the fidelity of dna synthesis catalyzed by derivatives of ... · the fidelity of dna synthesis...

Vol. 265, No. 23, Issue of August 15, pp. 13878-13887, 1990 Printed in U.S. A.

The Fidelity of DNA Synthesis Catalyzed by Derivatives of Escherichia coli DNA Polymerase I*

(Received for publication, February 26,199O)

Katarzyna BebenekS, Catherine M. Joycei, Mary P. Fitzgerald+, and Thomas A. KunkelSn From the SLaboratory of Molecular Genetics, National Institute of Environmental Health Sciences, Research Triangle Park, North Carolina 27709 and the SDepartment of Molecular Biophysics and Biochemistry, Yale University Medical School, New Haven, Connecticut 06510

The fidelity of DNA synthesis by an exonuclease- proficient DNA polymerase results from the selectivity of the polymerization reaction and from exonucleolytic proofreading. We have examined the contribution of these two steps to the fidelity of DNA synthesis catalyzed by the large Klenow fragment of Escherichia coli DNA polymerase I, using enzymes engineered by site- directed mutagenesis to inactivate the proofreading exonuclease. Measurements with two mutant Klenow polymerases lacking exonuclease activity but retaining normal polymerase activity and protein structure demonstrate that the base substitution fidelity of polymerization averages one error for each 10,000 to 40,000 bases polymerized, and can vary more than 30-fold depending on the mispair and its position. Steady-state enzyme kinetic measurements of selectivity at the initial insertion step by the exonuclease-deficient polymerase demonstrate differences in both the K, and the V ,,,a= for incorrect versus correct nucleotides. Exonu- cleolytic proofreading by the wild-type enzyme improves the average base substitution fidelity by 4- to 7-fold, reflecting efficient proofreading of some mispairs and less efficient proofreading of others.

The wild-type polymerase is highly accurate for -1 base frameshift errors, with an error rate of ~10-‘. The exonuclease-deficient polymerase is less accurate, suggesting that proofreading also enhances frameshift fidelity. Even without a proofreading exonuclease, Klenow polymerase has high frameshift fidelity relative to several other DNA polymerases, including eucaryotic DNA polymerase-cr, an exonuclease-deficient, 4-subunit complex whose catalytic subunit is almost three times larger.

The Klenow polymerase has a large (46 kDa) domain containing the polymerase active site and a smaller (22 kDa) domain containing the active site for the 3’ + 5’ exonuclease. Upon removal of the small domain, the large polymerase domain has altered base substitution error specificity when compared to the two-domain but exonuclease-deficient enzyme. It is also less accurate for -1 base errors at reiterated template nucleotides and for a 276-nucleotide deletion error. Thus, removal of a protein domain of a DNA polymerase can affect its fidelity.

Our understanding of the mechanisms that insure the faith-

* This work was supported by National Institutes of Health Grant GM28550 (to N. D. F. Grindley). 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 accord- ance with 18 U.S.C. Section 1734 solely to indicate this fact.

1 To whom correspondence should be addressed.

ful copying of genetic information stems from a wealth of studies, including those involving DNA polymerases such as DNA polymerase I of Escherichia coli. The overproduction (1) of the enzymatically active large (Klenow) fragment of this polymerase and the determination of its structure by x-ray crystallography (2) provides an opportunity to dissect fidelity mechanisms using protein engineering techniques. The Kle- now polymerase has a large (46 kDa) domain containing the polymerase active site and a smaller (22 kDa) domain containing the active site for the 3’ -+ 5’ exonuclease (3). Using site-directed mutagenesis to change amino acids within the exonuclease active site, two mutant polymerases were con- structed (4). The first (D424A) contains a single amino acid alteration, Asp --$ Ala, at codon 424. The second (D355A,E357A) contains two changes, Asp + Ala at codon 355 and Glu -+ Ala at codon 357. Both enzymes retained normal polymerization activity and were devoid of exonuclease activity. When examined by x-ray crystallography, neither mutant protein exhibited significant conformational altera- tions (4).

A third exonuclease-deficient form of E. coli DNA polymerase I was also overproduced and purified from E. coli cells containing a vector encoding the sequence for the large polymerase domain alone (5). This enzyme is the second smallest DNA polymerase described to date, after the 39-kDa eucaryotic repair enzyme, DNA polymerase-/3 (6). The purified large domain retains polymerization activity but has a lower specific activity and a lower apparent affinity for DNA and dNTP’ substrates than its parent.

In this study we use these structurally well-defined derivatives of E. coli DNA polymerase I to address several issues for understanding how genetic information is faithfully copied. First is a determination of the contribution of exonucleolytic proofreading to base substitution fidelity. Previous estimates with DNA polymerase I have employed reactions containing high concentrations of dNTPs or nucleoside monophosphate (7-9), or modified dNTPs (10) to reduce the activity of the exonuclease. These conditions have the disadvantage of pre- senting the polymerase with abnormal reaction conditions that in some cases only partly inhibit exonuclease activity, and/or possibly affect the base selectivity of the polymerase itself. The present studies compare the wild-type polymerase to its structurally normal but exonuclease-deficient derivatives using identical reaction conditions and normal dNTP substrates, permitting a direct determination of the contribution of proofreading to fidelity.

Second, DNA sequence analysis of M13mp2 ~CZCU complementation mutants generated in fidelity assays (11, 12) per- mits a description of proofreading specificity for both base

’ The abbreviations used are: dNTPs, deoxynucleoside triphos- phates; Pu, purine: Pyr, pyrimidine.

13878

Fidelity of Derivatives of E. coli DNA Polymerase 13879

TABLE II Base substitution fidelity of wild-type and mutant Klenow

polymerases in the opal codon reversion assay Reactions (100 ~1) were performed as described under “Experimen-

tal Procedures,” using 130 fmol of gapped DNA and 2.2 pmol of DNA polymerase. The background mutant frequency of uncopied DNA was 2 x 10-e.

DNA polymer- me

Wild-type D424A

D355A,E357A

dNTP

PM 1 1

1000 1

1000

Reversion frequency

X10-=

18 120 200 100 180

substitution and frameshift errors. This approach also de- scribes the error specificity of the exonuclease-deficient polymerase alone. Finally, we also examined the fidelity of the large polymerase domain after removal of the small domain. We did so because controlled removal of a protein domain is one approach for dissecting the structural determinants of fidelity and because proteolysis may be a physiologically rel- evant mechanism by which the properties of polymerase, including its fidelity, are altered in vivo.

EXPERIMENTAL PROCEDURES AND RESULTS’

Absence of 3’ + 5’ Exonuclease Activity-Measured by release of acid-soluble radioactivity from radiolabeled DNA, both the D424A and D355A,E357A mutant Klenow polymerases were previously found to be devoid of detectable exonuclease activity (4). Since the 46-kDa polymerase domain of E. coli DNA polymerase I does not contain the active site for the exonuclease, it too lacks exonucleolytic activity (5). Using two assays for excision of a mismatched base from a primer terminus, we confirmed that the mutant enzyme preparations used in this study contained no detectable exonuclease activity, but were able to add dNTPs onto several mispaired template-primer termini and onto a misaligned terminus as well (for details, see Fig. 1 and Table I and text in Miniprint Supplement).

Fidelity Measurements for Base Substitution Errors-The error rates for single-base substitutions by the wild-type and the mutant Klenow polymerases were measured using an Ml3mp2-based opal codon reversion assay (12). This assay detects eight base substitution errors at a TGA codon in the LucZa complementation coding sequence, as blue revertants of a colorless plaque phenotype. When polymerase reactions were performed using 1 pM dNTPs, the single-base substitution error rates of the exonuclease-deficient polymerases are 6- to 7-fold higher than that of the wild-type enzyme (Table II). For reactions performed using a IOOO-fold higher concentration of dNTPs, the reversion frequencies increased 1.7- and 1.8-fold, respectively, for the D424A and D355A,E357A mutant polymerases.

DNA sequence analysis of opal codon revertants (Table III) demonstrated that all revertants contained single-base substitutions at the opal codon and defined the base substitution specificity for the eight detectable errors.

Steady-state Kinetic Studies-Since the results in Table III demonstrate that, for exonuclease-deficient Klenow polym-

’ Portions of this paper (including “Experimental Procedures,” part of “Results,” Figs. 1-3, 5, and 6, and, Tables I, IV, and VII) are presented in miniprint at the end of this paper. Miniprint is easily read with the aid of a standard magnifying glass. Full size photocopies are included in the microfilm edition of the Journal that is available from Waverly Press.

TABLE III Sequence analysis of opal codon revertants produced by wild-type,

D424A, and D355A,E357A mutant polymerases The number of mutants having a particular sequence are given in

parentheses. The error rates were obtained by multiplying the per- centage of mutants having a particular sequence by the mutant frequency (Table II), then dividing by 0.6 to correct for expression of errors recovered by transfection of this (methylated) DNA (for the construction and transfection of these substrates, see Refs. 11, 12, and 14). When no mutants of a particular type were recoverd, “5” values were calculated based on one occurrence.

Error rate (X lo-‘)

Revertant Mispair codon 1 pi dNTP 1 rn~ dNTP

Wild type D424A D424A D355A,E357A CGA T.dGTP 1.9 (4) 120 (33) 180 (32) 120 (17) GGA T.dCTP 10.5 (0) 18 (6) 55.7 (0) 6.8 (1) AGA T.dTTP ~0.5 (0) ~3.6 (0) 12 (2) 20 (3)

TTA G.dATP 5.6 (12) 7.0 (2) 69 (12) 68 (10) TCA G.dGTP 8.0 (17) 3.6 (1) 12 (2) 34 (5)

TGG A.dCTP 3.3 (7) ~3.6 (0) 40 (7) 34 (5) TGT A.dATP ~0.5 (0) ~3.6 (0) 17 (3) 20 (3) TGC A.dGTP 11 (24) 47 (13) 55.7 (0) ~6.8 (0)

erase, the T. dGTP mispair was both made and extended at 1 pM dNTPs, while the G. d.ATP mispair was recovered at high frequency only at 1 mM dNTPs, we focused on these two errors. We examined the degree of discrimination by the polymerase at the insertion step using a quantitative gel electrophoresis assay (18). Using the exonuclease-deficient D355A,E357A polymerase, the kinetic constants (K,,, and V,,,,,) for insertion of (correct) dATP vers’sus (incorrect) dGTP opposite a template T and the insertion of (correct) dCTP versus (incorrect) dATP opposite a template G were determined as a function of substrate concentration (see Fig. 2 and Table IV in the Miniprint Supplement). Incorporation of dGTP opposite T is characterized by a K,,, that is l70-fold higher and a V,., that is g-fold lower than for the correspond- ing constants for the correct nucleotide, dATP. The K,,, for incorporation of dATP opposite template G is 260-fold higher than for incorporation of correct dCTP, while the V,., value is lower by 50-fold for dATP compared to dCTP. Discrimi- nation thus reflects substantial differences in both K,,, and V,,, values. The K,,, and V msr values can be used to calculate the misinsertion frequencies at the two sites (18). The enzyme misinserts dGTP opposite T at position 87 with a frequency of 6.5 X 10e4 relative to insertion of correct dATP. Misinser- tion of dATP opposite G at position 88 occurs with a frequency of 7.5 X 10e5 relative to correct insertion of dCTP.

We also examined the ability of the exonuclease-deficient polymerase to add the next correct nucleotide onto the two resulting mispairs, using a radiolabeled oligonucleotide substrate (see Fig. 3A in the Miniprint Supplement) and an excess of polymerase over DNA (an enzyme to DNA ratio similar to the Ml3 fidelity assay). The next correct nucleotide is incorporated much more effectively onto the T. G mispair than onto the G. A mispair (Fig. 3B).

Fidelity Measurements for Frameshift Errors-We examined frameshift fidelity using a highly sensitive reversion assay (Fig. 4) that scores -1 base errors within a template TTTTT run as well as at 36 other positions. The reversion frequency and error rates per detectable nucleotide polymer- izqd (calculated so as to correct for differences in target size) for DNA synthesis catalyzed by the wild-type and mutant Klenow polymerases are given in Table V. When reactions were performed at a dNTP concentration of 1 FM, all three

13880 Fidelity of Derivatives of E. coli DNA Polymeruse

Dark Blu.

TG ATT ACG AAT TCA CTG GCC GTC G-A C?+A CGT CGT GAC

Colarless

TG ATT ACG PAT TCA CTG GCC GTC GD m ACG ,?ZG TCiA

TG ATT ACG AAT TCA CTG GCC GTC Gm &CA ACG CGT GAC

FIG. 4. Reversion assay for frameshift errors. The assay uses an M13mp2 mutant DNA substrate containing a 361-base gap and scores blue revertants of a colorless plaque phenotype. The starting mutant (the middle line of DNA sequence) is colorless due to the addition of one extra T into the run of four consecutive T nucleotides at uositions 70 to 73 (underlined). Blue nlaaues result from frameshift errors that restore the reading frame. Minus one base errors within the TTTTT run restore the wild-type coding sequence and yield dark blue plaques (upper sequence). Minus one base errors at any of 36 other nucleotide positions also yield blue plaques. These are lighter blue because, although the reading frame is restored, 1 or more amino acids are no longer wild-type. For example, if the underlined 2’ in the middle line of sequence is lost, the revertant (lower line of sequence) differs from wild-type by three codons (bold face letters in the lower sequence). An unequivocal distinction between wild-type dark blue and pseudo-wild-type lighter blue revertants, and hence between frameshifts at the TTTTT run versus other sites, can be made bv comparing, on the same plate, candidate revertants generated in a polymerase reaction with authentic wild-type dark blue M13mp2 plaques. Both light and dark blue revertants can be quantitatively scored on plates containing as many as 10,000 colorless plaques (12), a density that was not exceeded in any of the experiments reported here.

TABLE V Minus one base frameshift fidelity of Klenow polymerases in the

frameshift reversion assay Reactions (100 ~1) contained 130 fmol of gapped DNA and 2.2

pmol of DNA polymerase. Reversion frequencies are the average of two determinations, after subtracting the background reversion frequency values of 2.1 X 10m5 and 3.2 X 1Om6 for dark blue (at the TTTTT run) and light blue (at the 36 other sites) mutants, respectively. Error rates are expressed “per detectable nucleotide polymerized,” and are calculated bv dividing the reversion frequencv by 0.6 (the’ frequency of error expression upon transfection, kef. -12i and then dividing by the number of target nucleotides for the type of error (5 nucleotides for the TTTTT run and 36 for the nonreiterated nucleotides). When the reversion frequency was less than 2-fold above the background, the error rate was calculated using the background reversion frequency.

Klenow polymerase

used

-1 TTTTT run -1 Nonreiterated

Reversion Error rate Reversion freauencv freouencv Error rate

1 _ 1 1

x10-5 x10-6 1 fiM dNTPs

Wild-type 52.1 ~1/140,000 5.3 1/4,100,000 D355A,E357A 52.1 ~1/140,000 ~3.2 ~1/6,800,000 D424A 4.2 l/71,000 8.8 l/2,500,000

1 mM dNTPs Wild-type 4.4 l/68,000 8.8 l/2,500,000 D355A,E357A 13.0 l/23,000 54.0 1/400,000 D424 12.0 l/25,000 130.0 1/170,000

forms of the polymerase were highly accurate. When the substrate concentration was increased lOOO-fold, all three polymerases generated errors at frequencies above the background frequency of uncopied DNA. For all three enzymes, the error rate was higher at the TTTTT run than the average

error rate at the 36 other sites. The exonuclease-deficient polymerases were 3-fold less accurate than the wild-type enzyme for errors at the TTTTT run. They were 6- to 15- fold less accurate than the wild-type enzyme at the other sites.

Fidelity Measurements with the Forward Mutation Assay- To obtain a more comprehensive estimation of accuracy than is possible when measuring errors at a limited number of template positions, we examined fidelity in the forward mutation assay (11). Here, base substitution, frameshifts, and more complex errors are scored within a 258-nucleotide target sequence, as light blue or colorless plaque phenotype mutants.

For reactions performed using 1 pM dNTPs, the wild-type polymerase is about 4-fold more accurate than the exonuclease-deficient polymerases (Table VI). The large polymerase domain required a dNTP concentration of 10 pM dNTPs to achieve complete gap-filling synthesis. At this concentration, its fidelity was similar to that of the two-domain, exonuclease- deficient polymerases.

At a dNTP concentration of 1 mM, the accuracy of the wild-type enzyme was reduced 4-fold, yielding fidelity similar to that of the exonuclease-deficient polymerases (Table VI). The error rate of the exonuclease-deficient Klenow polymerases was slightly increased relative to their accuracy at 1 pM, while the large polymerase domain had approximately the same fidelity at both substrate concentrations.

Next, we determined the sequence of independent mutants generated during DNA synthesis by each of the four enzymes (see Table VII, in the Miniprint Supplement). This was done for three reasons: (i) to describe the contribution of proofreading to fidelity for individual base substitution and frameshift errors, (ii) to examine discrimination at the polymerization step as a function of position and type of error but without interference from the proofreading activity and, (iii) to describe the fidelity of the large polymerase domain in the absence of the small domain.

The collection of mutants analyzed from the wild-type enzyme reaction consisted of 10 mutants containing single- base substitutions, one with a single-base frameshift and five others (Table VIII). The mutants generated by the D424A polymerase were similar to those produced by the D355A,E357A polymerase. This observation, plus the data in Tables II, III, and V demonstrating that the two polymerases gave similar results in the reversion assays, led us to combine their forward mutation data (Fig. 5, in the Miniprint Supple- ment). Most of these mutants resulted from single-base substitution and frameshift errors. Notably, six 2-base deletion were also recovered (Table X). Mutants generated by the

TABLE VI Fidelity of wild-type and mutant Klerww polymerases in the forward

mutation assay Reactions (50 ~1) contained 67 fmol of gapped DNA and 1.1 pmol

of DNA nolvmerase.

DNA polymerase Number of Mutant determinations frequency”

x1o-4 1 /.tM dNTPs

Wild-type Klenow 7 11 & 8 D424A 2 43 D355A,E357A 3 39k5 Large domain* 1 37

1 mM dNTPs Wild-type Klenow 3 40 f 4 D424A 2 69 D355A,E357A 1 63 Large domain 2 43

’ A background mutant frequency of 7 x lo-” has been subtracted. * This reaction was performed using 10 pM dNTPs.

13881 Fidelity of Derivatives of E. coli DNA Polymerase

TABLE VIII Classes of errors by wild-type and mutant Klenow polymerases and by the large polymerase domain in the forward

mutation assay

Wild-type Klenow Exo-deficient Klenow Large domain Type of mutation

Mutants M.F.” Error rate* Mutants M.F.” Error rate * Mutants M.F. Error rate* Base substitution 10 4.1 1/170,000 64 17 1/40,000 89 22 l/31,000 Frameshifts’ 1 0.2 s1/1,500,000~ 16 4.5 1/200,000 44 12 l/75,000 Other’ 5 1.3 14 2.0 33 6.8 Total 3 A 166

’ Mutant frequency (M.F.) x 10m4. The data in Table VII and the sequencing information were used to determine the mutant frequency for each class of errors, after which the following background frequencies for uncopied DNA (11) were subtracted: base substitutions, 3.4 X lo-‘; frameshifts, 0.6 X lo-$ other, 2.5 X lo-“. The overall mutant frequency was 12 X lo-’ for the wild-type Klenow polymerase, 30 X 10m4 for the mutant Klenow polymerases, and 47 x lo-’ for the large domain. Several light blue mutants were sequenced but had no change within the LucZcv sequence present within the gapped substrate; these were removed from the analysis.

‘Calculated using a target size of 114 for base substitutions and 150 for frameshifts, and a minus-strand expression value of 60% (12).

’ Includes +- and -1 base frameshifts only. dBecause the mutant frequency was less than 2-fold above the background frequency of uncopied DNA, this

error rate was calculated using the background mutant frequency. ’ This class includes mutants that have lost more than a single base and mutants containing two changes.

TABLE IX Base substitution error rates from the forward mutation assay

Error rates were calculated using the number of detectable sites listed for each individual mispair as a target size.

Change Known Wild-type Exe-deficient Large domain from + to Mispair sites Mutants Error rate Mutants Error rate Mutants Error rate T-C T.dGTP 22 0 11/180,000 19 l/22,000 35 l/13,000 C-+T C.dATP 23 1 l/180,000 13 l/33,000 11 l/44,000 G-+A G. dTTP 20 0 ~1/160,000 2 1/190,000 17 l/25,000 A-G A. dCTP 14 0 ~1/110,000 1 l/260,000 5 l/59,000 T-G T.dCTP 23 0 ~1/180,000 6 l/72,000 3 l/160,000 T-A T.dTTP 15 0 ~1/120,000 1 l/280,000 3 1/110,000

E? C.dCTP 8 0 ~1/64,000 0 s1/150,000 0 ~1/170,000 C . dTTP 14 0 ~1/110,000 0 s1/260,000 0 ~1/300,000

z: G.clATP 23 2 l/92,000 10 l/43,000 4 1/120,000 G.dGTP 17 1 1/140,000 10 l/32,000 4 1/90,000

A+T A.dATP 20 0 ~1/160,000 2 1/190,000 3 1/140,000 A-C A. dGTP 17 6 l/23,000 0 ~1/320,000 4 1/90,000

large domain (Table VIII) also comprised mostly single-base changes (Fig. 6, in the Miniprint Supplement). However, seven 2-base deletions and 14 identical (but independent) 276-base deletion errors were also recovered (Table X). The nucleotides lost included one of two g-base direct repeat sequences and all intervening nucleotides.

The sequencing information was used to calculate single- base error rates for both base substitution and frameshift errors (Tables VIII, IX, and X).

DISCUSSION

The fidelity of DNA synthesis depends on discrimination against errors during the initial nucleotide insertion step, discrimination against extension from mispaired or misaligned primer termini, and exonucleolytic removal of errors. The availability of well-defined derivatives of the Klenow fragment of E. coli DNA polymerase I has permitted the selectivity of these steps to be examined. The results presented in this study demonstrate that by far the greatest contribution to the base substitution and frameshift fidelity of the Klenow fragment of E. coli DNA polymerase I comes from discrimination during polymerization. Thus, the exonuclease-deficient polymerase discriminates against errors by a factor of 104- to 106-fold depending on the error considered (Tables III to V, VIII, IX, and X). This reflects the contri- butions of both the initial insertion and subsequent extension

steps. In contrast, using certain assumptions discussed below, we estimate from the data in Tables II, V, and VI that exonucleolytic proofreading improves the overall average fidelity of DNA synthesis by less than lo-fold.

Polymerase Selectivity in the Absence of Exonuckolytic Proofreading-A comparison of the two exonuclease-deficient Klenow polymerases shows that they have very similar error rates (Tables II, V, and VI) and error specificity (Table III). This is consistent with the observation that the amino acid substitutions introduced into the small domain to inactivate the exonuclease activity do not affect polymerase structure (4), and further suggests that these changes do not affect polymerase base selectivity. This conclusion is reinforced by the observation that the base substitution error rates (Table IX), ranging from a high of 4.5 X 10m5 (for T.dGTP errors) to a low of ~3.1 x 10m6 (for A.dGTP errors), are in the same range as previous base substitution fidelity estimates with intact Poll (8, 10) and the wild-type Klenow polymerase (9) when the exonuclease activity was reduced using altered reaction conditions. The base substitution values shown in Tables III and IX are also similar to estimates of wild-type Klenow polymerase selectivity at the polymerization step, from kinetic measurements with oligonucleotide substrates (19).

Specificity Rules for Base Substitution Errors-One goal of error specificity studies is to search for fidelity rules that could be useful in understanding the molecular basis for the

13882 Fidelity of Derivatives of E. coli DNA Polymeruse

TABLE X

Frequency of 1 and 2 base frameshift and 276 base deletion errors determined from the forward mutation assay Error rates were calculated using as a target size the number of detectable sites for each type of error.

Mutation

-1 in run -1 in non-run -2 bases 276-base deletion

Known sites

80 70

150

Wild-type Exo-deficient Large domain

Mutants Error rate Mutants Error rate Mutants Error rate

0 ~1/640,000 8 1/190,000 37 l/46,000 1 l/560,000 7 1/190,000 5 1/300,000 0 s1/1,200,000 6 1/470,000 7 1/450,000 0 17.5 x lo+ 0 53.2 x 1O-5 14 40 x 1o-5

fidelity of DNA synthesis. The average base substitution error rate of the exonuclease-deficient Klenow polymerase, l/ 40,000 (Table VIII), is 3- to &fold and 30-fold more accurate, respectively, than exonuclease-deficient eucaryotic DNA polymerases-a and -fi (14, 17). While the structural and/or kinetic basis of these polymerase-mediated differences has not been determined, they establish the obvious but important rule that the degree of selectivity of the polymerization step is strongly influenced by the polymerase itself.

The base substitution error specificities of the exonuclease- deficient Klenow polymerases (Fig. 5 and Tables IX and X) can be summarized as follows: (i) errors are found throughout the 258-base target sequence, (ii) these are not random with respect to the template position, (iii) even in this relatively small mutant collection, base substitution errors representing 9 of the 12 possible base mispairs are observed, with transversions (45%) occurring almost as frequently as transitions (55%), (iv) errors resulting from C. dCTP and C. dTTP mispairs are rare, (v) mispair symmetry differences are observed (e.g. compare C. dATP to A.dCTP mispairs or G .dATP to A .dGTP mispairs), and (vi) base substitution errors result primarily from misinsertion of purines (e.g. the order of dNMP substrate misinsertion is G, 45%; A, 39%, C, 11% and T, 5%).

Certain features of the exonuclease-deficient Klenow polymerase base substitution error specificity are remarkably similar to those of other exonuclease-deficient polymerases. Since the preference for misincorporation of purine rather than pyrimidine nucleotides (Table IX) has been seen with every exonuclease-deficient DNA polymerase examined in the M13mp2 system (13, 14, 17, 20), this seems to be a valid selectivity rule for DNA synthesis in vitro. This purine misinsertion preference results in lower fidelity for transversions via Pu.Pu mispairs rather than via Pyr. Pyr mispairs. With the Klenow polymerase, there is an asymmetry in error rates for the transition mispairs (Table IX), where misincorporation of dGTP opposite template T is more common than of dTTP opposite template G, and misincorporation of dATP opposite template C is more common than of dCTP opposite template A.

Since the base substitution error rates (Tables III and IX) reflect discrimination in the initial misinsertion reaction and during subsequent correct insertion from the mispaired terminus (because a misinsertion must be fixed by extension in order to be expressed upon transfection), the preference for purine misincorporation could reflect lesser discrimination at either step. When these two steps have been described for DNA polymerase-cy using gel electrophoresis assays, mispairs containing primer-terminal purines are not preferentially extended (21,22), but a distinct preference is seen for misinsertion of purines (23). Preferential purine misincorporation may reflect stronger base stacking interactions for purines than for pyrimidines. It has also been suggested (24) that purine nucleotides are more likely to occupy the proposed hydropho-

bit active site of a polymerase because they are more hydro- phobic than are pyrimidines.

Misincorporation of pyrimidines opposite template C residues is rare (Table IX). Since this observation has been made in a number of studies with other DNA polymerases (13, 14, 17, 20, 22, 23), it seems to be a reasonably firm polymerase fidelity rule, particularly for C . dCTP mispairs. However, the rule cannot be generalized to all four Pyr.Pyr mispairs, because T. dCTP errors are generated relatively frequently by the exonuclease-deficient Klenow polymerase (Tables III and IX) and also by DNA polymerases-a, $3 (14), and yeast DNA polymerase I (13).

The high error rate for purine misincorporation opposite template G residues (Tables III and IX) has also been seen with DNA polymerases-a and -/3 (13, 14). One possible expla- nation for transversions targeted to template purines is preferential misincorporation of dAMP or dGMP at sites where cryptic damage has eliminated template coding potential, one possible source being abasic sites. Although we cannot une- quivocally eliminate this possibility, we consider it unlikely for several reasons. The improved method of DNA substrate construction (12) does not require mistreatment of the DNA. Transversions at template A sites occur infrequently compared to transversions at template G sites. However, the rate of depurination of A, as extrapolated from high temperature incubations of DNA, is only P-fold lower than G (25). Most importantly, kinetic analyses of fidelity using oligonucleotide substrates depend on levels of misincorporation far in excess over the amount of damage that should be present in these substrates, yet the results of such studies (for the Klenow polymerase, Fig. 2 and Table IV in the Miniprint Supplement; for polymerase-a, Refs. 21 and 23) demonstrate that purine misinsertions at template G sites are quantitatively similar to the error rate estimates from the Ml3mp2 fidelity assays. Thus we favor the possibility that the high error rates for G. dAMP and G . dGMP mispairs reflect undamaged substrate interactions within the polymerase active site, although the structural and/or kinetic basis is not known.

As observed with all other DNA polymerases, site to site variations in base substitution fidelity are observed with the exonuclease-deficient Klenow polymerase (Fig. 5) and the large domain (Fig. 6). The largest difference is alO-fold (Fig. 6,10 T --, C errors at position -36, none at several other sites where this error can be scored). In the future, it may be possible to understand the interactions between the polymerase and the template-primer that are responsible for such neighboring nucleotide effects.

Misinsertion, Extension from Mispairs, and Steady-state Kinetic Analysis-Comparison of base substitution error rates for the D424A polymerase at 1 uersus 1000 PM dNTP concentrations reveals an increase in the error rate for several misincorporation events at the opal codon (Table III). The increase is rlO-fold for A. dCTP mispairs, g-fold for G .dATP errors, and r4.5-fold for A. dATP errors. The smaller effects observed for T.dTTP errors (23.1-fold), G.dGTP errors (3.1-

Fidelity of Derivatives of E. coli DNA Polymer-use

fold), and T. dGTP errors (I.&fold) may also be significant but could simply represent fluctuations due to sparse data. These differences are consistent with kinetic studies with several DNA polymerases which show that, following misinsertion, the rate of the subsequent extension step (21, 22, 26) depends on dNTP concentration.

Kinetic analysis of discrimination at the insertion step by the exonuclease-deficient Klenow polymerase shows differences in both the apparent K, (170- and 260-fold) and the V,,, (9- and 50-fold) for incorrect uersus correct insertions (Table IV). A similar conclusion was reached in a previous study with the wild-type Klenow polymerase (19), although in that study discrimination reflected larger differences in the relative V,., values (500-fold or greater) and small differences in the relative K,,, values (2- to 4-fold in Table II of Ref. 19, or 0- to IOO-fold, as described on page 6722 of that reference). These quantitative variations in relative kinetic constants may simply reflect site to site or mispair-dependent variation, as has been described with DNA polymerases-cy and avian myeloblastosis virus reverse transcriptase (23). Alternatively, differences in reaction conditions or the use of wild-type (19) uersw exonuclease-deficient Klenow polymerase (this study) may explain the differences. The steady-state kinetic analyses in Fig. 2 and Table IV are not sufficient to distinguish at which step in the complex polymerase reaction mechanism (the dNTP binding step, the chemical step, conformational changes) discrimination occurs for the base substitutions reported in Tables III and IX. However, the data with the exonuclease-deficient Klenow polymerase extend the general rule that, for every polymerase that has been examined, both K,,, and V,.. differences for incorrect versus correct nucleotide insertions are readily apparent.

For the same two mispairs at the same position, there is reasonable quantitative agreement between two very different methodologies for determining fidelity. The error rate calculated for the exonuclease-deficient Klenow polymerase from the steady-state kinetic analysis of the insertion step for the T-dGTP mispair at position 87 (Table IV) differed less than 4-fold from the value obtained using the Ml3 reversion assay and reactions performed with 1 mM dNTPs (Table III). For the G.dATP mispair at position 88, the two methods gave the same quantitative answer. The observation that the error rates obtained from the reversion assay for reactions performed with 1 pM dNTPs are somewhat lower than the misinsertion frequencies determined from the kinetic assay presumably partly reflects discrimination against incorporat- ing the next correct nucleotide onto a template-primer containing a terminal mispair (Fig. 3B). Nonetheless, in the Ml3 assays there is less than a 2-fold difference in total mutant frequency with a IOOO-fold change in dNTP concentration (Tables II and VII), clearly demonstrating that lack of extension of mispairs does not compromise reasonable estimates of average error rates with M13mp2-based fidelity assays.

Polymerase Selectivity for Frameshifts-At l/200,000, the average single-base frameshift error rate per nucleotide polymerized for the exonuclease-deficient Klenow polymerase is 5-fold lower than for single-base substitutions (Table VIII). As for other DNA polymerases, -1 errors are >lO-fold more frequent than +l errors. This may in part reflect a difference in the number of hydrogen bonds that must be disrupted to form a frameshift intermediate (l&27), and/or more effective protein-mediated constraints against +l than -1 frameshift intermediates.

Minus one base frameshift fidelity for the 6%kDa exonuclease-deficient Klenow polymerase is lo-fold greater than for exonuclease-deficient DNA polymerases (Y-DNA primase (13,

15), a 4-subunit enzyme complex with a 180-kDa polymerase catalytic subunit, and is IOO-fold greater than for the 40-kDa single-subunit DNA polymerase+ (15). Such large differences in discrimination against frameshift errors may result from substantial differences in polymerase contacts with the template-primer. Available structural and footprinting data suggest that the primer-binding cleft of the Klenow polymerase contacts 6 to 9 base pairs (3). The distance between the polymerase active site and the exonuclease active site on the small domain can accommodate 8 additional nucleotides, 4 of which probably become single-stranded to enter and bind to the exonuclease active site (28,29). It is feasible that the high frameshift fidelity of the exonuclease-deficient two-domain polymerase could result from the previously proposed “slide and melt” movement of the template-primer between the two domains (3), which could disrupt misaligned intermediates and reduce frameshift error rates. This concept is consistent with the observation (Table X) that the fidelity of the large polymerase domain for -1 errors in runs of a common nucleotide is lower than for the two-domain polymerase.

Similar to previous observations with other DNA polymerases, frameshift errors are not distributed randomly within the target sequence. This non random distribution is useful for considering possible mechanisms for frameshift errors. For example, in the frameshift reversion assay, the observation that the error rate is greater at iterated nucleotides than at other sites (Table V) suggests that a portion of these errors may result from template-primer slippage (30). This possibility is reinforced by the fact that most of the large domain errors in the forward mutational spectrum are also at template runs (Fig. 6 and Table X), as well as by observations with other DNA polymerases (see Ref. 27 for review). Alterna- tively, the higher frequency at reiterated sites could result from neighboring nucleotide effects on other frameshift mechanisms.

Several other mechanisms that could explain the frameshift errors in Table V and Figs. 5 and 6 have been proposed (15, 31-34). Included among these is our previous suggestion that frameshift errors could be initiated by nucleotide misinsertion (31); when the misinserted nucleotide is complementary to the next template nucleotide, realignment may occur to form a template-primer with a correct terminal base pair and an extra nucleotide. The results in Table I are consistent with this model, and demonstrate that a base substitution intermediate can be processed by the exonuclease-deficient Klenow polymerase into a -1 base error. The observation that 11 of the 12 non-run l-base frameshifts (Table X) were the loss of a purine nucleotide that has a template pyrimidine as a 5’ nearest neighbor suggests the possibility that these mutants result from Pu.Pu mispairs that realign. Six of seven non- run frameshift errors by the exonuclease-deficient Klenow polymerase were the loss of template A (template). Since A. dPuMP mispairs are extended with difficulty by the Klenow polymerase (35), it may be that those mispairs that are most difficult to extend are more likely to lead to formation of a misaligned substrate that can be extended more effectively by the polymerase. Experiments to determine if misinsertions during an ongoing polymerization reaction can produce frameshifts have provided support for this model (36).

Proofreading of Base Substitution Errors-A major objec- tive of this study was to describe the contribution of proofreading to fidelity during an ongoing polymerization reaction by comparing the wild-type polymerase to its exonuclease- deficient derivatives. The assumption is that misinsertions are proofread by the exonuclease activity of the wild-type enzyme in vitro rather than remaining unextended and then

13884 Fidelity of Derivatives of E. coli DNA Polymerase

being removed in uivo after transfection. Because some misinsertions by the exonuclease-deficient polymerase may not be extended and thus not recovered upon transfection, the comparison provides a minimum estimate. For errors at the opal codon in reactions performed with 1 pM dNTPs (Table III), the contribution of proofreading to fidelity was 63-fold for the T. dGTP mispair, 36-fold for the T . dCTP mispair, 4- fold for the A. dGTP mispair, and no effect for the G . dATP and G. dGTP mispairs.

A similar comparison of the error rate of the wild-type polymerase to the exonuclease-deficient Klenow polymerases from the forward mutation assay gives a second estimate of the contribution of proofreading to base substitution fidelity, a global view for all 12 mispairs at 114 template sites. Here the comparison is for reactions performed with 1 PM dNTPs with the wild-type enzyme and 1 mM dNTPs with the exonuclease-deficient polymerases. Considering all possible base substitution errors, the wild-type polymerase is 4-fold more accurate than its exonuclease-deficient derivatives (Table VIII). For individual mispairs, the difference in error rate between the wild-type and exonuclease-deficient enzymes var- ies (Table IX); the greatest effect is r8-fold for T.dGTP errors. Results from both the reversion and forward assays are thus in reasonable agreement. Similar estimates of the contribution of proofreading to fidelity have been made previously for a T. G mispair with the Klenow polymerase (9) and for A. G and A.C mispairs with intact PolI (8, 10). In those studies, it was estimated that proofreading enhanced fidelity by <7-fold and 5 and 30-fold, respectively.

The Klenow polymerase has been shown to proofread errors by two pathways, one an intramolecular process involving misinsertion followed by exonucleolytic removal without enzyme-DNA substrate dissociation (35, 37), and the other an intermolecular process that involves misinsertion, dissociation of the enzyme-DNA complex, reassociation with the exonuclease active site, and then excision (35). As discussed previously (35), which pathway is used depends on the relative rates of the polymerization and exonuclease reactions and the rate of dissociation of the enzyme-DNA complex. A kinetic analysis (19) with the wild-type Klenow enzyme has suggested that incorrect product dissociates from the enzyme 4- to 61- fold more slowly than does correct product, while incorporation of the next correct nucleotide onto the mismatch occurs 6- to 340-fold more slowly. Since rate reductions at either step may allow the 3’ + 5’ exonuclease to excise errors, determination of these rates may eventually explain the observations that some mispairs are more effectively proofread than others (Tables III and IX). It is interesting that the T. dGTP mispair at position 87 is more effectively proofread than is the G. dATP mispair at position 88 (Table III). Since the T.dGTP mispair is also more easily extended (Fig. 3), proofreading is not necessarily greatest for those mispairs that are difficult to extend. Discrimination may thus reflect the relative rates of incorrect versus correct product release. Slower release for the T. dGMP mispair-containing substrate may lead to more effective proofreading because the enzyme does not dissociate after the misinsertion event.

Proofreading of Frameshift Errors-The results in Table V suggest that proofreading contributes to the frameshift fidelity of the wild-type Klenow polymerase. The wild-type polymerase was 3-fold more accurate than either mutant polymerase for frameshifts at the TTTTT run. It was 6- and 15-fold more accurate, respectively, than the double or single mutant exonuclease-deficient polymerases, for errors at the 36 other positions.

If the -1 base errors at the TTTTT run indeed result from

template-primer slippage, the lesser proofreading effect for these errors compared to non-run frameshifts may result from protection of the misaligned heteroduplex by as many as four correct T. A base pairs. The ability of correctly paired bases to protect mispairs from exonucleolytic digestion is well known (e.g. see Ref. 38). It will be interesting to examine the contribution of proofreading to frameshift fidelity as a function of the length of the run. The greater contribution of proofreading to frameshift fidelity at non-run positions is consistent with lesser protection and the possibility (discussed above) that some frameshifts may result from misinsertions that can be proofread.

Fidelity of the Large Domain-The results in Figs, 5 and 6 and Tables VIII to X also describe the effect of removal of the small domain on the fidelity of the large polymerase domain, without interference from proofreading. The overall average fidelity of polymerization by the large domain is only slightly lower than for the exonuclease-deficient, two-domain polymerase (Table VIII), suggesting that discrimination against errors results primarily from substrate interactions with the amino acid residues important for dNTP binding, template-primer binding, and phosphodiester bond formation. In the case of E. coli DNA polymerase I, these are in the 46- kDa polymerase domain.

While the fidelity of the large domain and the exonuclease- deficient polymerase have similar base substitution specificity (Table IX), there are two interesting exceptions. The error rates for misincorporation of dTTP opposite G and of dCTP opposite A by the large domain were 4.4- and 7.6-fold higher, respectively, than for the two-domain polymerase. Both mispairs result from misincorporation of pyrimidines. This re- duces two error-symmetry biases observed with the two-domain polymerase. Thus, the 19:2 bias for misinsertion of dGTP opposite T compared to misinsertion of dTTP opposite G, and the 13:l bias for misinsertion of dATP opposite C compared to misinsertion of dCTP opposite A by the two- domain enzyme are both reduced to a 2:l biases for the large domain. It is not yet possible to distinguish among the possible explanations for these observations.

The fidelity of the large domain is 4-fold lower than the intact two-domain enzyme for frameshifts within runs of a common base (Table IX). It is also at least 12-fold more likely to generate a specific 276-nucleotide deletion error (Table X), involving loss of one of two nine-base direct repeats and all the intervening nucleotides (for a review of polymerase-mediated deletions by a direct repeat mechanism, see Ref. 27). The large domain also utilizes a misaligned template-primer as effectively as a mispaired template-primer (Table I), in contrast to the two-domain enzyme, which prefers the mispaired template-primer by a factor of 26:l.

Each of these observations demonstrate that the large domain differs substantially from its two-domain parent, and suggest that removal of a protein domain from a DNA polymerase can substantially alter its fidelity. This conclusion is consistent with observations on frameshift error specificity differences noted when intact E. coli DNA polymerase I and the wild-type Klenow polymerase were compared in a reversion assay (33), and with observations using other DNA polymerases (13,39,40), suggesting that proteolysis can affect fidelity.

Acknowledgments-We thank Kristin A. Eckert and Roe1 M. Schaaper for critical evaluation of the manuscript, Victoria Derby- shire for purified D424A protein, and Paul S. Freemont for purified large domain protein.


5.

6.

7.

8.

9.

10.

11. 12.

13.

14.

15. 16.

17.

18.

REFERENCES

Joyce, C. M., and Grindley, N. D. F. (1983) Proc. Nutl. Acud. Sci. U. S. A. SO, 1830-1834

Ollis, D. L., Brick, P., Hamlin, R., Xuong, N. G., and Steitz, T. A. (1985) Nature 313,762-766

Joyce, C. M., and Steitz, T. A. (1987) Trends Biochem. Sci. 12, 288-292

Derbyshire, V., Freemont, P. S., Sanderson, M. R., Beese, L., Friedman, J. M., Joyce, C. M., and Steitz, T. A. (1988) Science 240,199-201

Freemont, P. S., Oilis, D. L., Steitz, T. A., and Joyce, C. M. (1986) Proteins 1, 66-73

Wilson, S., Abbotts, J., and Widen, S. (1988) &o&m. Biophys. Acta 949, 149-157

Que, B. G., Downey, K. M., and So, A. (1978) Biochemistry 17, 1603-1606

Kunkel, T. A., Schaaper, R. M., Beckman, R., and Loeb, L. A. (1981) J. Biol. Chem. 256, 9883-9889

Fersht, A. R., Knill-Jones, J. W., and Tsui, W. -C. (1984) J. Mol. Biol. 156,37-51

Kunkel, T. A., Eckstein, F., Mildvan, A. S., Koplitz, R. M., and Loeb, L. A. (1981) PFOC. Natl. Acad. Sci. U. S. A. 78, 6734- 6738

Kunkel, T. A. (1985) J. Biol. Chem. 260,5787-5796 Kunkel, T. A., and Soni, A. (1988) J. Biol. Chem. 263, 4450-

4459 Kunkel, T. A., Hamatake, R. K., Motto-Fox, J., Fitzgerald, M.

P., and Sugino, A. (1989) Mol. Cell. Biol. 9, 4441-4458 Kunkel, T. A., and Alexander, P. S. (1986) J. Biol. Chem. 261,

160-166 Kunkel, T. A. (1986) J. Biol. Chem. 261, 13581-13587 Kunkel, T. A. (1984) PFOC. Nut/. Acad. Sci. U. S. A. 81, 1494-

1498 Roberts, J. D., Preston, B. D., Johnson, L. A., Soni, A., Loeb, L.

A., and Kunkel, T. A. (1989) Mol. Cell. Biol. 9,469-476 Boosalis, M. S., Petruska, J., and Goodman, M. F. (1987) J. Biol.

Chem. 262, 14689-14696

19.

20.

21.

22.

23.

24. 25. 26.

27.

28.

29.

30.

31.

32. 33.

34.

35. 36.

37.

38. 39.

40.

Kuchta, R. D., Benkovic, P., and Benkovic, S. J. (1988) Biochem- istry 27,6716-6725

Kunkel, T. A., and Tindall, K. R. (1988) Biochemistry 27,6008- 6013

Perrino, F. W., and Loeb, L. A. (1989) J, Biol. Chem. 264,2898- 2905

Mendelman, L. V., Petruska, J., and Goodman, M. F. (1990) J. Biol. Chem. 265,2338-2346

Mendelman, L. V., Boosalis, M. S., Petruska, J., and Goodman, M. F. (1989) J. BioE. Chem. 264, 14415-14423

Kypr, J. (1988) J. Theor. Biol. 135,125-126 Lindahl, T., and Nyberg, B. (1972) Biochemistry 11,3610-3618 Perrino, F. W., Preston, B. D., Sandell, L. L., and Loeb, L. A.

(1989) PFOC. Natl. Acad. Sci. U. S. A. 86,8343-8347 Kunkel, T. A., and Bebenek, K. (1988) Biochim. Biophys. Acta

951,1-1.5 Cowart, M., Gibson, K. J., Allen, D. J., and Benkovic, S. J. (1989)

Biochemistry 28, 1975-1983 Freemont, P. S., Friedman, J. M., Beese, L. S., Sanderson, M. R.,

and Steitz, T. A. (1988) Proc. Natl. Acad. Sci. U. S. A. 85, 8924-8928

Streisinger, G., Okada, Y., Emrich, J., Newton, J., Tsugita, A., Terzaghi, E., and Inouye, M. (1966) Cold Spring Harbor Symp. Quant. Biol. 31, 77-84

Kunkel, T. A., and Soni, A. (1988) J. Biol. Chem. 263, 14784- 14789

de Boer, J. G., and Ripley, L. S. (1988) Genetics 118, 181-191 Papanicolaou, C., and Ripley, L. S. (1989) J. Mol. Biol. 207,335-

353 Miller, M., Harrison, R. W., Wlodawer, A., Appella, E., and

Sussman, J. L. (1988) Nature 334,85-86 Joyce, C. M. (1989) J. Biol. Chem. 264, 10858-10866 Bebenek, K., and Kunkel, T. A. (1990) PFOC. Nutl. Acad. Sci. U.

S. A. 87,4946-4950 Mizrahi, V., Benkovic, P., and Benkovic, S. J. (1986) PFOC. Natl.

Acad. Sci. U. S. A. 83,5769-5773 Sinha, N. K. (1987) PFOC. Natl. Acad. Sci. U. S. A. 84,915-919 Brosius, S., Grosse, F., and Krauss, G. (1983) Nucleic Acids Res.

11,193-202 Reyland, M. E., and Loeb, L. A. (1987) J. Biol. Chem. 262,

10824-10830

” T. A. Kunkel, unpublished observations.

Fidelity of Derivatives of E. coli DNA Polymerase

i I ‘5-mwr P 4 -

_ - - -

‘4-nxr F’:: ‘3-m*r 1’::

I ; 23

\3 TCTTTT-5 CTTTT -*5

-TGACTGAGAAAA - -TGACTGAGAAAA- 87 sa

C A=-r - m-m C%---- T’A -

i!

0 Co. .*s .I .2 .s I

d CTPtmM, ! I ‘os ” 2 ‘5 ’ 2 5’ 1

dATP (n,M,


B

(..(J. ’ 6 I1 I \.I

(;r - \c (“G r.h 5 I.\* c.c* -

i! 1 I : 1

the fidelity of dna synthesis catalyzed by derivatives of ... · the fidelity of dna synthesis...

Documents