Improving the study of RNA dynamics through advances in RNA-seq with metabolic labeling and nucleotide-recoding chemistry

RNA metabolic labeling using 4-thiouridine (s4U) can be used to capture the dynamics of RNA synthesis and decay. The power of this approach is dependent on appropriate quantification of labeled and unlabeled sequencing reads, which can be compromised by the apparent loss of s4U-labeled reads in a process we refer to as dropout. Here we show that s4U-containing transcripts can be selectively lost when RNA samples are handled under sub-optimal conditions, but that this loss can be minimized using an optimized protocol. We demonstrate a second cause of dropout in nucleotiderecoding and RNA sequencing (NR-seq) experiments that is computational and downstream of library preparation. NR-seq experiments involve chemically converting s4U from a uridine analog to a cytidine analog and the apparent T-to-C mutations are then used to identify the populations of newly synthesized RNA. We show that high levels of T-to-C mutations can prevent read alignment with some computational pipelines, but that this bias can be overcome using improved alignment pipelines. Importantly, kinetic parameter estimates are affected by dropout independent of the NR chemistry employed, and all chemistries are practically indistinguishable in bulk, short-read RNA-seq experiments. Dropout is an avoidable problem that can be identified by including unlabeled controls, and mitigated through improved sample handing and read alignment that together improve the robustness and reproducibiltiy of NR-seq experiments. Overall design: NR-seq performed in HEK 293T cells with three published nucleotide conversion chemistries. RNA collection and data processing were performed with common protocols and those designed to minimize loss of s4U-labelled RNA and T-to-C mutation containing sequencing reads. All experiments performed in duplicate.