Background

Transcriptional errors due to ribonucleotide misincorporation are ubiquitous to all living organisms (Carey, 2015). Given that each messenger RNA (mRNA) can be translated 2-4 thousand times (Schwanhausser et al., 2011) and many special RNAs are expressed only once per cell at a given time (Islam et al., 2011; Pelechano et al., 2010), even a single transcription error at a critical residue can make large differences in a specific protein’s expression. In addition, transcriptional errors can accelerate protein aggregation leading to age-related diseases in humans (van Leeuwen et al., 1998). While transcription errors are conventionally held to have a random distribution across the genome, there is evidence indicating that transcription errors could be enriched at certain structural motifs and specific genomic regions (Imashimizu et al., 2013; van Leeuwen et al., 1998). These Transcription Error-Enriched genomic Loci (TEELs) have notable biological significance in various diseases such as Down syndrome and Alzheimer’s and are gaining attention in genetics research (Burns et al., 2010; Saxowsky et al., 2008). Unfortunately, there are major challenges that must be circumvented for the study of transcriptional error due to RNA polymerase unconfounded by RNA-editing processes such as those from post-transcriptional modifications. This requires purification of nascent RNA coupled with a highly accurate RNA sequencing method that can identify TEELs and elucidate transcriptional regulation and dysregulation contributing to transcriptional errors with implications to diseases.

There are several complications which impede the accurate detection of de novo transcription errors. The first challenge is eliminating the noise from post-transcriptional modifications, which requires the purification of nascent RNA freshly made by RNA polymerases. Hence, current RNA sequencing (RNA-seq) studies on transcriptional errors often overlook this requirement and therefore overestimate transcription error rates. The second challenge is rectifying the systematic noise from Next Generation Sequencing (NGS). NGS on average misreads approximately one base in every 1,000 (Minoche et al., 2011), and this is further compounded by the fact that reverse transcriptase (required for generating cDNA for NGS) misincorporates one base in every 10,000 (Ji and Loeb, 1992). The third challenge is computationally discerning TEELs from background noise. Even with accurate sequencing data, it is still difficult to computationally identify TEELs amongst background errors which are stochastically introduced by RNA polymerases (de Mercoyrol et al., 1992). Here, we present our background Error Model-coupled Precision nuclear run-on Circular-sequencing (EmPC-seq) method (Figure 1) to overcome these three main challenges. EmPC-seq consists of three core components: (1) a nuclear run-on assay to capture nascent RNA before post-transcriptional modifications (Mahat et al., 2016), (2) a circular-resequencing step that generates cDNA via rolling-cycle reverse transcription of circularized nascent RNA molecules (Acevedo and Andino, 2014) to improve sequencing accuracy by generating tandem cDNA repeats of the same circularized RNA molecule by rolling circle amplification so that the RNA molecule can be sequenced multiple times. (3) We also developed a background error model algorithmic analysis to remove stochastic background noise by simulating de novo sequencing data and subsequent error to serve as a control group carrying background alterations from sequencing noise, non-uniform sequencing depth, and alignment artifacts (Cheung et al., 2020). EmPC-seq aims to detect nascent transcriptional errors and elucidate their origins that may have implications to diseases.

Figure 1. Schematic of EmPC-seq. Real transcription errors are represented using orange dots. Dots in other colors represent systematic noise, including enzymatic errors and sequencing errors. (Step 1) Yeast cell is permeabilized. (Step 2) In vivo transcription is halted by adding all 4 kinds of biotinylated NTPs during the Nuclear Run-on assay. (Step 3) Yeast total RNA is extracted and purified via ethanol precipitation. (Step 4) RNA is fragmented with base hydrolysis into short (60-100nt) RNAs. (Step 5) Biotin-labeled nascent RNA is enriched through Streptavidin bead purification. (Step 6) Re-purified nascent RNAs are circularized by RNA ligase and processed into tandem copy cDNAs through rolling circle reverse transcription. (Step 7) Library DNA is prepared with a kit and then submitted for Next Generation Sequencing. (Step 8) Transcription errors are accurately detected by combining consensus sequence results with our background error model. This schematic is adapted from Cheung et al. (2020).