In vivo mutagenesis by Escherichia coli DNA polymerase I. Ile(709) in motif A functions in base selection.

The fidelity of DNA replication by Escherichia coli DNA polymerase I (pol I) was assessed in vivo using a reporter plasmid bearing a ColE1-type origin and an ochre codon in the beta-lactamase gene. We screened 53 single mutants within the region Val(700)-Arg(712) in the polymerase active-site motif A. Only replacement of Ile(709) yielded mutator polymerases, with substitution of Met, Asn, Phe, or Ala increasing the beta-lactamase reversion frequency 5-23-fold. Steady-state kinetic analysis of the I709F polymerase revealed reductions in apparent K(m) values for both insertion of non-complementary nucleotides and extension of mispaired primer termini. Abolishment of the 3'-5' exonuclease activity of wild-type pol I increased mutation frequency 4-fold, whereas the combination of I709F and lack of the 3'-5' exonuclease yielded a 400-fold increase. We conclude that accurate discrimination of the incoming nucleotide at the polymerase domain is more critical than exonucleolytic proofreading for the fidelity of pol I in vivo. Surprisingly, the I709F polymerase enhanced mutagenesis in chromosomal DNA, although the increase was 10-fold less than in plasmid DNA. Our findings indicate the feasibility of obtaining desired mutations by replicating a target gene at a specific locus in a plasmid under continuous selection pressure.

DNA polymerases catalyze chain elongation reactions guided by complementary base pairings opposite a single-stranded DNA template. These reactions are highly accurate, exhibiting error rates of about one base substitution error per 10 4 to 10 7 nucleotides polymerized (1). However, errors made by the polymerase, if not subsequently excised, can become fixed as mutations during subsequent rounds of replication. As a result, errors by DNA polymerases can be a major source of spontaneous mutagenesis and can contribute to the multiplicity of mutations found in cancer cells (2,3). Many DNA polymerases have intrinsic or associated 3Ј-5Ј exonucleases that preferentially hydrolyze non-complementary nucleotides immediately after formation of the phosphodiester bond and contribute from a few-to 100-fold to the fidelity of DNA synthesis (4 -6). In addition, errors introduced by DNA polymerases are subsequently corrected by a mismatch repair system, which contrib-utes an additional 2-3 orders of magnitude to the accuracy of DNA replication (7). However, base selection at the polymerase active site during both the nucleotide insertion and subsequent extension reactions, including Watson-Crick base pair formation between complementary bases and a conformational change of the active site during each incorporation step, is likely the most significant contributor to the fidelity of DNA polymerization (1, 8 -10). We have investigated the relationships between structure and function at the active site of DNA polymerases by substituting random sequences for nucleotides that encode residues at the active site and monitoring the effects of these substitutions on the fidelity of DNA synthesis (11,12).
Escherichia coli DNA polymerase I (pol I) 1 is involved in DNA replication, DNA repair, and genetic recombination (13); it is the most extensively studied of all DNA polymerases. Evidence indicates that pol I functions in DNA replication by removal of RNA primers and resynthesis of the resulting gaps between Okazaki fragments on the lagging strand (14,15). In addition, pol I participates in DNA repair by filling gaps resulting from the excision of damaged bases (16,17). Moreover, pol I is required for the initiation of synthesis at the origin of replication in certain plasmids (13,18). The crystal structure of the Klenow fragment of pol I (which lacks the 5Ј-3Ј exonuclease domain) reveals an architecture that is common among DNA polymerases and has been likened to a human right hand, with a fingers subdomain (which binds the incoming dNTP and interacts with the single-stranded DNA template), a thumb subdomain (which binds double-stranded DNA), and a palm subdomain (which harbors the catalytic amino acids and also interacts with the incoming dNTP) (19,20). Several mutant forms of the Klenow fragment of pol I (in which single amino acid substitutions have been introduced into the active-site fingers or palm subdomain (21)(22)(23) or a large 24-amino acid segment in the thumb subdomain has been deleted (24)) that exhibit altered fidelity in DNA synthesis in vitro have been investigated. Although many such mutant enzymes exhibit reduced fidelity in vitro, none has been shown to alter accuracy in vivo.
In a recent study, we examined the mutability of motif A, extending from Val 700 to Arg 712 , in the palm subdomain of E. coli pol I using random mutagenesis and a genetic complementation system (25). We established a library of 500,000 transfectants and sequenced 232 of 37,500 mutants that were active in the complementation assay. E. coli strains harboring the active mutants were fit to replicate repetitively, and the mutant polymerases, when purified, displayed 20 -190% of the wild-type specific activity. Thus, motif A is highly mutable while preserving wild type-like DNA polymerase activity in vitro and in vivo. The ease of substitutability of motif A residues revealed in this work, yielding highly functional variants, stands in sharp contrast to the marked conservation of the motif A sequence observed among prokaryotic DNA polymerases (26,27). Interestingly, we also found that certain substitutions of Ile 709 permit more efficient utilization of rNTPs as substrates in vitro.
In this study, we screened 53 mutations in motif A for infidelity of DNA synthesis in vivo and found that mutant enzymes harboring Ile 709 substitutions exhibited less accurate DNA replication. The mutator phenotype was enhanced when the Ile 709 substitutions were combined with deficiency of 3Ј-5Ј exonucleolytic proofreading activity. In subsequent in vitro experiments, we determined that the I709F substitution increased both insertion of non-complementary nucleotides as well as extension from primers with mismatched 3Ј-OH termini. To our knowledge, this is the first analysis of the effects of mutation in the polymerase active site of E. coli pol I on the fidelity of DNA synthesis both in vitro and in cells.

EXPERIMENTAL PROCEDURES
Plasmid Construction-The wild-type and mutant E. coli pol I genes were inserted into pHSG576 (28), placing them under the control of the lactose promoter. pHSG576 is low copy number plasmid that has a pol I-independent origin. To modify the gene, the wild-type pol I gene of E. coli DH5␣ was amplified by colony polymerase chain reaction and inserted into pHSG576 to create pECpol I as described previously (25). Site-directed mutagenesis was performed on pECpol I to introduce an A-to-C transversion at position 1271, thus changing Asp 424 to Ala and inactivating the 3Ј-5Ј exonuclease activity (29), to construct pECpol I-3Јexo Ϫ , which carries the 3Ј-5Ј exonuclease-minus pol I gene. Plasmids pECI709M, pECI709N, pECI709F, and pECI709A, which carry Ile 709 mutant pol I genes, were isolated from a mutant pol I library by genetic selection as described previously (25). Plasmids pECI709M-3Јexo Ϫ , pECI709N-3Јexo Ϫ , pECI709F-3Јexo Ϫ , and pECI709A-3Јexo Ϫ were constructed by substituting the 1.1-kb SacI-EcoRI fragment of pECpol I-3Јexo Ϫ for the corresponding fragment of pECI709M, pECI709N, pECI709F, and pECI709A, respectively.
The reporter plasmids for measuring the reversion frequency of the ␤-lactamase gene were constructed by modifying plasmid pGPS3 (New England Biolabs Inc., Beverly, MA), which contains a ColE1-type origin derived from pUC19. Site-directed mutagenesis was performed on pGPS3 to introduce a G-to-T transversion at position 76 of the ␤-lactamase gene, changing the codon GAA for Glu 26 to the ochre codon TAA. The resulting plasmid was designated pLA2800. The mutant ␤-lactamase gene containing its own promoter was amplified by polymerase chain reaction with the synthetic oligonucleotides 5Ј-GCACCCGACAT-ACATGTCCTATTTGTTTATT-3Ј and 5Ј-AAACTTGGTCGGATCCTTA-CCAATGCTTAATC-3Ј as primers and pLA2800 as a template, and the amplified fragment was cloned into pCRII (Invitrogen, Carlsbad, CA). The 1-kb AflIII-KpnI fragment containing the mutant ␤-lactamase gene was excised and cloned into the AflIII-KpnI site ϳ60 bp distant from the origin of pGPS3⌬LA (see below) to create pLA230. Plasmid pGPS3⌬LA was constructed by replacing the 1.2-kb BglII-BglI fragment of pGPS3 with the synthetic oligonucleotides 5Ј-GATCTGATCGCCCTTC-3Ј and 5Ј-GGGCGATCA-3Ј. A schematic representation of the reporter plasmids pLA230 and pLA2800 is shown in Fig. 1A.
␤-Lactamase Reversion Assay-A schematic representation of this assay is shown in Fig. 1B. E. coli JS200 (recA718 polA12 uvrA155 trpE65 lon-11 sulA1) (30, 31) harboring pLA230 or pLA2800 was transformed with plasmids carrying wild-type or mutant pol I genes. The recombinant strains were cultured at 30°C for 16 h in nutrient broth containing 50 g/ml kanamycin, 12.5 g/ml tetracycline, and 30 g/ml chloramphenicol. A 0.01 volume of the pre-cultured broth was inoculated into fresh medium; cultured at 37°C until an A 600 of ϳ1.0 was attained; and then plated onto LB agar plates supplemented with 50 g/ml kanamycin, 12.5 g/ml tetracycline, and 30 g/ml chloramphenicol in the presence or absence of 80 g/ml carbenicillin. After incubation at 37°C for 16 h, colonies were counted, and reversion frequencies were calculated as the ratio of carbenicillin-resistant to total colonies.
Plasmids prepared from revertants were re-transformed into E. coli BL21 and selected on LB plates containing 50 g/ml carbenicillin. The plasmids were prepared from the recombinant BL21 strain, and the nucleotide sequence of the entire ␤-lactamase gene was determined.
Trp ϩ Reversion Assay-The Trp ϩ reversion assay was performed basically following the method of Washington et al. (32). E. coli JS200 was transformed with plasmids pECpol I, pECpol I-3Јexo Ϫ , pECI709F, and pECI709F-3Јexo Ϫ , as indicated. The transfectants were cultured at 30°C for 16 h in nutrient broth supplemented with 30 g/ml chloramphenicol and 12.5 g/ml tetracycline. A 0.01 volume of each culture was inoculated into fresh medium, cultured at 37°C until an A 600 of ϳ1.0 was attained, and then plated onto M9 minimum agar plates supplemented with 30 g/ml chloramphenicol in the presence or absence of 40 g/ml tryptophan. After incubation at 37°C for 20 h, colonies were counted, and the frequency of appearance of the Trp ϩ strain was calculated.
The trpE gene in JS200 and in Trp ϩ revertant strains was amplified by colony polymerase chain reaction with 5Ј-CCATGCGTAAAGCAAT-CAGATACCC-3Ј and 5Ј-TTATCGAGCAGCAGAATGTCAGCCA-3Ј as primers, and the amplified fragment was cloned into pCRII; the nucleotide sequence of the entire trpE gene was then determined.
Kinetic Analysis-Steady-state kinetic analysis of misincorporation frequency was performed based on the method of Boosalis et al. (33). A 47-mer template (3Ј-GCGCGGCTTAAGGGCGATCGTTATAGCTTAA-GGCCTTTAAAGGGCCC-5Ј; the relevant template bases are underlined) was hybridized with a 32 P-5Ј-end-labeled 23-mer primer (5Ј-CGCGCCGAATTCCCGCTAGCAAT-3Ј) for analysis of misinsertion efficiency opposite dT and with a 25-mer primer (5Ј-CGCGCCGAATT-CCCGCTAGCAATAT-3Ј) for analysis opposite dG. Primer-template (5 nM) was incubated for 5 min at 37°C in a reaction mixture containing limiting amounts of purified recombinant Klenow (exo Ϫ ) protein (5 nM) prepared as described previously (25)  M dTTP for the I709F mutant enzyme. Following termination of the reaction by addition of 2.5 l of formamide solution, the products were analyzed by 14% polyacrylamide gel electrophoresis and quantified by phosphor image analysis (34).
Mismatch extension frequency was determined using a similar protocol, except that the sequence of the 24-mer primer was 5Ј-CGCGCC-GAATTCCCGCTAGCAATX-3Ј (where X was A, G, C, or T). Reaction mixtures contained dTTP, i.e. the correct dNTP for insertion opposite the next template base. The efficiency of dTTP incorporation opposite template dA was measured for each primer-template construct. The concentrations of the dTTP substrate were 0.5-3.5 nM for the T:A matched pair, 5-1000 nM for the T:G and T:C mismatches, and 0.025-14 M for the T:T mismatch.

Screening of pol I Mutants for
Mutator Activity-To measure errors in DNA synthesis by pol I in vivo, we established a two-plasmid system. The reporter plasmid pLA230 (Fig. 1A) contains a ␤-lactamase gene harboring an ochre mutation near the 5Ј terminus. Since evidence indicates that pol I is involved in initiation of DNA synthesis in ColE1-type plasmids (13,18), we introduced the ochre mutation ϳ230 bp from the ori sequence. The reporter plasmid, together with a second plasmid encoding the wild-type or mutant pol I gene, was transfected into E. coli JS200 (30, 31), a strain that contains a temperature-sensitive pol I. The reversion frequency at the ␤-lactamase locus was determined by measuring colony formation in the presence and absence of carbenicillin (Fig. 1B).
The pol I gene encoded by the second plasmid corresponds to the intact enzyme and hence encodes both 5Ј-3Ј and 3Ј-5Ј exonuclease activities as well as DNA polymerase activity (13). To identify amino acids in active-site motif A that affect the fidelity of DNA synthesis, we tested 53 different single motif A mutations within the segment spanning Val 700 -Arg 712 (25). Representative reversion frequencies obtained for mutants containing substitutions at each of the positions analyzed are shown in Table I. The reversion frequency observed for wildtype pol I was ϳ1 ϫ 10 Ϫ7 , as was that for all the mutants tested except the four with substitutions at position 709. Substitution of Met, Asn, Phe, or Ala for Ile 709 yielded 5.3-23 times higher reversion frequencies than that for wild-type pol I (Tables I and  II). We analyzed the nucleotide sequence of the ␤-lactamase reporter gene from five independent revertants harboring I709F mutant pol I and confirmed that the ochre mutation was converted twice to TTA, twice to TCA, and once to CAA. The enhanced mutagenesis observed for the Ile 709 mutants provides new evidence that E. coli DNA polymerase I is involved in plasmid replication by copying nucleotides near the ori sequence and demonstrates that Ile 709 is critical for accurate plasmid replication in vivo.
Effect of Exonucleolytic Proofreading Activity by pol I on Plasmid Replication-To analyze the contribution of the 3Ј-5Ј exonuclease ("proofreading") activity of pol I to the fidelity of plasmid replication, we mutated the 3Ј-5Ј exonuclease in the wild-type enzyme and in the Ile 709 variants by substituting Ala for Asp at position 424 in the exonuclease domain. When in-troduced into the wild-type construct, the D424A substitution enhanced the reversion frequency of the ␤-lactamase gene by 4.4-fold (Table II). Larger increases (up to 22-fold) were observed for specific amino acid substitutions of Ile 709 , indicating that discrimination at the active site can be more important for fidelity than exonucleolytic proofreading. Abolishment of the 3Ј-exonuclease activity in the mutants harboring the I709M, I709N, I709F, or I709A substitution resulted in a 29 -416-fold increase in mutation frequency relative to the wild-type enzyme. For each of the mutants, the increase in reversion frequency associated with inactivation of the exonuclease was greater than that observed for the wild-type enzyme. In the case of I709F, the increase was substantially greater than multiplicative, suggesting a functional interaction between the exonuclease domain and motif A.
To further evaluate the mutant polymerases, we measured the reversion frequency of the same ␤-lactamase gene on another plasmid, pLA2800 ( Fig. 1A and Table III). In this construct, the ␤-lactamase gene is located ϳ2.8 kb downstream of the origin of replication and thus is ϳ10-fold more distant from the origin than in pLA230. Introduction of the 3Јexo Ϫ mutation into the wild-type construct or substitution of Met, Asn, Phe, or Ala for Ile 709 in separate constructs resulted in at most a 1.8-fold increase in reversion frequency. In contrast, pol I harboring both the 3Јexo Ϫ mutation and an Ile 709 substitution showed 10 -87-fold higher reversion frequency than wild-type pol I; the elevations were not as large, however, as those observed for pLA230 (Table II and III). These results suggest that DNA synthesis by pol I is not necessarily limited to nucleotides near the origin, but can occur much farther downstream.
Effect of Mutator pol I on Replication of Chromosomal DNA-E. coli JS200 cannot grow in the absence of tryptophan since it carries the trpE65 (ochre) allele in the host chromosome. We investigated the effects of the mutant polymerases on the reversion frequency at the trpE locus (Table IV). The reversion frequency observed for wild-type pol I was 2.0 ϫ 10 Ϫ8 . Neither the 3Јexo Ϫ mutation nor the I709F substitution significantly increased this frequency. In contrast, the mutant pol I with both the 3Јexo Ϫ mutation and the I709F substitution exhibited a 40-fold increase in reversion frequency. We analyzed the nucleotide sequence of the trpE gene from three independent revertants and determined that the ochre mutation was converted once to TAC and twice to TCA. These results indicate that the mutant pol I bearing both an Ile 709 substitution and the 3Јexo Ϫ mutation participates in replication of the E. coli genome with less accuracy than the wild-type  reversion assay (B). A, the mutant ␤-lactamase gene harboring an ochre codon was inserted ϳ0.23 kb (pLA230) or ϳ2.8 kb (pLA2800) from the ColE1-type origin on pGPS3 as described under "Experimental Procedures". B, a pol I-deficient temperature-sensitive (ts) E. coli strain was transformed by a reporter plasmid carrying a ␤-lactamase mutant, after which the wild-type or mutant pol I gene on a plasmid was introduced into the bacteria. The recombinant strain was cultured in liquid medium containing kanamycin (Km), tetracycline (Tet), and chloramphenicol (Cm) and then plated on agar plates containing kanamycin, tetracycline, and chloramphenicol with or without carbenicillin (Carb). Colonies were counted after 16 h at 37°C, and reversion frequency was determined. enzyme. However, the effect of the mutator activity on chromosomal DNA synthesis was less than on plasmid DNA synthesis. In all of the in vivo situations examined, specific mutations at the polymerase active site and inactivation of the proofreading activity acted synergistically to increase the mutator activity of pol I.

Measurement of Polymerase Fidelity in Vitro-
To establish in vitro correlates of the mutator phenotype of the Ile 709 mutants, we purified the Klenow fragments of the wild-type and I709F exo Ϫ polymerases to apparent homogeneity (25). These fragments lack both 5Ј-3Ј and 3Ј-5Ј exonuclease activities. The 5Ј-3Ј exonuclease could remove the 5Ј-label from the primer, and the 3Ј-5Ј exonuclease could remove added nucleotides in extension experiments. We then analyzed the efficiency of misinsertion using a steady-state gel-based assay (33) to measure the kinetics of single nucleotide addition opposite template dT or dG. The primer was a 23-or 25-nucleotide oligomer that was labeled at the 5Ј-end with 32 P, and the 3Ј-terminal nucleotide was one residue upstream from the target. The wild-type and mutant enzymes showed typical Michaelis-Menten saturation kinetics when initial velocity was plotted against the concentration of each nucleotide (data not shown). Apparent kinetic parameters and relative insertion frequencies were determined for each dNTP (Table V). The I709F polymerase incorporated complementary nucleotides with a catalytic efficiency indistin-guishable from that of the wild-type enzyme. However, the catalytic efficiency of misincorporation of the non-complementary nucleotides was 6 -35 times greater than that of the wildtype enzyme; the enhancement was 6 -23-fold for misinsertion opposite template T and 8 -35-fold opposite template G. Increased misincorporation by the mutant enzyme was due almost exclusively to a lower K m for mispaired dNTPs. Notably, misincorporation opposite dT parallels our in vivo finding of A-to-C or A-to-T transversions among the plasmid-borne ␤-lactamase revertants. Based on current models of initiation at ColE1-type origins (13,18), these transversions putatively arise from T:C or T:T mispairs catalyzed by pol I during leading strand synthesis.
Both incorporation of mispaired nucleotides and extension of mispaired primer termini are required for base substitution mutations in vivo. We determined the efficiency of mispair extension using a series of primer-templates containing a 3Јterminal T:A, T:G, T:C, or T:T base pair and measuring the frequency of incorporation of the next correct nucleotide, dTTP (Table VI). All plots of initial velocity versus dNTP concentration exhibited saturation kinetics (data not shown). The I709F exo Ϫ polymerase extended the matched T:A pair with a catalytic efficiency indistinguishable from that of the wild-type enzyme. However, the catalytic efficiency of extension of the mismatched termini was 3-14 times greater than that of the wild-type enzyme, due predominantly to 7-17-fold lower K m values for the next correct nucleotide. The increases in catalytic efficiency and the reductions in K m are similar (i.e. are within a factor of ϳ2) to those observed for misincorporation opposite template T (Table V). These results indicate that the I709F mutation reduces discrimination against extension of mismatched primer termini as well as discrimination against incorporation of non-complementary nucleotides. DISCUSSION We have used random mutagenesis to establish a library of mutations in motif A of E. coli DNA polymerase I and utilized genetic complementation of a pol I-deficient temperature-sensitive E. coli strain to identify active mutants (25). By screening portions of this library with a reporter plasmid, we determined here that pol I mutants with an I709M, I709N, I709F, or I709A substitution in the catalytic palm subdomain exhibit a mutator phenotype. Enhanced mutagenesis was observed during both plasmid and chromosomal DNA replication. Thus, we have obtained pol I mutants that display low fidelity of DNA replication in vivo. We know of no other active-site mutants of pol I that exhibit reduced replication accuracy in cells, although Minnick et al. (23) have reported that a mutant Klenow (exo Ϫ ) enzyme with the single amino acid substitution E710A in motif A efficiently incorporates mismatched nucleotides in vitro. We have not detected the E710A mutation in the E. coli pol I motif A active mutant library (25) or in a corresponding Thermus aquaticus pol I library (35), presumably because the catalytic activity of the mutant is insufficient to permit complementation.
Measurements of the in vitro fidelity of the 3Јexo Ϫ Klenow fragment of E. coli pol I (9) and the 3Јexo Ϫ E. coli DNA polymerase III (pol III) holoenzyme (8) indicate that abolishment of the exonucleolytic activity increases the overall error frequency by 4 -7-fold for pol I and by Ͻ10-fold for pol III. These relatively modest increases indicate that the major component of accuracy of these enzymes, one error per 10 4 to 10 7 nucleotides polymerized, represents discrimination during polymerization, presumably including a conformational change in the enzyme at each nucleotide addition step (36,37). Our results demonstrating that nucleotide selection at the polymerase active site is the major contributor to fidelity in vivo provide a novel   confirmation of the findings for pol I. Interestingly, our results indicate that exonucleolytic proofreading may make roughly the same contribution to base substitution fidelity in vivo (4fold in this study) as in vitro (Ͻ10-fold) (9). In contrast, available data for the pol III holoenzyme suggest that proofreading may make a smaller contribution to base substitution fidelity in vitro (less than ϳ10-fold under most conditions) (8,38) than in vivo (Ͼ50 -400-fold) (39), perhaps due to interaction of the holoenzyme with other proteins at the replication fork. The importance of communication between the polymerase and exonuclease active sites for proofreading has been suggested by in vitro data (40,41). Our results show that an amino acid substitution in the polymerase active site, i.e. I709M, I709N, I709F, or I709A, together with 3Ј-5Ј exonuclease deficiency, produces an increase in mutation frequency that is more than additive and, in the case of I709F, that is more than multiplicative. Thus, the polymerase and exonuclease active sites of E. coli pol I may cooperate to achieve accurate DNA polymerization in vivo. Recent studies on mutations in bacteriophage RB69 DNA polymerase also provide evidence for coupling between the exonuclease and polymerase sites (42). In this enzyme, the contribution of the exonuclease to accuracy is much greater than in E. coli pol I. Nevertheless, the mutation rate of the double mutant is greater than the sum of the components (42).
In vitro kinetic analysis showed that one of our Klenow (exo Ϫ ) polymerase mutants, I709F, exhibited more efficient incorporation of non-complementary nucleotides and more efficient extension of mismatched 3Ј termini than the wild-type enzyme. Increased efficiencies were due almost entirely to ϳ10-fold reduction of K m values. We have previously observed that the same Ile 709 mutant efficiently incorporates ribonucleotides in vitro, also mediated by ϳ10-fold decreased K m values for incoming rNTPs (25). Taken together, our observations indicate that Ile 709 contributes to both base and sugar discrimination in wild-type pol I. Results from substitution of the corresponding residue of T. aquaticus pol I, Ile 614 , also indicate that this amino acid serves to maintain the fidelity of base selection and to exclude ribonucleotides in vitro (12,34). However, in contrast to E. coli pol I, hydrophobic substitutions in T. aquaticus pol I at position 614 do not reduce base discrimination. As discussed previously with respect to the T. aquaticus pol I⅐DNA⅐ddNTP ternary structure, Ile 614 packs near the sugar and base portions of the incoming nucleotide, and substitution of the isoleucine residue appears to result in loss of stable packing against incoming nucleotides, thus facilitating inaccurate polymerization (12,34). Studies of the Klenow fragment of E. coli pol I have shown that the coordination between the polymerase and exonuclease sites can be affected by changing amino acids between or within the active sites (43,44). The Ile 709 mutation in pol I might affect the ability of the mismatched primer terminus to slide into the exonuclease active site.
In ColE1-type plasmids, pol I initiates DNA synthesis from primers synthesized by RNA polymerase and RNase H and is replaced by the pol III holoenzyme to complete the replication of the plasmid (13,18). The detailed mechanism and location of the switch from pol I to pol III are not completely understood. We observed here that mutations occur in a ␤-lactamase reporter gene when the target is close to the origin at a frequency   4 -6-fold greater than when the same gene is located 2.5 kb from the origin. This suggests that pol I not only catalyzes synthesis near the origin, but also can participate in replication at a distance from the origin. Possibly, the increased expression of pol I in our recombinant host cells favors substitution for pol III. In E. coli chromosomal DNA replication, pol III is responsible for synthesis of both the leading and lagging strands. The role of pol I is limited and estimated to be responsible for less than ϳ1% of chromosomal replication by acting in joining of Okazaki fragments and in DNA repair (13). Thus, the 40-fold enhancement of the reversion frequency of a single codon in the chromosomal trpE65 gene was unexpected. At least three mechanisms can be invoked for this enhancement of chromosomal mutagenesis. 1) The trpE65 gene is a hot spot for mutagenesis, possibly due to unusual secondary structure; 2) the mutated site corresponds to a segment involved in the synthesis of an RNA primer; and 3) the contribution of pol I to chromosomal replication is greater than previously surmised. The high level of mutagenesis displayed by pol I mutants in copying genes located near the ColE1-type origin of replication suggests the feasibility of placing specific genes at this site and developing systems for progressive mutagenesis under continuous selection for mutants with desired properties. In the course of these investigations, Fabret et al. (45) reported a method for in vivo gene-targeted random mutagenesis. They showed that a targeted gene on a ColE1-type plasmid could be randomly mutated by 3Јexo Ϫ pol I when that gene was lysogenized in an E. coli strain lacking both wild-type pol I and mismatch repair. Our results indicate that I709F/3Јexo Ϫ mutator pol I displays 100 times more inaccurate DNA synthesis than 3Јexo Ϫ pol I. Taken together, these findings suggest the possibility of creating more efficient systems for targeted random mutagenesis and for selecting specific mutations in vivo.