Optimization of Circle-Seq improves transcription error profiling and transcriptome representation

Errors in transcription generate transient RNA variants that are amplified through translation, with potential consequences for cellular phenotypes and RNA-based therapeutics. Accurate detection of these rare events relies on high-fidelity sequencing approaches such as Circle-Seq, but technical artifacts introduced during RNA preparation can distort error profiles and frequencies. Here, we systematically evaluate the impact of bacterial RNA sample preparation on Circle-Seq outputs. We show that RNA extraction and ribosomal RNA depletion introduce artifactual oxidative damage signatures, particularly C>U and G>A substitutions, that inflate apparent transcription error rates. Optimized mRNA preparation conditions substantially reduce these artifacts. In addition, we identify an RNase III fragmentation bias caused by its preference for structured RNAs, resulting in uneven transcriptome coverage. Chaotropic monovalent ions mitigate this bias and improve coverage uniformity. Finally, our optimized protocol captured, with improved accuracy, the frequency and spectrum of transcription errors in absence of the transcription fidelity factor GreA. Together, these results define the technical limitations of Circle-Seq and provide practical solutions to improve the accuracy and interpretability of transcription error measurements in bacterial systems.