On the fidelity of DNA replication. Effect of divalent metal ion activators and deoxyrionucleoside triphosphate pools on in vitro mutagenesis.

The effect of assay conditions on the fidelity with which Escherichia coli DNA polymerase I copies natural DNA has been determined using a modification of the recently developed 4x174 fidelity assay system (Weymouth, L. A., and Loeb, L. A. (1978) Proc. Natl. Acad Sci. U. S. A. 75,1924). In this assay, the error rate of DNA synthesis in vitro is quantitated by measuring the frequency of reversion of copied +x174 am3 DNA to wild type DNA. Variations in the divalent metal ion activator used in the copying reaction markedly affect the reversion frequency of copied 1~x174 am3 DNA. Thus, the calculated error rate of 1 incorrect for every 17,100 correct nucleotides incorporated observed with 5 mM M&+ can be increased severalfold by the substitution of Mn2+ or Co’+ for Mg+. The error rate can also be increased by copying in the presence of inhibiting concentrations of M8+. In selective situations, the reversion frequency is dependent on the relative proportion of the different deoxyribonucleoside triphosphates present in the copying reaction. The error rate increases severalfold when the level of dCTP or dATP is increased during in vitro synthesis. Furthermore, there is a direct proportionality between relative increase in dATP pool size and error rate in Mn’+-activated reactions. These results with natural DNA substantiate our previous studies with synthetic polynucleotides and support the concept that the most likely substitutions observed in the +x174 fidelity assay system are a C for T transition and an A for T transversion at position 587, and a C for A transversion at position 586 of the am3 codon.

The effect of assay conditions on the fidelity with which Escherichia coli DNA polymerase I copies natural DNA has been determined using a modification of the recently developed 4x174 fidelity assay system (Weymouth, L. A., and Loeb, L. A. (1978) Proc. Natl. Acad Sci. U. S. A. 75,1924).
In this assay, the error rate of DNA synthesis in vitro is quantitated by measuring the frequency of reversion of copied +x174 am3 DNA to wild type DNA. Variations in the divalent metal ion activator used in the copying reaction markedly affect the reversion frequency of copied 1~x174 am3 DNA. Thus, the calculated error rate of 1 incorrect for every 17,100 correct nucleotides incorporated observed with 5 mM M&+ can be increased severalfold by the substitution of Mn2+ or Co'+ for Mg+. The error rate can also be increased by copying in the presence of inhibiting concentrations of M8+. In selective situations, the reversion frequency is dependent on the relative proportion of the different deoxyribonucleoside triphosphates present in the copying reaction.
The error rate increases severalfold when the level of dCTP or dATP is increased during in vitro synthesis. Furthermore, there is a direct proportionality between relative increase in dATP pool size and error rate in Mn'+-activated reactions. These results with natural DNA substantiate our previous studies with synthetic polynucleotides and support the concept that the most likely substitutions observed in the +x174 fidelity assay system are a C for T transition and an A for T transversion at position 587, and a C for A transversion at position 586 of the am3 codon.
Until recently, studies on the fidelity of DNA synthesis in vitro have involved the use of synthetic polynucleotides. These studies have clearly shown that the frequency of insertion of an incorrect nucleotide is markedly affected by the specific in vitro reaction conditions used. Of particular interest are the studies involving the use of mutagenic metals (I-7). For example, substitution of manganese or cobalt, both of which are carcinogenic (8,9) and mutagenic (lo-12), for magnesium as the divalent metal ion activator has been shown to increase the error rate of in vitro DNA synthesis by * This study was supported by grants from the National Institutes of Health (CA-24845) and the National Science Foundation . This is the tenth paper in a series, "On the Fidelity of DNA Replication." The first and ninth paper in this series are Ref. 14 and 7, respectively. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
$ Postdoctoral fellow of the National Institutes of Health (CA-06168).
We have recently developed an in vitro assay for measuring fidelity on a natural DNA containing all four bases (13). 6x174 DNA containing an amber mutation serves as a template. We can now address questions regarding the effects of mutagenic, or carcinogenic metals, or both, on fidelity using this natural template. In addition, in vitro studies with synthetic polynucleotides indicate that the error rate is related to the relative concentrations of correct and incorrect deoxynucleoside triphosphate present in the copying reaction (6,(14)(15)(16). Thus we can potentially identify the types of incorrect nucleotide insertions occurring at the amber mutation by altering substrate pool sizes, under any defined reaction conditions desired.
In order to measure the effects of changes in in uitro copying conditions, we have modified the original +x174 fidelity assay (13) so as to require approximately one-fiftieth the amount of template e primer complexes. Using this modified system, we now report the effects of variations in three divalent metal ion activators and in deoxyribonucleoside triphosphate pool sizes on the fidelity with which E. coli DNA polymerase I copies natural 4x174 DNA in uitro. A knowledge of these effects may be a required first step in the systematic investigation of the effects of replication proteins on the fidelity of DNA synthesis.
37°C with aeration to an Acm = 0.58 (4 x 10" cells/ml). The culture was made 5 mM in CaCb and +x174 an3 bacteriophage (spontaneous reversion frequency 3 x 10m7) were added at a multiplicity of infection of 5. and increased the number of infective centers observed by l-to P-fold.
Thus far, all attempts to increase the number of wild type revertants ("marker rescue") by using spheroplasts prepared from E. coli strains deficient in one or more nucleases, or DNA repair pathways, or both, have been unsuccessful. This is due to a large decrease in overall infectivity of spheroplasts prepared from these strains.

RESULTS
The basis of the 4x174 fidelity assay is the observation that infection of spheroplasts with a mutant $1~174 DNA circle hybridized to a wild type DNA fragment which will cover the mutation results in production of progeny which express the genotype of the wild type DNA fragment (26,27). The plan is to initiate in vitro DNA synthesis on +x174 am3 viral DNA at a fixed primer terminus, copy past an amber mutation, and then determine the frequency of insertion of an incorrect nucleotide by measuring the reversion of the amber mutation to wild type. In the experiments described here, we have chosen the amber mutation am3, which is in gene E, coding for host cell lysis, and in gene D, which overlaps gene E. Since the only nucleotide substitutions that will result in revertant, wild type, phage are those coding for both functional gene E and functional gene D proteins, it is likely that this assay measures only substitutions at the am3 codon itself. Dependence of Increase in Reversion Frequency on DNA Synthesis-The in vitro requirements for producing an increased reversion frequency with E. coli DNA polymerase I in a Mn"'-activated reaction are shown in Table I. Omission of metal activator, DNA polymerase, or any of the four deoxynucleoside triphosphates eliminates the production of revertants above the level obtained with the unincubated control. These requirements are identical with those needed for DNA synthesis on primed +x174 DNA (Table I) or other templates (28), and suggest that wild type revertants are indeed produced by in vitro DNA synthesis and not by some nonspecific effect of reaction components.
Metal Activation of DNA Synthesis-The ability of Mg2+, Mn'+, and Co'+ to act as divalent metal activators of DNA synthesis using specifically primed 4x174 am3 DNA is shown  Fig. 1. Optimal activation occurs with Mg'+ at 7.5 mM. Mn'+ and Co2+ can both substitute for Mg'+, with optima at 1 mM and 3.75 mM, respectively. The incorporation with Mn2+ and Co2+ was 68% and 48%, respectively, of that achieved with Mg2+ under the assay conditions described (Fig. 1). Concentrations above the optima were inhibitory for all three metal activators, most prominently with Mn2+. These metal ion activation curves are similar to results obtained with E. coli DNA polymerase I in copying synthetic polynucleotides (2). Effect of Metal Activators on Fidelity of DNA Synthesis-The fidelity of DNA synthesis on specifically primed +x174 am3 DNA in reactions containing various concentrations of Mg2+, Mn", and Co2+ was determined.
In these and all subsequent experiments, whenever metal activator concentrations above the optimum were employed, the enzyme/template ratio was increased to the extent necessary to achieve approximately the same level of synthesis observed at the optimum metal concentration.
Control experiments indicate that the error rate with equal concentrations of all four dNTPs, calculated as described previously (13), does not change significantly with enzyme concentration or primer utilized (Z,, Zx, or ZJ, provided synthesis has proceeded beyond the am3 site (data not shown).
As shown in Table II, increasing the concentration of Mg'+ in the copying reaction increases the reversion frequency of the copied DNA when compared to the uncopied control for each concentration of Mg"+. This results in an increase in the calculated error frequency from l/17,100 at 5 InM Mg" to 1/ 2,270 at 15 mM Mg'+, the highest concentration examined. Substitution of Mn2' for Mg2+ increases the calculated error frequency by severalfold over that observed with activating concentrations of Mg'+. The error rate of E. coli DNA polymerase I with 1 mM Mn'+ is l/2,060, 8-fold higher than that observed with 5 mM Mg2+. Similarly, Col+ increases the frequency of misincorporation by 2-to 4-fold over 5 mM Mg'+. The effect of Mn" on fidelity as measured in this assay system is independent of the concentration of Mn"' used, since the error rate in the presence of suboptimal, optimal, and inhibitory concentrations of MnP+ is approximately the same. Effect of Variations in Deoxynucleoside Triphosphate Pool Size on Fidelity-We previously reported that the reversion frequency of copied DNA, measured as progeny phage, was dependent on the relative amounts of the four deoxyribonucleoside triphosphates (dNTPs) present in the in vitro DNA synthesis reaction (13). An analysis of this phenomenon using an infective centers assay is presented in Table III. In reactions activated with 1 mM Mn'+, the error rate is dependent on the relative concentrations of dNTPs in the DNA polymerase reaction. The reversion frequency of DNA copied in the presence of a lo-fold excess of dATP was 13.6 x 10e4 as compared to 1.45 x 10m4 for DNA copied in the presence of equimolar concentrations of all four dNTPs. This calculates to approximately a 16fold increase in error rate of DNA synthesis. The effect is similar, although not as marked, with dCTP. A 5-to lo-fold excess of dCTP yields a l-to 3-fold increase in the calculated error rate (see also  For all error rates expressed as "less than" values, the error rate is calculated on the basis of a reversion frequency "less than" 5% above background.   of dGTP or dTTP in the copying reaction produced a reversion frequency less than that of the uncopied control. Uncopied control reactions containing increased concentrations of each dNTP independently yield phage with a reversion frequency identical with that obtained with the uncopied control. These controls indicate that increased dNTP concen- Thus, a lofold increase in dATP pool size gave little or no increase in error rate with 7.5 mM or 15 mM Mg"+ and only a 3-fold increase with 3.75 mM Co"+ (Table III).
Proportionality between Increase in dATP Concentration and Increase in Error Rate-The increase in error rate of E. coli DNA polymerase I when copying primed +x174 DNA in the presence of 1.0 mM Mn'+ is directly proportional to the relative increase in dATP concentration (Table IV). When dATP is present during DNA synthesis in 50-fold greater quantity than the other three dNTPs, the error rate, as calculated from reversion of the am3 locus, increases 45-fold. It was not possible to test higher concentrations of dATP, since synthesis was greatly reduced, presumably due to chelation of the divalent metal ion by the deoxynucleoside triphosphates. The direct proportionality of error rate to relative dATP concentration described here was observed only with Mn'+ and dATP. Use of increased dCTP with Mn"+ or Coy+ or increased dATP with Mg"+ or Co"+ did not result in this proportionality.
Reversal of Pool Size Effects-To further delineate the types of nucleotide substitutions that were occurring at the am3 locus when altered pools were present, the relative amounts of two different dNTPs were altered at the same time (Tables V and VI). In MI?'-activated reactions containing a 5-fold excess of dATP, in which the error rate increased 5-fold from l/2,600 to l/489, the relative concentrations of each of the other three dNTPs were varied independently in graded amounts. Under these conditions, dGTP had essentially no effect on the error rate, since a lo-fold increase in dGTP yielded an error rate of l/423, similar to the l/489 seen with 5 x dATP alone. Increasing the concentration of dTTP progressively reversed the effect of excess dATP, such that the "5 x dATP + 10 x dTTP" condition exhibited a higher fidelity than when all four dNTPs were equal. In contrast, increasing dCTP concentrations enhanced the error rate approximately P-fold over that observed with "5 x dATP" only. A similar experiment using a 5-fold excess of dCTP as baseline is shown in Table VI. Once again, a combination of increased dATP and dCTP gave enhancement of infidelity over the excess dCTP only. However, both dGTP and dTTP showed a marked ability to reverse the effect of increased Fidelity of DNA Synthesis 5723 errors with dCTP, as for example, with "5 x dCTP + 10 X dTTP" or "5 x dCTP + 10 X dGTP". DISCUSSION The modified biological assay of fidelity of DNA synthesis described here has achieved approximately a 50-fold increase in infectivity of copied 4x174 DNA molecules over that previously published (13). The use of protamine sulfate in spheroplast preparations provided the most dramatic increase in transfection efficiency, while the combination of protamine sulfate and spermidine stabilized the spheroplasts for up to 14 days. This latter phenomenon is particularly valuable when performing direct comparative experiments. The transfection efficiencies obtained in our experiments (10m4 to lo-") are somewhat lower than those obtained by Henner et al. (22) for untreated +x174 DNA (lo-' to lo-"). We attribute this difference in part to the fact that we are infecting spheroplasts with molecules that are partially double-stranded due to the copying reactions, and are thus intermediate between viral DNA and RF DNA. Also, manipulation of the DNA in the copying reactions, different suppression efficiency of the spheroplast strain and subtle differences in preparation of spheroplasts (e.g. use of Sigma Fraction V bovine serum albumin in some instances instead of "Povite" albumin) could account for the somewhat lower transfection efficiency. Control of these variables could therefore potentially increase transfection efficiencies even more.
Our modified assay allows one to perform a large number of experiments using relatively small quantities of primer and template DNA. More importantly, the assay can now be performed entirely as an infective center assay and still provide a significant number of revertants. This is an advantage over studies involving progeny phage, which require a large number of controls to determine possible differences in the extent of multiplication of wild type and mutant DNA. Calculations of error rates of the DNA polymerase depend on three assumptions (13). The first is that nucleotide substitutions at other sites do not mimic mutations at the am3 locus. This is reasonable since am3 is an amber mutation in gene E which overlaps gene D (29). Thus, only nucleotide substitutions which code for functional E and D proteins will produce revertant infective centers. The second assumption is that the efficiency of minus strand expression, or penetrance, is the same for copied DNA as for control heteroduplex DNA. The penetrance of 0.13 f 0.06 was determined using heteroduplex DNA formed from intact 4x174 am3 viral DNA annealed to Hue III Fragment Z7 (13). If frequency of marker rescue is indeed determined by a competition of replication versus repair of the heteroduplex, as suggested by Baas and Jansz (30,31), then the penetrance of the DNA could vary depending on the length of the minus strand DNA forming the heteroduplex.
The expression of the minus strand of heteroduplex DNA generated in the copying reaction could be potentially higher, lower, or equal to that determined for control heteroduplexes formed from the Hue III Z7 fragment. Thus, the calculated error rate could change somewhat, depending on the extent of DNA synthesis. To circumvent this problem, we have copied the DNA to approximately the same extent under all conditions. The third assumption is that every DNA molecule is copied past the am3 mutation. In the Zsprimed reactions used here, only 83 nucleotides per template must be added to reach the am3 locus (or 393 nucleotides per template for Zs-primed synthesis). On the basis of nucleotide incorporation data, it is probable that those molecules which have been initiated have been extended past the am3 locus. It also is reasonable that all molecules that have been primed are initiated, since the PoZ I was in 25-fold molar excess over template in the most limiting condition. However, there is no reason to expect that hybridization of primer to template is 100% efficient, even with the 5:l primer/template ratio used here. Thus, all molecules may not have been copied and the error rate may be greater than calculated. From such considerations, it is obvious that the error rates presented here represent initial estimations, and are subject to change. This does not, however, change the error rates relative to one another, and allows conclusions to be made from these direct comparative studies. Effect of Metal Activators on Fidelity-Both Mn'+ and Co*+ are known to be mutagenic and carcinogenic (32). We have previously shown that both metals substitute quite well for Mg"+ as divalent metal ion activators for DNA synthesis by several DNA polymerases (2, 5, 7). Furthermore, the fidelity of DNA synthesis has been shown to be dependent on the nature and concentration of the metal activator used. Thus, both Mn*+ and Co2+ have been found to decrease fidelity when compared to Mg'+-activated DNA synthesis on synthetic polynucleotides for PoZ I (2), avian myeloblastosis virus DNA polymerase (5), and human placenta DNA polymerases (Y and p (7). The data in Fig. 1 and Table II on metal ion activation and fidelity confirm these findings and extend them to DNA synthesis using a natural template. The error rates observed here are in fact quite similar to those found when PoZ I utilizes poly[d(A-T)] as a template (2). This represents an important confirmation that studies of fidelity utilizing synthetic polynucleotides are real and that valid conclusions can be made from such studies.
The data obtained here with natural DNA differ from data obtained using synthetic polynucleotides in two aspects. The first is the extent of the decrease in fidelity of DNA synthesis in reactions activated with optimal versus inhibiting concentrations of Mg2+. This decrease is much more pronounced with natural DNA (lo-fold at 1.0 mM versus 15 mM Mg"+, Table II)

than with poly[d(A-T)],
where the decrease is typically at most 2-fold with PoZ I.* The second is that the error rate in copying natural DNA is relatively constant at all concentrations of Mn*+ examined (Table II), contrasted with the large decrease in fidelity observed utilizing poly[d(A-T)] with increasing Mn*+ concentrations (2). While the exact reasons for these differences are not clear, such effects could be explained on the basis of metal-template interactions. Physical studies have shown that divalent metal ions interact with both phosphates and bases on polynucleotides (33). Thus, metal-template interactions may be crucial in determining the fidelity of DNA synthesis. Differences between synthetic polynucleotides and natural DNA with respect to metal ion binding constants could account for the differences in fidelity observed. The suggestion that metal-template interactions may be crucial to fidelity is supported by these studies, in which the error rate of DNA synthesis was invariant over a range of enzyme concentrations (from an enzyme/template ratio of 25:l to 500:1), when the amount of template used was kept constant. Conversely, PoZ I is reportedly much more faithful when the correct nucleotide is proportionately higher (25). Based on these observations, we determined which nucleotide substitutions are most likely to occur at the am3 locus under precisely defined reaction conditions.
The genetic consequences of various potential nucleotide substitutions at the am3 locus are shown in Table VII. Measured here as infective centers, increased reversion frequencies were observed in the presence of increased levels of dATP or dCTP, while the use of increased dGTP or dTTP resulted in reversion frequencies less than control values, copied at equimolar dNTP concentrations.
Qualitatively, these data confum our previous findings measured with the progeny phage assay (13). However, measured as progeny phage, increased dCTP yielded a much greater effect in Mn'+-activated reactions than did increased dATP. The reasons for this difference are not clear.
The most dramatic decrease in fidelity observed here was with increased dATP levels in Mn"'-activated reactions. The only probable direct substitution of A at the am3 codon is for T at position 587. The fact that this transversion is not nearly as frequent in Mg*+-or Co2'-activated reactions (Table III) suggests that Mn2+ promotes transversions more effectively than Mg'+ or Co*+. This possibility is supported by the report of Orgel and Orgel (11) of a class of Mn*+-induced Tq mutants which had the properties expected of transversion mutants. The findings here are in contrast to studies involving synthetic polynucleotides, which suggest that >95% of errors are transitions rather than transversions (25). Conceivably the nature of the metal activator, or the neighboring nucleotide sequences, or both, could play a major role in determining the nature of the substitution.
Further studies to test this possibility could utilize other amber mutations having different neighboring nucleotides. The increased reversion frequency of DNA copied in the presence of high C (Table III and VI) can be explained by a C for T transition at position 587, which would result in the original wild type DNA sequence (Table VII). A second possibility is a C for an A transversion at position 586, a possibility supported by the fact that the combination of high A and high C increases the reversion frequency over that observed when either is increased by itself (Tables V and VI).
The effects of increased dTTP and dGTP suggest that the substitutions of these nucleotides which could potentially produce increased revertants do not occur at detectable levels. This is even more striking when one realizes that one possible substitution, a G for T transversion at position 587, codes for an amino acid known to be functional (serine, which suppresses am3 in E. coli HF 4714), and would therefore produce wild type phage. Potentially, the inability to insert a G at this position could be a restriction by the neighboring nucleotides. The observation that both T and G actually decrease the reversion frequency below that of DNA copied at equimolar concentrations of all four dNTPs suggested that T and G were competing with C, or A, or both. This possibility is strongly supported by the data in Tables V and VI. T competes with both A and C, since the reversion frequency in the presence of increased A or C decreases as the level of T is increased. G competes with C, but does not compete with A. Interestingly, if G competes with C at position 587, it may do so without being incorporated, since a G substitution at 587 should be functional and would therefore produce an increased reversion frequency, which is not observed (Tables III and VI). Alternatively, it is possible that the G substitution at position 587 could be repaired in uivo and thus not be observed.
Taken together, the data in Tables III through VI suggest that there are three substitutions in the am3 codon which produce detectable wild type phage in our assay system. These are an A for T transversion or C for T transition at position 587, or a C for A transversion at position 586. These conclusions do not take into account the possibility of specific in uiuo repair processes in the spheroplast strain which may limit the detectable substitutions at the am3 locus. The use of several spheroplast strains with different repair capabilities and the use of amber mutations at other loci in the $x genome will be required to detect the range of possible nucleotide substitutions which may occur. It is important to note that the error rate was proportional to the relative concentration of incorrect to correct nucleotides (biased pools) in only one situation, Mn'+, and high dATP. A proportional change was not observed with Mg"+ or Co*+. Thus, measurements of fidelity which depend on extrapolation from biased pools could be invalid. In every case controls are required to demonstrate that the change in error rate is proportional to the relative concentrations of incorrect to correct nucleotides.
The observation that the error rate of DNA synthesis in vitro is dependent on the relative concentrations of each of the four nucleotide substrates is of particular interest in light of the recent report of Peterson et al. (34) that in viuo mutagenesis of Chinese hamster cells is increased 3-to IO-fold in the presence of biased thymidine and deoxycytidine pools. Our findings suggest one possible mechanism to explain such enhanced mutagenesis.
Highly accurate copying conditions produce copied DNA having a reversion frequency very similar to or less than uncopied control values (Tables II, III, V, VI). This sensitivity limitation would make any systematic study of factors which improve fidelity difficult. In order to circumvent this problem, we desired to define a set of error-prone conditions which could potentially be used as a model system to study factors which improve fidelity. The data in Table IV represent our initial attempts in this regard. The extremely error-prone copying conditions shown here allow one a wide range of detectable improvement in fidelity by added exogenous factors such as any of the numerous proteins thought to be involved in DNA replication, or repair, or both (28,35). The $x174 fidelity assay system described here thus allows one to address problems involving both increased and decreased fidelity. In addition, by choice of mutant DNA templates containing different neighboring sequences, one can address questions of specificity.