Lost in transcription: transient errors in information transfer (2024)

The publisher's final edited version of this article is available at Curr Opin Microbiol

High-resolution sequencing (HiRes-seq)

Imashimizu et al. [13] describe a high-resolution Illumina RNA-seq method that provides the first evidence that Escherichia coli transcription error rates can be assessed while taking into full consideration the generation of errors that occur during the process of converting RNA into DNA, the amplification of cDNA and subsequent NGS steps. RNA samples for HiRes-seq were generated by E. coli RNA polymerase (RNAP) transcription of the rpoBC operon on a multi-copy plasmid; RNA was produced either in vitro, with purified components, or in vivo in wild-type E. coli. Therefore, a target sequence approach was used akin to selection studies in the analysis of DNA mutation. The cDNA library preparation for NGS followed routine methodology including the addition of barcodes [14] that enable PCR and sequencing errors to be discounted during analysis. The HiRes-seq strategy is based on two key assumptions: by using high-fidelity reverse transcriptase (RT) and DNA polymerase (DNAP) the combined error rate can be reduced to ~10⁻⁵ b⁻¹; RNAP will generate errors with specificities different from those produced by RT and DNAP.

Statistically reliable detection of epimutations were observed through the uniquely mapped sequence reads, with an average 3×10⁶ read depth. Changes in base transition error rates in the range of 4×10⁻⁵ to 3×10⁻⁴ b⁻¹ were directly detected in vitro and in vivo. Importantly, it was shown that the presence of both GreA and GreB fidelity factors [15, 16] increased the level of fidelity of transcription in vitro to the level of fidelity observed in vivo. However, the level of resolution obtained in this study precluded analysis of transversion errors which occur at a lower level hindering detection by HiRes-Seq and frameshift errors were not analyzed to avoid read misalignments during bioinformatic analysis.

Circular sequencing (CirSeq)

To overcome the limitations of NGS error, Acevedo et al. [17••] developed a clever cDNA library preparation technique: RNA is fragmented and circularized to generate templates for rolling-circle reverse transcription, which yields cDNA arrays of tandemly repeated copies that then serve as substrates for NGS (Figure 2a). Epimutations originally present in the mRNA will be shared by all the repeats; differences between the linked repeats are enzymatic or sequencing errors and excluded from analysis. The redundant copies within each read are aligned to derive a consensus sequence of their initial RNA template. This approach reduces errors associated with Illumina NGS by up to 8 orders of magnitude, from 10⁻⁴ to 10⁻¹² per base, far below the estimated epimutation rates (10⁻⁴ to 10⁻⁶), facilitating rare variant detection and accurate measurement of low-frequency variants.

Open in a separate window

Figure 2

Novel cDNA library preparation for analysis of epimutation via next-generation sequencing (NGS). Red circles denote epimutation; other colored circles denote reverse transcription (RT), PCR and NGS errors. The majority of mRNA fragments would not contain transcription errors. (a) Circular sequencing (CirSeq) [17, 18]. Full-length mRNAs are processed into short (85–100 nt) circular RNAs (black circle). Rolling-circle RT yields tandem cDNA copies of the circular RNA template (each repeat is denoted by green, red and blue lines in the cDNA) that are cloned to generate a library of dsDNA molecules containing the tandem-repeat sequence including Illumina adapters for NGS. NGS reads are computationally processed to identify and align tandem repeats within each read. A consensus of the aligned reads reveals epimutation events. (b) Replicated sequencing (Rep-seq) [19]. Fragmented mRNAs, tagged by barcodes, are attached to a bead (black circle) and reverse transcribed. The newly generated cDNAs are washed away and the RT step is repeated (denoted by green, red and blue lines). The cDNAs produced after each of three rounds of RT are subjected to Illumina NGS, the RNA-seq reads are aligned to the genome and grouped into families. The epimutation is shared among all members of the family and is readily distinguished from processing errors.

CirSeq was initially used to characterize RNA variants in a poliovirus population [17••] but is well suited to the analysis of epimutation in mRNA transcripts [18]. In poliovirus it was found that mutation rates vary by more than 100-fold depending on mutation type, with transitions averaging 2.5×10⁻⁵ to 2.6×10⁻⁴ substitutions per site and transversions averaging 1.2×10⁻⁶ to 1.5×10⁻⁵ substitutions per site; within each type of base substitution, 10-fold differences between mutation sites were observed possibly reflecting the molecular mechanism of viral polymerase fidelity. Only consensus sequences devoid of indels were used for analyses to avoid artefacts.

Replicated sequencing (Rep-seq)

To overcome the limitations of NGS error, Gout et al. [19••] have developed a novel cDNA library preparation technique: individual mRNAs are barcoded and reverse transcribed multiple times; mutational changes in common to all cDNAs originating from a particular barcoded RNA represent epimutations (Figure 2b). This approach allows epimutation detection in natural transcripts at the transcriptome-wide level.

Rep-seq was used in wild-type, RNA editing-deficient, and nonsense-mediated mRNA decay-deficient strains of Caenorhabditis elegans yielding base substitution transcription error rates of 2.2×10⁻⁶, 3.3×10⁻⁶ and 5.2×10⁻⁶ per site, respectively; these three rates were judged not to be significantly different from each other, therefore Rep-seq ultimately revealed a base misincorporation rate in mRNAs of 4×10⁻⁶ per site. The most common type of transcription error is a C to U base substitution and transitions occur more frequently than transversion epimutation events, as has been found for spontaneous mutation [9, 11], therefore RNA polymerase base misincorporations appear to resemble DNA polymerase base misincorporations. A total of 26 indels were observed (18 insertions and 8 deletions), yielding a transcription indel rate of 1.2×10⁻⁶. These ±1 frameshifts tended to occur in hom*opolymeric nucleotide runs, again reminiscent of frameshift mutations.

An important advantage of Rep-seq is the identification of all types of errors, including base substitutions and frameshifts/indels, so the full molecular spectrum of epimutation is obtained. Moreover, Gout et al. provide the exact number of epimutations identified and a precise description of the type and position of each epimutation characterized [19••]. Rep-seq is readily applicable to other organisms, allowing study of epimutation throughout ‘life’s enormously arborescent bush’ [20].

Reverse epimutation

Recent reverse epimutation (partial phenotypic suppression or ‘leakiness’) systems in E. coli [21, 22•, 23•] and Saccharomyces cerevisiae have been described [24]. In E. coli, Zhou et al. constructed an A₉ run that is out of frame (−1) relative to a downstream lacZ gene on the chromosome to examine transcriptional slippage during elongation [22•]. An increase in slippage epimutations [+1(±3)_n base frameshifts] at the A₉ run will produce more in-frame lacZ transcripts that can be monitored by an increase in β-galactosidase levels. While this study is conceptually similar to previous frameshift reverse epimutation studies [25], the genetic system that was developed allowed, for the first time, the isolation and characterization of E. coli RNAP mutants with altered transcriptional slippage in vivo. Zhou et al. isolated six RNAP mutants, all rifampicin-sensitive, that exhibited an increased slippage phenotype both in vivo and in vitro. These amino acid substitutions were located in the β subunit of RNAP near where the RNA strand of the RNA-DNA hybrid is positioned in the elongation complex. Interestingly, this is also the binding site of the RNA inhibitor rifampicin. Analysis of a collection of RNAP mutants resistant to rifampicin (Rif^R) yielded 14 mutants affecting ten positions in the β subunit that altered elongation slippage compared with wild-type; eleven Rif^R mutants were observed to increase elongation slippage, and three Rif^R mutants were observed to decrease slippage [22•].

It appears that monotonous runs of adenines in DNA are universal hotspots for RNA polymerase slippage events during transcription. Adenine runs have been found to be slippery in vitro [26], in model systems in E. coli [3, 21, 22•, 25] and yeast [21, 24] and in genome wide transcriptome analysis of epimutation in C. elegans [19••]. In humans and canines suffering genetic disease due to constitutional frameshift mutations, transcription frameshift errors at A runs have been shown to provide partial function from a mutant gene [27–30].

A base substitution reverse epimutation approach that monitored the leakiness of a lacZ_amber mutation in a mutT strain of E. coli was used to assess how oxidative damage can impact the fidelity of RNA transcription [31]. The MutT protein removes 8-oxo-deoxyguanosine triphosphate (8-oxo-dGTP) and 8-oxo-guanosine triphosphate (8-oxo-GTP) from the nucleotide pools precluding incorporation into DNA and RNA, respectively. It was proposed that the absence of MutT can increase readthrough of the stop codon via 8-oxo-GTP incorporation into the nascent mRNA transcript and account for a reported 30-fold increase in β-galactosidase levels in mutT cultures compared to wild-type cultures [31]. However, both permanent DNA mutations and transient mRNA misincorporations at this codon contribute to the observed β-galactosidase levels and it has recently been shown that Lac⁺ DNA revertants can completely account for the increase in β-galactosidase levels in mutT lacZ_amber cultures, without invoking participation of transient transcription errors. It was therefore concluded that the absence of MutT produces a DNA mutator but does not equally create an RNA mutator [32•].

A recent study in E. coli [23•], monitored the partial phenotypic suppression of a different lacZ_amber mutation, and is conceptually similar to a previous base substitution reverse epimutation screen used to identify RNAP infidelity mutants in E. coli [33]. To study transcription fidelity under oxidative stress Inokuchi et al. also used a ΔmutT background in their lacZ_amber screening system [23•]. While the absence of MutT function produced a 1.8-fold increase in β-galactosidase levels over wild-type, mutations in other genes increased β-galactosidase levels over the ΔmutT background by up to 5-fold. Subsequent analysis revealed that specific amino acid changes in guanylate kinase and in the β (a Rif^R mutant was isolated) and β‘ (one mutation was located in the trigger loop which plays a role in nucleotide binding and elongation) subunits of RNA polymerase cause elevated levels of phenotypic suppression, specifically under aerobic conditions. Other proteins that appear to be involved in preventing phenotypic suppression were also identified including DnaB, DnaN and MsbA, which are involved in DNA replication and in preserving membrane structure [23•]. However, in light of the recent study that demonstrated the confounding influence of Lac⁺ revertants arising in the Lac⁻ ΔmutT population [32•], it will be of interest to resolve how much of the increased β-galactosidase levels are due to increased DNA mutation rather than due to RNA epimutation.

Forward epimutation

The one forward epimutation system described to date [34] monitors error events during transcription of the lacI gene which encodes the lac repressor (Figure 3a). The key point is that a lacI epimutation inactivates the lac repressor, thereby allowing activation of a downstream circuit that provides the system readout. The lac operon comprises an autocatalytic positive feedback loop allowing a heritable all-or-none epigenetic switch at a maintenance concentration of inducer (that concentration of inducer which does not activate transcription of the operon but allows an already induced cell to remain induced [34–36]; Figure 3b,c). The lac repressor is rare (~5 tetramers per cell) [37]. A transient depletion of repressor within a cell will lead to a transient derepression of the operon, producing a burst of lacY permease gene expression [38]. At the maintenance concentration of the inducer thio-methylgalactoside (TMG), the presence of permease will activate the positive feedback loop, so that the new induced state will be heritably maintained through cell division in a clonal cell population [3, 34] (Figure 3b). This classic lac memory-module was harnessed to capture and monitor the consequences of transient transcription errors in E. coli, providing a forward epimutation approach to study proteins and processes involved in modulating RNA fidelity [3, 32•, 34].

Open in a separate window

Figure 3

Heritable stochastic switching in the E. coli lac operon: a forward epimutation system. (a) Under maintenance inducer (mTMG) conditions, the lac operon is OFF when the lac repressor is bound to the lac operator (indicated by the solid red line) and the inducer TMG remains extracellular; stochastic events that lead to a transient derepression of the lac operon will result in a burst of lac operon functions and will initiate an autocatalytic positive-feedback response (indicated by solid blue lines), which will heritably maintain the ON state (TMG induces an allosteric transition in lac repressor, indicated by the dashed red line, so that it no longer binds to the lac operator), and the cell will exhibit green fluorescence [3, 34]. (b) Single E. coli cells in mTMG were grown into microcolonies in a microfluidic flow chamber. Comparison of bright field images (panel series on the left) with GFP fluorescence images (panel series on the right) allows clear distinction between OFF and ON cells in the microcolony. Top panels show mirocolonies that arose from single OFF and single ON cells that were grown in mTMG (ON single cells generate ON microcolonies; OFF single cells generate OFF microcolonies); bottom panels show mirocolonies that arose from single OFF cells that were grown in mTMG (within a growing OFF microcolony, stochastic events can turn a cell ON and this ON state is maintained in the cell lineage) [3]. (c) Wild-type cells that were originally ON (open circles) or OFF (closed circles) were sub-cultured and grown in media containing various concentrations of TMG. The shaded area highlights the maintenance concentration of 6 µM TMG for these strains; hysteresis and bistability in this system is observed [3, 34]. (d) Stochastic switching in the lac bistable gene network is increased when fidelity of transcription is decreased. When OFF cells are grown in the presence of mTMG, the absence of GreA and GreB (blue flow cytometry histograms) increases the proportion of ON cells with respect to wildtype cell levels (red flow cytometry histograms). Each blue and red line represents an independent histogram (20 independent cell populations are shown) representing the interrogation of 10⁴ cells; wild-type histograms are super-imposed over ΔgreAB histograms. The increase in stochastic switch frequency is 38-fold over wild-type level [3].

Using single-cell analysis, it was shown that the frequency of epigenetic switching from the OFF state to the ON state of the lac operon is increased when the fidelity of RNA transcription is decreased due to an error-prone RNA polymerase [34], an error-prone transcription sequence (an A₉ slippery run) [3] or to the absence of auxiliary RNA fidelity factors GreA and GreB (functional analogues of eukaryotic TFIIS) [3, 34]; see Figure 3d. This system was also used to show that the absence of MutT function does not create an epimutator phenotype [32•], which had previously been suggested [31]. Therefore, transcription infidelity contributes to molecular noise (fluctuation in protein numbers) and can effect heritable phenotypic change in genetically identical cells in the same environment. The steps in other microbial epigenetic phenotypic switches that are susceptible to molecular noise have already been highlighted [39], and epimutation may play a role in stochastic switching in these systems.

This forward epimutation system is conceptually different from the other epimutation systems so far discussed, in that errors in RNA and DNA can have the same heritable phenotypic consequence. This can be visualized when errors in information transfer are placed in the context of a transcriptional time series [40••] in conjunction with their phenotypic consequences (Figure 4). It has been shown in bacteria, yeast, and mammalian cells that gene expression, and the accompanying noise, occurs in stochastic bursts dominated by the production of mRNAs [38, 41, 42]; it should be noted that rarely made transcripts do not exhibit such burstiness [40••]. In contrast to the examples of chronic transcriptional errors producing a partial phenotype (reverse epimutation systems; Figure 4a), it was shown that one acute transcription error on a poorly transcribed mRNA may promote heritable phenotypic change due to a change of connectivity in a transcription network (Figure 4b). Therefore, one altered transcript within a multi-generational series of error-free transcripts can cause long-term phenotypic consequences. Thus, like DNA mutations, transcriptional epimutations can instigate heritable changes that increase phenotypic diversity.

Open in a separate window

Figure 4

Phenotypic consequences of epimutation in a transcription time series. Each plot shows the time series of mRNA production events (black lines indicate error-free transcripts, i.e. the transcript is an accurate complement of the template strand, be it wild-type or mutant); adapted from So et al. [40••]. Red lines indicate a stochastic error in information transfer (mutation or epimutation). Blue regions indicate the duration in time of the phenotypic consequences produced by the stochastic error event. The top panels show error-free transcription and the phenotypic consequence; the middle panels show transient RNA errors and the phenotypic consequence; the bottom panels show permanent DNA errors and the phenotypic consequence. (a) Constitutive expression of mutant gene. A transient transcription error that corrects the mutation will produce a burst of gene activity that will degrade over time (middle panel). A reversion mutation will produce maximal gene activity in the mutant cell and in all progeny in the mutant lineage (bottom panel). Therefore, RNA and DNA errors have different phenotypic consequences. (b) Lac repressor negatively regulates lacZ expression. Here two independent time series for lacI and lacZ are shown (this depiction does not imply any synchrony in the time series, nor a similar time scale for the two series). Even in the presence of wild-type repressor, the lac operon occasionally will be expressed producing a burst of β-galactosidase activity that will degrade over time (top panel) [46]. Epimutation in the lacI transcript will produce a non-functional repressor and the lac operon will be expressed (i.e., a noisy system will become noisier); in the presence of a maintenance level of inducer, an autocatalytic positive feedback response will initiate and the cell, and all progeny in the lineage, will produce maximal β-galactosidase activity even though error-free lacI transcripts are subsequently produced (middle panel). A lacI forward mutation will produce maximal β-galactosidase activity in the mutant cell and in all progeny in the mutant lineage (bottom panel). Therefore, RNA and DNA errors can have similar phenotypic consequences.

Epimutation and evolvability

The clarity of genetics (DNA) becomes blurry, or messy [43], when considering RNA, since transcription is bursty, error-prone, and the products are transient yet amplified (by translation). RNA transcripts have been likened to the disposable soma of genetic information compared to the heritable germ line, DNA [31]. Are the phenotypic consequences of transient transcription errors subject to natural selection and therefore of evolutionary consequence, or do transcription errors play a role in cell fate and differentiation or alternatively, aberrancy and decline, when gene regulatory networks start to break down (an error crescendo during the final acts of the cell) [44]?

Transcription errors, epimutation, create epigenetic phenotypic heterogeneity in a clonal cell population that can result in altered or aberrant cell behavior. The resulting phenotypic heterogeneity may provide the raw material upon which selection can act with the subsequent evolution of novel cell characteristics during, for example, antibiotic tolerance. It has been proposed that although epimutations are not individually subjected to natural selection, they collectively exert a direct and immediate effect on protein and bacterial fitness, expanding the capability of an isogenic population to face environmental challenges [8, 45]. Epimutation may therefore play a role in shaping protein traits such as expression levels, stability, and tolerance to genetic mutations [45].

The new epimutation systems described can aid in identifying and characterizing novel RNA fidelity factors and (in)fidelity RNAP alleles and assess the phenotypic consequences of epimutation in living cells. With the advent of the new NGS approaches described it will be possible to readily obtain epimutational spectra of the species or strain of choice. Of great interest will be how the epimutation spectra of the new RNA fidelity factors and infidelity RNAP mutants described (and those yet to be discovered) compare with the wild-type E. coli epimutation spectrum. From such studies the mechanisms of epimutation may be addressed. By comparing epimutation rates and spectra across bacterial species, the evolutionary forces acting on the fidelity of transcription itself may be discerned.

​

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Papers of particular interest, published within the period of the review, have been highlighted as:

• of special interest

•• of outstanding interest

1. Crick F. Central dogma of molecular biology. Nature. 1970;227:561–563. [PubMed] [Google Scholar]

2. Li G-W, Xie XS. Central dogma at the single-molecule level in living cells. Nature. 2011;475:308–315. [PMC free article] [PubMed] [Google Scholar]• This review describes powerful new approaches to the study, in real time, of dynamic processes such as transcription-factor binding, transcription and translation in single cells, the facts of gene expression as outlined in the sequence hypothesis of Crick. The central dogma of Li & Xie states that DNA is transcribed into mRNA, and mRNA is translated into protein; in contrast, the central dogma of Crick is a negative statement: the transfer of sequence information in protein to nucleic acid or protein does not exist.

3. Gordon AJE, Satory D, Halliday JA, Herman C. Heritable change caused by transient transcription errors. PLoS Genet. 2013;9:e1003595. [PMC free article] [PubMed] [Google Scholar]

4. Rando OJ, Verstrepen KJ. Timescales of genetic and epigenetic inheritance. Cell. 2007;128:655–668. [PubMed] [Google Scholar]

5. Ninio J. Transient mutators: a semiquantitative analysis of the influence of translation and transcription errors on mutation rates. Genetics. 1991;129:957–962. [PMC free article] [PubMed] [Google Scholar]

6. d'Ari R, Casadesús J. Underground metabolism. Bioessays. 1998;20:181–186. [PubMed] [Google Scholar]

7. Ninio J. Connections between translation, transcription and replication error-rates. Biochimie. 1991;73:1517–1523. [PubMed] [Google Scholar]

8. Meyerovich M, Mamou G, Ben-Yehuda S. Visualizing high error levels during gene expression in living bacterial cells. Proc Natl Acad Sci USA. 2010;107:11543–11548. [PMC free article] [PubMed] [Google Scholar]

9. Schaaper RM, Dunn RL. Spectra of spontaneous mutations in Escherichia coli strains defective in mismatch correction: the nature of in vivo DNA replication errors. Proc Natl Acad Sci USA. 1987;84:6220–6224. [PMC free article] [PubMed] [Google Scholar]

10. Miller JH. Mutational specificity in bacteria. Annu Rev Genet. 1983;17:215–238. [PubMed] [Google Scholar]

11. Halliday JA, Glickman BW. Mechanisms of spontaneous mutation in DNA repair-proficient Escherichia coli. Mutat Res. 1991;250:55–71. [PubMed] [Google Scholar]

12. Lee H, Popodi E, Tang H, Foster PL. Rate and molecular spectrum of spontaneous mutations in the bacterium Escherichia coli as determined by whole-genome sequencing. Proc Natl Acad Sci USA. 2012;109:2774–2783. [PMC free article] [PubMed] [Google Scholar]

13. Imashimizu M, Oshima T, Lubkowska L, Kashlev M. Direct assessment of transcription fidelity by high-resolution RNA sequencing. Nucleic Acids Res. 2013;41:9090–9104. [PMC free article] [PubMed] [Google Scholar]

14. Schmitt MW, Kennedy SR, Salk JJ, Fox EJ, Hiatt JB, Loeb LA. Detection of ultra-rare mutations by next-generation sequencing. Proc Natl Acad Sci USA. 2012;109:14508–14513. [PMC free article] [PubMed] [Google Scholar]

15. Erie DA, Hajiseyedjavadi O, Young MC, Hippel von PH. Multiple RNA polymerase conformations and GreA: control of the fidelity of transcription. Science. 1993;262:867–873. [PubMed] [Google Scholar]

16. Orlova M, Newlands J, Das A, Goldfarb A, Borukhov S. Intrinsic transcript cleavage activity of RNA polymerase. Proc Natl Acad Sci USA. 1995;92:4596–4600. [PMC free article] [PubMed] [Google Scholar]

17. Acevedo A, Brodsky L, Andino R. Mutational and fitness landscapes of an RNA virus revealed through population sequencing. Nature. 2014;505:686–690. [PMC free article] [PubMed] [Google Scholar]• • A novel cDNA library preparation technique for the elucidation of the spectrum of epimutation. The rate and nature of RNA errors in a poliovirus population was studied, but the method can be used to study epimutation in mRNA transcripts in any organism.

18. Acevedo A, Andino R. Library preparation for highly accurate population sequencing of RNA viruses. Nat Protoc. 2014;9:1760–1769. [PMC free article] [PubMed] [Google Scholar]

19. Gout J-F, Thomas WK, Smith Z, Okamoto K, Lynch M. Large-scale detection of in vivo transcription errors. Proc Natl Acad Sci USA. 2013;110:18584–18589. [PMC free article] [PubMed] [Google Scholar]• • A novel cDNA library preparation technique for the elucidation of the spectrum of epimutation. The rate and nature of RNA errors in C. elegans was studied, but the method can be used to study epimutation in mRNA transcripts in any organism.

20. Gould SJ. The evolution of life on the earth. Sci American. 1994;271:84–91. [PubMed] [Google Scholar]

21. Strathern JN, Jin DJ, Court DL, Kashlev M. Isolation and characterization of transcription fidelity mutants. Biochim Biophys Acta. 2012;1819:694–699. [PMC free article] [PubMed] [Google Scholar]

22. Zhou YN, Lubkowska L, Hui M, Court C, Chen S, Court DL, Strathern J, Jin DJ, Kashlev M. Isolation and characterization of RNA polymerase rpoB mutations that alter transcription slippage during elongation in Escherichia coli. J Biol Chem. 2013;288:2700–2710. [PMC free article] [PubMed] [Google Scholar]• Using a frameshift reverse epimutation system, RNA polymerase fidelity alleles were identified and characterized.

23. Inokuchi H, Ito R, Sekiguchi T, Sekiguchi M. Search for proteins required for accurate gene expression under oxidative stress: roles of guanylate kinase and RNA polymerase. J Biol Chem. 2013;288:32952–32962. [PMC free article] [PubMed] [Google Scholar]• Using a base substitution reverse epimutation system in a Δ mutT strain, RNA polymerase fidelity alleles as well as mutations in other genes that prevent phenotypic suppression were identified and characterized.

24. Strathern J, Malagon F, Irvin J, Gotte D, Shafer B, Kireeva M, Lubkowska L, Jin DJ, Kashlev M. The fidelity of transcription: RPB1 (RPO21) mutations that increase transcriptional slippage in S. cerevisiae. J Biol Chem. 2013;288:2689–2699. [PMC free article] [PubMed] [Google Scholar]

25. Wagner LA, Weiss RB, Driscoll R, Dunn DS, Gesteland RF. Transcriptional slippage occurs during elongation at runs of adenine or thymine in Escherichia coli. Nucleic Acids Res. 1990;18:3529–3535. [PMC free article] [PubMed] [Google Scholar]

26. Chamberlin M, Berg P. Deoxyribonucleic acid-directed synthesis of ribonucleic acid by an enzyme from Escherichia coli. Proc Natl Acad Sci USA. 1962;48:81–94. [PMC free article] [PubMed] [Google Scholar]

27. Linton MF, Pierotti V, Young SG. Reading-frame restoration with an apolipoprotein B gene frameshift mutation. Proc Natl Acad Sci USA. 1992;89:11431–11435. [PMC free article] [PubMed] [Google Scholar]

28. Linton MF, Raabe M, Pierotti V, Young SG. Reading-frame restoration by transcriptional slippage at long stretches of adenine residues in mammalian cells. J Biol Chem. 1997;272:14127–14132. [PubMed] [Google Scholar]

29. Young M, Inaba H, Hoyer LW, Higuchi M, Kazazian HH, Jr, Antonarakis SE. Partial correction of a severe molecular defect in hemophilia A, because of errors during expression of the factor VIII gene. Am J Hum Genet. 1997;60:565–573. [PMC free article] [PubMed] [Google Scholar]

30. Benson KF, Person RE, Li F-Q, Williams K, Horwitz M. Paradoxical hom*ozygous expression from heterozygotes and heterozygous expression from hom*ozygotes as a consequence of transcriptional infidelity through a polyadenine tract in the AP3B1 gene responsible for canine cyclic neutropenia. Nucleic Acids Res. 2004;32:6327–6333. [PMC free article] [PubMed] [Google Scholar]

31. Taddei F, Hayakawa H, Bouton M, Cirinesi A, Matic I, Sekiguchi M, Radman M. Counteraction by MutT protein of transcriptional errors caused by oxidative damage. Science. 1997;278:128–130. [PubMed] [Google Scholar]

32. Gordon AJE, Satory D, Wang M, Halliday JA, Golding I, Herman C. Removal of 8-oxo-GTP by MutT hydrolase is not a major contributor to transcriptional fidelity. Nucleic Acids Res. 2014;42:12015–12026. [PMC free article] [PubMed] [Google Scholar]• The absence of MutT function in E. coli results in a DNA mutator but not an RNA epimutator.

33. Blank A, Gallant JA, Burgess RR, Loeb LA. An RNA polymerase mutant with reduced accuracy of chain elongation. Biochemistry. 1986;25:5920–5928. [PubMed] [Google Scholar]

34. Gordon AJE, Halliday JA, Blankschien MD, Burns PA, Yatagai F, Herman C. Transcriptional infidelity promotes heritable phenotypic change in a bistable gene network. PloS Biol. 2009;7:e44. [PMC free article] [PubMed] [Google Scholar]

35. Novick A, Weiner M. Enzyme induction as an all-or-none phenomenon. Proc Natl Acad Sci USA. 1957;43:553–566. [PMC free article] [PubMed] [Google Scholar]

36. Ozbudak EM, Thattai M, Lim HN, Shraiman BI, van Oudenaarden A. Multistability in the lactose utilization network of Escherichia coli. Nature. 2004;427:737–740. [PubMed] [Google Scholar]

37. Gilbert W, Müller-Hill B. Isolation of the lac repressor. Proc Natl Acad Sci USA. 1966;56:1891–1898. [PMC free article] [PubMed] [Google Scholar]

38. Cai L, Friedman N, Xie XS. Stochastic protein expression in individual cells at the single molecule level. Nature. 2006;440:358–362. [PubMed] [Google Scholar]

39. Satory D, Gordon AJ, Halliday JA, Herman C. Epigenetic switches: can infidelity govern fate in microbes? Curr Opin Microbiol. 2011;14:212–217. [PubMed] [Google Scholar]

40. So L-H, Ghosh A, Zong C, Sepúlveda LA, Segev R, Golding I. General properties of transcriptional time series in Escherichia coli. Nat Genet. 2011;43:554–560. [PMC free article] [PubMed] [Google Scholar]• • What does it mean for a gene to be ON? The activity of a gene is described by a discrete time series of mRNA production events and this transcriptional time series is stochastic rather than deterministic. By using single-molecule mRNA fluorescence in situ hybridization in individual living E. coli cells, transcription is shown to be bursty, and the fundamentals of a transcription time series are explored.

41. Newman JRS, Ghaemmaghami S, Ihmels J, Breslow DK, Noble M, DeRisi JL, Weissman JS. Single-cell proteomic analysis of S. cerevisiae reveals the architecture of biological noise. Nature. 2006;441:840–846. [PubMed] [Google Scholar]

42. Raj A, Peskin CS, Tranchina D, Vargas DY, Tyagi S. Stochastic mRNA synthesis in mammalian cells. PloS Biol. 2006;4:e309. [PMC free article] [PubMed] [Google Scholar]

43. Tawfik DS. Messy biology and the origins of evolutionary innovations. Nat Chem Biol. 2010;6:692–696. [PubMed] [Google Scholar]

44. Orgel LE. The maintenance of the accuracy of protein synthesis and its relevance to ageing. Proc Natl Acad Sci USA. 1963;49:517–521. [PMC free article] [PubMed] [Google Scholar]

45. Goldsmith M, Tawfik DS. Potential role of phenotypic mutations in the evolution of protein expression and stability. Proc Natl Acad Sci USA. 2009;106:6197–6202. [PMC free article] [PubMed] [Google Scholar]

46. Choi PJ, Cai L, Frieda K, Xie XS. A stochastic single-molecule event triggers phenotype switching of a bacterial cell. Science. 2008;322:442–446. [PMC free article] [PubMed] [Google Scholar]