Spontaneous mutagenesis associated with nucleotide excision repair in Escherichia coli

Authors


  • Communicated by: Fumio Hanaoka

* Correspondence: Email: maki@bs.naist.jp

Abstract

The vast majority of spontaneous mutations occurring in Escherichia coli are thought to be derived from spontaneous DNA lesions, which include oxidative base damage. Systems for removing intrinsic mutagens and repairing DNA lesions contribute to the suppression of spontaneous mutations. Nucleotide excision repair (NER) is a general DNA repair system that eliminates various kinds of lesions from DNA. We therefore predicted that NER might be involved in suppression of spontaneous mutations, and analyzed base substitutions occurring spontaneously within the rpoB gene in NER-proficient (wild-type), -deficient and -overproducing E. coli strains. Surprisingly, the mutation frequency was lower in NER-deficient strains, and higher in NER-overproducing strains, than in the NER-proficient strain. These results suggest, paradoxically, that NER contributes to the generation of spontaneous mutation rather than to its suppression under normal growth conditions, and that transcription-coupled repair also participates in this process. Using E. coli strains that carried an editing exonuclease-deficient polA mutation, we further obtained data suggesting that unnecessary NER might account for these findings, so that errors introduced during repair DNA synthesis by DNA polymerase I would result in unwanted base substitutions. The repair system itself may thus be an important generator of spontaneous mutation.

Introduction

Spontaneous mutations arise under normal growth conditions, even in the absence of exogenous mutagens. Cells are exposed continuously to various kinds of unavoidable mutagenic events, including errors made by the DNA replication machinery, incorporation of mutagenic nucleotides into DNA, generation of spontaneous DNA lesions and deleterious actions of translesion DNA polymerases (Maki 2002). To counteract these hazards, cells possess elaborate mechanisms to correct replication errors, remove intrinsic mutagens, repair spontaneous DNA lesions and control translesion DNA polymerases. Thanks to the extremely high capacities of these mutation-avoidance mechanisms, the occurrence of spontaneous mutation is considered to be very low, and it is difficult to assess the extent to which an intrinsic mutagenic event culminates in a fixed spontaneous mutation. However, we have recently demonstrated that oxidative DNA damage is an important source of spontaneous mutations, especially hot-spot-type base substitutions, in aerobically growing Escherichia coli cells (Sakai et al. 2006). We speculated that cells might lack repair capacities against certain kinds of oxidative DNA lesions, generated specifically within particular sequence contexts and much less frequently than well-known DNA lesions such as 8-oxo-guanine, and that such rare types of DNA lesions might be an important cause of hot-spot-type base substitutions. Nevertheless, the origin of non-hot-spot-type base substitutions remains unclear.

Nucleotide excision repair (NER) is a general DNA repair system that consists of the global genome repair (GGR) and transcription-coupled repair (TCR) pathways (Selby et al. 1991; Sancar 1996; Hanawalt 2002). In E. coli, NER is a multi-step reaction facilitated by UvrA, UvrB, UvrC, UvrD, DNA polymerase I (Pol I) and DNA ligase. In the GGR pathway, the UvrA and UvrB proteins form a complex and recognize distorted DNA structures caused by DNA lesion. The UvrA comes off when UvrA2B2 complex is recruited to the damage site, and the UvrB stays there to form a stable pre-incision complex. Then, UvrC binds to the UvrB pre-incision complex, and incisions are made by the action of UvrC at the phosphodiester bonds on both 3′ and 5′ sides of the damage. The resulting oligo DNA with the lesion is removed by the helicase activity of UvrD protein, and the gap is filled by Pol I and DNA ligase. In the TCR pathway, DNA lesions are recognized by RNA polymerase. When the transcribed strand DNA contains a DNA lesion, RNA polymerase stops at the damage site. Mfd protein binds to the stalled RNA polymerase, removes the RNA polymerase from DNA, and stays at the damage site. The UvrA2B2 complex directly targets the Mfd protein at the damage site. NER was originally thought to be specific for bulky lesions, such as UV damage (Howard-Flanders & Boyce 1966; Howard-Flanders et al. 1966), but many kinds of DNA damage are now known to be repaired by this pathway (Lin & Sancar 1989; Branum et al. 2001). We therefore thought that NER might participate in suppression of spontaneous mutations. To test this, in the present study, we constructed E. coli strains carrying mutations in genes that function at each step of the NER pathways (uvrA, uvrB, uvrC and mfd), and analyzed a large number of base substitutions occurring spontaneously within the rpoB gene in NER-proficient (wild-type) and -deficient strains. Surprisingly, mutation frequencies in all of the NER-deficient strains were lower than that in the NER-proficient strain. UvrA, UvrB and UvrC participate in both pathways of NER, whereas Mfd functions only in TCR (Selby et al. 1991). The results suggest that both GGR and TCR contribute to generation of spontaneous mutations, rather than to their suppression, in normal growth conditions.

Escherichia coli NER exci-nucleases can excise oligomers of 12–13 nucleotides from undamaged DNA with a low but significant efficiency in vitro (Branum et al. 2001). This initially suggested two hypotheses: (i) that the threshold for DNA damage required to induce NER might be low, and the risk of unnecessary repair reactions occurring on undamaged DNA consequently high; and (ii) that if NER were induced unnecessarily on undamaged DNA, mutations might occur in proportion to the frequency of errors made during repair DNA synthesis by Pol I. We therefore predicted that a portion of spontaneous mutations would arise from unnecessary NER action, and that such mutations would reflect errors made during repair DNA synthesis. If NER can indeed occur unnecessarily under normal growth conditions, spontaneous mutations might be even more frequent in NER-overproducing strains. In agreement with this prediction, we found that more ssDNA regions were made in NER-overproducing strains than in a wild-type strain; moreover, the total mutation frequency was higher in NER-overproducing strains than that in wild-type, but decreased partially in a strain that carried an additional mfd-deficient mutation. To test the second hypothesis, we constructed a polA mutant lacking 3′→5′ exonuclease activity (editing Exo) and carrying either an additional NER-deficient mutation or an NER-overproducing plasmid. Using these strains, we analyzed spontaneous base substitutions occurring within the rpoB gene, and found that (i) the polA editing Exo mutant showed a relatively strong mutator phenotype, (ii) the additional NER-deficient mutation weakened the polA mutator activity, and conversely (iii) the additional NER-overproducing plasmid enhanced the polA mutator activity. These data indicate that NER can indeed occur unnecessarily; Pol I makes replication errors at a low but detectable frequency during repair DNA synthesis, and these errors are fixed as spontaneous mutations under normal conditions.

Results

Frequency of spontaneous base substitution in NER-deficient Escherichia coli strains is lower than that in an NER-proficient strain

To investigate the role of NER in spontaneous mutagenesis, we constructed NER-deficient E. coli strains and measured the mutation frequency in these strains under normal growth conditions using the target rpoB gene, which permits detection of many kinds of base substitutions leading to a rifampicin-resistant (Rifr) phenotype. To determine the Rifr mutation frequency, we carried out 20–40 independent experiments for each strain and calculated the median frequency. We expected that the mutation frequencies in the NER-deficient strains would be higher than that in the NER-proficient strain. On the contrary, however, the actual mutation frequencies in the NER-deficient strains ΔuvrA, ΔuvrB and ΔuvrC dropped to 15%–44% of that in the wild-type strain (Fig. 1, P < 0.01). Additionally, the mutation frequency in the TCR-deficient strain Δmfd was also decreased to 60% of that in the wild-type strain (Fig. 1, P < 0.05).

Figure 1.

Frequencies of Rifs→Rifr spontaneous mutation in wild-type (MK811) and NER-deficient (MK6008, MK6016, MK6024, MK6048, MK7607 and MK7610) E. coli strains. The rpoB mutation assay was performed as described in Experimental procedures. The data set for each strain is indicated as gray dots corresponding to mutation frequencies determined in 40 independent experiments, except that the data for MK7610 was from 20 independent experiments. Median values of the data sets are indicated as bars superimposed in the data sets and shown in a table below the data sets.

Among the NER-deficient strains, reduction in mutation frequency was in the following order: ΔuvrB > ΔuvrA > ΔuvrC > Δmfd. The largest mutation-reducing effect of ΔuvrB appeared to be dominant over ΔuvrA and ΔuvrC because the triple mutant strain ΔuvrABC showed a decreased mutation frequency similar to that in the ΔuvrB strain (Fig. 1). Escherichia coli possesses Cho protein, a homologue of UvrC. It was proposed that the Cho functions in NER for a particular type of DNA lesion which is not incised by the UvrC protein (Moolenaar et al. 2002). However, it seemed likely that the Cho protein was not suppressing the effect of ΔuvrC, because the spontaneous mutation frequency in a ΔuvrCΔcho double mutant was the same as that in the ΔuvrC strain (data not shown). The mutation-reducing effect of Δmfd was weaker than ΔuvrA, and the mutation frequencies in ΔuvrA Δmfd decreased to the same level as that in ΔuvrA.

Escherichia coli NER genes are antimutator genes

The reduction in spontaneous mutation frequency observed in the NER-deficient strains suggested an antimutator phenotype. However, there were two other possible explanations for the reduced frequency of Rifr mutations. One was that the NER-deficient strains might be hypersensitive to rifampicin, so that some of the rpoB mutations could lead to Rifr in an NER-proficient background but not in an NER-deficient background. To examine this possibility, we determined the minimal inhibitory concentration (MIC) of rifampicin for NER-proficient and -deficient strains. In fact, ΔuvrA and ΔuvrB strains showed a slightly lower MIC to rifampicin than that of the wild-type strain, while ΔuvrC and Δmfd strains displayed rifampicin sensitivity similar to that of the wild-type strain (data not shown).

The other possibility was that some rpoB mutations which led to Rifr in an NER-proficient background might be incompatible with NER-deficient mutations, so that NER-deficient cells carrying such an rpoB mutation could not form colonies on an LB plate containing rifampicin. To examine this possibility, we compared the site distribution of Rifr mutations in the rpoB target gene occurring in the NER-proficient strain with that found in the NER-deficient strains. Rifr mutations recovered from wild-type (180 mutations) and ΔuvrA (170) strains were identified by direct sequencing of the rpoB gene. As shown in Fig. 2, the vast majority of mutations in the NER-proficient strain were localized in a 200-bp segment starting from site 1525, 32 different base substitutions were mapped at 23 sites in this segment. The remainders were classified into two substitution types mapping at site 437 and 443. Of these 34 types of rpoB mutation, 20 were not found in the ΔuvrA spectrum (Fig. 2). We suspected that these 20 types might be incompatible with ΔuvrA, and analyzed colony-forming abilities and growth rates of Rifr strains carrying each of the rpoB mutations when the ΔuvrA mutation was introduced to the strains by P1-mediated transduction. Because we had no available stock of 437TG Rifr strain, other 19 types of Rifr mutations were examined. Six out of the 19 mutations showed no growth defect in the ΔuvrA background (class I, compatible mutations), whereas three were found to be incompatible with ΔuvrA (class II, incompatible mutations). The remaining ten types reduced the growth rate of ΔuvrA cells (class III, semi-incompatible mutations). A similar pattern of genetic interference with classes II and III mutations was observed for the ΔuvrB strain. Classes II and III mutations were not detected in the spectrum of Rifr mutations occurring in this strain. Therefore, the class II and perhaps some class III mutations are likely to be underscored when Rifr mutants are screened with the ΔuvrA and ΔuvrB strains. However, this genetic interference was not the only cause of the reduction of Rifr mutation frequency in the ΔuvrA and ΔuvrB strains: we also found that the frequencies of several class I mutations were lower in the ΔuvrA and ΔuvrB strains than in the wild-type strain.

Figure 2.

The site-distribution of Rifs→Rifr spontaneous mutations occurring in wild-type and ΔuvrA E. coli strains. Alterations in the rpoB sequence were determined for 180 and 170 Rifr mutants derived from MK811 and MK6008, respectively. The site and the type of base substitution for each sequence alteration is indicated at the bottom of the figure. For example, 436GT means G→Τ base substitution at the site 436. The mutation frequency for each site was calculated by multiplying the median mutation frequency shown in Fig. 1 by the relative occurrence of mutation at the individual site.

In contrast to the ΔuvrA and ΔuvrB strains, the ΔuvrC and Δmfd strains did not show genetic interference with classes II and III Rifr mutations. Consistent with this observation, the spectrum of Rifr mutations occurring in the ΔuvrC and Δmfd strains contained all classes of mutations, and the frequencies of most of the different mutations in these NER-mutant strains were lower than those in the NER-proficient strain (Fig. 3). From these data, we concluded that mutants of NER genes show an antimutator phenotype for various kinds of base substitutions, suggesting that some mutagenesis occurs via normal NER machinery, even in cells not exposed to exogenous DNA-damaging agents. It is known that Mfd takes part only in TCR, while UvrA, UvrB and UvrC are involved in both GGR and TCR pathways. Therefore, the antimutator phenotype of NER-deficient strains was readily expressed when the TCR pathway alone was hampered, and was further strengthened by additionally blocking the GGR pathway.

Figure 3.

The site-distribution of Rifs→Rifr spontaneous mutations occurring in ΔuvrC and Δmfd E. coli strains. Alterations in the rpoB sequence were determined for 100 Rifr mutants derived from MK6024 and MK7607. As a control, the data for wild-type strain (MK811) shown in Fig. 2 is added. Other descriptions are the same as Fig. 2.

NER reactions trigger the SOS response in NER-overproducing strains

The above findings suggested that NER reactions occur under normal growth conditions, raising the possibility that a significant amount of spontaneous DNA damage might be processed by NER or that needless NER reactions might occur on undamaged chromosomal DNA. The former possibility seemed unlikely because none of the NER-deficient strains showed any retardation of cell growth. Furthermore, cells defective in repairing spontaneous DNA damage often show a mutator phenotype, but the NER-deficient cells were found to be antimutator. If the latter possibility is the case, overproduction of NER proteins might lead to an increase of such needless NER reactions in cells. In the NER reaction, a ssDNA region of about 12 bp is made as a reaction intermediate and, if not quickly converted to dsDNA, directly or indirectly triggers the SOS response. We therefore examined whether the SOS response was induced when NER proteins were overproduced. The level of SOS response was monitored by β-galactosidase activity in cells carrying the plasmid pSK1002, in which the lacZ gene is placed downstream of the promoter and coding region of the umuDC gene, whose expression increases during the SOS response (Shinagawa et al. 1983). The results showed that the SOS response was markedly induced in UvrAB- and UvrABC-overproducing strains, but not in a UvrAB-overproducing strain harboring an additional ΔuvrC or Δmfd mutation (Fig. 4). These data suggest that the NER reaction is promoted in UvrAB- and UvrABC-overproducing strains even without DNA damage, and that Mfd is critical in this process. Therefore, how often the NER reaction occurs in normally growing cells correlated with the level of NER machinery under the circumstances where a very low but constant level of spontaneous DNA lesion was generated. This finding supports the idea that unnecessary NER reactions take place on undamaged chromosomal DNA.

Figure 4.

SOS response in NER-overproducing cells. β-galactosidase activities in wild-type strain (MK8000) and its derivatives carrying pBR322 (MK8002) or NER-overproducing plasmids (MK8004, MK8006, MK8014 and MK8016) were measured as described in Experimental procedures. As a control, UV-irradiated (100 J/m2) MK8000 cells were also subjected to the β-galactosidase assay. Four independent experiments were carried out with each strain, and an average of the β-galactosidase activity is indicated as a pillar. Bars on the pillars indicate the SDs.

Spontaneous base substitution frequency is higher in the NER-overproducing Escherichia coli strains than in an NER-proficient strain

If the NER reaction is a direct cause of mutation, cells with increased levels of NER proteins should show a mutator phenotype. To examine whether the frequency of spontaneous base substitution rises in NER-overproducing strains, we determined the mutation frequency using the same rpoB system. The mutation frequencies in UvrAB- and UvrABC-overproducing strains were 6.8-fold and 24-fold higher, respectively, than that in an NER-proficient strain (Fig. 5). The additional mfd mutation partially suppressed the increased mutation frequency in the NER-overproducing strain (Fig. 5). Considering these results, together with the data shown in Fig. 4, it is apparent that the mutator action caused by overproduction of NER proteins correlates with the level of NER reactions in the NER-overproducing strains.

Figure 5.

Frequencies of Rifs→Rifr spontaneous mutation in wild-type (MK6252) and NER-overproducing (MK6260, MK6268, and MK7686) E. coli strains. All the data sets were from 40 independent experiments. Other descriptions are the same as Fig. 1.

Since NER-overproducing cells readily induce the SOS response, there was a possibility that the mutator phenotype could be due to an SOS-induced error-prone DNA synthesis known as SOS mutagenesis (Hersh et al. 2004). However, the mutator phenotype of NER-overproducing cells was unchanged when the lexA1(ind) mutation, which cannot generate an SOS response, was introduced into the cells (data not shown). Thus, we concluded that the elevated mutation frequency observed in NER-overproducing strains was independent of SOS mutagenesis. Rather, a direct involvement of NER reactions in the elevated mutation frequency was strongly suggested from the pattern of site distribution of Rifr mutations induced in the NER overproducing strain. As shown in Fig. 6, NER overproduction resulted in more frequent occurrences of mutations at sites where the antimutator effect of NER-deficient mutation was observed.

Figure 6.

The site-distribution of Rifs→Rifr spontaneous mutations occurring in NER-overproducing E. coli strains. Alterations in the rpoB sequence were determined for 100 Rifr mutants derived from MK6252, MK6260 and MK6268. Other descriptions are the same as Fig. 2. The lower panel is the same as the upper panel but 10 times expanded in the vertical axis.

Replication errors during repair DNA synthesis in NER cause spontaneous mutations

If the unnecessary NER reaction is directly involved in spontaneous mutagenesis, there should be a discrete mutagenic step in the NER reaction. We hypothesized that errors made during repair DNA synthesis by Pol I might be a source of NER-dependent spontaneous mutations. To test this hypothesis, we constructed a polA (D424A) mutant lacking the 3′→5′ exonuclease activity of Pol I, in which errors during repair DNA synthesis are not proofread and would therefore be efficiently converted to mutations if not corrected by the mismatch repair system before the next round of DNA synthesis. Although the proofreading-defective polA mutant was previously reported to have no effect on the spontaneous mutation frequency, as determined by the trpE reversion system (Camps et al. 2003), we expected that the increased level of Rifr mutation frequency in NER-overproducing cells would be further enhanced by introduction of the polA mutant lacking 3′→5′ exonuclease activity into the cells. However, the polA (D424A) mutant strain itself showed a relatively strong mutator phenotype, yielding a 44-fold higher Rifr mutation frequency than that of a control (polA+) strain (Fig. 7). This unexpected finding prompted us to examine the effects of NER-deficient mutations and NER-overproducing plasmids on the mutator action of the polA (D424A) strain.

Figure 7.

Frequencies of Rifσ→Rifr spontaneous mutation in the proofreading-defective polA mutant E. coli strain (MK6236) and those with either NER-defective mutations (MK6240, MK6248, MK7676, and MK7682) or an NER-overproducing plasmid (MK6264). A control strain MK6220 carrying wild-type polA was also analyzed. All the data sets were from 40 independent experiments, except that the data for MK6248 strain was from 20 independent experiments. Other descriptions are the same as Fig. 1.

As shown in Fig. 7, the increased frequency of Rifr mutations in the polA (D424A) decreased to 30%–50% of the level observed when an additional NER-deficient mutation introduced into the polA mutator strain. This implied that 50%–70% of spontaneous mutations occurring in the polA mutator cells required the NER machinery, including Mfd protein. On the other hand, the mutator activity of polA (D424A) was further enhanced, about fourfold, by introduction of the NER-overproducing plasmid pUVRAB. Rifr mutations induced in the polA mutator strain were mapped within the rpoB gene by sequence analysis (Fig. 8). The pattern of site distribution of mutations overlapped partly with that observed in the polA+ NER-overproducing strain, but there were further new sites of mutation sites unique to the polA mutator strain. The former class of mutations and some of the latter class were suppressed when ΔuvrC was introduced into the polA mutator strain, indicating that these mutations are NER-dependent. From these results, we conclude that NER-dependent spontaneous mutagenesis involves errors made during repair DNA synthesis by Pol I.

Figure 8.

The site-distribution of Rifs→Rifr spontaneous mutations occurring in the proofreading-defective polA mutant E. coli strain (MK6236) and those with either ΔuvrC or Δmfd mutations (MK7676, and MK7682). Alterations in the rpoB sequence were determined for 100 Rifr mutants derived from each strain. Other descriptions are the same as Fig. 2.

Discussion

Among the numerous pathways of DNA repair in E. coli, NER is a unique and important mechanism for removing various kind of DNA damage, including bulky DNA adducts and UV lesions. Even low-dose UV irradiation leads to lethality in NER-deficient strains (Howard-Flanders et al. 1969). NER-deficient strains are much more susceptible to UV-induced mutagenesis than the NER-proficient strain. It is practically hard to induce mutation by even high-dose of UV-irradiation to the NER-proficient strain. Therefore, NER is a major factor that suppresses damage-induced mutagenesis when cells are exposed to exogenous mutagens. Although little is known about what kinds and extents of DNA damage are produced in normally growing cells in the absent of exogenous mutagens, it is likely that a proportion of any spontaneous DNA damage occurring under such conditions would be subjected to NER repair. Nevertheless, the role of NER in suppression of spontaneous mutations is not yet clearly understood. In the present study, we investigated whether NER contributes to suppression of spontaneous mutations. To our surprise, we found that NER relates to the generation of spontaneous mutation, rather than to its suppression. When compared with an NER-proficient E. coli strain, NER-deficient strains showed reduced frequencies of spontaneous base substitutions, and conversely, NER-overproducing strains displayed elevated frequencies of spontaneous base substitutions. We have provided evidence that the decreased level of spontaneous mutation frequency is due to an antimutator effect of NER-deficient mutations, which suggests that the NER system itself is a generator of spontaneous mutation.

A candidate for the of mutagenic activity underlying NER-dependent mutagenesis is repair DNA synthesis catalyzed by DNA polymerase I. Pol I is a high-fidelity enzyme with a proofreading function. We found a strong mutator phenotype in a polA mutant strain lacking the proofreading function of Pol I. This unexpected finding indicates that replication errors caused by Pol I could be a source of spontaneous mutations if they are not corrected by the built-in proofreading function. The elevated mutation frequency in the polA mutant strain was reduced by about 50% when NER-defective mutations were introduced. Thus, about half of the mutator effect of polA mutation is attributable to NER-dependent mutagenesis. It is probable that the remainder of the polA mutator effect is due to uncorrected replication errors that arise during repair processes other than NER. In contrast to NER-defective mutations, overproduction of NER proteins further exacerbated the polA mutator effect. From these observations, we conclude that NER-dependent mutagenesis is strongly affected by the fidelity of Pol I.

The very high fidelity of chromosomal DNA replication is achieved by the dam-dependent mismatch repair system in conjunction with the proofreading function of DNA polymerase III holoenzyme. The mismatch repair system is highly efficient, correcting more than 99.9% of mispairs and slippage errors. However, it acts on mismatches only for a short period after they are introduced during DNA replication. This is because MutH endonuclease specifically cuts hemimethylated GATC sequences surrounding the mismatch to initiate a long-patch repair process for mismatch correction; the mismatch repair system does not work once the GATC sequences are fully methylated by Dam methylase. Replication errors made during repair synthesis by Pol I for NER would therefore not be rectified by the mismatch repair system, unless the repair patch (12–13 nucleotides) contains GATC sequences. This notion predicts that the overall fidelity of DNA synthesis in NER reactions is much lower than it is in the chromosomal DNA replication.

A pre-requisite for NER-dependent spontaneous mutagenesis is that NER reactions do indeed take place in normally growing E. coli cells which are not exposed to exogenous mutagens. In the present study, we demonstrated that unnecessary NER reactions can occur in cells overproducing NER proteins. It seems plausible that such unnecessary NER reactions are occurring at a certain, albeit much lower, frequency in normally growing E. coli cells, which produce only very small amounts of UvrA and UvrB proteins. Although the targets of such “spontaneous” NER reactions in normally growing cells are unknown, we can consider several possible situations. First, unknown DNA lesions may exist on which NER acts preferentially in E. coli cells. If this is the case, NER-defective cells in which such lesions are unrepaired would show a growth defect, induce the SOS response, and express hyper-recombination and/or a mutator phenotypes. We observed neither growth defect nor SOS induction in a ΔuvrA strain (data not shown). The frequency of spontaneous recombination events, measured by a hemidiploid rpsL mutation assay (Kanie et al. 2007), in the ΔuvrA strain was identical to that in a wild-type strain (K. Hasegawa, S. Kanie, K. Yoshiyama and H. Maki, unpublished data). Therefore, the level of spontaneous DNA lesions subjected preferentially to NER seems negligible.

A second possible situation is that NER might target spontaneous lesions that are repaired preferentially by other systems, such as one of the many base excision repair (BER) pathways. In this hypothesis, such DNA lesions would be repaired properly in cells lacking the NER function. Oxidative DNA damage is the most abundant type of spontaneous DNA damage in aerobically growing E. coli cells, and is repaired with high efficiency by several BER pathways (Maki 2002, Sakai et al. 2006). Although it was previously shown that the UvrA–UvrB complex can bind to various kinds of oxidative DNA damage, it is unclear to what extent NER competes with BER pathways to repair such oxidative DNA lesions. Kamiya and colleagues reported very recently that an mutagenesis induced by 8-oxo-dGTP or 2-OH-dATP was weakened in uvrA and uvrB strains but not in a uvrC strain (Hori et al. 2007). Since they used the rpoB mutation assay, in which the detectable frequency of Rifr mutations in ΔuvrA and ΔuvrB strains is significantly lower than that in a ΔuvrC strain, their observation does not necessarily suggest an involvement of NER in 8-oxo-dGTP- or 2-OH-dATP-induced mutagenesis: they showed no incision on 8-oxo-dG-containing DNA after incubation with purified UvrABC exci-nuclease. Interestingly, in their control experiments without the mutagenic nucleotides, spontaneous Rifr mutation frequencies in uvrA and uvrB strains were 20%–30% of that determined with a wild-type strain. This is consistent with our observations, despite a different protocol for the mutation assay and a different genetic background of strains used in the two studies. However, the mutation frequency in a uvrC strain reported by Kamiya et al. was even slightly higher than the wild-type level. This is probably because of the relatively small number (5–11) of experiments that they repeated with each strain. To substantiate the nearly twofold difference in Rifr mutation frequency between wild-type and uvrC strains, we needed a data comprising 40 independent measurements.

A Third possible situation, which we consider more likely, is that a false recognition of DNA damage by NER might occur when undamaged chromosomal DNA undergoes a reversible structural distortion, upon DNA breathing or during various kinds of DNA transaction. The broad range of DNA-damage specificity of NER is based on the mode of damage recognition by NER proteins: the UvrA–UvrB complex recognizes a distorted DNA structure caused by DNA damage rather than the DNA damage itself (Ahn & Grossman 1996), and E. coli NER exci-nucleases can excise oligomers of 12–13 nucleotides from undamaged DNA with a low but significant efficiency in vitro (Branum et al. 2001). Furthermore, this hypothesis agrees well with our observation that overproduction of NER enzymes leads to an elevated spontaneous mutation frequency and a chronic induction of the SOS response. Finally, we should emphasize another possibility, namely that RNA polymerase might stall when it meets some specific DNA sequence or spontaneous DNA lesions, and the stalled RNA polymerase might then be recognized by Mfd protein, which initiates the TCR pathway of NER.

Whereas most pathways involved in spontaneous mutagenesis relate to chromosomal DNA replication (Maki 2002), NER-dependent mutagenesis may be independent from chromosomal DNA replication as well as from cell propagation. At a certain low level of occurrence, one round of spontaneous NER reaction could leave a mispair, as a repair DNA synthesis error, within a segment of 12–13 nucleotides spanning the repair patch. Such mispairs would accumulate with time in a cell even when it stops division, and mutations would become fixed upon the next round of DNA replication. Thus, NER-dependent mutagenesis would correlate with the lapse of time rather than the cycle of cell division. Although cells in logarithmic growth phase were used in this study, we expect that the contribution of NER-dependent mutagenesis would be a more significant component of spontaneous mutagenesis in stationary-phase cells, where chromosomal DNA replication is shut off but the genome is actively transcribed. According to a recent study with Bacillus subtilis, a time-dependent increase of the mutation frequency in mfd-deficient cells is lower than that in wild-type cells at stationary phase, and this concurs with the above-mentioned idea (Ross et al. 2006).

NER seems to be a double-edged sword: it acts to guard genome integrity when cells are exposed to exogenous mutagens on the one hand, but generates spontaneous mutations when it acts in cells unexposed to the mutagens on the other. It is known that the expression of NER genes is maintained at a low level (100 UvrA2UvrB complexes/cell) under normal growth conditions and is up-regulated sixfold by the SOS response when DNA replication is blocked by DNA damage (Crowley & Hanawalt 1998; Walker et al. 2000). Recently, two other mechanisms were reported to be involved in regulation of NER activity in E. coli cells. UvrA protein is very efficiently degraded by ClpXP protease (Neher et al. 2006). The two-component ArcA/ArcB system, which responds to the presence of oxygen, also negatively controls the expression of the uvrA gene (Ogasawara et al. 2005). The growth environment of bacteria in nature is basically inconstant, and the changes are often very rapid and severe. Such drastic environmental changes may alter the level of exogenous and spontaneous DNA damage, and cells might have evolved to survive this situation by developing an on-demand activation of NER. However, the basal level of NER activity, which is probably optimized for their survival strategy, appears to have the side-effect of causing spontaneous mutations in normally growing cells.

Experimental procedures

Bacterial strains and media

The bacterial strains used in this study are listed in Supplemental Table S1 and were all derivatives of an E. coli K12 wild-type strain, MG1655 (Guyer et al. 1981) obtained from the E. coli Genetic Stock Center, Yale University. An NER-proficient strain MK811 was previously described (Kanie et al. 2007). NER-deficient strains, MK6008 (ΔuvrA), MK6016 (ΔuvrB), MK6024 (ΔuvrC), MK6048 (ΔuvrA ΔuvrB ΔuvrC), MK7607 (Δmfd) and MK7610 (Δmfd ΔuvrA) were constructed by P1-mediated transduction (Miller 1992) with MK811 as a recipient, and P1vir phage lysates used were prepared with NER-gene deletion mutants obtained by the procedure of Datsenko and Wanner (2000). PCR primers used for the one-step mutagenesis are listed in Supplementary Table S2. NER-overproducing strains, MK6260 and MK6268 were constructed by transformation of MK811 with pUVRAB and pUVRABC plasmids, respectively. MK6264 and MK7686 were constructed from MK 6236 and MK7607, respectively, by introducing pUVRAB plasmid into the strains. As a control, MK6252 was constructed by introducing pBR322 into MK811. To construct strains carrying a polA mutation that hampers 3′→5′ exonuclease activity of Pol I, plasmid pPOLA2 was first introduced into MK811 and its NER-deficient derivatives, and then ΔpolA::Kmr was introduced into these strains by P1-mediated transduction with a P1vir lysate prepared with CJ278 (Joyce & Grindley 1984). As a control strain, MK6220 was constructed from MK811 by the same procedure except using pPOLA1. LB contained 1% (w/v) Bactotryptone (Difco, Detroit, MI), 0.5% (w/v) yeast extract (Difco) and 1% (w/v) NaCl. LB plates were solidified with 1.5% Bacto agar (Difco). Ampicillin and rifampicin were added to the medium when needed at concentrations of 100 µg/mL. Chloramphenicol, kanamycin, tetracyclin were added to the medium when needed at 25, 50 and 10 µg/mL, respectively.

Plasmids

NER-overproducing plasmids pUVRAB (carrying uvrA and uvrB) and pUVRABC (carrying uvrA, uvrB and uvrC) are derivatives of pBR322 and retain Tetr gene. To construct the plasmids, DNA fragments containing uvrA, uvrB, or uvrC gene with each promoter region and transcription terminator were separately amplified by PCR with MG1655 genomic DNA and primer sets shown in Supplementary Table S2. PvuI–uvrAAscI–NotI and NotI–ApaI–uvrBAseI DNA fragments were ligated with a larger PvuI–AseI segment of pBR322, and the resulting three-fragments recombinant plasmid is pUVRAB. pUVRAB was digested with restriction enzymes AscI and ApaI and ligated with an AscI–uvrCApaI fragment, resulting in pUVRABC. Plasmids pPOLA1 and pPOLA2 are derivatives of pACYC184 and retain Cmr gene and polA gene with its own promoter region. To construct plasmid pPOLA1, a BamHI–polA–HindIII fragment containing a whole coding sequence of polA gene, its promoter region and transcription terminator was amplified by PCR with MG1655 genomic DNA and a primer set shown in Supplementary Table S2 and ligated with a larger BamHI–HindIII segment of pACYC184. pPOLA2 was constructed from the pPOLA1 by replacing an A nucleotide at position 1271 in the polA coding sequence with a C nucleotide using a Quikchange Site-Directed Mutagenesis Kit (stratagene). The plasmid pSK1002(umuD-lacZ) (Shinagawa et al. 1983) used in the β-galactosidase assay is a gift from T. Hishida.

rpoB mutation assay

Escherichia coli cells were fully grown in 5 mL LB medium at 30 °C with agitation. After appropriate dilution of the cells, about 100 cells were inoculated into 5 mL of fresh LB medium. The cells were grown at 37 °C with agitation to OD600 = 1.0. After appropriate dilution or concentration of the cells, cells were plated on LB plates and selective plates (LB containing 100 µg/mL rifampicin) to determine the viable cell titer and a total number of Rifr cells in the culture, respectively. Colonies formed on the plates were counted after incubation for 24 h at 37 °C. The mutation frequencies were calculated by dividing the number of Rifr cells by that of total cells. For each strain to be examined, we repeated 40 times (in some cases, 20 times) the determination of mutation frequency, and a median of the mutation frequencies was obtained from the data. P values were calculated by the Mann–Whitney U-test (Siegel 1956). DNA sequence analysis of Rifr mutations was performed by a direct sequencing method (Kanie et al. 2007) using primer sets shown in Supplementary Table S2.

β-Galactosidase assay

Some modifications were added to the method described by Miller (1992). Escherichia coli cells were fully grown in LB at 37 °C with agitation. A measure of 80 µL of the culture were added to 5 mL of fresh glucose minimal medium. For SOS-uninduced cells, β-galactosidase activity was measured after incubation for 5 h at 37 °C. For SOS-induced cells, UV (100 J/m2) was irradiated to the culture after incubation for 3 h at 37 °C, and β-galactosidase activity was measured after further incubation for 2 h.

Acknowledgements

We are grateful to members of the Maki laboratories for helpful discussions and Dr Ian Smith for help in preparation of the manuscript. We thank Dr Takashi Hishida for providing a plasmid, Dr Kazuo Yamamoto and Dr Nora Goosen for providing E. coli strains. We acknowledge the financial support of Grants-in-Aid for Scientific Research on Priority Areas (17013060 to H. M.) from the Ministry of Education, Culture, Sports, Science and Technology of Japan.

Ancillary