Requirement of Watson-Crick Hydrogen Bonding for DNA Synthesis by Yeast DNA Polymerase η

Abstract

Classical high-fidelity DNA polymerases discriminate between the correct and incorrect nucleotides by using geometric constraints imposed by the tight fit of the active site with the incipient base pair. Consequently, Watson-Crick (W-C) hydrogen bonding between the bases is not required for the efficiency and accuracy of DNA synthesis by these polymerases. DNA polymerase η (Polη) is a low-fidelity enzyme able to replicate through DNA lesions. Using difluorotoluene, a nonpolar isosteric analog of thymine unable to form W-C hydrogen bonds with adenine, we found that the efficiency and accuracy of nucleotide incorporation by Polη are severely impaired. From these observations, we suggest that W-C hydrogen bonding is required for DNA synthesis by Polη; in this regard, Polη differs strikingly from classical high-fidelity DNA polymerases.

Classical DNA polymerases, such as T7 and Escherichia coli Klenow, synthesize DNA with high fidelity, and they are unable to replicate through DNA lesions (7). By contrast, the eukaryotic polymerase η (Polη) is a low-fidelity enzyme (16, 28, 45) with the ability to replicate through DNA lesions. Polη is unique among eukaryotic DNA polymerases in its proficient ability to replicate through UV-induced cyclobutane pyrimidine dimers (CPDs) (15). Remarkably, both yeast and human Polη insert A's opposite the two T's of a cis-syn thymine-thymine (TT) dimer with the same efficiency and accuracy as they insert A's opposite undamaged template bases (16, 44). In addition to TT dimers, UV light also induces the formation of CPDs at 5′-TC-3′ and 5′-CC-3′ sites, and genetic studies with Saccharomyces cerevisiae have indicated a role of Polη in the accurate bypass of these lesions as well (48). Because of its role in promoting the error-free bypass of CPDs, inactivation of Polη in humans causes UV hypermutability (43) and results in the cancer-prone syndrome, the variant form of xeroderma pigmentosum (14, 27). Yeast Polη also replicates through an 8-oxoguanine (8-oxoG) lesion efficiently and accurately, and genetic studies with S. cerevisiae have corroborated the requirement of Polη in the error-free bypass of this DNA lesion (11).

From studies done with nonpolar isosteric analogs of natural DNA bases which lack the ability to form Watson-Crick (W-C) hydrogen (H) bonds, it has been concluded that W-C H bonding is not needed for DNA synthesis by classical high-fidelity DNA polymerases (30, 32, 33); rather, the geometric fit of the incoming nucleotide with the templating base within the polymerase active site is the principal determinant of the efficiency and fidelity of DNA synthesis in these polymerases (7, 9, 22). Here we examine the effects of difluorotoluene (F), which is virtually identical in shape, size, and conformation to thymine (T) but lacks the ability to form W-C H bonds with adenine (A) (Fig. 1A), on DNA synthesis by yeast Polη. Because of the inability of F to form W-C H bonds in water (37, 38), this compound is ideal for examining the relative importance of hydrogen bonds and steric effects on DNA synthesis. Previously, it has been shown that high-fidelity DNA polymerases such as T7 and E. coli Klenow efficiently replicate DNA containing the F analog; in fact, these DNA polymerases incorporate an A opposite an F in the template with nearly the same efficiency and fidelity as they incorporate an A opposite template T (30, 32, 33).

FIG. 1.

(A) Structures of thymine (left) and difluorotoluene (right). (B) The running-start and standing-start DNA substrates. The 28-mer templates have a G, A, T, C, or F base at the N shown underlined and in boldface (position 24). The running-start substrate has an 18-mer primer, and the standing-start substrate has a 23-mer primer. (C) An autoradiogram of the running-start (lanes 1 to 3) and standing-start (lanes 4 to 6) assays with E. coli Klenow (exo⁻) and yeast Polη. Lanes 1 and 4 contain no protein, lanes 2 and 5 contain a template T residue, and lanes 3 and 6 contain a template F residue.

Here we show that Polη is ineffective at replicating DNA when F is present in the template or is used as the incoming nucleotide, and steady-state kinetic analyses indicate that the efficiency of incorporating an A opposite template F or the efficiency of incorporating an F opposite template A is reduced to almost the same level as the efficiency for incorporating the wrong nucleotides. We discuss these observations together with the other properties of Polη and conclude that W-C H bonding plays an important role in the efficiency and accuracy of DNA synthesis by Polη.

MATERIALS AND METHODS

Proteins and other materials.

Yeast Polη was expressed and purified as described previously (13), except that after binding of the glutathione S-transferase (GST)-Polη to a glutathione-Sepharose 4B column (Amersham) and washing, Polη was eluted by treatment with Precision protease (Amersham) for 16 h at 4°C to cleave off the GST tag. The resulting full-length Polη protein with a 7-amino-acid leader peptide was stored at −80°C.

Klenow fragment (exo⁻; 5 U/ml) was purchased from New England BioLabs. Solutions of all four deoxynucleoside triphosphates (dNTPs) (100 μM) were purchased from Roche Molecular Biochemicals. The difluorotoluene deoxynucleoside 5′-triphosphate (dFTP) and the F phosphoramidite were synthesized and purified as previously described (33). Two oligodeoxynucleotides were used as primers, and five oligodeoxynucleotides were used as templates (Fig. 1B), where N is a G, A, T, C, or F residue. Primer strands (1 μM) were 5′ ³²P end labeled by using polynucleotide kinase (Boehringer Mannheim) and [γ-³²P]ATP (Amersham) at 37°C for 1 h and were purified by using a Sephadex G25 spin column (Pharmacia). 5′ ³²P-end-labeled primer strands (250 nM) were annealed to the template strands (350 nM) by incubating them in 50 mM Tris HCl (pH 7.5)-100 mM NaCl at 90°C for 2 min, followed by slow cooling to room temperature for several hours.

DNA polymerase assays.

All DNA polymerase reactions were carried out in 35 mM Tris HCl (pH 7.5), 20 mM NaCl, 5 mM MgCl₂, 5 mM dithiothreitol, 100 μg of bovine serum albumin/ml, and 10% glycerol at 22°C. Running- and standing-start reactions contained all four dNTPs (2 μM each), 50 nM of 5′ ³²P-end-labeled DNA substrates, and either 1 nM yeast Polη or E. coli Klenow (exo⁻). Reaction mixtures were incubated for 30 min, quenched with 10 volumes of formamide loading buffer, and run on a 15% polyacrylamide sequencing gel containing 8 M urea. Single-nucleotide incorporation reaction mixtures contained only one dNTP (2 μM), a 50 nM concentration of 5′ ³²P-end-labeled standing-start DNA substrate, and either 1 nM yeast Polη or Klenow fragment (exo⁻). Reaction mixtures were incubated for 15 min, quenched with 10 volumes of formamide loading buffer, and run on a 15% sequencing gel.

Steady-state kinetics assays.

The kinetics of nucleotide incorporation were measured using the conditions described above except that 1 nM yeast Polη, 50 nM 5′ ³²P-end-labeled standing-start DNA substrate (Fig. 1B), and various concentrations of each nucleotide (0 to 1,000 μM) were incubated for 2.5 min before quenching with 10 volumes of formamide loading buffer. The extended and unextended primers were separated on a 15% sequencing gel and quantitated by using a PhosphorImager (Molecular Dynamics). The observed rate of nucleotide incorporation was graphed as a function of nucleotide concentration, and the k_cat and K_m steady-state parameters were obtained from the best fit to the Michaelis-Menten equation. The frequency of incorrect nucleotide incorporation, f_inc, was calculated using the following equation: f_inc = (k_cat/K_m)_incorrect/(k_cat/K_m)_correct (3).

RESULTS

DNA synthesis past a template F residue.

Because it has been shown previously (33) that H bonds are relatively unimportant in the case of the exonuclease-deficient Klenow fragment of E. coli DNA polymerase I, we used this enzyme as a basis for comparison in this study. First, we compared the abilities of Klenow fragment (exo⁻) and Polη to synthesize DNA on two substrates containing the template F residue (Fig. 1B). In the running-start substrate, the target T or F residue is the sixth available template base, while in the standing-start substrate, the target T or F residue is the first available template base. As shown in Fig. 1C, Klenow fragment (exo⁻) was able to synthesize DNA past a template F residue (left panel, lanes 3 and 6). In the running-start assay (Fig. 1C, lanes 3), no stall site was visible at position 23, which is immediately prior to the template F residue, but a stall site was visible at position 24, which is opposite the template F residue. This indicates that Klenow fragment (exo⁻) was not blocked for nucleotide incorporation opposite the F residue but was hindered in extending from the nucleotide inserted opposite the F residue (31). By contrast, Polη was unable to synthesize DNA past the target template F residue (Fig. 1C, right panel, lanes 3 and 6). In the running-start assay, a strong stall site was visible at position 23 (Fig. 1C, lanes 3), indicating that Polη was unable to insert nucleotides opposite the F residue.

Next, we compared the abilities of Klenow fragment (exo⁻) and Polη to insert dGTP, dATP, dTTP, dCTP, and dFTP opposite template G, A, T, C, and F residues. As shown in the upper panels of Fig. 2, Klenow fragment (exo⁻) incorporated the correct nucleotide opposite each of the template residues, dFTP was incorporated opposite both template A and template F residues, and both dATP and dFTP were incorporated opposite the template F residue (33). By contrast, for Polη (Fig. 2, lower panels), only the correct nucleotides were incorporated opposite each of the template residues. Neither the incorporation of dFTP opposite any of the template residues nor the incorporation of any dNTP opposite the template F residue was observed.

FIG. 2.

Incorporation of each dNTP opposite template G, A, T, C, and F residues by E. coli Klenow (exo⁻) and yeast Polη.

Steady-state kinetics of nucleotide incorporation.

We used steady-state kinetics assays to quantify the efficiency (k_cat/K_m) and the accuracy (f_inc) of dTTP and dFTP incorporation opposite each of the four template residues for Polη. dTTP was incorporated opposite template A with an efficiency which was approximately 90-, 70-, and 800-fold greater than that of the incorporation of dTTP opposite G, T, and C template residues, respectively (Table 1). However, dFTP was incorporated opposite template A with an efficiency that was ∼220-fold lower than that of dTTP incorporation opposite template A (Table 1). By contrast, dFTP was incorporated opposite templates G, T, and C with less than a 10-fold reduction in efficiency relative to dTTP incorporation opposite templates G, T, and C. Thus, overall, the incorporation of dFTP opposite template A occurred with very low fidelity, being only two- to threefold more efficient than that opposite templates G, T, and C (Table 1).

TABLE 1.

Steady-state kinetics parameters for the incorporation of dTTP and dFTP opposite all four template bases by yeast Polη

Base pair	kcat (min−1)	Km (μM)	kcat/Km (μM−1 min−1)	finc
dTTP · G	13 ± 1	140 ± 20	0.093	1.1 × 10−2
dTTP · A	7.0 ± 1	0.86 ± 0.24	8.1	1.0
dTTP · T	13 ± 1	110 ± 20	0.12	1.5 × 10−2
dTTP · C	NDa	>500	0.010b	1.2 × 10−3
dFTP · G	3.1 ± 0.4	260 ± 50	0.012	1.5 × 10−3
dFTP · A	10 ± 0.5	270 ± 30	0.037	4.6 × 10−3
dFTP · T	2.8 ± 0.4	190 ± 40	0.015	1.9 × 10−3
dFTP · C	2.6 ± 0.3	220 ± 40	0.012	1.5 × 10−3

^aND, not determined.

^bSince the rate of nucleotide incorporation was linear throughout the range of nucleotide concentration used, the k_cat/K_m value was obtained from the slope of the line.

We also used steady-state kinetics assays to determine the efficiency and accuracy of nucleotide incorporation opposite template T and template F residues (Table 2). dATP was incorporated opposite a template T residue with an efficiency which was 50-, 120-, and 30-fold greater than that of the incorporation of dGTP, dTTP, and dCTP, respectively. The better-than-expected incorporation of dCTP (45) likely reflects the fact that in the standing-start DNA substrate used for these experiments (Fig. 1B), dCTP can base pair with the next template G residue by “looping out” the T residue. Importantly, dATP was incorporated opposite template F with an efficiency that was 240-fold lower than that of dATP incorporation opposite template T (Table 2). Relative to the incorporation of dGTP and dTTP opposite template T, the incorporation of these nucleotides opposite F was reduced by 15- and 8-fold, respectively. The incorporation of dCTP opposite F was about the same as that opposite T, further supporting the looping out notion. Overall, dATP was incorporated opposite template F with only a three- to fourfold greater efficiency than that of the incorporation of dGTP and dTTP (Table 2), again demonstrating low fidelity in the absence of W-C hydrogen bonds.

TABLE 2.

Steady-state kinetics parameters for the incorporation of all four nucleotides opposite templates T and F by yeast Polη

Base pair	kcat (min−1)	Km (μM)	kcat/Km (μM−1 min−1)	finc
dGTP · T	13 ± 0.5	56 ± 8	0.23	1.9 × 10−2
dATP · T	9.1 ± 0.8	0.78 ± 0.18	12	1.0
dTTP · T	13 ± 1	130 ± 30	0.10	8.3 × 10−3
dCTP · T	9.5 ± 0.4	23 ± 7	0.41	3.4 × 10−2
dGTP · F	4.0 ± 0.3	250 ± 50	0.016	1.3 × 10−3
dATP · F	9.8 ± 0.6	200 ± 40	0.049	4.1 × 10−3
dTTP · F	5.3 ± 0.5	430 ± 100	0.012	1.0 × 10−3
dCTP · F	10 ± 0.5	18 ± 3	0.56	4.7 × 10−2

DISCUSSION

Depending on the sequence context and experimental conditions, the measured differences in standard free energy changes (ΔΔG°) for correct versus incorrect base pairs in solution range from 0.2 to 4.0 kcal/mol (1, 24, 36). These ΔΔG° values can provide as much as a several-hundredfold preference for incorporating the correct over the incorrect nucleotide. Since classical DNA polymerases, such as E. coli DNA polymerase I and T7, possess the ability to discriminate between correct and incorrect nucleotides by as much as 1 millionfold, they must achieve this discrimination by means other than mere W-C H bonding between the incoming nucleotide and the templating base. The currently accepted paradigm regarding DNA synthesis by classical polymerases holds that geometrical alignment of the incoming nucleotide with the template base is a major determinant of fidelity, whereas W-C H bonding makes relatively little contribution to base selectivity, and experimental support for these conclusions has come from studies using isosteric analogs of DNA bases (9, 30, 32, 33).

To determine the importance of W-C H bonding for the efficiency and fidelity of nucleotide incorporation by Polη, a low-fidelity lesion bypass polymerase, we examined its ability to utilize F as either the incoming nucleotide or the template base. Our results demonstrate that Polη is highly inefficient at forming F · A base pairs compared to T · A base pairs. Although the inability of Polη to efficiently utilize F could reflect the requirement of W-C H bonding between the incoming nucleotide and the template base, several alternative explanations need to be considered first.

Our observations that Polη poorly incorporates an F opposite template A, or an A opposite template F, could arise from a dependence of Polη on minor-groove H bonding interactions with both the incoming nucleotide and the templating base. The latter possibility arises from the fact that F lacks the O2 minor-groove bond acceptor and therefore cannot form H bonds with the protein (31, 37, 38). We examined the contribution of any such protein interactions with the minor groove of the incipient base pair by using 3-deazaguanine (3DG), which differs from G in having a carbon instead of a nitrogen at the 3 position, and replacement of a normal G by 3DG in the template or the primer abolishes the H bonding interaction of the polymerase with the DNA at the position of 3DG (41). From steady-state kinetics studies, we determined that the placement of 3DG at the position of the templating base has no significant effect on the efficiency or accuracy of C incorporation by yeast Polη; however, the incorporation of 3DG opposite a C template is 120-fold less efficient than the incorporation of a G opposite C (47). These results indicate that yeast Polη makes no functional H-bonding interaction with the minor groove of the templating residue but needs minor-groove H-bonding interactions with the incoming nucleotide. Thus, if the effects observed with F were due to Polη's interactions with the minor groove, then the efficiency and fidelity of nucleotide incorporation would have been eliminated only when F was used as the incoming nucleotide but not when F was the templating base. Since the efficiency and fidelity of nucleotide incorporation are eliminated regardless of whether F is the incoming nucleotide or the templating base, we consider this explanation to be unlikely.

Another possible explanation for the inability of Polη to form F · A base pairs is that the active site of Polη is very tight along the C1′-C1′ axis. Because F does not form W-C H bonds with A, the F · A base pair is not expected to contract in the same way that a T · A base pair does. Consequently, the F · A base pair should be slightly larger than the T · A base pair and might not fit well within the active site of Polη. We consider this explanation to be highly unlikely given the high-resolution structures of Polη and several other members of the Y family of DNA polymerases. These structures all reveal active sites that are much more open and much less constrained than the active sites of classical DNA polymerases (25, 40, 42, 49). If the slightly larger size of the F · A base pair is not problematic for the highly constrained active sites of Klenow or T7 DNA polymerase, then it should be no more problematic for the relatively unconstrained active site of Polη.

In view of the abovementioned considerations, the inability of Polη to synthesize DNA with F in the template or as the incoming nucleotide can best be explained by its dependence on W-C H bonding for efficient and accurate nucleotide incorporation. Furthermore, the lesion bypass properties of Polη also support such an inference. Polη replicates through a cis-syn TT dimer by inserting A's opposite the two T's of the dimer, and it does so with the same efficiency and fidelity as when replicating through undamaged T's (16, 44). Although a cis-syn TT dimer disrupts the DNA helix, bending it by ∼30° and unwinding it by ∼9°, this distortion does not affect the ability of the two T's in the dimer to form normal W-C base pairs with A's (2, 12, 18, 20, 35). Yeast Polη also replicates through an 8-oxoG lesion as efficiently and accurately as through an undamaged G (11). An 8-oxoG in the anti conformation pairs with C involving the same three hydrogen bonds as in the G · C base pair; however, the template strand is significantly distorted in the vicinity of the lesion in the 8-oxoG · C pair (23, 26, 29, 34). Because of the considerable template distortion in this base pair, replicative polymerases prefer to insert an A opposite 8-oxoG (11, 39). An 8-oxoG in the syn conformation mimics a T and has the correct geometry to pair with an A via two hydrogen bonds (23, 26, 29, 34). Thus, in spite of the greater geometrical distortion, Polη prefers to form the 8-oxoG · C base pair, presumably because of the ability of this base pair to form normal W-C H bonds. In contrast to the efficient and accurate bypass of these two DNA lesions, Polη is very poor at replicating through an abasic site, a prototypical noninstructional lesion. Classical high-fidelity replicative polymerases from prokaryotes as well as eukaryotes predominantly insert an A opposite the abasic site because this nucleotide causes no significant distortion of the helix, as evidenced from nuclear magnetic resonance studies indicating that DNA containing an A opposite the abasic site retains all aspects of B-form DNA, and the unpaired A and the abasic residue lie inside the helix (4, 5, 17). Polη, however, is very inefficient at inserting an A or any other nucleotide opposite an abasic site. Thus, Polη is ∼2,000-fold less efficient in inserting an A opposite an abasic site than in inserting this nucleotide opposite a T template, and the insertion of other nucleotides opposite this lesion site is also highly inefficient (10). We attribute the inefficiency of Polη in incorporating a nucleotide opposite an abasic site to the lack of W-C H bonding.

Also consistent with these findings is the recent observation that the blocking of thymine N3 with a methyl group in the template lowers the efficiency of dATP insertion by Polη by a factor of 50 (L. Sun, K. Zhang, L. Zhou, P. Hohler, E. T. Kool, X. F. Yuan, Z. Wang, and J. S. Taylor, submitted for publication). It is notable, however, that a nonpolar pyrene nucleotide is efficiently inserted at abasic sites by this enzyme. Thus, hydrogen bonding is not an absolute requirement with this large unnaturally shaped nucleotide. Presumably, its large size and strong stacking ability enable its favorable alignment in the active site.

Although we cannot as yet absolutely rule out the other factors described above, we propose that the most likely explanation for the present results with naturally shaped nucleotides is a requirement for W-C hydrogen bonding by this low-fidelity enzyme. Pre-steady-state kinetic analyses have indicated that yeast Polη discriminates poorly between the correct and incorrect nucleotide at both the initial nucleotide binding step (K_D^dNTP) and the subsequent nucleotide incorporation step (k_pol) (46). Since Polη shows a selectivity of only ∼5 at the initial nucleotide binding step (K_D^dNTP), which likely derives from W-C H bonding, the elimination of nucleotide discrimination at the initial binding step is to be expected with the T-to-F substitution. However, in view of the much greater selectivity (∼150-fold) at the k_pol step, which likely arises from geometric constraints, the elimination of nucleotide discrimination at this step requires explanation. Since the active site of Polη and related polymerases is relatively unconstrained and makes few intimate contacts with the incipient base pair (25, 40, 42, 49), we surmise that in the absence of W-C H bonding between the bases, the incoming nucleotide is able to move about in the active site of Polη, and as a result, its base and/or triphosphate are not positioned correctly for the subsequent induced-fit conformation change step or for the subsequent chemistry. By contrast, in the classical high-fidelity DNA polymerases such as T7 and Klenow fragment (exo⁻), W-C H bonding is not required at this step (30, 32, 33) because the active sites of these enzymes are highly constrained, providing for a tight fit and proper alignment of the incipient base pair (6, 8, 19).

Although the current paradigm which holds that the geometric-shape complementarity of the incoming nucleotide with the template base is a major determinant of fidelity and that W-C H bonding makes a much less important contribution remains valid for classical high-fidelity polymerases (21), Polη appears not to conform to this general rule. We attribute the exceptional nature of Polη to its more open and less highly constrained active site, which it needs to tolerate the geometric distortions in the DNA when efficiently bypassing lesions such as the cis-syn TT dimer and 8-oxoG. However, the geometry of the incipient base pair is also likely to be important for Polη. This is because for certain base pairs, Polη preferentially incorporates the correct nucleotide several thousandfold more efficiently than the incorrect nucleotide (45) and because Polη requires minor-groove interactions between the protein and the incoming nucleotide substrate (47). The latter observation raises the possibility that this interaction provides for a “geometry-sensing” mechanism similar to those proposed for classical DNA polymerases. Nevertheless, Polη differs strikingly from high-fidelity DNA polymerases in its strong dependence on W-C H bonding for efficient and accurate DNA synthesis.