Widespread DNA off-targeting confounds studies of RNA chromatin occupancy

The importance of long noncoding RNA (lncRNA) functions is recognized across biological systems, but their modes of action remain poorly understood. One mechanism proposed to be particularly common is gene expression regulation via recruitment to specific genomic regions. Several high-throughput sequencing methods were developed for studying the genome-wide chromatin occupancy of lncRNAs, including ChIRP-seq, CHART-seq, and RAP-seq. These methods utilize biotin-labeled probes targeting the RNA of interest to isolate and recover the chromatin associated with it. Surprisingly, many of the datasets obtained with these methods contain many thousands of binding sites, in apparent contradiction with the low abundance of the interrogated lncRNAs. We studied the chromatin interactome of NESPR lncRNA in cells with varying levels of endogenous expression, and then performed a meta-analysis using dozens of RNA-chromatin interaction datasets in human and mouse cells. We show that thousands of regions reported to bind lncRNAs in the published studies most likely arise from spurious recovery of DNA elements with partial complementarity of the ends of the recovered DNA fragments with the probes used for the pulldown. In addition, studies profiling RNA-chromatin interactions often lack crucial controls and replicates. Therefore, the vast majority of chromatin regions reported as bound by trans-acting RNAs in recent studies in mammalian cells appear to be technical artifacts. We provide suggestions for bioinformatic workflows for assessing RNA-chromatin dataset quality. ChIRP-seq of the long non-coding RNA NESPR in IMR32, RNase-treated IMR32, and SHEP neuroblastoma cells with a NESPR-targeting probe set, a LacZ-targeting probe set.