Replication fidelity in E. coli: differential leading and lagging strand effects for dnaE antimutator alleles

Abstract

DNA Pol III holoenzyme (HE) is the major DNA replicase of Escherichia coli. It is a highly accurate enzyme responsible for simultaneously replicating the leading- and lagging DNA strands. Interestingly, the fidelity of replication for the two DNA strands is unequal, with a higher accuracy for lagging-strand replication. We have previously proposed this higher lagging-strand fidelity results from the more dissociative character of the lagging-strand polymerase. In support of this hypothesis, an E. coli mutant carrying a catalytic DNA polymerase subunit (DnaE915) characterized by decreased processivity yielded an antimutator phenotype (higher fidelity). The present work was undertaken to gain deeper insight into the factors that influence the fidelity of chromosomal DNA replication in E. coli. We used three different dnaE alleles (dnaE915, dnaE911, and dnaE941) that had previously been isolated as antimutators. We confirmed that each of the three dnaE alleles produced significant antimutator effects, but in addition showed that these antimutator effects proved largest for the normally less accurate leading strand. Additionally, in the presence of error-prone DNA polymerases, each of the three dnaE antimutator strains turned into mutators. The combined observations are fully supportive of our model in which the dissociative character of the DNA polymerase is an important determinant of in vivo replication fidelity. In this model, increased dissociation from terminal mismatches (i.e., potential mutations) leads to removal of the mismatches (antimutator effect), but in the presence of error-prone (or translesion) DNA polymerases the abandoned terminal mismatches become targets for error-prone extension (mutator effect). We also propose that these dnaE alleles are promising tools for studying polymerase exchanges at the replication fork.

1. Introduction

Synthesis of genomic DNA is performed with high accuracy in all organisms. The frequency of replication errors in vivo in various organisms has been estimated to be as low as one error per 10^10–11 of replicated base pairs [1,2]. This low level of replication errors is carefully controlled by several mechanisms. One important factor is the accuracy of the replicative DNA polymerase, as determined by its base selectivity and proofreading activity [3]. In Escherichia coli, the major replicase is the multi-subunit DNA polymerase III holoenzyme (Pol III HE), which carries out the coordinated simultaneous replication of the two DNA strands. The Pol III alpha subunit (encoded by dnaE) has the DNA polymerase activity while the associated epsilon subunit (encoded by dnaQ) has exonuclease proofreading activity [4]. The postreplicative mismatch repair system (encoded by the mutHLS genes) is a critical additional fidelity factor that serves to remove replication errors that escaped the proofreading step [3].

Due to the antiparallel nature of the two DNA strands and the fact that synthesis proceeds only in 5’ to 3’ direction, the replication mechanisms of the two strands have to be different, with the leading strand being synthesized more continuously while the lagging strand is replicated discontinuously in the form of short Okazaki fragments [5]. Previously, using specific lac reversion and forward mutagenesis assays, our laboratories showed that in E. coli, where both DNA strands are replicated by the same DNA polymerase (Pol III HE), the in vivo fidelity of the two strands differed significantly. Further analysis indicated that it is the discontinuous lagging strand that is copied with significantly higher accuracy than the leading strand [6,7]. We proposed that the more dissociative character of the lagging-strand polymerase is the main factor contributing to the fidelity differences between the two strands [7]. In particular, dissociation from terminal DNA mismatches, which are potential mutations, would greatly reduce the ultimate mutation rate. In support of this model, an E. coli strain encoding a mutant dnaE gene (dnaE915) characterized by reduced processivity and reduced catalytic activity [7], displayed an antimutator phenotype.

At the same time, dissociation from terminal mispairs could also provide increased access for accessory DNA polymerases present in the cell. E. coli possesses four such accessory DNA polymerases. Among them, DNA Pol I (polA), DNA Pol II (polB) are considered accurate (error- free) polymerases, in part due to their 3’→5’ exonuclease activity. On the other hand, DNA Pol IV (dinB) and Pol V (umuDC) are error-prone polymerases lacking proofreading. In the previous studies, we showed that, depending on the conditions, each of the four accessory polymerases can indeed have occasional access to the replication point, with variable consequences for the mutation rate [8–12]. Interestingly, in each case this access was shown to be preferential to the lagging strand. Overall, it follows that chromosomal DNA replication fidelity, including its “strandedness’, is a complicated matter that depends on several factors, including strand-specific replication mechanisms, the participation of different DNA polymerases, polymerase ‘dissociability’, and presumably others.

In the present work, we have analyzed in detail the effects of the three dnaE alleles (dnaE911, dnaE915, and dnaE941) that exhibit strong antimutator phenotypes in a mismatch-repair-defective background, and hence are characterized by significantly lowered error rates [13–15]. They each carry a mutation in the alpha subunit of DNA Pol III [15,16]. We investigated (i) the strength of the antimutator effects, (ii) the strand dependence of these effects, and (iii) their behavior in the presence of one of several error-prone polymerases. The results show that each dnaE allele not only reduces the replication error rate, but does so preferentially in a normally more error-prone leading stand. Remarkably, in the presence of error-prone polymerases, these antimutator dnaE alleles display a mutator effect (higher overall error rate). We conclude that each of these phenotypes is in full support of the model in which increased ‘dissociability’ of the DNA polymerase serves as a fidelity factor, at least under conditions of normal DNA replication.

2. Materials and Methods

2.1. Strains construction and media

All strains were derivatives of E. coli MC4100 but have the lacIZYA operon inserted in the phage lambda attachment site (attB) in the two possible orientations (lacR and lacL) with respect to oriC [6]. The lac operon was in each case derived from strains carrying specific lacZ missense mutations that allow detection of specific base substitutions [17]. All strains were also mismatch-repair deficient (mutL::Tn5 allele transferred by P1 transduction from strain NR9559 (Fijalkowska and Schaaper 1995) to facilitate analysis of DNA replication errors [18]. Antimutator alleles of the dnaE gene were introduced by P1 transduction using selection for two flanking markers, zae-502::Tn10 and zae::Tn10d-Cam [14]. The sources of the dnaE911, dnaE915 and dnaE941 were strains NR9901, NR9905 and NR11185, respectively [15,16]. The presence of the antimutator alleles was confirmed by DNA sequencing. The recA730 allele was introduced by a two-step procedure described in Fijalkowska et al. 1995. In addition, all recA730 strains were made sulA366 [19]. Strains carrying the polBex allele were constructed as described in [10]. The AdinB::kan allele was transferred from YG7207 [20]. DNA Pol IV overexpression was achieved by transformation with low-copy plasmid pLO1, encoding dinB, or the empty vector plasmid pWSK129 as a control (described in [9]).

LB broth and minimal media (MM) were standard recipes [18]. MM plates contained 0.4% glucose or 0.4% lactose as the carbon source and 5 μg/ml of thiamine, and 50 μg/ml amino acids as necessary. Antibiotics, when required, were used at the following concentrations: ampicillin 50 μg/ml, kanamycin 50 μg/ml, tetracycline 15 μg/ml, chloramphenicol 25 μg/ml, rifampicin 100 μg/ml, and nalidixic acid 40 μg/ml.

2.2. Mutant frequency determination

To determine the frequencies of lac⁺, Rif^R or Nal^R mutants, individual cultures in 2 ml LB were grown overnight at 37°C. Appropriate dilutions were plated on MM glucose or LB plates to determine the number of viable cells and on the selective plates (MM lactose, LB Rif, or LB Nal) to determine the number of respective mutants. To calculate mutant frequencies, the number of mutants per plate was divided by the total number of cells. Tables present the median values of frequencies with 95% confidence intervals for 10–30 independent cultures. Leading/lagging strand assignments are based on the predominant mispairs for each base-pair substitution, as discussed previously [6–8,21].

3. Results

3.1. Three E. coli dnaE antimutator alleles show a preference for reducing the error rate of the leading strand of replication

To study the effect of several dnaE alleles on mutagenesis we used the lac reversion system described in [6]. In this system, several different lacZ alleles [17] were inserted into the E. coli chromosome in the two opposing orientations relative to the replication fork emanating from the replication origin, oriC. As a consequence, any given nucleotide in the lacZ target is replicated by leading-strand replication in one orientation and by lagging-strand replication in the other [6], and any differences in mutability can be correlated with strand-specific fidelity differences. The strains are also deficient in postreplicative DNA mismatch repair system (mutL) allowing direct analysis of mutability in terms of chromosomal replication fidelity. In Table 1A we list the mutant frequencies for four lac alleles reverting by G·C→A·T and A·T→G·C transitions, or G·C→T·A and A·T→T·A transversions, in the dnaE⁺ control and three dnaE alleles that were previously isolated as antimutator alleles [14,15]. The dnaE alleles used were dnaE911 and dnaE915, which carry the P357L and A498T missense mutation respectively, as previously reported [16]. Here we also include the dnaE941 allele which was initially isolated as an antimutator in a mutT-deficient background [15]. In the present study, we have sequenced the dnaE gene of this allele and found this to be a Leu to Phe substitution at dnaE residue 611 (L611F).

Table 1.

Comparison of mutant frequencies in strains carrying different dnaE antimutator alleles^a.

A.	Lac+/108	Strand	dnaE+	dnaE911	dnaE915	dnaE941
	G·C→A·T	leading	79.1 (72.1–96.2)	14.0 (13.1–17.3)	7.2 (5.9–8.5)	3.2 (2.8–5.3)
		lagging	24.2 (20.9–33.4)	5.0 (3.9–8.3)	4.5 (4.0–6.4)	3.7 (3.0–5.1)
	A·T→G·C	leading	50.2 (43.1–63.5)	9.1 (7.8–11.4)	7.1 (6.6–8.0)	2.1 (1.4–3.5)
		lagging	18.1 (16.1–23.1)	6.8 (5.8–7.6)	7.3 (6.0–8.9)	2.8 (2.7–4.2)
	G·C→T·A	leading	0.35 (0.33–0.46)	0.17 (0.15–0.25)	0.22 (0.21–0.36)	0.20 (0.16–0.30)
		lagging	0.24 (0.22–0.37)	0.13 (0.12–0.24)	0.33 (0.27–0.44)	0.31 (0.24–0.39)
	A·T→T·A	leading	0.31 (0.26–0.41)	0.10 (0.06–0.14)	0.06 (0.04–0.12)	0.03 (0.01–0.06)
		lagging	0.18 (0.14–0.28)	0.14 (0.10–0.25)	0.09 (0.06–0.10)	0.03 (0.01–0.05)
B.			dnaE+	dnaE911	dnaE915	dnaE941
	RifR/108		443 (368–580)	65 (61–71)	130 (111–164)	44 (36–82)
	NalR/108		116 (103–162)	14 (11–18)	19 (14–25)	6 (5–7)

^aMutant frequencies were calculated as described in Materials and Methods. Data are median frequencies with 95% confidence intervals. All strains were also mismatch-repair defective (mutL::Tn5). Leading/lagging strand assignments are based on the predominant mispairs for each base-pair substitution, as discussed previously [6–8,21].

The results of Table 1A corroborate previous observations that in the dnaE⁺ background there is a significant difference in the fidelity of replication between the two DNA strands, with the lagging strand replicated with higher accuracy than the leading strand [6]. For example, for the G·C→A·T transition, the mutation frequency on the leading strand is 3 -fold higher (79.1 × 10⁻⁸) than on the lagging strand (24.2 × 10⁻⁸). A similar bias is seen for all tested base substitutions. Table 1A further shows that strains carrying the dnaE911, dnaE915, and dnaE941 alleles display significantly reduced levels of mutagenesis, confirming their antimutator status. Strong antimutator effects (3- to 19-fold) are also observed for the case of the rifampicin (Rif^R) and nalidixic-acid resistance (Nal^R) forward targets (Table 1B), indicating that the effects are not in any way lac specific. Importantly, as can be seen from the calculated antimutator folds in Fig 1, the effects are generally larger for the more error-prone leading strand than for the more accurate lagging strand. The average effects are 8.5-fold (1.6–25) for the leading strand and 3.4-fold (0.7–6.5) for the lagging strand, while the average individual fold difference is nearly 2.5. As a result of the preferential antimutator effect for the leading strand, the fidelity difference between the two DNA strands is reduced. For example, for the dnaE941 allele, which is generally the strongest antimutator, the initial 3-fold fidelity difference (dnaE⁺) for the lac G·C→A·T and A·T→G·C transitions is essentially abolished (Table 1A). As noted above, the preferred explanation for these effects involves increased dissociability of the DnaE mutant polymerase (see also Discussion).

Fig 1. Antimutator effects of three different dnaE alleles for leading and lagging strand replication.

Reductions in mutant frequency (antimutator effect) for each of the indicated dnaE alleles for (A) transitions and (B) transversions were calculated by dividing the frequency of Lac⁺ mutants of Table 1 by that of the corresponding dnaE⁺ control. For example, for G·C→A·T transitions, the antimutator effects of dnaE941 were 25-fold for the leading strand (79.1 divided by 3.2) and 6.5-fold for the lagging strand (24.2 divided by 3.7).

3.2. Mutator effects of the dnaE911, dnaE915, and dnaE941 alleles in the presence of induced levels of DNA polymerase V

Next, we investigated the fidelity effect of the dnaE alleles in the presence of certain error-prone DNA polymerases. First, we analyzed the effects of DNA Pol V, a major E. coll translesion (TLS) polymerase, which is normally present at undetectable levels but is expressed at high levels upon induction of the SOS system [22]. To do so, we measured lac mutability in strains carrying, in addition, the recA730 allele (Table 2). This allele leads to constitutive expression of the SOS regulon and de-repression of the umuC and umuD genes that encode Pol V (UmuD’2C) [23]. Importantly, the constitutive presence of Pol V leads to a spontaneous mutator effect (SOS mutator), reflecting not only the error-prone nature of this polymerase but also its ability to gain access to the replication fork [19]. In Table 2 and Fig 2 we present the results obtained with the lac G·C→T·A and A·T→T·A transversions, which were previously shown to be sensitive indicators for Pol V induction [8,19,24,25].

Table 2.

Mutant frequencies (lac⁺ mutants per 10⁸ cells) in recA730 strains carrying different dnaE antimutator alleles.

Strain	Lac+/108	Strand	dnaE+	dnaE911	dnaE915	dnaE941
recA730	G·C →T·A	leading	1.8 (1.7–2.5)	1.8 (1.1–3.8)	5.3 (4.8–9.2)	2.7 (2.2–3.7)
		lagging	4.1 (3.9–6.5)	5.5 (5.0–7.7)	25.2 (20.6–32.4)	8.2 (7.1–13.8)
	A·T→ T·A	leading	0.9 (0.8–1.6)	1.3 (1.0–1.8)	3.3 (3.1–4.9)	1.6 (1.3–4.3)
		lagging	10.4 (8.5–18.5)	33.3 (24.5–38.7)	73.3 (63.6–79.6)	29.3 (26.2–33.0)

Mutant frequencies were calculated as described in Material and Methods. Data are median frequencies with 95% confidence intervals. All strains are mutL::Tn5.

Fig 2. Mutator effect of the recA730 allele for the two DNA strands in various dnaE ‘antimutator’ backgrounds.

Mutator effects were calculated by dividing the indicated lac⁺ mutant frequency as listed in Table 2 for the recA730 strain by that for the recA⁺ control from Table 1A. For example, for G·C→T·A transversion and the dnaE915 allele the recA730 mutator effects were 24-fold on the leading strand (5.3 divided by 0.22) and 76-fold on the lagging strand (25.2 divided by 0.33).

The data confirm that Pol V expression in the dnaE⁺ background results in a clear mutator phenotype (compare Table 1 to Table 2 or see the calculated folds in Fig 2). Specifically, the mutator effect is 3- to 5-fold for the case of the leading strand and 17- to 58-fold for the lagging strand. Thus, the SOS mutator effect occurs preferentially on the lagging strand, and this has been previously reported [8].

Most interestingly, under these conditions the antimutator effect of the dnaE alleles is no longer observed. Instead, we actually see an amplified SOS mutator effect, which occurs consistently across each of the lac and dnaE alleles (Fig 2). As only two examples, for the G·C→T·A allele, the SOS mutator effect in the dnaE⁺ background is 5- and 17-fold for leading and lagging strand, respectively, while for dnaE915 they are 24- and 76-fold (Fig 2). For the case of the A·T→T·A transversion, the SOS mutator effect in the dnaE⁺ strain is 3- and 58-fold for leading and lagging strand, respectively, but this is increased to 55- and 814-fold for dnaE915. We also note that in the dnaE ‘antimutator’ backgrounds the bias for a preferential lagging-strand SOS mutator effect is maintained. The simplest interpretation of these results is that when mutability is primarily governed by replication errors made by DNA Pol III, the dnaE antimutator effects prevail, but when mutability is primarily dictated by interference by PolV into the normal replication machinery, the dnaE antimutator effects are no longer seen. In the Discussion, we will relate these observations to the increased dissociative character of the dnaE ‘antimutator’ alleles.

3.3. Mutator effects of dnaE911, dnaE915, and dnaE941 in the presence of induced levels of DNA polymerase IV

We next tested the dnaE antimutator effects under conditions where DNA Polymerase IV (Pol IV) is engaged (Table 3). Like Pol V, Pol IV is a polymerase that is induced as part of the SOS response. However, in contrast to Pol V, it is present at a significant basal level under normal conditions. Pol IV is an error-prone, proofreading-deficient enzyme that can function in TLS and replication restart under conditions of DNA damage [26–28]. At its normal basal level, it has no significant impact on spontaneous levels of mutation [29,30] however, it was shown to contribute significantly to ‘adaptive’ mutations occurring in stationary phase cells [31]. When expressed and overproduced from a plasmid in growing cells, Pol IV does cause a mutator phenotype, which we have shown to occur preferentially on the lagging strand [9,25].

Table 3.

Mutant frequencies (lac⁺ per 10⁸) for strains overexpressing DNA Pol IV in various dnaE backgrounds.

Lac+/108	Plasmid	Strand	dnaE+	dnaE911	dnaE915	dnaE941
G·C→T·A	pWSK	leading	0.30 (0.12–0.71)	0.10 (0.08–0.32)	0.18 (0.12–0.31)	0.19 (0.06–0.42)
		lagging	0.18 (0.15–0.30)	0.23 (0.17–0.35)	0.33 (0.30–0.62)	0.15 (0.12–0.35)
	pLO1	leading	0.44 (0.36–0.70)	0.67 (0.65–1.58)	1.38 (1.19–2.02)	1.82 (1.25–2.63)
		lagging	0.54 (0.46–0.98)	2.81 (2.27–3.46)	19.6 (16.9–25.4)	3.67 (2.94–4.61)

Mutant frequencies were determined as described in Materials and Methods. Data are median frequencies with 95% confidence intervals. All strains are mutL::Tn5. Plasmid pLO1 is a low-copy plasmid carrying the dinB gene encoding Pol IV, whereas pWSK is the corresponding empty-vector control [9].

In Table 3 we present the results from experiments in which we overexpressed Pol IV from a (low-copy) plasmid.

As previously observed [9], Pol IV expression in the dnaE⁺ background yields a modest (1.5- to 3.0-fold) mutator effect for G·C→T·A transversions (a mutation most readily produced by Pol IV), with the lagging strand being the preferred target (Fig 3). For the dnaE ‘antimutator’ mutants, increased Pol IV production does lead to significant increases of the Pol IV mutator effect for both DNA strands. In all cases, the increased mutator effect is largest for the lagging strand.

Fig 3. Mutator effect induced by Pol IV overexpression for the two DNA strands in different dnaE ‘antimutator’ backgrounds.

The Pol IV mutator effects were calculated by dividing the lac⁺ mutant frequency from Table 3 for the pLO1 (dinB⁺) containing strain by that for the corresponding pWSK control.

3.4. Mutator effects of the dnaE911, dnaE915, and dnaE941 in the presence of proofreading-deficient DNA polymerase II

DNA polymerase II (Pol II), encoded by the polB gene, is an exonuclease-proficient polymerase that has been shown to play a role in the replication of both damaged [32,33] and undamaged DNA [34]. It was also assigned a role in preventing adaptive mutations in stationary-phase cells [35]. In our work [10], we have suggested Pol II to be a back-up fidelity factor for normal chromosomal DNA replication, presumably gaining access to the replication point under conditions when Pol III HE faces some difficulties. This could occur, for example, when continuation of DNA synthesis by HE is impaired or delayed at newly created terminal mispairs (polymerase error). In such cases, removal of the terminal mispair by the Pol II 3’-exonuclease would facilitate the continuation of DNA synthesis, and at the same time make a positive contribution to replication fidelity [10,12]. The same experiments showed Pol II to play an important role in protecting the replication fork against interference by error-prone Pol IV [10,12].

To probe the role of Pol II in the presence of the dnaE ‘antimutator’ alleles, we created corresponding strains in which the chromosomal wild-type copy of the polB gene is replaced by an exonuclease-deficient variant (polBex), as described before [10]. Using this variant, any normal error-free contribution of the enzyme might be captured as a mutagenic effect when polBex extends Pol III misinserted nucleotides. This will lead to an identification of a specific error-prevention role for Pol II [10,34]. The mutant frequencies presented in Table 4 show that the polB to polBex substitution leads to an increase in G·C→T·A mutations for all of the strains used.

Table 4.

Mutant frequencies (lac⁺ mutants per 10⁸ cells) for various dnaE alleles in polBex strains containing a proofreading-deficient Pol II.

Lac+/108	Strain	Strand	dnaE+	dnaE911	dnaE915	dnaE941
G·C→T·A	polB+	leading	0.35 (0.33–0.46)*	0.17 (0.15–0.25)*	0.22 (0.21–0.36)*	0.20 (0.16–0.30)*
		lagging	0.24 (0.22–0.37)*	0.13 (0.12–0.24)*	0.33 (0.27–0.44)*	0.31 (0.24–0.39)*
	polBex	leading	0.86 (0.69–1.17)	0.88 (0.77–1.07)	2.26 (2.12–3.29)	1.66 (1.25–2.49)
		lagging	2.97 (2.55–3.90)	4.30 (4.17–5.69)	7.42 (7.01–10.11)	6.90 (6.08–9.12)

Mutant frequencies were calculated as described in Material and Methods. Data are median frequencies with 95% confidence intervals. All strains are mutL::Tn5.

^*Data from Table 1, for convenience.

The fold-increase in mutant frequency is plotted in Fig 4. The results for the dnaE⁺ strain are consistent with our previous report [10], in which we reported that both DNA strands are subject to increases in mutability, but the lagging strand is subject to the largest effect. In strains carrying the dnaE ‘antimutator’ alleles, as in the case of Pol V and Pol IV overproduction, there is disappearance of any antimutator effects. Instead, the polBex mutator as seen in the dnaE⁺ control strain is extended and amplified, in dnaE antimutator backgrounds on both strands (Fig 4). Like in the control, the mutator effects are largest for the lagging strand.

Fig 4. Mutator effect of the polBex allele in different dnaE ‘antimutator’ backgrounds.

The polBex mutator effects were calculated by dividing the frequency of indicated lac reversions in the polBex strain by frequency in the corresponding polB⁺ control (Table 4).

These results, like those for the Pol IV and Pol V overexpression experiments, reveal the disappearance of the dnaE antimutator effects when additional pathways become predominant contributor to overall mutation production. Moreover, the dnaE alleles significantly increase mutability under these conditions compared to the dnaE⁺ control. This amplified mutagenic effect must be related to the altered enzymology of the Pol III HE.

4. Discussion

The present work was undertaken to better understand the factors that determine the fidelity of chromosomal DNA replication, in particular with regard to the question of what determines the differential fidelity of leading and lagging DNA strands. In E. coli, the simultaneous replication of the two DNA strands is performed by Pol III HE, which is composed of two joined DNA Pol III cores subassemblies (αεθ), one for each strand [36]. Alternative versions of HE may include a third Pol III core [37,38], but this has not been fully established [39]. Our previous studies showed that replication of the lagging strand was several-fold more accurate than of the leading strand [6–8,10,11]. We have previously proposed that this difference results from the greater dissociative tendency (dissociability) of the lagging-strand polymerase [6], which dissociates rapidly and efficiently from its primer terminus each time when reaching the end of an Okazaki fragment. The signal for release of the lagging-strand polymerase was shown to be not simply the enzyme reaching the end of the fragment, but rather the time-dependent availability of the next Okazaki fragment primer [40]. Indeed, early dissociation prior to completion of the fragment was shown to occur when the filling of the actual gap was delayed or slowed [40,41]. We have proposed that terminal base-base mismatches resulting from polymerase misinsertion errors are one such factor that can slow down the polymerase, as observed in multiple in vitro studies [42–45]. Therefore, premature polymerase dissociation at terminal mismatches should be considered a frequent event, at least in the lagging strand. Slowdowns at terminal mismatches will of course also occur in the leading strand, but the higher processivity of this polymerase will reduce the likelihood of such events. Finally, any dissociation from a terminal mismatch will expose the mismatch to multiple error-removal mechanisms (see below) and essentially prevent its chances of ever being converted into a mutation. Hence, dissociation is an important fidelity mechanism that will naturally favor the lagging strand.

Further evidence for polymerase dissociation as a fidelity mechanism came from the antimutator effect (reduced mutability) associated with the dnaE915 allele. This allele encodes a catalytic polymerase alpha subunit of Pol III (A498T) that was shown to have increased dissociative behavior (i.e., reduced processivity) [7]. Interestingly, it was also observed that this antimutator effect was unequal for the two strands; specifically, the normally more error-prone leading strand was reduced the most. This interesting finding may provide new insight into the mechanisms underlying the strand-specific fidelity differences. Here, we have further investigated this issue by extending the analysis to include two additional dnaE alleles, dnaE911 (P357L) and dnaE941 (L611F), isolated as antimutators for normal DNA replication [14,15], and by using four distinct alleles of lacZ reversion system. We found that in each of the cases the dnaE alleles provided clear antimutator effects, which again showed a preferential reduction of mutagenesis for the normally more error-prone leading strand (Figures 1A–B). Overall, these findings provide clear support for a dissociative fidelity mechanism operating in E. coli, although the reason why the leading strand is affected preferentially remains to be established (see also below).

Regarding the precise mechanisms underlying the dnaE antimutator effects, we have argued that these are not likely due to improved base selection of the polymerase. Instead, they are a consequence of more efficient error processing (removal) after the mismatch has been created [7,16]. In this view, the dnaE antimutator polymerases are, in fact, slightly impaired enzymes that suffer from reduced catalytic efficiency, reduced processivity, or both [7,12,18]. Due to these defects, the likelihood of extension of a newly created mispair (needed to convert the error into a potential mutation) is reduced, hence the antimutator effect. Instead of extension, the error has an increased probability of being removed by the proofreading exonuclease or abandoned by polymerase dissociation. In the case of a dissociation, the abandoned mismatch is unlikely to survive and will be subject to elimination by 3’-exonucleolytic removal (see below) before continuation of DNA synthesis. Importantly, previous results have shown that the dnaE antimutator effects are readily observed in proofreading-impaired backgrounds (dnaQ49, mutD5, dnaQ926) [18,46], indicating that not only proofreading, but also enhanced dissociability of the polymerase is to be considered an important fidelity factor.

While we use the broad term dissociation throughout, several different dissociation scenarios can be imagined that may lead to different fidelity outcomes for the two DNA strands. One is a full dissociation of the polymerase half, similar to what occurs in the lagging strand at the end of each Okazaki fragment, in which the polymerase simply continues further downstream at a newly laid-down primer, leaving behind a gap. The terminal mismatch left behind in the gap would be a ready target for a number of nucleases, most obviously any DNA polymerases containing a 3’-exonuclease. Such a type of dissociation can certainly also happen to the leading-strand polymerase, although this should presumably occur at reduced frequency due to its higher processivity. However, in the dnaE antimutator strains, this processivity restriction may be significantly alleviated, plausibly explaining the larger antimutator effect for this strand. This type of dissociation in the leading strand would also require additional priming of this strand, but several types of studies have revealed this not likely to be a major problem [47,48].

Second, polymerase dissociation could be followed by a rapid re-association of the polymerase with the mismatched primer terminus. In such new binding cases, DNA polymerases preferentially bind the 3’ terminus with their exonuclease domain [45,49,50], leading to removal of the error via an indirect proofreading mode. As DNA Pol III HE in certain studies has been described as a triple polymerase complex [37,38], this type of indirect proofreading could also readily occur via engagement of the third Pol III core present in the complex.

Third, polymerase dissociation could be coordinated with an exchange with DNA Pol II which has previously been demonstrated to act as a back-up proofreader for Pol III [10]

Finally, recent single-molecule studies on the dynamics of Pol III HE have shown that Pol III HE (or at least Pol III* - Pol III HE without β₂-clamp) exchanges much more frequently with Pol III* complexes from solution than previously assumed, hence indicating that even leading-strand synthesis is less processive than previously believed [51,52]. Note, that our model for differential leading/lagging strand fidelities does not require full processivity of the leading strand. Any significant processivity difference between the two DNA strands would suffice. If indeed DNA polymerase complexes are prone to frequent polymerase exchanges with Pol III molecules from the cellular environment, this type of exchange could also be regarded as a dissociative fidelity mechanism, that can be logically enhanced by the dnaE antimutator alleles, that might in fact be of greater significance to the more processive leading strand, as observed here.

In broad support of our dissociation model, we further demonstrate here that the dnaE antimutator alleles are increasingly sensitive to the presence of accessory DNA polymerases (Figures 2–4). Specifically, in the case of proofreading-deficient accessory polymerases (Pol V, Pol IV, PolBex) we did observe mutator effects for each of the ‘antimutator’ dnaE alleles, for both the leading and lagging DNA strand. Obviously, increased dissociation of Pol III in the dnaE strains will provide increased access to the error-prone enzymes. Their extension of the available mispairs and participation in synthesis will lead to the observed mutator phenotype. Importantly, while Pol V and Pol IV may not be present at sufficient levels under normal conditions to affect overall fidelity, Pol II is certainly present in its exonuclease-proficient form under normal conditions, and the mutator effect seen with its exonuclease-deficient form (PolBex) (Fig 4) may be envisioned to be the mirror image of an antimutator effect in the PolB⁺ configuration. It thus provides further evidence for the proposed back-up proofreader role of Pol II for DNA replication errors via one of the dissociation models outlined above [10].

The dnaE-mediated mutator effects shown in Figures 2–4 are not only larger than for the corresponding dnaE⁺ strain in terms of the fold-increase (i.e., compared to its own background in the absence of the accessory polymerase) but are, importantly, also significantly increased in terms of the absolute numbers of observed mutations. These data can be derived directly from the frequencies shown in the Tables 2–4. For example, for the case of PolV expression (recA730 background) in the dnaE⁺ strain, the mutant frequencies (× 10⁻⁸) for the lac G·C→T·A allele are 1.8 and 4.1 for the leading and lagging strand, respectively (Table 2), but for the dnaE915 strain they are significantly higher, 5.3 and 25.2, respectively. For the lac A·T→T·A allele in the same strains the frequencies are 0.9 and 10.4 for dnaE⁺, but 3.3 and 73.3, respectively, for dnaE915. Similar numbers can be derived for most other combinations. This provides further evidence for enhanced dissociation in the dnaE backgrounds.

Finally, we would like to note that the dnaE antimutators may be useful tools for analyzing important other aspects of chromosomal replication, such as the precise mechanisms mediating DNA polymerase exchanges at the replication fork, an area of significant current interest [51–55]. We note that there are subtle differences between the antimutator and mutator effects when comparing the various dnaE alleles. For example, dnaE941 appears to be generally the most efficient regular antimutator (Fig 1), but dnaE915 is particularly sensitive to Pol V and Pol IV overproduction and produces the largest mutator effects (Figures 2 and 3). Overall, these data support the possibility that the interactions of DNA Pol II, Pol IV and Pol V with the replisome may differ in important aspects, and the dnaE alleles might be helpful in further investigating these interactions.

Highlights.

• Pol III antimutator variants have higher fidelity than the wild-type enzyme

• antimutators increase the replication fidelity preferentially on the leading strand

• dnaE antimutators promote fidelity by increasing “dissociability”

• when antimutators are faced with error-prone polymerases they turn into mutators

• increased ‘dissociability’ of the Pol III serves as an important fidelity factor

Abbreviations

Pol

DNA polymerase

DNA Pol III HE

DNA Pol III holoenzyme