Search tips
Search criteria 


Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
Cell. Author manuscript; available in PMC 2014 July 3.
Published in final edited form as:
PMCID: PMC3924593

Observing a DNA Polymerase Choose Right from Wrong


DNA polymerase (pol) β is a model polymerase involved in gap-filling DNA synthesis utilizing two metals to facilitate nucleotidyl transfer. Previous structural studies have trapped catalytic intermediates by utilizing substrate analogues (dideoxy-terminated primer or non-hydrolysable incoming nucleotide). To identify additional intermediates during catalysis, we now employ natural substrates (correct and incorrect nucleotides) and follow product formation in real time with fifteen different crystal structures. We are able to observe molecular adjustments at the active site that hasten correct nucleotide insertion and deter incorrect insertion not appreciated previously. A third metal binding site is transiently formed during correct, but not incorrect, nucleotide insertion. Additionally, long incubations indicate that pyrophosphate more easily dissociates after incorrect, compared to correct, nucleotide insertion. This appears to be coupled to subdomain repositioning that is required for catalytic activation/deactivation. The structures provide insights into a fundamental chemical reaction that impacts polymerase fidelity and genome stability.


DNA polymerases faithfully copy template DNA during replication and repair. Although DNA synthesis errors are rare events, they can have dramatic biological consequences that are central to the etiology of diseases and aging. It is believed that central to DNA polymerase fidelity is the induced-fit mechanism, where good and bad substrates align or distort catalytic groups respectively (Johnson, 2008). Key structural checkpoints utilized by polymerases to select the right nucleotide from a pool of similar substrates has remained elusive due to the difficulty in correlating kinetic observations with crystallographic structures. This is due to the lack of structural intermediates during phosphodiester bond formation.

DNA polymerase (pol) β is the simplest eukaryotic DNA polymerase and serves as a model for kinetic and structural studies of DNA synthesis and substrate discrimination (i.e., fidelity) (Beard and Wilson, 2006). Pol β provides two enzymatic activities during repair of simple base lesions; DNA synthesis and deoxyribose phosphate lyase. These activities reside on carboxyl- and amino-terminal domains, respectively. Crystallographic structures of substrate and product complexes with correct and incorrect nucleotides in various liganded and conformational states have provided a framework to interpret kinetic properties. The polymerase domain of pol β is composed of three subdomains; catalytic (C), DNA binding (D), and nucleotide selection (N) subdomains (Beard et al., 2002). Structural studies of pol β bound to gapped DNA show the active site readily accessible with the N-subdomain in an open conformation (Sawaya et al., 1997). Upon binding a nucleoside triphosphate, the N-subdomain repositions itself with α-helix N sandwiching the nascent base pair between the duplex DNA (i.e., primer terminus base pair) and protein, forming a pre-catalytic closed complex. After chemistry, the resulting complex indicates that the polymerase has reopened (N-subdomain relocated) and released pyrophosphate (PPi).

From crystallographic structures of substrates bound to the Escherichia coli DNA polymerase I proofreading active site, a two-metal mechanism for the nucleotidyl transferase reaction in DNA synthesis was proposed (Beese and Steitz, 1991). The catalytic metal (metal A) lowers the pKa of the primer terminus O3′ that subsequently attacks the α-phosphate of the incoming nucleoside triphosphate (dNTP) resulting in dNMP insertion and formation of PPi. A second metal (metal B) coordinates non-bridging oxygens on the phosphates of the incoming dNTP facilitating binding, thus neutralizing the developing negative charge during chemistry. A similar mechanism has been proposed for pol β in which two metal ions are coordinated by the primer terminus, nucleoside triphosphate, and active site aspartate residues (D190, D192, and D256) (Figure 1A) (Batra et al., 2006). The catalytic metal facilitates deprotonation of O3′ (to D256) with O3′ attacking Pα of the incoming dNTP. This attack results in phosphodiester bond formation between O3′ and Pα and bond breakage between Pα and Pβ. Protonation of the PPi leaving group is also believed to occur in the transition state of the nucleotidyl-transfer reaction. This mechanism is consistent with structural (Batra et al., 2006) and computational studies (Lin et al., 2006). However, since substrate binding induces key protein/substrate conformational changes, it remains to be determined how alternate substrates (right and wrong) affect the geometry and electrostatics of the polymerase active site. This problem is especially acute regarding intermediate structures during the reaction path of DNA synthesis.

Figure 1
The Pol β Ground State Pre-catalytic Ternary Complex

Previous structural approaches to characterize substrate polymerase complexes have trapped the reaction by using alternate ligands that do not undergo chemistry or employing a divalent metal that does not support catalysis. Recently, Yang and co-workers (Nakamura et al., 2012) have crystallized a translesion synthesis Y-family DNA polymerase with natural substrates in the presence of Ca2+ that permits complex formation, but not catalysis. Soaking these crystals in a solution of Mg2+ initiates ion-exchange leading to insertion of a correct nucleotide. Using a similar approach, we have captured molecular snapshots of pol β inserting a correct or incorrect nucleotide that provides insights into structural strategies that hasten and deter insertion that have not been observed previously. Importantly, since pol β exhibits subdomain motions in response to substrate binding, the structures capture critical events that impact subdomain positioning and the role of these motions in product release, pyrophosphorolysis, and DNA fidelity.


Structural studies have shown that pol β binds 1-nucleotide (nt) gapped DNA to form an open binary complex, and that subsequent nucleotide binding and its associated metal results in repositioning of the N-subdomain closing upon the nascent base pair (Sawaya et al., 1997). Following ternary complex formation, the catalytic metal diffuses into the active site to form the catalytically competent complex coordinating the primer terminus (O3′) and Pα of the incoming nucleotide (Batra et al., 2006; Freudenthal et al., 2012). Those studies required substrate analogues that prevented catalysis thereby limiting observations to pre- and post-chemical events. Thus, well-populated structural intermediates that occur during phosphodiester bond formation have not been described. To overcome this problem, we utilized time resolved x-ray crystallography to capture intermediates during bond formation and visualize conformational events occurring in the subsequent product complex after correct or incorrect nucleotide insertion.

Correct nucleotide insertion

To follow the formation and decay of catalytic intermediates by crystallography, it is essential all molecules start from the same point. To ensure all pol β molecules were in a closed pre-catalytic ground state conformation, we soaked open binary complex crystals of pol β bound to 1-nt gapped DNA in a cryosolution containing the crystallization solution, ethylene glycol, dCTP, and CaCl2. Calcium does not support DNA synthesis. The resulting structure of the pol β ternary complex diffracted to 1.92 Å (Table S1) and indicated the pre-catalytic ground state conformation was closed. Figure 1B shows the active site of pol β with a Fo-Fc omit map for the incoming dCTP and Ca2+ ions. The map shows clear density for the incoming dCTP, Ca2+ ions, and the active site geometry is similar to that expected of the ground state (Batra et al., 2006). All the molecules in the crystal have closed around the nascent base pair with α-helix N forming one side of the binding pocket (Figure 1B). Figure 1C shows key distances for the pre-catalytic complex with the coordination states indicative of Ca2+ (i.e., slightly longer than those observed with Mg2+). A structural overlay of the pre-catalytic pol β ground states obtained with either a natural dCTP (+Ca2+) or a non-hydrolysable dUMPNPP (+Mg2+) is shown in Figure 1D. These structures are very similar (RMSD of 0.8 over 327 Cα) with only moderate differences localized to active site aspartate residues that coordinate the divalent cations. Critically, the incoming nucleotide and primer O3′ are in identical positions in both ground states. These observations indicate that Ca+2 is a good surrogate to prevent catalysis and establish the pre-catalytic ground state.

To initiate the reaction, closed pre-catalytic complex crystals were transferred to an identical cryosolution containing 200 mM MgCl2, but lacking CaCl2 and dCTP. High concentrations of MgCl2 were required to effectively exchange ions in the metal binding sites. Using this approach, we are able to follow phosphodiester bond formation by varying the soak time in MgCl2 prior to flash freezing at 100K. We determined crystal structures of the pol β ternary complex at 10 and 20 s to 1.97 and 1.85 Å, respectively (Table S1). These structures are similar except for varying amounts of reactant and product. Figure 2A shows a Fo-Fc omit map for the primer terminus, incoming dCTP, coordinating water molecules, and Mg2+ ions from the 20 s structure. Density corresponding for both the reactant and product states can be observed. The reactant state exhibits density between Pα and Pβ of the incoming nucleotide, while phosphodiester bond formation between the primer O3′ and Pα of the incoming nucleotide is visible for the product state. Occupancy refinement of the structure after a 20 s soak indicates the reaction is 50% complete with density for both the product and reactant state being modeled into the 2Fo-FC accurately, in comparison modeling of only the reactant or product state indicated a poor fit (Figure 2B and S1). The Mg2+ ions, water molecules, and active site aspartates remain in the ground state conformation during the reaction based on a structural overlay with the pre-catalytic complex (Figure 2C). The only significant movement occurs at O3′ of the primer terminus and Pα of the incoming nucleotide. This movement results from deprotonation of the primer O3′ and the subsequent nucleophilic attack on Pα of the incoming nucleotide (see Discussion). We are able to model an intermediate structure expected to resemble a transition state at the 20 s time point with no positive or negative density for the resulting Fo-Fc omit map (Figure 2D). While this approach cannot capture a transition state, the resulting average of the product and reactant states exhibits a distance between O3′ and Pα (2.2 Å) consistent with previous computational studies that calculated a probable transition state (Lin et al., 2006). While the structure at the 10 s time point was similar to that at 20 s, only 30% product was observed at the shorter time point precluding modeling of a good transition state structure (Figure S2 and Table S1).

Figure 2
Observing Pol β Phosphodiester Bond Formation after a 20 s Soak in MgCl2

Correct insertion product complex with a transient metal

DNA synthesis on the gapped substrate results in nicked DNA. Previous “product” structures of pol β determined by generating binary crystals of pol β with nicked DNA failed to capture events immediately following incorporation and is more reflective of the final complex prior to pol β dissociation from product DNA. To examine intermediate structures subsequent to the catalytic event and product complexes, the pre-catalytic ground state complexes were soaked in a MgCl2 containing solution for longer periods of time (40 s, 90 s, 5 min, 45 min, and 11 h; Table S1). By 40 s, all complexes had been converted to product. This is in contrast to solution studies where insertion occurs more readily (~1/s) (Batra et al., 2008). This 40-fold drop in activity is likely due to the viscous cyrosolution, presence of inhibitory Ca2+ ions, and restrained thermal motion in the crystal. A similar reduced rate of incorporation in the crystal was observed for pol η (Nakamura et al., 2012).

Following a 40 s soak in MgCl2 we determined the structure of the ternary product complex to 1.7 Å (Table S1). The Fo-Fc omit map for the catalytic and nucleotide metal binding sites, incorporated nucleotide, and previous primer terminus is shown in Figure 3A. The density shows phosphodiester bond formation between O3′ and Pα of the incoming nucleotide with a corresponding loss of bond density between Pα and Pβ resulting in PPi (Figure 3A). While we describe this early time point in detail, similar Fo-Fc omit maps are observed for the 90 s, 5 min, 45 min, and 11 h time points (Figure 3D and Figure S3). Overlaying the product complex with the pre-catalytic ground state shows no significant global structural changes following catalysis (RMSD of 0.24 Å over 331 Cα). The active site of the product complex indicates no localized structural changes with the active site aspartate residues and α-helix N remaining in the closed position (Figure 3B). The closed conformation was observed in all product complexes after a correct nucleotide insertion (Figure 3 and Figure S3). Notably, the crystal lattice permits subdomain motions (i.e., opening and closing) with a 1-nt gapped DNA substrate indicating that pol β is not prevented from re-opening due to experimental artifacts. In this closed product conformation, PPi remains stably bound in the active site with the occupancy only decreasing to 75% after 11 h (Figure S3C and Table S1).

Figure 3
A Transient Third Metal Binding Site in the Closed Pol β Product Complex

Although no large protein structural changes following catalysis are observed, there are subtle substrate changes in the active site. The O3′ of the primer terminus shifts 1.1 Å to undergo nucleophilic attack at Pα of the incoming nucleotide. This nucleophilic attack results in Pα shifting 1.0 Å upstream and inversion of its non-bridging oxygens (Figure 3B). The resulting phosphodiester bond alters the coordination state of the catalytic Mg2+ hastening dissociation after bond formation. Based on the coordination geometry, we have modeled Na+ into the catalytic metal binding site (Figure 3C and Figure S3), but recognize that it is difficult to unambiguously identify the metal at this site (Batra et al., 2006). Indeed, the cation that replaces the catalytic metal will be determined by the local cellular concentration. In contrast, the nucleotide associated Mg2+ remains in the active site coordinating the aspartates and PPi (Figure 3 and Figure S4). The PPi undergoes a subtle change as one phosphate (Pγ remnant) is stripped from the nucleotide metal with a competing water molecule (Figure 3). At 40 s, this phosphate can be modeled in two conformers; directly coordinating the metal or in an extended conformation with a water mediated coordination to this divalent ion (Figure 3A and 3C). After 90 s, only the water mediated contact is observed (Figure 3D and Figure S3). This indicates PPi dissociation is facilitated by metal solvation in the closed polymerase conformation in a time dependent manner.

The active site geometry and metal composition of DNA polymerases is consistent with a two metal mechanism for nucleotide incorporation. Recently, an additional metal binding site was identified in pol η between Pα and Pβ that was suggested to stabilize an intermediate state (Nakamura et al., 2012). We also observe a similar metal binding site with pol β that appears after catalysis in the product complex. Figures 3C and S4 show that after 40 s a third binding site for a divalent metal ion appears between Pβ and Pα. We were able to model Mg2+ into the Fo-Fc omit map based on the number of coordinating ligands and distances. This third metal binding site was verified by doing a 40 s soak in MnCl2 and determining the structure of the product complex to 1.88 Å (Table S1). Using the anomalous signal from MnCl2, we confirmed the third metal binding site. An overlay of the structures from the MnCl2 and MgCl2 soaks shows the divalent metal ions in identical positions (Figure S5). The time dependence of the third metal binding site shows the occupancy of this site decreases over time. The occupancy was 0.6, 0.5, and 0.4 at 40 s, 90 s, and 5 min, respectively (Table S1). At the longer incubations (45 min and 11 h), the third metal binding site was void of any observable Mg2+ (Figure 3D and Figure S3). This confirms that this additional metal is only present for short times after catalysis and is not condition dependent. We do not observe three Mg2+ bound in the active site con-currently since the product complex appears to have a Na+ in the catalytic metal binding site. Significantly, the additional Mg2+ is not observed in structures that have a substantial population of reactant species (Figure 2). It is only observed in product complexes.

Incorrect nucleotide insertion

Previous structural studies of pre-catalytic pol β closed ternary complexes indicated that pol β utilizes a unique strategy to avoid misinsertion by shifting the coding base upstream of the template-binding pocket (Batra et al., 2008; Beard et al., 2009). Concomitantly, the primer-terminus sugar (i.e., O3′) rotates away from the catalytic metal (Figure 4A). By using the same time scale crystallography technique used for correct insertion we are able to capture key steps both during and after nucleotide misinsertion. To this end, we soaked binary pol β DNA complexes with a templating guanine in a cryosolution containing CaCl2 and dATP. The reaction was then initiated by transferring these ternary crystal complexes to a cryosolution containing MnCl2. MnCl2 was utilized because previous studies have shown that binding of an incorrect nucleotide was enhanced (Batra et al., 2008). By varying the time, we were able to capture events both during and after incorporation of the wrong nucleotide.

Figure 4
Pol β Pre-catalytic Mismatch Ternary Ground State

The incorporation of dAMP opposite guanine in the crystal was significantly slower than in solution and much slower than correct incorporation (Batra et al., 2008). The first observable change within the active site did not occur until after a 2.5 min soak. This structure diffracted to 2.0 Å and formed a closed ternary complex (Figure 4B and Table S2). Figure 4B shows a Fo-Fc omit map for the incoming nucleotide, metal ions, and primer terminus. The anomalous density clearly shows two active site Mn2+ being coordinated by the active site aspartic acid residues and a water molecule. Although no phosphodiester bond formation has occurred, there does appear to be a localized increase in the dynamics of the primer terminus. The primer terminus exhibits poor electron density for O3′ (B-factor = 58 Å2; average of the rest of the primer strand = 32 Å2). In contrast, O3′ of the previously reported ternary complex generated using a non-hydrolysable analogue was reported to have a B-factor of 46 Å2 (average of the rest of the primer strand = 31 Å2) (Batra et al., 2008). A structural overlay of the 2.5 min structure with the previous ternary mismatch ground state shows very similar global conformations (RMSD of 0.26 Å over 326 Cα). The only significant difference is the base of the incoming nucleotide forms a weak base pair with the templating guanine (O6 and N1 of guanine with N7 and N1 of adenine, respectively) (Figure 4C). Other key participants (primer terminus, templating base, aspartic acid residues, α-helix N, and metal ions) are in similar locations in each structure.

The first discernable structural event was observed after a 5 min soak in MnCl2. Intermediate ternary complex structures actively inserting an incorrect nucleotide after 5 and 10 min were determined to 2.0 Å (Table S2). Catalysis had proceeded 30 and 40% based on occupancy refinement at 5 and 10 min, respectively. A Fo-Fc omit map for the incoming nucleotide, active site metals, and primer terminus is shown for 5 and 10 min soaks (Figure 5A and B, respectively). At these time points the polymerase is closed based on the position of α-helix N (Figure 5A and B). In this closed intermediate catalytic state, the base of the incoming nucleotide shifts away from the templating base towards α-helix N, while the phosphate (Pα) from the incoming nucleotide remains in the active site. In contrast to the matched intermediate catalytic state, incorrect insertion does not exhibit clear bond density during phosphodiester bond formation. Instead, the phosphate position of the incorporated nucleotide corresponding to the product state is unique due to flipping 120° resulting in a 4.1 Å displacement (Figure 5A-C). The templating base, aspartate residues, and metals do not undergo observable significant rearrangements during catalysis. However, the catalytic metal occupancy decreases to 70 and 60% after 5 and 10 min, respectively. This is consistent with the catalytic metal diffusing out of the active site following catalysis as observed for post-catalytic events following correct nucleotide insertion.

Figure 5
Observing a Misincorporation Event

Product complex of a base substitution error

In contrast to the product complexes after a correct insertion, the mismatched product complexes show significant rearrangements. The structure of a ternary mismatch complex after a 30 min soak diffracted to 1.98 Å (Table S2). Figure 5C shows a Fo-Fc omit map for key active site residues and highlights that only the product complex remains after 30 min with no apparent density between Pα and Pβ. While the phosphodiester bond between Pα (dNTP) and O3′ (primer terminus) can be clearly observed, the sugar and base of the incorporated nucleotide remains diffuse. This is likely due to the increased space in the active site from the polymerase partially opening after catalysis. Figure 5C shows an overlay of α-helix N from the 30 min soak (magenta), pre-catalytic ternary mismatch complex (cyan), and the open binary complex (tan) showing the intermediate location of α-helix N after a 30 min soak. Additionally, the active site aspartate residues are in alternate conformations indicative of partial opening. The opening of α-helix N also promotes the release of PPi and the catalytic metal from the active site. The PPi and nucleotide metal have occupancy of 60% while the catalytic metal appears diffuse and can occupy several positions. Similar to the 30 min soak, an overnight soak with MnCl2 resulted in the substrates undergoing complete turnover and the polymerase re-opening fully. In this conformation, PPi and active site metals have disassociated from the active site. Figure 5D shows a Fo-Fc omit map for the product complex resulting from an overnight soak of a ternary mismatch complex. There is clear bond formation between the previous primer terminus O3′ and Pα of the incoming nucleotide, while the sugar and base for the incorporated nucleotide have poor density. The only observed stable conformation for the incorporated adenine is where it stacks over the templating guanine. The phosphate of the incorporated incorrect nucleotide has undergone rotation and is coordinating two water molecules while the active site aspartate residues are in a conformation typically observed in the fully open state. Figure S6 shows the global protein conformational changes during mismatch incorporation with α-helix N and the lyase domain adopting an open conformation as the reaction proceeds.


Significantly, there is no density corresponding to a third metal binding site in either omit map or anomalous density analysis of any of the mismatched ternary complexes. This is probably due to the large rotational movement of Pα away from Pβ as the reaction proceeds. The 4.1 Å distance between these phosphates prevents the stable bridging of these phosphates by an additional metal as observed after correct nucleotide insertion. In an attempt to decipher the role of this additional metal ion, we hypothesized that it might be involved in hastening the reverse reaction, pyrophosphorolysis. To test this idea, we conducted an assay for pyrophosphorolysis reactions initiated with nicked DNA containing either a matched or mismatched base pair at the nick. The results indicate that only the nicked substrate with the matched base pair at the nick served as a good substrate for the reverse reaction (Figure 6A). A mismatch at the primer terminus deters pyrophosphorolysis.

Figure 6
Insight into the Role of the Product-Associated Metal and Pyrophosphorolysis

To gain additional insight into the pyrophosphorolysis reaction we generated binary pol β product complexes by annealing oligonucleotides to create a nicked substrate with a matched base pair at the nick prior to crystallization. After crystallizing the nicked open binary complex, as previously reported (Sawaya et al., 1997), we performed a 30 min soak in mother liquor containing ethylene glycol, PPi, and MgCl2. The resulting ternary complex diffracted to 1.85 Å (Table S1) and indicated the polymerase had undergone closure with PPi bound in the active site (Figure 6B). 2Fo-Fc and Fo-Fc density maps show the catalytic metal binding site appears to contain Na+ while the nucleotide and product metal binding sites contain Mg2+ based on coordination distances and geometry (Figure 6B and C). This closed conformation represents a ground state structure and indicates that PPi promotes closure of the polymerase with a matched nicked DNA substrate. Unfortunately, we were unable to observe a reverse reaction probably due to the unfavorable equilibrium for this reaction. An overlay of the closed product complexes generated from the forward reaction (i.e., 40 s) and reverse reaction (i.e., nicked DNA with PPi) shows that both ground state product complexes have a nearly identical active site arrangement with the same ions in the catalytic, nucleotide, and product metal binding sites with bound PPi (Figure 6D).


Structural studies have often relied on substrate analogues (non-hydrolyzable dNTP, dideoxy-terminated DNA primer) or inactive mutant proteins that stop the reaction prior to catalysis resulting in non-catalytic ‘dead’ complexes. To overcome this problem, we have trapped catalytic intermediates prior to and following correct/incorrect nucleotide insertion of the natural substrates by freezing polymerase/DNA crystals at specific time intervals during catalysis. These structures identify additional intermediates not appreciated previously, as well as provide insights into the role of active site conformational changes that hasten or deter nucleotide insertion. Importantly, correct and incorrect incorporation show subtle but critical differences during catalysis providing molecular insights into enzymatic checkpoints ensuring high fidelity DNA synthesis.

Polymerase catalysis

In general, all DNA polymerases rapidly bind nucleoside triphosphates and in most cases, this is accompanied by repositioning of a polymerase subdomain that closes around the nascent base pair inducing subtle active site re-arrangements that either hasten or deter nucleotide insertion (Batra et al., 2008). The PPi product is released either during or after polymerase subdomain opening. Figure 7A illustrates the catalytic steps during the polymerase reaction for insertion of the right nucleotide following polymerase closure. First, the catalytic metal facilitates deprotonation of the primer terminus 3′-OH (step 1). The lack of an apparent stable water molecule within hydrogen bonding distance of 3′-OH during early time points supports the suggestion that Asp256 serves as the proton acceptor during activation of the primer terminus, consistent with previous computational studies (Lin et al., 2006). Following deprotonation, Pα of the incoming nucleotide undergoes nucleophilic attack by the primer-terminus O3′ (step 2). The in-line attack results in inversion of the phosphate group during phosphodiester bond formation (step 4). Either during or following nucleophilic attack, the PPi product can be protonated by a nearby ordered water molecule (step 3). Pyrophosphate, and its associated metal, remain stably bound in the active site. Importantly, during the course of the reaction the active site aspartate residues and metal ions remain in a constant configuration.

Figure 7
Mechanistic Model for Correct/Incorrect Incorporation and Pyrophosphorolysis

Following correct nucleotide insertion, the product remains in a conformation very similar to the ground state, except that the catalytic metal has dissociated from the active site. During catalysis the catalytic metal ion is believed to act as an electron sink to hasten catalysis. Loss of a key coordinating ligand (primer O3′) following phosphodiester bond formation distorts metal coordination that facilitates loss of the catalytic metal. Unexpectedly, the product complex after correct nucleotide insertion (i.e., nicked DNA) remains in a closed conformation with bound PPi. This is in contrast to previously reported open binary pol β nicked DNA complexes (Krahn et al., 2004; Sawaya et al., 1997) or 1-nt gapped DNA structures (Sawaya et al., 1997) suggesting that PPi can impact the global polymerase conformation. In the case of a 1-nt gap, the open conformation is energetically favored since the dNTP binding pocket is empty thereby precluding interactions with α-helix N. After correct insertion, but before PPi release, the nascent base pair interactions with α-helix N stabilize the closed conformation.

In contrast to the events with correct nucleotide insertion, the structural intermediates during incorrect insertion exhibit moderate rearrangements both during and after catalysis. Nevertheless, both correct and incorrect insertions appear to undergo catalysis in the polymerase closed conformation. Figure 7B shows the steps observed during mismatch incorporation following ternary substrate complex formation. In this case, since the primer O3′ is poorly positioned, it must move 2.9 Å to coordinate the catalytic metal. While we are unable to capture this movement, the primer-terminus exhibits elevated dynamic behavior based on its high B-factors and lack of any significant rearrangements in the active site that might suggest other catalytic re-organizational steps. Following coordination of the catalytic metal, subsequent catalytic events are similar to those proposed for correct insertion (Figure 7B). Surprisingly, the phosphate backbone of the nascent primer terminus flips away from the active site metals and PPi (step 4). This conformational adjustment likely arises from the strain on the DNA due to the template shift and lack of Watson-Crick hydrogen bonds between the incoming nucleotide and templating base in the closed conformation. The strain and subsequent phosphate rotation could promote re-opening of pol β following incorrect nucleotide incorporation. In this open state, the remaining divalent metal ion and PPi are able to dissociate from the active site. This observation is consistent with the idea that PPi is released during or after the polymerase opens (Dahlberg and Benkovic, 1991; Patel et al., 1991).

Polymerase fidelity

DNA polymerases must select the right nucleotide from a pool of structurally similar molecules to preserve the Watson-Crick structure of DNA. Kinetic and structural studies are consistent with an induced-fit mechanism for DNA polymerase substrate discrimination (Johnson, 2008). The induced fit model for the origin of DNA polymerase fidelity states that the correct nucleotide binding will optimally align catalytic participants for chemistry. In contrast, incorrect nucleotide binding misaligns critical catalytic atoms. Consistent with this model, the structures indicate that correct nucleotide binding induces a closed conformation optimally aligned for chemistry. In this state, deprotonation/protonation events accompany catalysis with resulting atomic changes (phosphate inversion) and a stable closed conformation. In contrast, incorrect nucleotide binding results in a closed conformation that deters catalysis (O3′ is misaligned). Since the open or partially open state is not significantly populated, the structural results indicate that O3′ could move to a catalytic competent position by melting of the primer terminus. This is consistent with the elevated B-factor calculated for the primer terminus. A suboptimal catalytic orientation or catalytic metal coordination of O3′ would result in a reduced catalytic rate. Likewise, the poor density for the sugar and base of the incorrect incoming nucleotide highlights the lack of stabilizing contacts thereby resulting in poor binding of wrong nucleotides.

Another key difference between correct/incorrect incorporation is the conformation of the primer terminus backbone. After incorrect insertion, the phosphate backbone rotates away from the active site suggesting that the nascent phosphodiester bond may be strained following formation of a mismatched base pair. This is in contrast to formation of a matched base pair where the phosphate backbone only undergoes minor adjustments following bond formation. This indicates that nucleophilic attack of the primer O3′ and subsequent bond formation is suboptimal in the structure with the wrong incoming nucleotide. Both the strain during bond formation and the movement of O3′ necessary in the closed conformation would decrease the efficiency of incorrect incorporation and increase fidelity. These differences in right or wrong substrate incorporation are consistent with the induced-fit model for DNA polymerase fidelity. Additionally, the distorted backbone conformation of the incorrect primer terminus would discourage further DNA synthesis providing an opportunity for error correction (see below).

Product metal binding site

The appearance of a third metal binding site in the product complex was unexpected. Our results indicate the metal transiently binds to this site after chemistry. Interestingly, this metal binding site coincides with the loss of the catalytic metal ion and results in the polymerase having only two divalent metal ions in the active site at any given time. A previous study with a Y-family DNA polymerase, pol η, observed an additional metal ion in an intermediate structure that was suggested to stabilize the transition state (Nakamura et al., 2012). In contrast to that study, we only observe this metal ion in the product state suggesting that it is not involved in facilitating the forward reaction for correct or incorrect insertion with pol β. These differences may be attributed to the unique features of the respective polymerase active sites; pol η active site is more solvent exposed and does not exhibit large conformational adjustments upon substrate binding. Both structural studies, however, indicate that this metal appears to coordinate non-bridging oxygens on the phosphates of the scissile bond of the incoming nucleotide (Figure S7A). Interestingly, a conserved lysine residue in A- and B-family DNA polymerases (Figure S7B) is situated in the vicinity of where the product-associated metal is observed suggesting that either this residue moves to permit an additional metal to bind or serves a similar role as the product-associated metal.

The position of the third metal binding site suggests a possible role to promote the reverse reaction, pyrophosphorolysis (Figure 7C). In this location, the metal could stabilize O1 of PPi following its deprotonation by a water molecule, reminiscent of the catalytic metal stabilizing O3′ during the forward reaction. Following deprotonation, PPi would undergo nucleophilic attack at the backbone phosphate of the primer terminus. Either during or following the nucleophilic attack O3′ is protonated and the catalytic metal binding site is restored. In this pre-catalytic ground state with the product nucleoside triphosphate, the enzyme could quickly undergo the forward reaction. We have attempted to capture intermediates during the reverse reaction, but the equilibrium for the overall reaction at the active site in the closed polymerase complex significantly favor the forward reaction. We were, however, able to capture the ground state of the reverse reaction by soaking open nicked DNA binary complex crystals in MgCl2 and PPi. The resulting closed polymerase complex with bound PPi and Mg2+ occupying the third metal binding site (Figure 6B–D) indicates that the same structural snapshot can be observed starting from the forward (with substrates) or reverse (with products) direction. This supports a role for this metal in establishing an active site organization that could be necessary for pyrophosphorolysis. Consistent with this hypothesis, product complexes with an active site mismatch do not exhibit this metal and are unable to support pyrophosphorolysis.

Biological implications

Using the time resolved crystallography technique described here permitted us to isolate and characterize intermediate structures during correct and incorrect nucleotide insertion. Accordingly, the catalytic events for both right and wrong incorporation can be assessed and fidelity checkpoints identified and described at a structural level. Consistent with an induced-fit mechanism that controls substrate specificity, the open/closed polymerase conformation is sensitive to correct/incorrect nucleotide binding and catalysis (forward and reverse reactions). The intermediate catalytic structures reported here provide an impetus to better understand pyrophosphorolysis and the impact of open and closed DNA polymerase states on the reaction equilibrium. Additionally, pyrophosphorolysis and post-chemistry conformational changes would also alter the distribution of enzyme species thereby influencing catalytic efficiency. Kinetic characterization of the sensitivity of the mitochondrial DNA polymerase γ to nucleoside analogues has demonstrated that the reverse reaction and post-chemistry conformational changes can influence nucleotide insertion (Hanes and Johnson, 2007). Consistent with those observations, the results described here demonstrate that the closed product complex and pyrophosphorolysis are promoted after correct, but not incorrect, nucleotide insertion. Likewise, the replicative A-family T7 DNA polymerase (Patel et al., 1991) and B-family ø29 DNA polymerase (Blasco et al., 1991) that inter-convert between open and closed states also cannot remove a mismatched terminus by pyrophosphorolysis. As originally noted for HIV-1 reverse transcriptase (Meyer et al., 1998), pyrophosphorolysis can be a biologically important catalytic activity. In the case of reverse transcriptase, an enhanced reverse reaction can decrease chain-terminating nucleoside inhibitor sensitivity.

DNA polymerases are optimized for high-fidelity DNA synthesis and a central tenet of the ‘induced-fit’ hypothesis is that incorrect nucleotide binding/insertion induces a suboptimal conformation that deters wrong nucleotide insertion. The poor active site geometry observed after misinsertion of a wrong nucleotide is consistent with the idea of an additional fidelity checkpoint that takes advantage of this poor geometry by deterring further nucleotide insertions (i.e., extension of a mismatch) that would result in a base substitution error. In addition to the poor active site geometry after misinsertion, the structures reveal that the polymerase opens and releases PPi more rapidly than after correct nucleotide insertion. Accordingly, pyrophosphorolysis would be a poor proofreading mechanism since it would require an active site geometry that would also hasten extension of a mismatched terminus; i.e., strong proofreading would be counterproductive. However, dAMP misinserted opposite the mutagenic DNA lesion 8-oxoG can be removed through pyrophosphorolysis by pol λ, a X-family DNA polymerase (Crespan et al., 2012). Consistent with the model described above, the A/8-oxoG base pair mimics a matched A/T base pair (McAuley-Hecht et al., 1994) making it an ideal substrate for pyrophosphorolysis. At the same time, the ease at which this promutagenic lesion can be extended makes it highly mutagenic.

Another functionally significant observation was the stable nature of the pol β closed product complex following correct, but not incorrect, insertion. Biologically, this conformational difference could provide a structural signal during base excision repair that denotes successful high-fidelity gap-filling DNA synthesis. Specifically, a closed polymerase conformation after correct incorporation could protect the resulting nicked substrate and facilitate channeling to DNA ligase. In this model the closed ternary product pol β conformation with nicked DNA would remain until DNA ligase binds and promote pol β disassociation. In contrast, incorrect insertion promotes an open polymerase conformation that would result in pausing of DNA synthesis and could be channeled to a proofreading enzyme such as AP endonuclease 1 to excise the mismatched nucleotide (Chou and Cheng, 2002).

Experimental procedures

DNA sequences used in this study are described in Extended Experimental Procedures.

Protein expression and crystallization

Human pol β was overexpressed in E. coli and purified as previously described (Beard and Wilson, 1995). Binary complex crystals with a templating guanine in a 1-nt gapped DNA were grown as described in Extended Experimental Procedures (Batra et al., 2006). Binary pol β:DNA complex crystals were first transferred to a cryosolution containing 15% ethylene glycol, 50 mM imidazole, pH 7.5, 20% PEG3350, 90 mM sodium acetate, 2 mM dCTP or dATP, and 50 mM CaCl2 for 30 min. Ground state ternary complex crystals were then transferred to a cryosolution containing 50 mM imidazole, pH 7.5, 20% PEG3350, 90 mM sodium acetate, and 200 mM MgCl2 or MnCl2 for varying times. Binary product complex crystals containing a nick in the active site were generated with the same sequence as the 1-nt gapped binary crystals, but with a longer primer (11-mer). Crystals were grown as described in Extended Experimental Procedures. Binary pol β:DNA nick complex crystals were transferred to a cryosolution containing 15% ethylene glycol, 50 mM imidazole, pH 7.5, 20% PEG3350, 90 mM sodium acetate, 5 mM PPi, and 200 mM MgCl2 for 30 min. All reactions were stopped by freezing the crystals at 100K prior to data collection at the home source or the Advanced Photon Source (Argonne National Laboratory).

Data collection and refinement

All data was collected at 100K. In house data collection was done at a wavelength of 1.54 Å and remote data collection was done at 1.0 Å, as described in the Extended Experimental Procedures. Data were processed and scaled using the HKL2000 software package (Otwinowski and Minor, 1997). Initial models were determined using molecular replacement with the previously determined open (3ISB), closed (2FMS), or mismatch (3C2M) structures of pol β. All Rfree flags were taken from the starting model. Refinement was carried out using PHENIX and model building using Coot (Adams et al., 2010; Emsley and Cowtan, 2004). Partial catalysis models were generated with both the reactant and product species and occupancy refinement was performed. The figures were prepared in PyMol (Schrödinger) and all density maps were generated after performing simulated annealing with a carve set at 3 and 1.6 Å for Fo-FC and 2Fo-Fc, respectively. Ramachandran analysis determined 100% of nonglycine residues lie in allowed regions and at least 98% in favored regions.


The reverse reaction, pyrophosphorolysis, was measured under single-turnover conditions (E[dbl greater-than sign]DNA) using a 5′-[32P] labeled nicked DNA substrate that contained either a matched (template/primer, G/C) or mismatched (G/A) primer terminus. Reactions procedures and conditions are detailed in the Extended Experimental Procedures.


  • Crystal structures (15) of a DNAP captured during right or wrong nucleotide insertion
  • A third metal binding site observed during right, but not wrong, nucleotide insertion
  • A stable closed enzyme conformation is observed after right, but not wrong, insertion
  • Structures provide insights into the induced-fit mechanism for substrate selectivity

Supplementary Material




We thank Lars Pedersen, Juno Krahn, and the Collaborative Crystallography group at the NIEHS for help with data collection and analysis. Use of the advanced Photon Source was supported by the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences, under Contract No. W-31-109-Eng-38. This research was supported by Research Project Numbers ZO1-ES050158 and Z01-ES050161 in the Intramural Research Program of the National Institutes of Health, National Institute of Environmental Health Sciences and was in association with the National Institutes of Health Grant 1U19CA105010.


Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.


  • Adams PD, Afonine PV, Bunkoczi G, Chen VB, Davis IW, Echols N, Headd JJ, Hung LW, Kapral GJ, Grosse-Kunstleve RW, et al. PHENIX: A comprehensive Python-based system for macromolecular structure solution. Acta Cryst D. 2010;66:213–221. [PMC free article] [PubMed]
  • Batra VK, Beard WA, Shock DD, Krahn JM, Pedersen LC, Wilson SH. Magnesium induced assembly of a complete DNA polymerase catalytic complex. Structure (Camb) 2006;14:757–766. [PMC free article] [PubMed]
  • Batra VK, Beard WA, Shock DD, Pedersen LC, Wilson SH. Structures of DNA polymerase β with active site mismatches suggest a transient abasic site intermediate during misincorporation. Mol Cell. 2008;30:315–324. [PMC free article] [PubMed]
  • Beard WA, Shock DD, Batra VK, Pedersen LC, Wilson SH. DNA polymerase β substrate specificity: Side chain modulation of the “A-rule” J Biol Chem. 2009;284:31680–31689. [PMC free article] [PubMed]
  • Beard WA, Shock DD, Yang XP, DeLauder SF, Wilson SH. Loss of DNA polymerase β stacking interactions with templating purines, but not pyrimidines, alters catalytic efficiency and fidelity. J Biol Chem. 2002;277:8235–8242. [PubMed]
  • Beard WA, Wilson SH. Purification and domain-mapping of mammalian DNA polymerase ® Methods Enzymol. 1995;262:98–107. [PubMed]
  • Beard WA, Wilson SH. Structure and mechanism of DNA polymerase β Chem Rev. 2006;106:361–382. [PubMed]
  • Beese LS, Steitz TA. Structural basis for the 3′-5′ exonuclease activity of Escherichia coli DNA polymerase I: A two metal ion mechanism. EMBO J. 1991;10:25–33. [PubMed]
  • Blasco MA, Bernad A, Blanco L, Salas M. Characterization and mapping of the pyrophosphorolytic activity of the phage [phi] 29 DNA polymerase: Involvement of amino acid motifs highly conserved in α-like DNA polymerases. J Biol Chem. 1991;266:7904–7909. [PubMed]
  • Chou KM, Cheng YC. An exonucleolytic activity of human apurinic/apyrimidinic endonuclease on 3′ mispaired DNA. Nature. 2002;415:655–659. [PubMed]
  • Crespan E, Maga G, Hubscher U. A new proofreading mechanism for lesion bypass by DNA polymerase-λ EMBO Rep. 2012;13:68–74. [PubMed]
  • Dahlberg ME, Benkovic SJ. Kinetic mechanism of DNA polymerase I (Klenow Fragment): Identification of a second conformational change and evaluation of the internal equilibrium constant. Biochemistry. 1991;30:4835–4843. [PubMed]
  • Emsley P, Cowtan K. Coot: model-building tools for molecular graphics. Acta Cryst D. 2004;60:2126–2132. [PubMed]
  • Freudenthal BD, Beard WA, Wilson SH. Structures of dNTP intermediate states during DNA polymerase active site sssembly. Structure (Camb) 2012;20:1829–1837. [PMC free article] [PubMed]
  • Hanes JW, Johnson KA. A novel mechanism of selectivity against AZT by the human mitochondrial DNA polymerase. Nucl Acids Res. 2007;35:6973–6983. [PMC free article] [PubMed]
  • Johnson KA. Role of induced fit in enzyme specificity: A molecular forward/reverse switch. J Biol Chem. 2008;283:26297–26301. [PMC free article] [PubMed]
  • Krahn JM, Beard WA, Wilson SH. Structural insights into DNA polymerase deterrents for misincorportion support an induced-fit mechanism for fidelity. Structure (Camb) 2004;12:1823–1832. [PubMed]
  • Lin P, Pedersen LC, Batra VK, Beard WA, Wilson SH, Pedersen LG. Energy analysis of chemistry for correct insertion by DNA polymerase β Proc Natl Acad Sci USA. 2006;103:13294–13299. [PubMed]
  • McAuley-Hecht KE, Leonard GA, Gibson NJ, Thomson JB, Watson WP, Hunter WN, Brown T. Crystal structure of a DNA duplex containing 8-hydroxydeoxyguanine-adenine base pairs. Biochemistry. 1994;33:10266–10270. [PubMed]
  • Meyer PR, Matsuura SE, So AG, Scott WA. Unblocking of chain-terminated primer by HIV-1 reverse transcriptase through a nucleotide-dependent mechanism. Proc Natl Acad Sci USA. 1998;95:13471–13476. [PubMed]
  • Nakamura T, Zhao Y, Yamagata Y, Hua Yj, Yang W. Watching DNA polymerase η make a phosphodiester bond. Nature. 2012;487:196–201. [PMC free article] [PubMed]
  • Otwinowski Z, Minor W. Processsing of X-ray diffraction data collected in oscillation mode. Methods Enzymol. 1997;276:307–326.
  • Patel SS, Wong I, Johnson KA. Pre-steady-state kinetic analysis of processive DNA replication including complete characterization of an exonuclease-deficient mutant. Biochemistry. 1991;30:511–525. [PubMed]
  • Sawaya MR, Prasad P, Wilson SH, Kraut J, Pelletier H. Crystal structures of human DNA polymerase β complexed with gapped and nicked DNA: Evidence for an induced fit mechanism. Biochemistry. 1997;36:11205–11215. [PubMed]