|Home | About | Journals | Submit | Contact Us | Français|
Mutation in response to most types of DNA damage is thought to be mediated by the error-prone sub-branch of post-replication repair and the associated translesion synthesis polymerases. To further understand the mutagenic response to DNA damage, we screened a collection of 4848 haploid gene deletion strains of Saccharomyces cerevisiae for decreased damage-induced mutation of the CAN1 gene. Through extensive quantitative validation of the strains identified by the screen, we identified ten genes, which included error-prone post-replication repair genes known to be involved in induced mutation, as well as two additional genes, FYV6 and RNR4. We demonstrate that FYV6 and RNR4 are epistatic with respect to induced mutation, and that they function, at least partially, independently of post-replication repair. This pathway of induced mutation appears to be mediated by an increase in dNTP levels that facilitates lesion bypass by the replicative polymerase Polδ, and it is as important as error-prone post-replication repair in the case of UV- and MMS-induced mutation, but solely responsible for EMS-induced mutation. We show that Rnr4/Polδ-induced mutation is efficiently inhibited by hydroxyurea, a small molecule inhibitor of ribonucleotide reductase, suggesting that if similar pathways exist in human cells, intervention in some forms of mutation may be possible.
The stability of the genome is critical for life, but mutation occurs as a result of both exogenous and endogeneous mutagens . Mutations contribute to ageing , oncogenesis, tumor progression [3–5], and resistance to chemotherapy [6–8]. Ironically, because many chemotherapeutic drugs are mutagens, they may themselves play a role in inducing the mutations that give rise to resistance and therapy failure [9–11].
For many years, it was assumed that mutagens induced DNA damage, and the damaged DNA caused the replicative polymerases to make errors during DNA replication. However, it is now understood that most types of DNA damage inhibit DNA replication, and that the cellular response to this inhibition is critical for the induction of mutation. The cellular response to replication blocks in the yeast Saccharomyces cerevisiae has been intensively studied and is mediated in part by the intra-S checkpoint response  which helps to stabilize stalled replication forks, inhibit late origin firing, and induce an S-phase specific increase in the levels of dNTPs [13–16].
In addition, for many types of DNA damage the induction of mutation requires the nonessential specialized translesion synthesis (TLS) polymerases Rev1, Polζ, or Polη, encoded by REV1, REV3 and REV7, and RAD30, respectively. These genes are part of the RAD6 genetic epistasis group [17–19], which is responsible for post-replication repair (PRR), a process that converts low molecular weight DNA fragments into higher molecular weight DNA after genome replication is complete . PRR likely involves the filling in of DNA single-stranded gaps that are created after synthesis is re-initiated downstream of a replication block [21–23]. However, the TLS polymerases are not required for mutation induced by some mutagens, such as the alkylating agent EMS , suggesting that other pathways of induced mutation exist.
Using the complete set of systematic gene deletion mutants created by the Saccharomyces Genome Deletion Project [25,26] we have performed a screen for genes required for UV- and MMS-induced mutation. We identified genes that appear to act in two independent pathways: the known mutagenic PRR pathway and a novel pathway that appears to induce mutation at damaged DNA by up-regulating dNTP levels and facilitating TLS by the replicative polymerase Polδ. We discuss the importance of these pathways in inducing mutation and the possibility of inhibiting them to eliminate mutation induced by chemotherapy.
Compounds used in this study were obtained from commercial sources as follows: lcanavanine sulfate, US Biological; 5-FC, Acros; MMS, Aldrich; EMS, Acros; G418, RPI; and HU, US Biological. Yeast were cultured using standard methods at 30 °C in Yeast extract/Peptone/Dextrose (YPD) or Synthetic Complete (SC). Canr mutants were selected on SC-Arg media containing 60 mg/L l-canavanine sulfate. lys2ΔBgl and hom3–10 revertants were selected on SC media lacking lysine or threonine, respectively; 5-fluorocytosine (5-FC) resistant colonies were selected on SC media containing 100 µg/mL 5-FC. UV irradiations (UV-C, 254 nm) were performed with a G8T5 germicidal tube (Ushio America, Cypress, CA). Strains used in this study are listed in Table S1. Screening experiments were performed using a 4848 member library of nearly all the nonessential yeast gene deletion strains in the genetic background BY4741 obtained from Open Biosystems (Huntsville, Alabama). Characterization of individual gene deletions identified by the screens were performed in strain RDKY3615  (MATa ura3–52 leu2Δ1 trp1Δ63 his3Δ200 lys2ΔBgl hom3–10 ade2 Δ 1 ade8 hxt13URA kindly provided by R.D. Kolodner, UCSD). Characterization of induced mutation frequencies was performed in strains RDKY3615 and S18-1 (MATα ade5 lys2insE trp1–289 his7–2 leu2–3,112 ura3–52 kindly provided by D. Gordenin, NIH/NIEHS). All genomic DNA for strain constructions and mutation spectra determinations was prepared using a standard phenol/chloroform extraction method . Nucleotide pool measurements were determined as described previously .
The library of yeast deletion mutants was inoculated into YPD in 96-well microplates and incubated at 30 °C for 2 days. Five microliters of the resulting saturated culture of each strain were plated onto YPD, YPD containing 0.015% MMS, and YPD that was subsequently irradiated with UV light (50 J/m2). Plates were incubated at 30 °C for 2 days and photographed. The plates were replica plated onto SC-Arg containing canavanine, incubated at 30 °C for 3 days and photographed to score the number of CAN1 mutants.
The screen identified 305 deletion strains which qualitatively appeared less mutable than wild type cells. Thirty-eight strains were found to require arginine for growth and were not further characterized, as their isolation was assumed to be an artifact of the SC-Arg media required for Canr selection. Of the remaining 267 strains, 151 had an apparent UV-induced Canr forward mutation frequency that was more than 2-fold lower than that observed for wild-type. These 151 library deletion strains (BY4741 strain background) were re-constructed in strain RDKY3615. Genomic DNA from the selected library strains was isolated and used as a template for PCR amplification of the gene specific kanMX4 cassette flanked by ~50 bases of homology to the upstream and downstream region of the gene. Double stranded DNA cassettes were transformed into RDKY3615 and gene disruptants were selected on G-418 using standard protocols . The MMS- and UV-induced mutation frequencies of multiple, independent transformants were tested using the high throughput mutation assay described above. A representative transformant was selected for further quantitative mutation assays. For each strain, the spontaneous mutation frequencies were re-determined using three independent clones, and the UV-induced mutation frequencies were re-determined using one clone. Of the 151 library deletion strains selected, only 31 mutants had a UV-induced mutation frequency that was 2-fold lower than wild-type cells. The majority of the strains, ~60%, exhibited a UV-induced mutation frequency that was similar to wild-type despite the corresponding BY4741 strains exhibiting significantly reduced frequencies. For about 10% of the library deletion strains examined, PCR amplification with the gene flanking primers resulted in wild-type products rather than the kanMX4 cassette, indicating that the strains no longer carried the expected gene deletions. For the 31 strains that showed a ≥2-fold reduced induced mutation frequency in the experiment with a single clone, we next re-determined the mutation frequencies using 3 independent clones. A subset of the 31 strains showed induced mutation frequencies within error of wild-type frequencies, and these strains were eliminated from further analysis. Gene deletions were also confirmed at this stage by PCR for the presence of the gene deletion cassette and the absence of wild-type gene. A subset of strains did not pass the PCR confirmation test, and these strains were removed from further analysis. In all, 20 strains were identified with a properly integrated kanMX4 cassette and ≥2-fold average reduction in the UV-induced mutation frequency as compared to wild type. These 20 strains were subjected to more rigorous analysis by determining UV-induced (36 J/m2) mutation frequencies at other loci using reversion of the lys2ΔBgl allele to allow growth on media lacking lysine, reversion of the hom3–10 allele to permit growth in the absence of threonine, and forward mutation of genes involved in resistance to 5-FC.
To characterize spontaneous mutation, cells from saturated cultures (grown for 24 h) were harvested by centrifugation, washed and resuspended in saline, and plated on media containing canavanine to identify Canr mutants and on SC-Arg to determine the number of viable cells. To characterize induced mutation, similarly prepared cells were plated on SC-Arg plates and irradiated with varying doses of UV light. Cells were allowed to recover for 6 h at 30 °C after which they were washed off the SC-Arg plates and re-plated onto Canr selective media to identify mutants and onto SC-Arg to determine the number of viable cells. Colonies were counted after three days at 30 °C. Spontaneous mutation frequencies were determined by fluctuation analysis using the method of the median  from 9 independent cultures determined 2–3 times. Induced mutation frequencies were determined by dividing the number of Canr mutants observed by the number of viable cells and subtracting the contribution of the preexisting, i.e. spontaneous, mutations. Induced frequencies were determined from 3–8 independent cultures. Mutation frequencies at other loci were examined similarly using appropriate media for selection of mutants and SC for determination of viability.
Mutation spectra of canavanine resistant (Canr) clones were determined by DNA sequencing of the entire 1.8 kb CAN1 gene amplified from mutant colonies obtained from deletion strains in the RDKY3615 genetic background. Mutants were collected from plates generated during determination of induced mutation frequencies. Colonies were re-streaked to canavanine containing media, and genomic DNA was extracted. The primers used for PCR amplification were 5’-GTA AAC CGA ATC AGG GAA TCC and 5’-CTA CTC CGT CTG CTT TCT TTT C; and the resulting PCR product was sequenced using 5’-CGA AAT GGC GTG GAA ATG TG and 5’-CAA TGT AGA AGG TTA AGA TAC GG.
To identify non-essential genes involved in induced mutation, we performed a qualitative screen for deletion mutants that are unable to induce mutations in the CAN1 gene in response to UV or MMS treatment. Mutations in CAN1, typically substitutions or insertions/deletions, confer the easily selectable phenotype of canavanine resistance (Canr) . Strains were spotted onto permissive media, treated to induce DNA damage, and transferred to Canr selective media. Growth of each strain was evaluated and compared on the permissive and selective plates. Three different phenotypes were observed: (1) sensitivity to DNA damage, (2) little to no sensitivity to DNA damage and wild-type-like or greater growth on selective media, and (3) little to no sensitivity to damage and less than wild-type-like growth on selective media (Figure S1). Two hundred sixty-seven strains exhibiting the latter phenotype in both the MMS- and UV-induced mutation screens were identified and tentatively assigned as deficient in inducing mutation in the CAN1 gene (Table S2).
We next validated the reduced mutation phenotype of the identified strains by determining quantitative mutation frequencies under both spontaneous and UV-induced conditions. We found that 151 of the 267 identified strains had an apparent UV-induced Canr forward mutation frequency that is more than 2-fold lower than wild-type, with similar trends observed for the spontaneous mutation frequencies (Table S2).
We next examined whether the observed decreases in mutation frequencies resulted from disruption of the expected open reading frame. Deletion mutants corresponding to the 151 library strains were re-constructed in strain RDKY3615. Analysis of Canr mutation in these strains identified 20 with a UV-induced mutation frequency at least 2-fold lower than wild-type (Table S3). Included among these 20 strains are those previously implicated in induced mutation, including RAD5, RAD6, RAD18, REV1, REV3, REV7, POL32, and SRS2. These constitute all of the key non- essential components of mutagenic PRR, and as they have already been well characterized, they were not characterized further here.
For the 12 remaining genes, we next examined UV-induced reversion of the +4 lys2ΔBgl and +1 hom3–10 frameshift alleles, as well as spontaneous and UV-induced forward mutation of genes conferring resistance to the toxic cytosine analog fluorocytosine (5-FC) (Table 1 and and2).2). Resistance to 5-FC may be acquired through point mutations in FCY1, a cytosine deaminase or FCY2, a purine-cytosine permease (unpublished data). These assays are differentiated according to whether mutation results in a gain-of-function (lys2ΔBgl and hom3–10) or loss-of-function (CAN1 and 5-FC), and together these assays allowed us to differentiate between strains that are truly defective for induced mutation and those that are only apparently defective due to artifacts, such as polyploidy (multiple copies of a loss-of-function allele will result in apparently reduced mutation frequencies ). Eight deletion strains exhibited decreased mutation frequency in the loss-of-function assays but increased frequencies in the gain-of-function assays. Further analysis confirmed that these strains are deficient in the regulation of chromosome number and they will be described elsewhere.
Of the four remaining genes whose deletion results in reduced mutation in both gain- and loss-of-function assays (KAP123, VPS3, FYV6 and RNR4), KAP123 and VPS3 were not further characterized. For KAP123, the observed decreases in mutation frequency were restricted to the loss-of-function assays and were generally smaller than for the other identified genes. For VPS3, which encodes a protein involved in sorting and processing of soluble vacuolar proteins , wild-type mutation rates were observed when more time was allowed between UV irradiation and Canr selection (Figure S2), suggesting that the apparently reduced mutation rates are an artifact associated with phenotypic lag. This recovery of induced mutation was not observed for any of the other deletions strains identified. The remaining two genes, FYV6 and RNR4, appear to play a significant and general role in induced mutation.
The only previously proposed role of FYV6 that is obviously related to DNA damage tolerance or repair is regulation of non-homologous end joining . However, this function is unlikely to be related to the role of FYV6 in induced mutation as yku70Δ, lif1Δ, dnl4Δ and mre11Δ strains, all of which are known to be defective in non-homologous end joining, showed wild-type mutation rates (as expected since they were not isolated in the screen, data not shown). Thus, we determined whether FYV6 functions in the same pathway(s) as the known TLS polymerases. Genetic epistasis analysis using UV induced mutation of CAN1 shows that FYV6 and REV3 are not epistatic, suggesting that at least some of the mutations facilitated by Fyv6 are independent of Polζ (Figure 1A). Analysis of the Canr mutation spectra (Figure 2) showed that mutations induced in fyv6Δ are similar to those induced in wild-type cells (p>0.05) and clearly distinct from those induced in rev3Δ cells (p<0.02) . We also found that unlike deletion of REV3, deletion of FYV6 results in a small but significant reduction in EMS-induced mutagenesis (Figure 1J).
Rnr4 encodes one of the small subunits of RNR and is induced by DNA damage [36,37]. Previously, Brendel and co-workers demonstrated that deletion of RNR4 reduces UV-induced mutation and the authors suggested that this resulted from reduced translesion synthesis by Polζ . We first examined the UV sensitivity and mutability of a rad1Δ rnr4Δ mutant to determine whether Rnr4 is involved in NER, the primary repair mechanism responsible for removing UV-lesions. While the rnr4Δ rad1Δ double mutant was not significantly more sensitive to UV light (2.5 J/m2) than the rad1Δ mutant (as expected since the rnr4Δ single mutant is not sensitive to UV; Figure S3), it had a significantly lower UV-induced mutation frequency (Figure 1B). Thus, in the absence of NER, Rnr4 still appears to be required for the efficient induction of mutation.
rnr4Δ is also not epistatic with either rev1Δ or rev3Δ (Figure 1C and 1D). In fact, the reduction in induced mutation associated with deleting RNR4 is similar in rev3Δ cells and wild-type cells, suggesting that, like Fyv6, Rnr4 facilitates at least some mutations independently of mutagenic PRR. Deletion of RAD30 has no effect on UV-induced mutation in either the wild-type or the rnr4Δ strain (Figure 1E), thus no genetic interaction may be inferred. In addition, we were unable to construct the rnr4Δ pol32Δ mutant, which is consistent with a synthetic lethal interaction reported previously . However, while indirect, the HU data presented below suggest that Pol32 is not required for Rnr4-mediated mutation. Also as was observed with fyv6Δ, the UV-induced Canr mutation spectrum of rnr4Δ is clearly distinct from rev3Δ (p<0.01), but similar to wild-type cells (p>0.05) (Figure 2). Furthermore, in contrast to rev1Δ, rev3Δ or rad30Δ cells ( and Figure S4), EMS induces virtually no Canr mutations in rnr4Δ cells (Figure 1J). Given the phenotypic similarities, it is not surprising that FYV6 and RNR4 are epistatic for mutation at CAN1 (Figure 1F), possibly with the exception of high UV-irradiation conditions. These data indicate that RNR4 and FYV6 function in a pathway for induced mutation that is distinct from mutagenic PRR. However, the smaller effect of FYV6 deletion, relative to RNR4 deletion, suggests that Fyv6 participates in only a subset of the mutations induced by Rnr4.
Lastly, we examined the genetic interactions between RNR4 and RAD18. The double mutant was viable but grew much more slowly than either single mutant. In addition, we found that the rnr4Δ rad18Δ mutant is synergistically more sensitive to UV, MMS, or EMS exposure than either single mutant (Figure 3A). These results suggest that Rnr4 functions to help tolerate both spontaneous and induced DNA damage that may also be processed by PRR.
Since FYV6 appears to be involved in only a subset of the mutations mediated by Rnr4, we focused further efforts on the characterization of RNR4. At least some, and perhaps most of the mutations facilitated by Rnr4 are independent of the conventional TLS polymerases, suggesting that perhaps one or both of the replicative polymerases is responsible for the mutations induced by RNR4. To examine this possibility we characterized the pol3-1 and pol2-4 strains, in which the exonuclease proofreading activities of Polδ and Polε, respectively, have been disrupted. Both mutants exhibit increased spontaneous mutation (Table 1); and as reported previously [40,41], the effect is much larger with Polδ. In each case, deletion of RNR4 reduces the spontaneous hypermutability by ~3-fold (Table 1). The fact that deletion of RNR4 reduces mutation in both exonuclease deficient strains, but not in the wild-type strain, suggests that proofreading reduces the contribution of RNR4 to spontaneous mutation.
In the case of both UV- or EMS-induced mutation, disabling Polδ proofreading has a large effect, increasing mutation up to 70-fold. Interestingly, this hypermutability is dependent on RNR4 (Figure 1G and 1K); with the pol3-1 rnr4Δ double mutant showing UV- and EMS-induced mutation rates that are indistinguishable from those of wild-type cells. The mutation frequency of the pol3-1 rnr4Δ strain is not fully reduced to the level of the rnr4Δ strain which is consistent with the observation that deletion of RNR4 does not completely abolish the damage-mediated increase in dNTP levels (see below). Alternatively, this behavior may reflect an ability of exonuclease deficient Polδ to replicate through a lesion even in the absence of increased dNTP levels. In contrast, to RNR4, the hypermutability that results from rendering Polδ exonuclease deficient does not depend on Polζ as deletion of REV3 has no effect (Figure 1H and 1L). Deletion of REV3 in a pol3-1 rnr4Δ double mutant likewise has no effect on induced mutation (Figure 1G and 1H). Surprisingly, disabling Polε proofreading has no effect on UV- or EMS-induced mutation in wild-type cells (Figure 1I and 1K). Likewise, disabling Polε proofreading in rnr4Δ strains has no effect on UV-induced mutation (Figure 1I), however, it does result in a small increase in EMS-induced mutation (Figure 1K). These data suggest that RNR4 interacts specifically with Polδ, and that even in the absence of proofreading, Polε contributes little to induced mutation, and even then, only in the absence of RNR4.
To further understand the relationship between polymerase proofreading and induced mutation, we examined the effect of rendering each polymerase exonuclease deficient in a rad18Δ mutant lacking PRR (Figure 3B). Disabling Polε proofreading had no effect on the sensitivity of the rad18Δ strain to UV, MMS, or EMS. Again, in stark contrast, disabling Polδ proofreading strongly suppresses the sensitivity of the rad18Δ strain to all three damaging agents. These data suggest that Polδ acts upstream of PRR and that disruption of Polδ proofreading specifically facilitates a pathway involving Rnr4 that is distinct from PRR, but that acts to repair or tolerate the same lesions.
RNR4 mutants are known to progress more slowly through S-phase and show decreased RNR activity , which was suggested to result from reduced nucleotide pools in rnr4 cells [36,37]. To determine whether deletion of RNR4 reduces dNTP levels, we analyzed the amount of dNTPs that are present in wild-type and rnr4 strains, both in undamaged cells and in cells treated with 30 J/m2 UV light or 0.025% MMS (Figure 4 and Figure S5). The basal levels of dNTPs were the same in wild-type and rnr4Δ strains. However, while dNTP levels were elevated ~7-fold following UV-damage in wild-type cells, they were only elevated ~3-fold in rnr4Δ strains. Moreover, MMS damage induced a greater than 6-fold dNTP increase in wild-type cells, but only ~2-fold in rnr4Δ cells. This suggests that RNR4 may contribute to induced mutation by helping to mediate increases in dNTP levels when DNA is damaged.
Hydroxyurea (HU) is a small molecule chemotherapeutic that is thought to inhibit RNR and thus reduce intracellular dNTP concentrations . To confirm that decreased nucleotide pools can reduce mutation, and to determine whether induced mutation may be inhibited by a small molecule, we examined the effects of HU. Remarkably, HU significantly inhibits UV-induced mutation in a dose-dependent manner (Figure 5A). For example, at 100 mM HU, the Canr mutation frequency per viable cell after 67 J/m2 of UV light is reduced by ~100-fold, virtually to spontaneous levels. A similar reduction in UV-induced mutation frequency was observed with lys2ΔBgl reversion (Figure 5A), implying that HU inhibits the formation of both point mutations and frameshift mutations. In addition, HU also potently inhibits EMS-induced mutation (Figure S6), and its inhibition of UV-mutation is independent of RAD30, POL32, and REV3 (Table 3), further suggesting that HU specifically inhibits the same pathway that is mediated by Rnr4.
At high concentrations (~200 mM), HU induces S-phase arrest of S. cerevisiae , thus it is possible that an increase in repair during a prolonged S-phase is responsible for the effect of HU. To determine whether the HU-mediated inhibition of mutation (which occurs at concentrations as low as 10 mM), results from a prolonged S-phase, the UV-induced mutation frequency with or without HU was determined in an S-phase arrest deficient rad53Δ mutant. (Examining the same effect with the rnr4Δ mutant was not possible, since the rnr4Δ rad53Δ double mutant is inviable .) With or without Rad53, HU induces a profound reduction in mutation, suggesting that a checkpoint-mediated delay in S-phase is not required (Figure 5B).
To further support the suggestion that the anti-mutagenic effect of HU is mediated through reduced dNTP levels, we examined the effect of HU in the absence of SML1. Sml1 is a protein inhibitor of RNR, and nucleotide levels are known to be elevated in sml1Δ strains [44,45]. The anti-mutagenic effect of HU was clearly suppressed in the sml1Δ strain, consistent with the effect being mediated by a reduction in nucleotide pools (Table 3).
To identify proteins that are required for induced mutagenesis, we carried out a screen for non-essential genes in S. cerevisiae that when absent, render cells less able to mutate. The screen identified all of the key non-essential PRR genes previously known to be involved in mutation, RAD5, RAD6, RAD18, REV1, REV3, REV7, POL32, and SRS2. The screen also identified FYV6 and RNR4, which appear to cooperate in a second, distinct pathway of induced mutation. Fyv6 has no sequence homology to a known protein and its identification in high-throughput screens for mutants sensitive to the antifungal K1 killertoxin  and mutants with altered telomere length  suggest no obvious function. While it has been suggested that Fyv6 plays a role in regulation of non-homologous end joining in stationary cells , this activity does not appear to contribute to the protein’s role in induced mutation. However, FYV6 has also been isolated in a high-throughput screen for genes required for resistance to DTT . This screen also identified the genes encoding thioredoxin, which is an essential electron donor for RNR. Given the role of RNR in this pathway, one plausible role for Fyv6 is that it interacts in some manner with thioredoxin to help regulate RNR activity. Nonetheless, Fyv6 appears to play a role in only a subset of the mutations induced by Rnr4.
RNR4 encodes one of the four subunits of RNR. We find that deletion of RNR4 results in an up to 10-fold reduction in UV-induced mutation. Previously, Rnr3, another subunit of RNR has been implicated in MMS-induced mutation. Specifically, the inhibition of the TORC1 pathway, which controls transcription of RNR3 via Rad53, also results in a reduction of mutation in response to MMS . These data suggest that the effect on induced mutation may be generalized to other subunits of the RNR complex. We also demonstrated that RNR4 is required for UV- and MMS-induced increases in dNTP levels, consistent with the ability of the protein to stimulate TLS as proposed previously . Moreover, our data suggest that RNR4 functions in a previously unknown second pathway of induced mutation that involves TLS by the replicative polymerase Polδ. While this pathway contributes significantly to UV and MMS-induced mutation, it appears to be the major pathway responsible for EMS-induced mutation.
Interestingly, Polδ has been previously implicated in TLS across an abasic site in vivo, at least when the abasic site is located within a single-stranded region of a plasmid . Moreover, the hypermutability associated with rendering Polδ exonuclease deficient has been shown to depend on the activity of the S-phase checkpoint . The hypermutability is significantly diminished by deletion of DUN1, which encodes a substrate of Rad53 that regulates RNR activity, and this hypermutability is not affected by deletion of REV3. These data are consistent with an S-phase checkpoint-mediated upregulation of dNTP levels for Polδ mediated TLS.
A model consistent with all of the data is presented in Figure 6. In the case of UV-damage, the majority of photo-lesions are repaired by NER, but unrepaired photolesions, or lesions induced by other types of mutagens (i.e. MMS or EMS) that are not repaired prior to S-phase, cause replication forks to stall and the induction of the intra-S checkpoint response. As part of this response, Rnr4 is upregulated and the resulting increase in dNTPs facilitates error-prone TLS by Polδ, perhaps specifically during lagging strand synthesis where Polδ is thought to predominantly function [52,53]. This pathway is also favored by disabling Polδ exonuclease proofreading activity, presumably because this activity competes with TLS. If TLS does not occur, genome replication is completed by error-free or mutagenic PRR, perhaps involving gap filling after re-initiation of synthesis downstream of the blocking lesion. This explains why factors that favor Polδ TLS, including disabling its exonuclease activity or the presence of RNR4, strongly reduce the damage sensitivity of PRR deficient cells.
The data does not rule out a contribution of Rnr4 to Polζ-mediated mutagenic PRR, and examining this issue will require additional studies. What is clear from the data is that the Polζ and Polδ pathways are distinct and that the Polδ pathway relies on Rnr4, but not on Polζ. Both pathways, Rnr4/Polδ and mutagenic PRR, appear to contribute to mutation induced by many types of DNA damage, including UV and MMS, but the Rnr4/Polδ pathway alone appears to play an important role in EMS-induced mutagenesis. It seems likely that UV- and MMS-induced damage are more potent blocks of replication, and thus are more likely to require the specialized activities of the conventional TLS polymerases, while EMS induces damage that may be more easily replicated through by Polδ, at least in the presence of elevated levels of dNTPs. The effects of other mutagens may similarly partition between these two pathways depending on the ability of Polδ at elevated levels of dNTPs to synthesize through the associated lesion.
In addition to facilitating lesion bypass, the presence of elevated dNTP levels may also reduce the fidelity of Polδ acting on undamaged templates and may result in the induction of undirected mutations. In fact, elevated nucleotide pools have been associated previously with reduced fidelity of DNA replication possibly by making competitive the rate at which a mispair is extended with the rate at which it is excised . Additional work is required to deconvolute the contributions of lesion directed and undirected mutation mediated by the Rnr4/Polδ pathway.
The ability of HU to inhibit Rnr4/Polδ-mediated mutation is particularly interesting, as it suggests that if similar mechanisms are conserved in human cells, the mutations induced during chemotherapy might be inhibited. The potential therapeutic utility of such an approach depends on the proportion of mutations mediated by Rnr4/Polδ as opposed to PRR, which is currently under investigation. Whatever the result, the data suggest that there are a finite number of pathways, perhaps only two, that independently induce mutation. Thus, HU combined with an inhibitor of Polζ, might represent a therapeutic combination that efficiently eliminates all forms of chemotherapy-induced mutation. Studies directed towards reaching these goals promise to provide an unprecedented opportunity to understand genome instability in eukaryotic cells and also to intervene in mutation, and thus in the development and progression of cancer.
We thank Dinesh Kumar and Andrei Chabes (Umeå University) for performing the dNTP measurements. Financial support for this work was provided by the National Institutes of Health (GM68569 to F.E.R.) and a graduate fellowship from the California Breast Cancer Research Program (11GB-0004 to E.T.L.)
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.