Hypermutability and error catastrophe due to defects in ribonucleotide reductase Deepti Ahluwalia and Roel M. Schaaper1 Laboratory of Molecular Genetics, National Institute of Environmental Health Sciences, Research Triangle Park, NC 27709 Edited by Paul Modrich, Duke University Medical Center, Durham, NC, and approved October 1, 2013 (received for review June 7, 2013)

The enzyme ribonucleotide reductase (RNR) plays a critical role in the production of deoxynucleoside-5′-triphosphates (dNTPs), the building blocks for DNA synthesis and replication. The levels of the cellular dNTPs are tightly controlled, in large part through allosteric control of RNR. One important reason for controlling the dNTPs relates to their ability to affect the fidelity of DNA replication and, hence, the cellular mutation rate. We have previously isolated a set of mutants of Escherichia coli RNR that are characterized by altered dNTP pools and increased mutation rates (mutator mutants). Here, we show that one particular set of RNR mutants, carrying alterations at the enzyme’s allosteric specificity site, is characterized by relatively modest dNTP pool deviations but exceptionally strong mutator phenotypes, when measured in a mutational forward assay (>1,000-fold increases). We provide evidence indicating that this high mutability is due to a saturation of the DNA mismatch repair system, leading to hypermutability and error catastrophe. The results indicate that, surprisingly, even modest deviations of the cellular dNTP pools, particularly when the pool deviations promote particular types of replication errors, can have dramatic consequences for mutation rates.

T

he enzyme ribonucleotide reductase (RNR) is a critical cellular factor for the synthesis and control of the deoxynucleoside-5′-triphosphates (dNTPs), the building blocks for DNA replication and DNA repair (1). Specifically, RNR is responsible for the reduction of the ribonucleotides to the corresponding deoxynucleotides. In many well-studied cell systems, like the bacterium Escherichia coli, the yeast Saccharomyces cerevisiae, or mammalian cells, this reduction takes place at the level of the nucleoside diphosphates (NDP → dNDP), and each of four separate NDPs (ADP, GDP, CDP, and UDP) can serve as RNR substrate (1, 2). Following reduction, the dNDPs are converted to the corresponding dNTPs by nucleoside diphosphate kinase (NDK) (3). The DNA precursor dTTP is not generated directly through this pathway; instead, it is produced from dCTP via dUTP (4). RNRs have been subdivided into several classes, depending on the type of radical used in the catalytic reaction (1, 5–7). The class Ia RNR, as present in E. coli, yeast, and mammalian cells, employs a tyrosyl radical. Structurally, these RNRs are tetramers composed of a dimer of a large subunit (R1) and a dimer of a small subunit (R2). The large subunits contain the catalytic site and two allosteric regulatory sites (termed specificity site and activity site), whereas the small subunit contains the essential tyrosyl radical. The activity site is located at the N terminus of the R1 subunit and functions as an “on-off” switch: ATP binding leads to an active enzyme, whereas dATP binding inhibits the enzyme. Hence, by monitoring the ATP/dATP ratio, the enzyme aims to ensure an overall dNTP level that is presumably optimal for DNA replication (1, 8). The R1 specificity site, is a binding site for dATP, ATP, dGTP, and dTTP (1, 2, 9, 10); depending on which nucleotide is bound, the nearby catalytic site is conformationally primed to reduce a specific NDP substrate (ADP, CDP, GDP, or UDP). In this manner, this site regulates the enzyme such that the four dNTPs are maintained at their desired relative ratios. Although these two types of regulation were first described many decades ago,

18596–18601 | PNAS | November 12, 2013 | vol. 110 | no. 46

their precise mechanisms are still an issue of intense interest (1, 2, 9, 11, 12). New insights into the regulation of the dNTP pools inside the cell may be obtained through genetic studies by using mutants affected in their dNTP levels or ratios, including studies on RNR (11, 13). For E. coli, a large number of novel RNR mutator mutants were obtained after random mutagenesis of the nrdAB genes encoding the R1 (NrdA) and R2 (NrdB) subunits of RNR (11). Three distinct groups of mutators were found. The first group carried amino acid changes in the R1 activity-site domain, and these mutants were postulated to suffer from a lack of feedback inhibition by dATP, leading to significantly elevated dNTP pools, particularly dATP and dCTP (11). A second group was found, unexpectedly, in the R2 subunit, affecting amino acid residues near the R1/R2 interface directly adjacent to the R1 activity-site domain. Based on this proximity, a commonality of mechanism between the R1 activity mutants and the R2 mutants was postulated, which was supported by similar dNTP pool changes and similar mutational specificity. We postulated that upon dATP binding to the activity site, a joint R1/R2 conformational change occurs, leading to an inactive form of RNR, possibly by distorting the direct linkage between the radical center of R2 and the catalytic site in R1 (11). The third group of RNR mutator mutants was found at residues G295 and A301 in the R1 specificity site. These two residues are located in or near the base of an important regulatory loop (L2) as identified in recent structural studies (11, 14–16). Conformational changes of this loop upon binding of particular dNTPs are communicated to the catalytic site on the other side of the loop, thus determining which NDP substrate may bind (9). Because the dATP-mediated feedback mechanism at the activity site is still functional in these mutants, no large increases in the Significance The accuracy by which cells are able to replicate their chromosomal DNA is critical for the rate by which they produce mutations. To keep the mutation rate low, cells have developed several mechanisms including accurate base selection by DNA polymerase, exonucleolytic proofreading that can excise incorrectly inserted bases, and postreplicative DNA mismatch repair. Also important are the concentrations of the deoxynucleoside-5′-triphosphates (dNTPs) that serve as direct substrates for the DNA synthesis reaction. Here, we describe a case in which certain changes in the relative levels of the four dNTPs (dATP, dTTP, dGTP, dCTP) can profoundly disturb the overall replication fidelity by impairing the efficiency of each of the three main fidelity systems, leading to a case of extreme hypermutability. Author contributions: R.M.S. designed research; D.A. performed research; D.A. and R.M.S. analyzed data; and D.A. and R.M.S. wrote the paper. The authors declare no conflict of interest. This article is a PNAS Direct Submission. 1

To whom correspondence should be addressed. E-mail: [email protected].

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. 1073/pnas.1310849110/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1310849110

Results E. coli Mutator Mutants at the RNR Specificity Site: Exceptionally High Mutability for Rifr Forward Mutations. Among a group of 23 newly

discovered RNR mutators mutants, three were found to reside at the RNR specificity site: G295S, A301T, and A301V (11). As depicted in Fig. 1, the G295 and A301 residues are part of loop 2 in the RNR structure, which mediates the communication between the dNTP effector and the NDP substrate (9, 10). Presumably, the amino acid substitutions at G295 and A301 disrupt or alter this communication, leading to altered RNR output. The dNTP pool changes observed in these mutants, along with some of the activity-site mutants, are shown in Table 1. Clearly, although the activity-site mutators (exemplified here by the S37L and H46Y mutants) are characterized by robustly elevated dNTP levels, especially of dATP (fivefold) and dCTP (30-fold), the

Table 1. dNTP pools in nrdA mutator mutants nrdA allele Wild type G295S A301T A301V S37L H46Y

RNR site

dATP

dTTP

dGTP

None 11.9 ± 3.8 5.8 ± 2.0 Specificity 7.5 ± 1.4 10.3 ± 1.6 Specificity 6.9 ± 0.9 8.8 ± 1.1 Specificity 6.9 ± 2.2 9.9 ± 3.4 Activity 34 13 Activity 58 15

dCTP

3.1 ± 1.2 5.3 ± 2.0 2.3 ± 1.4 6.0 ± 2.6 4.7 6.9

2.4 3.7 8.0 8.4

n

± 2.4 11 ± 1.9 12 ± 2.7 5 ± 4.7 13 36 2 74 2

n, the number of independent measurements. All dNTP levels are relative to ATP, as described (14).

specificity-site mutators G295S, A301T, and A301V display only modest changes. In each of them, there is a decrease of dATP (∼twofold) and an increase of dTTP (∼twofold). The three mutants differ from each other in the dGTP and dCTP levels. Most notably, there is an increase in dGTP for G295S and A301V (∼twofold), and a dCTP increase (∼2.5-fold) for the two A301 mutants. These differential pool changes for the two groups of mutators result in differential mutational responses as shown in Table 2. The table shows the differential mutability for the two lac reversions, A·T → T·A and G·C → T·A, which occur at a single site in lacZ (17). These specificity differences can be rationalized based on the differential dNTP pool changes (11). However, a major unique feature of the G295S and A301V mutators reported here is the exceptionally high mutant frequency (a more than 1,000-fold enhancement) for the frequency of rifampicinresistant (Rifr) mutations, which are scored at a large number of sites in the rifampicin-resistant (rpoB) gene (18, 19). Similarly, large increases are observed when scoring for nalidixic acidresistant mutants mapping at multiple sites in the gyrA gene. These observations raised the question as to how to reconcile the high mutability with the modest dNTP pool changes. A further clue to the underlying mechanism was the generally small and heterogeneous colony sizes of the G295S and A301V mutants, while also including a subset of larger colonies. When checking these larger colonies, their mutator phenotype appeared to have been significantly diminished or even lost. These observations suggested a model in which G295S and A301V are exceptionally strong mutators, which because of their high mutation rate may suffer from accumulation of deleterious mutations. At the same time, the very high mutation rate is also responsible for the production of suppressor mutations, which reduce the mutation rate and restore normal viability. A precedent for this scenario of high mutability, reduced viability, and production of suppressor mutants exists in the example of certain proofreadingdeficient E. coli mutator mutants, as further discussed below. The Mismatch Repair Saturation Hypothesis. Exceptionally high mutability has been observed in certain other cases in which a primary increase in the number of replication errors leads to a compromised ability to perform postreplicative DNA mismatch repair (MMR) leading to a further, strong amplification of the mutator Table 2. Mutability of two classes of nrdA mutators nrdA strain

Fig. 1. Loop 2 region of E. coli NrdAB ribonucleotide reductase. The loop 2 is presented in dark purple color and is based on the RNR coordinates from refs. 14 and 6. Shown are the location of residues G295 and A301, which upon mutation (G295S, A301T, or A301V) produce a mutator phenotype (see Results for details). Also shown are a GDP substrate (Right) and the dTTP effector (Left), communication between which is mediated by loop 2 (9, 10).

Ahluwalia and Schaaper

nrd+ G295S A301T A301V S37L H46Y

lac G·C → T·A 1.4 2.3 3.0 2.3 25 53

± ± ± ± ± ±

0.6 1.1 1.2 0.9 4.0 23

lac A·T → T·A 0.9 7.4 9.6 22 5.5 12

± ± ± ± ± ±

0.6 3.7 4.7 7.0 0.8 3.9

Rifr 1.4 4,200 7.0 2,900 40 48

± ± ± ± ± ±

0.6 1,600 3.0 2,300 10 1.7

Mutants per 108 cells.

PNAS | November 12, 2013 | vol. 110 | no. 46 | 18597

GENETICS

overall dNTP levels are expected. Instead, changes in the relative ratios of the four dNTPs were found (11). In the course of further studies on specificity-site mutants, in particular the G295S and A301V mutants, we observed that despite their modest changes in dNTP levels, they displayed excessively strong mutator phenotypes (more than 1,000-fold increases) in certain mutational forward systems, such as resistance to rifampicin or nalidixic acid. In this work, we investigate the possible mechanism(s) by which the modest dNTP changes observed in the specificity-site mutants can lead to such dramatically enhanced overall mutation rates. We show that modest increases in certain specific replication errors can lead to an “error catastrophe” by compromising the ability of the cells to conduct postreplicative DNA mismatch repair. The studies also suggest reasons why cells need to keep dNTP levels sufficiently low and how to control in particular the dGTP concentration.

effect. The effect on MMR has been traced to titration of one of its limiting components, MutL protein (20). The phenomenon has been termed error catastrophe (20, 21) and can lead to poor growth or even death (21). The classic example of this phenomenon is the mutD5 mutator of E. coli, whose primary defect is in the proofreading activity of the PolIII holoenzyme (22). Although the contribution of the proofreading to fidelity is estimated to be in the range of 100-fold (23), mutD5 strains growing in rich medium display mutation rates for Rifr and other markers that are 1,000- to 10,000-fold elevated. An even more extreme case is the dnaQ926 mutation, which represents a complete knockout of the proofreading activity that renders the cell essentially inviable (21). Both mutD5 and dnaQ926 readily produce suppressors that are of larger colony size and have reduced mutator phenotype. They map in the dnaE gene encoding the α (polymerase) subunit of PolIII, and which have been broadly characterized as antimutators (21, 24). Error catastrophe has also been described in the yeast S. cerevisiae, where certain proofreading-deficient variants of DNA polymerase delta or epsilon are incompatible with a complete deficiency of DNA mismatch repair (25, 26). Suppressors of this incompatibility (termed eex) can also be obtained, which reduce the mutation rate sufficiently to permit life (25, 26). In the series of experiments described below, we demonstrate that the extreme mutability of G295S and A301V is due to MMR saturation. Four approaches are described, which have previously used in diagnosing error catastrophe: (i) analysis of mutant frequencies in MMR+ and MMR− backgrounds, (ii) effects of MutL overproduction, (iii) effects of the dnaE925 antimutator, and (iv) direct measurements of the capacity of the cells to conduct DNA mismatch repair upon transfection with heteroduplex molecules. Mutator Effects in MMR-Defective Background. In Fig. 2A, we show

the Rifr mutant frequencies for several RNR mutators in MMR+ (wild-type) and MMR− (mutS) backgrounds. The difference in mutability between the two backgrounds reflects the MMR capacity of the particular cells. As indicated, the effect is 255-fold

100000

Mutant per 108 cells

A

0.6 x

10000 1000

1300 x

6.6 x

94 x 255 x

100 10 1 0

B

100000

Mutant per 108 cells

10000 7x

1000 100

34 x

10 1 0

Fig. 2. Role of DNA mismatch repair in nrdA specificity-site mutators (mutant frequencies for rifampicin-resistant mutants). (A) Comparison of MMR+ and MMR− (mutS) derivatives of indicated nrdA mutants. Numbers above the bars indicate the fold increase due to the mutS defect, reflecting the efficiency of MMR. (B) Effect of MutL overproduction due to presence of plasmid pMQ350 (mutL+). Numbers above the bars indicate the fold reduction in MMR by the plasmid.

18598 | www.pnas.org/cgi/doi/10.1073/pnas.1310849110

for a wild-type strain, 94-fold for a representative activity-site mutator (H88Y), 1,300-fold for the A301T defect, but only about sixfold for A301V, whereas the frequency for G295S is essentially unchanged between MMR+ and MMR− backgrounds. These results are consistent with severely diminished or even total loss of MMR capacity in the latter two mutators. MutL Overproduction. Fig. 2B shows that increased production of MutL protein from plasmid pMQ350 (27) has only modest effects on the wild-type strain and on the A301T mutator (2.2and 1.5-fold reductions, respectively), but much larger effects on the A301V and G295S mutators (34- and sevenfold reductions, respectively), consistent with increased MMR capacity. A corresponding effect of pMQ350 for the mutD5 mutator was 5- to 14-fold (20, 21) and 4- to 30-fold for error catastrophe induced by exposure to the base analog dP (28). Effects of the dnaE925 Antimutator Allele. The dnaE925 antimutator

allele (F388L) encodes a more accurate DNA polymerase (29), and the associated reduction in replication errors is expected to occur also under conditions of dNTP pool alterations. If so, the stress on the MMR system will be reduced, leading to a disproportionately large antimutator effect, reflecting not only the reduction in primary errors but also the restoration of MMR. In Fig. 3, we show the effects. Indeed, a 95- to 70-fold antimutator effect for G295S and A301V is seen, which compares favorably to the previously reported 24-fold for the mutD5 mutator (21). Direct Measurement of DNA Mismatch Repair Using M13mp2 Heteroduplex DNA. A heteroduplex DNA molecule of the bacte-

riophage M13mp2 was created containing a T·G mismatch in the lacZα gene carried by this phage, as described (20, 30). T·G mismatches are among the best recognized mismatches by MMR (23, 31, 32), and they are a highly sensitive indicator of the state of MMR. If no correction of the mismatch occurs when the molecule is introduced into a cell, a mixed blue/colorless plaque will result, representing progeny from both the T (colorless) and G (blue) strands (30). However, repair of the mismatch will yield a pure plaque, either blue or colorless depending on which strand is corrected (30). The results presented in Fig. 4 indicate that in wildtype strains, only some 3% of the plaques are mixed, indicating efficient repair. In the MMR-defective mutS strain, some 35% are mixed, consistent with lack of repair (because of strand loss during phage replication, 35% of mixed plaques is near the maximum observed). Importantly, strongly elevated percentages of mixed plaques are observed for the G295S and A301V mutators. The approximately 15% mixed plaques observed suggest that only about half of the molecules are repaired, providing the most direct evidence that mismatch repair is severely compromised. DNA Sequence Analysis of Rifampicin-Resistant Mutants. Hypermutability resulting from impaired MMR should be accompanied by a characteristic mutational fingerprint, namely an excess of the two transition mutations (A·T → G·C and G·C → A·T base substitutions), which essentially reflect the nature of the primary (uncorrected) replication errors (19, 33). In contrast, MMR+ strains usually yield a mixture of transitions and transversions due to the strong preference of MMR to correct (remove) the transition errors (19, 32). DNA sequencing of the rpoB gene for approximately 100 rifampicin-resistant mutants for each of the relevant strains yielded the results listed in Table 3. The wildtype strain yielded a mixture of the two transitions combined with a significant contribution of transversions, as expected (18, 19). Also, as expected, the MMR-deficient mutS strain produced nearly exclusively transitions (18, 19). The A301T mutator displayed a mixture of transitions and transversions, and this pattern may be taken as a confirming sign that MMR is fully active in this strain. In contrast, for the A301V and G295S mutators, the Ahluwalia and Schaaper

Mutant per 108 cells

10000.0

1000.0 95 x

70 x

100.0

10.0

1.0

Fig. 3. Effect of the dnaE925 antimutator allele on mutability of nrdA specificity-site mutators. Shown are the frequencies of rifampicin-resistant mutants. Numbers above the bars indicate the fold reduction in MMR by the dnaE allele.

spectra are fully dominated by the transitions, consistent with a lack of MMR. The complete spectra for all strains are presented in the Figs. S1–S4. These data allow a further analysis of mutator effects at individual sites, including hotspot sites, as addressed in Discussion. Discussion The results of this work reveal that seemingly modest changes in the cellular dNTP concentrations can have dramatic consequences for the mutation rate. That such an effect can occur is a further indication of the importance of cellular dNTP control. As argued below, particularly the dGTP level may need to be controlled to avoid the possibility of error catastrophe. Our results show that the G295S and A301V mutator alleles of E. coli ribonucleotide reductase suffer from a 1,000-fold or more increase in forward mutations in the rpoB gene, while having modest dNTP pool changes that include approximately twofold changes in the concentration of dGTP (increased) and dATP (reduced). In this work, we provide several lines of evidence that this excessive mutability results from a compromised ability to perform postreplicative DNA MMR. MMR is an efficient fidelity mechanism responsible for reducing the replication error rate by 100- to 1,000-fold depending on the type of error (22, 23). At the same time, MMR has been found to have a limited capacity, which can become overwhelmed under conditions of high error production (20, 33). MMR has a defined specificity, correcting transition basepair mismatches (purine·pyrimidine mispairs, such as G·T or A·C) with much greater efficiency than those leading to transversions (purine·purine or pyrimidine·pyrimidine). In general, the highest affinity (and correction efficiency) of MMR is found for T·G mismatches (23, 31, 32). Hence, any increased production of this kind of error is expected to be the most taxing on the MMR capacity. For the present study, we note that the dNTP pool changes in the A301V and G295S mutants will specifically favor this type of mismatch: At template T sites, there will be competition at the insertion step between dGTP as the incorrect nucleotide and dATP as correct nucleotide. Therefore, both the observed increase in dGTP and the decrease in dATP will favor this particular error. The expected genome-wide production of these errors is likely to place considerable stress on the MMR system. Additional support that enhanced T·G mispairing underlies a majority of mutations can be found in an analysis of the spectra of rifampicin-resistant mutants in the rpoB gene. Table 3 indicates that the majority of mutations are A·T → G·C transitions. The detailed spectra (Figs. S1–S4) show that these transitions are located at four hotspots, rpoB positions 1532, 1538, 1547, and 1598. The sequence contexts of these sites are shown in Fig. 5, Ahluwalia and Schaaper

which also depicts the competition between dGTP (incorrect nucleotide) and dATP (correct nucleotide) opposite the template T, as well as the next dNTP to be incorporated following the misinsertion. Although mutations can be detected over some 70 distinct pathways throughout the rpoB gene, these four sites take up more than 95% of all observed mutations. Notably, each site is characterized by the requirement for incorporation of either dGTP (1532, 1547, 1598) or dCTP (1547) as the next nucleotide that needs to be incorporated following the T·G mismatch. The concentration of the next dNTP has been shown to be a critical determinant for the misincorporation rate, because elevated next dNTPs promote extension of the mismatch at the expense of its removal by the exonucleolytic proofreading (next-nucleotide effect) (34, 35). This phenomenon, initially developed from in vitro studies with DNA polymerases, has also been shown to operate efficiently in vivo (11, 13, 36). Thus, our observations of increased dGTP for both G295S and A301V and of increased dCTP for A301V (Table 1) are fully consistent with the specific enhancement of mutations at these indicated sites. At these sites, three parameters conspire to elevate the T·G mispairing errors for the two RNR mutators: elevated dGTP as misincorporating nucleotide, reduced dATP as correct nucleotide, and elevated dGTP or dCTP as extension-promoting next nucleotide. Thus, despite modest increases in these nucleotides, their effects can be multiplicative and, when involving impairment of MMR, catastrophic. Hence, we are describing a case where all three major fidelity mechanisms that cells use for their DNA replication (base selection, proofreading, and mismatch repair) become impaired by the change in DNA precursors. Since the first described cases of MMR saturation and error catastrophe (20, 22), new insight has emerged into the physiology of stressed bacteria, including stress due to carbon source deprivation (37), antibiotic exposure (38–40), impaired replication fork progress (38, 41), and dNTP pool depletion by the RNR inhibitor, hydroxyurea (42, 43). Some of these stresses have been reported to result in the production of oxygen radicals or a mutator phenotype (38, 39, 42, 44). In this light, it is relevant to ask whether the A301V and G295S mutants may also suffer from stress and whether their poor growth and hypermutator phenotype could result, at least in

GENETICS

100000.0

Strain wild-type A301V G295S mutS

Mixed plaques (%) 3.6 15 13 35

Pure plaques (%) 94.4 85 87 65

Fig. 4. Heteroduplex correction in various E. coli strains. A phage M13mp2 DNA heteroduplex molecule containing a T·G mismatch was introduced by electroporation into the indicated strains, and the resulting plaques were scored for pure (either blue or colorless) or mixed blue/colorless phenotype. One example of each phenotype is indicated by a red box. The T·G mismatch resides in the lacZα part of M13mp2, and lack of correction leads to a mixed blue/colorless phenotype, whereas correction will lead to a pure white of blue plaque depending on the direction of correction. Details of the assay can be found in refs. 20 and 30 (Materials and Methods).

PNAS | November 12, 2013 | vol. 110 | no. 46 | 18599

Table 3. Nature and number of identified RifR (rpoB) mutations in nrdAB mutator strains either proficient or deficient in DNA mismatch repair (mutS) Mutation Transitions A·T → G·C G·C → A·T Transversions G·C → T·A A·T → T·A A·T → C·G G·C → C·G Total

WT

WT mutS

A301T

A301T mutS

A301V

A301V mutS

G295S

G295S mutS

24 42

57 41

69 5

83 16

96 0

89 10

99 1

71 28

4 27 1 1 99

1 0 0 0 99

2 23 0 2 101

0 1 0 0 100

0 1 1 0 98

0 1 3 0 103

0 0 0 0 100

0 0 0 0 99

The complete spectra, as obtained by DNA sequencing, are provided in Figs. S1–S4.

part, from such alternative disturbances. The dNTP pool changes of G295S and A301V are modest and, thus, not directly predictive of major replication stress, although we cannot yet exclude the possibility that they may suffer from more severe bottlenecks in growth stages other than logarithmic growth, where the dNTPs were measured. Importantly, the suppressors of the poor-growth phenotype have uniformly lost their hypermutability, at least consistent with a direct link between the two phenotypes. However, the operation of additional mechanisms that contribute to hypermutability and poor growth cannot be excluded and should be considered. In particular, it is likely that the hypermutators accumulate mutated (misfolded) proteins that can trigger defined stress responses, contributing to the poor growth and mutability (37, 39, 40, 42, 44). In contrast to the high concentrations of the cellular rNTPs, which are generally in the millimolar range, those of the dNTPs are in the low- to mid-micromolar range (45–49). Thus, although high levels of the dNTPs would presumably facilitate rapid DNA

replication, it is clear that such use of high dNTP levels would be a bad strategy from a fidelity perspective. It thus appears that dNTP levels are kept low, so that exonucleolytic proofreading can be an effective fidelity mechanism and preserve, at the same time, the efficiency of the postreplicative MMR system. It is also clear from our study that elevations of dGTP combined with reductions of dATP might be particularly dangerous for the cell, as T·G mispairings, which are the most frequent polymerase errors, need to be kept in check. From this perspective, it is not surprising that among the four dNTPs, dGTP is normally the lowest while dATP is the highest (Table 1 and refs. 47, 50, and 51). Materials and Methods Strains and Media. E. coli strains used are all derivatives of strain NR12470, a Δ(gpt-lac)5 derivative of MG1655 (11). They also carry a chromosomal ΔnrdAB::kan deletion and a complementing nrdAB-containing plasmid (pHABamp, pHABcat, or pHABtet). Low-copy plasmids pHABamp and pHABcat have been described (11). pHABtet was created by replacing the cat gene of pHABcat by the tet gene of pBR322 (this work). The strains also carried the F’prolac from strains CC104 or CC105 (17), permitting measurement of lac reversion frequencies (11). Plasmids pHABcat-A301T, pHABcat-A301V, and pHABcat-G295S, are versions of pHABcat that contain the A301T, A301V, and G295S nrdA mutator alleles introduced into the strains by plasmid exchange (11). The dnaE925 antimutator allele (21, 29) linked with transposon zae-502::Tn10, and the mutL218::Tn10 allele (52) were introduced by P1 transduction. Plasmid pMQ350, a pBR322-derived plasmid containing the E. coli mutL+ gene, was provided by M. Marinus (University of Massachusetts, Worchester, MA). Liquid and solid LB and minimal glucose media (MM) have been described (29). MM-Lac plates used for determination of lac reversion frequencies contained lactose (0.2%) instead of glucose. LBRif plates contained 100 μg/mL rifampicin (Sigma-Aldrich). Mutant Frequency Determinations. Mutant frequencies were determined as described (36) from 12 or more 1-mL LB cultures, each started from a separate colony and grown overnight to saturation at 37 °C. For the strong mutators multiple, independent isolates (i.e., resulting from plasmid exchanges) were routinely used. To minimize the possible contribution of suppressors for the high-mutator strains, cultures were inoculated in most experiments directly from the transformation plates by resuspending the entire colony in 1 mL of LB without colony purification. dNTP Pool Measurements. dNTPs were extracted by using the procedure described by Buckstein (46) with minor modifications (11) and quantitated by HPLC analysis as described (11). Mutation rates were monitored throughout all stages of the cell cultures to avoid inadvertent accumulation of suppressors.

Fig. 5. Sequence contexts of hotspot sites for rifampicin-resistant mutants observed in nrdA mutator strains. Indicated are the positions in the rpoB gene where the mutations occur and the number of occurrences for each strain. The numbering of the sites is as in ref. 19. A·T → G·C transitions can also be observed at positions 1534, 1552, and 1577, where we detected a total of six mutations for A301V and G295S combined (of 200 total). The sequence context for these three sites is 5′-TGTC, GTTC and TGTG. See Discussion for details. The underline indicates the template base at which the base·base mispairing errors occur.

18600 | www.pnas.org/cgi/doi/10.1073/pnas.1310849110

Sequencing of rpoB mutations. The mutational specificity of the nrdA mutator alleles (G295 S, A301T, and A301V) was determined by sequencing the rpoB gene of approximately 100 independent rifampicin-resistant mutants in each strain in both MMR+ and MMR− (mutS) background. Each Rifr mutant was derived from a separate mutator plasmid exchange reaction (see above), assuring complete independence of the Rifr mutants. Only mutations occurring between rpoB positions 1516–1717 were considered during analysis (approximately 80% of all rpoB mutations are localized in this region; ref. 19).

Ahluwalia and Schaaper

Measurement of DNA Mismatch Repair Using M13mp2 Heteroduplex DNA. A bacteriophage M13mp2 heteroduplex molecule was used containing a T·G mismatch at position 90 of lacZα (30). The molecule is also hemimethylated, as the viral (+) strand was derived from phage grown in a dam+ strain, whereas the complementary (−) strand was derived from double-stranded DNA of phage grown in a dam− strain, as described (20, 30). The molecule was electroporated into competent cells of interest and aliquots of the transfection mix plated on MM-glucose plates containing IPTG and XGal and using strain CSH50 as indicator strain, as described (20, 30). The number of mixed (blue/colorless)

and pure (either blue or colorless) plaques was enumerated, and the efficiency of mismatch repair calculated from the percentage of mixed plaques.

1. Hofer A, Crona M, Logan DT, Sjöberg BM (2012) DNA building blocks: Keeping control of manufacture. Crit Rev Biochem Mol Biol 47(1):50–63. 2. Reichard P (2010) Ribonucleotide reductases: Substrate specificity by allostery. Biochem Biophys Res Commun 396(1):19–23. 3. Hama H, Almaula N, Lerner CG, Inouye S, Inouye M (1991) Nucleoside diphosphate kinase from Escherichia coli; its overproduction and sequence comparison with eukaryotic enzymes. Gene 105(1):31–36. 4. Wang L, Weiss B (1992) dcd (dCTP deaminase) gene of Escherichia coli: Mapping, cloning, sequencing, and identification as a locus of suppressors of lethal dut (dUTPase) mutations. J Bacteriol 174(17):5647–5653. 5. Holder PG, Pizano AA, Anderson BL, Stubbe J, Nocera DG (2012) Deciphering radical transport in the large subunit of class I ribonucleotide reductase. J Am Chem Soc 134(2):1172–1180. 6. Stubbe J, Nocera DG, Yee CS, Chang MC (2003) Radical initiation in the class I ribonucleotide reductase: Long-range proton-coupled electron transfer? Chem Rev 103(6):2167–2201. 7. Tomter AB, et al. (2013) Ribonucleotide reductase class I with different radical generating clusters. Coord Chem Rev 257:3–26. 8. Birgander PL, Kasrayan A, Sjöberg BM (2004) Mutant R1 proteins from Escherichia coli class Ia ribonucleotide reductase with altered responses to dATP inhibition. J Biol Chem 279(15):14496–14501. 9. Larsson KM, et al. (2004) Structural mechanism of allosteric substrate specificity regulation in a ribonucleotide reductase. Nat Struct Mol Biol 11(11):1142–1149. 10. Xu H, et al. (2006) Structures of eukaryotic ribonucleotide reductase I provide insights into dNTP regulation. Proc Natl Acad Sci USA 103(11):4022–4027. 11. Ahluwalia D, Bienstock RJ, Schaaper RM (2012) Novel mutator mutants of E. coli nrdAB ribonucleotide reductase: Insight into allosteric regulation and control of mutation rates. DNA Repair (Amst) 11(5):480–487. 12. Brignole EJ, Ando N, Zimanyi CM, Drennan CL (2012) The prototypic class Ia ribonucleotide reductase from Escherichia coli: Still surprising after all these years. Biochem Soc Trans 40(3):523–530. 13. Kumar D, et al. (2011) Mechanisms of mutagenesis in vivo due to imbalanced dNTP pools. Nucleic Acids Res 39(4):1360–1371. 14. Eriksson M, et al. (1997) Binding of allosteric effectors to ribonucleotide reductase protein R1: Reduction of active-site cysteines promotes substrate binding. Structure 5(8):1077–1092. 15. Nordlund P, Sjöberg BM, Eklund H (1990) Three-dimensional structure of the free radical protein of ribonucleotide reductase. Nature 345(6276):593–598. 16. Uhlin U, Eklund H (1994) Structure of ribonucleotide reductase protein R1. Nature 370(6490):533–539. 17. Cupples CG, Miller JH (1989) A set of lacZ mutations in Escherichia coli that allow rapid detection of each of the six base substitutions. Proc Natl Acad Sci USA 86(14): 5345–5349. 18. Curti E, McDonald JP, Mead S, Woodgate R (2009) DNA polymerase switching: Effects on spontaneous mutagenesis in Escherichia coli. Mol Microbiol 71(2):315–331. 19. Garibyan L, et al. (2003) Use of the rpoB gene to determine the specificity of base substitution mutations on the Escherichia coli chromosome. DNA Repair (Amst) 2(5): 593–608. 20. Schaaper RM, Radman M (1989) The extreme mutator effect of Escherichia coli mutD5 results from saturation of mismatch repair by excessive DNA replication errors. EMBO J 8(11):3511–3516. 21. Fijalkowska IJ, Schaaper RM (1996) Mutants in the Exo I motif of Escherichia coli dnaQ: Defective proofreading and inviability due to error catastrophe. Proc Natl Acad Sci USA 93(7):2856–2861. 22. Schaaper RM (1988) Mechanisms of mutagenesis in the Escherichia coli mutator mutD5: Role of DNA mismatch repair. Proc Natl Acad Sci USA 85(21):8126–8130. 23. Schaaper RM (1993) Base selection, proofreading, and mismatch repair during DNA replication in Escherichia coli. J Biol Chem 268(32):23762–23765. 24. Schaaper RM (1998) Antimutator mutants in bacteriophage T4 and Escherichia coli. Genetics 148(4):1579–1585. 25. Herr AJ, et al. (2011) Mutator suppression and escape from replication error-induced extinction in yeast. PLoS Genet 7(10):e1002282.

26. Williams LN, Herr AJ, Preston BD (2013) Emergence of DNA polymerase e antimutators that escape error-induced extinction in yeast. Genetics 193(3):751–770. 27. Aronshtam A, Marinus MG (1996) Dominant negative mutator mutations in the mutL gene of Escherichia coli. Nucleic Acids Res 24(13):2498–2504. 28. Negishi K, Loakes D, Schaaper RM (2002) Saturation of DNA mismatch repair and error catastrophe by a base analogue in Escherichia coli. Genetics 161(4):1363–1371. 29. Fijalkowska IJ, Schaaper RM (1995) Effects of Escherichia coli dnaE antimutator alleles in a proofreading-deficient mutD5 strain. J Bacteriol 177(20):5979–5986. 30. Schaaper RM (1989) Escherichia coli mutator mutD5 is defective in the mutHLS pathway of DNA mismatch repair. Genetics 121(2):205–212. 31. Modrich P (1991) Mechanisms and biological effects of mismatch repair. Annu Rev Genet 25:229–253. 32. Schaaper RM, Dunn RL (1991) Spontaneous mutation in the Escherichia coli lacI gene. Genetics 129(2):317–326. 33. Schaaper RM, Dunn RL (1987) Spectra of spontaneous mutations in Escherichia coli strains defective in mismatch correction: The nature of in vivo DNA replication errors. Proc Natl Acad Sci USA 84(17):6220–6224. 34. Fersht AR (1979) Fidelity of replication of phage phi X174 DNA by DNA polymerase III holoenzyme: Spontaneous mutation by misincorporation. Proc Natl Acad Sci USA 76(10):4946–4950. 35. Kunkel TA, Schaaper RM, Beckman RA, Loeb LA (1981) On the fidelity of DNA replication. Effect of the next nucleotide on proofreading. J Biol Chem 256(19):9883–9889. 36. Schaaper RM, Mathews CK (2013) Mutational consequences of dNTP pool imbalances in E. coli. DNA Repair (Amst) 12(1):73–79. 37. Al Mamun AA, et al. (2012) Identity and function of a large gene network underlying mutagenic repair of DNA breaks. Science 338(6112):1344–1348. 38. Foti JJ, Devadoss B, Winkler JA, Collins JJ, Walker GC (2012) Oxidation of the guanine nucleotide pool underlies cell death by bactericidal antibiotics. Science 336(6079): 315–319. 39. Gutierrez A, et al. (2013) β-lactam antibiotics promote bacterial mutagenesis via an RpoS-mediated reduction in replication fidelity. Nat Commun 4:1610. 40. Kohanski MA, Dwyer DJ, Wierzbowski J, Cottarel G, Collins JJ (2008) Mistranslation of membrane proteins and two-component system activation trigger antibiotic-mediated cell death. Cell 135(4):679–690. 41. Uchida K, et al. (2008) Overproduction of Escherichia coli DNA polymerase DinB (Pol IV) inhibits replication fork progression and is lethal. Mol Microbiol 70(3):608–622. 42. Davies BW, et al. (2009) Hydroxyurea induces hydroxyl radical-mediated cell death in Escherichia coli. Mol Cell 36(5):845–860. 43. Nakayashiki T, Mori H (2013) Genome-wide screening with hydroxyurea reveals a link between nonessential ribosomal proteins and reactive oxygen species production. J Bacteriol 195(6):1226–1235. 44. Kohanski MA, DePristo MA, Collins JJ (2010) Sublethal antibiotic treatment leads to multidrug resistance via radical-induced mutagenesis. Mol Cell 37(3):311–320. 45. Bochner BR, Ames BN (1982) Complete analysis of cellular nucleotides by twodimensional thin layer chromatography. J Biol Chem 257(16):9759–9769. 46. Buckstein MH, He J, Rubin H (2008) Characterization of nucleotide pools as a function of physiological state in Escherichia coli. J Bacteriol 190(2):718–726. 47. Ferraro P, Franzolin E, Pontarin G, Reichard P, Bianchi V (2010) Quantitation of cellular deoxynucleoside triphosphates. Nucleic Acids Res 38(6):e85. 48. Traut TW (1994) Physiological concentrations of purines and pyrimidines. Mol Cell Biochem 140(1):1–22. 49. Mathews CK (2006) DNA precursor metabolism and genomic stability. FASEB J 20(9): 1300–1314. 50. Mathews CK, Ji J (1992) DNA precursor asymmetries, replication fidelity, and variable genome evolution. Bioessays 14(5):295–301. 51. Wheeler LJ, Rajagopal I, Mathews CK (2005) Stimulation of mutagenesis by proportional deoxyribonucleoside triphosphate accumulation in Escherichia coli. DNA Repair (Amst) 4(12):1450–1456. 52. Siegel EC, Wain SL, Meltzer SF, Binion ML, Steinberg JL (1982) Mutator mutations in Escherichia coli induced by the insertion of phage mu and the transposable resistance elements Tn5 and Tn10. Mutat Res 93(1):25–33.

Ahluwalia and Schaaper

PNAS | November 12, 2013 | vol. 110 | no. 46 | 18601

GENETICS

ACKNOWLEDGMENTS. We thank Drs. Thomas Kunkel and Jessica Williams [National Institute of Environmental Health Sciences (NIEHS)] for their helpful comments on the manuscript for this paper and Dr. Rachelle Bienstock (NIEHS) for preparing Fig. 1. The NIEHS DNA Sequencing Core is acknowledged for the sequencing of the E. coli rpoB mutants. This work was supported by Project Z01 ES065086 of the Intramural Research Program, NIEHS.