DNA polymerase III from Escherichia coli cells expressing mutA mistranslator tRNA is error-prone.

Translational stress-induced mutagenesis (TSM) refers to the elevated mutagenesis observed in Escherichia coli cells in which mistranslation has been increased as a result of mutations in tRNA genes (such as mutA) or by exposure to streptomycin. TSM does not require lexA-regulated SOS functions but is suppressed in cells defective for homologous recombination genes. Crude cell-free extracts from TSM-induced E. coli strains express an error-prone DNA polymerase. To determine whether DNA polymerase III is involved in the TSM phenotype, we first asked if the phenotype is expressed in cells defective for all four of the non-replicative DNA polymerases, namely polymerase I, II, IV, and V. By using a colony papillation assay based on the reversion of a lacZ mutant, we show that the TSM phenotype is expressed in such cells. Second, we asked if pol III from TSM-induced cells is error-prone. By purifying DNA polymerase III* from TSM-induced and control cells, and by testing its fidelity on templates bearing 3,N(4)-ethenocytosine (a mutagenic DNA lesion), as well as on undamaged DNA templates, we show here that polymerase III* purified from mutA cells is error-prone as compared with that from control cells. These findings suggest that DNA polymerase III is modified in TSM-induced cells.

Autonomous organisms normally replicate its DNA accurately, but the fidelity of replication can be transiently decreased in response to environmental and physiological stimuli through a number of pathways (1). Although the Escherichia coli SOS response represents the best-described transient mutator response (2), emerging evidence indicates the existence of multiple inducible mutagenic pathways in E. coli (3). One especially intriguing pathway is provoked by increased translational errors resulting from mutations in tRNA genes (4,5), in genes specifying tRNA-modification enzymes (6), or from exposure to streptomycin, an antibiotic that promotes mistranslation (7). This pathway, dubbed translational stress-induced mutagenesis (TSM) 1 (3), does not require the induction of lexA/ recA-regulated SOS genes and is suppressed in cells defective for RecABC/RuvABC-dependent homologous recombination (5,(7)(8)(9). Available genetic (5,(7)(8)(9) and biochemical (10) evidence suggests that the TSM phenotype results from error-prone DNA replication rather than from defective DNA repair. On the basis of the effect of elimination of individual genes encoding "non-replicative" polymerases, we previously proposed that either DNA polymerase III or an unidentified new DNA polymerase is responsible for error-prone replication in TSM-induced cells.
In addition to pol III, four additional DNA polymerases (polymerases I, II, IV, and V) are known in E. coli. Even though these polymerases carry out important cellular functions, genes encoding each of these polymerases can be mutationally inactivated, implying that considerable functional redundancy is built into the replication apparatus. DNA polymerase I (pol I; encoded by polA), known as a "repair polymerase," normally functions to fill gaps that arise during lagging strand replication and during excision repair. The remaining three DNA polymerases, namely II (pol II; encoded by polB (25,26)), IV (pol IV; encoded by dinB (27)), and V (pol V; encoded by umuDC (28,29)), are induced as a part of the SOS system. pol II has been proposed to play a role in replication-restart following DNA damage, a process that bypasses DNA damage in both an error-free (30,31) and error-prone manner (32,33). pol IV is involved in certain types of untargeted SOS mutagenesis (27,34,35), whereas pol V, working in conjunction with a number of other factors including the RecA protein, is believed to be responsible for translesion DNA synthesis (28,29,36). How-ever, pol V may also be responsible for untargeted mutations at undamaged template sites (37,38).
The individual loss of polA, polB, dinB, or umuDC genes does not affect the expression of the TSM response (5,8,10). However, an analysis based on cells with defects in individual genes leaves open the possibility that two or more of these four nonessential DNA polymerases may have nonexclusive (redundant) roles in error-prone DNA synthesis. Here we show that a strain simultaneously defective for pol I, pol II, pol IV, and pol V can be constructed, proving that loss of all four non-replicative polymerases is compatible with viability. Analysis of the TSM response in this strain confirms that the four non-replicative polymerases are not required collectively or individually for the TSM response. To address the question directly whether pol III HE or an unknown 6th DNA polymerase is responsible for mistranslation-induced mutagenesis, we purified pol III* from TSM-induced and uninduced cells and analyzed its replication fidelity on both damaged and undamaged template DNA. Our data show that purified pol III* from TSM-induced cells shows elevated mutagenesis on undamaged DNA as well as at a site-specific mutagenic lesion.

MATERIALS AND METHODS
Reagents, Enzymes, and Replication Proteins-Taq DNA polymerase was from Roche Molecular Biochemicals; restriction endonuclease BglI, T4 DNA ligase, Vent DNA polymerase, and T4 polynucleotide kinase were from New England Biolabs. The ␤ subunit was purified as described by Johanson et al. (39). The primase DnaG was purified as described by Marians (40). The E. coli single-stranded DNA-binding protein (SSB) was purified according to Minden and Marians (41).
Plasmid and Strain Construction-Plasmid pMV22 was constructed from plasmid pHM22 (5) as follows. The SphI-HindIII 1604-bp fragment containing the lacI q gene and the wild type glyV tRNA gene under P trc promoter was inserted into the low copy number plasmid pMW119 (42) cut with SphI and HindIII. Similarly, plasmid pMV11 was constructed by inserting the SphI-HindIII 1604-bp fragment containing the mutant glyV tRNA gene (mutA) from pHM11 (5) into the SphI-HindIII sites of pMW119.
To enable the subsequent transduction of the polA1 allele in strain HSM83 (see Table I for genotypes), the metE::Tn10 marker from strain CAG18491 (43) was first transferred to HMS83 (44) by P1 transduction. The transductants were selected on LB agar containing 30 g/ml tetracycline (Sigma). One isolate, named AM132, was characterized by its inability to grow in minimal medium without methionine. E. coli strains AM134 and AM135 were constructed by co-transducing the polA1 allele along with the marker metE::Tn10 from strain AM132 into strains CC105 and CC105mutA, respectively. Strains AM134 and AM135 were characterized for the presence of polA1 allele by testing their sensitivity to 0.04% methyl methanesulfonate in LB plates and to UV irradiation at 7 J/m 2 as described previously (45). The E. coli strain AM130 was created by P1 transduction of the ⌬(umuDC)595::Cm r allele from strain RW82 to strain AM107 (CC105 polB (10)). The presence of the ⌬(umuDC)595 allele was confirmed by UV sensitivity at 30 J/m 2 . The E. coli strain AM146 was constructed by transferring the ⌬dinB::Kan r allele from strain YG7207 to strain AM130 by P1 transduction. The presence of the ⌬dinB::Kan r allele in AM146 was confirmed by PCR amplification (forward primer 5Ј-CGCTGTATCAATAC-TTTGGTCA; reverse primer 5Ј-AGGCGAATAAGTTTTGTTTTGA) followed by analysis of restriction digestion patterns of the PCR products. The E. coli strain AM147 was made by co-transducing the polA1 allele along with the marker metE::Tn10 from strain AM132 to AM146 by P1 transduction. Strain AM147 was characterized by sensitivity to 0.04% methyl methanesulfonate and to UV irradiation at 7 J/m 2 .
The FЈ factors contained in strains CC104 and CC107 (two strains related to CC105, the parental strain for AM147) contain a second copy of the dinB gene (46,47). To show that AM147 does not have a second (wild type) copy of dinB, we carried out a PCR-based analysis as summarized in Fig. 1. This analysis shows that an "external" primer set (F1/R1 in Fig. 1) yields a single band corresponding to a disrupted version of dinB in AM147 cells (lane 2) and that two "internal" primer sets (F2/R2 and F3/R3) do not yield a band (lanes 4 and 6), indicating the loss of sequences corresponding to dinB. To verify that we could have detected a second copy of dinB, we repeated the analysis with AM155 (CC104 ⌬dinB::Kan r ), a strain harboring an FЈ factor that was previously shown to have a second, episomal copy of dinB. Our analysis shows that the F1/R1 primer set yields two bands (lane 8), one corresponding to a disrupted allele (1862 bp) and the other corresponding to the wild type allele (1603 bp). As expected, internal primer sets F2/R2 and F3/R3 amplify the second copy of dinB in AM155 cells. Thus, these data confirm that AM147 does not have an undisrupted dinB allele.
Construction of Site-specific ⑀C Lesion-bearing ssDNA Constructs-M13 single-stranded DNA (ssDNA) molecules bearing an ⑀C residue were constructed following the procedures described elsewhere (49 -51) Fidelity of pol III from E. coli mutA Cells and summarized below. M13mp7L2 ssDNA was linearized by cutting with the restriction endonuclease EcoRI that cuts a hairpin DNA structure within the polycloning site of the vector. The linearized DNA was annealed to a 57-nt "scaffold" and a 5Ј-phosphorylated 17-nt insert containing a single site-specific ⑀C lesion. Annealing of the 57-mer draws the two ends of the linear M13 ssDNA together to form a noncovalently closed circular DNA with a 17-nt "gap" complementary to the lesion-containing 17-mer. The annealed DNA is subjected to DNA ligation to generate a covalently closed ssDNA circle containing the lesionbearing 17-nt insert. After the ligation step, the scaffold is removed by heat denaturation in the presence of a 10-fold molar excess of an "anti-scaffold" 57-mer (i.e. a 57-mer with a sequence complementary to that of the scaffold). The constructed circular ssDNA was purified by using a QIAquick gel extraction kit column (Qiagen).
To prime in vitro DNA synthesis, a 60-mer was annealed to the ssDNA (5Ј-TAACCAATAGGAACGCCATCAAAAATAATTCGCGTCTG-GCCTTCCTGTAGCCAGCTTTCA-3Ј; complementary to the M13 sequence from 6628 to 6687; Fig. 4A) as follows. One pmol of the ssDNA construct (with or without an ⑀C lesion) was mixed with 1.5 pmol of a 60-mer in 50 mM Tris-Cl (pH 7.8), 10 mM MgCl 2 , and 20 mM dithiothreitol (DTT), heated at 85°C for 5 min, and allowed to cool over 2 h to 30°C. The terms ⑀C-ssDNA and C-ssDNA, respectively, refer to the primed ssDNA bearing a site-specific ⑀C lesion or normal cytosine (as a control).
Purification of pol III*-pol III* was purified following the procedures described previously (23), as summarized below. Cells were grown in 300 liters of LB medium supplemented with ampicillin (50 g/ml) and IPTG (1 mM) as appropriate, and cells were harvested by centrifugation at 4°C in a Sharples centrifuge. The cell pellet was resuspended in an equal volume of 50 mM Tris-Cl (pH 7.5), 10% sucrose solution, quick-frozen in liquid nitrogen, and stored at Ϫ80°C.
The cell suspension (2 kg) was thawed on ice, and the A 595 was adjusted to 200 with 50 mM Tris-Cl (pH 7.5), 10% sucrose solution. One-ninth volume of lysis buffer (1.5 M NaCl, 0.2 M EDTA, 0.2 M spermidine, and 50 mM DTT) was added, and the pH was adjusted to 8.5 by adding solid Tris base. Lysozyme (10 mg/ml in water) was added to 0.2 mg/ml, and the contents were incubated for 30 min on ice, followed by 10 min in a 37°C water bath. The lysed cells were distributed into 250-ml Sorvall GSA bottles, kept on ice for 5 min, and centrifuged for 60 min at 13,000 rpm in a Sorvall SLA-1500 rotor. The supernatant (fraction 1a) was collected by gentle decantation. To precipitate nucleic acid, polymin P (1% v/v; Amersham Biosciences) was added slowly, with gentle stirring, to fraction 1a to a final concentration of 0.06%. Stirring was continued for an additional 30 min at 4°C. The suspension was centrifuged at 13,000 rpm for 30 min at 4°C in a Sorvall SLA-1500 rotor to collect the supernatant (fraction 1b).
Ammonium sulfate (0.25 g/ml) was added to fraction 1b over a 10-min interval while stirring. The contents were stirred for an additional 30 min at 4°C, transferred into Sorvall GSA bottles, and centrifuged at 27,000 ϫ g in a Sorvall SLA-1500 rotor for 45 min at 4°C. The precipitate was resuspended in 0.2 TBEB (1/10th of the fraction 1a volume), and the insoluble fraction was collected by centrifugation as above. This procedure was repeated with 0.17 TBEB (1/40th of the fraction 1a volume), and the final recovered insoluble fraction was dissolved in a small volume of TBEB (fraction II). Fraction II was quick-frozen in liquid nitrogen and stored at Ϫ80°C.
Fraction II was thawed and dialyzed overnight against 4 liters of buffer A ϩ 40 mM NaCl at 4°C. The dialysate was clarified by centrifugation for 20 min at 48,200 ϫ g at 4°C. The clarified solution was diluted with buffer A to the conductivity of buffer A ϩ 50 mM NaCl. The diluted fraction II was applied to a heparin-agarose (Sigma) column (1 ml of resin per 10 mg of protein) equilibrated with buffer A ϩ 50 mM NaCl. The column was washed with 2 column volumes of the equilibration buffer, and the activity was eluted with a 10-column volume of NaCl gradient (50 -400 mM) in buffer A. Fractions were pooled, and ammonium sulfate was added to 0.26 g/ml. The suspension was stirred for 2 h at 4°C and centrifuged at 37,000 rpm for 60 min at 4°C in a Sorvall A841 rotor. The precipitate was resuspended in a small volume (300 l) of buffer B (fraction III).
Fraction III was clarified by centrifugation for 5 min at 13,000 rpm in a microcentrifuge. The clarified fraction III was gel-filtered through a fast protein liquid chromatography Superose 6 column (HR 10/30; Amersham Biosciences) equilibrated with buffer B at a flow rate of 0.1 ml/min. Peak fractions were pooled, and glycerol was added to achieve a final concentration of 38% (fraction IV) and stored in aliquots at Ϫ80°C.
pol III* Assay-The assay mix (25 l) containing 50 mM HEPES-KOH (pH 8.0), 50 mM potassium glutamate, 10 mM magnesium acetate, 10 mM DTT, 10 g/ml rifampicin, 100 g/ml bovine serum albumin (BSA), 24 g/ml SSB, 3.2 g/ml M13 Gori ssDNA, 100 M each of CTP, GTP, and UTP, 80 M each of dGTP, dCTP, dATP, and [ 3 H]dTTP (200 -400 cpm/pmol), and 1 mM ATP was pre-warmed for 5 min at 30°C. Twenty ng of DnaG was added, and the contents were incubated for 5 min at 30°C. Thirty ng of the ␤ subunit were added, and the reactions were started by the addition of pol III*. If necessary, the proteins were diluted with 50 mM Tris-Cl (pH 7.5), 150 mM NaCl, 5 mM EDTA, 10 mM DTT, 500 g/ml BSA, and 20% glycerol. After incubation for 5 min at 30°C, the reaction was stopped by chilling and addition of 2 drops of 0.2 M sodium pyrophosphate, 2 drops of calf thymus DNA (1 mg/ml; Sigma), and 4 ml of 5% trichloroacetic acid. Acid-insoluble radioactivity was then measured. One unit of activity is the amount Fidelity of pol III from E. coli mutA Cells catalyzing the incorporation of 1 nmol of deoxyribonucleotide in 30 min at 30°C.
Replication of DNA-primed M13 ssDNA by Purified Proteins-Replication of M13 ssDNA primed with an annealed 60-mer by purified proteins was carried out following the procedures of Geider and Kornberg (53) and of Livneh (54) as follows. The standard reaction mixture (25 l) included 20 mM Tris-Cl (pH 7.5), 4% glycerol, 0.1 mM EDTA, 8 mM MgCl 2 , 8 mM DTT, 50 mM potassium glutamate, 80 g/ml BSA, 10 g/ml rifampicin, 1 mM ATP, 80 M each of the dNTPs, 20 fmol 60-mer primed C-ssDNA or ⑀C-ssDNA, 32 g/ml of SSB, 20 ng of ␤ subunit, and 4 units of pol III*. Reactions were allowed to proceed for 20 min at 37°C and were stopped by adding EDTA to 5 mM, Tris-Cl (pH 8.0) to 10 mM, and SDS to 0.5% (w/v), followed by heating at 65°C for 10 min. Proteinase K was added to a final concentration of 50 g/ml, and the mixture was incubated at 55°C for 30 min. Then the reaction mix was subjected to phenol extraction, followed by standard ethanol precipitation. The precipitated DNA was dissolved in a minimum volume of H 2 O, and an aliquot was checked by electrophoresis on an agarose gel (0.8%). Ethidium bromide staining revealed that about half of the input ssDNA (with or without ⑀C) was converted to replicative form II (RF II) DNA on the basis of electrophoretic mobility.
Analysis of in Vitro Mutation Fixation at ⑀C-Replication reactions were carried out as described above, and the newly synthesized minus strand was selectively amplified by ligation-mediated PCR (LM-PCR) following the procedures described in detail by Al Mamun et al. (10,55). The resulting gel-purified 309-bp DNA fragment was subjected to multiplex sequencing reactions as described by Al Mamun et al. products were electrophoresed on 16% polyacrylamide, 8 M urea gels, and the autoradiograms were analyzed by densitometry as described previously (49).
Construction of Gapped M13mRS65 DNA-Gapped DNA was prepared following the procedures described elsewhere (56,57) with some modifications. Double-stranded form I (RF-I) DNA of phage M13mRS65 was isolated using a Qiagen plasmid mini kit following the procedures described by the vendor. The DNA was digested with the restriction endonucleases MluI and SphI, which cut at positions Ϫ60 and ϩ379 of the lacI gene, respectively, producing two fragments of size 439 and 8040 bp. The 8040-bp fragment was purified from 0.8% TAE (40 mM Tris acetate, 1 mM EDTA (pH 8))-agarose gels using the QIAquick gel extraction kit (Qiagen) by procedures described by the vendor. The purified 8040-bp fragment (1850 fmol) was mixed with M13mRS65 ssDNA (1250 fmol) in the presence of 25 mM Tris-Cl (pH 7.5), 10 mM MgCl 2 , and 10 mM DTT, heat-denatured at 90°C for 2 min, cooled in an ice bath, incubated at 70°C for 5 min, and allowed to slow-cool to room temperature over several hours by switching off the water bath.
Gap-filling DNA Synthesis and Mutation Frequency Determination-Gap-filling DNA synthesis was performed following the procedures as described previously (56,57). The 25-l reaction mixture contained 30 mM HEPES-KOH (pH 7.6), 10 mM MgCl 2 , 8 mM DTT, 100 g/ml BSA, 200 M of each dNTP, 1 mM ATP, 20 nM ␤ subunit, 50 fmol of gapped DNA, and 5-15 units of DNA polymerase III* (pol III*). Where indicated, replication products were radiolabeled by including 7 Ci of [␣-32 P]dATP (10 mCi/ml; 6000 Ci/mmol; Amersham Biosciences). The reaction mix was incubated at 37°C for 10 min and quenched by adding EDTA to 15 mM. To assess gap-filling, 10 l of each reaction were mixed with 2.5 l of SDS/dye mix (20 mM Tris-Cl (pH 7.5), 5 mM EDTA, 5% SDS, 0.5% bromphenol blue, 25% glycerol) and subjected to electrophoresis in a 0.8% agarose gel in TAE buffer containing 0.5 g/ml ethidium bromide at constant voltage (70 V) for 16 h. Bands were visualized by illumination with UV light and photographed. The gel was dried and exposed to Kodak XRP film to produce autoradiograms.
To analyze mutagenesis, replicated DNA contained in the 25-l reaction mix (described above) was extracted with phenol-chloroform, precipitated with ethanol, and resuspended in 12 l of H 2 O. An aliquot was checked by electrophoresis to confirm the completeness of gapfilling. 1-2-l aliquots were used to transfect competent MC1061 cells by electroporation as described by Al Mamun et al. (55). Immediately following electroporation, 1 ml of cold SOC medium (0.5% yeast extract, 2% tryptone, 10 mM NaCl, 2.5 mM KCl, 10 mM MgCl 2 , 10 mM MgSO 4 , and 20 mM glucose) was added. An appropriate volume (50 -100 l) of the transfection mix, together with 3 ml of soft agar containing 2.5 mg of X-gal (Sigma) and 0.3 ml of indicator culture (E. coli NR9099), were mixed and the contents plated on minimal agar. The plates were incubated at 37°C for 24 h, and the lacI mutants were scored as light and dark blue plaques. The mutation frequency was determined by dividing the number of mutant plaques by the total number of plaques.
Determination of Percent Translesion DNA Synthesis-Translesion DNA synthesis was determined following the procedure as described by Pham et al. (36), as summarized below. The reaction mix (10 l) contained 20 mM Tris-Cl (pH 7.5), 5 mM DTT, 8 mM MgCl 2 , 0.1 mM EDTA, 25 mM potassium glutamate, 4% glycerol, 40 g/ml BSA, 1 mM ATP, 500 M dNTPs, 100 nM SSB (as tetramer), 40 nM ␤ (as dimer), 50 nM primer/template (see below), and 0.5 to 1 unit of pol III*. The reaction mix was incubated at 37°C for 10 min and terminated by adding an equal volume of 20 mM EDTA in 95% formamide. The reaction products were heat-denatured and separated on a 12% denaturing polyacrylamide gel, and band intensities were measured by PhosphorImaging using ImageQuant software (Amersham Biosciences). Percent translesion synthesis values were calculated as the number of primers extended to the lesion plus the number of primers extended past the lesion, divided by the total number of primers.

The TSM Phenotype Is Expressed in an E. coli mutA Strain
Simultaneously Deficient for pol I, pol II, pol IV, and pol V-Increased mistranslation results in the expression of an errorprone DNA polymerase that is demonstrable in cell-free extracts. This activity is detected as increased in vitro mutation fixation at a site-specific mutagenic lesion (3,N 4 -ethenocytosine; ⑀C) borne on phage M13 ssDNA (10). In this system, primed ssDNA bearing an ⑀C residue (⑀C-ssDNA) is replicated with cell-free extracts from mutA cells, and the newly synthesized complementary (minus) strand is selectively amplified using ligation-mediated PCR (LM-PCR). The amplified products are subjected to multiplex sequencing to characterize mutagenesis at the ⑀C lesion. Interestingly, the error-prone polymerase activity is undetectable in cell-free extracts prepared from recA mutA cells (10), consistent with the observations that expression of the TSM phenotype requires a functional recA gene (5,8).
We have shown previously, through a series of in vivo experiments, that the TSM response does not require the individual function of pol I, pol II, pol IV, or pol V (5,8,10). This observation, together with the crude extract results considered above, argued that the TSM phenotype is mediated by pol III. However, the available data did not exclude the possibility of a redundant function for the "non-essential" DNA polymerases such that one of the remaining three were able to compensate for the loss of any one polymerase. To address this possibility, we constructed a strain (AM147) that is defective for all four nonessential DNA polymerases (pol I, pol II, pol IV, and pol V), and we asked if the TSM phenotype, detected as increased lacZ Ϫ 3 lacZ ϩ papillation (58), is still displayed in such a strain in response to the expression of the mutA tRNA gene. When E. coli AM147 is transformed with pMV11, a plasmid bearing mutA, a strong mutator phenotype (increased papillation) is observed in the presence of IPTG, a condition in which the mutA allele is expressed (Fig. 2). The intensity of the papillation is comparable with that in strain CC105/pMV11, the positive control with wild type polA, polB, dinB, and Fidelity of pol III from E. coli mutA Cells umuDC alleles. Increased papillation is not observed for strains CC105/pMV22 (plasmid-borne wild type glyV) and AM147/pMV22 (plasmid-borne wild type glyV) in the presence of IPTG. These observations rule out a requirement for the individual or collective function of polA, polB, dinB, and umuDC for the expression of the TSM phenotype. It may be noted that a second copy of dinB has been found on the FЈ factor of some E. coli strains (46,47). We have tested for a second dinB allele in AM147 by PCR amplification using both an external primer set and two internal primer sets, and we confirmed that the strain did not have a second dinB allele (see "Materials and Methods"; Fig. 1).
Purification and Characterization of pol III* from Mutator and Control Strains-To address the question whether DNA polymerase III is error-prone in TSM-induced E. coli cells, we purified pol III* from two pairs of strains. The first pair consisted of strains AM134 (polA; non-mutator control) and AM135 (polA mutA; mutator); the second set consisted of strains AM147/pMV22 (polA polB dinB umuDC; non-mutator control) and AM147/pMV11 (polA polB dinB umuDC mutA; mutator). Purification was carried out as described under "Materials and Methods." Table II summarizes some characteristics of the four purified enzymes and shows that all four enzymes had similar specific polymerase activities by the M13Gori polymerization assay (59). Fig. 3 shows that the polymerization activity, based on an assay in which a 60-mer primer annealed to M13 ssDNA is elongated, is somewhat higher for pol III* from non-mutator strains as compared with that from mutator strains. Because polymerization activity for the four polymerase preparations is essentially similar using the DnaG-primed M13Gori ssDNA assay (Table II), it is not clear why the 60-mer primer elongation assay shows a slight reduction in polymerization by pol III* from mutator strains as compared with those from nonmutator strains.
Mutation Fixation at an ⑀C Lesion by Purified pol III*-We have shown previously (5) that mutation fixation at an ⑀C residue borne on transfected M13 ssDNA is significantly increased in mutA cells. Subsequently we showed that in vitro translesion DNA synthesis across a site-specific ⑀C residue by crude extracts prepared from mutA cells is error-prone (10). To determine whether translesion synthesis across ⑀C by purified pol III* from mutA strains is also error-prone, we used a modification of the previously described in vitro replication-in vitro mutation detection system (10,55). In the modified assay system for purified pol III, we created a primed template (⑀C-ssDNA) consisting of M13 ssDNA bearing a site-specific lesion to which a 60-mer primer is annealed so that the 3Ј-OH terminus is situated 388 nt upstream of the lesion site. The primer is elongated by the DNA polymerase, followed by selective amplification of the newly synthesized complementary strand by LM-PCR and subsequent multiplex sequencing analysis as outlined in Fig. 4. In the multiplex sequence analysis strategy, the specificity of base insertion opposite the lesion is determined by elongation of a labeled sequencing primer in the presence dGTP, dCTP, and ddATP. Correct insertion of a guanine opposite ⑀C yields a 23-mer limit elongation product, whereas insertion of a T (i.e. a C 3 A transversion) or an A (C 3 T transition) opposite ⑀C yields, respectively, 21-and 22-nt-long products. Densitometric analysis of the relative signal intensity of each band is used to calculate the frequency of each type of mutation as described previously. Fig. 4C shows that when a control template-primer in which normal cytosine replaces ⑀C (C-ssDNA) is replicated by pol III* from non-mutator strains (lanes 1 and 3) or from mutator strains (lanes 2 and 4), as expected, almost all of the signal is contained in the 23-nt band (i.e. no detectable 22-and 21-nt bands are observed). Lane 5 shows that ⑀C-ssDNA, when replicated by pol III* from non-mutator strain AM134, displays detectable mutagenesis. Replication of ⑀C-ssDNA by pol III* from mutator strain AM135 (lane 6) results in detectably increased mutagenesis, as indicated by the increased intensity of the bands corresponding to C 3 A (21 nt) and C 3 T (22 nt) events. Table III provides a quantitative summary of the results obtained from at least three independent replication for each of the pol III*, followed by mutation analysis. Translesion synthesis by pol III* from AM134 results in 17% mutagenesis, whereas translesion synthesis by the polymerase from AM135 results in 34% mutagenesis (Table III). A similar pattern of mutation is observed when ⑀C-ssDNA is replicated with pol III* from non-mutator strain AM147/pMV22 (Fig. 4C, lane 7) or from the mutator strain AM147/pMV11 (lane 8) (9.8 versus 26%; Table III). In vivo mutagenesis at an ⑀C lesion ranges from 4 to 13% for wild type strains and 40 to 60% in mutA cells (5,8,10). The lower fold elevation in mutagenesis by pol III* from mutator strains versus non-mutator strains (2-3-fold; Table III) is considered under "Discussion." To address the question whether the elevation of mutagenesis at ⑀C lesion is mediated by elevated translesion DNA synthesis by pol III from mutator strains, we used a templateprimer consisting of a 120-nt-long oligonucleotide bearing a site-specific ⑀C lesion (or normal C in controls) at nucleotide position 51 annealed to a 5Ј-end-labeled 30-mer primer whose 3Ј-OH terminus is 14 nt downstream of the lesion site (Fig. 5). The primer was elongated by pol III following the procedures described under "Materials and Methods," and the products were analyzed by high resolution gel electrophoresis. Fig. 5 shows that all of the tested pol III preparations were capable of significant levels of translesion DNA synthesis across ⑀C. Table  IV presents a quantitative analysis, based on densitometric analyses of the type of data represented by Fig. 5, and confirms that there are no significant differences in translesion DNA synthesis by pol III preparations from control and mutator strains. Therefore, elevated mutagenesis at ⑀C lesion by pol III FIG. 2. Colony papillation assay for mutator activity. Cells were streaked on papillation agar medium containing IPTG (see "Materials and Methods") and were incubated for 5 days at 37°C to allow the development of papillae. The plate shows that strains bearing the pMV11 (mutA) plasmid (from left, 2nd, 4th, and 5th streaks) show dramatically increased papillation, in contrast to the background level papillation observed for strains carrying pMV22 (glyV ϩ control; 1st and 3rd streaks). a One unit is defined as 1 nmol of nucleotide incorporation in 30 min at 30°C using the procedures described under "Materials and Methods."

Fidelity of pol III from E. coli mutA Cells
from mutator strains appears to result from increased errors rather than from increased translesion DNA synthesis.
Fidelity of pol III* from Mutator and Control Strains during in Vitro Gap-filling DNA Synthesis on Undamaged Templates-M13mRS65 phage is a derivative of M13 phage harboring, in addition to the lacZ␣ gene, the entire adjacent lacI gene (M13lacI ϩ Z␣ ϩ ) (60,61). Because phage M13mRS65 carries the lacZ␣ gene, it is capable of ␣-complementation on a lacZ⌬M15 host such as E. coli NR9099. However, because it also carries the lacI gene encoding the lac repressor, ␣-complementation depends on the state of the lacI gene. On plates containing X-gal, M13mRS65 (lacI ϩ ) will produce colorless plaques, whereas lacI mutants generated by DNA synthesis errors will produce blue plaques. Thus, this phage provides a forward mutational target where blue mutants can be readily observed against a background of colorless plaques. Fig. 6A outlines an in vitro assay system based on this phage. The assay consists of creating a 439-nt-long gap in the lacI region (Ϫ60 to ϩ379) following the procedures described under "Materials and Methods." The lacI gap is filled by DNA synthesis catalyzed by pol III* from non-mutator and mutator strains, and the resulting products are transfected into an appropriate host for analysis of mutagenesis that occurred during in vitro DNA synthesis. Fig. 6B (ethidium bromide fluorescence) and Fig. 6C (autoradiography) show that the mobility of the gapped DNA changes to that of replicative form II (RF-II) DNA after a 10-min gap-filling reaction by pol III. Table V  . After incubation at 37°C for 10 min (A) or for the indicated time (B), the reaction was stopped by chilling, and the addition of 2 drops of 0.2 M sodium pyrophosphate, 2 drops of calf thymus DNA (1 mg/ml), and 4 ml of 5% trichloroacetic acid. The resulting mixture was passed through a glass fiber filter (Enzo), and acid-insoluble radioactivity was measured following the procedure described by Wickner and Kornberg (68).

FIG. 4. Characterization of in vitro mutagenesis at an ⑀C residue by purified pol III preparations.
A, schematic representation of M13mp7L2 genome showing the position of the ⑀C lesion (or C in control constructs; position X at nucleotide 6240), the BglI restriction endonuclease site (at nucleotide 6402), and the 60-mer primer-annealing site (nt 6628 -6687). DNA synthesis on the primed template was carried out as described under "Materials and Methods," and a 319-nt sequence of the newly synthesized strand was selectively amplified by ligationmediated PCR as described in detail elsewhere (10) and summarized briefly under "Materials and Methods." B, summary of multiplex sequence analysis procedures for a 319-bp double-stranded DNA fragment derived by selective amplification of the newly synthesized strand as above. A 5Ј-32 P-labeled 19-nt-long primer (50) was mixed with the 309-bp DNA fragment and subjected to 20 cycles of (linear) elongation by Taq DNA polymerase (Taq pol) in the presence of dGTP, dCTP, and ddATP as described under "Materials and Methods." Under the described conditions, primer elongation on wild type (WT) and mutant templates results in dead-end products of characteristic lengths: 23-mer for wild type, 22-mer for C 3 T mutation, and 21-mer for C 3 A mutation. The elongation products were fractionated by high resolution gel electrophoresis and quantitated by computing densitometry. C, composite autoradiogram showing in vitro mutation fixation at ⑀C lesions (lanes 5-8) and, as controls, at normal cytosine (lanes 1-4). In vitro synthesized complementary (minus) strand was selectively amplified by LM-PCR followed by PCR, and the double-stranded DNA products were subjected to multiplex sequence analysis as outlined in B above. The resulting labeled elongation products were fractionated on 8% polyacrylamide, 8 M urea gels followed by autoradiography. Lanes 1-4, C-ssDNA (control, C) replicated with 4 units of pol III* from the indicated strains. Lanes 5-8, ⑀C-ssDNA replicated with 4 units of pol III* from the indicated strains. We have previously demonstrated, by reconstruction experiments, that the in vitro mutation analysis system described here measures mutation frequency and specificity with reasonable accuracy (10).
Fidelity of pol III from E. coli mutA Cells mutation frequencies obtained by transfection of the replication products into competent E. coli MC1061 cells, and shows that the mutation frequency for pol III from the non-mutator strain AM134 is 34.7 ϫ 10 Ϫ4 , whereas it is 83.4 ϫ 10 Ϫ4 for AM135, a 2.4-fold elevation. Similar results were obtained for pol III from strains AM147/pMV22 (non-mutator) and AM147/ pMV11 (mutator; Table V). Thus pol III* from mutator strains is error-prone against undamaged DNA also. Table V also shows that mutation frequencies for uncopied DNA, and for gap-filled (non-mutator polymerases) are, respectively, 12 ϫ 10 Ϫ4 , and 35-42 ϫ 10 Ϫ4 , values that are ϳ2-fold higher than those reported by Pham et al. (62) for uncopied DNA (6 ϫ 10 Ϫ4 ) and for DNA replicated with wild type pol III (15 ϫ 10 Ϫ4 ). It is possible that the higher background rates in our experiments are due to differences in the preparation of the template M13mRS65 gapped DNA. However, the 3-3.5-fold elevation in mutation frequency gap-filled DNA when compared with uncopied DNA in our experiments (Table V)   a Source for pol-III*. b pol-III* was purified from control (Ϫ) or mutator (ϩ) strains as described under "Materials and Methods." c Mutation frequencies were determined as summarized in Fig. 4 and described under "Materials and Methods." Each pol-III* preparation was used in at least two in vitro replication reactions. The newly synthesized strand from each replication reaction was amplified, and the resulting products were subjected to at least three rounds of multiplex sequence analysis to arrive at the numbers shown. S.D., standard deviation. d Fold elevation of mutation frequency for pol-III* from mutator strain over pol-III* from the corresponding control strain.
FIG. 5. Translesion DNA synthesis across an ⑀C residue by purified pol III*. The DNA template-primer used, schematically shown at right, consisted of a 120-nt-long oligonucleotide template bearing an ⑀C (or C for control experiments) at position 51, annealed with a 5Ј-32 P-end-labeled primer (30 nt) 14 nt downstream of the lesion. DNA synthesis by pol III* was carried out as described under "Materials and Methods," and the elongation products were analyzed on a 12% denaturing polyacrylamide gel. Locations of the full-length product, the site of the lesion, as well as the unelongated primer are marked on the left of the autoradiograph.  a %TLS (percent translesion DNA synthesis) was determined as described under "Materials and Methods" and is averaged from at least three independent replication reactions. S.D., standard deviation. Fidelity of pol III from E. coli mutA Cells observed for pol III from mutator strains in comparison to non-mutator strains (Table V) is similar to the 3-fold elevation in a lacZ forward mutagenesis assay based on replication of M13 ssDNA by crude extracts from mutA and wild type cells (10). Thus, pol III* from the mutator strains tested here is error-prone against undamaged DNA.

DISCUSSION
The present study was undertaken to investigate the mechanisms underlying the mutator phenotype observed in cells in which mistranslation levels have been increased either due to mutations in genes for tRNAs or tRNA-modifying enzymes or as a result of exposure to mistranslation-promoting agents such as streptomycin. Although mutator phenotypes can result from defects in repair, replication, or recombination, accumulating evidence suggests that mistranslation-mediated mutagenesis is mediated by error-prone DNA replication. In particular, our previous work (10) showed that a transiently expressed novel DNA polymerase or an induced alteration of a constitutively expressed DNA polymerase might be responsible for the TSM phenotype.
How exactly mistranslation activates error-prone replication is not understood. Two available hypotheses have focused on a modification of E. coli DNA polymerase III. Slupska et al. (4) have proposed that the phenotype results from direct mistranslation of the ⑀ subunit protein of pol III to create a small fraction of dominant-negative ⑀ proteins. Because Asp 3 Gly mistranslation by a missense suppressor tRNA is inefficient (1-2% (63)), a key prediction of this hypothesis is that the mutator phenotype is transient, being expressed in a small fraction (1-2%) of mutA cells at any one time. However, transfection of M13 ssDNA bearing a site-specific ⑀C residue results in a 5-20-fold elevation in mutagenesis, a result that suggests that a majority of mutA cells must constitutively express a mutator phenotype (5). To allow for a constitutive expression of the phenotype, the alternative hypothesis suggested that mistranslation triggered an "upstream" event activating a pathway that ultimately led to the constitutive expression of an error-prone replication activity (5). The molecular signal that turns on error-prone replication is presumed to be increased protein turnover due to an accumulation of misfolded proteins generated by mistranslation. Alternatively, mistranslation may result in the production of very small amounts of "gain-of-function" mutant proteins that activate a signal cascade leading to the production of an error-prone polymerase.
The work reported here shows that efficient induction of the TSM response occurs in cells simultaneously defective for pol I, pol II, pol IV, and pol V (Fig. 2). The construction of a viable, multiply deficient strain (AM147 polA polB dinB umuDC) suggests that all essential replication functions in E. coli can indeed be carried out by the remaining polymerase, namely pol III.
To address the question whether a modified form of pol III can account for the mutator phenotype, we purified pol III* from two pairs of TSM-induced and uninduced cells ( Fig. 3 and Table II) and analyzed the replication fidelity of the purified enzymes on both damaged (⑀C; Fig. 4 and Table III) and undamaged DNA templates ( Fig. 6 and Table V). Our results show that mutagenesis is elevated over 2-fold on both damaged and undamaged DNA when they are replicated by pol III from mutator strains as compared with control strains (Tables III  and V).
The enhanced mutagenesis by pol III* from TSM-induced cells is characterized by an increase in C 3 A and C 3 T mutations, with C 3 A mutations predominating over C 3 T mutations (Table III). This mutational specificity is in complete agreement with our in vivo studies (5,8) as well as our previous in vitro studies with crude extracts (10). The data here show that pol III is capable of extensive DNA synthesis across an ⑀C lesion (ϳ35% translesion synthesis; Fig. 5 and Table IV). Thus, our results suggest that both base insertion opposite an ⑀C residue as well as extension past the site can be carried out by pol III in E. coli cells and that "lesion-bypass polymerases" such as pol II, pol IV, and pol V (32) are not required for some noninstructive mutagenic DNA lesions. This finding resolves a number of early and previously puzzling observations on the mutagenic properties of ⑀C, including its noninstructive template characteristics (64), as well as its ability to induce mutations in SOS-deficient cells (65). Indeed, the unusual mutagenic properties of ⑀C were instrumental in uncovering UVM, an SOS-independent DNA damage-inducible pathway in E. coli (3,49,50,66).
Because essentially identical results are obtained with two different pol III* preparations from two different mutA strains (one of which was defective for all four of the remaining DNA TABLE V Fidelity of in vitro gap-filling pol-III* M13mRS65 gapped DNA was constructed as described ("Materials and Methods" and Fig. 6), and in vitro gap-filling synthesis by purified pol-III* preparations was carried out as described under "Materials and Methods." Approximately 20 ng of the replicated (RFII) DNA was transfected into electrocompetent MC1061 cells, and the transfected cells were plated on X-gal indicator plates as described under "Materials and Methods." Under the plating conditions, wild-type phage forms colorless plaques, whereas lacI mutant phage forms blue or light blue plaques. a Source of pol-III* used for replication. b Mutation frequencies were determined as the ratio of the number of mutant (blue) plaques to the total number of plaques as described under "Materials and Methods." For each pol-III* preparation, two independent gap-filling reactions were carried out. Each gap-filled DNA product was transfected at least twice. Thus, values shown are averages of at least four transfections. S.D., standard deviation.
c Fold elevation of mutation frequency for DNA gap-filled with pol-III* over uncopied (unfilled) DNA. d Fold elevation of mutation frequency for DNA gap-filled with pol-III* from mutator strain over that with pol-III* from the corresponding control strain.
Fidelity of pol III from E. coli mutA Cells polymerases thus ruling out any polymerase cross-contamination), these results are consistent with the conclusion that pol III* from mutA cells is error-prone. Nevertheless, the apparent magnitude of the mutator effect at ⑀C residues differs from that observed in vivo as well as previous in vitro replication studies based on crude extracts. Thus, mutagenesis is elevated 2-fold at ⑀C by pol III* from TSM-induced cells over controls, as compared with the 5-20-fold effect observed in vivo (5,8) and in vitro with crude cell extracts (10). This magnitude of difference between in vivo and in vitro results appears to be an inherent, as yet unexplained, property of highly purified pol III preparations. It is instructive to compare these results with the in vivo and in vitro magnitude of differences for wild type dnaQ and mutD5 pol III holoenzyme preparations (62). The dnaQ defect in mutD5 cells reduces the editing activity of ⑀ to Ͻ2% of the wild type level and leads to a 10,000-fold elevation in mutagenesis. A part of the mutator effect is attributable to saturation of the mismatch repair pathway; however, even in cells overexpressing mutL (the limiting component of mismatch repair system), there is only a 6 -20-fold reduction in the magnitude of mutator effect, so that residual mutagenesis is still about 2 to 3 orders of magnitude above that in wild type cells (67). Despite such a huge in vivo difference, the error frequency differences between the wild type and mutD5 holoenzymes (in the same experimental system as the one we have used) ranges from 2.2-to 4.5-fold at dNTP concentrations of 50 -1000 mM (we use 200 mM dNTPs) (62). In comparison to mutD5, mutA is a weaker mutator that elevates mutagenesis by about 5-20-fold depending on the assay (3,7,58), the most accurate determination being a 17-fold increase in mutation rate in streptomycin-induced rpsL1408 cells at a specific lacZ site (7). The above considerations suggest that the observed elevation in error levels mutA pol III is significant. These results, in conjunction with the demonstration that the mutA phenotype is expressed in cells defective for polA, polB, dinB, and umuDC (Fig. 2), suggest that the error-prone polymerase mediating TSM is a modified version of pol III. How exactly pol III is modified in TSM-induced cells is unknown. Among the possibilities are incorporation of one or more mistranslated subunits into the holoenzyme, chemical modification (such as phosphorylation or acetylation), and acquisition of a cofactor that can affect fidelity.