|Home | About | Journals | Submit | Contact Us | Français|
Reactive oxygen species are ubiquitous mutagens that have been linked to both disease and aging. The most studied oxidative lesion is 7,8-dihydro-8-oxoguanine (GO), which is often miscoded during DNA replication, resulting specifically in GC → TA transversions. In yeast, the mismatch repair (MMR) system repairs GO·A mismatches generated during DNA replication, and the polymerase η (Polη) translesion synthesis DNA polymerase additionally promotes error-free bypass of GO lesions. It has been suggested that Polη limits GO-associated mutagenesis exclusively through its participation in the filling of MMR-generated gaps that contain GO lesions. In the experiments reported here, the SUP4-o forward-mutation assay was used to monitor GC → TA mutation rates in strains defective in MMR (Msh2 or Msh6) and/or in Polη activity. The results clearly demonstrate that Polη can function independently of the MMR system to prevent GO-associated mutations, presumably through preferential insertion of cytosine opposite replication-blocking GO lesions. Furthermore, the Polη-dependent bypass of GO lesions is more efficient on the lagging strand of replication and requires an interaction with proliferating cell nuclear antigen. These studies establish a new paradigm for the prevention of GO-associated mutagenesis in eukaryotes.
Eukaryotic genome stability can be compromised by changes at the nucleotide level, alterations in chromosome structure, or changes in chromosome number. Although such changes are responsible for many human diseases, including cancer, a low level of instability is necessary to provide the raw material for evolutionary processes. Changes at the nucleotide level generally occur during replication, either as errors made when copying an undamaged DNA template or during the bypass of DNA lesions. Many types of DNA lesions are due to reactive oxygen species (ROS), which are generated by exposure to physical and chemical mutagens, as well as by normal metabolic processes, such as aerobic respiration (12, 32). Although cells contain multiple antioxidants and other proteins that protect the genome from oxidative damage, ROS have been implicated as causal agents of many diseases and of aging (11, 50).
The most common oxidized DNA lesion is 7,8-dihydro-8-oxoguanine, which is referred to here as a GO lesion. The mutagenic potential of this lesion is due to miscoding during DNA synthesis, with replicative DNA polymerases usually misinserting adenine across from the lesion to generate GO·A mispairs and ultimately GC → TA transversions (49). Studies examining the crystal structure of T7 DNA polymerase complexed with a GO·C base pair or a GO·A mispair indicate the basis of this mutagenic specificity. Whereas the GO·C structure physically resembles that of a mismatch, the GO·A mispair structurally resembles a normal Watson-Crick base pair and therefore is likely to escape polymerase-associated proofreading activity (6). A GO-containing nucleotide triphosphate (8-oxo-dGTP) can also be used by DNA polymerases during DNA synthesis, leading specifically to AT > CG transversion events (7).
There are three major proteins in Escherichia coli that work independently to prevent GO-associated mutagenesis: MutM (Fpg), MutY, and MutT (36). MutM is a DNA glycosylase that removes GO lesions in the GO·C base pairs created by oxidation of guanine in normal G·C base pairs, while MutY is an adenine-DNA glycosylase that removes adenines from the GO·A mispairs created by incorporation of adenine opposite a GO lesion. If DNA replication occurs before MutM can remove the GO lesion from a GO·C base pair, the lesion will likely generate a GO·A mispair, which is then subjected to the A-specific activity of the MutY protein. Once MutY removes the adenine from the newly synthesized strand, a cytosine can be inserted opposite the lesion, giving MutM another opportunity to excise the GO base. MutT is an 8-oxo-dGTPase that degrades 8-oxo-dGTP, thereby greatly reducing its incorporation into DNA. The postreplicative mismatch repair (MMR) pathway has also been implicated in preventing GO-associated mutagenesis in E. coli by functioning as an alternative to MutY or by helping MutY identify and remove mismatched adenines from GO·A mispairs (3, 60).
In the yeast Saccharomyces cerevisiae, the Ogg1 protein is the functional homolog of MutM (55) and thus removes GO lesions that are base paired with cytosine. The MMR machinery is functionally analogous to the MutY protein (37), excising adenines that are inserted opposite GO lesions during DNA replication. The mismatch recognition MutSα complex (a heterodimer of the Msh2 and Msh6 proteins) specifically recognizes GO·A mispairs and initiates removal of the portion of the newly synthesized strand containing the adenine (37). A homolog of MutT has yet to be identified in yeast, although one does exist in mammalian cells (23). It is possible that the MutT homolog has eluded discovery either because it is essential, because there is a redundant activity, or because 8-oxo-dGTP is not a significant mutagen in yeast.
A third mechanism that limits GO-associated mutagenesis in yeast involves the translesion synthesis (TLS) polymerase, polymerase η (Polη), which is a member of the Y family of DNA polymerases and is encoded by the RAD30 gene (18, 61). Y family polymerases have a large active-site pocket that can accommodate structurally deformed bases, enabling them to insert a nucleotide opposite a lesion (29). Not only is such lesion bypass potentially error prone, the larger active-site pocket of TLS polymerases imparts very low fidelity when copying undamaged DNA. Polη, for example, is error prone when bypassing some lesions, such as abasic sites (17), but has relatively high fidelity when bypassing GO lesions, usually inserting a cytosine across from the lesion (18, 61). At GO lesions, Polη is 10-fold more accurate and efficient than Polδ (34). When given an undamaged DNA template, however, the base substitution error frequency of Polη in vitro is 3 orders of magnitude greater than that of a typical replicative polymerase (35). In addition to Polη, there are two other TLS polymerases in S. cerevisiae (Polζ and Rev1), but neither has been implicated in the bypass of GO lesions (10, 48).
The most straightforward way for Polη to be involved in GO bypass would be for it to be recruited when a replicative polymerase stalls or leaves behind a gap. The replication-blocking potential of GO lesions, however, is unclear. Some in vitro studies have shown that replicative DNA polymerases stall or pause when encountering a template GO lesion (18, 47), but other studies have suggested that this is not the case (49). The currently accepted model is that Polη is specifically recruited to fill the gaps generated by MMR when it initiates correction of GO·A mispairs (18, 54). This model of MMR-Polη cooperation in preventing GO-associated mutagenesis is based on epistasis analysis performed using the CAN1 forward mutation assay (18). Although the relationship between msh2 and rad30 was concluded to be epistatic, the data are also consistent with an additive relationship and, hence, potentially independent roles of Msh2 and Polη in limiting GO-associated mutagenesis. How and why the MMR pathway might specifically recruit a generally error-prone polymerase to fill the gaps in what is normally an extremely accurate repair process is not obvious.
In the present study, a SUP4-o forward-mutation system was used to reexamine the relationship between MMR and Polη in preventing GO-associated mutagenesis in yeast. To enhance the accumulation of GO lesions, all experiments were conducted in mutants defective in removing GO from GO·C base pairs (an ogg1 background). In addition, both msh2 and msh6 mutants were analyzed. In an msh6 background, loss of the MutSα heterodimer eliminates the correction of GO·A mispairs, while retention of MutSβ (a heterodimer of Msh2 and Msh3) allows continued correction of most insertion-deletion loops. Finally, mutation spectra, as well as mutation rates, were considered in order to focus specifically on GC → TA mutations. The results reported here demonstrate that Polη can function independently of MMR to prevent GO-associated mutagenesis, presumably through its ability to bypass these lesions in an error-free manner. The data further indicate that the Polη-dependent bypass of GO lesions is more efficient on the lagging strand of replication and that it requires interaction with proliferating cell nuclear antigen (PCNA).
All strains were derived from strain SJR576 (MATa ura3ΔNco lys2-1oc can1-100oc ade2-1oc leu2-K). To insert SUP4-o into the HBN1 locus in the forward orientation (hbn1::SUP4-oF allele), the following primers were used to amplify the allele from plasmid JF1754 (43): forward primer, 5′ GGGAATGCAGCTGCGTACGCTGGGAAGTCAGCCTTTAGCTTTTCAGTTACCTTGGGATCCGGGACCGGATAATT, and reverse primer, 5′ GGCTATAGAAAGCCCTGCCGGTCAAAAGAGGCCTGCTTCAGCAAGGGATGAGGCCAATTCTTGAAAGAAATATTTC. The underlined portion of the sequence corresponds to sequence flanking SUP4-o, and the nonunderlined portion corresponds to the HBN1 locus. SUP4-o was amplified and inserted in the reverse orientation at the HBN1 locus (hbn1::SUP4-oR allele) using the forward primer 5′ GGGAATGCAGCTGCGTACGCTGGGAAGTCAGCCTTTAGCTTTTCAGTTACCTTGAATTCTTGAAAGAAATATTTC and the reverse primer 5′ GGCTATAGAAAGCCCTGCCGGTCAAAAGAGGCCTGCTTCAGCAAGGGATGAGGCCGGATCCGGGACCGGATAATT. SUP4-o was inserted into the AGP1 locus in the forward orientation (agp1::SUP4-oF allele) using the forward primer 5′ GCTTGATTAATTCTTCATCAAAGATTTGTCTATGAGAATCTAGGTCGATCTTGTCGGATCCGGGACCGGATAATT and the reverse primer 5′ GGTCGGTAACGGTACCGCGTTGGTTCATGCGGGTCCAGCTGGACTACTTATTAATTCTTGAAAGAAATATTTC. Finally, SUP4-o was inserted into the AGP1 locus in the reverse orientation (agp1::SUP4-oR allele) using the forward primer 5′ GCTTGATTAATTCTTCATCAAAGATTTGTCTATGAGAATCTAGGTCGATCTTGTCAATTCTTGAAAGAAATATTTC and the reverse primer 5′ GGTCGGTAACGGTACCGCGTTGGTTCATGCGGGTCCAGCTGGACTACTTATTGGATCCGGGACCGGATAATT. Following transformation of SJR576 with the appropriate PCR product, Lys+ colonies were selected, and the presence of SUP4-o was inferred by cosuppression of the ade2-101 and can1-100 ochre alleles. Insertion of SUP4-o at the correct location was confirmed by sequencing.
The OGG1, RAD30, MSH2, and MSH6 genes were deleted by transforming strains with a PCR-generated fragment containing a kanamycin resistance (2), KlURA3 (URA3 gene from Kluveromyces lactis) (15), or hygromycin resistance marker (14) with the appropriate flanking sequence of the target gene. ogg1Δ::kan, msh2Δ::hyg, and msh6Δ::hyg transformants were selected on YEPD medium (1% yeast extract, 2% Bacto peptone, 2% dextrose, and 250 mg/liter adenine) containing 200 μg/ml Geneticin or 300 μg/ml hygromycin. rad30Δ::KlURA3 transformants were selected on synthetic complete medium containing 2% dextrose and lacking uracil. Deletions were confirmed by PCR. The rad30(1-624) allele (16), lacking the last 8 amino acids of the Rad30 protein, was introduced using the delitto perfetto method (52) as previously described (1). Transformants were confirmed by sequencing them.
To determine mutation rates, four or five individual colonies were used to inoculate a 5-ml starter culture. Following overnight growth at 30°C, the starter culture was used to inoculate independent 5-ml secondary cultures to a concentration of 2.5 × 105 cells/ml. Two isolates were used for each strain, and at least 12 cultures were used for each isolate. These cultures were grown for 3 days at 30°C. Nonselective YEPGE medium (1% yeast extract, 2% Bacto peptone, 2% glycerol, 2% ethanol, and 250 mg/liter adenine) was used for both starter and secondary cultures. Appropriate dilutions of each culture were plated onto YEPGE medium to determine total cell numbers and on plates with synthetic complete medium containing 2% dextrose and lacking arginine supplemented with 60 μg/ml l-canavanine (Sigma) to select canavanine-resistant (Canr) colonies. Colonies were designated SUP4-o mutants if they were both resistant to canavanine and red (Ade−), indicating loss of suppression of both the can1-100 and ade2-1 alleles. Mutation rates were determined using at least 24 cultures and the method of the median (27), and 95% confidence intervals were calculated as previously described (51). Either comparison of the confidence intervals or the Mann-Whitney test (http://faculty.vassar.edu/lowry/utest.html) was used to determine whether two rates were significantly different. The mutation rates for specific mutation types were calculated by multiplying the proportion of an event in the corresponding spectrum by the total mutation rate.
To generate mutation spectra, DNA was extracted from purified red Canr colonies (http://jinks-robertsonlab.duhs.duke.edu/protocols/yeast_prep.html). The SUP4-o gene was amplified by PCR and sequenced using the HBN1 sequencing primer (5′ CCGCTTTCAACTCCCAGCC) or the AGP1 sequencing primer (5′ GGGTTATTGGTCGGTAACGG), as appropriate. Sequencing was performed by either the High-Throughput Genomics Unit (Seattle, Washington) or the Duke University DNA Analysis Facility (Durham, North Carolina).
Proportions of leading- and lagging-strand mutations were analyzed using chi-square analysis (http://faculty.vassar.edu/lowry/newcs.html). A P value of less than 0.05 was considered statistically significant.
In the experiments reported here, the SUP4-o forward-mutation system originally characterized by Pierce et al. (43) was used to examine the roles of MMR and Polη in preventing GO-associated mutagenesis. SUP4-o is a mutant tRNA that suppresses ochre stop codons by inserting a tyrosine, and two ochre alleles were used to monitor SUP4-o function: ade2-1 and can1-100. In the presence of SUP4-o, cells are phenotypically Ade+ (white) and Cans (canavanine sensitive). Mutations that inactivate SUP4-o can be identified by simultaneous loss of suppression of the ade2-1 and can1-100 alleles, resulting in Ade− (red), Canr colonies. SUP4-o is an ideal reporter to use for studying GO-associated mutagenesis because a mutation at virtually any site disrupts SUP4-o function and allows phenotypic detection; the GC content is 51%, which is higher than the average GC content of the yeast genome and makes detection of GC → TA mutations more efficient; the small size of the gene (89 bp) enables easy sequencing and determination of mutation patterns. It should be noted that the SUP4-o sequence reported throughout is that of the transcribed (noncoding) strand, which follows the convention established by Pierce and coworkers (43).
The model that Polη and MMR cooperate to prevent GO-associated mutagenesis was based on epistasis analysis between msh2 and rad30 alleles in an ogg1 background. In the CAN1 forward-mutation assay, Canr rates were elevated 11-fold, 21-fold, and 27-fold in ogg1 rad30, ogg1 msh2, and ogg1 rad30 msh2 mutants, respectively, relative to the wild type (WT) (18). Although it was concluded that the msh2 and rad30 alleles are epistatic, the use of mutation rate data to distinguish additive versus epistatic interactions becomes especially problematic when one rate is higher than the other. An additional issue with the analysis was that only total mutation rates, rather than the rates of GO-associated GC → TA mutations, were considered. The proportion of mutation spectra that are GC → TA transversions can vary dramatically in different genetic backgrounds (37), making total mutation rates a poor indicator of GO-specific GC → TA rates. Because of the inherent uncertainties associated with the earlier epistasis analysis, we reexamined the relationship between Polη and Msh2, focusing specifically on GO-associated GC → TA mutations in the SUP4-o reporter.
For initial analyses, we inserted SUP4-o into the nonessential HBN1 locus (the hbn1::SUP4-oF allele), which is closely linked to the well-characterized ARS306 origin of replication on chromosome III. In an ogg1 background, GO lesions persist in the genome, leading to GO·A mispairs during replication and increasing the relative contribution of GC → TA transversions to mutation spectra. Although the hbn1::SUP4-oF mutation rate was not significantly elevated in the ogg1 strain, the proportion of GC → TA mutations increased from 27% to 67% (Table (Table11 and Fig. Fig.1).1). Taking the proportional increase in GC → TA transversions into account, the rate of these mutations was 3.3-fold higher in the ogg1 mutant than in the WT. Although Polη loss had no effect on mutagenesis when Ogg1 was present, the rate of GC → TA transversions in the double mutant was 10-fold higher than in the WT and 3.1-fold higher than in the ogg1 single mutant (Table (Table1).1). Thus, as reported in other studies (10, 18), there was a synergistic effect of simultaneously removing Ogg1 and Polη, confirming that these proteins act in separate pathways to limit GO-associated mutations.
A similar synergistic interaction was evident when epistasis between ogg1 and msh2 was examined. Relative to the WT, the total SUP4-o mutation rate was elevated 1.3-fold in the ogg1 single mutant, 4.5-fold in the msh2 single mutant, and 19-fold in the ogg1 msh2 double mutant. The synergism was more striking when only the rates of GC → TA mutations were considered, with 3.3-fold, 4.7-fold, and 42-fold increases in the ogg1 single, msh2 single, and ogg1 msh2 double mutants, respectively. As observed in the CAN1 assay, the SUP4-o mutation rate in the ogg1 rad30 msh2 triple mutant was slightly higher than in the ogg1 msh2 mutant (5.5 × 10−6 and 3.0 × 10−6, respectively), but not significantly so when confidence intervals were considered. When mutation spectra (see Fig. S1 in the supplemental material) were used to calculate rates of GO-associated GC → TA transversions, however, the effect of combining msh2 and rad30 was clearly more than additive, with the GC → TA rate in the ogg1 rad30 msh2 triple mutant being more than twice that in the ogg1 msh2 double mutant (Table (Table1).1). The genetic data obtained here with the hbn1::SUP4-oF allele thus do not support a model in which Polη bypass occurs only downstream of Msh2-dependent GO·A mismatch removal. We suggest instead that Polη can directly bypass replication-blocking GO lesions in an error-free manner.
To further substantiate the MMR-independent role of Polη in preventing GO-associated mutagenesis, we also examined the epistatic interaction between rad30 and msh6 in an ogg1 background. Given the greater proportion of base substitutions in msh6 spectra than in msh2 spectra (37), we speculated that the relationship between Polη and MMR might be more evident in msh6 mutants. Loss of Msh6 led to a 2.3-fold increase in the overall forward-mutation rate of the hbn1::SUP4-oF allele, and simultaneous loss of Msh6 and Ogg1 resulted in a 9.4-fold increase in the overall mutation rate (Table (Table1).1). The ogg1 rad30 msh6 triple mutant exhibited a 28-fold elevation in the overall SUP4-o mutation rate, which is clearly a synergistic increase relative to the 3.4- and 9.4-fold increases in the ogg1 rad30 and ogg1 msh6 double mutants, respectively (Table (Table1).1). The synergism was again much more striking when only GC → TA transversions were considered, with 10-, 26-, and 88-fold rate increases for the ogg1 rad30 double-, ogg1 msh6 double-, and ogg1 msh6 rad30 triple-mutant strains, respectively, relative to the WT. In all subsequent experiments, msh6 mutants were used as the MMR-defective background for analysis of GO-associated GC → TA transversions.
To exclude a site-specific anomaly in our data, we inserted SUP4-o into the AGP1 locus, which positions the allele on the other side of ARS306, approximately 2.4 kb away from the HBN1 locus. The analogous ogg1, ogg1 rad30, ogg1 msh6, and ogg1 rad30 msh6 mutant strains containing the agp1::SUP4-oF allele were constructed and analyzed. Neither the overall mutation rates nor the proportions of GC → TA mutations were different from those in the equivalent strains with the hbn1::SUP4-oF allele (Table (Table2).2). Because the WT mutation rate was slightly lower in the agp1::SUP4-oF strain, however, the rate increases in the mutant strains were even more dramatic than those obtained in the analogous hbn1::SUP4-oF strains. Relative to the WT strain, the GC → TA mutation rate within the agp1::SUP4-oF allele increased 21-fold in the ogg1 rad30 strain, 49-fold in the ogg1 msh6 strain, and 241-fold in the ogg1 rad30 msh6 strain.
Polη contains a short, PCNA-interacting peptide (PIP) domain at its C-terminal end, and studies suggest that its bypass activity requires an interaction with PCNA (16, 54). To determine if this interaction is required for the Polη-dependent bypass of GO lesions, we replaced RAD30 with the rad30(1-624) allele, which lacks the last 8 amino acids of Rad30 and thereby removes the PIP domain, into the ogg1 and the ogg1 msh6 strains containing the hbn1::SUP4-oF allele. If an interaction with PCNA is required for Polη bypass of GO lesions, the rad30(1-624) allele should produce a phenotype similar to that of the rad30 null allele. As shown in Table Table1,1, the GC → TA mutation rates were indistinguishable in the strains containing the rad30Δ versus the rad30(1-624) alleles. Polη thus requires interaction with PCNA for the efficient bypass of GO lesions in the SUP4-o system.
Because the identity of the initiating lesion is known, GC → TA mutations in ogg1 mutants can be attributed to GO·A mispairs rather than to C·T mispairs. This property of GO-associated mutagenesis can be used to assign the strand of origin of mutations and, as first described by Pavlov et al. (41), to compare leading- and lagging-strand mutagenesis if the direction of replication is known. This approach was used previously to demonstrate that more GO-associated mutagenesis occurs during leading-strand synthesis (41) and that most of this bias results from more efficient MMR during lagging-strand synthesis (40). As illustrated in Fig. Fig.2A,2A, the transcribed strand is the lagging-strand template in the hbn1::SUP4-oF strain, meaning that G → T and C > A mutations arise from GO lesions on the lagging- and leading-strand templates, respectively. In the mutation spectrum of the hbn1::SUP4-oF ogg1 strain, there were many more C > A than G → T mutations (57% and 9.9%, respectively) (Fig. (Fig.1),1), consistent with most GO-associated mutations being generated during leading-strand synthesis. Also in agreement with earlier studies (40), this bias largely disappeared in the ogg1 msh6 double mutant, where the numbers of C > A and G → T mutations were more similar (46% and 30%, respectively).
Although an estimate of the ratio of leading- to lagging-strand errors can be obtained by comparing the numbers of G → T and C > A mutations in a given spectrum, a more accurate method is to examine exactly the same sequence when it is present on each of the two strands. The orientation of the SUP4-o reporter within the HBN1 locus was thus reversed to generate the hbn1::SUP4-oR allele, in which the transcribed strand is switched from the lagging- to the leading-strand template during replication (Fig. (Fig.2B).2B). As was done with the hbn1::SUP4-oF allele, mutation rates and spectra were generated for ogg1, ogg1 rad30, ogg1 msh6, and ogg1 rad30 msh6 strains containing the hbn1::SUP4-oR allele (see Table S1 and Fig. S2 in the supplemental material). Because the rate of SUP4-o mutations in a given strain background was independent of gene orientation, the proportions of G → T (or C > A) mutations that arose during leading-strand synthesis were directly compared to those generated during lagging-strand synthesis.
As shown in Table Table3,3, G → T leading-strand mutations accounted for 32% of the ogg1 spectrum while G → T lagging-strand mutations comprised only 9.9% of the spectrum. There is thus an approximately threefold leading-strand bias for GO-associated mutations (P < 0.0001). In the ogg1 msh6 strains, the proportion of both leading- and lagging-strand G → T mutations increased, but the proportion of lagging-strand mutations increased more than that of leading-strand mutations (9.9% to 30% for lagging-strand mutations and 32% to 41% for leading-strand mutations). Interestingly, a similar pattern was seen in the ogg1 rad30 strains, with a 9.9%-to-24% increase in lagging-strand mutations but only a 32%-to-39% increase in leading-strand mutations. The leading-strand bias for G → T mutations was thus greatly reduced by elimination of either Msh6 or Polη, indicating that both processes are more efficient at reducing GO-associated mutagenesis during lagging-strand synthesis. Importantly, however, a significant bias still persisted in each corresponding double mutant. This bias was entirely eliminated in the ogg1 rad30 msh6 triple mutant, with the proportion of G → T mutations being statistically the same on the two strands (P = 1). Similar results were obtained when the proportions of C > A mutations generated during leading- versus lagging-strand replication were analyzed; the strong leading-strand bias in the ogg1 mutant was reduced by loss of either Msh6 or Polη but was completely eliminated only when both were absent. These results confirm previous observations that MMR is more efficient at repairing GO·A mismatches that arise during lagging-strand synthesis (40) and also demonstrate a clear lagging-strand bias for the error-free bypass of GO lesions by Polη. In addition, because some leading-strand bias for mutations persists when either Polη or Msh6 is present, the data indicate that the Polη strand-specific bias is at least partially independent of MMR, and vice versa.
To examine whether genome position affects either the efficiency of GO·A mispair removal or error-free GO bypass, a similar analysis was done at the AGP1 locus on the other side of ARS306. At this location, the transcribed strand of SUP4-o was the leading-strand template in the agp1::SUP4-oF allele and the lagging-strand template in the agp1::SUP4-oR allele (Fig. 2C and D). The comparisons of G → T and C > A mutations when SUP4-o was placed at AGP1 are presented in Table Table44 (see the supplemental data for mutation rates and spectra). As in the hbn1::SUP4-o ogg1 strains, there was a strong leading-strand bias for both G → T and C > A mutations in the agp1::SUP4-o ogg1 strains (P = 0.0001 and P = 0.0004). The effects of Msh6 and Polη loss on this bias, however, were different at the AGP1 location. While there was only a weak reduction in the leading- to lagging-strand bias when Msh6 was eliminated in the ogg1 background (from 2.5 to 1.9 for G → T mutations and from 1.6 to 1.3 for C > A mutations), the bias for both G → T and C > A mutations was completely eliminated in the ogg1 rad30 mutants (P = 0.823 and P = 1, respectively). Thus, at the AGP1 position, the greater accumulation of GO-associated mutations during leading-strand synthesis in the ogg1 single mutant appears to be completely dependent on the more efficient bypass activity of Polη during lagging-strand synthesis.
It is important to note that the overall strand-related biases calculated in Tables Tables33 and and44 represent the cumulative effects at many sites across the SUP4-o sequence. When individual sites were analyzed in the hbn1::SUP4-o ogg1 strains, the ratio of leading- to lagging-strand mutations ranged from 0.43 to 31 (see Fig. S5 in the supplemental material). Although the small numbers of mutations at most of the individual sites preclude accurate statistical analysis, the huge site-to-site variation in the leading- versus lagging-strand bias demonstrates that mutagenesis is not equal at every site and that there is a wide range of variability in the GO-related strand bias for both MMR and Polη. When all sites were summed, the ratio of GC → TA leading-strand mutations to GC → TA lagging-strand mutations was 2.9. In the ogg1 rad30, ogg1 msh6, and ogg1 rad30 msh6 strains, the range of the leading- to lagging-strand ratios at individual sites narrowed, and the ratios determined by summing all sites decreased to 1.4, 1.4, and 1.0, respectively.
Although mutation spectra alone provide information about the distributions of mutations, a more quantitative assessment of site-to-site variation in GC → TA transversions can be obtained by calculating the mutation rates at individual sites. As described above, both Polη and MMR are more efficient on the lagging strand at the HBN1 locus (Table (Table3).3). Therefore, GO lesions on the lagging-strand templates of the hbn1::SUP4-o alleles were specifically examined. This corresponds to G → T mutations in the hbn1::SUP4-oF allele and C > A mutations in the hbn1::SUP4-oR allele. A small subset of these sites is shown in Fig. Fig.3;3; the complete data for all sites can be found in Fig. S6 in the supplemental material. Analysis of site-specific mutation rates revealed several important points. First, consistent with the overall rate measurements, there was a synergistic effect of removing Rad30 and Msh6 at the majority of sites. At some sites, however, the relative role of MMR or Polη was enhanced or diminished. At sites 8G and 10C, for example, Polη appeared to be required for all error-free bypass of GO lesions, with MMR playing a relatively minor role. In contrast, at other sites (e.g., site 30G), deletion of RAD30 had little, if any, effect on the mutation rate, and only the MMR machinery seemed to be important for limiting GO-associated mutagenesis. Interestingly, in an ogg1 rad30 msh6 background, the rates of GO-associated mutations at specific sites varied more than 20-fold. Even in the absence of repair, there clearly is significant site-to-site variability in mutagenesis.
As presented here, G → T mutations always reflect GO lesions on the transcribed strand of the SUP4-o gene and C > A mutations represent mutations generated in response to GO lesions on the nontranscribed strand. Interestingly, the proportion of mutations that were C > A transversions in the triple-mutant ogg1 rad30 msh6 background was greater than the proportion of mutations that were G → T transversions (Tables (Tables33 and and4).4). At both the HBN1 and AGP1 locations, approximately 50% of mutations were C > A transversions while only 40% were G → T transversions (P = 0.0002 for both sites). The greater abundance of C > A mutations suggests that there is more accumulation of GO lesions on the nontranscribed than on the transcribed strand of SUP4-o, regardless of whether the nontranscribed strand is the leading- or lagging-strand template during replication.
The mutagenic potential of GO lesions and their relevance to cancer and aging have been well documented (11, 32, 50). Despite decades of study, however, the various mechanisms that act to prevent GO-associated mutagenesis in eukaryotes have yet to be fully described. In addition to Ogg1, which excises GO lesions from GO·C mispairs, both MMR and Polη reduce the mutagenesis that results from the insertion of adenine opposite a template GO lesion during DNA synthesis (18, 37, 61). Based on epistasis analysis between msh2 and rad30, it was proposed that Polη works exclusively in the context of MMR to fill the gaps generated when the MMR system removes the adenine of GO·A mispairs (10, 18, 54). The data presented here demonstrate that the full relationship between MMR and Polη in limiting GO-associated mutagenesis was obscured in earlier studies because msh2 strains were used for epistasis analysis and because only total mutation rates were considered (18). Using spectra to focus on GO-associated GC → TA mutations, we were able to clearly observe synergism between msh2 and rad30. When msh6 mutants were used instead of msh2 mutants, the synergism with rad30 was evident in total mutation rate measurements and was further exaggerated when GC → TA mutation rates were calculated. Finally, a simple comparison of the ogg1 msh6 and ogg1 msh6 rad30 spectra provided visual confirmation that msh6 is not epistatic to rad30; if it were, no changes would have been expected upon the additional deletion of RAD30. Our data thus clearly demonstrate that MMR and Polη can function independently to limit GO-associated mutagenesis. As illustrated in Fig. Fig.4,4, we suggest that the MMR-independent role of Polη reflects its direct recruitment for error-free bypass of GO lesions that block the replicative DNA polymerases. Whether such recruitment involves a polymerase switch at the fork and/or occurs through the filling of gaps left behind the fork is not known.
While our data clearly demonstrate that Polη can act independently of the MMR machinery to reduce GO-associated mutagenesis, they do not exclude the possibility that Polη may sometimes be recruited to fill GO-containing gaps created by MMR, as originally proposed (18). We suggest, however, that Polη would only become involved in MMR if a replicative polymerase is blocked by a lesion during the gap-filling reaction. TLS in this context is likely no different from that triggered by any other replication-blocking lesion, and so it is not unique to the MMR process. It thus requires that the PIP domain of Polη be intact (Table (Table1)1) (16, 54) and may involve a specific interaction with the monoubiquitinated form of PCNA (39, 54, 58, 63). Aside from this specific scenario, however, it is difficult to imagine a more global involvement of Polη in MMR; the relatively low fidelity of this enzyme on undamaged DNA would lead to many nonspecific errors during MMR-associated gap filling (35). Indeed, the only unequivocal evidence of Polη functioning in MMR is in the specialized somatic hypermutation of B cells (9).
One advantage of sequencing a small target is that mutation patterns can be easily discerned. In the case of the SUP4-o allele, there are 45 positions where GO-initiated GC → TA transversions can occur, and transversions at most of these sites were detected (Fig. (Fig.1).1). As expected based on overall GC → TA rates, an analysis of mutation rates at individual sites within the hbn1::SUP4-o spectra confirmed that most sites were synergistically affected by the removal of both Polη and Msh6. There were sites, however, where loss of Polη had a greater than average effect and removal of Msh6 had little effect. We suggest that these sites correspond to locations where the replicative DNA polymerase is more efficiently blocked by a GO lesion. In contrast, other sites that were affected only by loss of Msh6 may be locations where the replicative polymerase rarely stalls, resulting in a high incidence of GO·A mismatches that are then subject to MMR. Because of this considerable site-to-site variation, our data underscore the great caution that needs to be exercised when using a single site (e.g., a reversion assay) in mutation analyses.
Even though mutations were much more evenly distributed in the ogg1 rad30 msh6 triple mutant than in other genetic backgrounds, site-to-site differences persisted, indicating significant variability in the susceptibilities of specific guanines to oxidative damage and/or in the propensity of the replicative DNA polymerases to insert A opposite GO lesions. This observed variability is not surprising, as previous studies have shown that mutation and repair rates vary across the genome (20) and even within small stretches of DNA (4, 19). Several studies, for example, have shown that GO lesions are less accessible to Ogg1 in areas of DNA that also contain AP sites, single-strand breaks, additional oxidative lesions, and other types of damage (8, 22, 42). Moreover, efficient nucleotide insertion opposite GO lesions and primer extension from these insertions are also affected by nearby lesions and neighboring bases (22, 62). Based on previous studies and the work presented here using SUP4-o, it is clear that multiple factors determine the relative involvements of Ogg1, MMR, and Polη in limiting GO-associated mutagenesis at a given site.
Possible differences in the frequencies or mechanisms of mutagenesis during leading- versus lagging-strand synthesis have been examined using a variety of assays, mutagens, and strain backgrounds (31, 38, 41, 53, 56, 59). These different studies have inferred a leading-strand bias, a lagging-strand bias, or the complete absence of a bias. A single, unifying explanation for these various findings is unlikely, as mutations result from a wide variety of initiating events and pathways. Our results with the hbn1::SUP4-o alleles demonstrate that, at this specific location, both MMR and Polη-dependent bypass are more efficient on the lagging strand of replication, leading to enhanced mutagenesis on the leading strand (Tables (Tables33 and and4).4). We considered the possibility that the greater efficiency of Polη activity during lagging-strand synthesis at the HBN1 location could simply reflect a role for Polη during MMR and hence the strand bias of MMR. It should be noted, however, that some leading-strand mutation bias persisted in the absence of Msh6 and that this residual bias was completely eliminated upon additional removal of Polη. These results are consistent with earlier work demonstrating that MMR removes errors more efficiently during lagging- than during leading-strand synthesis (40) and provide the first evidence that the MMR-independent Polη bypass is also more efficient on the lagging strand. Interestingly, all of the leading-strand bias in SUP4-o mutagenesis at the AGP1 location required the presence of Polη, with the contribution of MMR being relatively minor. This suggests an additional layer of complexity, with effects of the immediate chromosomal environment extending into the mutational target.
It has been suggested that the lagging-strand bias of MMR is due to the increased accumulation either of nicks or of PCNA that accompanies discontinuous DNA synthesis (40). Just as PCNA would be left behind when the DNA polymerase is recycled to extend the next primer during lagging-strand synthesis, PCNA would presumably be left behind to mark a gap created when a lesion blocks a replicative polymerase. Because lagging-strand DNA replication is an inherently discontinuous process, we speculate that lesion-triggered gaps are more readily formed on the lagging strand than on the leading strand. This would account for the greater efficiency of Polη during lagging-strand synthesis, as well as for the central role of PCNA in orchestrating the bypass reaction.
In addition to demonstrating replication-related strand differences in mutagenesis, the data presented here also indicate that there are more GO lesions present on the nontranscribed strand than on the transcribed strand of SUP4-o (Tables (Tables33 and and4).4). This could be explained by either more efficient removal of lesions from the transcribed strand or a greater accumulation of lesions on the nontranscribed strand. In the first scenario, GO lesions on the transcribed strand might block RNA polymerases, resulting in the activation of transcription-coupled nucleotide excision repair and a concomitant reduction in potential mutagenesis. In the second scenario, the process of transcription might lead to the formation of a DNA-RNA hybrid between the DNA template and the newly synthesized RNA, leaving the nontranscribed strand transiently single stranded and more chemically reactive (25). It should be noted that both CG > TA mutations resulting from cytosine deamination and GO-dependent GC → TA transversions have been shown to preferentially accumulate on the nontranscribed strand of a highly transcribed gene in E. coli (5, 24).
ROS are a constant threat to DNA integrity, and defining the mechanisms that limit their mutagenic effects is critical to understanding the regulation of eukaryotic genome stability. As in yeast, both MMR and Polη have been shown to prevent GO-associated mutagenesis in mammalian cells (28, 33, 45, 46). While purified yeast Polη is both more efficient and more accurate than Polδ in GO lesion bypass, recent data suggest that human Polη has much lower fidelity (34). Whether this reflects a lack of relevant accessory proteins in vitro or a less prominent role of Polη in the error-free bypass of GO lesions in mammalian cells is unclear. The exact role of MMR in mammals is also unclear, as the MutSβ complex does not bind GO mismatches at all and the MutSα complex does not bind or repair GO mismatches efficiently unless the mismatches are within repeats and are associated with slippage (26, 30, 33). Consistent with these in vitro data, oxidative lesions have been linked to frameshifts and microsatellite instability in many species (13, 21, 57). Interestingly, human cells lacking Polη were recently shown to have increased rates of spontaneous fragile-site instability (44). It is possible that replication is more likely to stall at GO lesions present in fragile sites and, therefore, that Polη is required for bypassing these lesions and maintaining genome stability. Future studies that further elucidate the mechanisms behind MMR-independent Polη bypass will provide additional insight into the role of TLS in the prevention of oxidative mutations and, specifically, the role of human Polη in limiting mutagenesis and preventing human disease.
This work was supported by NIH grants GM038464 and GM064769 awarded to S.J.-R.
We thank Jessica Franke and Amy Abdulovic for contributions made in the early phases of this work. We thank members of the S.J.-R. laboratory for helpful discussions.
Published ahead of print on 27 July 2009.
†Supplemental material for this article may be found at http://mcb.asm.org/.