Induced Fit in the Selection of Correct versus Incorrect Nucleotides by DNA Polymerase β

Abstract

DNA polymerase β (Pol β) repairs single-nucleotide gapped DNA (sngDNA) by enzymatic incorporation of the Watson−Crick partner nucleotide at the gapped position opposite the templating nucleotide. The process by which the matching nucleotide is incorporated into a sngDNA sequence has been relatively well-characterized, but the process of discrimination from nucleotide misincorporation remains unclear. We report here NMR spectroscopic characterization of full-length, uniformly labeled Pol β in apo, sngDNA-bound binary, and ternary complexes containing matching and mismatching nucleotide. Our data indicate that, while binding of the correct nucleotide to the binary complex induces chemical shift changes consistent with the process of enzyme closure, the ternary Pol β complex containing a mismatching nucleotide exhibits no such changes and appears to remain in an open, unstable, binary-like conformation. Our findings support an induced-fit mechanism for polymerases in which a closed ternary complex can only be achieved in the presence of matching nucleotide.

Cellular DNA is continuously damaged, at rates of up to 20 000 damaged base pairs per cell per day.¹ Base-excision repair (BER) pathways are responsible for identifying, removing, and replacing these damaged base pairs, in order to maintain the genome fidelity required for continuing cellular function and replication.² Disruption of this repair process leads to cancer.^3,4 The fidelity of short-patch BER is primarily governed by the 39 kDa enzyme DNA polymerase β (Pol β).^1,5 Pol β binds double-stranded DNA that contains a single nucleotide gap from which the damaged base has been previously removed, selects the matching deoxynucleotide triphosphate (dNTP) from the cellular pool, then catalyzes the insertion of the nucleotide into the gapped DNA sequence.⁶⁻⁸ Correct function of Pol β is critical to this repair process. Alterations to the cellular function of Pol β, either through over- or underexpression⁹ or through a variety of mutations to the wild-type enzyme,¹⁰⁻¹⁴ have been associated with, and in some cases, demonstrated to cause, cancer. As such, Pol β is a potential therapeutic target for chemotherapy.¹⁵ In addition to its biological importance, Pol β’s small size makes it an ideal model polymerase to use for studies of the mechanism of nucleotide insertion. Prior investigations into Pol β’s function have thus provided insights into the behaviors of error-prone members of the polymerase family.¹⁶

The mechanism and structure of Pol β have been extensively characterized (Figure 1; Scheme 1).^8,17 Crystallographic studies have shown that Pol β contains two domains, an 8 kDa lyase domain and a 31 kDa polymerase domain. The polymerase domain can be further divided into three subdomains, referred to sequentially as the “D” (DNA-binding), “C” (nucleotide transferase), and “N” (nucleotide selection) subdomains. (Figure 1a) The basic catalytic cycle is similarly well-defined (Scheme 1). The apo enzyme adopts an open conformation in solution in which the lyase domain exists in an extended orientation relative to the polymerase domain. This structure is highly flexible,^18,19 and NMR data suggest that the lyase domain samples a binary-like conformation even in the absence of the DNA substrate.²⁰ Upon binding of a single-nucleotide-gapped DNA (sngDNA) substrate, the lyase domain undergoes a 40 Å motion to generate a compact “open” binary structure in which the DNA substrate is bent by ∼90° at the gap position in order to facilitate nucleotide binding.^21,22

Figure 1.

Crystal structures of Pol β. (a) Crystal structure of apo Pol β, with domains, subdomains, and significant structural features labeled. The structures of the three catalytic asparate residues (D190, D192, and D256) shown in red define the active site. (b) Structural overlay of “open” binary complex (red) and “closed” matching ternary complex (blue), showing the motion of the ‘N’ subdomain in response to nucleotide binding. The two Mg²⁺ ions are shown in purple. (c) Structural elements of the ‘N’ subdomain that undergo significant structural changes in response to correct nucleotide binding and enzyme closure. The ‘N’ subdomains are colored as in panel b. (d) Shown is the “hydrophobic hinge” region that controls enzyme closure. (e) Overlay of the ‘N’ subdomains of Pol β crystal structures: An “open” binary complex (1BPX) is shown in orange, a “closed” matching ternary complex (1BPY)²² is shown in blue, together with “open” (4F5P)³³ red mismatching ternary Pol β and “closed” (3C2M)³⁴ purple mismatching ternary Pol β.

Scheme 1.

Simplified Mechanism of Pol β-Catalyzed DNA Repair Pathway, Depicting Sequential Binding of the sngDNA Substrate to apo Pol β (E_o), Initial Nucleotide Association To Form a Loosely Affiliated Complex (ED_o-N), Enzyme Closure (END_c), Nucleotide Insertion, and Sequential Release of Pyrophosphate (PP_i) and DNA Product

The process of nucleotide selection by this binary complex has been intensely studied, although details remain controversial. Initial binding of nucleotide into the active site appears to be both rapid and nonselective.²³ Association of matching nucleotides with the enzyme-bound templating base pair induces conformational changes characterized by a 10 Å change in the position of α-helix N within the ‘N’ subdomain, resulting a “closed” ternary complex in which the nucleotide is poised for insertion into the gapped substrate^21,24,25 (Figure 1b). This closure of the N-helix is either stabilized or initiated by binding of a second “catalytic” Mg²⁺ cation,^21,26−28 as well as by a cascading series of altered hydrogen bonds and side chain rotations,^26,27 located in part within a flexible loop between βsheets 6 and 7 (“loop 6/7”) (Figure 1c) as well as within the socalled “hydrophobic hinge” (Figure 1d). Both experimental^28,29 and computational^30,31 studies support the existence of a second “pre-chemistry” step involving small motions inside of the active site following enzyme closure. Pol β-catalyzed nucleotidyl transferase, in which the 3′-OH terminus of the DNA primer attacks the P_α position of the incoming Mg²⁺nucleotide triphosphate to generate a nicked DNA complex and a free pyrophosphate (PPi_i) group, occurs in the fully closed conformation.²² Catalysis depends on the participation of three crucial aspartic acid residues, D190, D192, and D256, which coordinate the two Mg²⁺ cations during the nucleotidyl transfer reaction.^24,26,32 Following nucleotide insertion into the gapped DNA substrate, the complex opens followed by the sequential release of the Mg²⁺ cations, the PPi_i, and the nicked DNA substrate.⁸ At the heart of the insertion fidelity of Pol β are the differences in this catalytic process when Pol β is confronted with a matching versus mismatching nucleotide.

Disagreements exist as to how Pol β selects the proper nucleotide and avoids nucleotide misinsertion. Experimental and computational studies have implied that multiple “checkpoints” may contribute to enzyme fidelity.^35,36 Kinetic studies have demonstrated that wild-type Pol β discriminates between matching and mismatching nucleotide both at the level of the differential transition state binding³⁷ and at the stage of chemistry (i.e., nucleotidyl transfer; k_pol).^38,39 While crystal structures of matching ternary Pol β typically adopt a “closed” conformation, crystal structures of mismatching ternary complexes with N-helices in “open”,³³ “intermediate”,⁴⁰ and “closed” conformations^34,40 have all been obtained (Figure 1e). Formation of the “closed” mismatching ternary complex requires Mn²⁺, which is known to both promote enzyme closure and decrease enzyme fidelity.⁴¹ Limited evidence suggests the crystal structure of the mismatching ternary complex may vary with the nature of the mispairing.⁴⁰ Crystal structures using postchemistry nicked DNA resembling the product of nucleotide misincorporation show an N-helix in intermediate position between open and closed structures.⁴² Solution-phase SAXS studies point toward a “semiclosed” ground state mismatching ternary conformation that is noted to be “on-pathway” to the fully closed complex.⁴³ Related stopped-flow Förster resonance energy transfer (FRET) studies suggest that the closed mismatching ternary complex is destabilized relative to the matching ternary complex,⁴⁴ possibly implying that the mismatching ternary complex exists in rapid equilibrium between open and closed conformations. Other stopped-flow FRET experiments, depending upon the labeling scheme and substrates employed during the analysis, have pointed toward either limited⁴⁵ or no enzyme closure in the mismatched ternary complex.²⁹ Several computational studies have implied that mismatching ternary complexes involve distorted binding pockets, thus emphasizing the role of postclosure, “prechemistry” stochastic rearrangements of the active site that aid nucleotide insertion.^30,31 Other computational results provide evidence for greater ground-state enzyme selectivity. One study concluded that the lowest energy state of the mismatching ternary complex remains that of an open, binary-like conformation,²⁷ and another that the transition state for nucleotide transferase with a mismatching nucleotide may involve one or more partially open conformations.⁴⁶ This is in agreement with theoretical arguments, which have postulated that any enzyme closure within the mismatching complex would reduce enzyme fidelity.^36,47 Simulations using path-sampling techniques have also suggested that correct and incorrect nucleotide incorporation occur by different pathways and involve different rate-limiting steps.^48,49

Recently, we demonstrated that the lyase domain of apo Pol β is flexible on the microecond-millisecond time scale and appears to sample a binary-like conformation on a catalytically relevant time scale.²⁰ Here we extend these experiments to study the differences between matching and mismatching ternary Pol β complexes to further illuminate the structural and dynamical properties in each enzyme. Our data indicate significant differences in structure and dynamics between matching and mismatching ternary complexes generated in the presence of Mg²⁺.

METHODS AND MATERIALS

Protein Expression and Purification.

As detailed previously,²⁰ the DNA construct of rat Pol β was transformed into BL21(DE3) cells and grown in uniformly labeled ²H,¹⁵NM9 minimal media (1 g ¹⁵NH₄Cl, 2 g glucose, 6.8 g Na₂HPO₄, 3.0 g KH₂PO₄, 0.5 g NaCl, 0.1 mM CaCl₂, 2 mM MgSO₄, and 0.1 equiv Gibco minimal essential media vitamin mix per 1 L of 99.9% ²H₂O, pH = 6.6). To obtain triple-labeled (²H,¹³C,¹⁵N) Pol β, uniformly labeled ¹³C-glucose was used in the growth media. Overexpression was induced by addition of 1 mM IPTG, and cells were allowed to express for 8−14 h at 24 °C prior to harvesting by centrifugation at 6000 rpm. Protein was purified as detailed previously.²⁰ Following purification, pure protein was buffer exchanged into NMR buffer (50 mM HEPES, 100 mM KCl, 2 mM DTT, 3 mM NaN₃, pH = 7.4 with 1:1000 Roche EDTA Free Protease Inhibitor Cocktail), concentrated to a final volume of 0.5 mL containing 7% ²H₂O. Final protein concentration was measured using a UV−vis instrument (ε = 21 100 M⁻¹ cm⁻¹) and typically ranged from 0.3 to 0.8 mM.

Preparation of ds-Gapped DNA and Ternary Complex.

DNA oligonucleotides were purchased from Integrated DNA Technologies, Inc. (IDT, Coralville, IA). The template, primer, and downstream sequences were 5′-CGACCGACGGCGCATCAGCC-3′, 5′-GGCTGATGCGC-3′, and 5′pGTCGGTCG-3′, respectively. The downstream sequence was phosphorylated at the 5′ position. This 20-base pair sequence is an extended version of the 16-mer sequence used in a prior publication.²⁰ The template, primer, and phosphorylated downstream oligonucleotides were mixed in a 1:1.1:1.1 ratio, buffered with a 10× solution of NMR buffer (vide supra), diluted with PCR-grade water, and annealed using a BioRad Peltier Thermocycler. The mixture was incubated at 95 °C for 10 min, slowly cooled to 50 °C over 30 min, held at 50 °C for 20 min, and then cooled to 4 °C before immediate transfer to ice and storage at −20 °C. The quality of the annealing was assessed by 18% native polyacrylamide gel followed by staining with ethidium bromide. The binary complex was formed by adding this 20-mer gapped ds-DNA to Pol β at a ratio of DNA:protein of ∼1.5:1, followed by immediate pH adjustment to pH = 7.4.

The formation of stable ternary complexes was performed by adding nonhydrolyzable deoxy-nucleotide triphosphate derivatives (dNpNHpp) (Scheme 2) to the preformed binary Pol β/sngDNA complex. The templating base opposite the gap is a guanine. Thus, formation of the matched ternary complex was obtained by adding 2-deoxycytidine-5′-[(α,β)-imido]triphosphate (dCpNHpp) (Jena Biosciences), whereas the mismatching ternary complex was formed by addition of 2′deoxyadenosine-5-[(α,β)-imido]triphosphate (dApNHpp) (Jena Biosciences). Initial estimates of the dNpNHpp dissociation constants (K_d) were obtained from ITC experiments (vide infra). Binding of dNpNHpp to form the ternary complex was performed by titrating aliquots from concentrated stock solutions of dNpNHpp in NMR buffer (44 mM for dCpNHpp, 39 mM for dApNHpp) containing 80 mM MgCl₂ to the Pol β binary complex and monitoring formation by following peak shifts in an ¹H−¹⁵N TROSY experiment.⁵⁰ Saturation was deemed complete when NMR resonances remained consistent upon further addition of nucleotide. NMR-derived K_D values were in good agreement with those obtained by ITC. For the matching complex dCpNHpp (stock solution 44 mM), spectra were acquired at ratios of dCpNHpp:binary complex of 0:1, 0.2:1, 0.4:1, 0.6:1, 0.8:1, 0.95:1, 1.1:1, and 2:1, whereas, for the mismatching dApNHpp (stock solution 39 mM), spectra were acquired at ratios of 0.15:1, 0.25:1, 0.4:1, 0.7:1, 1:1, 1.6:1, and 4.6:1 dApNHpp:binary complex.

Scheme 2.

Structures of Nonhydrolyzable Nucleotide Analogues 2′-Deoxycytidine-5′-[(α,β)-imido]triphosphate (dCpNHpp; top) and 2′-Deoxyadenosine-5′-[(α,β)-imido]triphosphate (dApNHpp; Bottom)

Isothermal Titration Calorimetry (ITC).

An isothermal titration calorimeter VP-ITC (Micro Cal LLC), equipped with a 200 μL tantalum sample cell, was used to determine the thermodynamics of binding of dApNHpp and dCpNHpp to the Pol β binary complex to form the matching and mismatching ternary complexes. For the ternary complex titration, both the binary complex and dNTP analogue were prepared in identical buffer (50 mM HEPES, 100 mM NaCl, 5 mM MgCl₂, and 2 mM DTT; pH 7.4). Protein concentrations in the apo form were measured at 280 nm. All ITC experiments were performed at 23 °C. For the matching ternary complex binding experiment, 0.2 mM dCpNHpp was injected into 0.013 mM Pol β binary complex. The titration experiments with dCpNHpp were carried out with a stirring speed of 1000 rpm and duration of 250 s between each 1 μL injection. The mismatched ternary complex binding experiments were carried out by titrating 6 mM dApNHpp to 0.106 mM Pol β binary complex. The experiments with dApNHpp were carried out at 23 °C stirring speed of 1000 rpm and 250 s duration between each 0.8 μL injection volume.

NMR Titration, Assignment, and Relaxation Experiments.

Previously we deposited the NMR assignments for apo Pol β. Using additional triple resonance experiments and higher static magnetic field strengths (900 MHz ¹H frequency), we have reevaluated and reassigned our published backbone amide assignments for apo Pol β such that the 82% of the amide backbone resonances are now assigned. These reassignments and new identifications particularly focus on residues in the lyase domain. These additional data do not alter the conclusions of the prior work,²⁰ and the new data has been uploaded to the BMRB under accession #18267.

The sample temperature was calibrated to 23 °C with a 100% methanol standard prior to each experiment. The mismatching ternary complex was characterized by monitoring the chemical shift changes during ligand titration. As nearly all residues in the matching ternary complex were in slow exchange, resonances in this complex were assigned through the comparison of limited triple-resonance data (HNCO, HNCA) to that acquired for the binary complex. Overall, the number of amino acid residues that could be confidently analyzed was 228 for the binary, 204 for the matching ternary, and 215 for the mismatching ternary complexes.

All NMR relaxation experiments were performed at 23 °C on an Agilent 800 MHz instrument located at Yale University and 900 MHz magnet located in the Department of Chemistry and Biochemistry at the University of Colorado. The TROSY-based CPMG relaxation dispersion experiments⁵¹ were acquired with a constant relaxation times⁵² of 20 ms with τ_cp delays of 0, 0.625, 0.714, 1, 1.25, 1.67, 2, 2.5, 3.33, 5, and 10 ms. All NMR relaxation data were processed using NMRPipe⁵³ and analyzed using Sparky.⁵⁴ Peak heights were quantified in Sparky using the average of nine points from a 3 × 3 grid centered on the peak maximum. Relaxation rates were determined from peak intensities using in-house written programs. Results are reported only for resonances that are not overlapped in the two-dimensional spectra and that have sufficient signal-to-noise such that reliable quantification of peak intensities is possible. CPMG relaxation data were analyzed with⁵⁵

$$R_2\left(1/{\tau }_{\text{cp}}\right)=R_2^0+R_{\text{ex}}\left[1-2\tanh \left(k_{\text{ex}}{\tau }_{\text{cp}}/2\right)/\left(k_{\text{ex}}{\tau }_{\text{cp}}\right)\right]$$ (1)

In eq 1 $R_{\text{ex}}=p_{\text{a}}{p}_{\text{b}}\text{Δ}{\omega }^2/k_{\text{ex}}$ where p_a/b are the equilibrium populations of conformation a and b, k_ex, the exchange rate constant, is the sum of the forward and reverse rates for interconversion of between conformation a and b and Δω is the chemical shift difference for an ¹⁵N nucleus when in conformation a and b.

Chemical shift differences between the various Pol β complexes before and after saturation with the designated ligand were quantified by

$$\text{Δ}=\sqrt{\left(\text{Δ}{\delta }_{\text{NH}}^2+\text{Δ}{\delta }_{\text{N}}^2/25\right)/2}$$ (2)

in which Δδ is the chemical shift difference between the free and ligand saturated complexes for the amide proton (NH) and the amide nitrogen (N) resonances.

NMR lineshape analysis was performed on select resonances from a titration series of the binary Pol β complex in the formation of the matching or mismatching ternary complexes. Resonances were analyzed if they remain resolved in both the ¹H and ¹⁵N dimensions during the entire titration. Fitting was performed as described previously^56,57 to a two-state “U” model using the Integrated Data Analysis Platform (IDAP) software provided by Professor Evgenii Kovrigin (Marquette University).⁵⁶

RESULTS

Characterization of Binary (sngDNA-Bound) Pol β.

As reported previously, titration of apo Pol β with a single-nucleotide-gapped DNA substrate induces substantial chemical shift changes in many residues, primarily located in the lyase domain (Figure 2a–c, Figure S1). The 10% trimmed mean composite chemical shift change (Δ; eq 2) was 0.074 ppm, with a standard deviation of 0.098 ppm. 31 residues (L19, F25, N28, Q31, I33, H34, K35, N37, E58, K60, L62, G64, I69, K72, I73, E75, F76, L77, T79, G80, L82, E86, K87, I88, R89, G105, E117, I119, K120, L259, and V303) exhibited Δ values above the 1.5 × standard deviation of the 10% trimmed mean value of 0.22 ppm (Figure 2). All of the above-mentioned residues are plotted onto the crystal structure of binary Pol β (1BPX) in Figure 2d, in order to better visualize the extent of the chemical shift changes. Formation of the binary complex also has a substantial impact upon the millisecond (ms) motions in Pol β. The millisecond motions of binary Pol β were probed using the TROSY-based ¹⁵N relaxation-compensated CPMG dispersion experiment. Of the 228 observable resonances, only a single residue, V115, exhibits dispersion at 21.1T. (Figure 2c) A nonlinear fit of the relaxation dispersion data to eq 1 yields an approximate rate of exchange of k_ex = 850 ± 430 s⁻¹ and R_ex = 10.1 ± 2.6 s⁻¹. These data are consistent with previous solution NMR studies indicating the binary Pol β is largely a stable complex lacking significant conformational exchange motions on the microsecond−millisecond time scale.²⁰

Figure 2.

Binary complex formation. (a) Overlay of HSQC spectra of apo (red) and binary (blue) Pol β complexes. (b) Composite chemical shift changes induced by binding to 20-mer single nucleotide gapped DNA. The black line represents the cutoff for large chemical shift changes exhibited upon binding (as defined in the text). (c) CPMG relaxation dispersion of V115. (d) Crystal structure of binary Pol β (1BPX).²² Subdomains are colored as in Figure 1a, and the three catalytic aspartic acids are shown as red stick representation. Residues with Δ values above the black line in Figure 2b are shown as gray spheres. The only residue to exhibit dispersion, V115, is shown in yellow.

Formation and Analysis of Matching Ternary DNA Polymerase β.

Formation of the ternary complex using a nonhydrolyzable deoxynucleotide triphosphate was monitored by ITC under conditions to mimic the subsequent NMR experiments. 2′-Deoxycytidine-5′-[(α,β)-imido]triphosphate (dCpNHpp) binds to the Pol β binary complex with a K_D = 0.7 ± 0.1 μM, ΔH = 93 ± 1 kJ/mol, and ΔS = 432 ± 5 J/mol−K (Figure S2). This value is in good agreement with K_d = 1.7 ± 0.3 μM obtained from kinetic data.⁵⁸ Titration of the matching dCpNHpp into samples of binary DNA-bound Pol β (containing a single guanidine at the gap and 10 mM MgCl₂) induced significant chemical shift changes in many of the well-resolved amide residues, as monitored by NMR spectroscopy (Figure 3a). As essentially all residues were in slow exchange between the binary and matched ternary complex, most resonances could not be assigned simply by following the peak shift during the titration. For these resonances we aided assignments of the ternary enzyme by acquiring triple resonance experiments (TROSY-based HNCO, HNCA). Overall, 204 residues were confidently assigned in the final ternary spectrum. Analysis of the chemical shift changes induced upon binding of dCpNHpp showed a 10% trimmed mean composite chemical shift change (Δ, eq 2) was 0.052 ppm, with a standard deviation of 0.14 ppm. Twenty-four residues (D170, A175, R182, D192, Q217, R256, K262, G268, V269, F272, G274, R283, F291, T292, I293, N294, E295, T297, I298, L311, V313, D314, S334, and E335) exhibited large (Δ > 1.5 × standard deviation from the trimmed mean value) chemical shift changes upon addition of matching nucleotide (Figure 3b; Figure S3). In addition to these 24 residues, six other intense, isolated amide resonances in the binary complex (C178, G179, M191, T201, C267, and K289) were in slow exchange during titration with matching nucleotide and could not be identified in the matching ternary complex. In order to better understand the impact of these titration results, residues exhibiting large chemical shift changes are plotted onto a crystal structure of the matching precatalytic complex (1BPY²²) in Figure 3c. Consistent with prior crystallographic studies showing that the majority of structural changes due to binding of matching nucleotide occur in this region,^22,26,33 the nucleotide-binding (‘N’) subdomain exhibits the largest chemical shift changes upon binding of correct nucleotide.

Figure 3.

Matching ternary complex formation. (a) Overlay of ¹H−¹⁵N TROSY spectra of binary (red) and matching ternary (blue) Pol β complexes. (b) Composite chemical shift changes induced in Pol β by binding of the matching nucleotide analogue, dCpNHpp. The black line represents the cutoff for large chemical shift changes (as defined in the text). (c) Crystal structure of closed matching ternary Pol β (1BPY). Subdomains are colored as in Figure 1a, and the three catalytic aspartic acids are shown as red stick representation. The matching nucleotide is shown in orange, while the two bound Mg²⁺ ions are in violet. Residues with Δ above the black line in Figure 3b are shown as gray spheres.

As with the binary enzyme complex, the matching ternary Pol β complex was assessed for evidence of microsecond-millisecond motions using the TROSY CPMG relaxation dispersion experiment. At a static magnetic field strength of 21.1 T, no residues in the matching ternary complex exhibited relaxation dispersion profiles that would indicate enzyme motions on this time scale. Together with the significant chemical shift changes described above, these data indicate that in solution the matching ternary Pol β adopts a distinct, stable, closed conformation.^{19,21,22,24,26,59}

NMR lineshape analysis has been shown to be invaluable for the characterization of protein folding⁶⁰ and for monitoring the kinetics and mechanisms of ligand protein interactions.^{56,57,61−64} NMR titration data for formation of the matching ternary complex were analyzed using IDAP,⁵⁶ a software program which extracts a 1D projection of lineshapes in either the ¹H or the ¹⁵N dimension in a two-dimensional spectrum and then treats the data using a variety of mechanistic models. Lineshape data for matching nucleotide binding are best fit to a two-site exchange model (Scheme 3), in which all chemical shift changes can be attributed directly to ligand binding.^56,65 Simultaneous analysis of lineshape titration data from nine residues (A38, A307, E153, G305, G308, I69, I150, S243, and V214) yields values of k_off = 5.1 s⁻¹ (95% confidence interval: 4.7−5.8 s⁻¹) and K_D = 3.8 μM (95% confidence interval: 2.6−6.9 μM) for dCpNHpp. Spectral traces and the fitted 1D projections of two of these residues (A307 and I69) are shown in Figure 4; the remaining data are found in Figures S4 and S5. These results from analysis of lineshape behavior are consistent with previous single-turnover kinetic data,³⁹ and the measured value of K_D is in reasonable agreement with the ITC data (vide supra).

Scheme 3.

Standard Two-Site Model Employed by IDAP^a

^aNucleotide binding occurs in a single step, which can be modeled using two parameters, K_D and k_off, the rate constant associated with nucleotide release.

Figure 4.

Chemical shift changes and line shape analysis of two representative residues, I69 (top) and A307 (bottom), during titration of binary Pol β with matching nucleotide dCpNHpp. Panels (a, c) show an overlay of HSQC spectra acquired at varying concentrations of nucleotide (0, 200, 300, 400, and 1000 μM dCpNHpp for red, orange, yellow, green, and blue traces, respectively). Panels (b, d) show the ¹H slices together with the global fits (solid lines). For clarity, not all titration points are shown. See Supporting Information, Figure 5 for complete data.

Ternary Mismatching Complex.

The binding of the nonhydrolyzable adenosine analogue 2′-deoxyadenosine-5′[(α,β)-imido]triphosphate (dApNHpp) to form the mismatching (A opposite G) complex in the presence of MgCl₂ was monitored by ITC measurements (Figure S2). Fitting of the calorimetric titration data resulted in a K_D = 211 ± 2.3 μM for formation of the mismatching ternary complex and is in good agreement with biochemical data for misinsertion of adenosine opposite the templating guanosine (K_d = 210 ± 50 μM).⁵⁸ The thermodynamic parameters for mismatching nucleotide binding are ΔH = 37 ± 9 kJ/mol and ΔS = 195 ± 32 J/mol−K. As expected, the mismatching nucleotide binds to the Pol β binary complex approximately 300-fold weaker than does the matching nucleotide based upon ITC data.³⁹

Identical NMR experiments were performed on the mismatching complex as was done to investigate the matching ternary complex. As with the matching complex, ¹H−¹⁵N TROSY⁵⁰ two-dimensional spectra were used to compare the differences between binary and ternary Pol β complexes (Figure 5a; Figure S6). Nearly all resolved residues were in fast to moderate exchange at 14.4 T and could easily be tracked throughout the dApNHpp titration process by following the resonance shift. Out of the 215 residues that could be monitored through the titration process, a 10% trimmed mean composite chemical shift change of 0.0245 ppm, with a standard deviation of 0.0195 ppm, is obtained (Figure 5b). This 10% trimmed mean value is half that measured upon titration with the matching complex. 33 residues (H51, S96, I150, R152, E154, I161, N164, K167, A175, V177, C178, R182, G184, A185, H197, F235, L259, K262, G268, V269, L270, G274, K280, N281, M282, R283, T292, I293, N294, E295, D314, K331, and E335) exhibited chemical shift changes that were 1.5 standard deviations above the trimmed mean value (≥0.054 ppm). These residues are plotted onto a crystal structure of an open ternary complex (4F5P³³) in Figure 5c. For better comparison of the magnitude of the chemical shift changes induced by binding of matching and mismatching ligands, Figure 6 shows a box plot containing a side-by-side comparison of the chemical shift changes induced by binding of matching and mismatching ligand as a function of subdomain.

Figure 5.

Mismatching ternary complex formation. (a) Overlay of ¹H−¹⁵N TROSY spectra of binary (red) and mismatching ternary (blue) Pol β complexes. (b) Composite chemical shift changes induced upon Pol β binding of mismatching nucleotide analog dApNHpp. The threshold for large chemical shift changes (defined in text) is not shown due to its proximity to the baseline. For purposes of comparison, the purple line represents the threshold value shown in Figure 3b obtained for the matching ternary complex. (c) Crystal structure of open mismatching ternary Pol β (4F5P). Subdomains and structural features are colored as in Figure 3c, while the mismatching nucleotide is shown in red. Residues with large Δ upon binding of mismatching nucleotide are shown as gray spheres.

Figure 6.

Box plot depicting composite chemical shift changes (Δ) upon binding of matching and mismatching nucleotide as a function of (sub)domain, for the whole protein (β), the lyase domain (LY), the “D” subdomain, “C” subdomain, and “N” subdomain (D, C, and N, respectively) for both matching and mismatching nucleotides. The bar inside the box indicates the median value of Δ, and the ends of the box mark boundaries of the upper and lower quartile of the data. The “whiskers” indicate the minimum and maximum values with the circles representing outlier data, which are statistical outliers defined as points whose value is 1.5 times greater than the interquartile value.

As noted, the majority of residues were in either fast or moderate exchange at 14.1 T, readily enabling both assignment of the final mismatching complex and lineshape analysis of titration data. Resonances for nine isolated residues (A38, D263, E335, F235, G274, I150, I293, R182, and V221) were integrated to yield 1D traces (in the case of I293, both ¹H and ¹⁵N traces were fit; for all other residues, only the ¹H trace was used) and analyzed using IDAP (Figure 7; Figures S7 and S8). Similar to the matching complex, the data for the mismatching ternary complex could be readily fit to a two-state model, yielding values of K_D = 125 μM (95% confidence window: 99−150 μM) and k_off = 280 s⁻¹ (95% confidence window: 200−480 s⁻¹). This K_D is within 2-fold of the value measured by ITC.

Figure 7.

Chemical shift changes and lineshape analysis of two representative residues, E335 (top) and I293 (bottom), during titration of binary Pol β with mismatching nucleotide (dApNHpp). (a, c) show an overlay of HSQC spectra acquired at varying concentrations of nucleotide (0, 200, 510, 765, and 3400 μM dApNHpp for red, orange, yellow, green, and blue traces, respectively). Panels (b, d) show 1-D slices through the ¹H dimension as bars with the global fits shown as solid lines. For clarity, not all titration points are shown, and some contour levels in (c) are increased in order to better observe the residue during intermediate exchange. See Figure S8 for all data.

We used ¹⁵N-CPMG relaxation dispersion experiments to test whether the mismatching ternary Pol β enzyme undergoes conformational exchange motion. Surprisingly, unlike either the binary or the matching ternary complex, 10 resolved resonances (Q8, D17, A23, F25, D226, N294, E323, R326, and four unidentified residues) as well as two overlapping sets of resonances (I88/E288 and K120/R149) in the mismatching ternary complex have elevated R₂ values at lower pulsing frequencies at 21.1 T, indicating motions on the ms time scale (Figure 8; Figure S9). Using eq 1, calculated rates of exchange of these residues vary from 500 s⁻¹ to 2700 s⁻¹. Residues Q8, D17, F25, D226 and three unidentified residues could be fit to a single k_ex value of 640 ± 150 s⁻¹ (Figure 8a), residue N294 exhibited dispersion with k_ex = 1240 ± 170 s⁻¹ (Figure 8b), and residues A23, E232, R326 and an unidentified residue all exhibited dispersion at k_ex = 2600 ± 500 s⁻¹ (Figure 8c). Mobile residues are plotted onto a binary, open crystal structure (4F5P)³³ in Figure 8d. These data indicate that the mismatching ternary complex is more flexible than the matching ternary complex.

Figure 8.

Dynamics of the incorrect ternary Pol β complex. Global fits of three clusters of dispersive residues to eq 1: (a) Q8 (brown), D17 (maroon), F25 (green), D226 (blue-green), and three unidentified residues (orange, light blue, and dark blue) exchange at k_ex = 640 ± 150 s⁻¹. (b) Residue N294 is in fast exchange at k_ex = 1240 ± 170 s⁻¹. (c) Residues A23 (blue), E232 (maroon), R326 (green), and an unidentified residue (orange) exchange at k_ex = 2600 ± 500 s⁻¹. (d) Residues exhibiting dispersion plotted onto the mismatching complex. (Red residues are fit to k_ex = 640 ± 150 s⁻¹. Orange residues are fit to k_ex = 2600 ± 500 s⁻¹. N294 (k_ex = 1240 ± 170 s⁻¹) is in gold. Overlapping residues are in dark green.

DISCUSSION

Apo Polymerase β Samples a Binary-Like Conformation.

Our original study concluded that the motions experienced by apo Pol β were such to allow sampling of a binary-like conformation even in the absence of DNA.²⁰ These additional assignments largely strengthen that original conclusion. Moreover, these assignments now more specifically locate these mobile residues in parts of the lyase domain that have been previously associated with DNA binding, including the first of two helix−hairpin−helix motifs that play roles in indiscriminate binding of single-nucleotide-gapped DNA substrates.^66,67 Additionally, the majority of residues still exhibit a linear correlation between R_ex and Δδ_N² (Figure S10), suggesting that they likely sample a binary-like conformation. Furthermore, the residues that do not exhibit this linear correlation appear to be predominantly located directly at or near the helix−hairpin-helix motif. Upon the basis of the large chemical shift changes exhibited by these residues relative to their predicted values (Figure S10b), we speculate that the measured chemical shifts in the binary system contain a substantial contribution from proximity to the DNA itself. In support of this is our observation of small but noticeable chemical shift differences between the binary complex containing a 20-mer substrate and a binary complex formed with a truncated 16-mer substrate (Figure S11).

Tentative evidence suggests that parts of the ‘D’ subdomain may also be in slow conformational exchange in apo Pol β. Inspection of the apo TROSY spectra reveals that many residues in the ‘D’ subdomain are far less intense than those found in other parts of the enzyme (Figure S12), consistent with the behavior of residues in slow or moderate exchange with one or more lower populated conformers. It seems unlikely that lack of ¹H back exchange from solvent is the cause of this reduced peak intensity. Unlike residues in other regions of Pol β with limited solvent exposure, the intensities of these residues in the ‘D’ subdomain does not increase during the weeks of data acquisition. The diminished intensity of residues in the ‘D’ subdomain has also limited efforts to conclusively assign this region of the protein.

Binary Pol β Is Static on the Microsecond and Millisecond Time Scale.

The addition of sngDNA substrate to the Pol β enzyme substantially alters the chemical shifts of residues within the lyase domain and, to some degree, the ‘D’ subdomain, while the “C’ and ‘N” subdomains in the polymerase domain are largely unaffected by DNA binding. (Figure 2d, Figure S13) These results are in good agreement with crystallographic studies that find the lyase domain but not the remaining enzyme structurally altered by sngDNA binding.^22,68 We also found that binary Pol β is largely immobile on the CPMG time scale.²⁰ These additional experiments confirm this finding.

The sole residue that exhibits dispersion in the binary complex at 21.1T, V115, is located within α-helix ‘G’, part of a helix−hairpin−helix motif in the ‘D’ subdomain associated with dsDNA binding.^66,67 Limited tumor sequencing data indicates that residues within this α-helix are the locations of a number of cancer-associated point mutations,^10,69 and one study has identified this region as a “hot spot” for mutations identified in esophageal cancers.¹⁰ However, there is no evidence for other ms motions within this motif in the binary complex, and therefore observed motion of this residue likely represents localized behavior or the dispersion profile is the result of motion of a nearby residue. There is no direct evidence linking motion of V115 with function in Pol β.

Matching Ternary Complex.

The magnitude and locations of the chemical shift changes in Pol β induced by binding to the nonhydrolyzable, matching dNTP confirm that the correct ternary complex adopts a “closed” conformation similar to that characterized by numerous crystallography studies.^8,24,33 The largest chemical shift changes (17 of 24 residues) are observed within the ‘N’ subdomain (Figure 3b, Figure 6, and Figure 9), which is significantly repositioned upon enzyme closure between the binary and ternary complexes (Figure 3). Additionally, many residues with large chemical shift changes upon formation of the matching ternary complex are specifically located in parts of the “C’ and ‘N” subdomain that are known to assist in enzyme closure and nucleotide positioning. Intriguingly, despite the conformational changes that appear crystallographically, there is only limited correlation between the residues that exhibit large chemical shift changes and those that exhibit large structural changes as measured by comparison of RMSD values between X-ray structures (Figure S14). This observation does not, however, preclude the possibility of large conformational changes, as many proteins that undergo significant structural perturbations nevertheless exhibit the largest chemical shift changes in residues adjacent and close to the active site.⁷⁰ This is expected to be particularly true in the case of Pol β, as ¹H secondary shifts in amide residues are significantly affected by hydrogen bonding interactions,⁷⁰ and both the formation and the stabilization of the closed ternary complex involves a cascading series of altered hydrogen bonds.^26,27 Alterations to the hydrogen bonding structures in either the backbone amides or the side chains upon nucleotide binding and enzyme closure have been studied in four of the five residues exhibiting the largest composite chemical shift changes, including N294 (Δ = 0.76 ppm),⁷¹ E295 (Δ = 0.64 ppm),⁷² D192 (Δ = 0.60 ppm),^26,73 and T292 (Δ = 0.46 ppm).²⁷ Nevertheless, the observation of considerable chemical shift changes in residues such as G268 (Δ = 0.5 ppm) and S334 (Δ = 0.4 ppm), which are located 8−10 Å from the bound ligand in most crystal structures points toward additional residues undergoing conformationally induced chemical shift changes.

Figure 9.

Locations of large chemical shifts for the matching Pol β complex. (a) Closed ternary crystal structure with residues exhibiting large chemical shift changes labeled. (b) Residues exhibiting large chemical shift changes (purple) located around the active site in the closed matching ternary crystal structure (1BPY, colored as in Figure 1a). For comparison, the open binary structure (1BPX) is shown in gray.

The largest chemical shift changes induced upon binding of matching nucleotide are located in a small loop containing residues N294, E295, Y296, and T297, which is flanked by β-sheets 6 and 7 (F291−I293 and I298−P300) (Figure 1a). With the exception of unassigned residue Y296 and residue R299, all residues in this region exhibit large chemical shift changes, with values of Δ as great as 0.8 ppm. Residues within this region are responsible for stabilizing both the templating DNA and the enzyme in the closed ternary conformation, and many point mutations in this region exhibit decreased catalytic activity.^{71,72,74−77} N294 is adjacent to α-helix N and is likely to change conformation during nucleotide binding and insertion,⁷¹ as do residues T292 and Y296 in the closed ground state.^24,27 Residues E295 and Y296 also form a salt bridge with R258 in the fully closed enzyme complex. The cancer-causing mutation E295K is incapable of forming this salt bridge and is unable to close even in the presence of matching nucleotide.^72,74,78,79 The signi^ficant chemical shifts exhibited by other residues in this region, such as F291, I293, and I298, suggest that they too may participate in or are significantly affected by the process of enzyme closure, but the functional roles of these residues have not been explored.

Numerous other residues that exhibit significant chemical shift changes are also clustered around or near the binding pocket (Figure 9b). Residues D192 and F272 both play crucial roles in the process of ligand binding and subsequent enzyme closure,^{24,26,73,80,81} while R258 likely participates in substrate binding.³² R283 appears to align with the templating strand of DNA, and mutations of this residue have resulted in both reductions in fidelity and activity,^38,82 likely due to its role in enzyme closure.³³ Residue R182 is both involved in template binding and in a triad interaction that controls packing within the hydrophobic hinge region.^83,84 However, as highlighted above, chemical shift changes in residues distant from the active site such as A175 (which is likely reporting on changes experienced by adjacent residue I174, which has in turn been linked to the hydrophobic hinge responsible for directing enzyme closure⁸⁵) can only be attributed to binding-induced conformational changes in the ground state structure of Pol β.

The chemical shift changes outlined above, coupled with the absence of motions measurable on the millisecond time scale, support a model in which the ground state conformation adopted by the correct precatalytic ternary complex is that of a closed and stable conformation. The atomic-level resolution of these NMR experiments allows us to identify the extent of the conformational changes that have occurred at the ground state of the enzyme. The ground state structure of the correct ternary complex is one in which key residues associated with enzyme closure such as F272^73,80 and D192^48,73 exhibit significant chemical shift changes, and residues such as E295 believed to stabilize the closed conformation only upon full enzyme closure⁷² have also been repositioned. Residues distant from the active site, such as those in the V303 loop employed during FRET studies,²⁹ have also been repositioned (vide infra). While the NMR data demonstrate that the large structural changes associated with enzyme closure have occurred, the exact degree to which more subtle motions hypothesized to be associated with either enzyme closure or the rate-limiting step, such as rotation of R258,²⁶ have occurred remains unclear.

Over the past two decades, extensive FRET-based studies analyzing the kinetics of ternary Pol β formation and turnover have been performed.^{16,21,28,43−45,59,86} Our spectra, together with our titration data, provide an intriguing corollary to these studies. In combination with a series of SAXS studies acquired as a complement to these FRET studies, ¹H−¹⁵N HSQC spectra of both matching and mismatching ternary Pol β structures (see S4 of Tang et al. (2008)⁴³) were obtained. While a residue-by-residue comparison of these spectra to our data is outside the scope of this paper, close inspection of resolved residues such as those in the V303 loop (vide infra) indicates that the ground state of the matching complex analyzed by Tang et al. exhibits the characteristic chemical shift changes we have associated with enzyme closure. Other FRET experiments measuring the rate of “reverse closure”, i.e., the rate of enzyme opening from the matching ternary complex, yielded a rate constant of 6.32 ± 0.02 s⁻¹ using an extendable gapped primer and a matching nucleotide analogue at 37 °C.⁴³ This value is comparable to our measured rate constant of k_off = 5.4 s⁻¹ at 23 °C, implying both that our studies measure the same process as that monitored in FRET studies as well as that substrate dissociation and enzyme opening are coupled processes for the matching ternary complex. These values invite comparison to values of k_pol that have been measured experimentally during single-turnover experiments. At room temperature, values of pre-steady-state k_pol have been measured at 10−15 s⁻¹,³⁹ 2− 3-fold greater than the k_off values measured during our experiments. This comparison implies that, while enzyme closure may not be the slowest precatalytic step during the process of matching nucleotide insertion, nucleotide insertion is faster than enzyme opening. Intriguingly, a recent study has posited that, under conditions in which enzyme opening is slower than nucleotide insertion, a prechemistry conformational change may contribute to or even determine enzyme selectivity even if it is not rate-determining.⁸⁷

Mismatching Ternary Pol β Complex.

Our NMR data indicate that the ground state structure of incorrect ternary Pol β is neither a misshapen closed conformer⁴⁹ nor a semiclosed ternary complex^43,46,86 but a flexible, binary-like complex containing a weakly bound incorrect nucleotide triphosphate. Furthermore, these data indicate that the limited conformational changes in Pol β induced by binding of mismatching nucleotide are, broadly speaking, not located along the closing pathway adopted by the correct enzyme but are unique to the mismatching ternary complex.

Several lines of evidence support this reasoning. Figure 6 indicates that chemical shift changes induced upon binding of the mismatching nucleotide are overwhelmingly smaller than those induced by binding to the matching ligand. The magnitude of these differences is striking: the maximum composite chemical shift change (Δ) induced upon binding of mismatching nucleotide (0.17 ppm) is 4-fold less than that induced upon binding of matching nucleotide (0.82 ppm). Furthermore, while the largest chemical shift changes due to binding of the matching nucleotide are primarily clustered in the ‘N’ subdomain, reflective of enzyme closure, binding of mismatching nucleotide modestly perturbs both the “C” and “N” subdomains to similar−and small−degrees. In addition, with the exception of residues S96 and K280, the magnitude of the individual chemical shift changes induced upon binding of mismatching nucleotide are roughly equal to or smaller than those induced by the binding of matching nucleotide. Residues such as V313, S334, and E335 that undergo large chemical shifts (Δ = 0.23−0.41 ppm) upon formation of the matching complex exhibit only minimal shifts (Δ = 0.02−0.07 ppm) upon formation of the mismatching complex.

Despite the minimal conformational changes induced in the ground state of ternary Pol β upon addition of mismatching nucleotide, titration data indicate that this process is nevertheless representative of nucleotide binding. As described above, the value of K_D measured by NMR is 125 μM, which is within a 2-fold range of values of K_D measured both in this work by ITC (211 ± 2 μM) and in prior studies using biochemical methods (210 ± 50 μM).⁵⁸ In addition, the residues that undergo comparatively large chemical shift perturbations upon binding of mismatching nucleotide are primarily concentrated in regions clustered around the binding pocket (Figure 5c). Both of these behaviors indicate that nucleotide docking−and, presumably, some degree of interaction between the mismatching nucleotide and the templating base pair−occurs within the ligand binding pocket, although they cannot indicate the exact nature of these interactions.

The directions of chemical shift changes during titrations can serve as “fingerprints” indicating whether related ligands induce similar structural perturbations.^64,70,88 Formation of the mismatching ternary complex from the binary Pol β complex as followed by NMR titration reveals that the resonances move in linear fashion until saturation with mismatching dNTP is achieved (Figure 7). This linear shift is characteristic of a two-site process where the observed shift represents a population-weighted average of the dNTP unbound and bound conformations. Thus, if titration of binary Pol β with mismatching nucleotide induced chemical shift changes in the direction of the matching ternary chemical shifts, the simplest conclusion would be that the mismatching complex is partially closed but on-pathway to the fully closed state. This is not what is observed in the NMR titration studies for formation of the ternary complexes. Key amino acid residues that respond to both matching and mismatching nucleotide, such as those in loop 6/7, not only exhibit far smaller chemical shift changes but also shift in entirely different directions (Figure 10a), indicating that the ground state conformation of the mismatching ternary Pol β complex around the nucleotide binding pocket is not on-path to the closed “matching-like” structure. Similarly, residues employed as “indicators” of enzyme closure, such as those in the V303 loop, which responds to but does not significantly participate in enzyme closure,²⁹ shift in the presence of a matching nucleotide but exhibit insignificant shifts in response to binding of mismatching nucleotide (Figure 10a, inset). Residues further from the active site behave similarly (Figure 10b). Of the residues that do shift in similar directions upon binding of matching and mismatching nucleotide, the relative ratios of the magnitude of chemical shift changes induced by nucleotide binding are unpredictable, indicating that this correlation does not reflect the global adoption of a partially closed enzyme structure. These results imply that incorrect nucleotide is unable to induce enzyme closure and that the conformational changes caused by binding of the incorrect nucleotide are not on the pathway of enzyme closure.

Figure 10.

Overlay of binary (red), saturated matching ternary (blue), and saturated mismatching ternary (green) ¹H−¹⁵N TROSY spectra. Blue and green arrows indicate chemical shift changes induced by binding of matching and mismatching nucleotides, respectively. (a) Chemical shift changes induced by binding of matching nucleotide to residue T292. The much smaller chemical shift changes induced by binding of mismatching nucleotide are shown for comparison. Inset: Overlay of select titration spectra during saturation with mismatching nucleotide (0 mM red; 0.3 mM orange; 1.2 mM yellow; 3.4 mM green) demonstrating the absence of enzyme closure as monitored by chemical shift changes in G308, located within the V303 loop.²⁹ (b) Similar data for G64 and D263 as shown in (a).

Relaxation dispersion experiments demonstrate that, unlike the matching nucleotide, binding of mismatching nucleotide results in an increase in enzyme flexibility on the microsecond-millisecond time scale (Figure 7). Nonlinear fitting of the relaxation dispersion data suggests the flexible residues cluster into three groups with exchange rate constants of k_ex = 650, 1250, and 2600 s⁻¹. However, the locations of these flexible residues suggest that these global fits may be largely a coincidence (Figure 7d). Of the six residues exhibiting dispersion at k_ex = 640 s⁻¹, three are close in space (Q8, D17, and F25), on α-helix A, which has been hypothesized to influence the packing of the lyase domain.^89,90 The other identified residue that exhibits dispersion at 640 s⁻¹, D226, is distant from Q8, D17, and F25. The residues that move at the faster rate of exchange (k_ex = 2600 s⁻¹) are similarly unevenly distributed throughout the protein. Residue N294, which alone exhibits an exchange of k_ex = 1240 s⁻¹, is located in the template binding region. None of these motions are reflective of the rate of nucleotide binding/dissociation within the binding cavity, which, in the presence of 3.0 mM dApNHpp, we estimate to yield a k_ex of ∼7000 s⁻¹, outside of the exchange regime measurable by CPMG. Nor do these motions reflect conformational sampling of either a matching ternary or binary-like complex, as values of R_ex measured during fitting do not linearly correlate with calculated values of Δδ_N (Figure S15) as would be expected were this the case. Instead, these data suggest that, in addition to the mismatched ternary complex being in an open conformation, this structure is also quite flexible, unlike the stable matching Pol β ternary complex. NMR line shape data similarly indicate that binding of mismatching nucleotide is transient. The value measured for k_off, approximately 250 s⁻¹, is approximately 50-fold greater than that measured for the matching nucleotide. This again points toward the general instability of the mismatching ternary complex.

The studies presented here are consistent with recent stopped-flow FRET experiments that provide evidence for Pol β closure upon binding of correct nucleotide but do not show any evidence for enzyme closure when Pol β binds to the incorrect dNTP.²⁹ Similar to our NMR experiments, the FRET experiments used double-stranded DNA substrate with a templating guanine, contained Mg²⁺ as the divalent cation, and monitored the repositioning of a residue (V303) distant from the active site. Our data are also consistent with computational studies suggesting the mismatched ternary complex will not form a closed complex²⁷ and that nucleotide misincorporation may proceed through a partially open conformation.⁴⁶

In contrast, our data suggest a different conclusion than that derived in some other experimental and computational studies. FRET-based studies either employing a 2-aminopurine template or exploiting the natural fluorescence of the single tryptophan residue (W325) have found that the binding of incorrect nucleotide elicits a minor but measurable response.^21,44,45,86 This process—analogous to other FRET studies of binding matching nucleotide^{21,28,29,45,59,91,92} has been interpreted as representative of enzyme closure. It has been cautioned, however, that direct interpretation of FRET data may be difficult.⁹³ Complementary studies, performed in 2008, characterized these same species using SAXS, and concluded that mismatching ternary Pol β adopts a ground state that is approximately 25% closed.⁴³ However, inspection of the NMR data from 2008 indicate that the ground-state structure adopted by the mismatching complex characterized by SAXS exhibits the same pattern of chemical shifts as we have detailed above for the mismatching complexes in this study. If this is the case, then the FRET signals that have been interpreted as reflective of partial enzyme closure may in fact correspond only to the minor, off-pathway conformational changes that we observe. We speculate that the limited intensity of the spectral changes induced by mismatching nucleotide binding may reflect alterations in enzymatic distances that result in only limited FRET signals, as was suggested in 2002.⁴⁵

By far the strongest experimental evidence for a “closed” mismatching ternary complex has come from X-ray crystallographic studies. As such, these results should be explored in depth. With few, highly artificial, exceptions,³³ all crystal structures of ternary wild-type Pol β containing a matching nucleotide and a catalytically active metal adopt a “closed” conformation, as defined by the position of α-helix ‘N’ (Figure 1e).^22,94,95 Similar crystals of “closed” mismatching ternary complexes have also been obtained, with backbone root mean squared deviation (RMSD) from the matching ternary complex of as little as 0.8 Å.^33,34,40 However, to obtain these structures, it has been necessary to replace Mg²⁺ by Mn²⁺ to stabilize the mismatching ternary complex. Manganese is known to both enhance the rate of mismatch incorporation³⁴ as well as to favor enzyme closure. Thus, these structures may not be representative of the physiological mismatched structure that binds Mg²⁺ and exhibits high nucleotide incorporation fidelity. This distinction has prompted one such study to identify these “closed” or “semiclosed” conformations as representative not of the ground state of the mismatching ternary complex but of the prechemistry transition state presumed to be required for insertion of mismatching nucleotide.⁴⁰ By contrast, crystal structures of mismatching nucleotide formed in the presence of Mg²⁺ have adopted conformations that vary from “open” to “semiopen” (as defined by the position of α-helix N), depending upon the nature of the mispair.^33,40 These structures are in closer agreement with the behaviors observed in this paper. Interestingly, as indicated above, both kinetic characterization of the parameters associated with ligand binding in the solution state and our reported relaxation dispersion data may point toward one explanation for the absence of truly “open”, structurally unperturbed mismatching ternary complexes: The ground state of the mismatching ternary complex appears to be both dynamic and unstable, which is likely reason crystallographic studies of this Mg-bound complex has proven difficult.

If the ground state of the mismatching ternary complex remains in an “open” conformation, is it possible that the rate-limiting step of insertion of mismatching nucleotide involves partial or complete enzyme closure? Surprisingly, prior studies can be more readily reconciled if it does not. Experimentally, it appears that nucleotide-induced conformational changes are not rate-limiting in the insertion of either matching or mismatching nucleotide.⁹² Efforts to reconcile these findings with a model in which conformational changes represent the, or one of the, fidelity-determining checkpoints during nucleotide insertion have been controversial,^36,87 with at least one compromise proposal suggesting that difficulties associated with enzyme closure are merely reflective of the truly fidelity-determining energetics of the transition state.³⁷ If the dominant—though not necessarily the only—pathway of nucleotide misincorporation involves an unfavorable “open” transition state, this would restore the process of formation of the ground state “closed” ternary conformation as a checkpoint for enzyme fidelity.⁴⁷

CONCLUSION

We have presented a solution NMR based study of the ground-state structures adopted throughout the precatalytic DNA polymerase β pathway. Spectroscopic characterization of these complexes indicates that, while binding of correct nucleotide induces chemical shift changes consistent with enzyme closure, binding of incorrect nucleotide does not significantly perturb the “open” conformation adopted by the binary enzyme but does serve to enhance the flexibility of the overall structure. Furthermore, the measured kinetics of enzyme closure, together with prior experimental results, strongly imply that mismatching nucleotide incorporation by Pol β may proceed through an open rather than a closed conformation.

ABBREVIATIONS

BER

base excision repair

Pol β

DNA polymerase β

TROSY

transverse relaxation optimized spectroscopy

NMR

nuclear magnetic resonance

HSQC

heteronuclear single quantum coherence

CPMG

Car-Purcell Meiboom-Gill

dNTP

nucleotide

FRET

Forster resonance energy transfer

DTT

dCpNHpp (2′-deoxycytidine-5′-[(α,β)-imido]triphosphate

dApNHpp

2′-deoxyadenosine-5′-[(α,β)-imido]triphosphate

sngDNA

single nucleotide gapped DNA