|Home | About | Journals | Submit | Contact Us | Français|
The accuracy of DNA synthesis depends on the accuracy of the polymerase as well as the quality and concentration(s) of the available 5′-deoxynucleoside-triphosphate DNA precursors (dNTPs). The relationships between dNTPs and error rates have been studied in vitro, but only limited insights exist into these correlations during in vivo replication. We have investigated this issue in the bacterium E. coli by analyzing the mutational properties of dcd and ndk strains. These strains, defective in dCTP deaminase and nucleoside diphosphate kinase, respectively, are characterized by both disturbances of dNTP pools and a mutator phenotype. ndk strains have been studied before, but were included in this study, as controversies exist regarding the source of its mutator phenotype. We show that dcd strains suffer from increased intracellular levels of dCTP (four-fold) and reduced levels of dGTP (two-fold), while displaying, as measured using a set of lacZ reversion markers in a mismatch-repair defective (mutL) background, a strong mutator effect for G·C→T·A and A·T→T·A transversions (27- and 42-fold enhancement, respectively). In contrast, ndk strains possess a lowered dATP level (four-fold) and modestly enhanced dCTP level (2-fold), while its mutator effect is specific for just the A·T→T·A transversions. The two strains also display differential mutability for rifampicin-resistant mutants. Overall, our analysis reveals for both strains a satisfactory correlation between dNTP pool alterations and the replication error rates, and also suggests that a minimal explanation for the ndk mutator does not require assumptions beyond the predicted effect of the dNTP pools.
Cells employ multiple mechanisms to ensure an appropriate low error rate during the replication of their genome [1-3]. The responsible DNA polymerases (replicases) avoid errors through a combination of error prevention at the insertion step (base selection) and error correction via the exonucleolytic removal of erroneous insertions (proofreading or editing). Following polymerization, postreplicative mismatch repair (MMR) is capable of further reducing the error rate by detecting and correcting mispairs that escaped the proofreading step. In addition, cells employ numerous DNA surveillance and repair pathways aimed at maintaining the DNA in a damage-free state, so as to minimize the entry of damaged, and likely mutagenic, DNA into the replication fork .
One important additional factor that is relevant to DNA replication fidelity and has been receiving increasing interest is the status of the DNA precursors, the 5′-deoxynucleoside-triphosphates (dNTPs), both qualitatively and quantitatively. Quality-wise, the cellular dNTPs are ‘sanitized’ to rid the precursor pool of modified or damaged dNTPs that upon incorporation into the DNA may be toxic and/or mutagenic [5, 6]. One example of such sanitizing activity is the MutT protein that removes 8-oxodGTP from the pools [5, 7].
In terms of quantities, each of the dNTPs is kept at relatively low concentration in the micromolar range (at only a few % of the corresponding rNTPs). The reason why the dNTPs are kept low is an interesting question that may well relate to the need to keep the error rate low. In addition to overall low levels, the relative ratios of the four dNTPs (dATP, dTTP, dCTP, dGTP) are kept under control, although not at equimolar levels. In general the dGTP level is the lowest in both prokaryotes and eukaryotes [8, 9] with the exception of mitochondria [10, 11].
The regulation of the cellular dNTP pool occurs at a number of levels. One critical step involves allosteric regulation of the enzyme ribonucleotide reductase (RNR), which performs the important reduction of the ribonucleotides to the corresponding deoxyribonucleotides in an allosterically controlled fashion that (i) limits overall dNTP production and (ii) balances the reduction of the four substrates such that the resulting dNTPs are present in the cell at the desired proportions . In both E. coli and human cells this reduction occurs at the level of the nucleoside diphosphates (NDP→dNDP). Once formed, the dNDPs are converted to the dNTP level by the enzyme nucleoside diphosphate kinase (NDK).
RNR is a unique enzyme containing two allosteric regulatory sites. The first one regulates overall activity through an (ATP/dATP-controlled) on/off switch (this site is termed the activity site). The second allosteric site determines which substrate can be reduced at any given time (called the specificity site) . Recent studies have described strains of E. coli  and S. cerevisiae  carrying amino acid substitutions at sites of RNR allosteric regulation and, consequently, displaying a mutator phenotype (increased mutation rates). Likewise, overproduction of RNR itself is mutagenic [8, 14], further supporting the important role of RNR in controlling the cellular mutation rate.
While RNR is a critical component in the dNTP metabolism, it is embedded within a larger network of nucleotide metabolizing pathways, the components of which may also affect the dNTP pools. We have outlined part of this network in Fig. 1. Of keen interest to our present study into the role of dNTPs and mutagenesis are two genes: dcd, encoding dCTP deaminase (Dcd) , and ndk, encoding nucleoside diphosphate kinase (Ndk) [16, 17]. As shown in Fig. 1, dCTP deaminase is responsible for converting dCTP to dUTP, representing the first step in the de novo synthesis of dTTP . dUTP is rapidly converted to dUMP by dUTPase (dut gene product); dUMP is converted to dTMP by thymidylate synthase, and dTMP is converted by two consecutive kinase reactions to the final product dTTP. Altered dNTP pools in E. coli dcd mutants lacking dCTP deaminase activity were reported early on [18, 19]. As expected, these strains accumulate dCTP and, when grown on synthetic medium, suffer from reduced dTTP levels [18, 19]. However, no studies on the mutation rates of dcd mutants have been reported, and therefore this strain was chosen presently to analyze relationships between dNTP pools and mutagenesis.
The second genetic defect that we chose for an in-depth analysis of mutagenesis and dNTP pools is ndk. Ndk is a multipurpose enzyme capable of converting each of the (d)NDPs to the corresponding triphosphates. Interestingly, in E. coli the enzyme is dispensable as its function can be substituted, at least in part, by other kinases, particularly adenylate kinase [21, 22]. Deletion of ndk caused both dNTP pool changes and a mutator phenotype, which was ascribed to the dNTP pool changes [22, 23]. However, this particular explanation for the ndk mutator effect has been controversial [24, 25]. One other study showed that complementation of the E. coli ndk defect with the human NDK homolog hNm23-H2 abolished the mutator phenotype but did not restore the dNTPs to their normal levels . Hence, the dNTP pool alteration was not considered the main cause of the mutator phenotype. Instead, an alternative explanation was forwarded in which the main cause of the mutator effect is related to an increased production of dUTP and its incorporation into DNA instead of dTTP [24, 25]. As noted (Fig. 1), dUTP resulting from the deamination of dCTP, is hydrolyzed by dUTPase (dut gene product) to yield dUMP, which is then converted to dTMP and eventually dTTP. As dUTP is a close analog of dTTP, differing only in the lack of the 5-methyl group, it is readily incorporated by DNA polymerase, and rapid destruction of dUTP is therefore necessary to prevent frequent incorporation. In additional support for the dUTP hypothesis, a synergistic interaction was reported between ndk and the dut-1 defect, resulting in further enhancement of mutagenesis . On the other hand, the intrinsic mutagenicity of uracil in DNA is low, as is the case for the abasic sites that result from its its removal by uracil DNA glycosylase. Therefore, the precise mode of mutagenesis in ndk strains remains open to interpretation and is addressed by this study.
In this paper, we investigate in some detail the mutational properties of the dcd and ndk strains, focusing on the precise types of mutations that are enhanced (mutational specificity). In parallel, we determine the dNTP pool changes in the strains, and subsequently attempt to make meaningful comparisons between the pool changes and the observed mutational specificity. The mutation measurements are made in mismatch-repair-deficient strains (mutL) in order to more accurately correlate the dNTP data with replication error rates. Overall, the data indicate that meaningful correlations between dNTP pool changes and the induced replication errors can be derived.
The E. coli strains used are derivatives of strain KA796 (Δpro-lac, ara, thi) . Strain constructions by P1 transduction were done using P1virA. Two mismatch-repair-deficient (mutL) derivatives of KA796 were used: NR9559 (mutL211::Tn5)  and NR12443 (mutL218::Tn10). The mutL markers in the two strains derive, respectively, from strains ES1301 and ES1481 . The dcd-12::Tn10dkan allele  was introduced into KA796 and NR9559 (mutL) from BW1040  (provided by Dr. Bernard Weiss, University of Michigan) yielding, respectively, NR12572 (dcd) and NR12573 (mutL dcd). The Δndk::cam allele was introduced into KA796 and NR12443 (mutL) from QL7623  yielding, respectively, NR11814 (ndk) and NR11816 (mutL ndk). Into each of these strains we introduced, by conjugation, the series of F’prolacZ episomes originally present in strains CC101 through CC106 .
Liquid LB and minimal media (MM) were prepared using standard recipes . Solid medium contained 1.5 % agar (Difco). MM medium contained 1x Vogel-Bonner salts , 0.4% glucose (MM-Glu) or 0.4% lactose (MM-Lac) as a carbon source, and 5 μg/ml of thiamine. For experiments with dcd strains the solid media (LB, MM-Glu, MM-Lac) contained additionally 50 μg/ml of thymidine to improve colony growth on the plates (larger colony sizes). Liquid media, used for generation of mutant frequencies and extraction of cellular dNTP pools (see below), did not contain any added thymidine. Antibiotics, when required during strain constructions, were added at 25 μg/ml (kanamycin), 12.5 μg/ml (tetracycline), or 20 μg/ml (chloramphenicol); LB-Rif plates used for the scoring of rifampicin-resistant mutants contained 100 μg/ml rifampicin (Sigma-Aldrich).
Mismatch-repair-deficient (mutL) strains were used for these series of experiments. For each strain, a total of 12-20 independent LB cultures (1 ml each) were initiated from single colonies (one colony per tube). The cultures were grown to saturation on a rotator wheel at 37°C. The total cell count in the cultures was determined by plating 0.1 ml of a 10−6 dilution on MM-Glu and LB plates. The number of lac revertants in each culture was determined by plating 0.1 ml of the undiluted cultures on MM-Lac plates. The number of rifampicin-resistant colonies in each culture was determined by plating 0.1 ml undiluted cultures on LB-Rif plates. The plates were incubated for 20 h (LB) or 40 h (MM) at 37°C. To calculate mutant frequencies, the number of mutants on each selection plate was divided by the total number of plated cells. Occasional jackpot cultures were removed from the analysis. Average frequencies with standard error (SE) were determined using the statistical software program Prism (GraphPad).
dNTP pool size experiments were performed on mismatch-repair-proficient derivatives of the wild-type (NR10835, NR10836), dcd (NR12578), and ndk (NR15669) strains. The dNTPs were determined in exponential-phase cells grown in LB medium. Cells were harvested at OD660 = 0.5, and dNTP pool data were recorded using the DNA polymerase-based assay, as described . A possible source of error in this assay, when using Klenow fragment as DNA polymerase, is its possible usage of GTP and CTP - when present at high concentration - leading to an overestimation of the dGTP and dCTP concentrations . However, we have found this to be a significant source of error only when the CTP/dCTP or GTP/dGTP ratios exceed about 100 . In E. coli B, which is typical of wild-type E. coli strains, the CTP/dCTP and GTP/dGTP ratio do not exceed 10 . Therefore, any errors resulting from the use of Klenow fragment to generate the current dNTP pool data are assumed to be insignificant.
The purpose of our studies was to investigate the correlations of the cellular dNTP pools (and changes therein) with the cellular mutation rate or, more precisely, with the rate by which cells produce specific DNA replication errors. As replication errors are normally subject to correction by the postreplicative mismatch repair system (mutHLS system), which imposes its own specificity on the individual errors (different errors are corrected with different efficiency), we performed our mutagenesis experiments in strains defective in mismatch repair (mutL). In this manner, the observed mutations may be assumed to reflect directly the rate and specificity of the replication error production. In addition, we investigated the specificity of mutagenesis (types of errors) using the set of lacZ reversion alleles developed by Cupples and Miller . This set of six lacZ missense alleles can each revert to lac+ by one of the six possible base pair substitutions: A·T→C·G, G·C→A·T, G·C→C·G, G·C→T·A, A·T→T·A, and A·T→G·C . In addition, we measured mutations resulting in resistance to the antibiotic rifampicin, for which mutations occur at a larger number of sites in the rpoB gene (forward system) .
The results presented in Table 1 show that a strain defective in dCTP deaminase (dcd) possesses elevated mutation rates, i.e., dcd strains are mutators. This mutator effect has a clear specificity: when calculating the increase in mutant frequency for each of the six lacZ alleles, the increase is most pronounced for the lac G·C→T·A and A·T→T·A transversions, 27- and 42-fold, respectively, compared to the control mutL strain. In contrast, the increase using the Rif forward assay is only very modest, around 2-fold.
The mutator effect of ndk strains has been reported before [22, 23]. Nevertheless, the strain was included in the present study to permit analysis of the mutational specificity in our strains, to be used alongside the simultaneously determined dNTP pool changes (see below).
In Table 2, we list the results for the ndk mutL strain along with its mutL parent strain. The results show a strong mutator effect for the lac A·T→T·A transversion, but no or only small effects for the other lac alleles. In contrast, a clear 10-fold increase is observed for the rifampicin-resistant forward mutations. These results are similar to those observed by Miller et al. .
In order to best correlate the observed mutational effects with dNTP pool changes, the pools were measured in cells growing in LB medium (mutant frequencies were determined from cells growing in LB). The results for a large number of measurements in the wild-type strain show that, under these conditions, dCTP possesses the largest pool (75 pmoles/108 cells), whereas the smallest is observed for dGTP (8.0 pmoles/108 cells) (see Table 3). In the dcd strain, there is a strong further increase of the dCTP pool (near 4-fold) and a reduction in the dGTP concentration (2-fold). The elevation of the dCTP pool, also reported in early studies [18, 19], is as expected in view of the lack of dCTP deaminase activity. No significant decrease of the dTTP pool [18, 19] was found in our experiments. This is presumably due to the use of LB medium, which includes nucleotide precursors such as thymine and thymidine, which permit synthesis of dTTP via salvage pathways . We did observe significant reductions in the dTTP pool when the dcd cells were grown in defined minimal media (not shown), consistent with the previous reports [18, 19].
In the ndk strain, we observe two types of effect, a significant reduction in the dATP pool (4-fold) coupled with a doubling of the dCTP pool (Table 3). For comparison, we also determined again the pools for the previously described QL7623 (ndk) strain and corresponding wild type (JC7623) . In this background, the ndk strain showed no significant decrease of the dATP pool; instead, a significant increase (6-fold) occurs in the dCTP level, as well as a modest increase (2.5-fold) for dGTP (Table 3). These results are more similar to the previously reported values for this strain background (a 20-fold increase in dCTP and a 7-fold increase in dGTP) . It is to be noted that the JC and QL strains were recB recC sbcB strains used for the chromosomal deletion of the ndk gene and contain a further 21 known mutations (E. coli Genetic Stock Center). Thus, it may perhaps not be surprising that different dNTP pool effects are noted. The finding also highlights the fact that strain backgrounds, as well as growth media, need to be taken into account when investigating dNTP pools and associated effects.
From in vitro studies with DNA polymerases, it has become clear that DNA polymerase error rates depend on a number of variables [1, 2]. At the nucleotide insertion step, each DNA polymerase has an intrinsic error rate, which represents the probability that it may insert a wrong nucleotide at a given template position. Insertion error rates generally vary with the type of error and also with the DNA sequence context. Broadly, transition errors (purine·pyrimidine mispairings, like T·G or A·C) are made more frequently than transversion errors (pyrimidine·pyrimidine or purine·purine mispairings). Importantly, since correct and incorrect dNTPs compete with each other at the nucleotide insertion step, a linear correlation is expected between the incorrect/correct dNTP ratio and the misinsertion rate. Thus, independent of the magnitude by which a DNA polymerase may discriminate against insertion of a particular incorrect nucleotide (intrinsic misinsertion fidelity), an x-fold change in the incorrect/correct ratio should lead to a corresponding x-fold change in the misinsertion rate.
A second critical aspect of dNTP pools regarding polymerase fidelity relates to the subsequent editing (proofreading) step, whereby the polymerase has an opportunity to excise the newly misinserted base. In E. coli this step is performed by the Pol III ε subunit (dnaQ gene product) [37, 38]. The fidelity contribution of the proofreading step in vivo is substantial (up to two orders of magnitude), based on the strong mutator effects associated with defective dnaQ mutants . As judged from polymerase studies in vitro, the efficiency of proofreading can be affected by the dNTPs, in particular those that are to be inserted directly following the mismatch: higher “next” dNTPs will act to promote extension of the mismatch at the expense of the reverse excision step [1, 39]. Higher dNTPs can therefore be mutagenic, and this “next nucleotide” effect is likely to operate in vivo as well [12, 13]. The combined effects of the currently observed dNTP changes on both polymerase misinsertion and proofreading are discussed below.
Our results show that dcd mutants display a specific enhancement of the dCTP pool (Table 3). At the same time, the associated mutational data (Table 1) show a specific enhancement of the lac G·C→T·A and A·T→T·A transversions (27- to 42-fold). In the scheme of Fig. 2 we analyze the six lacZ base substitutions in terms of the two possible mispairing errors that can lead to each substitution, as well as the next nucleotide to be incorporated for each case. The format used to describe each mispairing event is: “template•(incorrect/correct dNTP)”. For an optimal analysis one would like to know which of the two possible mispairs (1 or 2 in the scheme) is the more frequent one, because, obviously, an increase in error rate predicted for the predominant mispair will be readily observable as an increase in the mutation rate, whereas such an effect for the minor mispair could remain obscured. The predominant event for at least four of the lac base substitutions appears well established [40-42], and these are bolded and boxed in the diagram: G·T and T·G mispairs for the two transitions (G·C→A·T and A·T→G·C, respectively) and the T·T and C·T pyrimidine·pyrimidine mispairs (G·C→T·A and A·T→T·A transversions, respectively) [40-42]. As in the dcd strain dCTP is the nucleotide that is primarily affected, we have it presented in color: red, when its enhancement is predicted to mutagenic, and blue when expected to be antimutagenic (likewise for dGTP, which is two-fold reduced). As an example (Fig. 2), the analysis shows that the lac G·C→A·T transition would be disfavored at the insertion step (because dCTP is the correct nucleotide at this site), but would be promoted at the extension step (dCTP is the next nucleotide). A modest 5-fold increase in mutant frequency is observed (Table 1), not inconsistent with the two opposing effects. Importantly, a straightforward mutator effect is predicted for the G·C→T·A and A·T→T·A transversions, as dCTP is the next nucleotide in both cases. No particular effect is predicted for the A·T→G·C transition, consistent with the observation. It is possible that the weak 2.5-fold effect for this transition results from an enhancement of the minor A·C mismatch. Finally, the 3-fold increase in the A·T→C·G transversions may be due to increased T·C mispairing, although this remains speculative. Overall, it can be concluded that the lacZ reversion system is an informative tool to investigate the correlations between dNTP pools and replication error rates. We also wish to highlight that a four-fold increase in the dCTP concentration leads to a 30- to 40 fold in the observed error rate, signifying that in vivo exonucleolytic proofreading is a powerful fidelity instrument. The lack of a significant dcd effect on the rifampicin-resistant mutants is discussed in section 4.4.
The pool changes in ndk are distinct from that of dcd and, hence, mutational consequences are expected to be different, as is indeed observed (Tables 1 and and2).2). In the ndk mutant there is a 2-fold enhancement of the dCTP pool and a 4-fold decrease in the dATP pool (Table 3). Continuing the logic developed above, we use the scheme of Fig. 3 to make possible correlations between the mutational specificity and pool changes. It is clear that the strong ndk mutator effect for the A·T→T·A transversions can be rationalized by the decrease in dATP (correct nucleotide) and simultaneous increase in dCTP (next nucleotide). None of the other mispairings is subject to such a double promotion, and this is fully consistent with the unique enhancement of the A·T→T·A transversions. The remaining, modest frequency increases for the G·C→A·T, G·C→T·A and A·T→G·C substitutions are consistent with a single modest effector (2-fold enhancement of dCTP) or a combination of a positive and negative effects.
The correlations between dNTP pool changes and mutational specificity can also be investigated using the rifampicin-resistance marker, which yields mutations in the rpoB gene rendering RNA polymerase resistant to rifampicin. Our data show an only very modest (2-fold) mutator effect for the dcd mutant, but a strong (at least 10-fold) for the ndk mutant. While we have not sequenced the rifampicin-resistant mutants in the present study, conclusions on this issue can be readily drawn based on the studies by Miller et al.  and Garibyan et al. , which provide detailed information on the types of mutations and their sites of occurrence in the rpoB gene in the MMR-defective background, including the events specifically enhanced by the ndk mutator .
The main hotspot for rpoB mutants in the MMR-defective ndk strain is an A·T→G·C transition at position 1547 (22 events out of 33 total) . This is also the main hotspot in the single MMR-defective strain (29 out of 57) and therefore reflects the occurrence of the most frequent DNA replication error. The local sequence context for position 1547 is 5′-GTC-3′ (1547 underlined). Assuming that the A·T→G·C at this position reflects Ttemplate·G mispairing, the incorrect dGTP competes at the misinsertion step with the correct dATP, while dCTP is the next nucleotide. It is clear from the earlier analysis that both ndk-mediated dNTP pool changes (lower dATP, higher dCTP) will cooperate to drive this event. The second rpoB hotspot (position 1534) induced by ndk is also a 5′-GTC-3′ sequence and its increase can be similarly accounted for. We note here that, as opposed to the two indicated hotspots in rpoB, the A·T→G·C transition in the lac system (Table 2), is not significantly elevated, as in this case dATP is the next nucleotide hence providing for an antimutator effect rather than mutator effect at the proofreading step (dATP is 4-fold down).
In contrast to the ndk strain, the dcd strain is not a strong mutator for rifampicin-resistant mutations. As a strong mutator effect can only be expected if there is an enhancement of the existing main events in the MMR-defective background, one would again be looking for enhancement of the A·T→G·C mutations at the 5′-GTC-3′ of positions 1547 and 1534 . While the increase in dCTP would drive the extension reaction, the Ttemplate·G mispairing would not be enhanced and, in fact, would be reduced by the 2-fold lowering of the dGTP concentration (Table 3) (dGTP is the misinserting nucleotide). Thus, in order to enhance rifampicin-resistant mutations higher dCTP and either elevated dGTP or lowered dATP is needed. This requirement is met by the ndk mutator but not the dcd mutator.
Our analysis of dNTP pool changes and associated mutational effects indicates that the rules for the dNTP-dependence of replication fidelity, as uncovered based on studies with isolated DNA polymerases, can be applied to the behavior of in vivo replicases. Our case concerns the bacterium E. coli, but similar assumptions have been applied to the yeast S. cerevisiae . A notable feature that has emerged from our study is the strong fidelity role played by the ‘next-nucleotide’ dNTP: strong mutagenic effects are observed due to increased levels of dCTP in both the lac and rpoB systems when it serves as a next nucleotide. This is not only consistent with the notion that proofreading is an important fidelity in vivo mechanism [38, 43], but also suggests that a main reason for cells to keep the dNTP concentrations low (in the micromolar level, as opposed to the millimolar level for the NTPs) is to be able to control the error rate of the replication process.
Our data and analysis yield plausible correlations between dNTP changes and resulting mutational consequences. These correlations support the previously proposed model that, indeed, this particular mutator effect may result from the observed dNTP pool changes . This is important because it has been argued elsewhere that simple dNTP changes are not sufficient to explain the ndk mutator effect . The strongest argument in favor of this viewpoint is results in which a clone of the human NDK homolog nm23-H2 was able to reduce (to near wild-type level) the frequency of rifr mutants of an ndk strain  (as also demonstrated in ); however, when the dNTP pools were determined in the complemented strain, distorted pools could still be observed, seemingly unlinking the relationship between mutator effect and dNTP changes. However, caution is required with conclusions of this type. Correlations between dNTPs and mutations are complicated, requiring interpretation in terms of both correct and incorrect nucleotides as well as next nucleotides, and while these correlations can be drawn satisfactorily as shown here, the correlations are not necessarily quantitative, particularly when considering the next-nucleotide effect. For example, a four-fold increase in the dCTP concentration, as a next nucleotide, can be responsible for an up to 42-fold increase in A·T→T·A transversions (dcd, see Table 1, Fig. 2). Thus, while dNTP changes in nm23-H2-complemented ndk strains were judged to be small and possibly not significant , it cannot be excluded that they are, in fact, consistent with the observed loss of mutability. The conclusions of Nordman and Wright  are based on measurement of rifampicin-resistant mutants. As noted, these events, in nkd strains are dominated by A·T→G·C transitions at a few hotspot sites, in particular the two 5′-GTC-3′ sequences at positions 1534 and 1547 (Miller et al. 2002). For these events, in addition to the increased dCTP (the next nucleotide), the dGTP/dATP (incorrect/correct) ratio appears to be a critical factor (see section 4.4). Importantly, as complementation of ndk with the human NDK homolog nm23-H2 appeared to substantially restore the dATP concentration (from 0.27 to 0.76 of wild-type level) , this restoration may provide a plausible explanation for the reported loss of mutator activity.
As second aspect that requires consideration is the role of DNA mismatch repair (MMR). Our present studies were conducted in the MMR-defective strain background to focus directly on replication errors. The complementation studies of ndk with nm23-H2 were conducted in both MMR-proficient and MMR-defective backgrounds . While in the MMR+ background, complementation led to loss of mutability, this was not the case in the MMR- background (Table S1 in ). The latter fact clearly indicates that in the nm23-H2 complemented ndk strain the enhanced replication errors are still occurring, thus preserving the correlation between dNTP pool changes and mutability. These observations suggest that the functioning of the DNA mismatch repair system may be a critical factor in shaping the ndk mutator effect. It is to be noted that the efficiency of mismatch repair can be compromised, wholly or in part, by the occurrence of frequent mispairing errors [44-46], and as a consequence modest changes in error rates can have disproportional effects (both positive and negative) on the observed mutation rates. Thus, mismatch-repair proficient strains are not an optimal vehicle for studying the effects of altered dNTPs on replication fidelity.
As an alternative to the dNTP hypothesis it was postulated that the ndk mutator effect is due to elevated dUTP levels . This proposal was based on the strongly increased mutability observed for ndk dut strains, which are additionally defective in dUTPase activity (dut gene product). While this high mutability is intriguing, no direct evidence for this hypothesis is yet available. For example, no dUTP levels have been measured or reported for these strains. In addition, dUTP is not generally considered a mutagenic compound. While dUTP can go readily into DNA in place of dTTP and, once in the DNA, can be excised by uracil glycosylase generating repair intermediates such as abasic sites, neither uracils nor abasic sites are considered to be particularly mutagenic, as evidenced, for example, by the lack of mutability of dut strains . While abasic sites can be mutagenic under conditions of SOS-induction , this type of damage-dependent mutagenesis is not generally consistent with the A·T→G·C specificity observed in the rpoB (and lacI) gene target . In view of the above, we suggest that the direct mispairing model remain the simplest, minimal hypothesis to explain the ndk mutator effect. Obviously, additional experimentation is required to settle these issues.
The authors thank Linda Wheeler, Ronnie Dunn, and Pamela Allen for their excellent technical assistance in this project, and Damian Gawel and Scott Lujan of the NIEHS for their helpful comments on the manuscript for the paper. This work was supported by project Z01 ES065086 of the Intramural Research Program of the National Institute of Environmental Health Sciences (NIEHS), and NIH Grant R01 GM073744 and NIEHS Center Grant ES000210.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Conflict of interest
The authors declare that there are no conflicts of interest