Role of Escherichia coli DNA Polymerase I in chromosomal DNA replication fidelity

Summary

We have investigated the possible role of E. coli DNA polymerase I in chromosomal replication fidelity. This was done by substituting the chromosomal polA gene by the polAexo variant containing an inactivated 3'→5' exonuclease, which serves as a proofreader for this enzyme's misinsertion errors. Using this strain, activities of Pol I during DNA replication might be detectable as increases in the bacterial mutation rate. Using a series of defined lacZ reversion alleles in two orientations on the chromosome as markers for mutagenesis, 1.5- to 4-fold increases in mutant frequencies were observed. In general, these increases were largest for lac orientations favoring events during lagging strand DNA replication. Further analysis of these effects in strains affected in other E. coli DNA replication functions indicated that this polAexo mutator effect is best explained by an effect that is additive compared to other error-producing events at the replication fork. No evidence was found that Pol I participates in the polymerase switching between Pol II, III and IV at the fork. Instead, our data suggest that the additional errors produced by polAexo are created during the maturation of Okazaki fragments in the lagging strand.

Introduction

Chromosomal DNA replication is a highly accurate process. The fidelity of this process relies on faithful DNA synthesis by the replicative DNA polymerases. Escherichia coli possesses 5 distinct DNA polymerases. Replication of the chromosome is performed primarily by DNA polymerase III holoenzyme (HE), which is an asymmetric, dimeric, 17-subunit complex that is capable of coordinated high-speed and high-fidelity synthesis of leading and lagging strand at the replication fork (McHenry, 1991; Kelman & O'Donnell, 1995; McHenry, 2003; O'Donnell, 2006; Pomerantz and O'Donnell, 2007). The α, ε and θ subunits of HE constitute the polymerase core. The α subunit, responsible for polymerase activity, is encoded by the dnaE gene (Gefter et al., 1971). The ε subunit, encoded by the dnaQ gene (Takano et al., 1986), contains the 3' → 5' exonuclease that proofreads mismatches made by α subunit. The small (8 kDa) θ-subunit primarily serves as a stabilizer for ε subunit (Taft-Benz and Schaaper, 2004). Additionally, HE contains a pair of sliding-clamp processivity factors (β₂), one for each core, as well as the seven-subunit DnaX complex (τ₂γδδ'χψ) (O'Donnell et al., 2001; McHenry, 2003) responsible for loading and unloading the processivity clamp.

While the role of Pol III in chromosomal replication is clearly established, the precise role(s) of the additional DNA polymerases are still being investigated, including a possible role in chromosomal replication and its fidelity. DNA polymerase II (Pol II) encoded by the polB (dinA) gene belongs to the B-family of polymerases. Like eukaryotic replicative B family polymerases, Pol II is a high fidelity enzyme possessing 3' → 5' exonuclease activity (Cai et al., 1995). The constitutive intracellular concentration of Pol II (30–50 molecules/cell) is comparable to the estimated concentration of Pol III HE (30 molecules/cell) (McHenry and Kornberg, 1977; Qiu and Goodman, 1997). However, upon induction of the SOS response, Pol II levels can be further increased sevenfold (Qiu et al., 1997; Bonner et al., 1988, 1990). Our group (Banach-Orlowska et al., 2005) and others (Bonner et al., 1988; Escarceller et al., 1994; Foster et al., 1995; Rangarajan et al., 1997, 1999; Foster, 2000; Napolitano et al., 2000) have previously demonstrated that Pol II can perform certain functions at the replication fork. One such proposed function is to substitute for Pol III when HE has difficulties extending a mismatched primer terminus (Banach-Orlowska et al., 2005), thus serving as a proofreader for Pol III, as also suggested by Curti et al. (2008).

Two other polymerases, Pol IV and Pol V, are members of the Y family of DNA polymerases; they lack a proofreading activity and are generally considered low-fidelity DNA polymerases (Goodman, 2002). Both are expressed at elevated levels during SOS induction. Pol V (umuDC gene product) (Reuven et al., 1999; Tang et al., 1999) is the major polymerase involved in damage-induced mutagenesis. It performs translesion synthesis, creating mutations targeted at DNA damage sites, but also produces untargeted mutations at non-damaged DNA sites (Ichikawa-Ryo and Kondo, 1975; Fijalkowska et al., 1997). Importantly, Pol V is not significantly expressed in non-induced cells.

Pol IV is encoded by the dinB gene (Wagner et al., 1999). The basal level of Pol IV in normal, uninduced cells is relatively high (250 molecules per cell) (Kim et al., 2001). Pol IV can also participate in translesion synthesis (TLS), alone or in combination with Pol V, depending upon the nature of the lesion and its local sequence context (Napolitano et al., 2000; Wagner et al., 2002). However, despite its relative high basal level, Pol IV does not contribute significantly to the chromosomal mutation rate in growing cells (Kuban et al., 2004; Wolff et al., 2004).

DNA Polymerase I, encoded by the polA gene, the first DNA polymerase discovered, is the most abundant polymerase in E. coli (approximately 400 molecules per cell) (Kornberg and Baker, 1992). The Pol I polypeptide has two functional domains: a large domain (Klenow fragment) that contains the 5'→3' polymerase and 3'→5' proofreading exonuclease, and a small fragment that contains a 5'→3' exonuclease activity (Joyce and Grindley, 1984). The latter activity enables Pol I to remove RNA primers that were used to initiate the downstream Okazaki fragment (Okazaki et al., 1971). Other well-known functions of Pol I are to participate in DNA repair and recombination. During DNA repair, Pol I fills in DNA gaps that result from the removal of DNA lesions (for a review, see Friedberg et al., 1995, 2006). ΔpolA strains are inviable when grown in rich medium, but their viability can be restored by providing the 5'→3' exonuclease function in trans (Joyce and Grindley, 1984).

The mechanisms regulating the involvement of multiple DNA polymerases in DNA replication are a subject of active interest (Shcherbakova and Fijalkowska, 2006; Fuchs and Fujii, 2007; Jarosz et al. 2007; Lehmann et al., 2007; Curti et al., 2008). Evidence has accumulated indicating that DNA polymerases can collaborate and substitute for each other not only during translesion synthesis, but also during ongoing DNA synthesis. For example, Pol II was proposed to be able to act as proofreader for HE-produced misinsertion errors (Banach-Orlowska et al., 2005). These experiments made use of polBex1 mutants defective in the Pol II 3' proofreading activity, which were demonstrated to display a modest mutator effect. This mutator effect increased synergistically when combined with certain dnaE or dnaQ mutator alleles (Banach-Orlowska et al., 2005), suggesting that the level of Pol III-made errors is a determining factor for the mutagenic action of the proofreading-deficient Pol II. It was postulated that Pol II occasionally has access to mismatched 3' termini created by Pol III; these mismatches will normally be removed by Pol II by means of its proofreading activity but will be extended into replication errors by the proofreading-deficient version of the enzyme. The same model readily allows for the second postulated role of Pol II, namely protection of mismatched 3' termini against the mutagenic action of Pol IV. The latter (proofreading-deficient) polymerase is capable of producing mutations as evidenced by the mutator activity observed when the enzyme is overproduced (Kim et al., 1997; Wagner and Nohmi, 2000; Kuban et al., 2005). Likewise, this Pol IV mutator phenotype is proposed to result from the extension of mismatched primer termini created by Pol III HE (Kuban et al., 2005).

The data discussed above suggest that there is competition between Pol II, Pol III, and Pol IV [and in SOS-induced cells, also Pol V (Fijalkowska et al., 1989)] for a terminal mismatch that may become accessible to other polymerases upon dissociation of Pol III (Mo and Schaaper, 1996; Pham et al. 1998, 1999). In the present study, we have addressed the possible role of Pol I in this competition and, hence, in determining the chromosomal error rate. As Pol I is an abundant polymerase, contains a 3' proofreading activity and, like other polymerases, is capable of interacting with the β processivity clamp (Bonner et al., 1992; López de Saro and O'Donnell, 2001; Pham et al., 2001; Wagner et al., 2001; López de Saro et al., 2003; Sutton and Duzen, 2006), a role for this enzyme may be envisaged. To investigate this issue, we created a proofreading-deficient variant of Pol I (polAexo mutant) (Bebenek et al., 1990) and investigated whether this deficiency leads to increased mutant frequencies in a variety of genetic backgrounds. Based on the results we conclude that Pol I does affect the chromosomal mutation rate. However, this effect is not achieved by directly competing with Pol II, Pol III or Pol IV. Instead, we postulate that the role of Pol I at the replication fork is primarily the faithful filling of Okazaki fragment gaps associated with removal of RNA primers in the lagging strand.

Results

Experimental system for studying the fidelity role of Pol I

To study the possible role of Pol I in replication fidelity, we constructed strains in which the chromosomal polA gene was replaced by an error-prone form, polA-D424A (polAexo) (see Experimental Procedures for the construction). In this mutant, residue Asp424 in the Exo I motif of the 3' exonuclease domain is replaced by Ala. Biochemical experiments on this mutant polymerase have shown the enzyme to be specifically defective in the 3'→5' exonucleolytic proofreading activity, while retaining full DNA synthesis activity (Bebenek et al., 1990). In vivo, this substitution is expected to increase the error rate of Pol I during DNA synthesis in the cell. The underlying assumption for our experiments was that in a background of high-fidelity DNA synthesis even a small amount of error-prone DNA synthesis by the exonuclease-defective variant might be detectable as an increase in the bacterial mutation frequency.

In order to measure mutagenesis, we used a previously described lacZ reversion system employing pairs of strains in which a lacZ mutational target resides on the chromosome in the two possible orientations relative to the direction of replication (Fijalkowska et al., 1998). This approach permits analysis of the results in terms of potentially differential effects during leading and lagging strand replication, as discussed (Fijalkowska et al., 1998; Maliszewska et al., 2000; Gawel et al., 2002a b; Banach-Orlowska et al., 2005; Kuban et al., 2006; Gawel et al., 2008). In many of these experiments, the strains used are also mismatch-repair deficient (mutL) to more directly measure the effects on uncorrected replication errors.

In Table 1 we present the results of several experiments aimed at showing the effect of the polAexo allele on mutant frequencies for a set of four lac alleles, as well as on the frequency of rifampicin or nalidixic-acid resistant mutants. The lac alleles used permit the specific scoring of G·C→A·T, G·C→T·A, A·T→T·A, or A·T→G·C base substitutions, respectively (Cupples and Miller, 1989). For each allele, the mutant frequency was measured for the two chromosomal orientations (R and L) of the lac target (Fijalkowska et al., 1998). As before (Fijalkowska et al., 1998), significant differences are observed in the mutant frequency between the two orientations, particularly in the mismatch-repair-defective background (Table 1B), which have been interpreted to reflect the differential fidelity of leading and lagging strand replication (Fijalkowska et al., 1998). To illustrate one example, for the lac G·C→A·T allele, in section B, there is a 3-fold difference between the two orientations: 150 × 10⁻⁸ (L-orientation) versus 43 × 10⁻⁸ (R-orientation) (Table 1B). As G·C→A·T mutations are assumed to be mediated primarily by G·T mispairings (rather than C·A), the location of the template G is in the leading strand for the L-orientation and in the lagging strand for the R-orientation (Fijalkowska et al., 1998), thus leading to the conclusion that for this lac allele lagging-strand replication is several fold more accurate than leading-strand replication. Using similar reasoning for the three other lac alleles, the primary strand whose fidelity is measured for each lac orientation can be deduced (see Fijalkowska et al., 1998 and the Legend to Table 1 for more details). Based on these data, reproduced in multiple studies from our laboratory, we have suggested that on the E. coli chromosome, lagging strand replication is more accurate. Thus with regard to the present issue of the role of Pol I, the use of this lac system provides additionally an opportunity to investigate any strand-specific effects of Pol I.

Table 1.

Mutator effects of polAexo and polBex alleles in various strain backgrounds.^a

	lac orientation (strand)c	lac → lac+ (per 108)
lac alleleb		Pol+	polAexo	polBex	polAexo polBex
A. mutL+
G·C → A·T	R(lagging)	2.5 ±0.7	2.3 ±0.9 [0.9]	ND	ND
	L(leading)	2.1 ±0.8	1.8 ±0.8 [0.8]	ND	ND
G·C → T·A	R(lagging)	0.8 ±0.7	3.3 ±1.0 [4.1]	ND	ND
	L(leading)	1.8 ±1.0	1.8 ±0.9 [1.0]	ND	ND
A·T → T·A	R(lagging)	0.9 ±0.5	1.5 ±0.5 [1.6]	ND	ND
	L(leading)	0.6 ±0.5	1.6 ±0.5 [2.7]	ND	ND
A·T → G·C	L(lagging)	0.1 ±0.1	0.1 ±0.1 [1.0]	ND	ND
	R(leading)	0.1 ±0.2	0.1 ±0.1 [1.0]	ND	ND
RifR		1.2 ±0.9	1.8 ±1.3 [1.5]	ND	ND
NalR		0.2 ±0.2	0.2 ±0.3 [1.0]	ND	ND
B. mutL
G·C→A·T	R(lagging)	43 ±11	69 ±19 [1.6]	50 ±24 [1.2]	79 ±26 [1.8]
	L(leading)	150 ±35	167 ±26 [1.1]	200 ±41 [1.3]	163 ±38 [1.1]
G·C→T·A	R(lagging)	1.9 ±0.8	3.9 ±1.6 [2.1]	4.5 ±3.2 [2.4]	6.1 ±1.6 [3.2]
	L(leading)	3.8 ±0.7	2.6 ±0.8 [0.7]	3.6 ±1.4 [0.9]	3.3 ±1.6 [0.9]
A·T→T·A	R(lagging)	1.0 ±0.7	1.7 ±1.0 [1.7]	1.6 ±0.9 [1.6]	ND
	L(leading)	3.0 ±1.3	5.5 ±4.0 [1.8]	2.4 ±1.2 [0.8]	ND
A·T→G·C	L(lagging)	17 ±3	23 ±5 [1.4]	26 ±12 [1.6]	ND
	R(leading)	48 ±10	48 ±9 [1.0]	36 ±22 [0.8]	ND
RifR/108		277 ±115	244 ±166 [0.9]	311 ±170 [1.1]	261 ±74 [0.9]
C. mutL dnaE915
G·C→A·T	R(lagging)	9.0 ±3.6	34 ±31 [3.8]	63 ±32 [6.4]	86 ±26 [8.7]
	L(leading)	30 ±4	29 ±3.0 [1.0]	67 ±13 [2.2]	51 ±12 [1.7]
G·C→T·A	R(lagging)	1.6 ±0.6	4.1 ±0.6 [2.6]	6.0 ±2.0 [3.8]	8.8 ±2.8 [5.5]
	L(leading)	0.7 ±0.3	0.7 ±0.3 [1.0]	2.5 ±0.9 [3.6]	2.9 ±0.8 [4.1]
A·T→T·A	R(lagging)	0.5 ±0.6	1.7 ±3.4 [3.4]	3.8 ±1.9 [7.6]	ND
	L(leading)	0.3 ±0.4	9.5 ±4.4 [32]	2.2 ±0.7 [7.3]	ND
A·T→G·C	L(lagging)	7.0 ±1.1	18 ±2.4 [2.6]	35 ±7 [5.0]	ND
	R(leading)	5.5 ±2.0	8.2 ±3.1 [1.5]	12 ±7 [2.2]	ND
RifR/108		62 ±29	87 ±35 [1.4]	250 ±190 [4.0]	290 ±130 [4.7]

^aThe three strain backgrounds are: (A) mismatch-repair proficient (mutL⁺), (B) mismatch-repair deficient (mutL::Tn5), and (C) mismatch-repair deficient (mutL::Tn5) plus dnaE915 (antimutator allele, see text). Mutant frequencies are based on the median value for 10–20 independent cultures (see Experimental procedures). In [brackets] the polAexo or polBex mutator effect defined as the increase in mutant frequency relative the pol⁺ strain.

Bold numbers indicate statistically significant values (P < 0.05, see Experimental procedures).

ND – not determined.

^blac indicates the lac allele which reversion is being tested (see text).

^cThe L and R orientations of the lac operon are as defined in Fijalkowska et al. (1998). The DNA strand in parentheses denotes the strand in which the presumed dominant mispair occurs for each of the four base substitutions (G·T for G·C→A·T, C·T for G·C→T·A, T·T for A·T→T·A, and T·G for A·T→G·C) (template base indicated first). See text and Fijalkowska et al. (1998) for a discussion.

Expected signature(s) for Pol I-mediated errors

Previous studies have investigated the in vitro fidelity of the D424A exonuclease-deficient form of Pol I Klenow fragment, assaying overall error rates or the rates of misinsertion and mismatch extension in various sequence contexts (Bebenek et al., 1990; Joyce et al., 1992). In general, the observations conform to the pattern observed with many other DNA polymerases (Fijalkowska et al., 1998). Among transition mismatches, T·G and G·T are more frequent than the complementary A·C and C·A at both the misinsertion and extension step. For transversion mismatches the data are more varied, but generally adhere to the observations for other DNA polymerases, but with one exception. It was found that Pol I generates A·A mismatches at relatively high frequency, similar to or higher than the reciprocal T·T frequency, which has been assumed to be preferred by Pol III (Fijalkowska et al., 1998). Thus, for the lacZ allele that reverts through A·T→T·A transversion, any synthesis by Pol I (D424A) might lead to an apparent strandedness that is the opposite of what is routinely observed in our assay (see below).

polAexo mutator effects

The experiments of Table 1 in both mismatch-repair proficient (A) and mismatch-repair-deficient (mutL) strains (B) show that the presence of the polAexo allele indeed increases several of the mutant frequencies, although by modest extents. In the mutL⁺ background, the polAexo allele increases the mutant frequency for the G·C→T·A allele 4.1-fold specifically for the R(lagging) orientation. For the A·T→T·A allele the frequency was enhanced 1.6- to 2.7-fold for both orientations. We also measured the frequency of mutants resistant to the antibiotics rifampicin and nalidixic acid. The Rif^R frequency was increased 1.5-fold, while no effect was detected for the Nal^R mutants. In the mutL background (B) increases were observed of similar nature and extent. For the lac G·C→T·A allele the increase is again specific for the R(lagging) orientation, while for the lac A·T→T·A allele both orientations are affected. In this case we also observe a small, but statistically significant increase for the A·T→G·C transition for the L(lagging) orientation. Interestingly, no increase could be detected for the Rif^R (or Nal^R) mutations (Table 1B). Table 1 contains representative experiments, but the modest but significant effects of polAexo were observed in many repeated experiments.

polAexo effects in the dnaE915 antimutator background

In previous studies investigating the mutator effects of an exonuclease-deficient Pol II allele (polBex1), increased effects of the Pol II proofreading defect were noted when the experiment was performed in strains containing the dnaE915 allele (Rangarajan et al., 1997; Banach-Orlowska et al., 2005). This allele represents an antimutator variant of DNA Polymerase III (Fijalkowska et al., 1993), although its precise mechanism is not yet established. It was argued that the reduction in background mutation frequency in the antimutator strain made it easier to observe the mutator effect resulting from the Pol II proofreading defect (Rangarajan et al., 1997; Banach-Orlowska et al., 2005). Likewise, the mutator effect of the polAexo variant might be more clearly noted in the dnaE915 background, and we have pursued this possibility.

The results of these experiments (Table 1C) confirm the previous observations (Banach-Orlowska et al., 2005) that dnaE915 significantly lowers the mutant frequency for each of the lac alleles, particularly for the two transitions (G·C→A·T and A·T→G·C). Further, the antimutator effect is clearly dissimilar for the two strands, the strongest reductions occurring for the presumed leading strand (Banach-Orlowska et al., 2005). Although not the direct subject of the present study, it is likely that this strand asymmetry of the dnaE915 antimutator will be helpful in better understanding the differential fidelity mechanisms for leading and lagging strand replication in E. coli. The results show that, indeed, in the mutL dnaE915 background several of the mutator effects of polAexo are increased (Table 1C). For example, a 3.8-fold increase in the frequency of G·C→A·T transitions is observed for the R(lagging) orientation (Table 1C) (compared to 1.6-fold in mutL) (Table 1B). For the G·C→T·A transversion, a 2.6-increase for the R(lagging) orientation of the G·C→T·A transversion is observed. For the lac AT→TA allele both lac orientations are enhanced (3.4- and 32-fold), while the two orientations of the A·T→G·C allele are enhanced by 2.6- and 1.5-fold for L (lagging) and R (leading), respectively.

The combined results for both mutL and mutL dnaE915 indicate not only a significant mutator effect of the polAexo allele, but also that the effects are unequal for the two DNA strands. Overall, the effect seems to be largely specific for the lagging strand. The one exception may be the case of the A·T→T·A transversions, where both orientations are enhanced and, in fact, the L(leading) orientation is most strongly enhanced. However, this assignment of "leading" is based on assumed T·T mispairing being responsible for the majority of A·T→T·A transversions. As indicated above, Pol I has been shown to have a strong tendency to produce A·A mispairings (Joyce et al., 1992). Hence, the strong enhancement of mutations in the L orientation by polAexo is most likely fully consistent with a preferred action of Pol I in the lagging strand.

Comparison of polAexo with polBex

The experiments of Table 1 also include for comparison the results with the polBex mutant and the polAexo polBex double mutant. The mutator effects of polAexo and polBex are generally comparable in magnitude, although some differences are seen. In the mutL dnaE915 background, most of the effects of polBex are slightly larger, and include a clear (4-fold) effect on the frequency of Rif^r mutations. PolBex effects are also observed for both orientations of the lac alleles, consistent with our proposal that Pol II can operate in both strands (Banach-Orlowska et al., 2005).

An interesting picture emerges when looking at the results for the polAexo polBex double mutator strains (Table 1, last column). It appears that the two mutator effects operate independently and that, quantitatively, the joint mutator effect might be most easily described by simple additivity of the two separate mutant frequencies. This suggests that the pathways by which Pol I and Pol II contribute to replication fidelity are separate and do not involve a common intermediate and, hence, reflect a lack of direct competition between the two polymerases. The next series of experiments was conducted to more precisely delineate the possible role of Pol I.

Pol I does not process mismatches created by Pol III HE

One model that we have advanced for the involvement of accessory polymerases in the DNA replication process, especially its fidelity, is a switching of polymerases when Pol III HE makes a misinsertion error (Fijalkowska et al., 1987; Maliszewska-Tkaczyk et al., 2000; Banach-Orlowska et al., 2005; Kuban et al., 2005; Gawel et al., 2008). As mismatches are generally difficult to extend, a temporary stalling may occur which may provide an opportunity for additional polymerases to gain access to the DNA growing point. This interference may be mutagenic if access is gained by an exonuclease deficient enzyme or antimutagenic if access is gained by a proofreading-proficient enzyme. This mechanism has been used to explain the mutator effect of Pol IV overexpression, as well as the back-up proofreading role that Pol II may play for Pol III (Banach-Orlowska et al., 2005; Kuban et al., 2005; Gawel et al., 2008). Experiments in support of this mechanism have generally employed the use of additional deficiencies in Pol III (DnaE) and its accessory subunits, such as the Pol III proofreading subunit (DnaQ), the Pol III τ subunit (DnaX), and the Pol III β subunit (DnaN). In the present study, we conducted similar experiments with the polAexo allele.

In Table 2, we analyze the effect of polAexo in a series of dnaQ, dnaE, dnaN, and dnaX mutants. Some of these were investigated in the mutL background, while others, due to strain construction constraints, where examined in the mutL⁺ background. The studies were also conducted specifically with the lac G·C→T·A allele in the lagging orientation, as this configuration generally displays one of the stronger effects for the polAexo allele. Two dnaQ mutator mutants were used (dnaQ928 and dnaQ49), which carry a defect in the Pol III epsilon subunit (Takano et al., 1986; Taft-Benz and Schaaper, 1998). Two dnaE alleles were used (dnaE486 and dnaE511), carrying defects in the Pol III α subunit resulting in a mutator phenotype (Vandewiele et al., 2002). The dnaX36 mutant carries a defect in the Pol III τ subunit that leads to a very specific mutator effect (enhancement of transversions and −1 frameshift mutations) (Pham et al., 2006; Gawel et al., 2008). Finally, the dnaN159 allele represents a defect in the β subunit of HE, which also specifies a mutator phenotype that has been ascribed to an altered polymerase usage (Sutton, 2004; Sutton and Duzen, 2006). In parallel, we compared the results of the polAexo mutation with those of the polBex1 mutation. The general logic of these experiments is that if Pol I were to extend mismatches created by the mutator HEs, a synergistic (ie, more than additive or even multiplicative) effect of the dna and polAex mutator effects is predicted. On the other hand, if the dna and polAex alleles were to act entirely separate, then the combined mutator effect would likely be simply the sum of the individual effects.

Table 2.

Effect of polAexo and polBex on lac G·C→T·A mutant frequency in strains with impaired Pol III HE.^a

	lac → lac+ (per 108)
Strain	polA+	polAexo	polBex
mutL+	0.4 ± 0.2	1.3 ± 0.3	1.6 ± 1.0
mutL+ dnaQ928	9.2 ± 2.2	11.6 ± 1.4	16 ± 7.0
mutL+ dnaE486	2.8 ± 1.4	4.0 ± 1.6	251 ± 23
mutL+ dnaE511	2.0 ± 1.4	3.9 ± 1.8	163 ± 35
mutL	1.4 ± 0.7	4.6 ± 2.3	8.4 ± 11
mutL dnaQ49	44 ± 23	42 ± 12	120 ± 11
mutL dnaN159	5.6 ± 1.7	9.8 ± 3.2	34 ± 7.0
mutL dnaX36	56 ± 12	81 ± 16	ND

^aThe experiments were performed with the lac G·C→T·A allele in the R(lagging) orientation. Mismatch repair deficient strains were mutL::Tn5. Mutant frequencies were determined as described in Experimental Procedures. Each entry is based on the median value for 10–20 independent cultures grown at 30°C (dnaQ49, dnaN159), 34°C (dnaQ928, dnaE486, dnaE511), or 37°C (dnaX36).

ND – not determined.

Table 2 shows that the interaction of polAexo with each of these replication/fidelity defects is very slight or even absent; all polAex-mediated enhancements of the dna mutator effects are less than 2-fold. Instead, it appears that simple additivity of the various mutator effects is the most straightforward interpretation. For example, for the mutL dnaN159 strain, the enhancement of the mutant frequency by the polAexo allele is 4.2 × 10⁻⁸ (9.8 × 10⁻⁸ minus 5.6 × 10⁻⁸), which corresponds reasonably to the polAex effect in the mutL dnaN⁺ strain (4.6 × 10⁻⁸ minus 1.4 × 10⁻⁸ = 3.2 × 10⁻⁸). Overall, these data suggest that role of Pol I is not related to any of the error-producing pathways studied here. This finding contrasts to very substantial interactions displayed by the polBex1 allele seen in the same experiment (see also Banach-Orlowska et al., 2005). We conclude that the mode of mutation production by polAexo, in contrast of that of polBex, is unlikely related to the extension by Pol I of Pol III-mediated misinsertion errors.

Competition between accessory polymerases

In Table 3 we present results of one additional experiment in which we analyzed the effects of the polAexo deficiency in strains lacking Pol II, Pol IV, or both. It was argued that if Pol I were to compete with either polymerase, then the effect of polAexo might be amplified in such deficient backgrounds. The results clearly indicate that the polAexo effect is not changed in either the ΔpolB, ΔdinB, or ΔpolB ΔdinB backgrounds (Table 3). In the mutL background, the deletion of Pol II, Pol IV, or both, does not significantly affect the mutant frequency for the tested G·C→T·A allele, as observed before (Kuban et al., 2006; unpublished data), and the mutator effect of the polAexo allele remains unaltered (2.7- to 3.5-fold effect, Table 3A). An additional set of measurements was performed in the dnaE486 background, which was previously found to be suitable for demonstrating increased participation of Pol II and Pol IV (Banach-Orlowska et al., 2005; Kuban et al., 2005; Curti et al., 2008). This experiment, performed in the mutL⁺ background, clearly demonstrates that even under conditions where Pol II and Pol IV actively contribute to determining the mutation rate, the effect of polAexo still remains unchanged (0.8-to 1.6-fold effect, Table 3B).

Table 3.

Competition between Pol I, Pol II and Pol IV DNA polymerases.^a

		lac G·C→T·A (per 108)
Strain	lac orientation (strand)	polA+	polAexo	polAexo effect
A. mutL strains
dnaE+	R(lagging)	3.3 ± 0.9	9.1 ± 2.8	2.7
dnaE+ ΔdinB	R(lagging)	4.0 ± 1.2	11.5 ± 1.3	2.9
dnaE+ ΔpolB	R(lagging)	3.0 ± 1.2	10.5 ± 2.7	3.5
dnaE+ ΔdinB ΔpolB	R(lagging)	4.6 ± 1.8	12.4 ± 2.7	2.7
B. mutL+ strains
dnaE486	R(lagging)	4.2 ± 1.7	5.9 ± 2.2	1.4
dnaE486 ΔdinB	R(lagging)	2.7 ± 1.3	3.1 ± 1.4	1.1
dnaE486 ΔpolB	R(lagging)	8.4 ± 2.5	6.3 ± 3.9	0.8
dnaE486 ΔdinB ΔpolB	R(lagging)	0.8 ± 0.6	1.3 ± 0.7	1.6

^aThe experiments were performed with the lac G·C→T·A allele in the R(lagging) orientation. The mutL strains were mutL::Tn10. Mutant frequencies were determined as described in Experimental procedures. Each entry is based on the median value for 10 independent cultures grown at 37°C (dnaE⁺) or 34°C (dnaE486). No statistically significant effects are observed for the ΔpolB, ΔdinB, or ΔpolB ΔdinB strains in the polA⁺ dnaE⁺ series (both mutL⁺ and mutL) (P > 0.05). The corresponding differences in the polA⁺ dnaE486 series are significant (P < 0.05).

Statistically significant polAexo mutator effects are indicated in bold.

Lack of interaction of polAexo with Nucleotide Excision Repair (NER)

Pol I is the primary polymerase to fill in the gaps during NER, and the possibility that the polAexo mutator effect results from the error-prone filling of those gaps should be considered. In addition, Hasegawa et al. (2008) recently proposed an active role of NER in the production of spontaneous mutations, based on their observation that uvrA and uvrB deficiencies led to a reduction in spontaneous mutations. It was proposed that the NER system frequently engages in unnecessary "repair" of undamaged DNA and that the gapfilling synthesis by Pol I is responsible for a large fraction of spontaneous mutations (Hasegawa et al., 2008). To test this hypothesis in our present system we investigated the polAexo effect in strains deficient in NER. In Table 4 we present results for the uvrA and uvrC strains, along with the single mutant control. The experiments were conducted in both mismatch-repair proficient (Table 4A) and deficient backgrounds (Table 4B). Our data clearly indicate that the NER deficiency does not affect the polAexo mutator effect. In the uvr backgrounds, the mutant frequencies for the G·C→A·T and G·C→T·A lac alleles are essentially unchanged, and, most importantly, the magnitude of the polAexo mutator effect is unaltered and remains confined to the lagging strand orientation. The NER-deficiency did also not significantly change the polAexo mutator effect for the lac A·T→T·A transversion and did not affect the level of the lac A·T→G·C transitions (data not shown). Likewise, no altered polAexo effects were noted in the ΔuvrB background (not shown). Therefore, we conclude that the additional mutations observed in the polAexo strains are not a result of errors made by Pol I in excision repair tracts.

Table 4.

Mutator effects of proofreading-deficient polAexo mutant in NER-deficient background.^a

			lac → lac+ (per 108)
		lac orientation (strand)	polA+	polAexo	polAexo effect
A. mutL+ strains
uvr+	G·C→A·T	R(lagging)	0.6 ± 0.3	0.5 ± 0.3	0.8
		L(leading)	0.8 ± 0.4	0.5 ± 0.2
	G·C→T·A	R(lagging)	1.0 ± 0.7	2.4 ± 1.0	2.4
		L(leading)	1.0 ± 0.4	0.9 ± 0.4
uvrA	G·C→A·T	R(lagging)	1.4 ± 0.6	1.1 ± 0.9	0.8
		L(leading)	1.1 ± 0.4	1.3 ± 0.4
	G·C→T·A	R(lagging)	1.3 ± 0.6	3.3 ± 0.9	2.5
		L(leading)	1.4 ± 0.5	1.45 ± 0.6
uvrC	G·C→A·T	R(lagging)	1.2 ± 0.5	1.2 ± 0.6	1.0
		L(leading)	1.0 ± 0.4	1.0 ± 0.5
	G·C→T·A	R(lagging)	0.85 ± 0.6	2.2 ± 0.5	2.6
		L(leading)	0.86 ± 0.4	1.7 ± 1.4
B. mutL strains
uvr+	G·C→A·T	R(lagging)	33 ±15	39 ± 20	1.2
		L(leading)	139 ± 28	153 ± 13
	G·C→T·A	R(lagging)	1.4 ± 0.5	4.1 ± 1.5	2.9
		L(leading)	4.1 ± 0.9	2.8 ± 1.2
uvrA	G·C→A·T	R(lagging)	20 ± 7	26 ± 5	1.3
		L(leading)	129 ± 18	149 ± 16
	G·C→T·A	R(lagging)	2.1 ± 0.8	4.4 ± 1.2	2.1
		L(leading)	3.2 ± 1.6	3.9 ± 1.7
uvrC	G·C→A·T	R(lagging)	21 ± 6	37 ± 6	1.8
		L(leading)	126 ± 22	124 ± 27
	G·C→T·A	R(lagging)	2.1 ± 0.6	5.8 ± 2.8	2.7
		L(leading)	3.6 ± 0.9	3.6 ± 2.0

^aMutant frequencies were determined as described in Experimental Procedures. Each entry is based on the median value for 10 independent cultures grown at 37°C. The mismatch-repair-deficient strains were mutL::Tn10. The polAexo effect was calculated dividing the frequency of lac⁺ mutants in the polAexo and polA⁺ strains. All polAexo mutator effects shown in bold were statistically significant (P < 0.05, see Experimental procedures).

Discussion

Two major roles for DNA polymerase I have been established previously: (i) the maturation of Okazaki fragments by removal of the RNA primers in the lagging strand and filling the resulting gaps, and (ii) filling excision repair tracts during excision repair (NER and BER). In addition to those roles, another possible function of Pol I that we investigated in this work is its active role in polymerase switching at the replication point, analogously to the proposed roles of Pol II and Pol IV. In particular, a correction of Pol III-mediated misinsertion errors by proofreading-proficient polymerases such as Pol I (and Pol II) might constitute a chromosomal fidelity mechanism. Our observations of a mutator effect associated with the exonuclease-deficient form of Pol I indicate that Pol I plays a fidelity role under normal wild-type conditions. This fidelity role appears to be modest, up to about 4-fold, depending on the base-pair substitution investigated. With regard to the precise mechanism underlying this fidelity function, three findings are most relevant: (i) preferential mutator effects for lagging strand mutational events, (ii) lack of synergism between the polAexo mutator effect and other tested polymerase or replication defects, and (iii) apparent lack of competition of Pol I with the accessory polymerases Pol II and Pol IV. Our overall conclusion based on the current data and analysis following below is that this fidelity role is likely limited to the error-free processing Okazaki fragments and associated gap-filling, and does not relate to any direct polymerase exchanges at the replication fork nor involves excessive NER activity.

The lagging strand preference of the polAexo effect is obviously consistent with the established role of Pol I in processing the Okazaki fragments. Assuming an average Okazaki fragment length of 1000 nucleotides and the length of RNA primers of 10–20 nt, Pol I is likely responsible for 1% to 2% of all synthesis in the lagging strand. While a relatively small fraction, this mode of synthesis, if significantly less accurate than that of the Pol III HE, may still contribute measurably to the overall error rate. For example, if the exonuclease-deficient Pol I would be 50-fold less accurate than the proofreading-proficient Pol III HE, a not unreasonable assumption, then a 2% contribution may produce a doubling of the chromosomal error rate.

We have previously presented evidence suggesting that other DNA polymerase defects or disturbances can lead to mutator effects specific for the lagging strand, such as the overproduction of Pol V (Maliszewska-Tkaczyk et al., 2000) or Pol IV (Kuban et al., 2005; Gawel et al., 2008) or from the proofreading-deficient form of Pol II (Banach-Orlowska et al., 2005; Gawel et al., 2008). These results were interpreted to indicate preferential (although not exclusive, such as in the case of Pol II) access of these polymerases to the lagging strand. In all cases, the model considered that the action of these polymerases occurred on terminal mismatches created by misinsertions made by the replicative Pol III HE. These mismatches might be particularly accessible to alternative polymerases in the lagging strand due to the increased dissociative character of HE in this strand. The present study was undertaken to probe whether Pol I might also be involved in this type of terminal error processing. The results do not support this idea: the observed lack of synergism of the polAexo mutator with other replication-associated mutator effects as well as the lack of competition of Pol I with Pol II and/or Pol IV, suggest clearly that the action of Pol I resides in a distinctly different pathway from that envisioned for Pol II and Pol IV. While the latter three enzymes may compete with Pol III for mispairs at the replication point, this is not the case for Pol I. Instead, the operation of Pol I in remedying Okazaki fragment gaps provides a pathway that is separate from the polymerase switching at the replication point and readily accounts for the nonsynergistic (i.e., additive) effects of the polAexo mutator, as well as, of course, its preference for the lagging strand.

Our results with the NER-deficient uvr strains do not support a significant contribution to spontaneous mutation by Pol I-mediated gap filling during NER, a pathway recently proposed (Hasegawa et al., 2008). Notably, no significant differences were observed in the mutant frequencies (for either pol⁺ or polAexo strains) when comparing NER⁺ and NER⁻ strains. Secondly, it is unlikely that the action of Pol I during NER-mediated gap filling will lead to a strand bias favoring the lagging strand. Finally, most of our experiments were performed in the mismatch-repair-defective background, where uncorrected replication errors are expected to largely outnumber mutations from other sources. Obviously, errors made by Pol I during Okazaki fragment filling (in contrast to errors made during NER away from the replication point), are expected to be subject to correction by MMR, as is observed (see Table 1–Table 4).

To possibly address the discrepancy with the Hasegawa et al. (2008) study, it is worth pointing out that, in addition to any strain background effects, there are at least three significant differences between the two studies. Hasegawa et al. (2008) analyzed rifampicin-resistant mutants, occurring at a large number of sites in the rpoB gene. Secondly, they used exclusively mismatch-repair-proficient strains. Thirdly, the role of Pol I was assayed by expressing the polAexo gene from a multicopy plasmid, likely leading to significant overexpression. Each of these factors may affect the outcome and interpretation, and further studies will be needed to clarify these issues. Nevertheless, our present analysis of the strand-specific mutability of a series of lac alleles in the mismatch-repair defective background provides solid support for our proposed role of Pol I at the replication fork.

In Fig. 1 we present a scheme incorporating the current results on Pol I as well as previous data on the participation and contribution of other DNA polymerases to E. coli chromosomal replication fidelity (Kuban et al., 2004; Banach-Orlowska et al., 2005; Kuban et al., 2005). While Pol III, as the major replicative enzyme, provides for high overall fidelity, occasional participation by Pol II can lead to a further reduction in error rate by functioning as a back-up proofreader in cases where HE has difficulty negotiating a terminal mismatch (Banach-Orlowska et al., 2005; Gawel et al., 2008). Under those conditions, Pol II also acts to prevent access by Pol IV, whose action would lead to an increase in the chromosomal error rate. The precise mechanism underlying the hand-off to Pol II is not yet clear but may involve the action of the Pol III τ subunit (Pham et al., 2006; Gawel et al., 2008). The apparent exclusion from this process of Pol I, present in abundant quantities in the cell, as indicated here also suggests that this hand-off is not likely a simple mass-action driven process. When overproduced, Pol IV and Pol V can also gain access to the fork, leading to increased error rates. Their contributions seems largely confined to the lagging strand, and one may presume that there is a fundamental difference in the ways by which accessory polymerases obtain access to the two strands. Finally, Pol I is limited to the filling of Okazaki fragments generated during lagging strand DNA synthesis, a role for which it is well adapted due to its associated 5'→3' exonuclease. Its 3'→5' (proofreading) exonuclease ensures that this gapfilling occurs with sufficient accuracy so as not to produce an increase in the overall replication error rate.

Fig. 1.

Participation of DNA polymerases at replication fork in E. coli. Shown is a scheme delineating events that might occur upon Pol III HE creating a 3' terminal mismatch (M·m) by misinsertion in lagging or leading strand. The HE is depicted in a simplified form as two Pol III cores connected by the dimeric τ subunit. The most likely outcome of a misinsertion is removal of the misinserted nucleotide by the Pol III proofreading activity. However, occasionally proofreading might not occur, and the resulting stalling of HE may provide an opportunity for other polymerases to participate. Here, we have shown a dissociation of the Pol III core from the mismatch in the lagging strand, which may occur in a manner similar to the dissociation normally occurring at the end of an Okazaki fragment. This provides an opportunity for Pol II to act as a back-up proofreader (Banach-Orlowska et al., 2005) or for Pol IV or Pol V to extend the mismatch (mutator activity). Polymerase exchange is also possible in the leading strand, but possibly through a different mechanism that strongly favors action by Pol II. The function of Pol I appears limited to fixing the Okazaki fragment gap, generally in an error-free way. See Text for more details.

Experimental procedures

Media

Solid and liquid media were as described (Fijalkowska and Schaaper, 1995). Minimal plates were supplemented with 0.4% glucose or 0.4% lactose as a carbon source and 50 µg ml⁻¹ amino acids, as required. Antibiotics were added as follows: tetracycline, 12.5 µg ml⁻¹; chloramphenicol, 24 µg ml⁻¹; spectinomycin, 20 µg ml⁻¹; kanamycin 50 µg ml⁻¹; ampicillin, 25 µg ml⁻¹; rifampicin, 100 µg ml⁻¹; nalidixic acid, 30 µg ml⁻¹.

Construction of plasmids

Generally, plasmids were constructed according to the standard protocols as described by Sambrook et al. (2001). Propagation of plasmids was performed in E. coli DH5α. PCR products were confirmed by DNA sequencing.

Cloning of the polA gene

Genomic DNA from strain MC4100 was used as a template to clone the polA gene by PCR amplification (using Pfu Turbo polymerase) with primers: UP-polA: 5'-CGCTTAAAGCTTTTGTCATTGATGTAG-3' (HindIII site underlined) and LOW-polA: 5'-GCATAGGGAATTCTAATAGCCATCAC-3' (EcoRI site underlined). The 4.1-kb HindIII/EcoRI fragment was then inserted into the chloramphenicol-resistance low-copy plasmid pHSG576 (Takeshita et al. 1987) resulting in plasmid ppolA (~7.7 kb). Correctness of the construct was confirmed by sequencing with several primers homologous to the sequence of polA gene.

Construction of the polAD424A (polAexo) allele

To create plasmid ppolAex containing the exonuclease-deficient D424A mutation in polA (positions 1270–1272), PCR amplification was performed using plasmid ppolA as a template and as primers D424A-F, 5'-CAAAACCTGAAATACGCGCGCGGTATTCTGGCGAACTACGGC-3', and D424A-R, 5'-GCCGTAGTTCGCCAGAATACCGCGCGCGTATTTCAGGTTTTG-3'. The PCR reaction was performed using the Stratagene Site-Directed Mutagenesis Kit and protocols provided by the manufacturer. Replacement of the GAT codon by GCG at polA positions 1270–1272 leads to substitution of Asp424 by Ala (D424A). Among transformants, the presence of the D424A mutation was verified by sequencing the polA region. The D424A alteration also leads to the loss of a PvuI restriction site.

Gene replacement cassette for substitution of chromosomal polA⁺ by polAexo

The cassette used for replacing the chromosomal polA gene by the polAexo variant contained the cat gene from transposon Tn9 inserted into the intergenic region between the polA gene and the adjacent spf gene. The cassette was assembled in three sections, in linear order: (i) a KpnI-EcoRI fragment containing the polA⁺ or polA(exo) gene, (ii) an EcoRI-Hind III fragment gene containing the cat gene from transposon Tn9, and (iii) a HindIII-BamHI fragment containing part of the spf gene, which is located downstream of polA on the chromosome. The polA alleles were amplified by PCR from plasmid ppolA or ppolAex, respectively, using as primers: UP-polA, 5'-GGTACCCTCTCATACCAGCTGGCGACG-3' (KpnI_site underlined) and LOW-polA, 5'-GCATAGGGAATTCTAATAGCCATCAC-3' (EcoRI_site underlined). The resulting KpnI-EcoRI PCR products (~2.1 kb) were cloned into pBluescript II SK (+) (Stratagene) using the KpnI and EcoRI sites, yielding plasmids pSKpolA and pSKpolAD424A, respectively. The cat gene of transposon Tn9 was PCR amplified using plasmid pHSG576 (Takeshita et al. 1987) as a template and as primer pair: up_Tn9 5'-CTAGAATTCGCGCCGAATAAATACCTGTGACG-3' (EcoRI site underlined) and lw_Tn9 5'-GCACTTTTTGTAATTTTTTTTCAGTTGTTGCATAGGAAGCTTAACTGGCCTCAGGCAT TTGAG-3' (HindIII site underlined) resulting in 1097-bp PCR product. Primer lw_Tn9 is partially homologous to the spf gene (double underline). The third fragment was prepared from chromosomal DNA of strain MC4100 by amplification of the spf gene downstream of polA using as primers: up_spf (homologous to lw_Tn9) 5'-GTTAAGCTTCCTATGCAACAACTGAAAAAAAATTACAAAAAGTGC-3' (HindIII site underlined) and lw_spf 5'-GCAGGATCCTTTCTTGCCTTAATGCTTGTGCC-3' (BamHI site underlined), yielding a 311-bp PCR product. The 1097-bp and 311-bp PCR fragments were then joined by an amplification reaction using primers up_Tn9 and lw_spf described above, yielding a 1408-bp Tn9-spf fragment. This fragment was ligated into EcoRI/BamHI digested plasmids pSKpolA and pSKpolAD424A, yielding plasmids pSKpolAint and pSKpolAD424Aint, carrying the complete gene replacement cassettes. KpnI/BamHI restriction of pSKpolAint and pSKpolAD244Aint yielded the linear cassettes used for replacement (see below).

Gene replacement

To replace the chromosomal polA⁺ gene with the cat-containing polA⁺ or polAexo cassette described above, we used a modification of the method of Datsenko and Wanner (2000). Strain BW25141 carrying plasmid pKD46 (providing the λ-Red recombination proteins) was transformed with the linear KpnI-BamHI DNA fragments described above. Upon transformation of the fragment into BW25141 (pKD46), homologous recombination between the polA and spf regions of chromosome and the fragment led to integration of the cat-carrying cassette into the chromosome, permitting selection of the desired recombinants by chloramphenicol resistance. P1 transduction using P1virA was then used to transfer the new alleles into strain KA796 (Schaaper et al., 1985). The resulting strains were proven to carry the polA⁺ or polAexo allele by sequencing of the polA gene. Replacement of the chromosomal polA region by the Tn9cat-containing cassette was confirmed by two PCR reactions. The presence of the D424A mutation was confirmed by single-colony PCR using as primers 5'-CGGACTGGATACGCTGTATGC-3' (forward) and 5'-CTTCCAGTACCTCTTCCGACG-3' (reverse), yielding a ~1.2 kb product. As the D424A alteration leads to the loss of a PvuI restriction site, the PCR product was digested with PvuI. Only in case of the wild-type polA⁺ allele the digestion results in 646-bp and 574-bp products. The second colony PCR was performed using one primer internal to the Tn9cat sequence (5'-AACGTGGCCAATATGGAC-3' forward) and a second primer complementary to a sequence in the adjacent yihA gene (5'-ACACGCTGACTAACCAGA-3' reverse) (gene order polA-cat-spf-yihA). A ~1.2-kb PCR product was obtained only if the integration cassette was introduced into the expected chromosomal position of the polA locus.

Expression analysis

To check whether expression of the polA gene from the cat-containing chromosomal construct was altered in any way from the normal chromosomal configuration, we performed real-time Quantitative PCR analysis of polA mRNA. These experiments are detailed in the Supporting Information. The results (Fig. S1) indicate that expression of the polA gene is unaffected by the presence of the nearby cat gene. Expression of the polAexo gene was slightly reduced to 76 % of that of the polA⁺ gene, possibly within the margin of error.

Strain constructions

For the purpose of mutagenesis studies, the new KA796-derived strains containing the polA⁺ or polAexo (polAD424A) alleles (see above) were used as donors in P1virA-mediated transductions into a set of lacZ tester strains. These strains, described before (Fijalkowska et al., 1998), contain defined lacZ alleles in the two possible orientations on the chromosome. With the one exception indicated below, they are derivatives of MC4100 [Δ(argF-lac)169] containing the lac operon derived from strains CC102, CC104, CC105, and CC106 (Cupples and Miller, 1989) inserted into the phage λ attachment site in two orientations (Fijalkowska et al., 1998). The four lacZ missense alleles allow scoring of mutagenesis via reversion to lac⁺ by a specific base substitution mutation: G·C→A·T, G·C→T·A, A·T→T·A, or A·T→G·C, respectively (Cupples and Miller, 1989).

The mutL::Tn5 or mutL::Tn10 markers, specifying a deficiency in mismatch repair, were introduced by transduction from strains NR9559 (Fijalkowska and Schaaper, 1995) or NR9161 (Schaaper et al., 1989) using selection for kanamycin or tetracycline resistance, respectively. The polB1 deletion allele [Δ(araD-polB)::Ω] was transferred by transduction from strain SH2101 (Escarceller et al., 1994) using spectinomycin resistance as a selective marker. The polBex1 derivatives were created as described (Banach-Orlowska et al., 2005). The dinB::kan marker was introduced by transduction from YG7207 (Kim et al., 1997).

The dnaQ49 and dnaQ928 alleles, specifying Pol III proofreading defects, were introduced by transduction from strain NR9695 (Schaaper and Cornacchio, 1992) or NR11641 (Taft-Benz and Schaaper, 1998), respectively, using selection for tetracycline resistance and scoring for mutator phenotype (~ 40% linkage of dnaQ with zae-502::Tn10).

The dnaN159 (TS) allele, specifying a defect in the HE β-clamp, was transferred from strain HC194, obtained from the Coli Genetic Stock Center (Yale University). First, the dnaN allele was linked in HC194 with transposon zid-501::Tn10 (Singer et al. 1989), yielding strain NR11824 (dnaN159, zid-501::Tn10). NR11824 was then used as donor for subsequent transfers of dnaN159 (selection for tetracycline resistance and scoring for impaired growth at 42°C).

The dnaE486 (TS) and dnaE511 (TS) alleles were transferred from strains CS115 and AR2 (Vandewiele et al., 2002), respectively, kindly provided by Dr. R. Woodgate (NICHHD, NIH). Selection was for tetracycline resistance conferred by transposon zae-502::Tn10, followed by screening for impaired growth at 42°C (~ 40% linkage).

The dnaE915 (antimutator) allele was transferred from strain NR9905, in which dnaE915 is flanked by two nearby markers, zae-502::Tn10 and zae::Tn10d-Cam (Fijalkowska et al. 1993). An intermediate strain containing dnaE915 linked only to zae-502::Tn10 was constructed, which was then used to transfer dnaE915 into the desired lacZ strains carrying polA⁺ or polAexo, using tetracycline resistance. In all cases, the presence of dnaE915 allele was confirmed by sequencing.

The ΔuvrA::mini-Tn10kan, ΔuvrB::mini-Tn10kan and ΔuvrC::mini-Tn10kan mutations were transferred from strains JD25374, JD26189, and JD26681, respectively, obtained from the National Institute of Genetics, Japan (http://www.shigen.nig.ac.jp/ecoli/strain/top/top.jsp). Selection was for kanamycin resistance, followed by screening for UV sensitivity.

Experiments with the dnaX36 allele were performed in a slightly different strain background. Strain NR11927 (ara, thi, Δprolac, mutL::Tn5, dnaX36), identical to the previously described strain NR11928 (Pham et al., 2006), was transduced to carry the lac operon of strain CC104 in two orientations in the λ att site (Fijalkowska et al., 1998) using ampicillin resistance selection, followed by introduction of the polAexo allele.

Mutant frequency measurements

Mutant frequencies were determined by starting, for each strain, 10–20 cultures (2 ml of LB) from single colonies and growing them to saturation at 37°C (30°C or 34°C, where indicated) with agitation. Colonies were taken from two or three independent lac integrants for each lac orientation and usually from several independent transductants (e.g., polA⁺, polAexo, etc.) for each integrant. As a polA⁺ control for all polAexo experiments, the analogously constructed polA⁺ strains were used that also carried the cat cassette (see construction, above). For lac reversions assays appropriate dilutions were plated on minimal glucose plates to determine the total cell count and on minimal lactose plates to determine the number of lac⁺ mutants. For the forward mutagenesis assays toward rifampicin or nalidixic acid resistance, appropriate dilutions were plated on LB plates to determine the total cell count and on LB plates containing rifampicin or nalidixic acid to determine the number of antibiotic-resistant mutants. To calculate mutant frequencies, the number of mutants per plate was divided by the total number of cells. The non-parametric Mann-Whitney criterion (Sokal and Rohlf, 1981) was applied to the mutant frequency distributions for the purpose of comparing any given set of two strains using Statistica 5.5 (StatSoft) analysis software.