DNA Polymerase Insertion Fidelity GEL ASSAY FOR SITE-SPECIFIC KINETICS*

A quantitative assay based on gel electrophoresis is described to measure nucleotide insertion kinetics at an arbitrary DNA template site. The assay is used to investigate kinetic mechanisms governing the fidelity of DNA synthesis using highly purified Drosophila DNA polymerase a holoenzyme complex and M13 primer-template DNA. K , and V,,, values are reported for correct insertion of A and misinsertion of G, C, and T opposite a single template T site. The misinsertion frequencies are 2 x lo-* for GOT and 5 X for both C-T and T-T relative to normal A-T base pairs. The dissociation constant of the polymerase- DNA-dNTP complex, as measured by K,, plays a dom-inant role in determining the rates of forming right and wrong base pairs. Compared with K , for insertion of A opposite T (3.7 f 0.7 p ~ ) , the K , value is 1100- fold greater for misinsertion of G opposite T (4.2 f 0.4 mM), and 2600-fold greater for misinsertion of either C or T opposite T (9.8 +I 4.2 mM). These K , differences indicate that in the enzyme binding site the stability of A-T base pairs is 4.3 kcal/mol greater than G-T pairs and 4.9 kcal/mol greater than C-T or T-T pairs. In contrast to the large differences in K,, differences in V,,, are relatively small. There is only a 4-fold reduc- tion in V,,, for insertion of G opposite T and an %fold reduction for C or T opposite T,

Gel electrophoresis in conjunction with DNA sequencing strategies makes it possible to resolve each base in an arbitrary DNA sequence and permits the analysis of a wide variety of localized phenomena including site-specific mutations. Sequencing gels have recently been used to detect base substitution errors by DNA polymerase in vitro (1)(2)(3). In the present study of polymerase fidelity, we develop a quantitative gel assay to measure kinetic parameters (V,,, and IC,,,) for nucleotide insertion a t arbitrary single sites along a DNA template strand. We illustrate the method with Drosophila DNA polymerase a (4) for the misinsertion of G, T, and C opposite a template T site relative to normal insertion of A, on primed M13 DNA. The misinsertion frequency is evaluated in terms of relative Vmax and K,,, values found with each nucleotide substrate. Recently, we have used this assay to measure nucleotide misinsertion kinetics of Drosophila polymerase a *This research was supported by National Institutes of Health Grants GM21422 and GM33863. 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. opposite template abasic (apurinic/apyrimidinic) sites (5).
The kinetic measurements are compared with the predictions of a model for DNA synthesis fidelity based on K , discrimination (6-9). In this model, the relative rates of inserting matched uersus mismatched nucleotides are governed by the relative residence times of dNTP substrates in the enzyme-DNA-dNTP complex, and the fidelity of DNA synthesis is attributed primarily to a much higher dissociation constant for wrong versus right dNTP substrates in the enzyme complex. Differences in catalytic rate constants for insertion resulting in V,,, discrimination are assumed to play a much less significant role in fidelity. In K , discrimination, the degree of fidelity is dependent on free energy differences between right and wrong base pairs at the binding site of polymerase (10) as opposed to mechanisms involving nucleotide selection or rejection at the catalytic site.

EXPERIMENTAL PROCEDURES
Materials-Purified Drosophila DNA polymerase a, consisting of at least three polypeptide subunits including primase (4), was a generous gift of Dr. I. R. Lehman, Stanford University. The primer, a 23-base deoxynucleotide (5"GGCCTTGATATTCACAAACGAAT-3') complementary in sequence to bases 2248-2225 in wild-type M13 DNA ( l l ) , was synthesized by conventional solid-phase methods in an automatic Microsyn 1460 synthesizer (Systec, Inc.) and was provided by Dr. R. E. Eritja, City of Hope. The template was singlestranded DNA isolated from wild-type M13 phage grown in Escherichia coli strain JM103. HPLC'-purified dNTP substrates were purchased from Pharmacia Biotechnology, Inc.
There seemed to be no detectible cross-contamination of any of the dNTPs. The absolute level of cross-contamination of a given dNTP with another was less than 1 part in 10'. This upper bound was established by HPLC chromatography of each dNTP substrate, first alone and then in the presence of added dNTP "contaminant" at a concentration just sufficient for detection. The ratio of detectable contaminant to substrate concentration was used to define the upper bound. AmPur deoxynucleotide (dAmPurTP) was prepared as described (7). Radioactive nucleotides, [y3'P]ATP and [cY-~'P]~ATP, were purchased from ICN Radiochemicals; T, polynucleotide kinase, from U. S. Biochemicals.
Primer-Template Annealing-The 32P-labeled primer was annealed to M13 DNA template in an annealing buffer (230 pl) of 50 mM Tris. HCI (pH 8.0), 2 mM 0-mercaptoethanol, and 50 rg/ml acetylated bovine serum albumin, containing 74 nM primer and 81 nM template. The solution was incubated at 100 "C for 6 min and then cooled slowly to room temperature.
Based on a computer analysis of the base sequence in M13 DNA, The abbreviations used are: AmPur, 2-aminopurine; dAmPurTP, 2-aminopurine deoxyribonucleoside triphosphate; HPLC, high performance liquid chromatography. the primer was expected to form productive hybrids only with the site (bases 2225-2248) to which it was perfectly homologous. No other 23-base sequence showed greater than 60% homology with the primer. Furthermore, no other sequence showed perfect homology to the last six bases at the 3'-end of primer. In our experiments, therefore, primer extension by DNA polymerase was unlikely to occur at any other site in M13 DNA.
DNA Polymerase Reactions-Equal volumes of solution A containing enzyme-primer-template complex and solution B containing dNTP substrates were mixed to start polymerization reactions for kinetic studies. Solution A was made by adding 2 pl of Drosophila polymerase a stock solution (2,000 units/ml) and 9 pl of concentrated bovine serum albumin (25 mg/ml) to 80 pl of the original primertemplate annealing solution. Solution B contained various concentrations of dNTPs in 52 mM Tris . HCl (pH 8.0), 16 mM MgCl,, and 5 mM dithiothreitol. A 10-pl reaction mixture of solutions A and B contained approximately 0.2 unit of polymerase activity (specific activity, 50,000 units/mg, a unit being capable of incorporating 1 nmol of dNTP into acid-soluble material in 60 min at 37 'C (4)). The estimated ratio of primer-template to enzyme is 301. The reaction mixture was incubated at 37 "C for time periods of 1 or 4 min. Reaction was terminated by adding 20 pl of 20 mM EDTA in 95% formamide.
Reaction conditions for time course experiments to determine a suitable reaction time for kinetic studies were similar except that reaction times varied between 1 and 16 min.
Gel Electrophoresis and Autoradiography-Samples (5 81) of DNA polymerase reaction mixture were denatured at 100 "C for 5 min, cooled on ice, loaded on 16% polyacrylamide gel (30 X 40 cm X 0.4 mm) containing 8 M urea, and electrophoresed for 4 h at 2000 V (50 V/cm) to obtain good resolution of extended primers. For autoradiography of the 32P-labeled primer bands, gels were vacuum-dried on Whatman No. 3MM filter paper and overlayed with blue medical xray film (Kodak GPB-1) for 1-12-h exposures.
Densitometry-The bands in each lane of a gel autoradiograph were scanned on a Hoeffer GS300 densitometer at 6.5 cm/min to obtain maximum resolution of absorbance changes. Band intensities were evaluated with a Hoefer GS350 data system by integrating the area under the absorbance curve for each band, above a base line drawn by linearly connecting points of minimum absorbance on each side of the band.
Calibration against a photographic step tablet purchased from Kodak showed the densitometer readings to be proportional to absorbance up to A = 2.0. Varying quantities of 32P-labeled primer were run on a gel and exposed over 3-h to 2-day time periods to determine the range over which film response (absorbance after development) was linearly related to counts of radiation. Exposed films were selected so that peak absorbance for each labeled primer and product band lay within the linear response range of the film. Plots of film exposure time versus integrated band intensity in arbitrary units showed (a) linearity up to 12,000 units, representing a peak absorbance around 2.0 and band width around 6 mm, and ( b ) a threshold value of 150 units. Accordingly, a band intensity measured as n units was assigned a corrected value of n + 150 units. In kinetic experiments, film exposure time was selected so that the highest band intensity measured on film did not exceed 10,000 units.
Competition Assay-The same primer-template and DNA polymerase conditions were used except that primer was not labeled and added substrates included 0.75 p~ [ o~-~' P ]~A T P (3000 Ci/mmol The ranges of unlabeled dNTP substrates used to compete out [a-32P]dATP were: 0.7-5.0 p~ dATP, 0.7-2.2 mM dGTP, 1.5-7.5 mM dCTP, and 4 pM-6 mM dTTP. Low concentrations of dTTP were used to observe the kinetics of extending the primer template to the template A site next to the target T ( Fig. 1, site 3 ) .

RESULTS
The experimental design allows one to compare the kinetics of inserting right versus wrong nucleotides at a single target site along a DNA template strand. Here we investigate the insertion of A, G, C, and T opposite a template T site ( Fig. 1) using highly purified Drosophila DNA polymerase CY, which is devoid of any detectible 3' + 5' proofreading exonuclease activity when present as an intact polymerase-primase holoenzyme complex (4, 12). The target T site, designated as site 3, is located three bases downstream from the original primer. Before reaching the target site, polymerase must first insert two properly paired Gs at sites 1 and 2. The primer is labeled with 32P at its 5'-end so that primers extended by one, two, or three nucleotides can be observed as discrete bands on an autoradiograph of the polyacrylamide gel.
The experimental objective is to measure v2,, the velocity of primer extension from site 2 to site 3, for correctly paired and mispaired nucleotide substrates. The expression relating v2, t o band intensities on the gel (see "Theoretical Analysis," below) is where I, and I, are the integrated intensities at sites 2 a n d 3, expressed as percentage of total primer, and t is the reaction time (1-16 min). With 50 pM dGTP in the reaction mixture, (I2 + I,) is close to its maximum value at a given t and does not change significantly with added dNTP (including additional much larger amounts of dGTP). When t is held constant and (I2 + I,) stays constant over the range of dNTP substrate concentrations used ( Table   I), t h e 13/12 term in Equation 1 measures the relative value of u2, at each substrate concentration.
An advantage in using v = 13/12 in place of u2, (Equation 1) is that there is no need to correct for loading error (inadvertant loading of different sample volumes on the gel). Proper evaluation of v23 requires normalization by dividing by the total integrated intensity of all bands on the gel (Io + I , + Iz + Is). Usually, two gel exposures are needed to photograph the very dark primer band (Io) and much lighter extension bands within the linear region of t h e film. However, 13/12 can be evaluated without normalization or dual exposure. As shown below, 13/12 at t = 4 min provides a suitable relative measure of vZ3 for kinetic studies.
Time Course Studies-The evolution of each band intensity as a function of t up to 16 min is illustrated in Figs. 2 and 3. When 3 p~ dATP is used for insertion of A opposite T ( A . T ) in site 3, along with 50 p~ dGTP for insertion of G opposite C (G. C) in sites 1 and 2, there is a rapid increase in I, (Figs.

TABLE I Examples of gel assay results showing the behavior of the insertion velocity (v23) as a function of time and substrate concentration
The value of v23 is the product of (I2 + Z3)/t and Z3/Z2, as measured on gels. Experimental data are shown for substrate dCTP at concentrations of (a) 1 mM, (b) 3 mM, and (c) 6 mM, in the presence of 50 p~ dGTP. At each dCTP concentration, as t is increased from 2 to 16 min, one observes that Z3/Z2 increases while (Z2 + Z3)/t decreases, so that their product ~2 3 is nearly constant. On the other hand, if t is held constant as [dCTP] is increased from ( a ) to (c), (Iz + Z 3 ) / t stays nearly constant so Z3/Z2 itself becomes a relative measure of v23. The value of Z3/Z2 at t = 4 min is used to measure relative velocity (v) for V,,, and K,,, determinations (Table 11).  If instead of adding 3 W M dATP one adds much more dGTP (3 mM), then G is misincorporated in place of A in site 3 (Figs. 2b and 3b), but at a slower rate. Comparing Fig. 3b with Fig. 3a, one sees that at each time point 1 3 is now lower but I2 is higher, such that their sum (Iz + 13) stays approximately the same. In other words, when either dATP or much more dGTP is added to 50 W M dGTP, (I2 + 13) remains constant while &/I2 changes, A similar observation is made when either a high concentration of dCTP or dTTP is added to 50 W M dGTP (Figs. 3c and 34. Kinetic Studies: u uersus [dNTPJ-The behavior of u23 is examined as a function of the concentration of dNTP substrate used for incorporation opposite T in site 3 of the primertemplate system (Fig. 1). With 50 p~ dGTP in the reaction mixture, DNA polymerase LY incorporates G opposite C in sites 1 and 2 at near maximum velocity, without causing detectible incorporation of G opposite T in site 3. At each concentration of added dNTP, u23 as found by Equation 1 remains essentially constant for t up to 8 min in the case of dATP and up to 16 min in the case of other substrates, e.g. dCTP (Table I). This result indicates that the enzyme-DNAsubstrate complex at site 3 is in steady state for such time The K, value found for misinsertion of G opposite T (4.2 mM) is 1100 times greater than the value (3.7 p~) for correct insertion of A opposite T (Table 11). In contrast, the V,,, value is less by only a factor of 4, being 3 for dGTP compared with 13 for dATP ( Table 11).
The ratio, V,,,/K,, which is the initial slope in a plot of u uersus [dNTP] (Fig. 5), measures the efficiency of nucleotide insertion by polymerase (13). The ratio of the insertion efficiencies for wrong uersus right base pairs indicates the misinsertion frequency ( f ) and the fidelity ( l / f ) of polymerase. Our results ( Table 11) show Drosophila DNA polymerase CY has f = 2.1 X for misinsertion of G opposite T in the site selected on M13 DNA (Fig. 1).
The f value for transversion mismatches C. T and T -T is about one-fourth that for the transition mismatch G. T ( Table   11). The observation of only a slight curvature in u with increasing [dCTP] or [dTTP] (Fig. 5 ) allows only rough estimates of K , and Vmax for these substrates. On the basis of Hanes-Woolf and Lineweaver-Burk plots, we estimate K, -10 mM and relative V,,, -1.7 in each case ( To carry out a direct competition assay using the gel system, we start with the same primer template configuration (Fig. 1) and use 50 p~ dGTP for primer extension to site 2, but use unlabeled primer and 0.75 p~ [ c Y -~' P ]~A T P as labeled substrate for insertion of A opposite T in site 3. Insertion of labeled A results in a band whose integrated intensity (Io) can be measured as before. In the presence of added dNTP, the total band intensity (I) diminishes as N competes with A for insertion opposite T (data not shown).
The ratio of [cx-~'P]~ATP concentration to the dNTP concentration when f = *h Io provides a measure of the misinsertion frequency ( f ) for d N T P relative to dATP. The following The relative velocity of insertion was measured as the ratio of band intensities, u = Z3/Zz, observed by gel electrophoresis after 4 min reaction time. A Hanes-Woolf plot of [dNTP]/u versus [dNTP] was fitted linearly by least-squares to determine the intercept (K,,,/V,,,d and slope (1/VmaX). Shown are the mean K, and Vmax values k S.D. found in repeated experiments with each substrate (number in parentheses). Also shown is the corresponding insertion efficiency, Vmax/K,,,, and resultant misinsertion frequency relative to dATP. The f values obtained by competition assay are shown in parentheses for comparison.   Table I1 are obtained by least squares fits to Equation 2 over [dNTP] ranges given for the competition assay under "Experimental Procedures." Within experimental error, these results agree with the corresponding evaluations from Vmax and K, measurements (Table 11).
A significant amount of data has been published comparing the insertion of the mutagenic base analogue AmPur in competition with A in different sequences (14, 15), and K, and VmaX measurements comparing AmPur with A have also been made (7). We find, using the gel assay, that V,,, values for AmPur and A are equal ( Fig. 5 and Table 11), consistent with the standard DNA polymerase assay (7). The gel assay also shows a 10-fold higher K, for AmPur than for A, in agreement with expectations when G is present as the 5"nearest neighbor base on the primer strand (16).

THEORETICAL ANALYSIS
Target Velocity-Consider the DNA primer-template system illustrated in Fig. 1. With dGTP present at 50 p~ concentration, DNA polymerase adds G to primer rapidly in sites 1 and 2, but only very slowly in site 3. The objective is to derive an expression for the velocity of insertion a t target site 3.
With [dGTP] = 50 p~, primers are extended from 0 to 2 a t close to maximum velocity, because K , for G insertion opposite C is only about 5 p~. As [dGTP] or any other [dNTP] is increased, the velocity for 0-2 extension (uo2) remains nearly constant, while that for 2-3 extension (u23) increases. To derive a general expression for uZ3, we proceed as follows.
Note that this derivation makes no assumptions about the processivity of polymerase.
Consider the time required on average to extend a primer molecule from 0 to 3, namely 703. This is equal to the time from 0 to 2 ( T~~) plus the time from 2 to 3 (723). Because times for elongation are inversely related to velocities, one can also write the following.

Rearrangement of Equation 3
leads to the following expression for the velocity at the target site.
Velocity Measurements-The reaction is started by adding a certain concentration of substrate, [SI = [dGTP] for example, to a fixed amount of enzyme and preformed primertemplate complex with 32P on the 5'-end of primer. After a short reaction time ( t ) , EDTA is added to stop the reaction. The amount of primer extension in this time period is observed by gel electrophoresis, autoradiography, and densitometry. In addition to band 0, corresponding to unreacted primer, one observes bands 1, 2, and 3. We will now show that u23, the velocity of primer extension from site 2 to 3, is propor-tional to Z3/Z2, where Z2 and Z3 are the integrated intensities of bands 2 and 3, respectively.
For primer extension from 0 to 3 the velocity corresponds to Z3 divided by reaction time.

u03 = 1 3 / t (6)
On the other hand, uo2 is given by uo2 = ( 1 2 + 13)/t (7) since every primer that has reached position 3 has also reached position 2. Substituting Equations 6 and 7 in Equation 5 , we find Equation 1 or its equivalent, Equation 8. u23 = u02(13/12) (8) Now at [dGTP] = 50 p~, uO2 is already near maximum and changes very little as either [dGTP] or any other [SI is increased. As seen in Table I for the case of S = dCTP, the value of [Z2 + Z3]/t at each t does not change significantly with increasing [SI. As long as this condition holds, the ratio Z3/Z2 is itself an appropriate relative measure of u23. Furthermore, the relative velocity, u = Z3/Z2, has the advantage of being insensitive to loading error. While Z2 and Z3 increase with time, Z, remains low and nearly constant a t all times in the 2-16 min range (Fig. 3). One expects Z, to be low if the enzyme is processive, i.e. has low tendency to dissociate from primer template. The more processive the enzyme, the lower the value of Z, should be. The near constancy of ZI at 2-3% implies that polymerase (Y is moderately processive en route to the target site. Kinetic Analysis-Initially, a t t = 0, only the original primer band is present (Io = 100%). As t increases, Zo decreases while the extended primer bands (Zl, Z2, Z3) rise to different levels. A plateau is first reached by I,, then by Z2, and finally by Z3 (Fig. 3). Because Z2 levels off before Z3, the ratio Z3/Z2 increases with t while (Z2 + I3)/t decreases. The product of 13/Z2 and (Z2 + Z3)/t is nearly constant for several minutes (up to t = 16 min as seen in Table I). This means that for several minutes the amount of substrate-enzyme-primer-template complex contributing to u23 remains in steady state (see "Appendix" for kinetic analysis). Under steady-state conditions, a Michaelis-Menten equation is expected to hold for uZ3 as [SI is increased, namely uz3 = (Idzz)(Iz Id/t v z 3 [ s ] / ( K m -t [SI) (9) where Vz3 is the maximum value of u23 and K , is the value of [SI when u 2 3 = V23/2. As long as (Z, + Z3) remains constant for a given t, we also expect u = 13/12 = Vmax[SI/(Km + [SI) (10) where V, , , = (13/12)max. The relationship between V, , , and v 2 3 is as follows.
v 2 3 = Vmax(Z2 + L ) / t  (12) recognizing that V2, remains constant in the range t = 1-16 min ( Table I).  dCTP] or [dTTP] varied, it is theoretically possible to measure V,,, and K, for either C or T insertion opposite T. In practice, however, a problem arises if K, is above 7 mM, because above this concentration substrate inhibition becomes significant. If we limit [SI to 7 mM, we also limit our ability to determine V,,, and K,. However, the ratio V,,,/ K,,, can still be determined from the initial slope of u uersus [SI (Fig. 5).

DISCUSSION
We have developed a quantitative assay to measure DNA polymerase insertion kinetics at any specific template site by using polyacrylamide gel electrophoresis to resolve primers extended by polymerase. The procedure makes it possible to evaluate nucleotide insertion velocity by simply measuring the intensity ratio of adjacent bands in a gel lane. Measurements at different substrate concentrations allow relative Vmax and K,,, values for correct and incorrect base pairs to be determined with natural or synthetic DNA templates and normal or analogue dNTP substrates. The present study focuses on enzymatic mechanisms underlying DNA polymerase insertion fidelity, using highly purified Drosophila DNA polymerase a. In future work, we intend to use the same approach to analyze the kinetic basis for base substitution mutational hot spots in DNA.
Before discussing experimental results, let us examine theoretical mechanisms by which a DNA polymerase ( E ) may discriminate between competing nucleotides for insertion opposite a given DNA template site. The enzymatic steps leading to the extension of primer DNA from n to R + 1 nucleotides by insertion of either a complementary (right) or noncomplementary (wrong) dNTP substrate may be described simply as follows (see "Appendix" for a more comprehensive treatment). , is the total concentration of enzyme-DNA and enzyme-DNA-dNTP complex involved in the reaction. As a function of the concentration of enzyme-DNA and dNTP, the velocity is as follows (13).
Suppose two dNTP substrates, one right ( r ) and the other wrong (w) are competing for reaction under the same conditions. If u(r) and u(w) are their respective velocities of insertion, the misinsertion frequency ( f ) is given by the following.

Km (14)
In other words, discrimination between right and wrong dNTP substrates can occur theoretically because of K,,, and/ or V,,, differences (8,9). Base mispairs are expected to show a higher K,,, because of their greater tendency to dissociate and lower VmaX because of their less favorable geometry. Normal base pairs are stabilized by hydrogen bonding between complementary bases and stacking interactions with neighboring bases in B DNA geometry. Base mispairs are less stable because of lack of complementary and unfavorable geometry for stacking. Hence, a mispair between a noncom-plementary substrate and template target base should be relatively unstable, rapidly dissociating to release the mismatched dNTP from the enzyme-DNA complex. A higher tendency to dissociate means a higher dissociation rate constant k-, and therefore a higher K, value. An enzyme in which misinsertion frequencies are governed primarily by relative K , values for wrong versus right dNTPs, i.e. K,,, discrimination, achieves fidelities limited by the differences in dissociation energy for correct and incorrect base pairs in the environment of the active site (6,9,10).
Once dNTP is bound in the active site, a second discrimination step can occur based on geometrical constraints that affect the rate of phosphodiester bond catalysis and thereby alter the rate constant kc,,. Binding of a noncomplementary dNTP might occur in an unfavorable geometry. An enzyme that has very rigid geometrical constraints in the active site is capable, in principle, of attaining a higher level of fidelity than is possible by K, discrimination alone (9). Such an enzyme should exhibit a marked decrease in VmaX for wrong dNTPs.
Using the gel assay for the template T site on M13 DNA ( Fig. l), we compare the correct insertion of A and misinsertion of G, C, and T by highly purified Drosophila polymerase a. A misinsertion frequency of 2.1 X is found for G and 5 X for C and T (Table I). These low f values are principally the result of high K , values for wrong base pairs. Compared with A . T (A insertion opposite T ) , G . T shows a 1100-fold higher K, but only a 4-fold lower V,,,. Similarly, C.T and T.T both show a 2600-fold higher K, but only an 8-fold lower VmaX. These results indicate that K, , , discrimination is predominantly the mechanism determining the fidelity of DNA synthesis by Drosophila polymerase a.
The large K,,, values for wrong nucleotides are attributed primarily to a large increase in the dissociation rate constant (k-l) in Equation 13. A high k-, means a low affinity between substrate base and template base. For wrong bases, k-l is much larger than kc,,, so that K, is very close to the dissociation equilibrium constant k-,/k1. Even for a right base, such as A opposite T, k-, is likely greater than kc,, (6), so that K, is not much different from the dissociation equilibrium constant. The binding energy for a wrong base at the polymerase active site is less than that for a right base by the amount, RT ln[K,,,(w)/K,,,(r)], From the K,,, values found experimentally (Table I), we find t h a t t h e G -T mispair is 4.3 kcal mol-' less stable than the normal A. T pair, whereas C . T and T.T mispairs are 4.9 kcal mol" less stable. These free energy differences are 2-4 times as large as those indicated by melting temperature differences between DNA oligomers containing matched uersus mismatched base pairs in aqueous solution (17). We have proposed that amplifications of this magnitude may be explained by the partial exclusion of water at the polymerase active site (10). The geometry of G . T mispairs has been shown by NMR (18) and x-ray crystallography (19) to be that of a wobble base pair, which differs slightly in dimensions from a normal A . T pair. The influence of this steric difference on catalysis (kcat) could be responsible for the 4-fold reduction in V,,, that we observe with polymerase a. Whether or not other enzymes are catalytically more sensitive to differences in base-pair geometry remains to be seen. In any event, the large difference in stability in the binding site seems to be the main reason why polymerases select A -T over the mismatches (G.T, C.T, and T.T) for incorporation into DNA. The latter two mispairs are expected to be less stable and more distorted than G .T, consistent with our finding of a further 2-fold increase in K, and %fold decrease in V, , , ( Table 11).
Data obtained by our gel assay seem consistent with results obtained using a standard DNA polymerase assay in which incorporation of radioactive nucleotides into acid-insoluble DNA is measured by liquid scintillation counting. First, for Drosophila DNA polymerase a complex (consisting of at least three subunits including primase), K,,, for the insertion of A opposite T was reported to be 3.7 /IM (4), the same as obtained here (Table 11). Second, we have previously shown with polymerase from KB cells (8) and also with bacteriophage T4 polymerase (7) that insertions of base mispairs involving the mutagenic base analogue AmPur are governed by K,,, discripination in agreement with gel assay results for Drosophila polymerase a (Table I1 and Fig. 5 ) .
An estimate of the absolute velocity for inserting A opposite T can be deduced from the kinetic model ("Appendix"). Our estimate of kc,, = 2.2 nucleotides/s/enzyme is similar to a value of 2 nucleotides/s/enzyme previously determined on singly primed 4x174 DNA for the same preparation of Drosophila polymerase a (20).
The misinsertion frequency of 2.1 x found for G.T, relative to normal A. T pairs, implies that Drosophila a polymerase-primase complex exhibits an accuracy similar to a polymerases purified from other organisms. Loeb and Kunkel (21), using synthetic polynucleotide templates, report misinsertion frequencies (G.T) in the range 1-3 x for various a polymerases.
In another type of fidelity assay, a natural DNA template such as 9x174 or M13 is copied by DNA polymerase i n uitro and then transfected into E. coli, where single-site revertants (22) or different classes of forward mutants (23) are selected. Drosophila polymerase a has been observed to revert the am3 site on 4x174, by forming A -A or G -A mispairs, at a frequency of about 2 x lO"j, similar to the reversion frequency by E. coli polymerase I11 holoenzyme complex (20). The polymerase I11 holoenzyme contains a 3' -5' proofreading exonuclease activity (24, 25), whereas there is no detectable exonucleolytic activity found in the intact polymerase-primase multisubunit complex of Drosophila polymerase a (4,12). The fidelity of Drosophila polymerase inferred from the 4x174 transfection assay seems to be about 25-fold greater than indicated in the gel assay. Although this difference in fidelities might be attributable to higher probabilities to form the mispairs C . T or T. T measured here as opposed to A. A or G-A (20), we suggest another reason for possible discrepancies in the two assays.
The reversion frequency in the 4x174 transfection assay is determined by the product of two probabilities, the nucleotide misinsertion probability multiplied by the probability of primer extension beyond the misinsertion site to generate a nonrepaired, viable DNA molecule. Primers having a misinserted 3"terminal nucleotide are less likely to be extended efficiently than primers having properly paired termini. Thus, error frequency measurements based on the transfection assay may be lower than "true" nucleotide misinsertion values, because DNA strands containing a mismatch may either be repaired or elongated inefficiently following transfection.
Finally, it is important to verify whether misinsertion ratios based on Vmax and K , values of individual dNTP substrates are similar to those obtained under conditions where competing wrong and right dNTPs are simultaneously present in the assay. In a competition experiment, incorporation of hot nucleotide ([a-"PIdATP) opposite template T is measured by extending an unlabeled primer and evaluating the band intensity of the extended primer by gel electrophoresis and autoradiography. Increasing concentrations of cold dNTP are then added to compete with hot dATP for insertion at the target T site. The band intensity is reduced by a factor of 2 when a competing dNTP is misinserted opposite T at a rate equal to the correct insertion of A. Misinsertion frequencies obtained from the competition experiment are in agreement with those found from the Vmax and K , values (Table 11). APPENDIX A kinetic model for the elongation of primer from site 0 to site 3 (Fig. 1) is desired to describe velocity u23 in terms of kinetic constants for individual reaction steps. The model is required to include each polymerization reaction and to allow for dissociation of polymerase from primer-template and yet be amenable to solution by the steady-state approximation. The following simple model is proposed for this purpose: where E represents DNA polymerase devoid of exonuclease proofreading activity. Do is the original primer bound to template DNA; Dl, D2, and D3 are primers extended by the addition of 1, 2, and 3 nucleotides, respectively. ED,,, ED1, and ED, are the corresponding enzyme-DNA complexes. EDJV,, EDIN,, and ED,N3 are the enzyme-DNA-substrate complexes with substrates N1, N2, and N3, respectively. In our experiments, Nl and Nz are dGTP, and N 3 is either dATP (right substrate) or dGTP, dCTP, or dTTP (wrong substrate). The side pathways (downward pointing arrows) leading to dissociation of complexes EDJV1, EDIN,, and ED2N3 with rate constants kd, kc, and k.?, respectively, are assumed to be irreversible. The reverse association requires the simultaneous interaction of three components, an unlikely event. Reassociation of polymerase and extended primer (Dl and D2) can occur via ki and k: (upward pointing arrows).
The velocity of primer extension at the target site is ~2 3 = kc,, [ED2N3]/[DIc, expressed as a fraction of total primer [Dl,. A general expression for the steady-state concentration of ED2N3 is Some of the terms in Equation A2 are small enough to be neglected. Because the substrate for insertion in sites 1 and 2 is correct (N1 = N2 = dGTP), dissociation constants kd and kc are very small in relation to catalytic constants kf and kg for a processive enzyme. Furthermore, [dGTP] is set high enough (50 p M ) to approach the maximum rate of incorporation in sites 1 and 2, so that K F / [ N I ] and KG/[N2] are each <<1. Also, for short time periods, the concentration of extended primers (Dl and D2) is very small in relation to original primer (Do)-Other terms are sufficiently similar to be equated. Because association of polymerase with Do, Dl, and D, occurs with similar rate constants, we can set k i = k: = KO. Similarly, for dissociation, k; = ki = k 0 . Furthermore, because the same substrate is incorporated in positions 1 and 2, kg = kp A further simplification can be made by setting kc,, + kg = k,. This can be justified as follows. For correct insertion at site 3 (N3 = dATP), kc,, is similar to k!, whereas ks is very small in relation to kf. On the other hand, for incorrect insertion (N3 = dGTP, dCTP, or dTTP), k, should rise as kcat falls. The rise in k,, resulting from a less stable enzyme-DNAsubstrate complex, may compensate the fall in kc,.,, resulting from the less favorable geometry. Hence, kcat + k, may be close to kf even for wrong nucleotides in position 3.
The resultant simplified version of Equation A2 is given here.
This leads to a Michaelis-Menten equation for uz3 as follows: where VZ3 is the maximum velocity, and K M app is the apparent Michaelis constant, Note that KM app is identical to K , when k-o = kf (moderately processive enzyme) but is less than K , when k 0 << kf (highly processive enzyme). In the case of DNA polymerase a, a moderately processive enzyme (26), we expect k o to be comparable to kf, in which case the observed K,,, should be close to the true K,. For the insertion of A opposite T, the K , observed by gel assay with Drosophila DNA polymerase a is the same as reported (3.7 p~) by conventional assay (4).
In Equation A l l for the maximum velocity ( Vz3), the term

kcat [EI1/[D], is divided by the factor (3 + k f / b [ D o ] ) .
This factor has a minimum value of 3, when enzyme is saturated with primer, i.e. [Do] >> k f / b . For lower primer concentrations, the factor will be larger.