Failing to account for RNA quantity inflates background and leads to the misleading appearance that PRC2 and GFP bind to RNA in vivo

Abstract

We recently published biochemical and quantitative evidence that challenges the widespread claims that PRC2 binds directly to many RNAs in vivo. A recent preprint performs a re-analysis of some of our data using a different metric and claims to identify contradictory results - that PRC2 binds to more RNAs than the well-characterized PTBP1 and SAF-A RNA-binding proteins. The preprint also reports other counterintuitive observations, including that PRC2, and all proteins examined, bind to RNAs that do not exist within the same cell and therefore could not interact in vivo nor could they be UV-crosslinked. Because these observations are incongruent with orthogonal, non-PCR based, measurements as well as fundamental RNA biology, we explored why their analysis would lead to these unexpected observations. After carefully examining the methods used and regenerating the reported results, we found a very simple explanation for these discrepancies: the authors effectively ignore the vast majority of RNAs in the sample. Here we show that this failure to properly account for the total RNA in each sample leads to inaccurate estimates of the true proportions of each RNA in the sample and therefore the enrichment calculated. We find that the number of RNA reads discarded is dependent on the RNA binding properties of the protein and therefore dramatically inflates background signals (e.g., RNA detected with GFP) while deflating specific signals (e.g., RNA detected with PTBP1). Importantly, we show that this issue alone explains the discrepancy between their observations and our previously reported results. The fact that PRC2 binding can only be observed when analyzed in this way further reinforces the fundamental point of our original paper: the existing evidence in support of direct PRC2-RNA binding requires critical re-evaluation.