Spontaneous deamination of cytosine to uracil is biased to the non-transcribed DNA strand in yeast

Depósito de investigación de la Universidad de Sevilla https://idus.us.es/ “This is an Accepted Manuscript of an article published by Elsevier in DNA Repair on June 2023, available at: https://doi.org/10.1016/j.dnarep.2023.103489 .” Spontaneous deamination of cytosine to uracil is biased to the non-transcribed DNA strand in yeast Jonathan D Williamsa,1, Demi Zhua, María García-Rubiob, Samantha Shaltza, Andrés Aguilerab, Sue Jinks-Robertsona,* aDepartment of Molecular Genetics and Microbiology, 213 Research Dr., Duke University Medical Center, Durham, NC 27710 bCentro Andaluz de Biología Molecular y Medicina Regenerativa (CABIMER), Universidad de Sevilla-CSIC, Seville, Spain Abstract Transcription in Saccharomyces cerevisiae is associated with elevated mutation and this partially reflects enhanced damage of the corresponding DNA. Spontaneous deamination of cytosine to uracil leads to CG>TA mutations that provide a strand-specific read-out of damage in strains that lack the ability to remove uracil from DNA. Using the CAN1 forward mutation reporter, we found that C>T and G>A mutations, which reflect deamination of the non-transcribed and transcribed DNA strands, respectively, occurred at similar rates under low-transcription conditions. By contrast, the rate of C>T mutations was 3-fold higher than G>A mutations under hightranscription conditions, demonstrating biased deamination of the non-transcribed strand (NTS). The NTS is transiently single-stranded within the ~15 bp transcription bubble, or a more extensive region of the NTS can be exposed as part of an R-loop that can form behind RNA polymerase. Neither the deletion of genes whose products restrain R-loop formation nor the over-expression of RNase H1, which degrades R-loops, reduced the biased deamination of the NTS, and no transcription-associated R-loop formation at CAN1 was detected. These results suggest that the NTS within the transcription bubble is a target for spontaneous deamination and likely other types of DNA damage. Keywords transcription; mutagenesis; RNA polymerase; R-loops; transcription bubble; deamination 1. INTRODUCTION Mutations generally have negative consequences but are required for adaptation under stress conditions and for evolutionary change. Most mutations arise during genome duplication and reflect either rare errors of the high-fidelity replicative DNA polymerases or the bypass of DNA damage by error-prone translesion-synthesis DNA polymerases. In addition to *Communicating author: Department of Molecular Genetics and Microbiology, 213 Research Dr., Duke University Medical Center, Durham, NC 27710, Phone: 919-681-7273, [email protected]. 1Current address: Labcorp, Burlington, NC HHS Public Access Author manuscript DNA Repair (Amst) . Author manuscript; available in PMC 2023 September 11. Published in final edited form as: DNA Repair (Amst) . 2023 June ; 126: 103489. doi:10.1016/j.dnarep.2023.103489. Author Manuscript Author Manuscript Author Manuscript Author Manuscript replication-associated changes in DNA sequence, mutations also arise in the context of transcription. This may be a particularly important source of genetic alterations in the post-mitotic cells of metazoans, such as those in neural tissue [1]. In microorganisms, transcription-associated mutagenesis (TAM) is operationally defined as an increase in the mutation rate of a target gene when it is transcribed at a high versus low level (reviewed in [2]). Early studies in Saccharomyces cerevisiae demonstrated that TAM increases when excision-repair or error-free damage-bypass pathways are inactivated, and decreases when translesion synthesis is disrupted [3]. There also is synergism between the stimulatory effects of exogenous mutagens and transcription on recombination rates [4]. Collectively, these genetic data indicate that transcription increases the exposure of the underlying DNA template to endogenous and exogenous damage. Mutagenic damage can also be inflicted by topoisomerase I (Top1), which resolves the torsional stress that accumulates when DNA strands are separated during transcription [5, 6]. Top1-dependent mutagenesis specifically generates small deletions and reflects sequential cleavage of one DNA strand, with the initial incision occurring at an embedded ribonucleotide [7, 8]. Both replication and transcription impart an asymmetry between the two strands of duplex DNA. During yeast replication, complementary strands are copied by distinct DNA polymerases and the efficiencies of replication-error removal differ, which can lead to mutation asymmetries on the leading and lagging strands (reviewed in [9]). During transcription, only one strand typically serves as the template for RNA synthesis. The template strand is referred to as the transcribed strand or TS, while the complementary strand is the non-transcribed strand or NTS. Strand-related biases are readily evident when mutations are reported with reference to only one DNA strand and both replication- and transcription-associated mutation biases have been reported in cancer genomes [10]. Strand-related biases associated with transcription can reflect more robust repair of the TS than NTS and/or more damage to the NTS than TS. A lesion on the TS that blocks/ stalls RNA polymerase, for example, triggers more efficient removal by the nucleotide excision repair pathway than does the same lesion on the NTS (transcription-coupled repair; reviewed in [11]). In addition, single-stranded DNA is more susceptible to chemical and enzymatic damage than is double-stranded DNA (reviewed in [12, 13]). More damage of the NTS relative to the TS presumably reflects its enhanced single-stranded character. During transcription, 10–15 nt of the NTS are exposed within the RNA polymerase (RNAP) transcription bubble, whereas the TS is protected through its pairing with the nascent transcript. A much larger segment of the NTS can be exposed when the RNA threads back after exiting RNAP and pairs with the template strand [14]. The extended RNA-DNA hybrid, together with the single-stranded NTS, is referred to as an R-loop. R-loops are important for transcription termination, but conflicts with the replisome can generate DSBs that are potent initiators of homologous recombination and genetic instability (reviewed in [15]). The contribution of R-loops to the accumulation of spontaneous DNA damage is less clear. While transcription-associated genetic instability is generally problematic, it is physiologically important in the maturation of the vertebrate immune system. Antigenstimulated somatic hypermutation (SHM) improves antibody affinity, while class-switch Williams et al. Page 2 DNA Repair (Amst) . Author manuscript; available in PMC 2023 September 11. Author Manuscript Author Manuscript Author Manuscript Author Manuscript recombination (CSR) diversifies antibody effector functions (reviewed in [16]). Both processes are limited to B cells and require enzymatic deamination of cytosine to uracil by the single-strand specific enzyme AID (activation-induced cytosine deaminase). In vitro , AID travels with phage RNAP [17] and primarily deaminates the NTS, leading to the suggestion that it also targets the transcription bubble. A more recent study with eukaryotic RNAPII, however, demonstrated equal deamination of the two DNA strands by AID and it was suggested that both become accessible in a backtracked transcription bubble [18]. By contrast, other reports have shown that deamination during CSR involves R-loop formation [19]. AID expression in yeast was reported to have an NTS bias in an RNAPII-transcribed gene, but only in a mutant background where R-loops were expected to accumulate at a high level [20]. The metazoan-specific APOBEC (apolipoprotein B mRNA editing enzyme, catalytic peptide) family of cytosine deaminases target viral genomes and are an important component of innate immunity [21]. APOBECs also robustly deaminate single-stranded genomic DNA, however, and are potent drivers of tumorigenesis [22]. In yeast, APOBEC3G primarily targets the NTS of tRNA genes [23], which are strong sites of R-loop accumulation [24]. The NTS bias in tRNA genes was exacerbated by loss of RNase H, which degrades the RNA component of RNA-DNA hybrids, consistent with a primary targeting of single-stranded DNA in the context of R-loops. An APOBEC from lamprey also targeted the NTS in a genome-wide analysis in yeast, but with preferential deamination of transcription start sites that contain stable pre-initiation complexes [25]. While the single-stranded DNA of R-loops can be clearly targeted by cytosine deaminases, a role for R-loops in spontaneous deamination has not been reported. A primary role for the transcription bubble in NTS-biased cytosine deamination, however, was previously inferred in bacterial cells [26], where co-transcriptional translation is assumed to limit R-loop formation. The rapid assembly of yeast transcripts into ribonucleoprotein particles similarly sequesters transcripts to limit R-loop formation [27]. Although most work to date has suggested greater susceptibility of the NTS to various types of DNA damage, the reverse bias was reported at a bacterial promoter when the transcription and replication forks were in a head-on orientation [28]. It was suggested that this reflected a protective interaction of sigma factor with the NTS in a stabilized RNAP open complex. In the current study, we examined the spontaneous deamination of cytosine to uracil in yeast as a function of lowversus high-transcription of the CAN1 forward-mutation reporter. Mutations attributable to cytosine deamination occurred at similar rates on the NTS and TS strands under low-transcription conditions but were strongly biased to the NTS under high-transcription conditions. No correlation between the transcription-associated mutation bias and R-loop accumulation was detected, however, suggesting that the NTS exposed in the transcription bubble may be the primary target of DNA damage in some contexts. Williams et al. Page 3 DNA Repair (Amst) . Author manuscript; available in PMC 2023 September 11. Author Manuscript Author Manuscript Author Manuscript Author Manuscript 2. MATERIALS AND METHODS 2.1. Strain construction All strains used for mutational analyses (Table S1) were derived from SJR282 ( MATα ade2–101OC his3 Δ 200 ura3 Δ Nco gal80 Δ ::HIS3 ) by transformation. As previously described [5], high transcription of CAN1 was achieved by replacing the endogenous promoter with the GAL1 promoter ( pGAL ) linked to the bacterial kan marker [29]. FCY1, MSH6 , MFT1 , RNH1 , RNH201 , TOP1 or UNG1 was deleted by one-step allele replacement using PCR- generated cassettes containing an appropriate selectable marker. For experiments involving RNase H1 overexpression, the high-txn ung1 Δ strain was transformed with the URA3-CEN vector pRS316 [30] or a derivative containing a pGAL1-RNH1 fusion [31]. 2.2. Mutation rates and spectra A minimum of 24 cultures of each strain was used for determining canavanine-resistance (Can-R) rates. Each culture was inoculated with a single colony and was grown to saturation (three days) in YEP-GE (1% yeast extract, 2% Bacto-peptone, 2% glycerol and 2% ethanol) on a roller drum at 30°C. In a gal80 background, pGAL is constitutively active in YEP-GE medium [32]. Cultures were washed with water and the number of Can-R mutants in each was determined by plating an appropriate dilution on synthetic complete medium (SC; 0.17% yeast nitrogen base, 0.5% ammonium sulfate, 2% agar and 0.13% Hartwell’s complete amino acid mix) lacking arginine and supplemented with 60 μg/ml canavanine (Sigma). Total cell number was estimated by plating an appropriate dilution on non-selective YPD medium (1% yeast extract, 2% Bacto-peptone, 2% dextrose and 2% agar). Mutation rates were calculated using method of the median [33], and 95% confidence intervals (CIs) were determined as previously described [34]. Independent Can-R mutants for sequencing were obtained using a prong-plating technique that generates ~100 mini-cultures/plate. Briefly, cells were diluted in water and a 100-count, flat-tipped custom pinning device was used to transfer ~103 cells/pin onto non-selective YEP-GE plates. Cells were also spotted onto SC-Arg+Can plates to ensure there were no pre-existing Can-R mutants. After three days of growth, cells were replica-plated onto SC-Arg+Can plates and incubated for two days to allow the appearance of Can-R colonies. A single Can-R colony was picked from each spot and following genomic DNA isolation the CAN1 locus was amplified using MyTaq (Bioline) polymerase. For PacBio analysis, 16-nt barcode (a list available is at https://github.com/PacificBiosciences/Bioinformatics-Training/blob/master/barcoding/ pacbio_384_barcodes.fasta) was attached to the 5′ end of each CAN1 -specific primer and DNA of each mutant was amplified using a unique pair of barcodes. Barcodes for Can-R mutants isolated from the low-transcription, pCAN- CAN1 strains or high-transcription pGAL-CAN1 strains were attached to forward primers 5′-CAGTTTTTAATCTGTCGTCAATCGAAAG or 5′-GCCGAGCGGGTGACAG, respectively. Barcodes were attached to single CAN1 reverse primer (5′- GGGAGCAAGATTGTTGTGGT). Following product quantification using ImageJ software (https://imagej.nih.gov/ij/), barcoded amplicons were pooled in equal amounts for Williams et al. Page 4 DNA Repair (Amst) . Author manuscript; available in PMC 2023 September 11. Author Manuscript Author Manuscript Author Manuscript Author Manuscript subsequent SMRT library preparation. SMRT libraries were constructed and sequenced at Duke Center for Genomic and Computational Biology using the PacBio RSII and Sequel systems. Circular consensus reads were sorted and analyzed using an in-house pipeline [35]. The rates of C>T and G>A mutations were calculated as the product of the total Can-R rate and the proportion of the mutation type in the corresponding spectrum. 95% CIs were assigned to mutation proportions using the Vassarstats web site (vassarstats.net). To assign upper and lower confidence intervals to a C>T or G>A mutation rate, the 95% CIs of both the corresponding Can-R rate and the mutation proportion were combined using the root of the square of the sums [36]. For experiments involving the over-expression of RNase H1, the high-transcription ung1 Δ mutant was transformed with empty vector or with a plasmid in which RNH1 was regulated by pGAL1 . Mutation rates were calculated and Can-R mutants were isolated as described above, except that all pre-growth was in SC medium containing 2% glycerol and 2% EtOH as carbon sources and missing uracil in order to maintain plasmids. Following DNA isolation and PCR amplification of CAN1 (forward and reverse primers 5′- TTGGTATGATTGCCCTTGGT and 5′-ACAGCGTTGAAGATATGTGGC, respectively), nonsense mutations between positions 412 and 1056 were captured by Sanger sequencing (primer 5′-TATCCACACCTCTGACCAAC). This region contains six CAG/CAA codons and nine TGG codons to monitor C>T and G>A nonsense mutations, respectively. 2.3. DNA-RNA Immunoprecipitation (DRIP) Cultures (100 ml) grown to mid-log phase in YEP containing 2% galactose were washed with chilled H2O, resuspended in 1.4 ml spheroplasting buffer (1 M sorbitol, 10 mM EDTA pH 8, 0.1% β-mercaptoethanol, 2 mg/ml Zymolase 20T) and incubated at 30° for 30 min. Spheroplasts were pelleted (5 min at 4000 rpm), rinsed with H2O and resuspended in 1.65 ml of buffer G2 (800 mM guanidine HCl, 30 mM Tris-Cl pH 8, 30 mM EDTA pH 8, 5% Tween-20, 0.5% Triton X-100). Samples were treated with 10 μl of 10 mg/ml RNase A for 30 min at 37° and then with 75 μl of 20 mg/ml proteinase K (Roche) for 1 h at 50°. Genomic DNA was extracted gently with chloroform:isoamyl alcohol (24:1). DRIP was performed as previously described [37] with the following modifications. Precipitated genomic DNA was spooled onto a glass rod, washed twice with 70% EtOH, resuspended gently in 250 μl TE (10 mM Tris-HCl pH 8, 1 mM EDTA pH 8) containing 2 mM spermidine and 2.5 μl of 10 mg/ml BSA, and digested overnight with 50U each of Hind III, EcoR I, BsrG I, Xba I and Ssp I. Half of the DNA was treated with 3 μl RNase H1 (Invitrogen Ambion; 1U/μl) overnight at 37° for an RNase H1-treated control. RNase H1-treated and untreated samples were incubated with 10 μl of Dynabead-coupled S9.6 monoclonal antibody (1 mg/ml) in 500 μl binding buffer (10 mM NaPO4 pH 7.0, 140 mM NaCl, 0.05% Triton X-100) overnight at 4°C and then washed three times with binding buffer. The S9.6 antibody was coupled to Dynabeads Protein A (Invitrogen) for 2 h at 4°. DNA was eluted in 100 μl elution buffer (50 mM Tris pH 8.0, 10 mM EDTA, 0.5% SDS), treated 45 min with 7 μl proteinase K (20 mg/ml) at 55° and purified with the Macherey- Nagel DNA purification kit. Real-time quantitative PCR was performed using iTaq Universal SYBR Green (Biorad) and a 7500 Real-Time PCR machine (Applied Biosystems). Williams et al. Page 5 DNA Repair (Amst) . Author manuscript; available in PMC 2023 September 11. Author Manuscript Author Manuscript Author Manuscript Author Manuscript Forward and reverse primers for PDC1 were 5′-GAAGGTATGAGATGGGCTGGTAA and 5’ CCTTGATACGAGCGTAACCATCA, respectively. CAN1 -specific forward and reverse primers were 5′ GGTTTTCTTGGGCAATCACTT and 5′ AATTGAATGACTTGGCCAACTACA, respectively. Three biological replicates were performed for each strain; mean and standard deviations are reported. As a positive control for R-loop detection in DRIP experiments, genomic DNA extracted from W303 and an spt16–11 derivative was used. 2.4. Whole genome analysis of spontaneous deamination sites The data for 162 C>T mutations detected from whole-genome sequencing of an ung1 strain passaged for 81 days was downloaded from Chen et al. [38]. The reported genome location and flanking sequences were used to map the exact location of mutations in the Saccharomyces Genome Database (version 6CA_00146045.2). The deaminated strand (NTS or TS) was determined for mutations in RNAPII-transcribed genes using release 91 of the Ensembl genome browser. The transcript level of each gene was downloaded [39, 40] and genes were ranked based on expression level. 2.5. Additional statistical analyses The numbers of C>T versus G>A mutations observed in different strain backgrounds were compared using either a 2×2 contingency chi-square test or a one-tailed Fisher exact test, as appropriate (vassarstats.net). These tests were also used to compare the relationship between transcript level and NTS/TS deamination in the whole-genome analyses. Chisquare goodness of fit was used to compare observed to expected numbers of events. Rates were significantly differently if the rate in each individual strain did not overlap the corresponding rate-associated CI of the other strain. 3. RESULTS Relative levels of spontaneous DNA damage associated with transcription of a target gene can be inferred by examining the mutagenic consequences. An inherent complication with most base damages is that they or their repair intermediates slow or stall subsequent DNA replication and transcription. This can trigger specialized bypass mechanisms during replication and/or template-specific engagement of nucleotide excision repair during transcription, which complicates subsequent analyses and inferences. The hydrolytic deamination of cytosine to uracil is an exception, however, because the base-pairing properties of uracil are identical to those of thymine [41]. In wild-type (WT) yeast cells, uracil is efficiently removed by the Ung1 uracil-DNA N-glycosylase and is only rarely mutagenic. In the absence of Ung1 deaminated cytosine persists as part of a U:G mismatch and the uracil is replicated as a thymine, leading to CG>TA transition mutations [42]. 3.1. Transcription increases cytosine deamination in CAN1 As a reporter for spontaneous cytosine deamination we used the yeast CAN1 gene, which encodes an arginine permease that transports the toxic analog canavanine and confers canavanine sensitivity. Any forward mutation that inactivates the encoded protein results in canavanine resistance (Can-R) on drug-containing medium. To examine effects of Williams et al. Page 6 DNA Repair (Amst) . Author manuscript; available in PMC 2023 September 11. Author Manuscript Author Manuscript Author Manuscript Author Manuscript transcription on spontaneous deamination, strains were used in which the CAN1 gene was either under control of its native promoter or fused to the highly active GAL1 promoter (low- and high-transcription conditions, respectively; [5]). Under low-transcription (low-txn) conditions, 31% (69/220) of Can-R mutants isolated in an UNG1 background contained a CG>TA base substitution (Tables S2-S3). We note that this is higher than the CG>TA proportion among Can-R mutants reported previously (60/277; p=0.018; [43]). This difference could reflect either strain background differences or the use of ten times the standard amount of canavanine in the previous study. Deletion of UNG1 was accompanied by a 1.6-fold increase in the Can-R rate (Table S4) as well as a proportional increase in CG>TA mutations from 31% to 88% (217/246; p<0.0001). This translates into a 4.6-fold increase in the CG>TA rate in the absence of Ung1 (Figure 1) and confirms that most CG>TA mutations arising in an ung1 Δ background reflect cytosine deamination. Highly elevated transcription resulted in an ~12-fold increase in Can-R rate the WT background and deletion of UNG1 was associated with an additional 1.9-fold rate increase. As under low-txn conditions, there was a proportional increase in CG>TA mutations in the ung1 Δ background under high-txn conditions: from 6% (41/663) to 45% (249/554; p<0.0001). The rate of CG>TA mutations was 7-fold higher under high-txn than under low-txn conditions in the ung1 Δ background, demonstrating that transcription increases deamination of the corresponding DNA (Figure 1). The lower proportion of CG>TA mutations under highthan low-txn conditions (45% and 88%, respectively; p<0.0001) reflects the large contribution of Top1-dependent 2–5 bp deletions to the spectrum under high high-txn conditions [5, 6]. 3.2. Transcription elevates nonsense mutations more than missense mutations An inherent problem with forward-mutation assays is that missense mutations, which alter a single amino acid, are often of no functional consequence. Changes that convert a sense to a nonsense/stop codon truncate the corresponding protein product, however, and almost always disable function. At the CAN1 locus it was previously estimated that only ~5% of missense mutations confer canavanine resistance [43]. The missense-detection issue potentially can be exacerbated under high-txn conditions where there may be enough of a highly dysfunctional protein to confer a canavanine-sensitive phenotype. Among Can-R mutants analyzed, we indeed observed proportionally fewer missense mutations among CG>TA changes detected under highthan under low-txn conditions (83/249 and 114/217, respectively; p<0.0001; Tables S2 and S3). Because of the inability to reliably detect missense mutations under high-txn conditions, only CG>TA mutations that created a TAA, TAG or TGA stop codon were considered in subsequent analyses. 3.3. Transcription-associated deamination is biased to the NTS By convention, mutations are reported in relation to the coding/sense strand of a gene, which has the same sequence as the mRNA and corresponds to the non-transcribed strand (NTS). Although reporting mutations in this way allows a strand bias to be easily discerned, it often provides no information as to the source of the bias. This can be inferred, however, using DNA polymerases with specific misincorporation patterns [44], disabling a repair pathway that is highly specific for damage to only one of the complementary bases [45] or creating a Williams et al. Page 7 DNA Repair (Amst) . Author manuscript; available in PMC 2023 September 11. Author Manuscript Author Manuscript Author Manuscript Author Manuscript single-stranded DNA target [46]. Thus, deamination of cytosine on the NTS leads to a C>T change while deamination of a cytosine located on the transcribed strand (TS) results in a G>A mutational read-out (Figure 2A). C>T mutations in sense codons generate all three stop codons: CAA>TAA, CAG>TAG and CGA>TGA. By contrast, G>A mutations create only TGA and TAG stop codons, both of which are derived from the TGG sense codon (TGG>TGA and TGG>TAG); a TAA stop codon arises only via a G>A change in a preexisting TAG or TGA stop codon. Within the CAN1 ORF, there are 12 CAA/CAG/CGA codons plus 16 TGG codons (Figure S1) and most were frequent sites of C>T or G>A mutations, respectively. Exceptions were the four TGG codons nearest the 3′ end of the gene, mutation of which presumably generates a truncated but functional protein, and a TGG located 5′ to other positions where nonsense mutations were frequent. Mutations at these five codons were also absent from the spectra of Lang and Murray [43]. Excluding these five TGG codons, there were 12 and 22 positions (22 guanines within the remaining 11 TGG codon targets) in CAN1 where C>T and G>A mutations, respectively, were monitored. Even though proportionally fewer C>T than G>A mutations that created stop codons were expected (12/34 and 22/34, respectively), similar numbers of C>T and G>A mutations were observed in the low-txn ung1 Δ background: 55 and 48, respectively (Table S4; p=0.02). On a per site basis, there were 4.6 C>T changes at cytosines where deamination was monitored (55 mutations at 12 sites) and 2.2 G>A mutations at monitored guanines (48/22). There is thus an approximately 2-fold C>T bias even under low-txn conditions. The corresponding rates of C>T and G>A events, however, were very similar and are shown in Figure 2B. In contrast to similar numbers of C>T and G>A mutations under low-txn conditions, there was a 3.4-fold excess of C>T relative to G>A mutations in the high-txn ung1 Δ background: 128 and 38, respectively (p<0.0001). On a per site basis, there were 10.7 C>T changes at cytosines where deamination was monitored (128 mutations at 12 sites) and 1.7 G>A mutations at monitored guanines (38/22). In terms of rates, C>T mutations were elevated 14.5-fold under high-txn conditions while G>A mutations were elevated only 4.8-fold (Figure 2B). 3.4. The transcription-associated C>T bias reflects neither enzymatic deamination nor biased U:G repair Cytosine in single-stranded DNA spontaneously deaminates more rapidly than cytosine in duplex DNA [47], suggesting that the strand-specific deamination bias observed under hightxn conditions reflects the enhanced single-stranded character of the NTS. An alternative possibility, however, is strand-biased deamination of cytosine by the yeast Fcy1 protein, which has been reported to deaminate DNA as well as RNA [48]. If Fcy1 were the primary source of cytosine deamination, then the rate of C>T mutations should be reduced in its absence. There was no change in the rates (Figure 2B; Table S4) or distribution (p=0.12; Tables S2-S3) of C>T and G>A mutations in the ung1 Δ fcy1 Δ double mutant relative to the ung1 Δ single mutant, consistent with strand-biased hydrolytic deamination of cytosine. The mismatch repair (MMR) system removes errors that arise during DNA replication by initiating excision and re-synthesis of the newly synthesized strand (reviewed in Williams et al. Page 8 DNA Repair (Amst) . Author manuscript; available in PMC 2023 September 11. Author Manuscript Author Manuscript Author Manuscript Author Manuscript [56]. Wang D, Landick R, Nuclease cleavage of the upstream half of the nontemplate strand DNA in an Escherichia coli transcription elongation complex causes upstream translocation and transcriptional arrest. J. Biol. Chem 272 (1997) 5989–5994. [PubMed: 9038220] [57]. Barnes CO, Calero M, Malik I, Graham BW, Spahr H, Lin G, Cohen AE, Brown IS, Zhang Q, Pullara F, Trakselis MA, Kaplan CD, Calero G, Crystal structure of a transcribing RNA polymerase II complex reveals a complete transcription bubble. Mol. Cell 59 (2015) 258–269. [PubMed: 26186291] [58]. Churchman LS, Weissman JS, Nascent transcript sequencing visualizes transcription at nucleotide resolution. Nature 469 (2011) 368–373. [PubMed: 21248844] Williams et al. Page 15 DNA Repair (Amst) . Author manuscript; available in PMC 2023 September 11. Author Manuscript Author Manuscript Author Manuscript Author Manuscript Figure 1. Loss of Ung1 preferentially elevates CG>TA mutations in CAN1 under low- and high-txn conditions. White bars correspond to the total mutation rate and gray bars to the rate of CG>TA mutations. Error bars are the 95% confidence interval (CI). Williams et al. Page 16 DNA Repair (Amst) . Author manuscript; available in PMC 2023 September 11. Author Manuscript Author Manuscript Author Manuscript Author Manuscript Figure 2. Strand-specific mutation accumulation. (A) Cytosine deamination on complementary DNA strands yields C>T and G>A mutations. In an ung1 Δ background, a uracil that results from cytosine deamination persists and templates dAMP incorporation during subsequent DNA synthesis. After two rounds of replication, a deaminated cytosine on the NTS (blue) or TS (yellow) gives rise to a C>T or G>A mutation, respectively, when the NTS sequence is reported. Dotted lines represent the starting DNA strands and arrowheads correspond to 3′ ends. (B) In the absence of Ung1, high levels of transcription elevate C>T more that G>A mutations (blue and yellow bars, respectively) demonstrating a bias for NTS deamination. This bias reflects neither Fcy1 nor MMR activity. Error bars are 95% CIs. Williams et al. Page 17 DNA Repair (Amst) . Author manuscript; available in PMC 2023 September 11. Author Manuscript Author Manuscript Author Manuscript Author Manuscript Figure 3. Biased deamination of the NTS under high-txn conditions is not exacerbated in genetic backgrounds where R-loops are expected to (A) increase or (B) decrease. Blue and yellow bars reflect the rates of C>T and G>A mutations, respectively, in the indicated genetic backgrounds. Error bars are the 95% CIs. Williams et al. Page 18 DNA Repair (Amst) . Author manuscript; available in PMC 2023 September 11. Author Manuscript Author Manuscript Author Manuscript Author Manuscript Figure 4. Relative levels of RNA-DNA hybrids at CAN1 and PDC1 . Samples were or were not (+ or −, respectively) treated with RNase H1, which degrades RNA-DNA but not RNA-RNA duplexes, prior to immunoprecipitation with the S9.6 antibody. Values plotted are arbitrary units (A.U.) that represent the mean of three biological replicates after immunoprecipitation (IP) with S9.6 relative to the IP input. Error bars are standard deviations. The spt16–11 derivative of W303, which accumulates R-loops, was used as positive control for RNA-DNA detection and was analyzed in parallel with experimental strains; duplicate samples were used for WT but only a single sample was analyzed from the spt16–11 background. Williams et al. Page 19 DNA Repair (Amst) . Author manuscript; available in PMC 2023 September 11. Author Manuscript Author Manuscript Author Manuscript Author Manuscript Figure 5. Vulnerability of the NTS to DNA damage during transcription. A small portion of the NTS is rendered single-stranded as part of each transcription bubble; re-formed duplex DNA and the nascent transcript leave RNAP through separate channels. If the transcript threads back and pairs with the TS to form an R-loop, a larger portion of the NTS is exposed. In the case of CAN1 , a contribution of R-loops to the biased deamination of the NTS was detected neither genetically or physically. Damage to the NTS that occurs within the transcription bubble is indicated by a star. Williams et al. Page 20 DNA Repair (Amst) . Author manuscript; available in PMC 2023 September 11. Author Manuscript Author Manuscript Author Manuscript Author Manuscript