Fidelity of Eucaryotic DNA Polymerase δ Holoenzyme fromSchizosaccharomyces pombe *

The fidelity of Schizosaccharomyces pombe DNA polymerase δ was measured in the presence or absence of its processivity subunits, proliferating cell nuclear antigen (PCNA) sliding clamp and replication factor C (RFC) clamp-loading complex, using a synthetic 30-mer primer/100-mer template. Synthesis by pol δ alone was distributive. Processive synthesis occurred in the presence of PCNA, RFC, and Escherichia coli single strand DNA-binding protein (SSB) and required the presence of ATP. “Passive” self-loading of PCNA onto DNA takes place in the absence of RFC, in an ATP-independent reaction, which was strongly inhibited by SSB. The nucleotide substitution error rate for pol δ holoenzyme (HE) (pol δ + PCNA + RFC) was 4.6 × 10−4 for T·G mispairs, 5.3 × 10−5 for G·G mispairs, and 4.5 × 10−6 for A·G mispairs. The T·G misincorporation frequency for pol δ without the accessory proteins was unchanged. The fidelity of pol δ HE was between 1 and 2 orders of magnitude lower than that measured for the E. coli pol III HE at the same template position. This relatively low fidelity was caused by inefficient proofreading by the S. pombepolymerase-associated proofreading exonuclease. The S. pombe 3′-exonuclease activity was also extremely inefficient in excising primer-3′-terminal mismatches in the absence of dNTP substrates and in hydrolyzing single-stranded DNA. A comparison of pol δ HE with E. coli pol IIIα HE (lacking the proofreading exonuclease subunit) showed that both holoenzymes exhibit similar error rates for each mispair.

The enzymes principally responsible for catalyzing procaryotic and eucaryotic DNA replication share many common elements. Eucaryotic DNA polymerases ␦ and ⑀ have the ability to proofread replication errors using pol 1 -associated 3Ј 3 5Ј-exo-nuclease activity (1,2), a property shared with Escherichia coli pols I-III (3). Each of these enzymes copies DNA with extremely low processivity, typically adding less than 30 nt before dissociating. There are closely analogous groups of eucaryotic and procaryotic polymerase accessory proteins that interact with the non-processive core pols, forming highly processive polymerase holoenzymes. E. coli pol III HE and pol II HE bind to the ␤ dimeric sliding clamp (4 -6), whereas eucaryotic pol ␦ HE and pol ⑀ HE bind to the PCNA trimeric sliding clamp (7). The processivity clamps are loaded on and off the DNA by clamp loading complexes, ␥ complex in E. coli (8 -11) and RFC in eucaryotic cells (7,(12)(13)(14).
Extensive studies on the fidelity properties of core DNA polymerases have been reported over the past 3 decades focusing on biochemical and kinetic analysis of deoxynucleotide insertion specificity and the reduction in pol-generated errors by proofreading exonucleases (15)(16)(17)(18), whereas there are but a paucity of experiments reporting on the fidelity properties of the more biologically relevant pol HE systems. Previous experiments employing pol HE systems have attempted to probe the fidelity of leading versus lagging strand synthesis using mutational reporter sequences (e.g. lacZ) (19,20) and to visualize synthesis past DNA damage sites using two-dimensional gel electrophoresis (21,22). We have recently generalized a gel kinetic assay originally designed to measure polymerase fidelity in the absence of proofreading (23,24), enabling fidelity measurements to be made at arbitrary p/t DNA sites in the presence of proofreading and pol accessory proteins (25)(26)(27).
There are a variety of questions regarding the fidelity properties of holoenzymes that can be investigated systematically using gel kinetic methodology. Recently, measurements on the fidelity of calf thymus pol ␦ were made in the presence and absence of PCNA at normal (28) and abasic (29) template sites. In this paper, we report on the base substitution error rate of the Schizosaccharomyces pombe pol ␦ HE and core for comparison with data with the E. coli pol III HE-catalyzed error rates (26) determined in the same sequence context.

Materials
Proteins cloned S. pombe pol ␦, PCNA, and RFC were purified as described (30). The enzyme reaction buffer contained 40 mM Tris⅐HCl, pH 7.8, 170 g/ml bovine serum albumin, 0. 5 1 The abbreviations used are: pol, DNA polymerase; HE, holoenzyme comprised of DNA polymerase ϩ processivity subunits, proliferating cell nuclear antigen (PCNA) sliding clamp, and replication factor C (RFC) clamp loading complex for S. pombe, and ␤ sliding clamp and ␥ clamp loading complex for E. coli; SSB, E. coli single strand DNAbinding protein; RPA, replication protein A, eucaryotic single strand DNA-binding protein and bovine serum albumin were purchased from Amersham Pharmacia Biotech.
DNA Substrates-The p/t DNA was made up of a synthetic 100-mer template annealed to complementary 30-or 35-mer primers or to a 35-mer primer containing a single noncomplementary base at its 3Јend. The 30-mer primer was annealed at the middle of the template leaving equal length (35 nt) ssDNA overhangs on each side. The matched 35-mer primer was annealed to the template leaving 35 nt of ssDNA at the 3Ј-end of the template and 30 nt of ssDNA at 5Ј-end. The mismatched 35-mer primer was identical to the matched 35-mer except that the nucleotide at the 3Ј-end contained an A in place of C. All oligomers were synthesized on an Applied Biosystems 392 DNA/RNA synthesizer (Perkin-Elmer) and gel-purified. The 100-mer was synthesized as two half-length oligomers and then ligated together.
The sequences for the 30-mer primer/100-mer template were as follows: where G is the target site where misincorporation frequencies were measured.
The sequences for the matched 35-mer primer/100-mer template were as follows: The sequences of the mismatched 35-mer primer/100-mer template differed only in that A replaced C at the primer-3Ј-end.

Methods
The primer was 5Ј-end-labeled with 32 P using T4 polynucleotide kinase in enzyme reaction buffer at 37°C for 60 min. p/t DNA was annealed in enzyme reaction buffer using a ratio of 1 primer to 1.2 templates by heating to 90°C and gradually cooling to room temperature. The concentration of p/t DNA after annealing was 100 nM (primer termini).
Assay for 3Ј-Exonuclease Activity of S. pombe pol ␦-10 nM either matched or mismatched 35/100-mer DNA were incubated at 37°C with 10 g/ml (0.2 unit/l) S. pombe pol ␦ in reaction buffer in the presence and absence of all 4 dNTPs (0.5 mM each if present) in separate reactions containing 20 l. One unit of pol ␦ supports the incorporation of 1 nmol of dTMP under the conditions specified above. The 35-mer primers were 5Ј-end-32 P-labeled. The mismatched 35-mer primer (10 nM) was used as single-stranded DNA substrate and incubated at 37°C with 10 g/ml (0.2 unit/l) S. pombe pol ␦ in reaction buffer (20 l). Aliquots (4 l) were removed from each reaction and quenched by mixing with 10 l of 20 mM EDTA, 95% formamide at different time points. Reaction products were separated on a 12% denaturing polyacrylamide gel run for 2 h at 2,000 V. The amount of primer extension catalyzed by pol ␦ (gel bands above the primer band) or degradation catalyzed by pol ␦ 3Ј-exonuclease activity (gel bands below the primer band) was measured as percentage of total gel band intensity in each lane using a PhosphorImager (Molecular Dynamics, Sunnyvale, CA).
Gel Kinetic Fidelity Analysis-A gel fidelity assay was used to determine the kinetics of incorporation of each of the four dNTPs opposite the target site (26,31). Primer extension reactions for pol ␦ in the absence of processivity accessory proteins were performed as follows. p/t DNA and the dNTP to be incorporated opposite the target were first mixed together in the reaction buffer. A mixture of S. pombe pol ␦ and the running-start nt (dATP) in the same reaction buffer was then added to initiate the reaction. Reactions were run for 5 min to measure the incorporation of dCMP opposite G and for 20 min to measure the misincorporation of dTMP opposite G. The assay conditions for correct incorporation satisfied single-completed hit conditions, whereby most of the p/t DNA molecules that undergo extension encounter a polymerase only once (27,31). However, multiple hit conditions were required to detect dTMP⅐G misincorporations for pol ␦ in the absence of the processivity factors, and the minor modifications required to analyze properly multiple encounter kinetics were made as described in Ref. 27. The final concentrations were 1 mM dATP, 10 nM p/t DNA, 1.0 g/ml S. pombe pol ␦, and dNTP concentrations as indicated in the figures. Control reactions were run for 5 min using just the running-start dATP to verify that misincorporation of the running-start nt opposite the target G site did not occur.
Primer extension reactions for pol ␦ in the presence of processivity accessory proteins were performed as follows. Solution A contained 33 nM p/t DNA, 150 g/ml RFC, 270 nM (PCNA) 3 , 1 M SSB, 2 mM ATP, and 4% glycerol in enzyme reaction buffer. Solution B consisted of the enzyme reaction buffer containing various concentrations of the dNTP to be incorporated opposite target site. Solution C contained 0.5 g/ml pol ␦, running start dATP (188 M), and 4% glycerol in the enzyme reaction buffer. The reaction was performed as follows: solution A (3 l) was mixed with solution B (3 l) and incubated at 37°C for 1 min to allow RFC to load PCNA onto the DNA; then solution C (4 l) was added to the mixture of A ϩ B to initiate the primer extension reaction. The final concentrations in the 10-l reaction mixture were 10 nM p/t DNA, 0.2 g/ml pol ␦, 75 M dATP, 45 g/ml RFC, 80 nM (PCNA) 3 , 300 nM SSB, 0.6 mM ATP, and various concentrations of dNTP for incorporation opposite the target site. Control reactions were run with the running-start dATP only to ensure that it did not misincorporate opposite G. The reactions, run at 37°C for 2 min for both correct incorporation and misincorporations opposite G, approximately satisfied single-completed hit conditions, in which about 20% of the primers were extended, so that no further corrections were required in the kinetic analysis. Reactions were quenched by addition of formamide/EDTA (20 l) to the reaction mixture. The samples were heated to 100°C for 6 min, placed on ice for 3 min, and then loaded on a 16% polyacrylamide denaturing gel. The gel was run at 2000 V for 4 h to separate reaction products.
Integrated polyacrylamide gel band intensities were measured on a PhosphorImager using ImageQuant software (Molecular Dynamics, Sunnyvale, CA). The nucleotide incorporation efficiency opposite the target site was obtained by measuring I T ⌺ /I TϪ1 , where I T ⌺ is the integrated gel band intensities of primers extended to the target site and beyond, and I TϪ1 is the integrated gel band intensity of primers extended to the site just prior to the target site (26,31).
A plot of the relative incorporation rate, I T ⌺ /I TϪ1 as a function of the dNTP substrate concentration, results in a rectangular hyperbola whose slope in the initial linear region is the apparent V max /K m . Apparent K m and relative V max values were obtained using a least squares fit to a rectangular hyperbola. The relative V max value is equal to the maximum value of I T ⌺ /I TϪ1 . In reactions where misincorporation opposite the target site was relatively inefficient, a plot I T ⌺ /I TϪ1 versus dNTP concentration showed little or no curvature, and apparent V max /K m values were obtained by a least squares fit of the data to a straight line. Apparent V max /K m values that were obtained under multiple-completed hit conditions were corrected to single-completed hit conditions as described in Ref. 27, but these corrections were essentially negligible. The misincorporation efficiency, f inc , which is the inverse of the fidelity, is given by the ratio shown in Equation 1, where the subscripts W and R refer to wrong and right incorporations, respectively. Measurement errors for V max /K m are Ϯ 30% and for f inc are Ϯ 40% (1 S.D.).

RESULTS
pol ␦, thought to be one of the principal eucaryotic replication polymerases (32), forms a processive holoenzyme complex in the presence of the trimeric PCNA sliding clamp and RFC, which is responsible for loading PCNA onto DNA (7). Eucaryotic pol ␦ is analogous to E. coli pol III core which uses both the ␤ sliding clamp and the clamp loading ␥ complex to form the highly processive pol III HE. In this study, we have measured the fidelity of S. pombe pol ␦ HE in the same sequence context that was used previously to study the fidelity of E. coli pol III HE (26). Although similar in some respects, the biochemical properties of the E. coli and S. pombe pol holoenzymes revealed several unanticipated differences, particularly with respect to the contribution of proofreading to fidelity.
A 30/100-mer p/t DNA Serves as a "Minimal" Substrate Supporting Processive Synthesis by S. pombe pol ␦ HE-We chose to measure the fidelity of S. pombe pol ␦ HE when copying a synthetic 100-mer DNA template. The advantage of using relatively short synthetic oligonucleotide minimal p/t DNA is that pol fidelity can be measured at arbitrary template target sites in defined sequence contexts that can be easily varied. However, before such a system can be used, it is necessary to show that it recapitulates the properties observed with much longer biological substrates, e.g. SV40 DNA (33).
A time course showing extension of a 32 P-labeled primer is arranged in seven groups of lanes to test the effects of PCNA, RFC, and SSB on pol ␦ processivity (Fig. 1). pol ␦ copied the synthetic 100-mer DNA template in a completely distributive manner in the absence of PCNA (Fig. 1, groups 1 and 3). A marked stimulation in pol ␦ processivity occurred in the presence of PCNA, PCNA ϩ RFC, or PCNA ϩ RFC ϩ SSB (Fig. 1,  group 2, group 4, and group 5, respectively). The observation that PCNA stimulated pol ␦ synthesis in the absence of RFC suggests that the processivity clamp can load onto the short DNA by itself and stabilize the pol ␦-p/t DNA complex. Thread-ing of PCNA onto linear DNA in the absence of RFC has been reported previously for Saccharomyces cerevisiae PCNA (34). We observed that SSB strongly inhibited synthesis by pol ␦ in the presence of PCNA (Fig. 1, group 6), but processive synthesis was restored by the addition of RFC (Fig. 1, group 5), clearly demonstrating that RFC was active in the assay.
The apparent stimulation of the RFC-dependent reaction by SSB (compare Fig. 1, groups 4 and 5) was caused most likely by the inhibition of a 3Ј-exonuclease contaminant present in our purified preparation of RFC. This adventitious 3Ј-exonuclease appeared to digest the primer extension products causing a uniform reduction in the gel band intensities in group 4 bands relative to either groups 5 or 2, while maintaining similar processivity patterns for these three groups.
The assays shown in Fig. 1 were performed in the presence of ATP (1 mM) required for RFC-mediated loading of PCNA onto p/t DNA (12,(35)(36)(37). Further characterization of the effects of PCNA and RFC on pol ␦ synthesis was carried out by performing similar primer elongation experiments in the absence of ATP. We found that processive synthesis observed in the presence of PCNA (Fig. 1, group 2) was retained in the absence of ATP (data not shown), whereas ATP must be present to observe processive synthesis in presence of PCNA, RFC, and SSB ( Fig.  1, groups 5 and 7). We conclude the following: (i) ATP is required for loading of PCNA onto DNA by RFC but that PCNA can also "thread" itself onto short linear p/t DNA in the absence of RFC, in a "passive" reaction not requiring ATP; (ii) SSB significantly inhibits the ATP-independent passive loading reaction but does not affect loading of PCNA by RFC.
S. pombe pol ␦ Base Substitution Fidelity-The nucleotide misincorporation value, f inc , which is the reciprocal of the fidelity, was determined using a gel kinetic assay suitable for measuring fidelity in the presence of proofreading and polymerase processivity proteins (27,31). The assay measured the relative rates of incorporating either a right (R) or wrong (W) nucleotide opposite a template target site base. Integrated gel band intensities corresponding to primers extended opposite a template target site and beyond were compared with extended primers terminating 1 base before the target site and were plotted as a function of dRTP and dWTP substrate concentrations to determine V max /K m values, in accordance with Equation 1 (27,31) (see "Experimental Procedures").
Base substitution fidelity measurements were performed using pol ␦ alone (Fig. 2) and pol ␦ HE, i.e. pol ␦ ϩ PCNA, RFC, and SSB (Fig. 3). A 32 P-labeled primer was extended by incorporation of four running-start As prior to reaching the template target site G where fidelity was measured. A sketch of the p/t DNA sequence is shown at the top of Fig. 2.
pol ␦ incorporated four running-start As and then C opposite the target G in a thoroughly distributive manner in the absence of the processivity proteins (Fig. 2, dCTP lanes). The appearance of faint primer extension bands corresponding to misincorporation of T opposite G allowed us to compute a T⅐G misincorporation ratio of 5.6 ϫ 10 Ϫ4 (Fig. 2, dTTP lanes). G⅐G and A⅐G misincorporations were not detectable (data not shown).
Each of these misincorporations was, however, readily detected for the pol ␦ HE (Fig. 3), which synthesized DNA processively (compare Figs. 3 and 2). Note the presence of the high intensity primer extension bands terminating opposite the target site G and continuing further downstream (Fig. 3). The T⅐G error rate of 4.6 ϫ 10 Ϫ4 was similar to that for pol ␦ alone, suggesting that the processivity proteins have essentially no effect on the fidelity for this mispair. The error rates for G⅐G and A⅐G mispairs were 5.3 ϫ 10 Ϫ5 and 4.5 ϫ 10 Ϫ6 , respectively.
It is important to emphasize that the inability to detect G⅐G and A⅐G misincorporations in the absence of PCNA, RFC, and SSB does not imply that the 4-subunit pol ␦ core has higher fidelity than pol ␦ HE. Rather, the absence of target site misincorporation bands for the case of pol ␦ alone was caused by the distributive nature of the enzyme. The residence time on the p/t DNA was simply too short to allow pol ␦ to catalyze the most difficult misincorporation events during a single pol-p/t DNA encounter. The data indicated that the apparent K m values for incorporation of C opposite G were 72 M for pol ␦ and 1.6 M for pol ␦ HE (Figs. 2 and 3). The K m values for misincorporation of T opposite G were ϳ6900 M for pol ␦ and only 370 M for pol ␦ HE (Figs. 2 and 3). The observation that the apparent K m values were much higher for both right (C⅐G) and wrong (T⅐G) incorporations for pol ␦ alone is consistent with reduced processivity. Thus, a much higher concentration of dNTPs is required to attain one-half V max when the pol ␦-p/ tDNA dissociation occurs rapidly, as is the case for a highly distributive synthesis. Because f inc (Equation 1) is expressed as the ratio of V max /K m for wrong versus right incorporations, the sensitivity of the gel kinetic assay is reduced in proportion to the reduction in (V max /K m ) R for pol ␦ in the absence of processivity factors. Thus, the assay is roughly 100-fold more sensitive for pol ␦ HE, enabling detection of misincorporation ratios on the order of about 10 Ϫ6 to 10 Ϫ7 .
S. pombe pol ␦-associated 3Ј 3 5Ј-Exonuclease Activity Is a Weak Proofreader-We measured 3Ј 3 5Ј proofreading exonuclease activity for the 4-subunit pol ␦ core under synthesizing and non-synthesizing conditions on p/t DNA, using either matched or mismatched primer-3Ј-ends (Fig. 4A). Excision of the primer-3Ј-end containing an A⅐G mismatched base pair occurred more rapidly than removal of a C⅐G correctly matched pair both in the presence and absence of dNTP substrates. However, the pol ␦ exonuclease activity appeared extremely weak. Removal of a terminal A⅐G mismatch was detectable in a 3-min incubation in either the presence or absence of dNTP substrates (Fig. 4A), whereas in the presence of dNTPs, a low level of incorporation of a next correct dGMP⅐C onto an A⅐G mismatched base pair was observed in a 7-min incubation. The S. pombe pol ␦ exonuclease-to-polymerase ratio, is about 1 to 30. That is, the rate of extending a correct dCMP⅐G terminus is roughly 30 times greater than the rate of removal of a dAMP⅐G mismatched terminus in the absence of dNTP substrates. Indeed, the excision of dAMP from a terminal A⅐G mispair was remarkably inefficient with greater than 90% of the input p/t DNA remaining following a 40-min reaction. In contrast, degradation of ssDNA occurred more rapidly than p/t DNA (Fig. 4B). The degradation reaction appeared to be dis- The misincorporation efficiency f inc was computed using Equation 1 (see "Experimental Procedures"). The template sequence corresponding to individual primer extension bands is indicated at the right-hand side of the gel. The dNTP concentration values have been rounded to two significant figures. tributive, showing removal of about 6 nt during a 6-min reaction.
The extremely low nuclease/polymerase ratio suggests that 3Ј-exonuclease of the pol ␦ may not be effective in eliminating nucleotide substitution errors. We tested this supposition by measuring f inc (dTMP⅐G) for pol ␦ HE at different concentrations of the next correct dGTP substrate. We found no measurable change in the T⅐G misincorporation ratios (f inc ϭ 4.6 ϫ 10 Ϫ4 ) for pol ␦, when varying dGTP concentrations between 0 and 160 M (data not shown). Since a decrease in fidelity with increasing next correct dNTP concentration is a well established hallmark of proofreading (38,39), the absence of a dependence of fidelity on dNTP concentration implies that the 3Ј-exonuclease of pol ␦ may not be effective in eliminating polymerase-catalyzed base substitution errors. DISCUSSION pol ␦ is believed to be the primary replicative enzyme in eucaryotic cells responsible for carrying out processive DNA synthesis in the presence of PCNA, RFC, and RPA (32). Despite the importance of this enzyme, little is known regarding its fidelity properties in vitro and in vivo. In this paper, we have used a gel kinetic assay (24,26,31) to measure fidelity at an arbitrary template G site using the pol ␦ HE purified from S. pombe (30).
Processive Synthesis by pol ␦ Using a Synthetic p/t DNA Oligomer-It is convenient to synthesize relatively short DNA templates to investigate DNA polymerase fidelity using defined sequence contexts. However, prior to performing a fidelity analysis using S. pombe pol ␦ HE on a 30/100-mer p/t DNA, it was necessary to demonstrate that the PCNA sliding clamp stimulated pol ␦ processivity, dependent on the presence of RFC and ATP, since the presence of ATP is required for loading of PCNA onto DNA by the RFC clamp loading complex (12,(35)(36)(37). This requirement is potentially important because PCNA can also diffuse onto linear but not circular DNA in the absence of RFC and ATP (34).
Synthesis by pol ␦ alone was distributive on the 30/100-mer p/t DNA with the addition of about 6 nt following a 3-min reaction and increasing to just 7 nt at 8, 18, and 40 min (Fig. 1,  group 1). The enzyme remained active during the 40-min time course as shown by the increased primer extension band intensities at the later time points. In contrast, synthesis by the pol ␦ HE was much more processive, with the addition of 35 nt to reach the end of the template strand well within the first time point taken at 3 min (Fig. 1, group 5). Processive synthesis does not occur in the absence of either PCNA or ATP (Fig. 1, groups  3 and 7, respectively). One can also clearly observe the PCNAindependent passive clamp loading reaction, with full-length synthesis also occurring in less than 3 min (Fig. 1, group 2). However, it is important to note that the passive clamp loading reaction failed to occur in the presence of SSB (Fig. 1, group 7), ensuring that our fidelity measurements made with pol ␦ HE in the presence of SSB, required PCNA, RFC, and ATP to carry out processive primer elongation. Experiments in which RPA (human or S. pombe RPA) was substituted for E. coli SSB showed no significant differences in either the rates or fidelity of DNA synthesis (data not shown). Therefore, a specific requirement for eucaryotic SSB has not been demonstrated in our in vitro model system and remains an open question requiring further investigation.
Fidelity of S. pombe pol ␦ HE-Nucleotide misincorporation values for S. pombe pol ␦ were found to be f inc ϭ 4.6 ϫ 10 Ϫ4 (T⅐G), 5.3 ϫ 10 Ϫ5 (G⅐G), and 4.5 ϫ 10 Ϫ6 (A⅐G) (Fig. 3 and Table  I). The pol ␦ HE error rates can be compared with values obtained with E. coli pol III HE and proofreading-defective E. coli pol III␣ HE (26) containing the ␤ sliding clamp (analo-gous to PCNA), ␥ clamp loading complex (analogous to RFC), and SSB in the same p/t DNA sequence context (Table I). The fidelity of pol ␦ HE is considerably lower than pol III HE for each mispair. The reduction in fidelity compared with pol III HE is 82-fold (T⅐G), 76-fold (G⅐G), 2 and 11-fold (A⅐G).
The higher nucleotide misincorporation rates for pol ␦ HE appear to be attributable almost entirely to a severely compromised ability to proofread insertion errors made by the polymerase catalytic subunit. Indeed, a comparison of f inc for pol III␣ 2 In the fidelity comparison for G⅐G misincorporations, the template used for E. coli pol III HE contains the base A in place of C immediately downstream from the target G site because pol III HE can incorporate dGMP opposite the downstream C by a primer-template slippage mechanism called "dNTP-stabilized" misalignment (26). In contrast, S. pombe pol ␦ HE misincorporates dGMP directly opposite G when C is located at the 5Ј-side of the target.  HE (containing the ␣ polymerase subunit in the absence of the ⑀ proofreading and subunits) shows that the nucleotide misinsertion rates for pol ␦ and pol III are essentially the same (Table I). The reduction in fidelity for pol ␦ HE compared with pol III␣ HE is only 2.6-and 1.5-fold for T⅐G and G⅐G mispairs, respectively, whereas pol ␦ HE may be slightly (1.2-fold) more accurate in forming A⅐G mispairs. These small differences are not statistically significant.
The apparent "absence" of effective proofreading for S. pombe pol ␦ in the in vitro experiments is quite puzzling. By using the same assay and p/t DNA sequence to measure E. coli pol III fidelity, we observed an 8-fold reduction in fidelity as proofreading of mispaired A⅐G termini were reduced in the presence of high concentrations of a next-correct "rescue" dNTP (26). The "next nucleotide" reduction in fidelity is a well established hallmark of a proofreading polymerase (38,39) and confirms that the gel kinetic assay can be used to analyze the effects of proofreading on fidelity. We observed no significant differences in pol ␦ HE fidelity using a wide concentration range of next-nucleotide dNTP (data not shown), and we concluded, therefore, that S. pombe pol ␦ is unable to effect a significant reduction in polymerase insertion errors. The p/t DNA sequence, requiring incorporation of four As prior to reaching the target G site, was chosen to maximize proofreading, i.e. "all things being equal" proofreading is most effective in removing misinserted nucleotides adjacent to relatively unstable DNA regions (40,41). This latter point serves to emphasize the inability of pol ␦ to carry out effective error correction.
The absence of a next-nucleotide effect is consistent with our observation that pol ␦ has an extremely weak associated 3Јexonuclease activity (Fig. 4). This activity barely degraded p/t DNA containing mismatched primer 3Ј-ends in the absence of dNTP substrates (Fig. 4A), and although it was able to degrade the single-stranded 35-mer primer somewhat more effectively (Fig. 4B), its activity was far lower than that observed using either E. coli pol III core or HE (data not shown). The pol ␦-catalyzed DNA degradation rate was essentially unchanged in the presence of PCNA (data not shown). A small increase in the rate of DNA hydrolysis was, however, observed in the presence of PCNA ϩ RFC, which can be attributed to a low level of nuclease contamination in our most highly purified RFC fraction (data not shown).
In an earlier in vitro study, Kunkel and co-workers (42) were also unable to demonstrate a proofreading contribution to accuracy when using calf thymus pol ␦ to copy a lacZ␣ reporter gene sequence in the absence of processivity subunits, although pol ⑀ proofreading was readily apparent using the same gap filling assay. Although we are unaware of any in vivo mutational data for proofreading-deficient S. pombe mutants, there are such data in S. cerevisiae showing that proofreading for pols ␦ and ⑀ can substantially reduce base substitution errors (43,44). In view of this dichotomy, it seems reasonable to speculate that associated exonuclease activity of pol ␦ may be masked in vitro. In this regard, it is interesting to note that the S. pombe and S. cerevisiae pol ␦s differ in their subunit structure. In S. cerevisiae, this complex has been shown to be a dimer of the three-subunit complex (M r 125,000, 58,000, and 55,000 (45,46)) whereas the S. pombe pol ␦ has been shown to be a dimer of the four-subunit complex (M r 125,000, 55,000, 54,000, and 22,000 (30)). Whether the unique 22-kDa subunit found in S. pombe pol ␦, the product of the non-essential cdm1 ϩ gene (47) affects the proofreading function of the S. pombe pol␦ in vitro, remains to be explored. In addition, perhaps some other protein cofactor may be required to stimulate proofread-ing. As indicated here, the gel fidelity assay might prove useful as a means to identify and purify proofreading stimulatory factors from cell lysates. On the other hand, the presence of an alternative excision repair pathway in S. pombe that has been shown to excise mispaired bases (48), in addition to damaged DNA bases, raises the possibility that this repair pathway might compensate for a lack of effective proofreading by pol␦.