Mutator mutants of Escherichia coli carrying a defect in the DNA polymerase III τ subunit


  • Phuong T. Pham,

    1. Laboratory of Molecular Genetics, National Institute of Environmental Health Sciences, Research Triangle Park, NC 27709, USA.
    Search for more papers by this author
    • Present address: Department of Molecular Biology, University of Southern California, Los Angeles, CA 90089, USA.

  • Wei Zhao,

    1. Laboratory of Molecular Genetics, National Institute of Environmental Health Sciences, Research Triangle Park, NC 27709, USA.
    Search for more papers by this author
  • Roel M. Schaaper

    Corresponding author
    1. Laboratory of Molecular Genetics, National Institute of Environmental Health Sciences, Research Triangle Park, NC 27709, USA.
    Search for more papers by this author

*E-mail; Tel. (+1) 919 541 4250; Fax (+1) 919 541 7613.


To investigate the possible role of accessory subunits of Escherichia coli DNA polymerase III holoenzyme (HE) in determining chromosomal replication fidelity, we have investigated the role of the dnaX gene. This gene encodes both the τ and γ subunits of HE, which play a central role in the organization and functioning of HE at the replication fork. We find that a classical, temperature-sensitive dnaX allele, dnaX36, displays a pronounced mutator effect, characterized by an unusual specificity: preferential enhancement of transversions and −1 frameshifts. The latter occur predominantly at non-run sequences. The dnaX36 defect does not affect the γ subunit, but produces a τ subunit carrying a missense substitution (E601K) in its C-terminal domain (domain V) that is involved in interaction with the Pol III α subunit. A search for new mutators in the dnaX region of the chromosome yielded six additional dnaX mutators, all carrying a specific τ subunit defect. The new mutators displayed phenotypes similar to dnaX36: strong enhancement of transversions and frameshifts and only weak enhancement for transitions. The combined findings suggest that the τ subunit of HE plays an important role in determining the fidelity of the chromosomal replication, specifically in the avoidance of transversions and frameshift mutations.


Recent advances in understanding the molecular mechanisms of inherited cancer susceptibilities have increased the interest in mutators, i.e. cell lines with higher mutation rates than wild-type cells. The bacterium Escherichia coli has been a productive model system for investigating mutators and their underlying mechanisms. Those studies have revealed mutators affected in the fidelity of DNA replication, postreplicative mismatch repair, and the repair of a wide range of DNA damages (for review, see Miller, 1996). Of central importance in mutagenesis is the fidelity of DNA replication, because it controls the production of mutations on normal and damaged DNA templates. Replication of the E. coli chromosome is performed by DNA polymerase III holoenzyme (HE), a large dimeric complex composed of two polymerase III core assemblies and a number of accessory factors. The accessory subunits are responsible for high processivity and for co-ordinating simultaneous synthesis of leading- and lagging-strands (for reviews, see Kelman and O’Donnell, 1995 or McHenry, 2003). The core polymerase contains three tightly bound subunits: α, the polymerase encoded by the dnaE gene; ɛ, the 3′→5′ proofreading subunit encoded by the dnaQ gene; and θ, a subunit encoded by the holE gene, which may function to stabilize the ɛ subunit (Taft-Benz and Schaaper, 2004). The accessory proteins include the β subunit, which forms a ring-shaped structure (sliding clamp), tethering the core to the DNA (Kong et al., 1992) to ensure high processivity, and the seven-subunit DnaX complex (τ2γδδ′χψ) (Dallmann and McHenry, 1995; Onrust et al., 1995; McHenry, 2003). Within the DnaX complex, the products of the dnaX gene (τ and γ subunits) play important roles. The complex functions as a clamp loader for the β-sliding clamp (Naktinis et al., 1996). The τ subunit (τ2) serves to dimerize the two Pol III cores (McHenry, 1982; Stukenberg and O’Donnell, 1995; Gao and McHenry, 2001) enabling a physical coupling of the leading- and lagging-strand polymerase (Kim et al., 1996; McInerney and O’Donnell, 2004).

With regard to fidelity, two of the HE subunits have received most of the attention. The α (polymerase) subunit is responsible for discrimination against incorrect nucleotides during the insertion step (base selection), while the ɛ subunit is responsible for proofreading misinserted nucleotides using its 3′→5′ exonuclease. θ subunit many play an indirect role by stabilizing the ɛ subunit (Studwell-Vaughan and O’Donnell, 1993; Taft-Benz and Schaaper, 2004). Together, base selection and proofreading are thought to maintain the accuracy of polymerization at about one error per 107 bases replicated (Schaaper, 1993a). Although genetic and biochemical studies have revealed insights into the organization and functions of the accessory proteins, little is known about their role in replication fidelity. To gain insight into this matter, we have initiated studies of E. coli mutants affected in genes encoding the HE subunits. Here, we show that a classical temperature-sensitive allele, dnaX36, encoding a mutant τ subunit, increases significantly the frequency of frameshift and base pair substitution replication errors. Among the base substitutions, transversions are enhanced preferentially. A search for new transversion mutators by localized mutagenesis targeting the dnaX gene yielded six new, τ-specific dnaX mutator alleles, which, like dnaX36, specifically enhance transversions and (−1) frameshifts. Thus, we have discovered a new class of E. coli mutators. The possible mechanism(s) by which the τ subunit contributes to high replication fidelity are discussed. Specifically, we develop a model in which τ may act as a sensor of certain Pol III HE misinsertion errors.


Mutator phenotype of dnaX(Ts) mutants

Among the accessory proteins of E. coli DNA polymerase III HE, the τ and γ subunits are of particular interest. They are the products of the same gene, dnaX (Kodaira et al., 1983; Mullin et al., 1983; Flower and McHenry, 1986). τ, a 71.1 kDa protein, is the full-length 643-amino-acid product, while γ, a 47.5 kDa product, results from early termination within the dnaX reading frame by a (−1) ribosomal frameshift over codons 428–430 (Blinkowa and Walker, 1990; Flower and McHenry, 1990; Tsuchihashi and Kornberg, 1990). Full-length τ consists of five domains, three of which (domains I, II and III) are common to both γ and τ, while the last two (IV and V) are unique to τ (Dallmann et al., 2000; Gao and McHenry, 2001). These two additional domains allow τ to perform several unique functions, such as to dimerize the two α subunits and to co-ordinate fork progress with the DnaB helicase (McHenry, 2003). Two conditionally viable (temperature-sensitive) dnaX alleles have been described: dnaX2016 and dnaX36 (Henson et al., 1979). dnaX2016 contains a glycine (GGT) to aspartate (GAT) change at residue 118 that affects both τ and γ, while dnaX36 contains a glutamate (GAG) to lysine (AAG) change at residue 601 which affects only τ (Blinkova et al., 1993).

We investigated the possible effects of dnaX2016 and dnaX36 on replication fidelity in a mutL strain background. The mutL mutation abolishes the MutHLS mismatch repair system, which corrects DNA mismatches resulting from replication errors (for review, see Modrich, 1991). Thus, all or most of the mutations observed in a mutL strain represent uncorrected replication errors, facilitating direct analysis of effects on the replication error rate. Table 1 shows the effects of the dnaX alleles on the frequency of rifampicin or nalidixic acid-resistant mutants, which arise by base pair substitution, and of revertants of the trpE9777 frameshift marker. The latter represent the loss of an A·T base pair (−1 frameshift) from a run of six consecutive A·T pairs (Siegel and Vaccaro, 1978). The results indicate that both dnaX alleles confer a mutator phenotype. The effect is temperature-dependent for dnaX2016, as it is observed essentially only at 37°C, at which temperature it affects all three tested markers. In contrast, the effects of dnaX36 are not dependent on the temperature and are largely restricted to the trp frameshift marker (11- to 14-fold enhancement). These data suggest that the two dnaX mutator effects may have defined, and different, specificities. This specificity was further investigated as described below.

Table 1.  Mutant frequencies (per 108 cells) in dnaX(Ts) strains at 30°C and 37°C.
  1. Entries for the mutL and mutL dnaX strains represent the average mutant frequency from two independent experiments, each comprising 12–24 independent cultures. In parentheses, the fold increase (mutator effect). Asterisks (*) indicate mutator effects that were statistically significant in both experiments (P < 0.05, see Experimental procedures). The data for the mutL+ control strain are from published experiments (Schaaper, 1993b; Fijalkowska et al., 1997) and are shown for comparison purpose. RifR and NalR indicate resistance to rifampicin and nalidixic acid respectively. Trp+ indicates revertants of the trpE9777 frameshift allele. Strains used were NR9464 (mutL), NR11920 (mutL dnaX2016) and NR11921 (mutL dnaX36). nd indicates that no data are available.

mutL+ndndnd   4.2  2.0   0.46
mutL330140 170 435120 200
mutL dnaX2016170 (0.5) 91 (0.6) 495 (2.9)2840 (6.5)*310 (2.6)*3300 (17)*
mutL dnaX36435 (1.3)110 (0.8)1900 (11)* 720 (1.7)130 (1.1)2700 (14)*

Mutational specificity of the dnaX(Ts) mutators

The mutational specificity of the two dnaX alleles was examined using the lacZ reversion system developed by Cupples and Miller (1989). It uses a set of six strains each carrying a different F′prolac (CC101 through CC106) containing a lacZ gene carrying a different mutation at codon 461. Each can revert to lac+ via only one of the six possible base pair substitutions, restoring the glutamic acid residue that is essential for growth on lactose. Table 2 shows that dnaX2016 moderately increases both transitions and transversions (two to fivefold). In contrast, dnaX36, while modestly affecting the two transitions (two to fourfold), strongly enhanced the G·C→T·A and A·T→T·A transversions (38- and 17-fold respectively). An enhancement by dnaX36 is also seen for the other two transversions, A·T→C·G and G·C→C·G, although the fold increase is difficult to estimate because these events were too infrequent to be detected in the control strain. The frequency of RifR mutations was also measured in this series with similar effects as in Table 1: dnaX2016 increases the frequency by 15-fold, but dnaX36 increases it less than twofold. Overall, these data indicate that while both dnaX2016 and dnaX36 confer a clear mutator effect, their specificities appear very different and, hence, their underlying mechanisms may be different. In the following, we have focused on the dnaX36 mutator for two reasons. One, its mutational specificity, enhancing predominantly transversions and frameshifts, is rather unusual and likely results from a specific defect. Second, the interpretation of its mutator effect is facilitated, as only the τ subunit is affected, in contrast to dnaX2016, which presumably has impaired τ and γ subunits. Below, we investigate the specificity of the dnaX36 mutator in detail by using the lacI forward system.

Table 2.  Specificity of dnaX (Ts) mutators (mutants per 108 cells).
Genotypelaclac+ via indicated base pair substitutionRifR
  1. Strains used were NR9559 (mutL), NR11926 (mutL dnaX2016), NR11928 (mutL dnaX36), but carrying F′CC101 through F ′CC106 (see Experimental procedures), which allow specific scoring of the indicated base pair substitutions (Cupples and Miller, 1989). The strains were grown at 37°C. Entries for the mutL and mutL dnaX strains represent the average (±SD) of frequencies from three independent experiments at 37°C. Numbers in parentheses are fold increases over the mutL strain (mutator effect). RifR, resistance to rifampicin. A < sign indicates that no revertants were observed. Tv, transversion; Ti, transition. The entries for the mutL+ control strain are historical data from our laboratory based on three to ten individual experiments and are shown here for comparison.

mutL+  0.33  0.54< 0.01  0.690.31 0.28   4.2
mutL< 0.2 57 ± 2.0< 0.02 3.4 ± 0.70.60 ± 0.2817 ± 16 350 ± 84
mutL dnaX2016< 0.5150 ± 37 (2.6)< 0.08 17 ± 11 (5.0) 2.1 ± 1.7 (3.5)38 ± 16 (2.2)5200 ± 1700 (15)
mutL dnaX36  0.69230 ± 22 (4.0)  0.27120 ± 10 (35) 8.9 ± 2.3 (15)35 ± 10 (2.1) 650 ± 175 (1.9)

Specificity of the dnaX36 mutator in the lacI forward system

The lacI gene encodes the repressor of the lac operon and presents a forward target in which multiple classes of mutations (base substitutions, −1 frameshifts, deletions, duplications, and more complicated rearrangements) can be scored at multiple sites. The target for these studies is the lacI gene residing on an F′prolacI+Z+ (F′128-27) (Schaaper and Dunn, 1991). LacI mutants are selected based on constitutive expression of the lac operon (see Experimental procedures) and the precise nature of the mutations is determined by DNA sequencing. Routinely, this analysis is restricted to the subgroup of the dominant mutations (lacI d), which provides a target of about 200 nucleotides with a high density of scorable sites (Schaaper and Dunn, 1991; Schaaper, 1993a). The system has been used previously to analyse the nature of E. coli replication errors in various genetic backgrounds (Schaaper, 1988; 1993a; 1993b; Schaaper and Dunn, 1991).

The lacI d mutant frequencies were 46 × 10−6 for the mutL strain and 188 × 10−6 for the mutL dnaX36 strain (see Experimental procedures), a 4.1-fold overall mutator effect. The sequencing of 300 or 280 independent mutants for the mutL and mutL dnaX36 strain, respectively, yielded the spectra presented in Fig. 1. As before (Schaaper, 1993a), the spectrum of mutations in the mutL strain is dominated by transition base substitutions (yellow). In addition, frequent −1 frameshifts occur at two previously identified sites where an A is lost from a run of five or four As (green triangles). The mutL dnaX36 spectrum is more diverse, containing a significantly greater proportion of transversions (red) and more −1 frameshifts (green). Also, the −1 frameshifts no longer occur primarily at the two mutL hot spot sites, instead shifting to sequences that are generally of the non-run type. In Table 3, we list the precise spectral compositions. In the final column we calculate the specific mutator effect for each of the mutational classes. The results show that dnaX36 is only a modest mutator for transition errors (2.5-fold) but a strong mutator for transversion errors (> 20-fold). In addition, the mutator effect for frameshift mutations is not only large (80-fold) but also specific for one class of frameshifts. Overall, these data from the forward system fully confirm the results gleaned from the reversion assays. In addition, they provide new information on the nature of the (−1) frameshift mutations – occurring at non-run sequences, which is likely relevant to the mechanism by which dnaX36 enhances mutagenesis.

Figure 1.

Spectra of sequenced lacId mutations in the mutL (top) and mutL dnaX36 strain (bottom). See text and Experimental procedures for details.

Table 3.  Summary of sequenced lacI d mutations in mutL and mutL dnaX36 strains.
MutationmutLmutL dnaX36dnaX36 mutator effect
  1. Listed are the aggregate numbers for each of the indicated classes of sequenced lacI d mutations as shown in Fig. 1. The last column calculates the specific mutator for each class based on the overall mutator effect of 4.1-fold. For example, for the G·C→T·A transversions, the effect can be calculated as [(46/241)/(7/267)] × 4.1 = 29.8.

Base substitutions
 A·T→G·C137 59  2.0
 G·C→A·T 96 78  3.7
 Transitions233137  2.7
 A·T→C·G  3 12 18
 A·T→T·A  3  1  1.5
 G·C→T·A  7 46 30
 G·C→C·G  0  3>14
 Transversions 13 62 22
−1 frameshifts
 Runs 19  5  1.2
 Non-runs  2 37 84
Total267241  4.1

Isolation of new dnaX mutators

The dnaX36 mutator effect, shown above to be specific for transversions and frameshift mutations, represents a specificity not reported before in E. coli. This prompted us to look for additional mutators in the dnaX gene that might have similar properties. We performed localized mutagenesis of the dnaX region of the E. coli chromosome (see Experimental procedures) using linkage with the zba-2321::mini-Tn10Cam transposon, which is tightly linked with dnaX (8590%). We scored for increased levels of papillation (Fig. 2) in mismatch repair-defective strains NR11939 (F′CC104) or NR11940 (F′CC105), which detect G·C→T·A and A·T→T·A transversions respectively. Putative dnaX mutators were selected as colonies having at least threefold more papillae than the mutL control. In three separate experiments, involving some 48 000 transductants, 27 candidate mutators were obtained. These were further tested by direct measurements of lac reversion frequencies and backcrosses to strains NR11939 or NR11940 to test for linkage with the transposon. Six closely linked mutators (70–90% cotransduction) were obtained: five resulting from the screening with the G·C→T·A transversion (dnaX983 and dnaX986-989) and one from the screen with the A·T→T·A transversion (dnaX985). Figure 2 shows, as an example, the increased papillation pattern of the dnaX985 mutator.

Figure 2.

A dnaX mutator on an XPG plate. NR11939 (mutL, F′CC104) displays ∼5–10 papillae per colony, while the dnaX985 mutator shows greatly increased number of papillae (∼500). Papillae represent minicolonies of lac + revertants, which have a selective advantage over the parental lac strain due to the presence of P-gal in the XPG medium (see Experimental procedures). The presence of X-gal facilitates their visualization. The plate is the results of a P1 transduction in which dnaX985 was introduced linked with zba-2321::mini-Tn10Cam (∼90% linkage) into NR11939. The plate was incubated for 3 days at 37°C followed by 3 days at room temperature.

The location of the responsible mutation in the dnaX gene was confirmed by complementation using pWTX-1, a plasmid carrying the dnaX+ gene (see Experimental procedures). In the presence of this plasmid, all mutant frequencies returned to the control level (data not shown), indicating that the new mutator alleles are recessive and reside in the dnaX gene.

Characterization of the novel dnaX mutators

The new dnaX mutators showed healthy normal growth at 37°C and appeared without significant defects, like the dnaX36 strain. The base pair substitution specificity of the new dnaX mutators was examined using the lacZ reversion system (at 37°C). The results (Table 4) show that all exhibit a similar phenotype: a strong and specific enhancement of the G·C→T·A (10- to 64-fold) and A·T→T·A (three to 32-fold) transversions with little enhancement (one to fourfold) of the two transitions (G·C→A·T and A·T→G·C). Similarly, the RifR frequency, another indicator for transition mutations in a mutL background (Rangarajan et al., 1997; Garibyan et al., 2003), was increased less than twofold. The mutators also enhanced significantly (3.5- to 17-fold) the (−1) frameshifts of the trpE9777 system. This pattern is similar to that seen for the dnaX36 mutator.

Table 4.  Mutator effects of newly isolated dnaX alleles for base pair substitution and (−1) frameshift mutations (fold increases over dnaX +).
Genotypelaclac+ via indicated base pair substitutionRifRTrp+
  1. Strains used for the determination of lac and RifR mutant frequencies were derivatives of NR9559 (mutL) carrying the corresponding dnaX alleles and F′CC102, F ′CC104, F ′CC105 or F ′CC106 for measuring the indicated base pair substitution. Strains were grown at 37°C. Strains used for measuring reversion of the trpE9777 frameshift allele were derivatives of NR9464.

  2. ND, not determined; Tv, transversion, Ti, transition.

mutL1.0 1.0 1.0
mutL dnaX9833.119 7.01.0ND 8.2
mutL dnaX9854.064322.21.717
mutL dnaX9863.040122.11.314
mutL dnaX9873.4 9.4 9.01.2ND 3.9
mutL dnaX9882.712 6.8
mutL dnaX9892.615 3.5

The six new dnaX mutator genes were sequenced along with the wild-type gene. The locations of the observed base pair changes and corresponding amino acid substitutions are listed in Table 5. All result from G·C→A·T changes, consistent with the mutational specificity of MNNG (N-methyl-N′-nitro-N-nitrosoguanidine), the mutagen that was used to induce them. Each carried at least one mutation in the C-terminal part of dnaX that is unique to the τ subunit. Specifically, they are located in domain V (residues 497–643) (Gao and McHenry, 2001), which is unique to τ, and which constitutes the α binding domain. Three alleles, dnaX983, dnaX985 and dnaX989, contained only a single mutation, while the other three carried more than one. [The latter is likely due to the strongly mutagenic action of MNNG, which is known to induce multiple mutations in bacteria (Guerola et al., 1971)]. The fact that all dnaX mutators exhibited a similar phenotype and each of them carries one or more mutation(s) in the part of dnaX gene unique to τ, strongly suggests that the amino acid changes in τ are responsible for the mutator phenotype. Thus, like dnaX36, the novel dnaX mutants carry a specific defect in the τ subunit, which leads to an unusual mutator effect characterized by the specific enhancement of transversion base pair substitutions and (−1) frameshifts.

Table 5.  Sequence changes in newly isolated dnaX mutators.
AlleleCodon change (nucleotide position)Amino-acid change (position)Subunit affected
  1. The mutated base is underlined. The numbering system is as in Yin et al. (1986).

dnaX983GAT→AAT (1861)Asp→Asn (621)τ
dnaX985CCG→CTG (1796)Pro→Leu (599)τ
dnaX986GCT→ACT (118)Ala→Thr (40)τ + γ
GGT→GAT (1742)Gly→Asp (581)τ
GAT→AAT (1861)Asp→Asn (621)τ
dnaX987GCA→ACA (1543)Ala→Thr (515)τ
GAT→AAT (1906)Asp→Asn (636)τ
dnaX988CGC→CAC (551)Arg→His (184)τ + γ
GCG→ACG (1897)Ala→Thr (633)τ
dnaX989CGC→TGC (1669)Arg→Cys (557)τ

dnaX mutator effects and DNA mismatch repair

The data presented above show that the dnaX alleles exert a mutator effect in a mismatch repair-defective (mutL) background, as expected for a mutator for DNA replication errors. The data in Table 6 comparing the effect of the dnaX36 and dnaX985 alleles in mut+ and mutH, mutL or mutS backgrounds demonstrate that the dnaX-induced errors are subject to correction by the mutHLS-dependent mismatch repair system. The corrective effect is seven to 23-fold for the G·C→T·A transversion, and about twofold for the A·T→T·A transversion, values consistent with the general (modest) correctability of transversion mismatches (Schaaper and Dunn, 1991).

Table 6.  Effect of the mutHLS mismatch repair system on the mutability of dnaX mutators (mutants per 108 cells).
Genotypelac→ lac+ via indicated base pair substitution
G·C → T·AA·T → T·A
  1. Strains used are: NR10834 and NR10835 (wild type), NR11939 and NR11940 (mutL), NR12440 and NR12441 (dnaX36), NR12451 and NR12453 (dnaX36 mutL), NR12452 and NR12454 (dnaX36 mutS), NR12455 (dnaX36 mutH), NR12436 and NR12437 (dnaX985), NR12445 and NR12448 (dnaX985 mutL), NR12446 and NR12449 (dnaX985 mutS), and NR12447 and NR12450 (dnaX985 mutH) (see Table 7). Strains were grown at 37°C. Numbers in parentheses indicate the fold increase due to loss of mutL, mutS or mutH function and reflect the susceptibility of the particular mutations to mismatch repair.

  2. ND, not determined.

Wild type 0.520.29
mutL 2.4 (4.6)0.42 (1.4)
dnaX36 3.41.9
dnaX36 mutL62 (18)3.5 (1.8)
dnaX36 mutS77 (23)2.9 (1.5)
dnaX36 mutHND2.0 (1.1)
dnaX985 2.60.9
dnaX985 mutL44 (17)2.4 (2.7)
dnaX985 mutS38 (14)2.1 (2.3)
dnaX985 mutH18 (7)2.0 (2.2)


The Pol III τ subunit is a determinant of the fidelity of DNA replication

We have uncovered a role of the dnaX gene products in controlling the fidelity of chromosomal DNA replication in E. coli. In particular, we have observed a mutator activity for an established dnaX mutant, dnaX36, carrying a specific defect in the τ subunit. We also succeeded in isolating a new set of dnaX mutator alleles based on the unusual properties of the dnaX36 strain. This new type of mutator has a unique specificity, enhancing preferentially transversions and frameshift mutations. That these novel mutators exert their effect by enhancing DNA replication errors is consistent with the intimate role that the τ subunit plays in the functioning of the replication fork, and is experimentally supported by our observations that dnaX enhancement of mutations is readily seen in mismatch repair-defective strains and that the mutator-induced mutations are subject to correction by the mutHLS mismatch repair system. Other data, not shown here, have indicated that the dnaX mutator effects are independent of a variety of DNA repair systems, such as the recA, umuDC, mutY and mutM systems, arguing further against the possibility that dnaX-induced mutations result from errors induced by DNA lesions. In the following, we interpret the dnaX mutator effects to indicate that the τ subunit has a specific error-prevention function, which is lost or impaired in the dnaX mutants.

A specific role of τ in preventing mutations: τ as a sensor for replication errors?

dnaX36 and dnaX983-989 carry mutations in the part of dnaX that is specific to τ, and they have the unique property of specifically enhancing transversion and frameshift mutations. This is an unusual combination not reported before for E. coli mutators. How does τ function to prevent these replication errors? It is unlikely that τ directly affects the insertion fidelity of the polymerase. Instead, most likely, the action of τ involves the processing of the mismatch once a misinsertion error has been made. How τ might affect this processing is unclear at this time. However, in Fig. 3, we propose a model as to how τ might be involved in the error-free processing of Pol III HE created misinsertion errors. This model is consistent with the current data as well as several other observations on the fidelity of Pol III HE described below. We propose that upon committing a misinsertion error, Pol III may be temporarily stalled. Extension of mispairs is generally difficult for many polymerases, particularly for more accurate, replicative enzymes (Beard and Wilson, 2003), and temporary stalling is likely to occur. We further propose that the stalling constitutes a signal that is sensed by the τ subunit, and τ will then initiate action to remedy the stalled state. The simplest way to achieve this is for HE to proofread the error and, in fact, τ may be necessary for efficient proofreading from stalled complexes (see below). Alternatively, τ may direct the mismatch to a proofreading-capable accessory polymerase, such as Pol I or Pol II, which has been proposed as one fidelity mechanism for certain replication errors (Fijalkowska et al., 1998). In either case, following removal of the mismatch, DNA synthesis can resume and no mutation will result. However, in dnaX36 and related mutants, τ's sensing ability is impaired, and HE may be forced to eventually extend the mismatch, yielding a mutation (Fig. 3). Alternatively, the τ defect may lead to abandonment of the growing point to an error-prone polymerase, such as Pol IV, leading to high probability of mutation. The precise mechanisms by which other DNA polymerases obtain access to the DNA growing point are of active current interest (Wagner et al., 2002; López de Saro et al., 2003a), and it should be considered that τ plays a role in the polymerase switching at the replication fork. In the following, we outline supportive evidence for our hypothesis of τ as a sensor for replication errors.

Figure 3.

τ as a sensor for misinsertion errors by Pol III HE. Diagram of the pathways by which DNA polymerase III and its τ subunit may process certain terminal mismatches within an in vivo replication fork. The model proposes that HE may enter a temporarily stalled state upon producing a misinsertion error that distorts the 3′ terminus. This stalled state is sensed by τ, which then mediates removal of the mismatch (top line). Mismatch removal could occur by facilitating the Pol III proofreading function, but may occur by different means. In the absence of the τ sensing function, as in dnaX36 or other dnaX mutants with a domain V defect (see text), the terminal mismatch is not properly removed, and a mutation may result (stippled pathways). Extension of the mismatch could lead to a base substitution (by direct extension) or frameshift (by misalignment extension). Extension could be by HE itself or by an accessory polymerase that may gain access upon dissociation of HE from the mismatch (see text). The circle denotes HE; the yellow circle, a stalled HE complex; the red square, an (error prone) accessory polymerase, such as Pol IV. This scheme is most pertinent for the processing of transversion mismatches, which have the largest tendency to produce a stalled complex. Direct, τ-independent proofreading of the mismatch without an intervening stalled complex may also occur (not indicated here). This may be a preferred pathway for the removal for the transition mismatches, which are generally less distorting.

τ as functional organizing centre of HE

A fidelity function of τ fits within the array of functions performed by τ as a structural and functional organizing centre of HE. Functions of τ recognized so far include: (i) binding to α subunits, effecting the physical coupling of the two (leading- and lagging-strand) polymerases; (ii) binding to the DnaB helicase, activating the helicase for high-speed fork movement; (iii) functionally tethering the χψ unit to the α subunit, permitting efficient synthesis of the lagging-strand in the presence of otherwise inhibitory SSB; (iv) protection of the leading-strand polymerase β clamp against unloading by the γ complex, ensuring high processivity in this strand; and (v) polymerase cycling in the lagging-strand by actively detaching the polymerase from the β clamp when the HE reaches the end of an Okazaki fragment, described as a DNA sensing function (Leu et al., 2003). Most or all of these functions depend on the α–τ interaction. The precise interaction has not been described in detail, but may be mediated by through multiple contacts that may, in fact, be different depending whether or not the HE is bound to DNA or not (Kim and McHenry, 1996a; López de Saro et al., 2003b). The residues in τ responsible for the τ–α interaction reside in DnaX domain V (residues 497–643), exactly where the amino-acid changes for dnaX36 and related mutants are located. Direct measurements of the binding interaction between α and τ have shown that the mutant τ proteins have a significantly impaired α–τ interaction (D. Gao and C.S. McHenry, pers. comm.). On the other hand, the generally robust health of dnaX36[it only ceases DNA replication around 43°C in salt-free medium (Henson et al., 1979)] makes it unlikely that leading- and lagging-strand replication would be permanently uncoupled, which would be the case if the mutant τ subunits were to fail to dimerize the cores. Thus, it is likely that in the τ mutants a subset of τ–α interactions is lost. This may impair functional communication between the two proteins when fork movement is compromised, such as at a stall site resulting from a misinsertion error.

Polymerase stalling

That polymerases are prone to stall, at least temporarily, at sites of misinsertion can be gathered from experiments indicating that following a misinsertion many DNA polymerases have difficulty with the subsequent extension step because the new primer terminus contains a mismatched base (Perrino and Loeb, 1989; Mendelman et al., 1990; Joyce et al., 1992). The possibility of stalling is likely to be particularly high for high-speed polymerases, such as Pol III, which are characterized by a ‘tight-fit’ catalytic site. These enzymes use careful geometric selection for correct base insertion and are inhospitable to aberrant primer structures (Ling et al., 2001; Beard et al., 2002; Beard and Wilson, 2003). Consistent with this, Pol III α subunit was shown incapable of synthesizing from A·A and G·A terminal mismatches (Kim and McHenry, 1996b). In the case of the proofreading-proficient HE, one would expect the mismatch to be readily removed exonucleolytically. However, in vitro fidelity assays using the M13 ss→RF DNA synthesis assay have shown little fidelity difference between proofreading-proficient and proofreading-deficient forms of HE (Pham et al., 1998; P.T. Pham et al. unpublished), suggesting that, at least under those conditions, the proofreading activity does not operate efficiently. Proofreading involves a conformational switch by which the primer terminus is relocated from the polymerase active site to the distant 3′ exonuclease site on the separate ɛ subunit, which may be kinetically limiting (Johnson, 1993). Thus, while proofreading-proficient Pol III actively removes a 3′ terminal base upon first binding to the primer end, a mismatch generated during ongoing synthesis may have a much longer life time, consistent with the production of stalled complexes even for proofreading-proficient Pol III (Johnson, 1993; Creighton and Goodman, 1995). One may speculate that τ is required for efficient occurrence of the conformational switch between the polymerase and exonuclease site. Data from our laboratory have indeed suggested the presence of a long-lived stalled mismatched intermediate in vitro (P.T. Pham and R.M. Schaaper, unpublished). Under conditions of a functional replication fork in vivo, this stalled state will be rapidly remedied, but a functionally operating τ subunit is likely required.

The fidelity and error specificity of Pol III HE in vitro

In vitro analysis of the fidelity of pol III using the M13 ss→RF system revealed an unusual pattern of DNA synthesis errors by both α subunit and HE: a high level of (−1) frameshifts with relatively few base substitutions (Mo and Schaaper, 1996; Pham et al., 1998). These data were proposed to be consistent with a model (Bebenek and Kunkel, 1990) in which (i) the polymerase has difficulty extending base·base mismatches directly – accounting for the relative paucity of base substitutions – but (ii) has an increased efficiency for continuing synthesis from a misaligned intermediate. Such intermediate can be created by forward slippage of the misinserted base on the next template base (if complementary), producing a (−1) frameshift, typically in a ‘non-run’ context (Bebenek and Kunkel, 1990). As an alternative, this kind of frameshift may be generated by a dNTP-mediated misalignment mechanism (Bloom et al., 1997). Importantly, the careful side-by-side comparison of the error rate of HE in vitro and in vivo suggested that additional mechanisms must operate in vivo to improve fidelity above the level seen in the ss→RF system. In particular, this improvement needs to abolish the frequent production of frameshifts (Pham et al., 1998; 1999). In view of the present data on the dnaX mutators, it is likely that this additional fidelity mechanism requires the action of τ within the context of a functional replication fork and is missing in dnaX36 and related mutators. This model predicts a similarity of replication errors between those observed in vitro and those in dnaX36 in vivo. This is definitely the case for the (−1) frameshifts, which occur specifically in non-run sequences in both cases. Further, a connection between the production of transversions and non-run frameshifts, both in vitro and in dnaX36 in vivo, can be readily made based on the consideration that transversion mismatches (purine·purine or pyrimidine·pyrimidine) are most distorting and most difficult to extend (Perrino and Loeb, 1989; Mendelman et al., 1990; Joyce et al., 1992). They are thus most likely to give rise to stalled complexes and to require help of the τ subunit for their resolution in a faithful manner. If not resolved, these mismatches are likely to lead to either a transversion or, in the appropriate sequence context, a non-run (−1) frameshift, as exemplified in Fig. 3. Thus, while our proposal for τ subunit acting as a sensor of mispairing errors is speculative, it is consistent with a larger body of observations regarding DNA polymerase III suggesting the existence of a such a fidelity pathway.

Experimental procedures

Media and strains

Solid and liquid media have been described (Fijalkowska and Schaaper, 1995; 1996). XPG plates used for papillation assays are minimal medium plates containing glucose (0.2%), phenyl-β-d-galactopyranoside (P-gal) (0.5 mg ml−1) and X-gal (40 µg ml−1) (Fijalkowska and Schaaper, 1996). All strains used are listed in Table 7. Strains were constructed by P1 transduction using P1virA and/or F′ crosses introducing F′prolac from strains CC101 through CC106 (Cupples and Miller, 1989) (for convenience we refer to this series of F′prolac as F′CC101 through F′CC106). The source of the dnaX36(TS) and dnaX2016(TS) alleles were strains GM36 and AX727 (obtained from the E. coli Genetic Stock Center, Yale University) respectively. In these strains, the dnaX alleles were linked up with the purE79::Tn10 marker from strain CAG12171 (Singer et al., 1989). The presence of the dnaX alleles was confirmed by testing for temperature sensitivity [lack of growth at 42°C on Luria–Bertani (LB) plates for dnaX2016 or at 44°C on LB plates lacking NaCl for dnaX36]. Linkage with purE79::Tn10 (∼2%) was then used to transfer the alleles to other backgrounds. In each case, the purE::Tn10 marker was removed by subsequent transduction to pur+. In this manner, transfer into NR9464 (ara, thi, trpE9777, mutL::Tn5, Δprolac) (Schaaper, 1993b) produced strains NR11920 (dnaX2016) and NR11921 (dnaX36); transfer into NR9559 (ara, thi, mutL::Tn5, Δprolac) (Fijalkowska and Schaaper, 1995) produced NR11926 (dnaX2016) and NR11928 (dnaX36). Strains NR11939 and NR11940 are NR9559, but carrying F′CC104 and F′CC105 respectively. Transposon zba-2321::mini-Tn10Cam is tightly (85–90%) linked to dnaX. It was obtained by transducing a dnaX2016 strain to chloramphenicol resistance at 42°C using a P1 lysate grown on a pool of bacteria carrying random insertions of mini-Tn10Cam derived from λNK1324 (Kleckner et al., 1991). λNK134 was obtained from the American Type Culture Collection (Rockville, MD). NR12425 (dnaX985) and NR12426 (dnaX36) were prepared by linking the dnaX alleles with zba-2321::mini-Tn10Cam followed by transfer into wild-type strain KA796 (ara, thi, Δprolac) (Schaaper et al., 1985). The source of the mutHLS alleles were ES1293 (mutL::Tn5) (Siegel et al., 1982), ES1301 (mutS::Tn5) (Siegel et al., 1982) and GM6470 (mutH::Tn5) (M. Marinus, University of Worchester). Plasmid pWTX-1 (dnaX+) was constructed by cloning the HindIII–PstI fragment of Kohara lambda clone 152 (12H5) (Kohara et al., 1987) containing the dnaX gene into the multicloning site of low-copy vector pMBL18 specifying ampicillin resistance (Nakano et al., 1995).

Table 7.  Strains used in this study.
StrainRelevant genotypeConstruction, source or reference
  1. The strains used for mutagenesis measurement are all derived from KA796. See Experimental procedures for details on strain constructions. F′CC101 to F ′CC106 indicate the F ′prolacIZ originally present in strains CC101 to CC106 (Cupples and Miller, 1989). Constructions by P1 transduction are indicated in the format ‘recipient × P1/donor’. Constructions by conjugation are indicated in the format ‘recipient × F ′ donor’.

Used for construction:
 AX727dnaX2016E. coli Genetic Stock Center
 CAG12171purE79::Tn10Singer et al. (1989)
 CC101…CC106F′prolac (F′CC101…106)Cupples and Miller (1989)
 ES1293mutL::Tn5Siegel et al. (1982)
 ES1301mutS::Tn5Siegel et al. (1982)
 GM36dnaX36E. coli Genetic Stock Center
 GM6470mutH::Tn5M. Marinus
 KA796ara, thi, ΔprolacSchaaper et al. (1985)
 NR9102F′prolac (F′128-27)Schaaper and Dunn (1991)
 NR11907dnaX2016, purE79::Tn10AX727 × P1/CAG12171
 NR11908dnaX36, purE79::Tn10GM36 × P1/CAG12171
Used for mutagenesis measurements:
 NR9464trpE9777, mutL::Tn5Schaaper (1993b)
 NR9559mutL::Tn5Fijalkowska and Schaaper (1995)
 NR10834F′CC104KA796 × F′CC104
 NR10835F′CC105KA796 × F′CC105
 NR11506sulA366, mutL::Tn5Fijalkowska et al. (1997)
 NR11911trpE9777, mutL::Tn5, dnaX2016, purE79::Tn10NR9464 × P1/NR11907
 NR11912trpE9777, mutL::Tn5, dnaX36, purE79::Tn10NR9464 × P1/NR11908
 NR11913mutL::Tn5, dnaX2016, purE79::Tn10NR9559 × P1/NR11907
 NR11914mutL::Tn5, dnaX36, purE79::Tn10NR9559 × P1/NR11908
 NR11918sulA366, mutL::Tn5, dnaX2016, purE79::Tn10NR11506 × P1/NR11907
 NR11919sulA366, mutL::Tn5, dnaX36, purE79::Tn10NR11506 × P1/NR11908
 NR11920trpE9777, mutL::Tn5, dnaX2016NR11911 × P1/KA796
 NR11921trpE9777, mutL::Tn5, dnaX36NR11912 × P1/KA796
 NR11926mutL::Tn5, dnaX2016NR11913 × P1/KA796
 NR11928mutL::Tn5, dnaX36NR11914 × P1/KA796
 NR11930sulA366, mutL::Tn5, dnaX2016NR11918 × P1/KA796
 NR11932sulA366, mutL::Tn5, dnaX36NR11919 × P1/KA796
 NR11939mutL::Tn5, F′CC104NR9559 × CC104
 NR11940mutL::Tn5, F′CC105NR9559 × CC105
 NR11965sulA366, mutL::Tn5, dnaX2016, F′CC104NR11930 × CC104
 NR11973sulA366, mutL::Tn5, dnaX36, F′CC104NR11932 × CC104
 NR12061sulA366, mutL::Tn5, F′CC104, zba-2321::mini-Tn10CamThis work
 NR12093mutL::Tn5, F′CC105, dnaX985, zba-2321::mini-Tn10CamThis work
 NR12159sulA366, mutL::Tn5, dnaX36, zba-2321::mini-Tn10Cam, F′CC104NR11973 × P1/NR12061
 NR12244sulA366, mutL::Tn5, dnaX983, zba-2321::mini-Tn10CamThis work
 NR12246sulA366, mutL::Tn5, dnaX985, zba-2321::mini-Tn10CamThis work
 NR12247sulA366, mutL::Tn5, dnaX986, zba-2321::mini-Tn10CamThis work
 NR12248sulA366, mutL::Tn5, dnaX987, zba-2321::mini-Tn10CamThis work
 NR12249sulA366, mutL::Tn5, dnaX988, zba-2321::mini-Tn10CamThis work
 NR12250sulA366, mutL::Tn5, dnaX989, zba-2321::mini-Tn10CamThis work
 NR12425dnaX985, zba-2321::mini-Tn10CamKA796 × P1/NR12093
 NR12426dnaX36, zba-2321::mini-Tn10CamKA796 × P1/NR12159
 NR12436dnaX985, zba-2321::mini-Tn10Cam, F′CC104NR12425 × CC104
 NR12437dnaX985, zba-2321::mini-Tn10Cam, F′CC105NR12425 × CC105
 NR12440dnaX36, zba-2321::mini-Tn10Cam, F′CC104NR12426 × CC104
 NR12441dnaX36, zba-2321::mini-Tn10Cam, F′CC105NR12426 × CC105
 NR12445dnaX985, zba-2321::mini-Tn10Cam, F′CC104, mutL::Tn5NR12436 × P1/ES1293
 NR12446dnaX985, zba-2321::mini-Tn10Cam, F′CC104, mutS::Tn5NR12436 × P1/ES1301
 NR12447dnaX985, zba-2321::mini-Tn10Cam, F′CC104, mutH::Tn5NR12436 × P1/GM6470
 NR12448dnaX985, zba-2321::mini-Tn10Cam, F′CC105, mutL::Tn5NR12437 × P1/ES1293
 NR12449dnaX985, zba-2321::mini-Tn10Cam, F′CC105, mutS::Tn5NR12437 × P1/ES1301
 NR12450dnaX985, zba-2321::mini-Tn10Cam, F′CC105, mutH::Tn5NR12437 × P1/GM6470
 NR12451dnaX36, zba-2321::mini-Tn10Cam, F′CC104, mutL::Tn5NR12440 × P1/ES1293
 NR12452dnaX36, zba-2321::mini-Tn10Cam, F′CC104, mutS::Tn5NR12440 × P1/ES1301
 NR12453dnaX36, zba-2321::mini-Tn10Cam, F′CC105, mutL::Tn5NR12441 × P1/ES1293
 NR12454dnaX36, zba-2321::mini-Tn10Cam, F′CC105, mutS::Tn5NR12441 × P1/ES1301
 NR12455dnaX36, zba-2321::mini-Tn10Cam, F′CC105, mutH::Tn5NR12441 × P1/GM6470
 NR12536trpE9777, mutL::Tn5, dnaX983, zba-2321::mini-Tn10CamThis work
 NR12537trpE9777, mutL::Tn5, dnaX985, zba-2321::mini-Tn10CamThis work
 NR12538trpE9777, mutL::Tn5, dnaX986, zba-2321::mini-Tn10CamThis work
 NR12539trpE9777, mutL::Tn5, dnaX987, zba-2321::mini-Tn10CamThis work
 NR12540trpE9777, mutL::Tn5, dnaX988, zba-2321::mini-Tn10CamThis work
 NR12541trpE9777, mutL::Tn5, dnaX989, zba-2321::mini-Tn10CamThis work
 NR12840mutL::Tn5, F′prolac (F′128-27)NR9559 × NR9102
 NR12841mutL::Tn5, dnaX36, F′prolac (F′128-27)NR11928 × NR9102

Mutant frequency determinations

Twelve or more colonies from each strain were toothpicked into 1 ml of LB medium and grown overnight to saturation with agitation at 30°C or 37°C. Colonies were taken from three or more independent isolates of each strain. Aliquots of undiluted cultures or appropriate dilution were plated on selective and non-selective plates at 30°C to determine mutants and total cell counts. RifR and NalR colonies were counted after 24 h incubation, trp+ and lac+ revertants after 48 h. Mutant frequencies were determined by dividing the median mutant count by the median total cell count. Statistical significance of mutant frequency differences of Table 1 were determined using the non-parametric Mann–Whitney criterion applied to the mutant frequency distributions for each strain using Prism 4 software (GraphPad, San Diego, CA).

lacI spectra

lacI mutant frequencies were determined from 15 independent cultures grown overnight in LB medium for strains NR12840 (mutL) and NR12841 (dnaX36 mutL). Aliquots were plated on P-gal medium to determine the total number of lacI mutants per cultures and aliquots of a 106 dilution were plated on MM glucose plates to determine the total number of cells. Mutant frequencies were 122 × 10−6 for the mutL strain and 750 × 10−6 for the dnaX36 mutL strain. Independent lacI mutants for sequencing were obtained as described (Schaaper and Dunn, 1991; Schaaper, 1993a). In brief, 10 independent overnight cultures for each of the two strains were diluted to about 103 cells ml−1 in LB medium. This solution was used to start 96 cultures (0.2 ml each) in wells of 96 well microtiter dishes (960 cultures total for each strain). The cultures were grown to saturation at 37°C on a shaking platform. A 1 µl aliquot of each culture was spread on a quarter section of P-gal plates, which were incubated at 37°C for 60 h. One mutant was picked randomly from each quarter section. These independent mutants (960 for each strain) were placed by toothpicking in a gridded pattern (96 per plate) on P-gal plates. The F′prolac of these mutants was then transferred by replica-mating into strain CSH52 for determination of the subclass of lacId mutants (Schaaper and Dunn, 1991) and strain S90C as a source for sequencing the identified lacI d mutations. The fraction of lacI d mutants, as determined in CSH52, was 0.43 for the mutL strain and 0.29 for the mutL dnaX36 strain. For sequencing, total DNA of the lacI d mutants (in strain S90C) was isolated by lysozyme/proteinase K treatment from 0.5 ml cultures. This DNA was polymerase chain reaction (PCR) amplified using two primers, 5′-GTGATTGGCTGTCTGAATCTGG-3′ and 5′-AGGCAGCTTCCACAGCAATGG-3′, yielding a 661 bp fragment that included the lacI region of interest (nucleotides 1–250). Sequencing was performed using an ABI Prism dRhodamine Terminator Cycle Sequencing Kit (Applied Biosystems) and manufacturer-recommended protocols. Out of 300 mutants for the mutL strain a total of 267 mutant sequences were identified; for the mutL dnaX36 strain, 241 mutant sequences were identified out of 280 sequenced. Based on these results, the adjusted lacI d mutant frequencies were calculated to be 46 × 10−6 and 188 × 10−6 respectively, as used in the calculations of Table 3.

Localized mutagenesis of the dnaX gene

Strain NR12061 (dnaX+, zba-2321::mini-Tn10Cam) was treated with 50 µg ml−1 of MNNG for 45 min as described by Miller (1992). A P1 lysate was prepared on the mutagenized culture, which was used to transduce strain NR11939 or NR11940 to CamR. In one set of experiments, transductants were obtained on LB-Cam plates, gridded on LB-Cam plates (100 colonies per plate) and replica-plated on XPG-Cam plates. On XPG plates (Fijalkowska and Schaaper, 1995; 1996) the mutation rate in growing colonies can be readily monitored based on the number of blue (lac+) papillae (Nghiem et al., 1988). In one other set of experiments, the transductants were plated directly on XPG-Cam plates to yield 200–500 transductants per plate. Typically, strains NR11939 and NR11940 produce 3–5 and 0–1 papillae per colony respectively, after 3–4 days at 37°C. Colonies with at least threefold more papillae than the control were selected as putative mutators. Backcrosses into NR11939 or NR11940 and complementation with plasmid pWTX-1 were used to confirm the location of the mutator mutation in the dnaX gene.

Sequencing of dnaX mutators

Mutations in the dnaX gene were identified by direct sequencing from genomic DNAs of mutator strains isolated using Easy-DNA kit (Invitrogen). The dnaX gene was amplified from the genomic DNA by PCR (24 cycles: 1 min 94°C, 1 min 47°C, 2 min 72°C) using a pair of 17-mer primers (PCR primer 1: 5′-TCACCAGCTTACTGGAA-3′; PCR primer 2: 5′-GACGTGCTGCGTCGTTG-3′). The resulting PCR products were used in a second series of PCR reactions (16 cycles: 1 min 94°C, 1 min 55°C, 2 min 72°C) with PCR primer 2 to generate ss DNA for sequencing. The products of the second PCR were purified by Bio-Gel P-30 chromatography columns (Bio-Rad) and concentrated using Microcon-30 microconcentrators (Amicon). Mutant and wild-type dnaX alleles were sequenced from the ss DNA PCR products using nine primers covering the entire gene. Sequencing was carried out by the dideoxy chain-termination method using Sequenase version 2.0 according to a protocol from Amersham Pharmacia Biotech.


We thank Mary Berlyn (E. coli Genetic Stock Center) and Martin Marinus for providing bacterial strains, Toshihiko Koga for plasmid pMBL18, Yuji Kohara for the Kohara lambda clones, and Ronnie Dunn of NIEHS for the pool of random mini-Tn10Cam insertions. We thank Tom Kunkel, Youri Pavlov and Charles McHenry for helpful comments on the manuscript for this paper.