The E295K Cancer Variant of Human Polymerase β Favors the Mismatch Conformational Pathway during Nucleotide Selection*

Background: Human DNA polymerase β is mutated in a high percentage of cancers with specific variants impacting enzymatic activity and/or fidelity. Results: In the E295K carcinoma variant, dCTP competes with cognate dTTP with dA as templating base. Conclusion: Structures verify that the E295K variant favors the mismatch over cognate dNTP. Significance: Relevant mutations may have the potential to lower the mismatch energy barrier to catalysis. DNA polymerase β (pol β) is responsible for gap filling synthesis during repair of damaged DNA as part of the base excision repair pathway. Human pol β mutations were recently identified in a high percentage (∼30%) of tumors. Characterization of specific cancer variants is particularly useful to further the understanding of the general mechanism of pol β while providing context to disease contribution. We showed that expression of the carcinoma variant E295K induces cellular transformation. The poor polymerase activity exhibited by the variant was hypothesized to be caused by the destabilization of proper active site assembly by the glutamate to lysine mutation. Here, we show that this variant exhibits an unusual preference for binding dCTP opposite a templating adenine over the cognate dTTP. Biochemical studies indicate that the noncognate competes with the cognate nucleotide for binding to the polymerase active site with the noncognate incorporation a function of higher affinity and not increased activity. In the crystal structure of the variant bound to dA:dCTP, the fingers domain closes around the mismatched base pair. Nucleotide incorporation is hindered because key residues in the polymerase active site are not properly positioned for nucleotidyl transfer. In contrast to the noncognate dCTP, neither the cognate dTTP nor its nonhydrolyzable analog induced fingers closure, as isomorphous difference Fourier maps show that the cognate nucleotides are bound to the open state of the polymerase. Comparison with published structures provides insight into the structural rearrangements within pol β that occur during the process of nucleotide discrimination.

DNA polymerase ␤ (pol ␤) is responsible for gap filling synthesis during repair of damaged DNA as part of the base excision repair pathway. Human pol ␤ mutations were recently identified in a high percentage (ϳ30%) of tumors. Characterization of specific cancer variants is particularly useful to further the understanding of the general mechanism of pol ␤ while providing context to disease contribution. We showed that expression of the carcinoma variant E295K induces cellular transformation. The poor polymerase activity exhibited by the variant was hypothesized to be caused by the destabilization of proper active site assembly by the glutamate to lysine mutation. Here, we show that this variant exhibits an unusual preference for binding dCTP opposite a templating adenine over the cognate dTTP. Biochemical studies indicate that the noncognate competes with the cognate nucleotide for binding to the polymerase active site with the noncognate incorporation a function of higher affinity and not increased activity. In the crystal structure of the variant bound to dA:dCTP, the fingers domain closes around the mismatched base pair. Nucleotide incorporation is hindered because key residues in the polymerase active site are not properly positioned for nucleotidyl transfer. In contrast to the noncognate dCTP, neither the cognate dTTP nor its nonhydrolyzable analog induced fingers closure, as isomorphous difference Fourier maps show that the cognate nucleotides are bound to the open state of the polymerase. Comparison with published structures provides insight into the structural rearrangements within pol ␤ that occur during the process of nucleotide discrimination.
DNA polymerases (pol) 3 experience significant conformational events within their multidomain architecture during each catalytic cycle of deoxynucleotide incorporation (1). These large motions are also accompanied by smaller reorganizations of key residues during establishment of the pre-catalytic state, and these changes are expected to occur in a specific order during faithful replication of DNA. Discrimination between cognate and noncognate base pairs at the polymerase active site occurs with major contributions at the level of affinity (K d ) and secondarily at the level of nucleotidyl transfer (k pol ), although the latter may include steps other than those directly related to catalysis. The induced-fit paradigm for substrate or ligand recognition by proteins has long been employed to describe the mechanisms of catalysis or modes of functional inhibition by small molecules. In the case of DNA polymerases, concerted large and small conformational events are observed upon binding dNTP. It was hypothesized from the early polymerase structures that the large domain movements accompanying dNTP binding (or during product release) could be the rate-limiting step (1)(2)(3). The development of fluorescencebased techniques to observe conformation independent of the chemical reaction suggested that structural changes occurred in at least two discernable phases after nucleotide binding and that the later phase was rate-limiting for correct nucleotide incorporation (4) with the chemical step becoming rate-limiting in the event of misincorporation. More recent studies using similar techniques demonstrate that the large conformational events occur at a rate significantly faster than that of the ratelimiting step as was shown using FRET analysis of the fingers subdomain of klentaq1 (5). For example, studies with the replicative DNA polymerases from bacteriophages RB69 (6) and T4 (7) imply that the alternative isomerization steps occurring either before or after the phosphoryl transfer reaction constitute rate-limiting events. A similar interpretation was proposed in related studies with DNA pol ␤ (8), and follow-up experi-ments suggested that the chemical step may still be rate-limiting (9,10). Although a number of different crystal structures are available for pol ␤, limited information is available to assign specific conformational states during the open to closed state transition associated with binding dNTP.
The human gap-filling DNA repair polymerase DNA pol ␤ contains a DNA duplex binding domain (residues 90 -150), a catalytic domain (151-260, and a C-terminal nucleotide binding domain (261-335) (11), analogous to the thumb, palm, and fingers domains, respectively, of the right hand replicative DNA polymerases (12)(13)(14)(15), along with an N-terminal lyase domain. Upon binding the DNA duplex, the fingers domain cycles between conformations during incorporation of dNTPs, with few published examples of a fingers domain closing on a mismatched base pair in the absence of a damaged base (16,17). Molecular dynamics simulations have shown that in pol ␤ the induced-fit model (18) is a multistep process that includes gross changes (domain movements) that are likely the measurable component described above, and subsequent minute changes (side chain conformations after or in conjunction with metal ion binding) that may be considerably slower and not observable in previous fluorescence measurements (19,20).
The E295K variant of pol ␤ was first identified in a gastric carcinoma and subsequently found by us in a colon carcinoma (21,22). Expression of the E295K variant in mouse mammary epithelial cells induces sister chromatid exchanges and cellular transformation (23). The mutation lies within the fingers domain of the enzyme proximal to the template strand near the nucleotide-binding pocket. Although E295K possesses significantly reduced polymerase activity, it binds to DNA with an affinity similar to that of wild-type (WT) pol ␤ and has normal 5Ј-deoxyribose phosphate lyase activity (23). E295K is unable to support base excision repair of hypoxanthine and is unable to restore resistance to methylmethane sulfonate in mouse embryo fibroblasts depleted of pol ␤. In fact, E295K appears to compete with WT pol ␤ for binding to single nucleotide gaps (23). Accumulation of base excision repair intermediates in cells expressing E295K is likely due to its ability to bind to single nucleotide gaps in place of WT pol ␤. Because E295K has very low polymerase activity, fewer gaps are filled, resulting in double strand breaks upon collision with the replication fork. This leads to genomic instability and cellular transformation.
Here, we structurally demonstrate that, for the E295K variant, the conformational change of the fingers domain is observed only under noncognate conditions, with a nearly closed conformation achieved for the dA:dCTP mismatch. Similarly, NMR studies of this variant previously reported no conformational event under dT:dATP cognate conditions (24). The dUMPNPP or dTTP cognate complexes reported here support those initial findings. We show that dCTP inhibits incorporation of the cognate dTTP through equivalent binding affinity but not through increased incorporation of the nucleotide into the DNA. The conformation of residues in the active site is distinct from a cognate base pair complex but shares attributes with those of other mismatches obtained with WT pol ␤, suggesting that the ground state for the E295K variant is that of a mismatch binding mode.

EXPERIMENTAL PROCEDURES
Protein Expression and Purification-Untagged wild-type (WT) human pol ␤ was subcloned into a modified pET28a vector. The E295K variant was generated using the polymerase chain reaction (PCR) from the WT plasmid followed by sitedirected mutagenesis using the QuikChange kit (Stratagene). Both WT and E295K plasmids were transformed into Escherichia coli Rosetta 2 DE3 (Novagen) and grown in Luria broth (LB). Expression was induced with 1 mM isopropyl ␤-D-1-thiogalactopyranoside for 2 h at 37°C. Cells were harvested and resuspended in 10 ml of Buffer A (50 mM HEPES, pH 7.6, 100 mM NaCl, 1 mM EDTA, 2 mM DTT), plus EDTA-free protease inhibitor mixture (Roche Applied Science) followed by five rounds of sonication for 30 s each. The soluble fraction was then separated by centrifugation at 19,800 ϫ g for 30 min at 4°C. The sample was passed through a 0.45-m filter prior to being loaded onto a Hi-Trap heparin (GE Healthcare) column. The protein preparation was separated by fast protein liquid chromatography (FPLC) with a NaCl gradient by mixing Buffer A with Buffer B (50 mM HEPES, pH 7.6, 2 M NaCl, 1 mM EDTA, 2 mM DTT). Fractions containing pol ␤ were subsequently concentrated using centrifugation (Amicon Ultra-15, Millipore) and diluted in Buffer A for further purification by FLPC using a HiTrap SP column (GE Healthcare). Protein fractions were pooled and analyzed by 10% SDS-PAGE and Coomassie staining for purity greater than 95%. The samples were concentrated using centrifugation and flash frozen in the presence of 15% glycerol for long term storage (Ϫ80°C).
Generation of DNA Substrate-The one-base gapped DNA substrate used for characterization of nucleotide incorporation by pol ␤ is similar to that described in published work (25,26). The DNA substrate was created using three separate oligodeoxynucleotides (Keck Oligo Synthesis Resource, Yale University) that were purified on a reverse phase cartridge. Two oligonucleotides (5Ј-32 P-10-mer primer and 5Ј-PO 4 -5mer downstream) were annealed to the complementary 16-mer template. Oligonucleotides were heated at 95°C for 10 min, cooled to 23°C for 60 min, held at 23°C for 70 min, and then cooled for 1 h at 4°C.
Single-turnover Kinetics-Single-turnover kinetics experiments of WT and E295K were performed at 23°C in the same buffer used for the crystallographic studies (50 mM HEPES, pH 7.6, 200 mM sodium acetate, pH 9, 14% (w/v) PEG 3350). Experiments with WT pol ␤ characterizing correct insertion of nucleotide opposite template A were carried out on a KinTek RQF-3 Rapid Chemical Quench Flow apparatus. The amount of pol ␤ used for single-turnover experiments was determined by primer extension assay where 50 nM DNA substrate, 10 mM dTTP (correct), and various concentrations of E295K were used. The optimal ratio was determined to be 50:1, protein to DNA. Therefore, 2500 nM of pol ␤ was used in the single-turnover experiments along with 50 nM single-base gapped DNA and various concentrations of dNTP at different times, typically between 60 s and 2 h. Experiments with E295K were conducted manually and required up to 2 h to observe primer extension with correct nucleotide, and up to 96 h with incorrect nucleotide. Nucleotide concentrations varied from 0 to 2,000 M with WT pol ␤ and 0 to 20,000 M for E295K. All reactions were quenched with 0.25 M EDTA, and products were separated on a polyacrylamide sequencing gel. Radioactive products were observed with Storm 860 phosphorimager and quantified by ImageQuant software. Data were fitted using Prism 6 (GraphPad Software, Inc) to the single exponential Equation 1 below, where k obs is the observed rate constant at each concentration of dNTP. Collectively, k obs is plotted against concentration of dNTP and fit to the hyperbolic Equation 2, where k pol is the maximum rate of polymerization, and K d (dNTP) represents the equilibrium dissociation constant for nucleotide.
Competition Assay with dTTP and dCTP-E295K competition assay was performed at 23°C in the same buffer used for crystallographic studies and the single-turnover kinetics assay. Both dTTP and dCTP were present in the reaction mixture with a total nucleotide concentration of 10,000 M. The experiment was conducted manually for up to 1 h and quenched with EDTA after each time point. Radioactive products were separated on a polyacrylamide sequencing gel and quantified using the method described above.
Crystallization and Data Collection-Binary complexes of pol ␤ bound to a single nucleotide gapped DNA duplex substrate were similarly prepared as described in Ref. 27 by mixing 10 mg/ml protein and 2-fold excess DNA duplex with the DNA sequence context used in the biochemical studies as follows: 5Ј-CCGA-CAGCGCATCAGC-3Ј template, 5Ј-GCTGATGCGC-3Ј primer, and 5Ј-GTCGG-3Ј downstream. The oligonucleotides used for crystallization were synthesized by Midland Certified Reagent Co. (Midland, TX), PAGE-purified, and mixed in a 1:1:1 ratio. Annealing was performed by heating to 95°C for 10 min, cooling to room temperature, and then placing on ice for 10 min prior to use. The crystallization reservoir contained 12-16% (w/v) PEG 3350 with 50 mM HEPES, pH 7.5, and 150 -300 mM sodium acetate, and crystals were grown at 18°C. Cryoprotection was achieved by increasing the final PEG 3350 concentration to 20% with addition of 14% ethylene glycol. Ternary complexes containing natural nucleotides were produced by soaking crystals for ϳ15 min at 18°C in cryoprotection solution containing 2 mM dNTP and 20 mM MgCl 2 . Crystals soaked with the nonhydrolyzable analog dUMPNPP were soaked at 2 mM nucleotide for 2 h. Crystals were flash cooled in liquid nitrogen and collected on a Rigaku RUH-3R rotating anode system utilizing a Mar345 (Marresearch GmbH) image plate. Crystallographic data were processed with HKL2000 (28).
Structure Solution and Refinement-The binary complex for E295K was solved by isomorphous replacement using a WT binary model (PDB code 3ISB (29)) with a Cross R on amplitudes of 23% with PDB 3ISB and 17% for the equivalent WT binary complex collected in-house (30). The ternary complexes of E295K with dTTP, dATP, or dUMPNPP were solved by isomorphous replacement with the E295K binary complex (Cross R 17:11% on intensities/amplitudes, 29:19%, and 32:22%, respectively). The E295K ternary complex with dCTP was solved by molecular replacement using Phaser (31) within CCP4 (32) and E295K pol ␤ devoid of nonprotein atoms as the search model, which was separated into the four structural domains (33). A preliminary model was achieved (R-factor of 42%) from the molecular replacement solution with translation function Z-score of 30.1 and included the palm domain (Z-score 11.3), lyase domain (Z-score 25.6), thumb domain (Z-score 45.3), and the first 8 bp of DNA duplex (Z-score 12.8).
All initial stages of model building were performed in the absence of the fingers domain to reduce model bias. As expected for the E295K mutation, disorder is observed for the fingers domain with weak composite omit electron density but adequate residual map prior to model building. Only the binary and dUMPNPP data sets showed unambiguous electron density for the Lys-295 side chain. Away from the active site, the template strand backbone proved difficult to refine due to flexibility at the phosphodiester linkage between positions T(Ϫ2) and T(Ϫ3), specially for the dUMPNPP structure. This position juxtaposes Tyr-296 and Lys-295 in differing orientations depending on fingers domain conformation. Models were built using Coot (34) and refined using Phenix (35) with Ͼ98% of residues in the preferred Ramachandran regions as determined by MolProbity (36). All figures were generated with PyMOL (The PyMOL Molecular Graphics System, Version 1.5.0.4 Schrödinger, LLC).

E295K Preferentially
Binds dCTP over dTTP-To characterize the biochemical properties of E295K, single-turnover kinetics experiments were performed. The data show that E295K is unusual in that it has a higher affinity for incorrect versus correct dNTP. At the level of K d(dNTP) , E295K has a 260-fold loss in discrimination compared with WT. Discrimination between correct (dTTP) and incorrect (dCTP) incorporation for E295K occurs at the level of k pol , as shown in Table 1. The rate of polymerization for E295K was determined to be 220 times slower for correct incorporation and 1,300-fold slower for incorrect incorporation compared with WT. E295K is 43 times less faithful than WT and is more likely to select incorrect over correct dNTP substrate compared with WT. A similar effect was reported by another group with E295A where a significant increase in the K d value for cognate base pairs, particularly A:T and T:A, resulted in reduced fidelity for the enzyme (37). Addition of incorrect (dCTP) dNTP to the reaction for E295K results in inhibition of the primer extension activity (Fig. 1). This finding suggests that E295K preferentially binds to incorrect dCTP over correct dTTP. Taken together, these results hinted at possible structural changes within the active site of E295K leading to the binding of the incorrect dCTP, rather than correct dTTP.
Noncognate dCTP Binds Opposite Template dA in the E295K Variant-The E295K mutation had minimal structural impact on the binary complex of the polymerase bound to DNA as illustrated by the low Cross R values and the F o Ϫ F o isomorphous difference Fourier map calculated between WT and E295K binary complexes shown in Fig. 2 (30). The weak polymerase activity of the E295K variant allowed for the soaking of DNA-bound crystals with natural rather than nonhydrolyzable nucleotides. Addition of dTTP (2.4 Å, R free ϭ 22.8%) as cognate to the template dA position resulted in low occupancy binding of the nucleotide (Fig. 3A), as shown in the F o Ϫ F o difference Fourier map (30). The residual maps justified the addition of only the triphosphate tail at 60% occupancy, with no observed metal coordination by the active site aspartates. The poor binding of the nucleotide is reflected in the high isomorphism and low Cross R between the binary (2.01 Å, R free ϭ 22.4%) and dTTP-soaked ternary complex crystals (17% on intensities) and is indicative of the absence of conformational change that would be expected in the event of productive nucleotide binding. We did, however, detect a partial disruption of the Arg-258/Asp-192 salt bridge, as indicated by higher B-factors and a dual conformation for Arg-258, even in the absence of metal coordination for dTTP. Both residues have side chain atoms with B-factors approximately twice that of the corresponding C␣. Extended soaking (2 h) with dUMPNPP (2.04 Å, R free ϭ 23.0%) resulted in an increase in nucleotide occupancy (0.6 for the base and 0.8 for the ribose and triphosphate, Fig. 3B) but likewise failed to induce closing of the fingers domain as indicated by the isomorphous difference Fourier map between the dTTP and dUMPNPP data sets (Fig. 3E). The Cross R on intensities for the dUMPNPP with the binary data were 32.6% and 23.6% when compared with dTTP. Data collection and refinement statistics are shown in Table 2.
In contrast to the dTTP complex, soaks with either dATP or dCTP produced significantly better electron density maps for each nucleotide after a 20-min soak, with dATP achieving 0.5 occupancy for the adenine base and 0.7 occupancy for the ribose and triphosphate upon refinement (Fig. 3C) and full occupancy for the dCTP. The dATP (2.28 Å, R free ϭ 22.7%) is seen in a staggered conformation with respect to the templating dA (difference Fourier map between dUMPNPP and dATP shown in Fig. 3D) and is accompanied by dual conformations of catalytic Asp-192 (omitted from Fig. 3 for clarity, shown as orange carbon coloring in Fig. 5A) resulting in the absence of metal ion A. Comparison with recent studies of the pol ␤ R283K variant (38) shows that the dATP complex structure is similar to that of the dG:dAMP(CH 2 )PP open one-metal mismatch (PDB code 4F5P), with a root mean square deviation of 0.32 Å for all C␣ resulting in identical orientation of the nucleotide. Soaking of dCTP, however, resulted in an active site conformation closer to that of a pre-catalytic complex (discussed below). Significant differences in global conformations between the nucleotide-bound forms of E295K are illustrated by the high Cross R with the binary complex (29% for the dATP and 70% for the dCTP, on intensities), such that the dCTP structure solution required phasing via molecular, rather than isomorphous, replacement. Final refinement produced full occupancy of the dCTP (2.09 Å, R free ϭ 23.1%) along with coordination of metal ion B as Mg 2ϩ and metal ion A as Na ϩ (39). The metal binding in site A is likely to be sodium due to the lack of coordination by both Asp-192 and the 3Ј-OH of the primer terminus (26,27).
. g Fidelity ϭ ͓efficiency (correct) ϩ efficiency (incorrect) ͔/efficiency (incorrect) . h x-fold, change in fidelity ϭ fidelity (WT) /fidelity (E295K) .  E295K Adopts a Unique, Nearly Closed Conformation in the dA:dCTP Complex-The transition from the binary state to a ternary pre-catalytic state involves three basic steps as follows: global repositioning of domains within the polymerase, conformational changes to specific residues in the fingers and palm domains, and shifting of the DNA duplex. The ribbon diagrams in Fig. 4 demonstrate the unique domain changes adopted by E295K in the dA:dCTP complex. There is a clear global conformational change of the fingers domain upon binding dCTP that is not observed in either the dTTP or dATP structures, which remain in a binary-like conformation (Fig. 4A). The conformation of the fingers domain more closely resembles that observed for the native closed complex (27) or that of a previously reported dC:dAMPNPP mismatch (16) with respect to the placement of ␣-helix N (Fig. 4B). The conformation of the lyase domain, however, is similar to that seen in binary complexes.
In addition to the global domain rearrangements, conformational changes to specific residues in the catalytic and fingers domains are also observed. The shift in ␣-helix N upon dCTP binding results in movement of the guanidinium group of Arg-283 into a planar 3.7 Å stacking with the purine ring for the template dA (Fig. 5A). The Arg-283 position is further stabilized by N⑀ and NH, which are both within 3.2 Å of the nonbridging phosphate oxygen of the templating dA. Tyrosine 271 rotates (2.4 Å at the OH relative to E295K binary) into a position to interact with the template dA, although Phe-272 undergoes a smaller (1.2 Å) shift as measured at C␥. Closer to the catalytic center, Arg-258 shifts away from its interaction with Asp-192, which in turn rotates to interact with the dCTP-liganded metal although the position of Asp-256 is unchanged. The rotation of Asp-192 to coordinate the catalytic metal ion along with the repositioning of Arg-258 has also occurred under extended soaking with the cognate analog dUMPNPP. Tyr-271, Arg-283, and helix N, however, remain in the binary-like state also observed for the dATP complex. Comparison of the E295K dA:dCTP with the WT cognate dNTP, dC:dAMPCPP (16), or  (30) and contoured at ϩ2.5 (green) and Ϫ2.5 (red). A, E295K ϩ dTTP (red, PDB code 4M9H) showing the isomorphous difference with E295K binary (gray, PDB code 4M9G). B, E295K ϩ dUMPNPP (salmon, PDB code 4M9J) showing the difference with binary. C, E295K ϩ dATP (orange, PDB code 4M9N) showing difference with binary. D, difference map between dATP and dUMPNPP. E, difference map between dTTP and dUMPNPP. dG:dATP (17) complexes (Fig. 5C) shows that the positions of Arg-258 and Tyr-271 are in reasonable agreement among mismatch complexes but that Phe-272 has flipped into the closed ternary-like state in the dC:dAMPCPP WT complex.
The orientation of DNA in the active site is critical in precatalytic complex assembly. The binding of dATP in a staggered open fingers orientation for the E295K results in no significant repositioning of the templating dA, the T(Ϫ1) base, or the primer terminus relative to that of the binary or dTTP structures (Fig. 5B). There is, however, a backbone shift of the template dA and T(Ϫ1) toward the minor groove for the dUMPNPP relative to the binary complex, as well as a tilt of the primer terminal ribose toward the ␣-phosphate of the incoming nucleotide (3.8 Å between 3Ј-OH and ␣-phosphate). The DNA for the E295K mismatch complex experiences both a backbone shift from T(ϩ1) to T(Ϫ1) along with a significant rotation of the templating dA toward Tyr-271 in the minor groove. The primer terminus is also rotated away from the incoming nucleotide (5.6 Å between 3Ј-OH and ␣-phosphate). This DNA orientation is distinct from that of either the WT cognate dNTP or mismatch complexes (Fig. 5D). Both WT and E295K mismatches result in improper orientation of the 3Ј-OH for catalysis, although their specific orientations differ. The template shift also differs between these types of structures where the WT mismatches reposition the template further from the active center than that of a closed complex. This is the case for both dC:dAMPCPP, which has a rotation of the templating dC such that the Watson-Crick face is flipped into the major groove due to the pyrimidine base adopting a syn conformation, and for dG:dATP mismatch, which displays two conformations of the incoming nucleotide (staggered or tilted).
Multiple Interactions Stabilize the dA:dCTP Mismatch in the E295K Active Site-An extensive hydrogen bond network stabilizes the dA:dCTP mismatch in the active site of E295K (Fig.  6). The stacking position of Arg-283 allows coordination of O2 of dCTP via a water molecule along with an interaction with N␦ of Asn-279. The rotation of dA allows for the purine N6 to interact with N3 of dCTP, whereas both the N6 of dA and N4 of dCTP coordinate a water molecule. The mismatch in E295K pol ␤ is distinct from that observed for the equivalent base pair in DNA polymerase I from Bacillus (40) as it is nonplanar and has not formed the dCTP-O2 to dA-N1 bond via tautomerization (41). Analysis of DNA parameters by the 3DNA server (42) shows the dCTP to have 1.8 Å stagger and 2.4 Å shear with 23.8 o propeller twist about the Watson-Crick plane relative to the templating dA. Within the nucleotide-binding pocket in WT pol ␤, Glu-295 partakes in a salt bridge interaction with Arg-258 upon nucleotide binding (43). The mutation of Glu-295 into a positively charged lysine eliminates the binding partner for Arg-258 in the pre-catalytic complex, requiring Arg-258 to find an alternative energy minimum. In the dA:dCTP mismatch complex, Arg-258 shares a water-mediated interaction with Asp-256 as part of a noncatalytic complex (Fig. 7). Furthermore, three water molecules interact with both the primer 3Ј-OH and the O␦ of Asp-256, which does not participate in metal coordination.

DISCUSSION
Although DNA polymerase fidelity can be simplified to a ratio of catalytic efficiencies, it should be viewed in context of the individual terms (k pol and K d ) in the equation along with the individual constants along a reaction pathway. The lower overall fidelity of pol ␤ in the context of a C:A mispair relative to A:A or G:A has been reported (44 -46) and is affected by numerous mutations involved in the conformational events prior to phosphodiester bond formation. This is particularly pertinent to the case of the human gastric carcinoma variant of pol ␤ studied show the orientations of residues undergoing conformational changes upon nucleotide binding, and C and D focus on the orientation of the DNA and incoming nucleotide. Both views were generated after least squares superposition of the catalytic domains using Coot (34). The top panels are superpositions of the binary complex of E295K (gray, PDB code 4M9G), ternary complexes of E295K with dTTP, dUMPNPP, dATP, or dCTP (red, PDB code 4M9H; salmon, PDB code 4M9J; orange, PDB code 4M9N; cyan, PDB code 4M9L), and ternary complex of WT pol ␤ (magenta; PDB code 2FMS) (27). The bottom panels compare three mismatch structures: dC:dATP (blue; PDB code 3C2L) (16) and dG:dATP (black; PDB code 4KLQ) (17), both obtained with WT pol ␤ and E295K dA:dCTP (cyan).  here. E295K exhibits infidelity as a stable intermediate along the conformational re-arrangement pathway rather than at a pre-chemistry state or enhancement of incorporation. The lower k pol in combination with the improved affinity for the mismatch suggests that the mutant has trapped the polymerase prior to the activation step favoring the nucleotide binding (47) or ajar (48) state described for replicative DNA polymerases. The conformation captured by this variant allows us to address significant questions regarding the mechanism of pol ␤.
Finger Domain Fails to Close with Cognate Nucleotides-The biochemical phenotype for the E295K variant of pol ␤ has been noted for its weak polymerase activity. The crystal structures in this study with either natural (dTTP) or nonhydrolyzable (dUMPNPP) cognate nucleotides display weak levels of binding of the nucleotide and metal with failure of the fingers domain to close down on the substrate and form the active site. These findings are in agreement with labeled methionine NMR studies that did not observe conformational events for the same variant with cognate nucleotide (24). Single-turnover kinetics revealed a 225-fold increase in the K d value between WT and E295K for cognate nucleotide. These results support the hypothesis that ground state nucleotide binding occurs for pol ␤ with the fingers domain open and that nucleotide selection occurs during closure.
Previous structures of pol ␤ have shown that in the binary complex Asp-192 and Arg-258 participate in a salt bridge (11). Upon binding of the proper incoming nucleotide, the fingers domain closes, and the phenyl ring of Phe-272 disrupts the salt bridge. Asp-192 is then free to bind a metal, and Arg-258 engages in a salt bridge with Glu-295. Both cognate complexes, and particularly dUMPNPP due to its longer soak time, reveal that the Asp-192/Arg-258 salt bridge is broken, even though the fingers domain remains open. This is most interesting as this step is generally not expected to occur prior to the conformational event of helix N of the fingers domain in WT pol ␤ according to the reaction scheme modeled from fluorescence data (9) or molecular dynamics simulations (20). Under these previous reaction schemes, the fingers domain closure succeeds dNTP binding but precedes binding of the metal ion, indicating that the conformational changes of the metal binding residues occur as a consequence of fingers closure. This shift was, however, predicted in molecular dynamic simulations with the E295K variant (49).
The comparison between the open cognate dUMPNPP and closed mismatched dCTP complexes provides insight into the mechanism of nucleotide selection by pol ␤ when the normal conformational pathways are disturbed. The dUMPNPP bound in the open active site is poised for hydrogen bonding with the templating dA, leaving Tyr-271 free to progress on the pathway. However, the failure of Arg-258 to complete its exchange from Asp-192 to Glu-295 due to the Lys substitution prevents the fingers domain closure. In contrast, dCTP does not properly bond with the templating dA making it free to interact with Tyr-271 and allowing for the alternative conformational pathway discussed below.
Conformations of Active Site Residues Arg-283 and Tyr-271-Arginine 283 is known to contribute to the fidelity of pol ␤. During catalysis, its conformation changes to opposite faces of the nascent base pair and has been proposed to modulate the A-rule (29). In studies where error rates are published in lieu of kinetic parameters, the dA:dCTP error rates were calculated as 0.3 ϫ 10 Ϫ3 for WT pol ␤ and 38 -60 ϫ 10 Ϫ3 for the Arg-283 (Ala/Lys) mutants (46). In the E295K dA:dCTP mismatch, Arg-283 is stacked between the templating dA and Lys-280 (Fig. 6), a residue proposed to stabilize templating purines (50). The Arg-283 mutation enhances error rates through both frameshift in dT stretches and higher occurrence of dCTP incorporation opposite dA (44,46). Importantly, although dCTP induces fingers domain closure for E295K, there is still poor nucleotide incorporation. This observation may speak to the dual role of Arg-283 in the context of the E295K variant where, although the arginine contributes to the binding stability of this mismatch intermediate, the orientation in which the nucleotide is held prevents final active site assembly for phosphodiester bond formation. The stacking orientation of Arg-283 above the templating dA hinders the final shift of the template toward the active site, as well as prevents the final rearrangements within helix N required to establish the pre-chemistry state.
Tyrosine 271 has been suggested to play a role in monitoring the structural state of the DNA via the minor groove of the nascent base pair (33), although mutagenesis studies do not provide a clear role for the residue in the catalytic pathway (45) as mutation to Phe or Ser reduced the fidelity for A:C but mutation to Ala did not. However, the conformation of the tyrosine is restricted in the E295K variant and becomes an infidelity determinant. The conformation of Tyr-271 contributes to a preference for binding dA in a non-Watson-Crick orientation via its interaction with N1. At a distance of 3 Å, the hydrogen bond between the cytosine N3 and the adenosine N6 has been established for pre-insertion, whereas the water coordinated between Arg-283 and the cytosine O2 is primed to allow for a water-mediated hydrogen bond between the cytosine O2 and the adenosine N1 during the insertion step. Tyr-271 has been predicted to be the last residue within helix N to assume the transition state or insertion conformation (49), which is consistent with what is observed for the dA:dCTP mismatch in E295K as well as the WT pseudo-abasic dC:dATP mismatch previously reported (16). We note that the orientation of both Arg-283 and Tyr-271 in combination with the rotation of the templating dA for the E295K mismatch is observed even in the absence of incoming nucleotide for the pol ␤ homolog pol (51), where the residues Tyr-505/Arg-517/Glu-529 are equivalent to Tyr-271/Arg-283/Glu-295. Arg-258 corresponds to Ile-492, which was suggested to contribute to the closed conformation observed for this polymerase in the absence of dNTP (51).
During DNA replication, the contribution to K d by the extra hydrogen bond of a G:C versus A:T base pair should be consistent across experiments for a given polymerase with the general expectation that the G:C base pair has a slightly lower K d . In most pol ␤ studies, a G:C base pair results in a 2-3-fold decrease in K d regardless of the incoming nucleotide being purine or pyrimidine. There is also a 1.5-3-fold decrease in K d values for pyrimidine incorporation over purine when compared under cognate conditions (37,44,45). Mutants along the reaction or conformational pathways show interesting deviations. For example, there is a 10-fold decrease for the dG:dCTP over dA:dTTP with both E295A (37) and Y271F (45) but a 2.8-fold increase for dG:dCTP for the same pairings in R283A (44). For each of these mutants, the fidelity of the polymerase for the A:C mismatch is 3-fold lower than that determined for WT. These studies verify the lower overall kinetic fidelity measured with a dA:dCTP mispair for WT and R283A (44), as well as Y271A (45). Although the rate of incorporation is slower for noncognate than with cognate nucleotide, the relative fidelity of E295K is decreased for dA:dCTP compared with WT due to a lack of difference in apparent K d values between the two nucleotides (see Table 1). The apparent favorability of the pol ␤ residue conformations without requiring the templating base to flip into the syn conformation as seen in the dC:dAMPCPP structure (16) provides structural insight into the previously observed lower fidelity of WT pol ␤ in the context of a dA:dCTP mismatch (52). In these studies, the fidelity of pol ␤ was evaluated with all possible template:nucleotide base pairs; only with dT:dGTP and dT:dCTP did pol ␤ show lower fidelity than with dA:dCTP.
Conformation of Arg-258 and Active Site Assembly-An interesting observation for the mismatch described here is the orientation of Arg-258 in relation to active site assembly. During the conformational transition from the binary to nucleotide-bound state in WT pol ␤, Arg-258 changes interacting partners from the metal coordinating residue Asp-192 to Glu-295. It has been predicted that the conformational change of Arg-258 would be one of the slowest (19), and likely one of the last, events in the pathway for active site assembly. This prediction is consistent with the observation that the mutation of Arg-258 to Ala results in a modest 1.5-2-fold increase in enzymatic activity (20,53), even with significant increases in K d and K m values for the dNTP or DNA duplex substrates. Previously published mismatched structures reported an intermediate conformation for Arg-258 (16,17), and for the E295K variant, the side chain appears to be in a position to block nucleophilic attack on the ␣-phosphate via coordination of the 3Ј-OH of the primer terminus. Although Arg-258 has shifted to this intermediate conformation, Phe-272 remains in the conformation seen in the binary complex or the dG:dATP nascent mismatch and has not flipped as it does for cognate ternary complexes of WT pol ␤ or when the mismatch mimics an abasic site (see Fig. 5). Combined observations from the dUMPNPP and dCTP complex structures suggest that binding of the nucleotide and the initial conformational change of the fingers domain provide the necessary force to break the Arg-258/Asp-192 interaction and that flipping of the phenyl group of Phe-272 hinders the reverse conformational event after active site assembly.
The high energy barrier proposed for the conformational change of Arg-258 (19,20) may be due to the alternative local minima observed in the mismatch and that it is a function of three distinct conformational events prior to active site assembly somewhere between transition states 3 and 5 (TS3 and TS5) predicted from molecular dynamics simulations (20,49,54). From these studies, domain closure and repositioning of Asp-192 constitute TS1 and TS2, respectively. It was predicted for both E295K under cognate conditions, as well as WT with a dG:dA mispair, that the Phe-272 conformational event (TS3) precedes the Arg-258 rotation (TS4), and the opposite was expected for the WT under cognate conditions with Arg-258 rotation occurring during TS3. However, the structures presented here, as well as the recently reported dG:dATP for WT in the presence of Mn 2ϩ , suggest that the mismatch conformational pathway involves an alternative Arg-258 conformation that precedes that of Phe-272. A summary of the distances between Asp-192 and Arg-258 or Arg-258 and Glu-/Lys-295 is shown in Table 3.
Implications for a pol ␤ Cancer Variant-Discrimination between cognate and noncognate dNTPs occurs via contributions from both binding and chemistry. The biochemical data for the E295K variant show that although there is a significant increase in affinity for dA:dCTP (3.7-fold change in K d over dTTP), there is still very poor incorporation efficiency due to the 130-fold decrease in k pol between the two nucleotides. Under these same conditions, WT pol ␤ shows a 22-fold decrease in k pol but a 71-fold increase in K d for the mismatch. This observation suggests that variants that affect the conformational pathway have the potential to shift the mode of selection from the level of native ground state binding to that of an alternative ground state impacting the chemical step. This can be related to the homologous pol , which lacks both the residue analogous to Arg-258 in pol ␤ and the large conformational event for the fingers domain during nucleotide binding. Fidelity studies with pol show very minor (Ͻ4-fold) differences in estimated K d values between cognate and noncognate base pairs but significant (20 -300-fold) differences in k pol (55) indicating that, like E295K, discrimination possibly occurs through an alternative ground state impacting the chemical step. In the case of the E295K variant, the resulting phenotype is a combination of failure of the induced-fit contribution for cognate (fingers open) and an induced wrong fit for the noncognate.
Concluding Remarks-We present here a cancer variant of pol ␤, E295K, with an unusual preference for a noncognate base pair. The failure to form a phosphodiester bond suggests that the residue conformations in the nucleotide binding pocket do not allow nucleotide incorporation during polymerase domain closure and dNTP sampling. We suspect that other cancerlinked mutations of pol ␤ could impact the enzyme conformation during dNTP sampling in a manner that favors progress to the chemical step with an incorrect incoming nucleotide. This could occur by lowering the energy barrier between a mismatch conformation like that described here and the active state of the enzyme.