Nearest Neighbor Influences on DNA Polymerase Insertion Fidelity*

The kinetics of forming all possible single base sub- stitution errors are measured for Drosophila melanogaster DNA polymerase a and avian myeloblastosis virus reverse transcriptase. Seventeen sites along bacteriophage M13 DNA are investigated so that effects of nearest neighbor base stacking on misinsertion kinetics can be evaluated. Polymerase a appears to be more error prone than reverse transcriptase. Polymerase a forms transversion mispairs at rates compa- rable to transition mispairs with two exceptions; A*A and C*C are formed with significantly higher and lower efficiencies, respectively. Reverse transcriptase forms transversions with lower efficiencies than tran- sitions, especially low being A*G, G-G, and C*C. For both enzymes, misinsertion frequencies vary typically by 10-fold for the same mispair in different locations. Misinsertion frequency can be expressed as a product of two components, one based on K , and the other on Vmm. DNA polymerase a appears to use primarily K,,, discrimination (100-5000-fold) to achieve insertion fidelity while reverse transcriptase shows a greater balance between K , and V,, discrimination. Nearest- neighbor base stacking interactions appear to have opposite effects on the two discrimination components. The 5”nearest neighbor influence on K , is greater for correct insertions than for incorrect, while the influence on Vmar is greater for the incorrect base. Target sites that have pyrimidine as the 5”nearest neighbor to incoming nucleotides show a higher than average misinsertion component based on K,, but a lower than

locus (1). DNA slippage is an example of a physical event wherein displacement of DNA strands relative to each other can serve as a source of single base substitutions (2) as well as frameshift mutations (3,4).
Enzymatic mechanisms leading to nonrandom error accumulation include error correction by mismatch repair enzymes and proofreading by 3' + 5' exonuclease activities. In the case of methylation-directed mismatch repair (5), the enzymes involved in repair exhibit different efficiencies for eliminating different mismatches. In general, transitions are corrected more efficiently than transversions, and within a specific mismatch class there appears to be sequence specificity (6, 7). Proofreading by 3' + 5' exonuclease (8,9) is another example of a sequence-dependent error correction process (10,11). These enzymes appear to act on unstable regions of duplex DNA; hence, they are less efficient in excising errors in G-Crich regions of duplex DNA (11,12).
The initial enzymatic step in error induction is misinsertion of a nucleotide by DNA polymerase. Nucleotide insertion errors occur nonrandomly in the DNA by mechanisms that are not well understood. In this paper, we investigate criteria governing site specificity of nucleotide insertion by comparing two very different polymerases, avian myeloblastosis virus reverse transcriptase and Drosophila rnelanogaster DNA polymerase a. Two basic questions are addressed. First, what are the qualitative and quantitative differences in the efficiencies of misinsertion at identical DNA sites? Second, for each polymerase, how are misinsertion efficiencies affected by the types of mismatch and by different nearest neighbors? EXPERIMENTAL PROCEDURES Materials-Purified D. melanogaster DNA polymerase CY holoenzyme, lacking detectable exonuclease activity and consisting of at least three subunits including primase (131, was a generous gift of Dr. time over which products accumulate linearly as a function of time. Two kinds of primer extension are examined (see Fig. 1). One, called extension from a standing (s) start, is obtained with only one dNTP substrate; the other, extension from a running (r) start, is obtained with up to two dNTP substrates. In an s-start experiment, nucleotide insertion is examined at the first template site adjacent to the original primer (position 1). In an r-start experiment, nucleotide insertion is examined at position 3, one dNTP being used a t "saturating" concentration for correct insertion into positions 1 and 2 and either the same dNTP or another at various concentrations for insertion into position 3.
DNA Polymerase Reactions-For each r-start kinetic study, an appropriate "saturating" concentration of correct dNTP is used, i.e. the highest concentration enabling insertion into positions 1 and 2 without significant insertion into position 3. Also, a reaction time is used during which the rise in intensity of the extended primer is linear with time and less than 20% of the original primer band (lo) is extended.
Each polymerase reaction is started by mixing equal volumes of Solution A containing primer-template-enzyme complex and Solution B, containing dNTP substrate(s), followed by incubation a t 37 "C. For Pol 01 kinetic studies, in accordance with previous work (16), incubation times of 3 min are found satisfactory for s-starts, and 8 min for r-starts. For AMV RT, shorter reaction times are necessary to avoid significant depletion of original primer; 1 and 2 min are used for s-and r-starts, respectively. For Pol 01 studies, Solutions A and B are prepared as described previously (15,16 AMV RT reactions for determining kinetic constants are begun by mixing 3-pl aliquots of Solutions A and B and incubating at 37 "C for 1 min (s-start) or 2 min (r-start). The 6-pl reaction volume contains 0.6 unit of AMV RT (specific activity, 50,000 units/mg, 1 unit being defined as the incorporation of 1 nmol of [3H]TTP into acid-insoluble material in 10 min at 37 "C). The ratio of primer-template to enzyme is 3:l. Reactions are stopped by adding 12 pl of 20 mM EDTA in 95% formamide.
Gel Electrophoresis, Autoradiography, and Densitometry-Samples of reaction mixture are heat-denatured at 100 "C for 5 min, cooled on ice, and 4-pl aliquots are loaded on a 16% polyacrylamide gel containing 8 M urea. Gels (40 X 30 X 0.04 cm) are run at 2000 V for 5 h to obtain the best resolution of elongated primers. Autoradiographs of gels were made and scanned as described (15,16).
Velocity Determination for a Standing Start-The velocity of nucleotide incorporation when extending the primer (band 0) by inserting a nucleotide opposite the first template base (band 1) is expressed in terms of the integrated gel band intensities, after reaction time t, as uol = (100 I1)/(1o + 0.5 I&, where t is in minutes (17). Typically, no more than 15% of total primer is used in the reaction.
Velocity Determination for a Running Start-The expression for the velocity of nucleotide incorporation opposite the DNA template target site (band 3) is given by the equation u23 = ( 1 3 / 1 2 ) . (13 where I2 is the integrated intensity of band 2 ahead of the target band (16). The value of (12 + 13)/t is approximately constant at all dNTP substrate concentrations; thus the ratio 13/12 is an appropriate relative measure of uz3. To allow a direct comparison to be made with s-start reactions, the relative VmaX (maximum value of 13/12) is multiplied by the constant value of (Z3 + Z2)/t, expressed as a percent of total primer/min.

RESULTS
To investigate the biochemical basis for base substitution "hot" and "cold" spots, we measure the kinetics of primer extension at selected sites by electrophoretic assay of extended primers (15,16). Here we report the results of studying the misinsertion properties of two polymerases, Drosophila Pol (Y and AMV reverse transcriptase, at 17 distinct sites on M13 DNA. The sites are chosen to explore nearest neighbor influences on misinsertion kinetics. Kinetic studies are made at each site to determine the relative efficiencies of inserting matched and mismatched nucleotides. Thirteen of the sites are labeled "r" for ''running'' start ( Fig. la), since polymerase must add two nucleotides to a. Running Start dGTP dNTP 5' "~-pr~rner-A / . . . . template C C T . . . . . . I primer before reaching the template target site; four are called "s" for "standing" start ( Fig. lb), since the insertion efficiency is determined at the site immediately adjacent to starting primer.
Kinetics of Misinsertion-When two nucleotide substrates compete for insertion at the same site, one being a proper match (right) and the other a mismatch (wrong), the misinsertion efficiency is defined as (1) where u(W) and u(R) are the insertion velocities for wrong and right nucleotides, respectively, at equivalent concentrations. The value off can be determined by allowing a wrong substrate to compete against a right substrate, and measuring the ratio of wrong to right insertions directly using different radioactive labels in the two substrates (19). Alternatively, by using labeled primers and unlabeled substrates, we can evaluate f from kinetic measurements for each substrate without having the two substrates present simultaneously (16). Kinetic measurements provide a more precise determination of f than competition measurements when f is small.
To evaluate f, we measure the velocity of insertion as a function of dNTP concentration for each nucleotide separately and determine apparent VmaX and K, in accordance with Michaelis-Menten kinetics. The f value is determined from the ratio of apparent V,,,.,/K, values for wrong and right dNTPs.
For each dNTP we use a linear least squares fit to a Hanes-Woolf plot (y = [dNTP]/u uersus x = u ) to obtain K,/Vm,, as the y-intercept and l/Vmax as the slope. Individual V, , , and K, values are not needed to obtain f or fidelity (l/f), only their ratio; however, both parameters are evaluated in order to relate fidelity to discrimination mechanisms (20-22). No attempt is made to extrapolate velocities to infinite DNA concentration, so only apparent rather than true kinetic constants are reported (Table 11). However, it is straightforward to show that for an ordered bisubstrate reaction of the type expected for DNA polymerase (23,24), the f value does not depend on DNA concentration.
Although V,,,/K, for each substrate does depend on DNA primer-template concentration, the dependence is the same for numerator and denominator in Equation 2 and thus cancels from the ratio at equivalent DNA concentrations.
Differences in Misinsertion Efficiency-Significant differences are observed in the efficiencies of mispair formation by the two polymerases at most of the target sites examined. The differences are evident in gel autoradiograms showing the misincorporation of dAMP opposite A at two different sites ( Fig. 2a) and misincorporation of dGMP opposite G at two different sites (Fig. 26). Pol CY is seen to form A. A mispairs a t sites p2A and p14 (Fig. 2a, left side), while AMV R T shows detectable formation of A.A mispairs only at site p2a ( Fig.  2a, right side). In the case of G.G mispairs (Fig. 2b), Pol a shows higher misinsertion rates a t site p5 than at p8, while AMV RT makes almost no detectable G . G mispairs at either site. Fig. 3 shows the range of misinsertion efficiency (f values) found for Pol a and AMV R T for all possible mispairs at the target sites. The base mispairs are arranged from left to right, starting with transition mispairs in order of decreasing f value, followed by transversion mispairs. In general, Pol n shows higher f values than AMV RT, i.e. Pol a shows lower fidelity. Both polymerases are influenced by neighboring bases as seen in Fig. 3 by the bars indicating f value ranges for mispairs at different sites. The largest differences in f for the two polymerases are found for transversion mispairs of type A. A, A. G, and G. G. In these cases, the Pol n values are higher by almost two orders of magnitude. Differences in K , and V,,,,,-A compilation of kinetic data for AMV RT and Pol a is given in Table 11. Apparent K, and V,,, and resultant misinsertion ratio (f value by Equation 2) are shown for each possible insertion (one right and three wrong) at the various template sites. Consider first the kinetics of correct insertions, which forms normal base pairs. Apparent K, values vary by more than a  factor of 100 in the case of AMV RT and somewhat less in the case of Pol a; the lowest K , values appearing in s-start sites. For example, the K , for insertion of A opposite T (Table  11, a) is lowest for AMV RT (0.09 p~) and for Pol a (0.28 p~) at the s-start, pl7T-X. The comparable r-start site, pl5-TTX, gives a much higher K , for X.T = A.T for both AMV RT as an s-start, K , is about 40-fold higher for AMV RT and 5fold higher for Pol a. A similar result is obtained for other correct insertions as well (Table 11, b, c, and d).
For each insertion, we have examined only one s-start site but at least three r-start sites. The latter provide different 5'nearest neighbors for the incoming nucleotide, allowing us to explore nearest neighbor influences on K , and V,,,,.
In r-start sites, we see that K , for correct insertion varies more from site to site than does K , for  ( I 1 :  (Table 11, b). The r-start sites that show high K , values for correct insertions tend to be the same for both polymerases, the sole exception being the insertion of C opposite G (Table 11, b), where the trend is reversed and the highest K , site for AMV RT is now the lowest for Pol a (site p8b-TTX).
Kinetic Components of Fidelity-The kinetic components, K , and V,,,, make separate contributions to f, namely fK and fv, respectively, as seen in Equation 3. The f value is the product of f~ and fv, each having values less than or equal to 1. The reciprocals off^ and fv are the K , and Vmax components of fidelity. The smaller the values of fK and fv, the larger the contributions to fidelity; values close to 1 giving practically no contribution.
In the case of Pol a, K , values for inserting a wrong nucleotide are 100 to 3,000 times greater than for the right nucleotide, while Vmax values are only 2-50 times lower (Table  11). Therefore, f~ is much smaller than fv for every mispair studied, i.e. Pol a fidelity is determined much more by K , discrimination than by Vmax discrimination. At site p15-TTX (Table 11, a), for example, the Pol a f values for X -T = G-T, C-T, and T.T (1.7 x 1.9 x and 1.0 x respectively) correspond to a fidelity of 600,5,300 and 10,000, respectively. For G T, K , discrimination contributes a factor of about 120 to fidelity while Vmax discrimination contributes only a factor of 5. Similarly, for C.T and T-T, K , discrimination factors are 450 and 360 while V,,, discrimination factors are only 12 and 27, respectively.
The properties of AMV RT differ significantly from those of Pol a. While showing a comparable K , component of fidelity at all sites, AMV RT also shows a large V,, component at many sites. For example, at site p7-CCX (Table 11, a), for insertion of X = G, C, or T opposite template T, K , is 100-150-fold greater than for insertion of A opposite T.
However, in contrast to Pol a, the V,, discrimination approaches K , discrimination, 80-and 150-fold, in the cases of C .T and T-T mispairs, respectively. At site p9-TTX (Table  TABLE 11 Relative K,,, and Vmax values for nucleotide insertions by AMV reuerse transcriptase and Rrosophila DNA polymerase LY at various DNA primer-template sites
Nearest Neighbor Effects on Kinetic Components-In order to determine the nearest neighbor influence on fK or fv at rstart sites, we evaluate the geometric mean of the component for each mispair at different sites and then evaluate the ratio for each site with respect to the mean. For example, consider fK for a specific mispair. First we determine its geometric mean value (_fK) in all the sites examined. At each site, the ratio of fK to f is either greater or less than 1. Values greater than one indicate the incorrect base is misinserted more efficiently, values below one, less efficiently. By plotting the ratio, fK/fK, for each mispair with respect to the 5"nearest neighbor at the insertion site (Fig. 4), any systematic nearestneighbor influence on the K, component off can be seen. A similar plot can be made for fv/fv to show the influence on the V,,,,, component.
The plot for AMV RT (Fig. 4A) shows that fK tends to be lower than the mean when the primer 5"nearest neighbor base is a purine and higher when a pyrimidine. The same trend is seen for transition mispairs (solid circles) and transversion mispairs (open circles). Conversely, fv is higher when a purine is 5"nearest neighbor and lower when a pyrimidine (Fig. 4B). Thus, upon going from purine to pyrimidine neighbors, AMV RT shows a distinct upward trend for the K, component off and a downward trend for the V,,, component.
Pol a shows less of an influence of 5"nearest neighbors, but the trends are similar (Fig. 4, C and D ) . The K, and Vmax components off generally range no more than 5-fold from the mean, compared to 10-fold or more for AMV RT. There is an upward trend in the K, component on going from 5"nearest neighbor G to A to C and a downward trend for the Vmax  (Table 11) as fK = K,,,(R)/K,(W) and fv = VmaX(W)/Vmax (R), where R and W refer to right and wrong insertions at the site. Foceach misinsertion at different sites, geometric msan values ( f K and fv) are determined and the ratios, f K / f K and fv/fv, are evaluated. At sites having the same 5"nearest neighbor, the ratios tend to be similar, as shown by the average values and standard deviation (error bars) for misinsertions leading to transitions (0) and transversions (0). WhenJhe 5'nearest neighbor is base G , the fg falls below the mean ( f~/ f K < 1) while fv rises above the mean (fv/fv > 1). This reciprocal relationship between K, and V,,, components is much more evident for AMV RT than for Pol a.
there is a reversal relative to C, largely because of the low K, and Vmax values for correct insertions at sites p8-TTX and p8b-TTX (Table 11, d).
It is important to note that although the ratio of fK or fv to the corresponding mean is generally between 0.1 and 10, the value of the mean for each component is different for each mispair for each polymerase. The mean K , discrimination values ( l / ?~) range 80-1,500 and 190-1,000 for AMV RT and Pol a, respectively. The values of l/Fv, on the other hand, range 5-680 for AMV RT and only 3-23 for Pol a. The ranges in means are consistent with the observation that Pol a uses mainly K, discrimination to select against incoming mispairs, and that AMV RT shows much more V,,, discrimination than Pol a while maintaining a similar K, discrimination.
Misinsertion Hot Spots and Cold Spots-The ratios (fK/fK and fv/fv) for each 5"nearest neighbor show a dependence on ?K and ?v. Based on a least squares fit to our data, this depsndence is described approximately by fK = (fK)'+' and fv = (fv)"', where c is a constant for each neighbor. The fK and f v values obtained from the data in Table I1 are shown for each mispair in Table 111. By plotting log fK uersus log rK and log fV uersus log ?V for all mispairs at sites having the same 5"nearest neighbor, we evaluate c from the slope of the line obtained by linear least squares fit. Such plots (data not shown) yield the following results for AMV RT: c = 0.  (Table 111) can be predicted within a factor of two (on average) using the equation, In Fig. 5 , we illustrate the corresponding nearest neighbor properties of misinsertion hot spots and cold spots. The nearest neighbor influences have the opposite effect in the two extreme cases, when K , discrimination is domjnant ( f~ << fv) and when VmaX discrimination is dominant (fv << f K ) .
Misinsertion hot spots occur next to 5"nearest neighbor pyrimidine when K, discrimination is dominant and next to purine when VmaX discrimination is dominant. When K, and VmaX are nearly equal, there is practically-no d_ependence off on nearest neighbors, as seen by setting fK = f v in Equation

DISCUSSION
The frequency of base substitution mutations (transitions and transversions) are controlled in vivo by various enzymatic mechanisms, such as misinsertions by polymerase and removal of mismatched nucleotides by exonuclease acting at the replication fork. There are also postreplication repair enzymes that recognize errors in newly synthesized DNA ( 5 ) . To analyze the initial step, misinsertion by polymerase, we use a gel assay recently developed in our laboratory (15,16) to measure the kinetics of misinsertion for all possible base mismatches, Pur. Pyr (transitions) and Pur. Pur, Pyr . Pyr (transversions).
Here we analyze the kinetic data for two polymerases, Drosophila Pol a and AMV RT, having no detectible exonucleolytic activity.
A main objective of this work is to examine kinetic components at a large enough number of sites to find relationships between misinsertion hot and cold spots and nearest neighbor base stacking partners. By comparing the data for the two polymerases at the same sites, we can distinguish general trends from those specific to each polymerase.
An initial point to note is that Pol a has a higher misinser- tion tendency than AMV R T (Fig. 3). Although it is generally believed that transversion mispairs are more difficult to form than transition mispairs, the highest misinsertion rate for Pol (Y appears to involve insertion of dAMP opposite template A. On average, Pol 01 makes transversion mismatches at rates similar to transitions; mean error rates at different sites range from 4 x for A. A mispairs to about 1 X for most of the other mispairs (Fig. 3). The sole exception is the C.C mispair which occurs at significantly reduced rates (<lo+).
The viral reverse transcriptase shows a much lower tendency to form transversion mispairs (Fig. 3). AMV R T forms A.A and A. G mispairs a t about lo-fold lower efficiency than Pol a and G. G at about a 100-fold lower efficiency. C. C mispairs are barely detectable ( f < lo+) for AMV R T in running start primer-template configurations. Although earlier reports, suggesting that AMV R T has lower fidelity than Pol 01 (25-28), are not in agreement with our data, a recent study using a transfection assay system (29) shows AMV R T to be as accurate as Pol 01 from human and calf thymus. While finding that AMV R T exhibits a higher insertion fidelity than Pol a f r o m Drosophila, we also find that AMV RT eitends preparation. mismatched primer termini more efficiently than Pol cy for most mispairs.' Our study includes standing starts as well as running starts. These differ in that the site at which fidelity is measured is located either immediately adjacent to original primer terminus or three nucleotides downstream, respectively. A comparison of nucleotide insertion fidelity at s-and r-starts is of interest for two reasons. First, it has been reported that for mouse DNA Pol a, E. coli DNA Pol I, and human immunodeficiency reverse transcriptase on synthetic templates, chain termination is more likely after the first nucleotide addition than after subsequent additions, suggesting a difference in binding of enzyme to primer-template for initial and subsequent incorporations during processive synthesis (24,30,31). Second, in the "energy relay" model for DNA polymerase fidelity, the first nucleotide inserted at the primer terminus (standing start), is predicted to be added with lower fidelity than all subsequent nucleotides during processive DNA synthesis (32). Pol a and AMV R T can be characterized as moderately processive; Pol a adds an average of about 10 nucleotides and AMV R T slightly more before dissociating from primer-template (33, 34).
Our data (Table 11) indicate no striking differences in overall fidelity between standing and running starts. However, there are significant differences in the individual kinetic components of fidelity. Apparent K,,, and V,,, values for correct nucleotides are both lower at s-starts, up to 50 times lower than at an r-start with the same 5"nearest neighbor (Table 11). Since, in an r-start configuration, the polymerase must traverse two template sites before reaching its target site, higher K,,, values associated with such starts may reflect a need for higher dNTP concentrations to achieve processivity, i.e. to avoid dissociation of enzyme from primer-template during translocation. By avoiding dissociation, the polymerase reaches higher V,,, values at r-starts than at s-starts. Since both V,,, and K,,, are increased, their ratio does not L. Mendelman, J. Petruska, and M. F. Goodman, manuscript in change much; hence, insertion efficiency ( Vmax/Km) and resultant fidelity are little affected.
In r-starts, the two polymerases show differences in the degrees of K, and V,,, discrimination at different sites. pol a appears to have much less V,,, discrimination a t all of the target sites investigated. The K, for inserting a wrong nucleotide is generally larger than for a right nucleotide by about 2-3 orders of magnitude for both polymerases. However, for Pol a, V, , , for insertion of a wrong compared to right nucleotide is usually less by factors of only 3-6 for transitions and 10-20 for transversions (Table 11). Previous kinetic data based on an entirely different assay system reached similar conclusions concerning Pol a fidelity (20)(21)(22). In contrast to Pol a, AMV R T has V,,, discrimination that even exceeds K,,, discrimination at some sites. Since its K, discrimination is similar to that of Pol a and its V,,, discrimination is higher, AMV R T has a higher fidelity overall.
An important reason for investigating each base mispair at several different primer-template locations is to ask whether local DNA sequences can influence nucleotide misinsertion rates and whether these effects might be similar or different for the two polymerases. An analysis of the data in Table I1 reveals the existence of an interesting inverse correlation between K, and V,,,,, components of misinsertion efficiency ( Fig. 4). Consider all transitions (Fig. 4, closed circles) and transversions (open circles) for each nearest neighbor base stacking partner located on the primer strand adjacent to the target site. Target sites that have purine as the 5"nearest neighbor to incoming nucleotides show a lower than average K, component of misinsertion ( f~/ f~ < 1) but a higher than average v,,, component (fV/?v > I). Conversely, target sites with pyrimidine as 5"nearest neighbor have a higher than The correlations shown in Fig. 4 and described empirically by Equation 4 imply that nucleotide misinsertion hot spots will occur next to pyrimidines when K, discrimination is dominant as illustrated in Fig. 5. This result is supported by experiments of Pless and Bessman (10) using bacteriophage T 4 polymerase to measure incorporation of the mutagenic analogue 2-aminopurine deoxynucleotide in competition with dATP opposite T a t 57 sites in $X174 DNA. The T4 enzyme shows only K, discrimination when utilizing the Z-aminopurine substrate (20,35). After taking into account the effects of the 3'-exonucleolytic activity of T4 polymerase, it is found (11) that hot spots for 2-aminopurine insertion occur next to T and C and cold spots next to A and G, in agreement with present data for K, discrimination in Fig. 4, A and C.
As seen in Fig. 4, the nearest neighbor influences on K, and V, , , discrimination are considerably greater for AMV RT than for Pol a. However, because the two components are countercorrelated in response to nearest-neighbors, the components tend to cancel each other, thereby reducing the dependence of misinsertion efficiencies on nearest neighbor. As shown in Fig. 3 and Table 111, the range off values for each mispair from hot spots to cold spots is usually less than a factor of 10 for each polymerase.
It has been proposed that the relative rates of nucleotide insertion are proportional to the residence times of dNTP substrates in the active site of polymerase (21, 35,36). The dNTP residence times, which reflect the stability of polymerase-DNA-dNTP complex, should be sensitive to short range (base stacking) interactions between primer-template and dNTP. According to a simple model for polymerase fidelity (35,36), the misinsertion ratio, f, is determined primarily by free energy differences between correct and incorrect base pairs in the polymerase active site.
Base pairing free energy differences, which are reflected in the K, component (-RT In f~) , have contributions from at least two kinds of interactions, hydrogen bonding and base stacking. Hydrogen bonding interactions occur between substrate base and template base, and stacking interactions (Van der Waals and hydrophobic) between substrate base and primer base on the 5'-side of substrate. Both types of interactions appear stronger in the active site of the polymerase than in aqueous medium, probably because of exclusion of competing interactions with water molecules (17,37).
The stacking interactions between bases in aqueous medium are measured in terms of the melting temperature (T,) of base pair nearest neighbor doublets in DNA. For correct base pairs (X.Y) in DNA, the T , value is observed to depend on the 5"nearest neighbor of X as follows: GX > AX > CX > T X (37). This trend supports our results for K,,, discrimination by AMV R T (Fig. 4A) and to a lesser extent, by Pol a (Fig. 4C). When G is 5"nearest neighbor base stacking partner, we find that K, for correct insertion is lowest, indicating that the enzyme-primer-template-substrate complex is most stable. A rise in K, upon going from nearest neighbor purine to pyrimidine (Fig. 4A) indicates a drop in stability of the complex, in keeping with the observed drop in T,. Since K, values for incorrect nucleotides are less dependent on nearest neighbor (Table 11), probably because of less favorable stacking, the nearest neighbor influence on K, discrimination is largely the result of stacking influences on correct base pairs. A somewhat surprising finding is that V,,, discrimination shows an opposite correlation with nearest neighbors (Fig. 5). When a purine is the 5"nearest neighbor, Vmax is not reduced as much for incorrect nucleotides relative to correct as when the neighbor is a pyrimidine. It appears that the stronger purine stacking interaction helps to stabilize the incorrect nucleotide in an orientation favorable for phosphodiester bond formation. As a result, the nearest neighbor influence on V,,, discrimination tends to cancel the influence on K, discrimination, so that overall fidelity (Table 111) shows a lower and less consistent dependence on nearest neighbors.