efficient lncRNA chromatin dissociation and its functional impact on enhancer activity and target gene expression, remain to be resolved. Here, we combine pulse-chase metabolic labeling of nascent RNA with chromatin fractionation and transient transcriptome sequencing to follow nascent RNA transcripts from their co-transcriptional state to their release into the nucleoplasm. By incorporating functional and physical characteristics in machine learning models, we find that parameters like co-transcriptional splicing contributes to efficient lncRNA chromatin dissociation. Intriguingly, lncRNAs transcribed from enhancer-like regions display reduced chromatin retention, suggesting that, in addition to splicing, lncRNA chromatin dissociation may contribute to enhancer activity and target gene expression.