Structure of PolC reveals unique DNA binding and fidelity determinants

Abstract

PolC is the polymerase responsible for genome duplication in many Gram-positive bacteria and represents an attractive target for antibacterial development. We have determined the 2.4-Å resolution crystal structure of Geobacillus kaustophilus PolC in a ternary complex with DNA and dGTP. The structure reveals nascent base pair interactions that lead to highly accurate nucleotide incorporation. A unique β-strand motif in the PolC thumb domain contacts the minor groove, allowing replication errors to be sensed up to 8 nt upstream of the active site. PolC exhibits the potential for large-scale conformational flexibility, which could encompass the catalytic residues. The structure suggests a mechanism by which the active site can communicate with the rest of the replisome to trigger proofreading after nucleotide misincorporation, leading to an integrated model for controlling the dynamic switch between replicative and repair polymerases. This ternary complex of a cellular replicative polymerase affords insights into polymerase fidelity, evolution, and structural diversity.

DNA polymerases are the enzymes responsible for DNA synthesis. Cellular organisms typically use multiple DNA polymerase types. The “replicative” polymerase performs the bulk of genome duplication, whereas various specialty polymerases repair damaged DNA and resolve Okazaki fragments. Across every kingdom of life, replicative polymerases exhibit certain hallmarks such as high fidelity, speed, and processivity (1). Polymerase holoenzyme accessory proteins play an integral role in achieving the extraordinary efficiency and accuracy of the replicative polymerase complex. These include a “sliding clamp” that encircles the DNA and increases processivity (2).

Bacterial replicative polymerases comprise the C family of DNA polymerases (3) and differ significantly from the replicative polymerases of eukaryotes, bacteriophage, and archaea, which belong to the B family. The major C family replicative polymerases are DnaE (PolIII), found primarily in Gram-negative bacteria, and PolC (PolIIIC), found primarily in Gram-positive bacteria (4). Apoenzyme crystal structures of DnaE have revealed surprising structural differences in the catalytic center of the enzyme compared with B family polymerases, suggesting a separate evolutionary origin for the C family (5, 6).

As the core component of the replicative polymerase complex in Gram-positive pathogens such as Staphylococcus aureus, PolC has received considerable attention as a potential target for antibacterial drug discovery (7, 8). No currently marketed antibiotics target the central replication apparatus, making PolC a novel target for antibacterial development. Gram-negative and Gram-positive bacteria are separated by >1 billion years of evolution, and PolC and DnaE share <20% sequence identity; PolC is further differentiated from DnaE by domain rearrangements and by the presence of an intrinsic 3′–5′ proofreading exonuclease domain. Notably, PolC and DnaE frequently exhibit differential sensitivity to inhibition by nucleotide analogs (8), suggesting significant differences in active-site structure. Thus, the available structural information on DnaE provides limited insight into critical interactions governing PolC substrate binding and inhibition.

A significant body of literature is available on substrate-bound and apoenzyme structures of viral and bacteriophage replicative DNA polymerases (9, 10). These polymerases have specialized features geared toward genome duplication within the context of a host infection. Among cellular replicative polymerases, substrate ternary structure information has only recently become available with a low-resolution structure of Thermus aquaticus (Taq) DnaE (11). However, at 4.6-Å resolution, that TaqDnaE ternary complex did not reveal much detail regarding key DNA interactions. The present work describes a much higher-resolution structure of a cellular replicative polymerase, PolC, bound to nucleotide substrate and DNA, thus providing insights into critical DNA interactions for this unique class of DNA polymerase.

Results and Discussion

Three-Dimensional Structure of PolC.

We determined the structure of Geobacillus kaustophilus PolC (GkaPolC) to 2.4-Å resolution in a ternary complex with primer-template DNA and incoming nucleotide (Fig. 1). To protect the DNA from degradation during crystallization, we deleted the 3′–5′ proofreading exonuclease domain. We also removed a poorly conserved N-terminal domain from the construct. Importantly, these truncations did not compromise core polymerase function (Fig. 1B). The structure spans residues 233-1444, with the only disordered regions of the complex being the linker region that replaces the 3′–5′ exonuclease domain, 2 nearby loops, and the distal 4 bp of the DNA. We refined 3 complexes [supporting information (SI) Table S1] that differ in the divalent metal ions included during the crystallization (Mg²⁺/Zn²⁺, Mn²⁺ only, and Mn²⁺/Zn²⁺). Except where noted, we describe the Mg-bound structure because the 3 structures are nearly identical apart from the metal-binding sites. In all cases, the 3′ end of the primer strand DNA was terminated with a dideoxynucleoside to prevent catalysis.

Fig. 1.

PolC structure and domain organization. (A) Ribbon representation of PolC structure showing DNA primer strand (white) and DNA template strand (orange). dGTP is shown in sphere representation, as are bound Mg²⁺ (green) and Zn²⁺ (gray) ions. (B) Domains are shown for full-length PolC and for the crystallized truncation. Relative polymerase primer extension and 3′–5′ exonuclease activities are indicated. (C) PolC surface representation with domains colored as in B and modeling ssDNA (pink) binding to the OB fold in PolC based on the alignment of RPA70-ssDNA (1JMC) to the PolC OB fold (Fig. S4).

PolC exhibits the canonical polymerase configuration resembling a right hand. The central region of PolC forms the polymerase core (residues 828-1293), with fingers, palm, and thumb domains that are defined by their interactions with the DNA substrate. The DNA duplex is held between the thumb and the C-terminal domain of PolC, which we have termed the duplex-binding (DB) domain. The single-stranded DNA (ssDNA) template enters the polymerase active site through a crevice formed between the fingers and DB domain. An oligonucleotide/oligosaccharide-binding (OB) domain (Pfam PF01336) and a polymerase and histidinol phosphatase (PHP) domain (12) (Pfam PF02811) are located N-terminal to the polymerase domain.

Despite having limited sequence similarity, individual domains of PolC and DnaE (5, 6) superimpose with rmsds ranging from 1.19 Å for the OB domain to 2.55 Å for the index finger domain (Fig. S1). One unique feature of the PolC palm domain is that it contains a zinc finger (Fig. S2) consisting of a tetrahedral cluster of highly conserved cysteines that are essential for activity (13). Although the zinc finger points away from the polymerization active site and is unlikely to be directly involved in catalysis or DNA binding, it may play a structural role that stabilizes the palm domain. Zinc fingers have not been found in DnaE or other DNA polymerase structures.

ssDNA Template Interactions and a Role for the OB Domain.

Although the OB domains of PolC and DnaE are located on opposite sides of the polymerase domains, they may play similar roles in binding ssDNA. In fact, the OB motifs of PolC and ssDNA-binding (SSB) proteins show a high degree of similarity (Fig. S4). In the DnaE ternary complex (11), the OB domain appears to form a track guiding the template strand into the active site. In the PolC structure, the short ssDNA downstream template occupies the channel formed by the DB and fingers domain but does not reach the OB domain. PolC OB is located where it could bind the template strand ≈15–20 nt ahead of the polymerase active site (Fig. 1C). PolC OB plays an important role in intrinsic polymerase function; truncation to remove PolC OB results in an elevated K_m^DNA and a 20-fold reduction in polymerase activity (data not shown). SSBs play a critical role in melting template secondary structure in advance of the lagging-strand DNA polymerase. N-terminal fusion of RB69 SSB to polymerase enhances intrinsic processivity (14). Thus, it seems quite plausible that C family polymerases could benefit from an intrinsic SSB-like function. Interestingly, few other DNA polymerases exhibit OB folds.

PolC PHP Domain Has a Metal-Binding Cluster Yet Lacks Catalytic Activity.

The PHP motif is universally found in C family polymerases, but rarely in other polymerases (3). The diverse PHP family is associated with a range of hydrolase activities, with many PHPs still uncharacterized. Like other PHP structures, the PolC PHP exhibits a distorted (βα)₇-barrel and coordinates up to 3 metals and a phosphate (Fig. S3). All of the metal-coordinating residues are highly conserved in PolC (Fig. S5), suggesting a functional importance for metal binding. PHP domains found in DnaEs of thermophilic origin exhibit 3′–5′ exonuclease activity (11). In contrast, PolC PHP lacks detectable nuclease activity (Fig. S6). Site-directed mutagenesis of 2 PolC metal-coordinating residues did not alter thermal denaturation or core polymerase or exonuclease activity (data not shown). Interestingly, DnaE orthologs from proteobacteria such as Escherichia coli lack several of the key residues involved in metal binding (12) and thus appear unlikely to harbor an active exonuclease. Nonetheless, genetic evidence points to a functional importance of this domain in both PolC and DnaE, with mutations adjacent to metal-chelating residues exhibiting slow-stop and template-slippage phenotypes (15, 16). These phenotypes may reflect a role for the PHP domain in holoenzyme interactions or in facilitating efficient conformational changes necessary for polymerase fidelity.

Structural Clues to the Evolution of the 3′–5′ Exonuclease in C Family Polymerases.

Unlike DnaE, native PolC has an intrinsic exonuclease domain that bisects the PHP motif. Most PHP structures exhibit a uniformly parallel β-barrel, with alternating α-helices girding the barrel. PolC PHP has 1 antiparallel β-strand intervening at the exact point of the exonuclease insertion (Fig. 2); DnaE exhibits the same discontinuity in the orientation of the β-sheets yet lacks the intrinsic exonuclease. This suggests a 2-step evolutionary path in which the last common ancestor of PolC and DnaE arose via insertion of an exonuclease motif into a parallel β-barrel motif. Divergent evolution could then lead to PolC with its intrinsic exonuclease, and DnaE with its separate proofreading subunit (ε). This divergent evolution step is supported by sequence analysis of exonuclease domains (17).

Fig. 2.

PHP β-barrel architecture. Schematic representations are shown of canonical PHP (αβ)₇-barrel (Left) (see Fig. S3), the PolC PHP domain (Center), and the TaqDnaE PHP domain [Right, based on Protein Data Bank (PDB) ID code 2HPI]. Red circles indicate the locations of metal-chelating residues. Filled red circles indicate the locations of metal-chelating residues that are conserved between PolC and TaqDnaE, open circles indicate the locations of metal-chelating residues in PolC and TaqDnaE that are not conserved in EcoDnaE.

Active-Site and Nascent Base Pair Binding Pocket.

The polymerase active site is cradled between the palm and fingers domains (Fig. 3A). A strong electron density peak, assigned as Mg²⁺, was observed between the dGTP triphosphate and 2 absolutely conserved aspartates in the palm (Asp-973 and Asp-975), identifying these as catalytic residues. Octahedral coordination of the metal is completed by nonbridging oxygens of the α-, β- and γ- phosphates of the incoming nucleotide and by a water molecule. The Mg²⁺ is equivalent to Metal B that is observed in the ternary complexes of other polymerase families (10, 18). No electron density was observed at the putative Metal A site, possibly because of a lack of the 3′-hydroxyl in the primer. A third aspartate (Asp-1098) would likely assist Asp-973 and Asp-975 in the coordination of a second metal, if present, but in the current structure forms a hydrogen bond with Lys-1096. The DNA duplex has standard B form geometry except at the primer terminus, where the ribose is in a C3′-endo conformation that would position the 3′-hydroxyl (if it were present) for an inline attack on the α-phosphate of the nucleotide substrate.

Fig. 3.

PolC active site and DNA interactions. Protein is colored as in Fig. 1; the metal ion and DNA are colored by element. (A) Active site. Residues interacting with the triphosphate of the incoming nucleotide (dGTP), with Mg²⁺, and with the primer terminus are indicated. The C3′-atom of the primer terminus is 3.8 Å from the α-phosphate, comparable with the distance observed in other polymerase ternary complexes. The F_o − F_c omit map (green; 3 σ level) for the dGTP and the Mg²⁺ is shown. (B) Nascent base pair-binding pocket. dGTP and templating base are shown as sticks; DNA and residues within 4 Å of the nascent base pair are shown as spheres. (C) Thumb elements Tβ1–Tβ2 interacting with the DNA phosphodiester backbone at primer positions −3, −4, and −5 and template positions −5, −6, −7, and −8.

The nascent base pair fits into a tightly constrained pocket composed of the terminal base pair and residues from the fingers and palm (Fig. 3B). The triphosphate of the incoming dGTP is held into position by direct and water-mediated hydrogen bonds to Arg-1238, Arg-1218, and Tyr-1269 in the fingers and to Lys-970 and Ser-895 in the palm and through coordination of Metal B (Fig. 3A). Ser-895 is part of a conserved Gly–Ser motif, which is conserved in the β-nucleotidyltransferase (β-NT) superfamily (19). The nascent base pair stacks against Tyr-1269 in the fingers, whereas Thr-1188, also in the fingers, protrudes over the phosphate and ribose of the templating base, sterically constraining the solvent-exposed edge of the nascent base pair. All residues within 4 Å of the bound dGTP are strictly conserved among PolC orthologs (Fig. S5), suggesting that the GkaPolC structure would provide an excellent active site model for PolC orthologs from pathogens such as S. aureus and Bacillus anthracis.

It is difficult to compare the PolC and DnaE active-site interactions in much detail because the lower-resolution diffraction of the DnaE complex (11) only allowed limited rigid-body refinement of domains. The primer–template duplex is a straight B form helix when bound to PolC but appears to have a significant bend when bound to DnaE (Fig. 4A). Additionally, superposition of the PolC and DnaE palms brings the active-site residues and the phosphates of the incoming nucleotide into alignment, but the base pairs at the primer terminus are 20–30° degrees out of alignment. It appears that in the DnaE complex, the fingers are not as tightly closed around the nascent base pair as in PolC. This could account for the differences in DNA base orientation.

Fig. 4.

Comparison of polymerase active sites from the β-NT and classical superfamilies. The active sites and DNA substrates are shown from PolC (colored as in Fig. 1) and TaqDnaE (white; PDB ID code 3E0D) C family polymerases (A), Polβ (PDB ID code 2FMP), an X family polymerase (B), and RB69 (PDB ID code 1IG9), a B family polymerase (C). (B and C) Palms are colored magenta, and fingers are colored blue.

Duplex DNA Interactions.

PolC interacts extensively with the DNA substrate further away from the active site (Fig. 1C). The thumb and DB domains contribute the majority of these DNA-binding residues. A surface area of ≈1,880 Ų is buried at the protein–DNA interface, of which ≈28% is contributed by the thumb domain. Two highly conserved antiparallel β-strands of the thumb (Tβ1–Tβ2; see Figs. S5 and S7) track along the minor groove and make hydrogen bonding and van der Waals contacts to the phosphodiester backbones of both strands at primer positions −3, −4, and −5 and template positions −5, −6, −7, and −8 (Fig. 3C). This represents a unique DNA interaction motif not seen in DNA polymerase structures. The potential role of the thumb in replication fidelity is discussed below.

In the DB domain, 2 helix–hairpin–helix (HhH) motifs (residues 1386–1403 and residues 1409–1430) contact the DNA. The first of these HhH motifs contacts the backbone phosphates of the primer strand at 8 and 9 residues upstream of the nascent base pair, whereas the second HhH motif contacts backbone phosphates of the template strand 12 and 13 residues upstream of the nascent base pair. These interactions are conserved in the DnaE complex (11). Even a 17-aa C-terminal truncation of GkaPolC, which would partially disrupt the second HhH motif, results in a detectable decrease in polymerase activity (data not shown); C-terminal truncation of Bacillus subtilis PolC by 44 aa (corresponding to residue 1400 in GkaPolC) results in complete loss of activity (20), consistent with disruption of both HhH motifs.

Structural Basis for High-Fidelity DNA Synthesis.

The steric constraints on the nascent base pair (Fig. 3B) would strongly select against mispairs that deviate from Watson–Crick geometry (10, 21). Additionally, a network of direct and water-mediated hydrogen bonds contact the minor groove at the primer terminus and the nascent base pair (Fig. S8), which would be disrupted by mispaired nucleotides. By analogy with other polymerases (21), we expect that this would weaken binding of the DNA to the polymerase domain and facilitate proofreading by the 3′–5′ exonuclease domain. Misincorporation of ribonucleotides would likely be prevented by His-1275, which contacts the ribose of the incoming nucleotide. This residue is positioned so that it would obstruct binding of any nucleotide with a 2′-OH, thus acting as a steric gate (22).

The thumb of PolC may contribute to fidelity in a unique way, by sensing the width of the minor groove. β-Strands of the thumb (Tβ1–Tβ2) track along the minor groove, contacting the phosphodiester backbones of both strands but not the DNA bases (Fig. 3C). Additionally, the positive dipole at the N terminus of thumb helix Tα1 points directly at the primer strand phosphate 4 bases upstream of the incoming nucleotide and, together with Lys-1011, makes potential electrostatic contributions to DNA binding. In the event that a mispaired nucleotide eluded detection in the nascent base pair-binding pocket, the mispair would disrupt the helical geometry and could potentially be sensed up to 8 bp after incorporation, via disruption of these thumb domain interactions. This could, in turn, trigger movement of the 3′ end of the primer from the polymerase active site to the exonuclease active site (discussed below). These interactions are expected to be conserved throughout the C family polymerases because all of the secondary structure elements in the PolC thumb are also found in DnaE. The DnaE thumb does, however, contain an additional helix–loop–helix motif (TaqDnaE 527–565), which provides additional contacts with the DNA duplex and may reach across to contact the downstream template DNA.

C Family Polymerases Belong to the β-NT Superfamily.

The PolC palm domain has the same topology as the catalytic domain of human DNA polymerase β (Fig. 4 A and B), providing further evidence that the C family bacterial replicative polymerases are not evolutionarily related to the B family eukaryotic and phage replicative polymerases (5, 6, 19) typified by RB69 polymerase (Fig. 4C). In addition, conserved residues in the PolC and Polβ palm domains interact with substrate DNA in very similar ways. In contrast to TaqDnaE and Polβ, however, the plane of the primer terminal base pair in PolC makes a ≈45° angle relative to the β-strand (P β4) bearing catalytic residues Asp-973 and Asp-975 (Fig. 4A). Modeling of DNA in the 2 DnaE apo structures (5, 6) relied on assumptions regarding the alignment of the catalytic aspartates with Polβ and on the expected parallel plane of the base pairs relative to the β-sheet. We believe that the resulting discrepancies arose from unexpected flexibility in the palm domain of the C family polymerases, as discussed below.

Conformational Flexibility in C Family Polymerases.

In the absence of a PolC apoenzyme structure, we have compared the ternary complex of PolC with the DnaE apoenzyme structures to determine the types of conformational changes that might occur upon binding of DNA and incoming nucleotide. If PolC and DnaE apoenzymes adopt a similar conformation, the index finger would need to rotate by ≈20° as the polymerase transitions between the apoenzyme (open) conformation and the ternary complex (closed) conformation (Fig. 5A). By analogy with A and B family polymerases, binding of incoming nucleotide to a PolC binary complex with primer-template DNA is expected to trigger closure of the index finger, whereas the release of pyrophosphate after nucleotide incorporation is expected to trigger its opening (23).

Fig. 5.

Conformational flexibility in C family polymerases. Conformational changes predicted to occur in index finger (A), palm (B), and DB domains (C). PolC (colored as in Fig. 1) and TaqDnaE (white; PDB ID code 2HPI) were aligned by superimposing the middle finger because the movements can be most easily described relative to the middle fingers and because this alignment brings the catalytic residues into close proximity and the palm domains superimpose less well (see Fig. S1). Arrows indicate rotations required to bring additional PolC and DnaE domains into alignment. Functionally important residues that would undergo large movements are shown in stick representation (magenta in A). (B) DnaE residues Arg-767 and Arg-458 are shown in 2 conformations, from TaqDnaE structures without (2HPI) and with (2HPM) bound nucleotide. When nucleotide is present, these residues directly contact the triphosphate and are oriented to point toward the active site and may be functionally equivalent to Arg-1238 and Lys-970 in PolC.

Surprisingly, it appears that part of the palm may move in concert with the index finger (Fig. 5B). PolC residue Lys-970, which interacts with the γ-phosphate of the incoming nucleotide, is located in a short loop of the palm domain. The equivalent loop in the DnaE apoenzyme structure is positioned further away from the polymerase active site, suggesting an unexpected flexibility in the palm domain of the C family polymerases. Movement of this loop could readily extend to 2 of the catalytic residues, Asp-975 and Asp-973, raising the possibility that the active conformation of the palm is only induced upon binding of incoming nucleotide.

Superposition of the PolC and DnaE ternary complexes based on the conserved 4-stranded β-sheet and underlying α-helices suggests that the conformational flexibility may be even more extensive. Not only does the index finger of DnaE appear not to be fully closed, it appears that the middle finger and the N-terminal portion of the palm also must move in toward the DNA (Fig. 4A). Although these differences may be an artifact of the lower resolution limit of the DnaE structure, the potential for movement of these domains is supported by the conformational differences in the TaqDnaE and E. coli DnaE (EcoDnaE) palm domains (5, 6) and would account for the difficulty of modeling DNA binding to EcoDnaE by superimposing the palm domains of Polβ ternary complex and DnaE apoenzyme (6). Comparison of the PolC and DnaE structures also indicates that the DB domain undergoes a large (≈30°) rotation, as seen with DnaE (5, 11). This motion probably occurs upon binding of primer–template DNA to form a binary complex (Fig. 5C). The implications of this are discussed below.

Holoenzyme Interactions and Implications for Switching from Replication to Repair.

Replicative polymerases generally operate in association with a sliding clamp (e.g., the PolIII β-subunit) that encircles the DNA and greatly enhances processivity. The structure of DNA-bound β has recently been reported (24), thus facilitating modeling with PolC (Fig. 6A). This model places the β-binding motif of PolC near the polymerase-binding motif on the clamp. Because DNA passes through β at an angle, there is sufficient space between β and the PHP domain to accommodate the 3′–5′ exonuclease domain in the intact PolC enzyme. Similar models have been proposed for DnaE (5, 6, 11).

Fig. 6.

Modeling of PolC holoenzyme. (A) Polymerization mode. (B) Proposed exonuclease mode. The position of β (pale green and blue ribbons) was modeled as described in SI Materials and Methods. The PolC β-binding motif (Q1440-F1444; ref. 29) is shown with orange spheres, and the polymerase peptide binding to each β subunit is shown in magenta. The structure of a homologous exonuclease domain (red) from Thermotoga maritima PolC (175 residues; PDB ID code 2P1J) is shown to illustrate where the PolC exonuclease domain (205 residues) is expected to be located. The point of the exonuclease insertion in the PHP domain is shown with red surfaces. PolC is colored as in Fig. 1. In B the DB domain, DNA, clamp, and clamp-binding peptides were rotated as a rigid body to align the DB domain with the position of the DB domain in the TaqDnaE apoenzyme structure (see Fig. 5C).

These models provide a static view of the holoenzyme, but do not suggest a dynamic mechanism for the enzyme to switch from polymerization mode to either exonuclease proofreading or translesion synthesis modes. In the A and B family DNA polymerases, the thumb domain maintains contact with the DNA duplex in both the polymerizing and proofreading modes (10, 25). Given the location of the C family polymerase thumb between the polymerase and exonuclease active sites, it is difficult to envision how the thumb could guide the movement of the nascent DNA strand between the polymerase and 3′–5′ exonuclease active sites, especially while β remains bound to the DB domain. For the C family polymerases, we propose instead that the switch between polymerase and exonuclease active sites is triggered by release of active site and thumb–DNA contacts (as discussed above) with concomitant rotation of the DB domain away from the polymerase domain (Figs. 5C and 6B), resulting in the primer terminus moving toward the exonuclease domain by ≈20 Å.

An intriguing implication of the PolC replication complex model is that the switch between replicative and repair polymerases could also be modulated by movement of the DB domain. In bacteria, the homodimeric clamp utilizes twin peptide-binding pockets on the monomer subunits to facilitate dynamic switching between the replicative polymerase and various repair polymerases (26, 27). During active DNA synthesis by the replicative polymerase, our model suggests that a repair polymerase bound to the second peptide-binding site on the clamp would be physically blocked from accessing the primer–template junction by the exonuclease domain (Fig. 6A). A 30° rotation of the DB, possibly triggered by damaged DNA-disrupting contacts with the polymerase domain, would provide space for a repair polymerase to access the β-clamp and the DNA (Fig. 6B). These alternate conformations would explain why, in the case of E. coli replication, PolIV can only take over DNA synthesis from PolIII (DnaE) when DNA synthesis stalls. Movement of the DB domain would thus provide a mechanism for coordinating high fidelity DNA replication with both translesion DNA synthesis and DNA repair.

Genetic evidence supports the role of the thumb domain in modulating C family polymerase fidelity and the switch to repair polymerases. An EcoDnaE A498T thumb mutation has an antimutator phenotype that is independent of exonuclease activity and facilitates displacement of DnaE by SOS polymerase PolII (28). The altered residue would be predicted to weaken the thumb contacts with the DNA, predisposing the mutant polymerase to the conformational change needed for polymerase switching.

Conclusions

The ternary structure of GkaPolC provides insights into key interactions governing substrate binding in a cellular replicative polymerase and reveals DNA binding and fidelity determinants. The high degree of sequence conservation around the nascent base pair-binding site suggests that the structure of GkaPolC will be an excellent model for PolC orthologs from bacterial pathogens and for structure-based design of inhibitors that target the polymerase active site. Finally, the GkaPolC ternary complex provides a solid structural framework for understanding how the replisome can coordinate the activities of multiple polymerases and DNA repair enzymes through interactions with the sliding clamp.

Materials and Methods

Protein Expression and Purification.

GkaPolC was PCR-cloned from genomic DNA and overexpressed in E. coli with a C-terminal His₈ tag. The exonuclease domain (residues 425–617) and N-terminal residues 1–227 were deleted. PolC was purified by using Ni–nitrilotriacetic acid–Sepharose (Qiagen), Mono Q (GE Healthcare), and Superdex-200 (GE Healthcare) chromatography (see SI Materials and Methods for details).

Protein Crystallization, X-Ray Data Collection, and Phasing.

Ternary complexes of PolC (15 mg/mL), DNA (130 μM), and dGTP (2 mM) were mixed in solution. The oligonucleotide used in crystallization corresponded to a primer–template pair where the dideoxy-terminated primer strand was 17 residues long and the template strand was 22 residues long (see SI Materials and Methods for details). GkaPolC complexes were crystallized by sitting-drop vapor diffusion (see SI Materials and Methods for details). The structure of GkaPolC was originally solved by single anomalous dispersion phasing from data collected from a crystal containing selenomethionine protein (see SI Materials and Methods for details). The crystals used for initial structure solution contained the same GkaPolC construct and nearly identical DNA oligonucleotides as the structures described here, but they included a proprietary small-molecule inhibitor instead of dGTP. The structures of the GkaPolC–DNA–dGTP complexes were solved by molecular replacement using the protein and DNA portions of the inhibitor-bound ternary complex as a search model (see SI Materials and Methods for details). Figures were prepared with PyMOL (Delano Scientific).

Data were collected from 3 separate crystals grown in the presence of different divalent metal ions: 10 mM MgCl₂ plus 2 mM ZnCl₂ (Mg²⁺/Zn²⁺; 2.4 Å), 2 mM MnCl₂ (Mn²⁺ only; 2.5 Å), and 2 mM MnCl₂ plus 1 mM ZnCl₂ (Mn²⁺/Zn²⁺; 2.5 Å). All data were collected by using synchrotron radiation at a wavelength of 0.99 Å. At this energy, manganese and zinc are expected to produce a measurable anomalous signal, but magnesium will have a negligible anomalous signal. Anomalous difference maps were calculated for all 3 datasets to aid in identification of metals bound to PolC. The 2.4-Å data for the Mg²⁺/Zn²⁺ structure had an overall R_merge of 10.1%, and the refined structure had R_cryst/R_free values of 22.9%/27.3%. Complete data collection and refinement statistics can be found in Table S1.