Increase in dNTP pool size during the DNA damage response plays a key role in spontaneous and induced-mutagenesis in Escherichia coli

Abstract

Exposure of Escherichia coli to UV light increases expression of NrdAB, the major ribonucleotide reductase leading to a moderate increase in dNTP levels. The role of elevated dNTP levels during translesion synthesis (TLS) across specific replication-blocking lesions was investigated. Here we show that although the specialized DNA polymerase PolV is necessary for replication across UV-lesions, such as cyclobutane pyrimidine dimers or pyrimidine(6-4)pyrimidone photoproduct, Pol V per se is not sufficient. Indeed, efficient TLS additionally requires elevated dNTP levels. Similarly, for the bypass of an N-2-acetylaminofluorene-guanine adduct that requires Pol II instead of PolV, efficient TLS is only observed under conditions of high dNTP levels. We suggest that increased dNTP levels transiently modify the activity balance of Pol III (i.e., increasing the polymerase and reducing the proofreading functions). Indeed, we show that the stimulation of TLS by elevated dNTP levels can be mimicked by genetic inactivation of the proofreading function (mutD5 allele). We also show that spontaneous mutagenesis increases proportionally to dNTP pool levels, thus defining a unique spontaneous mutator phenotype. The so-called “dNTP mutator” phenotype does not depend upon any of the specialized DNA polymerases, and is thus likely to reflect an increase in Pol III's own replication errors because of the modified activity balance of Pol III. As up-regulation of the dNTP pool size represents a common physiological response to DNA damage, the present model is likely to represent a general and unique paradigm for TLS pathways in many organisms.

Maintenance of optimal intracellular concentrations of dNTPs is critical for faithful DNA synthesis. Ribonucleotide reductase (RNR) performs the key step in de novo synthesis of dNTPs from the corresponding ribonucleotides (1). This reaction is the rate-limiting process in DNA precursor synthesis and is controlled by multiple complex mechanisms, including transcriptional regulation and subcellular localization of RNR. Failure to control the dNTP pool results in genetic abnormalities (2–6).

In Escherichia coli, transcription of the nrdAB operon is coordinated to initiation of DNA replication with a burst of RNR synthesis every time a new replication fork is initiated (7, 8). The main RNR enzyme, NrdAB, is controlled by feedback regulation through binding of nucleoside triphosphate to two allosteric sites located in NrdA. The specificity site regulates substrate specificity and ensures that the enzyme produces a roughly equilibrated pool, but the activity site regulates the overall activity of RNR by negative feedback (1). Steady-state concentration of dNTP during replication only allows a small fraction of the chromosome to be duplicated. The E. coli's genome encodes an additional RNR enzyme, NrdEF, that, unlike NrdAB, is silent under normal growth conditions (5, 9, 10). In response to DNA damage, both NrdAB and NrdEF mRNA expression levels are increased in a damage-specific manner (1, 9): nrdAB is up-regulated following UV light exposure (11, 12), whereas nrdEF is induced under oxidative-stress conditions (9, 13). In a transcriptome analysis, the nrdAB locus appeared to be induced by UV irradiation in a LexA-independent manner (11), but a transcriptional fusion-based study concluded that the SOS response is partially involved (12).

DNA damage-dependent up-regulation of RNR transcription is a widely used regulatory pathway not only in bacteria but also in yeast and Drosophila (14, 15). In these organisms, there is a detectable global increase in dNTP pool size in whole extracts. It should be stressed that the local dNTP concentration (i.e., in the vicinity of the replication machinery) may be higher than the increase measured globally because it was suggested that RNR is part of a “replication hyperstructure” in E. coli (16). When measured globally in extracts, mammalian cells do not show a detectable increase in dNTP pool size in response to DNA damage (17). However, as suggested recently, in mammals dNTP biosynthesis may also be compartmentalized close to the damage sites during the DNA damage response (DDR) (18).

In E. coli, up-regulation of dNTP pool size via RNR overexpression increases the level of spontaneous mutagenesis (4, 5). Similarly, inactivation of the nucleotide diphosphate kinase (ndk) gene leads to dNTP pool imbalance, in particular a significant increase of dCTP, and to a 20- to 30-fold increase in spontaneous mutagenesis (19, 20). In yeast, DNA damage leads to an increase in dNTP pool size that is responsible for improved survival and increased mutagenesis (3). Moreover, a genome-wide search for yeast mutants with decreased induced-mutagenesis identified the rnr4 gene, which encodes a damage-inducible small subunit of RNR (21). Although the basal dNTP levels are similar in wild-type and rnr4Δ strains, following DNA damage the dNTP levels increase nearly twofold less in rnr4Δ and wild-type strains, respectively. This reduction in dNTP concentration accounts for a distinct reduction in UV and ethyl methanesulfonate-induced mutagenesis in the rnr4Δ strain (21, 22). As shown recently, a slight increase in dNTP concentrations in yeast allows the replicative DNA polymerases to bypass 4-nitroquinoline-oxide (NQO)–induced lesions, as evidenced by increased mutagenesis (23). The culprit lesion induced by NQO may be 8-oxo-G, because in vitro Pol epsilon can bypass 8-oxo-G lesions at DNA–damage-state dNTP concentrations, but it is blocked at S-phase dNTP concentrations (23).

In the present work, we analyze the effect on mutagenesis of the increase in dNTP pool size that occurs during the DDR. For this purpose, we express the RNR gene, nrdAB, from a low copy-number plasmid, leading to increased dNTP concentrations similar to the levels reached following UV irradiation. Under these conditions, we observe a ∼ 10-fold increase in spontaneous mutagenesis. This spontaneous mutator phenotype does not depend upon any of the specialized DNA polymerases. We also investigate the role of an increase in dNTP pool size on induced mutagenesis using three replication-blocking lesions: the two major UV light-induced DNA lesions, cyclobutane pyrimidine dimer (TT-CPD) and thymine-thymine pyrimidine(6-4)pyrimidone photoproduct [TT(6-4)], and the C8-guanine adduct formed by the chemical carcinogen N-2-acetylaminofluorene (G-AAF). We show that a physiologically relevant increase in dNTP pool size acts in strong synergy with the presence of the cognate specialized DNA polymerase to ensure efficient translesion synthesis (TLS). With respect to TLS, elevated dNTP pool size can be mimicked by the genetic inactivation of the proofreading function of Pol III, strongly suggesting that the dNTP pool size increase induces transiently a status of Pol III with attenuated proofreading function.

Results

UV-Light Treatment Increases RNR Protein Expression and dNTP Pool Size.

Because nrdAB gene expression is up-regulated at the transcriptional level by UV irradiation (11, 12), to measure the net effect of irradiation on dNTP pool size, a logarithmically growing E. coli culture of strain MG1655 was UV-irradiated at 40 J/m². This dose led to an approximately threefold increase in the level of protein NrdB (Fig. 1A). We observed a distinct increase in the concentration of all four dNTPs, ranging from 1.8- to 3.7-fold (average increase 2.7-fold) compared with the control level (Fig. 1B). In untreated cells, dNTP pools are not fully balanced, as dGTP and dATP are underrepresented and dCTP overrepresented, as observed previously in E. coli using a previously described method (4). In addition to NrdAB, E. coli encodes NrdEF, which lacks the domain for dATP feedback inhibition (24). The dNTP pool size was measured in logarithmically growing cells expressing NrdAB or NrdEF genes under the control of an isfopropyl-β-d-thiogalactopyranoside (IPTG) inducible promoter on a pBR-derived plasmid (5). The average increase in pool size in cells overexpressing NrdAB or NrdEF is ∼3.3- and 7.6-fold, respectively (Fig. 1B). The dNTP levels reached in UV-irradiated cells are similar to the levels reached in cells with nrdAB plasmid, except for dCTP, which is about twofold higher when expressed from the plasmid (Fig. 1B). In cells overexpressing NrdA_H59AB, a mutant that suppresses dATP feedback inhibition (25), the dNTP pools further increased twofold compared with wild-type NrdAB (Fig. 1B).

Fig. 1.

dNTP levels increase upon UV light exposure but are restricted by dATP feedback inhibition in E. coli. (A) Equal amounts of total cellular proteins from wild-type MG1655 strain grown to midexponential phase and subjected (+) or not (−) to UV light exposure (40 J/m²) were separated by SDS-10% PAGE. NrdB protein and control TrxA protein were detected by Western blotting. (B) dNTP pools were measured in a wild-type strain (MG1655) treated or not by UV light (40 J/m²) or containing a plasmid carrying either nrdAB, nrdEF, or nrdA_H59AB genes. The RNR genes under the control of the ptrc promoter were induced by addition of 1 mM of IPTG at mid-exponential growth. The relative fold-increase in dNTP levels relative to untreated cells is indicated on top of each bar. (C) Increased dNTP levels enhance spontaneous mutagenesis. The spontaneous mutation frequencies to rifampicin (Rif^R) is determined and plotted against the average dNTP concentration in strains expressing the different RNR genes as indicated B. (Inset) The relative fold-increase in spontaneous mutagenesis relative to untreated cells is indicated on top of each bar.

Increased dNTP Levels Enhance Spontaneous Mutagenesis: The “dNTP mutator” Phenotype.

Previous reports showed that in E. coli an increase in RNR levels enhances spontaneous mutagenesis (4, 5, 19, 20). To correlate dNTP pool size and spontaneous mutagenesis more directly, we measured the spontaneous mutation frequency to rifampicin resistance (Rif^R) in cells overexpressing nrdAB, nrdA_H59AB, or nrdEF. The increase in spontaneous mutagenesis is thus essentially proportional to dNTP pool concentration (Fig. 1 B and C).

To characterize the mutator phenotype associated with an increase in dNTP pool size, hereafter referred to as the “dNTP mutator” phenotype, we compared the frequency of spontaneous mutagenesis caused by elevated pool levels in strains defective in TLS polymerases, polB, dinB, and umuDC, either singly or in combinations, to the corresponding wild-type strain. The level of Rif^R mutagenesis is similar in all of these strains (Fig. S1), demonstrating that the dNTP mutator pathway does not involve the specialized DNA polymerases. We wanted to further investigate whether the SOS response in general is involved in the dNTP mutator phenotype. For this purpose, we compared a wild-type strain to a strain defective in SOS induction [lexA(ind-)] or a constitutively SOS-induced strain [lexA(Def)]. The level of spontaneous mutagenesis triggered by nrdEF overexpression is similar in all three strains, suggesting that no SOS-controlled function is involved for the dNTP mutator pathway (Fig. S2B). It should also be noted that overexpression of nrdEF has no effect on the SOS response itself, as judged from the level of RecA protein expression in strains that are either wild-type, lexA(ind-), or lexA(Def) (Fig. S2A). These results indicate that the SOS response in not involved in spontaneous mutagenesis in contrast to previously published work (4).

Effect of Increased dNTP Pool Size on the Efficiency of Specific TLS Pathways.

Outline and strategy.

In E. coli, the process of TLS has been dissected by means of plasmid probes containing single DNA lesions. It was found that TLS across replication-blocking DNA lesions absolutely requires prior irradiation of the E. coli host by UV light (26–29). It is generally accepted that the purpose of prior UV irradiation is to induce the SOS response to trigger expression of Pol V, the major TLS polymerase in E. coli (30). However, in addition to SOS induction, UV irradiation triggers many other responses, including dNTP pool size up-regulation. In an attempt to dissect the various contributions of UV irradiation to TLS, we choose to express the specialized DNA polymerases and the RNR genes separately from plasmid vectors to levels comparable to the levels achieved following UV irradiation. For Pol V we use a low copy-number plasmid vector pGB2, derived from pSC101 that expresses UmuD′ and UmuC from their natural promoter (pRW134) (31). To mimic the increase in dNTP pool size mediated by UV irradiation, we use plasmid pNrdAB (5), which increases the dNTP pool size by 3.3-fold on average, a value similar to the average fold-increase (2.7-fold) triggered by UV light at 40 J/m² (Fig. 1B).

Tools to measure TLS in vivo.

Over the years we have developed double-stranded ColE1-derived plasmid probes containing site-specific DNA lesions to study TLS pathways in vivo (26). In a typical experiment, a given plasmid probe is introduced into various E. coli strains that have or not been treated by UV irradiation before transformation (26–28). In the present work, we used three distinct plasmid constructs, each containing a representative replication blocking lesion, as detailed below, to investigate the effect of increased dNTP pool size on TLS. The single lesions are located in the N-terminal part of the lacZ gene in a way such that a specific TLS event yields a blue colony on an X-Gal plate. Replication of the undamaged strand yields lac-plasmid progeny because of a short-sequence heterology located opposite the lesion (see details in Materials and Methods and Fig. S3).

G-AAF: A Pol II-mediated −2 frameshift pathway within the NarI site.

The chemical carcinogen N-2-acetylaminofluorene forms stable covalent adducts at the C8 position of guanine (G-AAF) that trigger −2 frameshift mutation hot spots in sequences such as the NarI site (GGCG^AAFCC) (26, 32). In E. coli, these frameshift mutation pathways are mediated by Pol II (encoded by polB) via replicative slippage (28, 33). Surprisingly, this mutation pathway requires activation of the DNA damage response by UV irradiation (28, 33), despite Pol II being expressed constitutively at about 30 copies per cell. We wanted to explore the effect of increased dNTP pool size on this TLS pathway. When introduced in a strain overexpressing NrdAB from a low copy-number plasmid, the level of −2 frameshifts increases to over 40%, compared with 0.7% with the empty vector (Fig. 2A). However, −2 frameshift mutagenesis still heavily depends upon Pol II, as the level of TLS drops from ∼ 40% to ∼ 2% when polB is inactivated (Fig. 2A). Similar data are observed upon overexpression of NrdEF or NrdA(^H59A)B. In conclusion, physiologically relevant levels of dNTP pools, such as those reached by moderate UV irradiation, act in synergy with Pol II to trigger high levels of TLS across G-AAF lesions. Overexpression of NrdAB had no effect on Pol II expression levels.

Fig. 2.

Translesion synthesis pathways. (A) Effect of increased dNTP pool size on TLS pathways. For all three lesions, G-AAF, TT-CPD, and TT(6-4), we have compared the efficiency of TLS under the following conditions: (i) Strains expressing normal levels of dNTP (designated “low”) in the presence of the cognate specialized DNA polymerase. For G-AAF, Pol II is expressed from the chromosomal polB gene (∼50 molecules per cell); for TT-CPD and TT(6-4), Pol V is expressed from a low copy number pRW134 that produces 50–60 Pol V molecules per cell. (ii) Strains expressing NrdAB from a low copy-number plasmid, leading to ∼threefold higher than normal concentrations of dNTP (designated “high”) without the cognate specialized DNA. (iii) Strains expressing NrdAB from a low copy-number plasmid in the presence of the cognate specialized DNA polymerase. For all three lesions, a similar trend can be observed: efficient TLS requires, simultaneously, the presence of the cognate DNA polymerase and high dNTP levels. (B) Effect of proofreading inactivation (mutD5 allele) on TLS pathways. For all three lesions we have compared the efficiency of TLS under the following conditions: (i) Proofreading-proficient strains in the presence of the cognate specialized DNA polymerase. For G-AAF, Pol II is expressed from the chromosomal polB gene (∼30 molecules per cell); for TT-CPD and TT(6-4), Pol V is expressed from a low copy number pRW134 that produces 50–60 Pol V molecules per cell. (ii) Proofreading-deficient strains (mutD5) without the cognate specialized DNA. (iii) Proofreading-deficient strains in the presence of the cognate specialized DNA polymerase. For all three lesions a similar trend can be observed: efficient TLS requires, simultaneously, the presence of the cognate DNA polymerase and the inactivation of proofreading function of Pol III. Overall, the effect of proofreading inactivation mimics high dNTP levels (as seen in A).

TT-CPD: A replication-blocking lesion that is essentially bypassed in an error-free manner by Pol V.

We wanted to investigate the potential role of an increase in dNTP pool size on the bypass of a pyrimidine dimer using plasmid pSP-TT, which contains a single TT-CPD lesion across from a short sequence heterology (34). This plasmid specifically monitors both error-free and base-substitution TLS events. When either (i) the dNTP pool size is increased via plasmidic expression of NrdAB or (ii) Pol V is expressed from a low copy-number plasmid, we observe low levels of TLS compared with the TLS level reached when there is simultaneous increase of dNTP pool size and Pol V expression (Fig. 2A). The synergy between increased dNTP pool size and Pol V expression appears to mimic the response seen upon UV irradiation of a wild-type strain. Indeed, in a wild-type strain, the level of TLS raises from 0.3% up to 9.6% upon UV irradiation at 40 J/m².

TT(6-4): A highly distorting DNA lesion that is highly mutagenic.

We used a plasmid probe that specifically monitor the T->C transitions at the 3′-T position, an event that represents ∼90% of all TLS events (35). In vitro and in vivo TLS across TT(6-4) requires Pol V, the umuDC gene product (30, 35, 36). The extent of TLS events was measured in a strain expressing either Pol V or NrdAB separately or simultaneously from plasmid vectors. In a strain expressing Pol V at a level close to the physiological level in UV-irradiated cells, the extent of TLS across the TT(6-4) adduct is close to 1% (Fig. 2A). In a strain with increased dNTP pool size, the TLS level remains low (0.13%). However, when expressing both Pol V and NrdAB, the level of TLS reaches about 10%, a value similar to the level reached in UV-irradiated cells (33). A similar result is obtained upon expression of NrdEF instead of NrdAB. It can thus be concluded that the full capacity to bypass TT(6-4), as observed in UV-irradiated cells, can be recapitulated by the concomitant expression of Pol V and of NrdAB or NrdEF.

Stimulation of TLS by elevated dNTP levels can be mimicked by genetic inactivation of the proofreading function of Pol III holoenzyme (mutD5 allele).

At high dNTP levels, Pol III exhibits an enhanced capacity to misincorporate and extend from the resulting mismatch, a property reminiscent of a proofreading defect. To investigate the effect of a proofreading defect on lesion bypass, we compared the efficiency of TLS in the following strains: (i) in a proofreading-proficient strain expressing the specialized DNA polymerase required for a given lesion [i.e., Pol II for G-AAF or Pol V for TT-CPD and TT(6-4)]; (ii) in a proofreading-deficient strain (mutD5 allele) in the absence of the cognate DNA polymerase; and (iii) in a proofreading-deficient strain (mutD5 allele) in the presence of the cognate DNA polymerase.

For the G-AAF adduct, the cognate polymerase, Pol II, is expressed from the chromosomal polB locus, leading to ∼30 molecules per cell. Efficient TLS is only observed when Pol II is present in the context of a proofreading-deficient strain (Fig. 2B). Neither inactivation of proofreading in a polB strain nor the presence of Pol II in a proofreading-proficient strain supports efficient TLS in good agreement with our previous observations (Fig. 2B) (37, 38).

As far as TT-CPD and TT(6-4) lesions are concerned, the specialized polymerase that is required is Pol V encoded by the umuDC locus. The functional form of Pol V, the UmuD′₂.UmuC hetero-trimer (39), is essentially not expressed in bacteria unless the SOS response is induced (40). We thus use a low copy-number plasmid that produces Pol V at a level of ∼ 50–60 molecules per cell (31). We find that neither inactivation of proofreading (mutD5 allele) nor expression of Pol V per se promotes efficient TLS (Fig. 2B). In contrast, simultaneous expression of Pol V in a proofreading-deficient background brings about efficient TLS of both lesions.

We conclude that efficient bypass of all three lesions requires the presence of the cognate specialized DNA polymerase in a proofreading-deficient background. Proofreading inactivation appears to mimic the effect of elevated dNTP levels and it is thus tempting to conclude that high dNTP levels induce a transient proofreading-deficient status of Pol III (see model below).

Discussion and Conclusion

During Genotoxic Stress, the Increase in dNTP Concentration Enhances the Fixation by Pol III of its Own Replication Errors (dNTP Mutator Phenotype).

In the Pol III holoenzyme, there is a subtle balance between the 5′-3′ polymerase and 3′-5′ exonuclease activities that depends upon the dNTP concentration (41, 42). Under normal dNTP conditions, it has been estimated that this balance is such that on average 3–4% of incorporated dNTPs are released as dNMP during the processive DNA synthesis by the Pol III holoenzyme (43). In response to DNA damage, the increased dNTP concentration shifts the exo/pol balance of Pol III HE into a “more elongation/less proofreading” mode, to which we will refer as the “Pol III^TLS” mode (Fig. 3A). Data presented here show that the dNTP mutator pathway neither depends upon the specialized DNA polymerases (Fig. S1) nor upon any other SOS-controlled function (Fig. S2). The observed dNTP mutator pathway is thus likely to represent Pol III's own replication errors. The Pol III^TLS mode triggers a modest increase in misinsertions and their subsequent elongation (41). Similarly, it was previously suggested that in ndk strains the increase in spontaneous mutagenesis represents polymerase errors caused by limited editing capacity (20). However, more recently it was suggested that Ndk acts by preventing accumulation and incorporation of dUTP in vivo (44). In addition, to a net increase in the average pool size, it should be noted that all dNTPs are not increased to the same extent (from 1.8- to 3.7-fold), following UV-light induction. Mutations caused by pool distortion may also contribute to the observed mutator effect, as recently shown to be the case in yeast (45).

Fig. 3.

How does dNTP pool size expansion increase spontaneous and induced mutagenesis? (A) Spontaneous dNTP mutator pathway. Under normal dNTP concentrations, if Pol III makes a misinsertion error, the mispaired intermediate is removed by the proofreading function and T is correctly inserted leading to high-fidelity replication. In response to DNA damage, the increase in dNTP concentration modifies the activity of Pol III holoenzyme by shifting the balance between the exonuclease and polymerase functions into a more elongation/less proofreading mode (Pol III^TLS hatched oval). The arrow lengths represent the relative pol/exo activities of Pol III. Pol III^TLS represents a transient modification of Pol III's activity and not a distinct biochemical form of Pol III. In this mode, the rates of insertion errors and subsequent mispair elongation are increased compared with the normal mode of Pol III. (B) Induced mutagenesis. The cartoon pertains to TLS across a replication blocking lesion that requires Pol V [for example: TT-CPD or TT(6-4)]. For the sake of simplicity, neither the β-clamp nor the RecA filament are represented (51, 57). The thick dotted line represents the limited patch of DNA synthesis made by Pol V (TLS patch). Under normal dNTP conditions, the TLS patch generated by Pol V is too short to support elongation by Pol III. Thus, upon rebinding Pol III, exo degrades the TLS patch leading to an aborted TLS event. Under high dNTP conditions, the attenuated proofreading activity associated with Pol III^TLS (hatched oval) contributes to successful TLS by allowing elongation, rather than degradation, of short TLS patches. The overall scheme also applies for Pol II-mediated TLS across G-AAF adduct. In that case, Pol III^TLS specifically acts at the level of the first switch by inserting a C across the lesion site. The resulting replication intermediate undergoes a slippage reaction before Pol II association and elongation (58–60).

The dNTP mutator pathway should not be mistaken with the previously described SOS mutator phenotype (46–48), which pertains to specific E. coli strains that carry SOS constitutive recA alleles, such as recA730. In such strains, Pol III-mediated replication errors are fixed into the newly synthesized strand via Pol V-mediated TLS reaction (47, 49, 50).

In conclusion, the dNTP mutator phenotype is defined as a physiological response that occurs in cells that experience genotoxic stress. The increase in dNTP pool size that accompanies genotoxic stress enhances the risk of misinsertion and mispair extension by Pol III on lesion free template DNA (Fig. 3A).

Transient Attenuation of Pol III's Proofreading Function Is Critical for Efficient TLS.

Can we derive a model that accounts for the effect of increased dNTP pool size on TLS across replication-blocking lesions? Enhanced dNTP pool size may exert a subtle effect of Pol III's activity, turning it into more elongation/less proofreading mode, referred to as Pol III^TLS. This hypothesis is strongly supported by the demonstration that for all TLS pathways tested, the effect of high dNTP pools can be mimicked by the genetic inactivation of Pol III's proofreading function (mutD5 allele) (Fig. 2B). We hypothesize that Pol III^TLS, because of its attenuated proofreading function, slightly increases the steady-state concentration of TLS intermediates (Fig. 3B). We define TLS intermediates as the replication products that are generated from the limited replication activity of the specialized DNA polymerases in the vicinity of a DNA lesion. The size of the TLS patch, made by the specialized DNA polymerase when it dissociates, will determine whether the replicative DNA polymerase extends or degrades the TLS intermediate. The increase in dNTP pool size that is part of the DDR, by shifting the activity balance of the replicative polymerase toward the elongation mode, stimulates the overall process of TLS (Fig. 3B). However, we cannot rule out the possibility that increased dNTP levels stimulate TLS by acting on the specialized DNA polymerases themselves, in addition to their action on the replicative DNA polymerase.

For both TT-CPD or TT(6-4) lesions, TLS data show a robust synergy between an increase in dNTP pool size and the necessity of Pol V expression (Fig. 2A). In fact, UV irradiation of wild-type E. coli cells, a condition that is necessary for efficient bypass, brings about simultaneously these two critical factors: (i) the increase in dNTP pool size via transcriptional induction of the nrdAB locus in an SOS-independent way, and (ii) the expression of Pol V, the specialized DNA polymerase, that is encoded by the SOS-controlled umuDC operon. Although simultaneous expression of Pol V and NrdAB synergistically increase induced-mutagenesis (Fig. 2A), we have not formally established that the present reconstruction is equivalent to the response triggered by UV irradiation.

As far as G-AAF bypass is concerned, we see a similar response despite the fact that the specialized DNA polymerase that is involved is Pol II, not Pol V. Indeed, a robust increase in TLS is observed upon dNTP pool size increase (Fig. 2A). For G-AAF, the expression of Pol II from its chromosomal polB gene is necessary and sufficient (Fig. 2A). The key replication intermediate generated by Pol III^TLS is likely to be the insertion product of C across G-AAF, a replication intermediate that has been visualized in vivo when the proofreading function is genetically inactivated (38). Subsequently, this intermediate isomerizes into a slipped structure that constitutes the specific Pol II substrate for −2 frameshift mutation (51). In a wild-type strain, the level of −2 frameshift mutagenesis is low unless the cells have been pretreated by UV irradiation (33). We suggest that the role of UV irradiation is not to further enhance the expression of Pol II, but instead to induce the dNTP pool size increase, which in turn triggers the activity change from Pol III to Pol III^TLS to increase the steady-state concentration of the key replication intermediate that will subsequently be elongated by Pol II.

For all three lesions, an increase in dNTP pool size turns out to be phenotypically equivalent to proofreading inactivation (Fig. 2). A typical TLS pathway entails at least two DNA polymerase switches, first from Pol III to a TLS Pol and second from the TLS Pol back to Pol III (Fig. 3B). High dNTP levels generate a form of Pol III that has reduced proofreading/enhanced elongation properties (Pol III^TLS). The concept that proofreading attenuation may in principle play a key role to facilitate TLS has been suggested many years ago (52, 53). Here we provide experimental evidence that increased dNTP pool size may lead to proofreading attenuation of Pol III, thus contributing to replication across lesions by preventing key TLS intermediates to be degraded. These replication intermediates are either substrates for or products of the specialized DNA polymerases that allow the TLS process to be completed successfully (Fig. 3B).

The present study shows that, in vivo, efficient TLS requires a synergistic cooperation between a specialized DNA polymerase and a transient form of the replicative DNA polymerase, with modified activity balance between synthesis and proofreading. This change Pol III activity, mediated by an increase in dNTP pool size, leads to proofreading attenuation. Although the work is performed in E. coli, we believe that it represents a unique and general paradigm for TLS pathways in many organisms that, like E. coli, induce dNTP pool levels either globally or locally during DDR.

Materials and Methods

Bacterial Strains and Plasmids.

The bacterial strains and plasmids used in this study are listed in Table S1. The RNR genes expressed under the control of the ptrc promoter were induced by addition of 1 mM IPTG. Mutation CAC to GCC that changes His into Ala at position 59 of NrdA was introduced by PCR oligonucleotide site-directed mutagenesis on pSMG12, leading to plasmid pSMG20 using a QuikChange Kit (Stratagene). Plasmid pRW134, a pSC101 derived plasmid, expresses UmuC and UmuD′ from the natural promoter of the umuDC operon in ref. 31. Strain construction: polB, umuDC, and dinB genes were deleted in the MG1655 strain by gene replacement, as previously described (54).

Irradiation by UV Light.

Mid-exponentially growing cells resuspended in 10 mM MgSO₄ were irradiated at 40 J/m² with 254-nm UV light. The cells were allowed to recover for 30 min in fresh LB at 37 °C. All experiments were conducted under yellow light to prevent photo-reactivation of UV lesions.

Measurement of dNTP Levels.

Mid-exponential phase cells were filtered with nitrocellulose filters (Millipore; 0.45 μm) that were immersed in 700 μL of ice-cold extraction solution (12% trichloroacetic acid, 15 mM MgCl2). An internal standard, 10 nmols of dITP, was added to all samples to monitor loss during extraction. The following steps were carried out at 4 °C, as previously described (3). Separation and quantification of NTPs and dNTPs were carried out as previously described (55) on a PolyWAX LP column (PolyLC) using an HPLC system (Agilent 1100) equipped with an HP G1315A UV detector (55). Nucleotides were isocratically eluted with 2.5% acetonitrile/0.5 M potassium phosphate (pH 5).

Spontaneous Mutation Frequencies.

Spontanious mutation frequencies to rifampicin resistance (Rif^R) was determined as follows (19): For each strain, an average of 10 independent colonies were inoculated at 37 °C with agitation in 5 mL of LB medium overnight. Appropriate amounts were plated on selective plates containing rifampicin (100 mg/mL) or on nonselective LB plates to determine CFUs after appropriate dilution. Mutation frequencies (mean values ± SD) were calculated.

Translesion Synthesis Assays.

The efficiency of a given TLS pathway was measured using single-adducted plasmids as outlined in Fig. S3. The experimental protocol was described previously (28). Briefly, 1 ng of plasmid containing either a single G^AAF adduct in the NarI sequence (56), or TT(6-4) (35) or TT-CPD (34) lesions in the lacZ′ were used to electroporate E. coli strain before selecting the transformants on X-Gal–containing ampicilin plates. Great care is taken to prevent photo-reactivation of the TT-CPD lesion by visible light. The fraction of TLS is determined as the number of blue colonies over the total number of colonies. Each determination is the average of three independent experiments.