Large-scale detection of in vivo transcription errors

Abstract

Accurate transmission and expression of genetic information are crucial for the survival of all living organisms. Recently, the coupling of mutation accumulation experiments and next-generation sequencing has greatly expanded our knowledge of the genomic mutation rate in both prokaryotes and eukaryotes. However, because of their transient nature, transcription errors have proven extremely difficult to quantify, and current estimates of transcription fidelity are derived from artificial constructs applied to just a few organisms. Here we report a unique cDNA library preparation technique that allows error detection in natural transcripts at the transcriptome-wide level. Application of this method to the model organism Caenorhabditis elegans revealed a base misincorporation rate in mRNAs of ~4 × 10(-6) per site, with a very biased molecular spectrum. Because the proposed method is readily applicable to other organisms, this innovation provides unique opportunities for studying the incidence of transcription errors across the tree of life.

Keywords: C. elegans; RNA polymerase fidelity; base substitution; evolution.

Fig. 1.

Overview of the method. (A) Fragmented mRNAs are tagged by attaching random 8-mers (bar codes) to their 5′ ends. In this example, one of the four mRNAs contains a transcription error (red star). (B) The tagged RNA fragments are attached to a bead and reverse transcribed. An error introduced by the reverse transcriptase is represented with a green star in one of the cDNAs. (C) The newly generated cDNAs are washed away. (D) After repeating the steps in B and C two more times and sequencing the cDNAs produced after each of the three rounds of reverse transcription, the RNA-seq reads are aligned to the genome and grouped into families according to the combined information of their bar code and breakpoint mapping position. The transcription error (red star, leftmost family) is shared among all members of the family and therefore is easily distinguished from the occasional RT (green star) and sequencing (blue stars) errors.

Fig. 2.

Molecular spectrum of transcription errors. The values shown in this graph are conditional error rates (i.e., A → U gives the probability of a U to be inserted at a position where an error-free mRNA should contain an A) for all (n = 96) transcription errors detected. Error bars represent the 95% confidence interval of the error rate.

Fig. 3.

Comparison of the genomic and transcription base substitution spectra. The genomic base substitution data are from ref. . The total numbers of base substitutions are n = 448 for genomic mutations and n = 96 for transcription errors. Because genomic mutations are not polarized (an A-to-G mutation is equivalent to a T-to-C mutation), transcription errors are merged into groups of complementary mutations to compare them to genomic mutations. This graph shows, for all possible types of transcription and genomic base substitutions, the fraction of base substitutions of a given type. For example, the blue bar A:T → G:C shows that ∼24% of all transcription errors are A → G or T (U) → C errors.