Effect of dNTP pool alterations on fidelity of leading and lagging strand DNA replication in E. coli

Abstract

The fidelity with which organisms replicate their chromosomal DNA is of considerable interest. Detailed studies in the bacterium Escherichia coli have indicated that the fidelity of leading- and lagging-strand DNA replication is not the same, based on experiments in which the orientation of certain mutational targets on the chromosome was inverted relative to the movement of the replication fork: different mutation rates for several base-pair substitutions were observed depending on this orientation. While these experiments are indicative of differential replication fidelity in the two strands, a conclusion whether leading or lagging strand is the more accurate depends on knowledge of the primary mispairing error responsible for the base substitutions in question. A broad analysis of in vitro base-pairing preferences of DNA polymerases led us to propose that lagging-strand is the more accurate strand. In the present work, we present more direct in vivo evidence in support of this proposal. We determine the orientation dependence of mutant frequencies in ndk and dcd strains, which carry defined dNTP pool alterations. As these pool alterations lead to predictable effects on the array of possible mispairing errors, they mark the strands in which the observed errors occur. The combined results support the proposed higher accuracy of lagging-strand replication in E. coli.

1. Introduction

High accuracy DNA replication is a fundamental process that allows the transfer of genetic information to the next generation with minimal information loss. Although the fidelity of DNA replication is high, errors during this process are a significant source of spontaneous mutations. High fidelity DNA replication is achieved by a combination of activities, among which are (i) the discrimination between correct and incorrect nucleotides during the insertion step of DNA polymerization, (ii) removal of incorrectly inserted nucleotides by an exonucleolytic proofreading step, and (iii) postreplicative DNA mismatch repair, in which newly synthesized DNA is surveyed for mismatches, which are then corrected using the sequence of the (correct) parental strand as template [1–4]. These three processes, which are broadly conserved among all forms of life, are capable of reducing the error rate per base pair replicated to values 10⁻⁹ to 10⁻¹¹ [4, 5].

While these processes comprise a broad outline, several important aspects still need to be understood. One particular issue is whether DNA replication occurs with similar fidelity in the leading and lagging strands of the replication fork. This issue is pertinent because of the slightly different enzymology of leading and lagging strand replication: due to the antiparallel nature of the two DNA strands, synthesis of the leading strand proceeds, continuously, in the same direction as the moving replication fork, while the lagging strand is synthesized discontinuously in the form of small sections called Okazaki fragments [6]. Because of this enzymatic dichotomy, fidelity differences may occur. Also, in eukaryotic organisms the two strands are replicated by different DNA polymerases, like polymerases δ and ε that replicate the eukaryotic lagging and leading strands, respectively [7, 8].

Studies from our laboratories have focused on understanding DNA replication fidelity in the bacterium Escherichia coli. In this model system, both strands are replicated by DNA polymerase III [6]. Specifically, the chromosome is replicated by a dimeric DNA polymerase complex termed DNA polymerase III holoenzyme (HE), whose composition is (αεθ)₂β₂τ₂(γδδ'χψ) [6]. The complex contains two core polymerases (αεθ), one for each strand. Within the core complex, α is the polymerase subunit, ε the associated proofreading subunit, and θ a stabilizing subunit for ε [9, 10]. Importantly, despite the fact that in E. coli the leading and lagging strand are replicated by an identical polymerase (Pol III core), our studies have revealed that the two strands are replicated with differential fidelity [11–16], thus revealing an intrinsic strand-dependent phenomenon, the basis for which we would like to understand.

Our results are based on measurements of lac reversion frequencies in pairs of strains in which the lac operon is inserted in the two opposite orientations relative to the replication fork [11]. In this manner, a DNA sequence of interest will be synthesized, for example, as part of the leading strand in one orientation, but as part of the lagging strand in the other orientation. The results using a series of four lac base substitution alleles revealed in each case a different lac reversion frequency between the two orientations [11–16]. These experiments were performed in a strain background lacking the postreplicative DNA mismatch repair system (mutHLS system) [3], facilitating interpretation of the mutant frequencies in terms of DNA replication error rates. It was concluded that on the E. coli chromosome leading and lagging strands have a different fidelity [11]. The differences varied by the type of base substitution investigated, but were in the 2- to 4-fold range [11–16].

While these experiments indicate differential replication fidelity for the two strands, they do not yet provide direct information about which of the strands is replicated more accurately. For example, when scoring A·T → T·A transversion mutations, these could be arising from A·A mispairings during copying of the A-containing strand or from T·T mispairings during copying of the T-containing strand, and unless a decision is made which of the two mispairs (A·A or T·T) is the main contributor to the observed mutations, a decision which strand is more accurate cannot be made. In a previous report we analyzed the properties of a variety of DNA polymerases that had been studied in vitro with regard to their misinsertion preferences and ability to extend a variety of base·base mismatches (extension of mismatches is needed for replication errors to be expressed as mutations), and made predictions for the corresponding properties of E. coli DNA polymerase III in vivo. On this basis, we postulated that A·T→T·A transversions would derive primarily from T·T errors (template base stated first and underlined), G·C→T·A transversions from C·T errors, G·C→A·T transitions from G·T errors, and A·T→G·C transitions from T·G errors [11]. Using these assumptions it emerged that, in four out of four cases examined, the lagging strand is the more accurate strand. A model as to why lagging-strand replication might be more accurate was discussed [11].

In the present work, we have further investigated this question of strand assignment. We make use of certain mutant of E. coli that possess alterations of their dNTP pools. The dNTPs (5'-deoxynucleoside triphosphates) are the DNA precursors used by the polymerases, and they have been shown to be critical factors in the ability of DNA polymerases to synthesize DNA accurately. First, at any given DNA template position, the one correct dNTP has to compete for insertion with the three incorrect ones [17, 18]. While polymerases perform this discrimination fairly well, the actual rate of misinsertion is directly proportional to the ratio of incorrect to correct dNTPs. Hence, mutants containing altered dNTP pools are expected to be mutators for certain base substitution mutations, as is indeed the case [19–21]. A second reason why dNTP pool mutants could display a mutator phenotype is because of the so-called “next nucleotide effect” [1, 22]. Here, elevated dNTP levels can enhance extension of base·base (or other) mismatches by promoting insertion of the next nucleotides following the mismatch. As the mismatch extension step competes directly with the removal of the erroneous base by proofreading, elevated dNTPs can be mutagenic through this mechanism. When the specific changes in the dNTP pool mutants are known, clear predictions can be made as to the mispairings that will be promoted [19, 21]. These predictions can then be used to “mark” the strands and to verify the earlier conclusion that lagging strand replication is more accurate than leading-strand replication. For this purpose, we use ndk and dcd mutants of E. coli, defective in nucleoside diphosphate kinase (ndk gene) and dCTP deaminase (dcd gene), respectively, both of which are characterized by defined dNTP pool changes and a corresponding mutator phenotype [21]. The strains were also defective in the mutHLS DNA mismatch-repair system to facilitate the interpretation of the mutation rates in terms of replication error rates. Our analysis of the mutational responses for several lac alleles in the two orientations on the chromosome fully supports our previously deduced higher accuracy of lagging-strand replication.

2. Materials and methods

2.1 Media

Solid and liquid media were as described [23]. Solid minimal medium (MM) contained 0.5% glucose or 0.4% lactose as a carbon source, 50 μg/ml of L-proline, and 5 μg/ml of thiamine. MM solid medium was further supplemented with 50 μg/ml of thymidine for experiments involving dcd strains. Antibiotics, where required, were added at the following final concentrations: ampicillin, 25 μg/ml; chloramphenicol, 20 μg/ml; kanamycin, 25 μg/ml; rifampicin, 100 μg/ml; and tetracycline, 12.5 μg/ml.

2.2 Bacterial strains

The E. coli strains used in the mutagenesis experiments are all derivatives of strains EC3114, EC3120, EC3126, EC3132, EC3138, EC3144, EC3150, and EC3156, which have been described [11]. They contain a Δlac deletion at the normal lac location near 8 min on the chromosome, but carry a reinserted lac operon at the bacteriophage lambda attachment site (attB) near 17 min. This lac operon was derived from strains CC102, CC104, CC105 and CC106 [24], which each carry a defined lacZ missense mutation that allows scoring of mutagenesis via reversion to Lac⁺ by G·C→A·T, G·C→T·A, A·T→T·A or A·T→G·C base substitution [24]. Each lac operon is inserted in the two possible chromosomal orientations (termed L and R, arbitrarily) [11] to permit investigation of the effects of gene orientation on mutagenesis and DNA replication fidelity. To more directly on focus replication fidelity, all strains used were also defective in DNA mismatch repair (mutL::Tn5 or mutL::Tn10) [11]. The dcd allele (dcd-12::Tn10dkan) was inserted by P1 transduction from strain BW1040 [25] using kanamycin resistance as selective marker. The ndk allele (Δndk::cam) was inserted by P1 transduction from NR11814 [21] using chloramphenicol resistance as selective marker.

2.3 Lac and rif^r mutant frequency measurements

For each strain, a total of 20 independent LB cultures (2 ml each) were initiated from single colonies (one colony per tube). The colonies were taken from three to nine independent isolates (transductants) for each strain. The cultures were grown to saturation at 37°C on a rotator wheel. One hundred microliters of the cultures, diluted to 10⁻⁶, were plated on MM glucose plates to determine the total cell count, while 100 μl of the undiluted cultures were plated on MM-lactose plates to determine the number of Lac⁺ mutants. To measure the frequency of rifampicin-resistant cells, 100 μl of undiluted or 10⁻¹ diluted cultures were plated on LB plates containing rifampicin. Plates were incubated for 20 h (LB) or 40 h (MM) at 37°C. To calculate mutant frequencies, the average number of mutants per plate was divided by the number of total cells plated. Statistical analysis was performed using the software program Prizm (GraphPad).

3. Results

3.1 The chromosomal lac reversion system

In Fig. 1, we present a set of diagrams outlining the DNA polymerization events that can lead to reversion of each of the lacZ alleles used in this study. The alleles are specific in the sense that, for each allele, the indicated missense codon can be reverted by only a single, defined base substitution to the GAG codon encoding the critical Glu461 residue of β-galactosidase (lacZ gene product) [24]. These lac alleles were first created on an F'prolac [24], but were transferred to the chromosome by insertion of the corresponding lac operons into the attB insertion site of bacteriophage lambda in the two possible orientations [11]. The two orientations, arbitrarily termed L and R [11], are shown in Fig. 1 along with the specific mispairing events that would revert the missense codons to the wild-type GAG codon. For example, in Fig. 1A, the lac allele originally present in strain CC104 [24] contains the GCG missense codon in either leading (L-orientation, left side of diagram) or lagging strand (R-orientation, right). Reversion of the GCG to GAG can occur by either C·T or G·A mispairing (both yielding the necessary G·C→T·A transversion). The C·T mispairing would be promoted by increased levels of dTTP (incorrect nucleotide) relative to dGTP (correct nucleotide), and/or increased levels of dCTP, the next (correct) nucleotide. The requirements for promoting the G·A mispairing are different: enhanced dATP (incorrect nucleotide) relative to dCTP (correct nucleotide) and by enhanced dGTP (next correct nucleotide). In view of these distinguishing requirements for promoting the possible mispairs, alterations of the dNTP pool are likely to be a useful tool for deciding which particular mispair is primarily responsible for the lac reversions and, hence, for deciding which strand is replicated more accurately. Below, we describe results with the dcd and ndk pool-bias mutants.

Fig 1.

Diagram for analyzing the effect of dNTP alterations in the dcd strain of E. coli. Panels A, B, C, and D show, respectively, the potential mispairing events responsible for the indicated G·C→T·A, A·T→T·A, G·C→A·T or A·T→G·C base-pair substitutions in the lacZ gene. These events revert the indicated lacZ missense codon (GCG, GTG, GGG, and AAG, respectively) to GAG, restoring the wild-type genotype. The lacZ missense alleles were first described residing on an F'prolac episome by Cupples and Miller [24], but in the current study they are present on the E. coli chromosome in the two possibly orientations (indicated as L or R) relative to the replication fork, as described [11]. For each base substitution, we present the two possible mispairs that can give rise to the required base substitution. It can be seen that, when comparing the two lac orientations, a given mispair occurring in the leading strand in one orientation occurs in the lagging strand in the other orientation, and vice versa. The mispair that we have proposed in a previous study [11] to be the primary contributor to the base substitution is indicated in bold and surrounded by a rectangle (for example, the C·T mispair for the G·C→T·A transversion of panel A). With arrows, we indicate the correct dNTP for this particular position as well as the next dNTP to be incorporated following the mismatch. The analysis focuses on the effects of the changes in dCTP and dGTP levels, which are primarily affected in the dcd strain (Table 1). When the dCTP and dGTP changes are predicted to be mutagenic, the dNTP is represented in RED color, while dCTP or dGTP changes predicted to be antimutagenic are presented in BLUE. As described in the text, both the competition between the incorrect and the correct nucleotide at the misinsertion step, and the concentration of the next dNTP are determining factors for the mutational outcome. In panel D, we also show the 5' T that precedes the AAG missense codon, because it is the template base that needs to be copied following an A·C mispairing event. See text for more details.

3.2 Effect of dcd on lac reversions

The E. coli dcd mutant is defective in dCTP deaminase, which performs the deamination of dCTP to produce dUTP, an important step in de novo dTTP biosynthesis [25]. The dUTP is converted to dUMP by E. coli dUTPase, which is then converted to dTMP by thymidylate synthase, followed by conversion to dTDP and dTTP. Due to their lack of deaminase activity, dcd mutants accumulate dCTP [21, 26, 27], which is one contributor to the elevated mutation level that they display (mutator phenotype) [21]. Despite their deaminase defect, dcd strains still retain some ability to synthesize dTTP through alternative pathways [28], although the dTTP concentrations are significantly reduced when cells are grown in minimal medium [21, 26, 27]. In the present series of experiments, cells were grown in LB medium, and dNTP measurement have shown the dTTP concentration to be essentially normal, presumably due to the presence of thymine and/or thymidine in the medium, which permit synthesis of dTTP through salvage pathways [28]. The observed dNTP levels in dcd strains are shown in Table 1 (from [21]). It is seen that the dCTP concentration is increased nearly four-fold, while the dGTP concentration is decreased by about 2-fold.

Table 1.

dNTP pools in dcd and ndk strains (pmoles/10⁸ cells)^*

Strain	dATP	dTTP	dCTP	dGTP
wild-type	16 ± 7	15 ± 6	75 ± 19	8.0 ± 2.4
dcd	11 ± 5	13 ± 7	277 ± 39	3.8 ± 5
ndk	4.0 ± 4.8	23 ± 7	150 ± 18	7.5 ± 1.7

^*Data from Schaaper and Mathews [21]

In Table 2, we present the mutant frequencies of the four studied lac reversions in dcd⁺ (wt) and dcd strains. The strains are also mismatch-repair deficient to facilitate interpretation of the mutant frequencies in terms of replication errors. For the wt strain the results reiterate the previous findings that the two orientations yield different reversion frequencies. For G·C→T·A, A·T→T·A, and G·C→T·A, the L-orientation yields the higher frequency, while for the A·T→G·C the R-orientation is higher. Typically, the frequency differences are 2- to 4-fold, as observed in multiple repeated experiments [11–16].

Table 2.

Mutator effects of the dcd deficiency for lac reversion of two chromosomal orientations (L or R) of the lac operon

		lac+ per 108 cells in indicated strain		Mutator effect
lac reversion	lac orientation	wt	dcd
G·C→T·A	L	1.5 ± 0.2	8.2 ± 1	5.5
	R	0.75 ± 0.2	4.2 ± 1	5.6
A·T→T·A	L	2.2 ± 0.2	8.5 ± 0.4	3.9
	R	1.6 ± 0.2	5.0 ± 0.1	3.1
G·C→A·T	L	78 ± 4	113 ± 51	1.5
	R	31 ± 2	20 ± 21	0.6
A·T→G·C	L	34 ± 3	69 ± 6	2.2
	R	83 ± 4	147 ± 8	1.8

All strains are mutL. The mutator effect indicates the increase in mutant frequency between dcd and wild-type (wt) strains.

Table 2 also clearly shows the mutator effect of the dcd strain, most prominently for the G·C→T·A and A·T→T·A transversions (3.1- to 5.6-fold). Overall, these dcd mutator effects are smaller than those reported recently for the same lac alleles residing on F'prolac [21], likely reflecting certain features specific to the F' replication. Nevertheless, the dcd mutator effects observed here for the chromosomal lac location should be informative with regard to the preferences of in vivo mispairings and, hence, the strand bias of DNA replication fidelity. We also note that the mutator effects are similar or near-identical regardless of the lac orientation. A full analysis will be presented in the Discussion.

3.3 Effect of ndk on lac reversions

The ndk gene encodes the enzyme nucleoside diphosphate kinase (Ndk), a non-specific enzyme responsible for phosphorylating all cellular nucleoside diphosphates (NDPs and dNDPs) to the corresponding triphosphates [29, 30]. Initially thought to be essential, an ndk deletion mutant was found to be viable [31]. However, not unexpectedly, the strain showed altered dNTP pools as well as increased mutability (mutator phenotype) [32]. It was assumed that another cellular kinase could replace Ndk, although not with the same efficiency and specificity. This alternative activity was shown to be adenylate kinase [32].

The dNTP levels observed in ndk strains are presented in Table 1: the dATP level is reduced by about four-fold, while there are increases for dCTP (2.0-fold) and, possibly, dTTP (1.5-fold). Table 3 also clearly shows the mutator effect of the ndk strain. Like is the case on the F', the chromosomal mutator effect is strongly specific for the A·T→T·A transversion (23- to 26-fold), with only small effects for the G·C→T·A transversion [21]. The two transitions were not monitored in these experiments. As in the case of the dcd mutator, we note that the increases are occurring similarly for both orientations.

Table 3.

Mutator effects of the ndk deficiency on lac reversion for two chromosomal orientations (L and R) of the lac operon

		lac+ per 108 cells in indicated strain		Mutator effect
lac reversion	orientation	wt	ndk
G·C→T·A	L	1.1 ± 0.1	1.7 ± 0.3 (1.5)	1.5
	R	0.33 ± 0.1	1.1 ± 0.2 (3.3)	3.3
A·T→T·A	L	1.4 ± 0.2	36 ± 1 (26)	26
	R	0.78 ± 0.1	18 ± 0.5 (23)	23

All strains are mutL. The mutator effect indicates the increase in mutant frequency between ndk and wild-type (wt) strains.

3.4 Frequency of rifampicin-resistant mutants in lac-inverted strains

As a control in the above-described experiments, we also routinely determined the frequency of rifampicin-resistant (Rif^r) mutants, which result from mutations in the rpoB gene [33]. As shown in Table 4, the dcd and ndk defects result in a mutator effect for Rif^r mutations (1.5-fold and 20-fold, respectively). However, as the rpoB gene is not subject to inversion, the frequency of rifampicin-resistant mutants is not significantly changed between strains with L-and R-oriented lac operons (Table 4), as expected.

Table 4.

Unchanged mutant frequencies towards rifampicin resistance in strains carrying the lac operon in L or R orientation on the chromosome

lac		Rifr per 108 cells in indicated strain
allele	orientation	wt	dcd	ndk
G·C→T·A	L	184 ± 5	240 ± 10	nd
	R	170 ± 4	240 ± 12	nd
A·T→T·A	L	190 ± 15	nd	5900 ± 800
	R	180 ± 14	nd	6200 ± 1000

nd = not determined. All strains are also mutL.

4. Discussion

The purpose of the current series of experiments was to provide support for our previous conclusion that on the E. coli chromosome the fidelity of lagging-strand replication is higher than that of leading-strand replication. We used the tool of alterations in the dNTP DNA precursor concentrations to make predictions as to which specific mispairings will be promoted or disfavored. The responsiveness of the tested lac base substitutions to the known pool changes should then be an indication whether the choice of preferred mispair is correct., page

In the Figs. 1 and 2 we present diagrams of replication forks traversing the relevant section of the lac operon in the two (L or R) orientations. Shown is the missense codon for each of the four monitored lac missense alleles: GCG, GTG, GGG, and AAG [24], as well as their complement in the opposite strand. For each case, we have also indicated the particular dNTPs that are relevant for the fidelity of the copying step. These include the correct dNTP competing with the creation of each mispair (its elevation will be antimutagenic at the misinsertion step) and the dNTP that follows the misinsertion (its elevation is predicted to be mutagenic). The mutagenic action of increased levels of the next nucleotide, generally known as the next-nucleotide effect, is well established from in vitro DNA polymerase reactions and has also been deduced to occur in vivo [19, 20]. Mechanistically, high levels of the next nucleotide promote extension of a mismatch and reduce the probability of removal of the misinserted base by proofreading [1, 22].

Fig 2.

Diagram for analyzing the effect of dNTP alterations in the ndk strain of E. coli. This analysis focuses on the effects of changes in dATP and dCTP levels, which are primarily affected in the ndk strain (Table 1). RED colors for dATP or dCTP indicate that the change in this dNTP is predicted to be mutagenic. BLUE colors indicate an antimutagenic effect. See the Legend to Fig. 1 and the main text for more details.

We will first analyze the results obtained with the dcd strain (Fig. 1). Diagram 1A concerns the production of the lac G·C→T·A transversions, which were enhanced by 5- to 6-fold in the dcd strain (Table 2). In our previous studies we have proposed that the C·T mismatch was the primary mismatch responsible for these events [11]. In the diagrams we have indicated those dNTPs whose changes are predicted to be mutagenic in RED color, while the antimutagenic dNTP effects are indicated in BLUE. In the dcd strain the dCTP level is enhanced while the dGTP level is reduced (Table 1). Hence, it is easy to see how the production of C·T errors would be enhanced: dGTP is the correct nucleotide and its reduction is predicted to be mutagenic, while dCTP, the next nucleotide, is enhanced, which should also be mutagenic. In contrast, the reciprocal G·A error is not expected to be enhanced. In fact, it is expected to be reduced: the dCTP elevation will be antimutagenic, as it is the correct nucleotide, while the decrease in dGTP, the next nucleotide, will disfavor extension of the mismatch. Thus, these data are fully consistent with our earlier assumption that the primary cause of the G·C→T·A transversions is C·T mispairing (boxed). The similar enhancement (mutator effect) for the two gene orientations is further consistent with the implicit assumption that the same mispair underlies the mutation in either strand.

The dcd mutator also enhances the lac A·T→T·A transversion (Table 2). Here, the T·T mismatch (boxed) was proposed to be the major contributor [11]. The diagram Fig. 1B indicates that there is no predicted effect for T·T mispairing at the misinsertion step (both the correct dATP and incorrect dTTP are unchanged), but extension of the T·T will still be promoted by the increase in dCTP, the next nucleotide. In contrast, any A·A mismatches that might occur are not predicted to be promoted. In fact, the lowering of dGTP would disfavor any extension. Thus, the previous choice of T·T [11] is supported by the dcd mutator data.

Modest mutator effects are observed for the two transitions in the dcd strain. Fig. 1C for the G·C→A·T transition, for which we favor the G·T mismatch (boxed) [11], shows that the elevation of dCTP would have two opposing effects on the G·T mispair, as dCTP is both the correct nucleotide (antimutagenic effect) and the next nucleotide (mutagenic effect). An identical effect is predicted for the reciprocal C·A error. Here, dGTP is both the correct nucleotide and the next nucleotide, and the effects of its reduction in the mutant may cancel out. Thus, no strong conclusions regarding this base-pair substitution can be drawn.

For the A·T→G·C transitions of Fig. 1D, where we favor the T·G mismatch (boxed) [11], somewhat similar arguments apply. For the T·G events, the dGTP decrease of the dcd mutant would be antimutagenic (it is the incorrect nucleotide). For the reciprocal A·C mismatch, the increase in dCTP is predicted to be mutagenic (it is the incorrect nucleotide). Thus, while no direct positive evidence is obtained for the prevalence of the T·G mismatch, indirect evidence still supports this notion. Indeed, if the A·C mismatch were the major contributor to the A·T→G·C transitions, then the strongly elevated dCTP concentration (incorrect nucleotide) would be predicted to have a strong mutagenic effect, which is not observed.

The data obtained for the ndk strain are also informative. Between the two tested transversions, the A·T→T·A transversion is strongly enhanced (Table 3). The analysis of the ndk effects is shown in the diagrams of Fig. 2. As the ndk strain is characterized by elevated dCTP and lowered dATP (Table 1), it can be clearly seen that for the A·T→T·A transversion (Fig. 2B) the T·T mismatch is favored by both the higher dCTP (next nucleotide) and the lowered dATP (correct nucleotide). In contrast, the reciprocal A·A mismatch would be disfavored by the low dATP (incorrect nucleotide). Thus, these data strongly support the proposed T·T mismatch as the main contributor to the A·T→T·A transversions.

With regard to G·C→T·A transversions, only a small ndk mutator effect is observed (Table 3). From the diagram of Fig. 2A, one might argue that extension of the C·T mismatch would be promoted by the elevated dCTP level (next nucleotide), predicting a possible mutator effect. No mutator effect is predicted for the reciprocal G·A mismatch, as high dCTP (correct nucleotide) and low dATP (incorrect nucleotide) would both disfavor the event. It is possible that the modest dCTP elevation in the ndk strain is not substantial enough for a strong enhancement of this event. We note that the G·C→T·A is significantly enhanced in the dcd strain, which is characterized by a larger dCTP increase than in the ndk strain and a decreased dGTP level (both favorable).

In summary, the current data provide solid support for several of our assumptions about the predominant mispair responsible for the lacZ base substitutions used in our leading/lagging strand system. We also note that the dcd or ndk mutator effects are found to be very similar or near identical for the lac L or R orientation, consistent with the notion that the same mismatch is responsible for the mutational events in leading and lagging strands. These data strengthen our conclusion that on the E. coli chromosome the fidelity of lagging-strand replication is higher than that of the leading strand. From Tables 2 and 3 we see that for the G·C→T·A, A·T→T·A and G·C→T·A substitutions the L orientation has the highest mutant frequency, while for the A·T→G·C the R orientation has the highest frequency. In all four cases (see Figs. 1 and 2), this corresponds with the preferred (boxed) mispair residing in the leading strand, thus indicating that leading-strand synthesis is more error prone. Although this strand-dependent fidelity difference is modest, about 2-fold in the present study, it has been observed consistently, at the 2- to 4-fold level, in a series of extended studies [11–16].

It is our contention that this fidelity bias does not reflect a difference in the intrinsic polymerase misinsertion error rate in the two strands, but rather a difference in the subsequent processing of the error [11, 12]. How there can be more effective error processing in the lagging strand is a question that remains to be answered. It is attractive to speculate that improved lagging-strand fidelity results from the more dissociative properties of the polymerase in this strand, permitting more efficient proofreading (note that proofreading is essentially a dissociative process in which the primer terminus is moved from the polymerase active site to the distant exonuclease site). Alternatively, the dissociative character of the lagging-strand polymerase may lead to increased removal of the terminal mismatch by the exonuclease of Pol II, which has been shown to proofread for Pol III under some conditions [12, 13, 34, 35] or by a third copy of Pol III that has been proposed to be present at the replication fork [36].

Highlights

• In E. coli leading and lagging strand are replicated with different fidelity

• Lagging strand replication is more accurate

• dNTP pool alterations can be used to mark the strands

• The fidelity difference may relate to differential removal of terminal mismatches