DNA Polymerase β Ribonucleotide Discrimination

DNA polymerases must select nucleotides that preserve Watson-Crick base pairing rules and choose substrates with the correct (deoxyribose) sugar. Sugar discrimination represents a great challenge because ribonucleotide triphosphates are present at much higher cellular concentrations than their deoxy-counterparts. Although DNA polymerases discriminate against ribonucleotides, many therapeutic nucleotide analogs that target polymerases have sugar modifications, and their efficacy depends on their ability to be incorporated into DNA. Here, we investigate the ability of DNA polymerase β to utilize nucleotides with modified sugars. DNA polymerase β readily inserts dideoxynucleoside triphosphates but inserts ribonucleotides nearly 4 orders of magnitude less efficiently than natural deoxynucleotides. The efficiency of ribonucleotide insertion is similar to that reported for other DNA polymerases. The poor polymerase-dependent insertion represents a key step in discriminating against ribonucleotides because, once inserted, a ribonucleotide is easily extended. Likewise, a templating ribonucleotide has little effect on insertion efficiency or fidelity. In contrast to insertion and extension of a ribonucleotide, the chemotherapeutic drug arabinofuranosylcytosine triphosphate is efficiently inserted but poorly extended. These results suggest that the sugar pucker at the primer terminus plays a crucial role in DNA synthesis; a 3′-endo sugar pucker facilitates nucleotide insertion, whereas a 2′-endo conformation inhibits insertion.

To maintain faithful DNA synthesis, DNA polymerases have evolved to select a dNTP from a pool of structurally similar molecules that preserve Watson-Crick base pairing. This is facilitated by geometric constraints (size, shape, and hydrogen bonding potential) imposed by the template strand, primer terminus, and polymerase. Although the fidelity of base substitution errors, and their correction, has been extensively studied, the fidelity of sugar discrimination has received much less attention. It is well recognized that the dNTP pool imbalances influence DNA polymerase fidelity. In this context, cellular rNTP levels are far greater than their dNTP counterparts (1,2). To prevent significant levels of RNA synthesis during replication and repair, DNA polymerases must inherently discriminate against nucleotides with a ribose sugar (i.e. possessing a 2Ј-OH) and select 2Ј-deoxyribose triphosphates. Previous studies with A-and B-family polymerases have found significant effects on nucleotide incorporation as a result of modifying the deoxyribose ring (3)(4)(5)(6)(7)(8). DNA polymerases insert ribonucleotides with a much lower efficiency than deoxynucleotides with the same base due to a slower rate of insertion and weaker binding.
DNA polymerase (pol) 2 ␤, an X-family member that also includes pol , pol , and terminal deoxyribonucleotidyltransferase (TdT), has been well characterized kinetically, structurally, and biochemically (9) making it a model DNA polymerase to probe sugar specificity. DNA polymerase ␤ is a critical component of the base excision repair (BER) pathway, which operates to repair simple DNA lesions. It is best suited for filling short DNA gaps (1-5 nucleotides) after a lesion-specific DNA glycosylase and apurinic/apyrimidinic endonuclease have removed the damaged base and incised the abasic site (9). DNA polymerase ␤ is composed of an 8-kDa lyase domain and a 31-kDa polymerase domain with functionally distinct subdomains as follows: DNA binding (D), catalytic (C), and nascent base pair binding (N) (10). Several global conformational changes occur when pol ␤ binds substrates. The most notable change occurs when the N-subdomain of the binary enzyme-DNA complex closes around the nascent base pair upon binding a correct or incorrect dNTP. This subdomain re-positioning from an open binary to a closed ternary complex is accompanied by subtle protein side chain and DNA-nucleotide re-adjustments (11)(12)(13)(14).
DNA polymerase ␤ has moderate fidelity, typically misinserting 1 nucleotide for 10 4 -10 7 insertions (15). To accurately replicate DNA, polymerases must stabilize the coding template base and the correct, but not the incorrect, incoming nucleotide (14). Correct dNTP binding induces closure of the N-subdomain that results in several key nucleic acid-protein interactions (10,16,17). Importantly, two ␣-helices (M and N) of this subdomain provide key interactions with the sugar and base moieties of the incoming nucleotide (Fig. 1). In addition, the primer terminus provides significant interactions that influence insertion efficiency (11,15,18). The C-subdomain includes acidic side chains , and Asp-256) that coordinate two essential divalent magnesium ions that are required for catalysis. Among other roles, these metals coordinate nonbridging oxygens of the triphosphate moiety of the incoming nucleotide. Several other side chains of the C-subdo-* This work was supported, in whole or in part, by National Institutes of Health main also participate in metal coordination or binding of the triphosphate portion of the incoming nucleotide.
One approach to dissect the mechanism of nucleotide selectivity is to employ a series of analogs that are strategically modified to resemble the natural substrate. This popular approach has been successfully applied to examine incoming nucleotide base attributes (size, volume, hydrogen bonding capacity, and charge) that influence nucleotide binding, insertion, and fidelity (19). To a lesser extent, this approach has also been used to probe triphosphate attributes that influence binding with pol ␤ (20 -24).
Relative to base substitution fidelity, much less is understood about sugar discrimination as it pertains to X-family DNA polymerases. The polymerase domain of members of this family is structurally homologous; however, many subtle structural differences exist that fine tune their catalytic activities to be able to utilize specific substrates necessary to fulfill their biological function (25). Here, we investigate the effects of modifying the deoxyribose sugar of the incoming nucleotide on insertion efficiency and fidelity of pol ␤. More importantly, we examine whether insertion of a ribonucleotide influences further synthesis or whether a ribonucleotide in the template (coding) position alters DNA synthesis or fidelity. The biological significance of these results is discussed.

EXPERIMENTAL PROCEDURES
Nucleotides-Ultrapure rNTP and dNTP solutions were purchased from Sigma, and [␥-32 P]ATP was obtained from PerkinElmer Life Sciences. The ddNTPs were purchased from Amersham Biosciences, and araCTP was purchased from Jena Biosciences (Germany).
DNA Preparation-A 34-mer oligonucleotide substrate with a single-nucleotide gap was prepared by annealing three high pressure liquid chromatography-purified oligonucleotides (Integrated DNA Technologies) to create a single-nucleotide gap at position 16. Each oligonucleotide was dissolved in 10 mM Tris-HCl, pH 7.4, and 1 mM EDTA, and their concentrations were determined by UV absorbance at 260 nm. The upstream primer was 5Ј-labeled with [␥-32 P]ATP using Optikinase (United States Biochemical Corp.), and free radioactive ATP was removed using a Bio-Spin 6 column (Bio-Rad). The downstream oligonucleotide was synthesized with a 5Ј-phosphate. The DNA substrates were annealed (1:1.2:1.2, primer/template/downstream oligonucleotide) as described previously (see Fig. 2A) (15).
Kinetic Assays-Steady-state kinetic parameters for singlenucleotide gap-filling reactions were determined as described previously (15). Reactions typically contained 50 mM Tris-HCl, pH 7.4, 100 mM KCl, 5 mM MgCl 2 , 200 nM single-nucleotide gapped DNA, and varying concentrations of nucleoside triphosphate. Reactions were initiated by the addition of enzyme and run at 37°C. Enzyme concentrations and reaction time FIGURE 1. DNA polymerase ␤ nucleoside triphosphate binding pocket and key protein-nucleic acid interactions. A, dNTP binding pocket is composed of nucleic acid (primer terminus and templating base, n t , yellow) and protein (purple). The incoming nucleotide, 2Ј-deoxyuridine-5Ј-[(␣,␤)-imido] triphosphate, is shown hydrogen-bonded (green dashed lines) with the templating nucleotide (dA). The two active site Mg 2ϩ ions are illustrated as light blue spheres. B, stacking of the nascent base pair with the primer terminus positions O3Ј of the primer terminus for optimal attack on the ␣-phosphate of the incoming nucleotide. Two ␣-helixes (M and N) provide key interactions with the sugar and base moieties of the incoming nucleotide. In addition, Arg-183 (R183) and O3Ј of the incoming nucleotide hydrogen bond to a nonbridging oxygen on the ␤-phosphate (dashed green lines). C, Asp-276 (D276) of ␣-helix N is positioned above the sugar ring and approaches C2Ј. The incoming nucleotide is represented as a semi-transparent surface (gray) with C2Ј highlighted in pink. The side chain of Asp-276 is also represented as a surface (purple) just above C2Ј and would potentially block araCTP that has a hydroxyl at this position. D, primer terminal base pair is highlighted (yellow), and the other bases are gray. The backbone (rather than the side chain) of Tyr-271 (Y271) of ␣-helix M would potentially block a ribonucleotide with a hydroxyl at C2Ј (magenta). Additionally, the side chain of Tyr-271 hydrogen bonds with the minor groove edge of the primer terminal base (dashed green line). Arg-283 (R283) of ␣-helix N hydrogen bonds with the nucleotide opposite the primer terminus, (n Ϫ 1) t , at O4Ј of the sugar ring.
intervals were chosen so that substrate depletion or product inhibition did not influence initial velocity measurements. Reactions (30 l) were quenched with 15 l of 0.3 M EDTA and 45 l of 95% formamide dye (bromphenol blue and xylene cyanol). Products were separated on 20% denaturing polyacrylamide gels and quantified using phosphorimagery and ImageQuant software. Steady-state kinetic parameters were determined by fitting the rate data to the Michaelis-Menten equation.

RESULTS
Insertion and Misinsertion of Dideoxynucleotides-To determine how the structure of the deoxyribose sugar affects nucleotide incorporation by pol ␤, we examined various nucleotide sugars modified at C2Ј and C3Ј of the furanose ring and measured how efficiently pol ␤ inserted them in a single-nucleotide gapped DNA substrate (Fig. 2).
A dideoxynucleoside triphosphate (ddNTP) lacks the 3Ј-OH on the deoxyribose ring and thus prevents further DNA synthesis once it has been inserted (Fig. 2B). These modified nucleotides are the basis for Sanger DNA sequencing and are commonly employed to capture crystallographic ternary (polymerase-dideoxyterminated DNA-dNTP) substrate complex structures of DNA polymerases poised for catalysis. Steady-state kinetic analysis revealed that pol ␤ efficiently inserted ddCTP in a single-nucleotide gap with a templating dG, which was only 3.3-fold less efficient than insertion with dCTP (Table 1 and Fig. 3B). Similar results were obtained for the reciprocal base pair, i.e. pol ␤ incorporated ddGTP only 2-fold less efficiently than dGTP opposite template dC (data not shown).
Likewise, the absence of 3Ј-OH on the deoxyribose sugar had little effect on misincorporation. DNA polymerase ␤ misinserted dATP and ddATP opposite template dG with catalytic efficiencies of 6.0 ϫ 10 Ϫ6 and 2.5 ϫ 10 Ϫ6 s Ϫ1 M Ϫ1 , respectively. Similarly, dTTP and ddTTP were misincorporated opposite template dG with comparable catalytic efficiencies (Table 1). Thus, pol ␤ does not discriminate between dideoxy-and deoxyribose sugars. As expected, however, incorrect dNTPs and ddNTPs were inserted much more slowly and bound more weakly than when forming a Watson-Crick base pair.
DNA Polymerase ␤ Strongly Discriminates against Ribonucleotides-Although 3Ј-OH had little effect on nucleotide incorporation, adding an additional hydroxyl at C2Ј had a much   Fig. 2A, where X ϭ dC and Y ϭ Z ϭ dG. The results represent the mean Ϯ S.E. of at least two independent determinations.   Table 1. greater impact. DNA polymerase ␤ inserted rCTP more than 3 orders of magnitude less efficiently than dCTP opposite template dG in a single-nucleotide gapped substrate (Table 1 and Fig. 3B) consistent with previously reported results (27). The loss in efficiency was due to a lower binding affinity (70-fold) and slower rate of insertion of rCTP (220-fold). 3 Arabinofuranosylcytosine triphosphate (araCTP) is the active form of a common chemotherapeutic and antiviral drug. araCTP and rCTP have 2Ј-OH on opposite sides of the plane of the sugar (2Ј-OH is nearer cytosine for araCTP, see Fig. 2B). araCTP was incorporated much more efficiently than rCTP opposite dG (900-fold relative to rCTP and 9-fold lower than dCTP, see Table 1 and Fig. 3B), despite its close structural resemblance to rCTP. Products formed from the addition of either araCTP or rCTP were easily distinguishable from the incorporation of dCTP due to differences in gel mobility (data not shown). Thus, the simple inversion of the 2Ј-OH configuration results in a significant recovery of catalytic efficiency for pol ␤.
Significantly, despite the relative inefficient incorporation of rNTPs, pol ␤ inserted a nucleotide with an incorrect sugar (correct base, i.e. rCTP) more efficiently than misinserting a nucleotide with an incorrect base (correct sugar, i.e. dTTP, 4 see Table  1 and Fig. 3B) opposite template dG. This indicates that pol ␤ prefers incorporating the nucleotide that maintained Watson-Crick base pairing rather than the nucleotide that would maintain the identity of the sugar suggesting that the incorrect base distorted the active site more than a nucleotide with an incorrect sugar.
Efficient Extension of rNMP-terminated Primers-Because pol ␤ can incorporate rNTPs, albeit inefficiently, we probed how detrimental a single ribonucleotide could be in various DNA gap contexts on pol ␤ insertion and fidelity. For a ribonucleotide to persist in DNA, the incorporated ribonucleotide needs to be extended to bury this aberrant residue. DNA polymerase ␤ does not possess an intrinsic 3Ј 3 5Ј-exonuclease activity that might remove a nucleotide with the wrong sugar. To determine how efficiently pol ␤ could extend a ribonucleotide, we designed an upstream primer with a single rCMP at the 3Ј-end that correctly base-paired with the upstream template base dGMP (Fig. 4A), and we examined the ability of pol ␤ to add the next correct nucleotide. We maintained dG as our template nucleotide to directly compare with the kinetic parameters obtained for ribonucleotide incorporation ( Table  1). The catalytic efficiencies for dCTP incorporation were comparable for rNMP-and dNMP-terminated primers, 1.0 and 0.90 s Ϫ1 M Ϫ1 , respectively ( Fig. 4B and Table 2). Moreover, we determined catalytic efficiencies for misinsertion of dTTP opposite template dG for both dNMP-and rNMP-terminated primers. The results indicate that a ribonucleotide at the primer terminus does not influence misinsertion or fidelity (Fig. 4B).
The relative discrimination for dTTP compared with dCTP was 9 ϫ 10 4 for both primers. These results suggest that a ribonucleotide at the 3Ј-primer terminus does not significantly perturb the conformation of the primer terminus or the incoming nucleotide.
To determine whether pol ␤ could insert two consecutive ribonucleotides, we measured the catalytic efficiency for rCTP incorporation opposite template dG using the rNMP-terminated primer. DNA polymerase ␤ was able to incorporate rCTP  Tables  1 and 2.

TABLE 2 Incorporation and misinsertion efficiencies on various ribonucleotide-containing oligonucleotides
All assays were performed as outlined under "Experimental Procedures." The sequences of the single-nucleotide DNA substrates used are provided in Fig. 2A, where X is the primer terminal nucleotide; Y is the template nucleotide opposite the primer terminus, and Z is the template (coding) nucleotide. The position where a ribonucleotide has been inserted in the oligonucleotides is underlined. The results represent the mean Ϯ S.E. of at least two independent determinations. on a ribo-terminated primer (i.e. rCMP) similar to that with a dCMP-terminated primer ( Fig. 4B and Table 2). These results demonstrate that ribonucleotide discrimination for the nascent base pair is independent of the identity of the sugar (i.e. deoxyribose or ribose) at the primer terminus. This also suggested that pol ␤ could insert consecutive ribonucleotides with the same efficiency as long as proper Watson-Crick base pairing was maintained.

Incoming nucleotide
In contrast to efficient insertion of araCTP, pol ␤ was extremely slow at extension of an araC-terminated primer (Fig.  5). DNA polymerase ␤ was preincubated with araCTP and a two-nucleotide gapped DNA substrate to generate a single-nucleotide gapped substrate. Correct insertion of dATP opposite the templating dT was measured (Fig. 5A). The efficiency of dATP addition on an araC-terminated primer (k cat /K m ϭ 8.8 ϫ 10 Ϫ3 s Ϫ1 M Ϫ1 ; Fig. 5B) was 110-fold lower than insertion opposite a single-nucleotide gap dT annealed to a dC-terminated primer (k cat /K m ϭ 1.0 s Ϫ1 M Ϫ1 , data not shown). Therefore, araCTP is inserted much more efficiently than rCTP, but pol ␤ prefers to add the next correct deoxynucleotide on a riboterminated primer rather than an arabino-terminated primer.
Influence of the Sugar Identity of the Template-Primer Terminus on Extension-Because the presence of a single ribonucleotide at the primer terminus had little effect on extension, we examined if other termini would also be tolerated by pol ␤. We modified the upstream template sugar from dGMP to rGMP and annealed the resulting template to the dC-and rC-terminated primers to create rG-dC and rG-rC (template-primer) 5 termini, respectively. We maintained correct Watson-Crick base pairing at the terminus in all sequence contexts because mismatched termini have been shown to have variable effects on dNTP extension (15). The kinetic parameters for extension of these termini are tabulated in Table 2, and their catalytic efficiencies for extension are plotted in Fig. 6.
Altering the upstream template strand sugar from dGMP to rGMP decreased the catalytic efficiency of correct incorporation (i.e. dCTP opposite template dG) 8.2-fold and decreased the efficiency of dTTP misinsertion 10-fold (Table 2 and Fig.  6B). Surprisingly, changing the terminus from a deoxynucleotide to a ribonucleotide base pair (i.e. dG-dC to rG-rC) had less of an effect on both correct incorporation of dCTP and misinsertion of dTTP (2-and 3-fold, respectively; see Table 2 and Fig.  6B), suggesting that a ribonucleotide base pair at the terminus does not significantly distort the DNA duplex relative to a deoxy-terminal base pair. The relative discrimination of dTTP compared with dCTP was not significantly affected by rG-dC or rG-rC termini compared with the dG-rC and dG-dC termini (Figs. 4B and 6B and Table 2).
We also examined incorporation of rCTP opposite template dG using both rG-dC and rG-rC termini. Interestingly, when the upstream template nucleotide was modified from dGMP to rGMP, rCTP insertion decreased 25-fold (Table 2 and Fig. 6B), 5 The terminus nomenclature is aY Ϫ bX, where X and Y are the bases A, T/U, C, or G, and a and b are the sugars, deoxyribose (d) or ribose (r), as illustrated in Fig. 2A.   but when the terminus became a ribonucleotide base pair (i.e. rG-rC), rCTP efficiency decreased only 7.3-fold relative to when the terminus was a deoxynucleotide base pair (i.e. dG-dC, see Table 2 and Fig. 6B). Together, these data suggest that a ribonucleotide immediately upstream from the templating base has the greatest effect on the efficiency of extension by pol ␤ in all of the termini sequence contexts examined. Templating Nucleotide Sugar Alters Nucleotide Insertion-A persistent ribonucleotide in duplex DNA may serve as a templating residue during DNA replication and repair. Thus, we investigated whether a ribonucleotide templating residue influences dNTP insertion and fidelity. To determine the effect of a ribose sugar in the templating position, we changed the dG at the templating position of our single-nucleotide gapped DNA substrate to rG, to maintain the identity of the base while altering the sugar (Table 2 and Fig. 7A). As shown in Fig. 7B, changing this template position to a ribonucleotide generally decreased catalytic efficiency for correct insertion of dCTP ϳ8-fold, whereas misinsertion of dTTP was 11-fold less efficient with a template rG relative to dG (Table 2). Accordingly, the relative discrimination for dTTP was unaltered because both correct and incorrect efficiency decreased to about the same extent. Thus, a ribonucleotide in the template (coding) position results in a modest distortion of the nascent base pair for both correct and incorrect insertions. Interestingly, pol ␤ was able to insert rCTP opposite a ribonucleotide at the template position (i.e. creating a ribonucleotide base pair), but this process was 18-fold less efficient than inserting rCTP opposite a deoxynucleotide template sugar ( Fig.   7B and Table 2). This was surprising given the fact that pol ␤ is a DNA-dependent DNA polymerase, yet it apparently can read a ribonucleotide in a template to incorporate an incoming rNTP (albeit slow) and make a proper Watson-Crick base pair.

DISCUSSION
Dideoxynucleotide Sensitivity-To determine the contribution of the deoxyribose ring on nucleotide insertion, we examined the ability of pol ␤ to incorporate sugars modified at the 2Ј and 3Ј positions. Crystallographic structures of substrate complexes of polymerases from several families indicate that 3Ј-OH of the incoming nucleotide is within hydrogen bonding distance to a nonbridging oxygen on the ␤-phosphate (pro-(S p )). For pol ␤, and other members of the X-family, a conserved arginine residue (Arg-183 in pol ␤, see Fig. 1B) also forms a hydrogen bond with this nonbridging oxygen. DNA polymerase ␤ inserted ddNTPs only 2-3-fold less efficiently than dNTPs in a single-nucleotide gapped DNA template indicating that the polymerase did not require a hydroxyl at C3Ј of the incoming nucleotide for efficient nucleotide incorporation. More importantly, this result suggests that Arg-183 was sufficient to stabilize the triphosphate moiety of the incoming nucleotide in the absence of O3Ј.
The sensitivity of A-family DNA polymerases toward ddNTPs depends on the identity (phenylalanine or tyrosine) of a conserved aromatic residue. Escherichia coli DNA polymerase I (Klenow fragment, Phe-762) and TaqDNA polymerase (Phe-667) discriminate against ddNTPs, whereas T7 DNA polymerase (Tyr-526) and pol ␥ (Tyr-951) do not. Changing this residue to the alternative aromatic side chain alters dideoxynucleotide discrimination accordingly; phenylalanine discriminates against and tyrosine facilitates ddNTP incorporation (3,28,29). These results suggest that loss of hydrogen bonding capacity between a protein side chain and the nonbridging pro-(S p ) oxygen on the ␤-phosphate destabilizes the triphosphate moiety and greatly reduces ddNTP insertion.
RB69 gp43 also exhibited strong discrimination (ϳ10 5fold) against ddNTP incorporation (8), and Vent DNA polymerase discriminated against ddNTPs 270-fold (5). These B-family DNA polymerases have an asparagine at the equivalent position of the aromatic residue observed with A-family polymerases. In the RB69 gp43 structure, this residue (Asn-564) interacts weakly through a water molecule with the nonbridging oxygen on the ␤-phosphate of the incoming dNTP. These observations are consistent with the idea that the lack of a strong interaction between the polymerase and ␤-phosphate precludes efficient ddNTP insertion.
Ribonucleoside Triphosphate Discrimination-In contrast to dideoxynucleotide discrimination, there were strong similarities between polymerases from different families for ribonucleotide discrimination. In all cases, DNA polymerases insert rNTPs with very low catalytic efficiency, and many of these efficiencies fall within a narrow range (Table 3). Yet, discrimination factors varied greatly depending on the fidelity of the polymerase, suggesting that discrimination is a reflection of how efficiently a polymerase inserts dNTPs rather than how poorly it inserts rNTPs (Table 3) (30). Thus, enzymes that discriminate rNTPs poorly also exhibit low dNTP insertion effi-  Tables 1 and 2. ciency (e.g. TdT). Although the efficiency for ribonucleotide insertion appears to be independent of polymerase identity, the strategy employed by X-family polymerases used to sterically deter ribonucleotide insertion is unique. DNA polymerases from other families employ a protein side chain to sterically exclude ribonucleotides (4,8,(31)(32)(33). In contrast, this exclusion is primarily provided by the protein backbone for X-family members (Fig. 1D) (34,35).
Significantly, however, pol ␤ inserted araCTP opposite template dG only 9-fold less efficiently than dCTP, indicating that inversion of the configuration at C2Ј (Fig. 2B) has a strong impact on nucleotide binding and chemistry. Similarly, RB69 and pol ␣ (B-family) and pol (X-family) demonstrated facile incorporation of araCTP while strongly discriminating against rCTP (8,34,36). The structure and conformation of an incoming dNTP in the pol ␤ active site suggest that O2Ј of a ribonucleotide would sterically collide with the backbone of ␣-helix M. In contrast, O2Ј of arabinonucleotide could clash with a polymerase side chain (e.g. Asp-276, Figs. 1C and 2B). However, the facile insertion of araCTP observed kinetically suggests that this side chain can adjust to accommodate a hydroxyl at C2Ј.
DNA polymerase ␤ inserts the wrong sugar/correct base (i.e. rCTP) opposite template dG more efficiently than a correct sugar/wrong base (i.e. dTTP) indicating that mispair geometry is more perturbing than when a ribose occupies the incoming nucleotide binding pocket. Similarly, pol also inserted rNTPs more efficiently than misinserting dNTPs (37). Although pol ␤ was also able to misinsert dTTP opposite a template rG (Fig.  7B), thus generating an rG-dT mismatch, we did not observe misinsertion of rATP, rGTP, or rUTP opposite a single-nucle-otide gapped dG with pol ␤ under our reaction conditions. Even in the presence of manganese, a metal ion known to reduce DNA polymerase fidelity by enhancing nucleotide binding (18,38), pol ␤ did not misinsert a ribonucleotide (data not shown). However, a recent study showed that pol (X-family) was able to misinsert a ribonucleotide (34). Taken together, these results denote a hierarchy of nucleotide substrate preference for DNA polymerases ␤ and , where given a template deoxynucleotide, pol ␤ (or pol ) will insert the correct sugar/correct base most favorably, followed by the incorrect sugar/correct base, and correct sugar/incorrect base and will very poorly (or not at all) insert an incorrect sugar/incorrect base.
Influence of Ribonucleotides on Extension and Templating-Although pol ␤ efficiently inserted araCTP and discriminated against rCTP, the reciprocal effect was observed for primer extension; pol ␤ efficiently extended an rNMP-terminated primer ( Table 2 and Fig. 4B) but could not efficiently extend the araCMP-terminated primer (Fig. 5B). Interestingly, similar results were observed for pol ␣ (B-family) (36,39). Thus, O3Ј of the rNMP-primer terminus must be well positioned for further catalysis. It has been noted previously that the DNA duplex near the polymerase active site assumes an A-like conformation typically observed with duplex RNA rather than B-form (35, 40 -42). Indeed, a high resolution structure of pol ␤ indicates that the sugar pucker of the 3Ј-primer terminus is 3Ј-endo like that observed for A-form DNA or RNA (12). Because araC prefers a 2Ј-endo sugar pucker (43), these results support the idea that the sugar pucker of the primer terminus and incoming nucleotide would be expected to have a strong influences on catalytic efficiency (44). Additionally, these results provide a molecular explanation for the strong chain termination activity of the anti-leukemia agent araC.
An arginine residue of ␣-helix N, Arg-283, is known to interact with the minor groove of the templating strand. In the closed ternary substrate complex, it provides van der Waals contact with the templating base and can hydrogen bond to the sugar of the upstream template nucleotide (i.e. nucleotide opposite the primer terminus; see Fig. 1D). Alanine substitution for this residue dramatically decreases catalytic efficiency and fidelity (16,45,46). Thus, the modest loss of catalytic efficiency (Table 2 and Fig. 6B) when a ribonucleotide is positioned opposite the primer terminus may be related to an altered sugar conformation that precludes hydrogen bonding with Arg-283.
Despite being a DNA-synthesizing enzyme, pol ␤ can insert a second rNTP on a ribonucleotide-terminated primer with comparable catalytic efficiency to the first ( Table 2 and Fig. 4). It remains to be determined how long a ribonucleotide chain pol ␤ can synthesize before nucleic acid binding would be diminished. According to one study (47), pol ␤ could synthesize an 8-nucleotide-long RNA product. Similarly, Klenow fragment (exo Ϫ ) could also incorporate 4 -7 successive rNTPs before RNA synthesis was dramatically reduced (4). Using a single-nucleotide gapped substrate where the template strand was RNA and the primer and downstream oligonucleotides were DNA, correct dNTP insertion for pol ␤ was hindered by 5 orders of magnitude. 6 Thus, the duplex hybrid nucleic acid would be structurally altered, having both A-and B-like qualities (48), relative to that normally encountered by pol ␤ and thereby interfering with proper binding. Similarly, catalytic efficiency of pol was severely diminished on substrates containing either a complete RNA primer or template (27,49). Consequently, although one or two ribonucleotide(s) in the upstream primer is(are) tolerated by X-family DNA polymerases, a stretch of ribonucleotides is expected to interfere with nucleic acid binding and DNA synthesis. DNA polymerase ␤ efficiently inserted a dNTP opposite a DNA template containing a single ribonucleotide (Table 2); thus, it appears that a templating ribonucleotide does not alter polymerase fidelity. Because pol ␤ was able to efficiently extend rNMP-terminated primers and incorporate dNTPs relatively efficiently opposite a single ribonucleotide in the template position, it is apparent that the insertion step is critical for discriminating against ribonucleotide contamination of DNA. In general, DNA polymerases bind ribonucleotide triphosphates weakly and insert them slowly. Mutational and structural studies with A-, B-, RT-, and Y-family DNA polymerases have indicated that a protein side chain provides a "steric gate" to discourage binding of ribonucleotides (50). For X-family DNA polymerases, C2Ј of the incoming nucleotide interacts with the backbone of the polypeptide rather than a specific side chain (Fig. 1D). Within the X-family, this backbone interaction is contributed by Tyr-271 and Tyr-505 for pol ␤ and pol , respectively. In contrast, the equivalent residue for pol and TdT is glycine (25). Importantly, pol ␤ and pol differ from pol and TdT in that they strongly discriminate against ribonucleotide triphosphates (27,34,49). It has recently been shown that the discrimination exhibited by pol can be relaxed with an alanine substitution for Tyr-505 even more than the Y505G mutant (34). Because rNTP insertion is similar for all members of the X-family with glycine or tyrosine at the structurally equivalent position (Table 3), discrimination is much more complex than a simple polypeptide steric clash. For pol ␤, Tyr-271 also interacts with the minor groove edge of the primer terminus base suggesting that this interaction may also influence ribonucleotide discrimination.
Biological Consequences-DNA polymerase fidelity, specificity, or discrimination represents relative kinetic terms used to describe the propensity of a polymerase to bind and insert an alternative substrate (e.g. insert a wrong nucleotide leading to a base substitution error). DNA polymerase specificity may be quantified in vitro by measuring the insertion kinetics of a single nucleotide (e.g. correct or incorrect; ribonucleotide or deoxynucleotide) opposite a defined templating base. The absolute rate or probability that a DNA polymerase inserts a nucleotide follows Michaelis-Menten kinetics. A steady-state kinetic approach defines substrate specificity as catalytic efficiency, k cat /K m , for formation of a specific base pair.
DNA polymerase specificity is commonly characterized by determining the misinsertion frequency (Equation 1) (51). f ϭ V a V a ϩ V c ϭ ͑k cat ͞K m ͒ a ͓dNTP͔ a ͑k cat ͞K m ͒ a ͓dNTP͔ a ϩ ͑k cat ͞K m ͒ c ͓dNTP͔ c (Eq. 1) The misinsertion frequency is the relative rate of incorporation of an alternative (i.e. rNTP) nucleotide (v a ) to the sum of the rates of incorporation of the ribo-and deoxynucleotides. When the concentration of competing substrates is the same, then f is the ratio of the specificity constant for ribonucleotide insertion over the sum of the specificity constants for ribo-and deoxynucleotide substrates (Equation 2).
f°ϭ ͑k cat ͞K m ͒ a ͑k cat ͞K m ͒ a ϩ ͑k cat ͞K m ͒ c (Eq. 2) We denote the misinsertion frequency when competing substrates are at the same concentration as f°. Fidelity is the reciprocal of the misinsertion frequency (1/f°). In general, the specificity constants for alternative nucleotides are much lower than for the correct nucleotide ((k cat /K m ) a Ͻ Ͻ (k cat /K m ) c ), so that f°is simply the ratio of specificity constants, (k cat /K m ) a / (k cat /K m ) c , and has been referred to as the relative misinsertion efficiency (f ins ) (52). Typically, f°is reported or tabulated (Tables 1 and 2). However, Equation 1 is useful when considering substrate pool bias and when considering cellular levels of ribonucleotides, because they are present at much higher concentrations than their deoxy-counterparts (2).
Although several DNA polymerases show remarkable discrimination against ribonucleotides (Table 3), if we take the nucleotide pool imbalance into account (rCTP/dCTP ϳ100) (1), then using Equation 1, f ϳ82, suggesting pol ␤ inserts an rNTP every 81 dNTP-insertion events. Because spontaneous depurination occurs at a rate of ϳ10,000/cell/day (53), the formation of apurinic sites is expected to be much greater than this (e.g. glycosylase-initiated pol ␤-dependent BER) for proliferative and nonproliferative cells. Thus, it is expected that DNA polymerases may insert a significant number of ribonucleotides during repair and replication. In fact, a recent study reported that the yeast replicative DNA polymerases (␣, ␦, and ⑀) inserted a large number of ribonucleotides during in vitro DNA synthesis where the ribo-and deoxynucleotide pools mimicked their physiological concentrations (54). If a large number of ribonucleotides were present in the genome, there could be structural repercussions that could modify nucleic acid-protein interactions and ultimately have deleterious cellular effects.
In addition, a single ribonucleotide makes the DNA backbone susceptible to cleavage by a general base or an RNase. Alternatively, ribonucleotides may be removed from DNA through a base excision repair pathway. Two groups have proposed independent pathways that may be responsible for removing aberrant ribonucleotides. In one case, topoisomerase type I was able to cleave an RNA/DNA duplex to near completion, and the presence of a single ribonucleotide was sufficient for strand cleavage (55). Also, RNase H-type II and flap endonuclease I have been shown to recognize a single ribonucleotide in duplex DNA and make 5Ј and 3Ј incisions, respectively, producing a BER intermediate, a single-nucleotide gap (56). Flap endonuclease I has also been implicated in pol ␤-dependent long patch BER (57). Therefore, a BER-type pathway may exist to ensure that ribonucleotides do not persist in DNA.