Structure of a high fidelity DNA polymerase bound to a benzo[a]pyrene adduct that blocks replication.

Of the carcinogens to which humans are most frequently exposed, the polycyclic aromatic hydrocarbon benzo[a]pyrene (BP) is one of the most ubiquitous. BP is a byproduct of grilled foods and tobacco and fuel combustion and has long been linked to various human cancers, particularly lung and skin. BP is metabolized to diol epoxides that covalently modify DNA bases to form bulky adducts that block DNA synthesis by replicative or high fidelity DNA polymerases. Here we present the structure of a high fidelity polymerase from a thermostable strain of Bacillus stearothermophilus (Bacillus fragment) bound to the most common BP-derived N2-guanine adduct base-paired with cytosine. The BP adduct adopts a conformation that places the polycyclic BP moiety in the nascent DNA minor groove and is the first structure of a minor groove adduct bound to a polymerase. Orientation of the BP moiety into the nascent DNA minor groove results in extensive disruption to the interactions between the adducted DNA duplex and the polymerase. The disruptions revealed by the structure of Bacillus fragment bound to a BP adduct provide a molecular basis for rationalizing the potent blocking effect on replication exerted by BP adducts.

Of the carcinogens to which humans are most frequently exposed, the polycyclic aromatic hydrocarbon benzo[a]pyrene (BP) is one of the most ubiquitous. BP is a byproduct of grilled foods and tobacco and fuel combustion and has long been linked to various human cancers, particularly lung and skin. BP is metabolized to diol epoxides that covalently modify DNA bases to form bulky adducts that block DNA synthesis by replicative or high fidelity DNA polymerases. Here we present the structure of a high fidelity polymerase from a thermostable strain of Bacillus stearothermophilus (Bacillus fragment) bound to the most common BP-derived N 2guanine adduct base-paired with cytosine. The BP adduct adopts a conformation that places the polycyclic BP moiety in the nascent DNA minor groove and is the first structure of a minor groove adduct bound to a polymerase. Orientation of the BP moiety into the nascent DNA minor groove results in extensive disruption to the interactions between the adducted DNA duplex and the polymerase. The disruptions revealed by the structure of Bacillus fragment bound to a BP adduct provide a molecular basis for rationalizing the potent blocking effect on replication exerted by BP adducts.
Structural studies of DNA polymerase complexes (for review, see Refs. [23][24][25][26][27] in combination with extensive enzyme kinetic studies (Refs. 28 -32; for review, see Ref. 33) have revealed the dominant mechanistic and structural features that contribute to accurate DNA replication, which are shared in large part by all polymerases. These features are well represented by our model system, the thermophilic Bacillus stearothermophilus DNA polymerase I large fragment (BF), that is capable of catalyzing DNA replication in a crystal. Using this system, we have previously studied replication of undamaged DNA (34,35), damaged DNA (36,37), and DNA mismatches (38) revealing mechanisms of replication and structural features that are not compatible with binding and replication of a bulky DNA adduct like [BP]dG. During replication, steric and geometric constraints are imposed on the template base as it moves from a pre-insertion site (see schematic, Fig. 2), where the template base is sequestered before pairing opposite an incoming dNTP, to the insertion site, where base pairs that exhibit the shape and geometry of correct Watson-Crick base pairs are selected in favor of base pairs that do not (39,40). The steric and geometric constraints at the pre-insertion and insertion sites contribute to the accuracy of replication achieved by high fidelity polymerases and would not be expected to accommodate [BP]dG without significant distortion of the polymerase active site. After nucleotide incorporation at the insertion site, the newly formed base pair translocates to the post-insertion site. Here and throughout the DNA duplex binding region, the polymerase engages in extensive interactions with the DNA minor groove, whereas the DNA major groove is predominantly solvent-exposed. Interactions between the DNA duplex and the polymerase are inevitably compromised by [BP]dG, a minor groove adduct.
Structures of [BP]dG adducts at single strand-double strand junctions in template-primer complexes in aqueous solution have been determined by NMR methods. These structures re-veal that [BP]dG can adopt either a syn conformation (41) or an anti conformation (42) depending on its position in the duplex and whether it pairs with an opposing base. Because these structures are determined in the absence of protein, they do not address [BP]dG conformation in the context of the polymerase active site. Recently, structures of DNA polymerases bound to the bulky major-groove DNA adducts (ϩ)-cis-[BP]-adenine (43), a less prevalent product of BPDE attack on DNA, and aminofluorene (37) have been reported. However, the structure of a DNA polymerase bound to a bulky minor-groove DNA adduct has yet to be determined.
Here we report a high resolution structure of the minor groove (ϩ)-trans-[BP]dG adduct, the major product of the reaction of (ϩ)-anti-BPDE with cellular DNA, at the BF active site. The [BP]dG adduct is observed in an anti conformation at the post-insertion site, where it forms Watson-Crick base pairs with an opposing cytosine. The polycyclic aromatic BP ring system protrudes into the nascent minor groove and induces distortions to the polymerase and both strands of the DNA duplex. Such distortion has not previously been observed in structures of DNA mismatches or lesions at the polymerase active site. The extensive nature of these distortions accounts for the blocking effect of [BP]dG on DNA replication.

EXPERIMENTAL PROCEDURES
Preparation of BP-modified DNA Duplexes-A Biosearch Cyclone automated DNA synthesizer was used to synthesize the oligonucleotides for this study. The sequence 5Ј-d(ACTCGCACCATCCCT) phosphorylated at the 3Ј-end was synthesized by automated methods using standard phosphoramidite derivatives of 2Ј-deoxynucleotides and 3Јphosphate controlled-pore glass supports. The direct synthesis approach (44,45) that was initially used to generate the structural features of the (ϩ)-trans-anti-BP-N 2 -dG adduct in an oligonucleotide duplex by NMR methods (46) was utilized here to generate BPDEmodified sequence 5Ј-d(ACTC-[BP]dG-CACCATCCCT). The oligonucleotide was dissolved in 3.2 ml of 0.05 M triethylamine acetate, 0.3 M NaCl, pH 10, buffer solution and treated with a 0.02 M BPDE tetrahydrofuran solution (initial molar ratio of [BPDE]/[oligonucleotide] was 2:1) overnight at room temperature and neutralized to pH 7.0 by the addition of 20 mM sodium phosphate. Modified and unmodified oligonucleotides were separated by reverse-phase HPLC on a PRP-1 column with a 10 -45% acetonitrile gradient 20 mM triethylamine acetate. The oligonucleotides containing the four different stereoisomeric anti-BP-N 2 -dG adducts were separated by reverse-phase HPLC on a 4.6 ϫ 250-mm C18 Microsorb-MV-89 column and a 13-25% 1:1 acetonitrile: methanol/triethylamine acetate (20 mM, pH 7.0) gradient. The modified oligonucleotides eluted in the order (10R(ϩ)-cis-, 10R(Ϫ)-trans-, 10S(Ϫ)-cis, and 10S(ϩ)-trans-anti-BP-N 2 -dG) as described (45). The oligonucleotide containing the single (ϩ)-trans-anti-BP-N 2 -dG adduct was purified by repeated HPLC cycles with similar protocols. The stereochemical properties of the modified oligonucleotide containing the (ϩ)-trans adduct were verified as described (7,44).
Primer Extension Assays-In vitro primer extension reactions were carried out with 43-mer template strands where the (ϩ)-trans-anti-BP-N 2 -dG lesions are positioned at the 25th nucleotide (counting from the 3Ј-side) as described (14). The 43-mer templates were annealed with a 32 P-labeled 19-mer primer (Fig. 1B). A time course of primer extension assays of the adducted templates with BF was determined at 37°C in 10 l of buffered solutions containing 50 mM Tris-HCl, pH 8.0, 5 mM MgCl 2 , 1 mM dithiothreitol, 50 g ml Ϫ1 bovine serum albumin, 4% glycerol, 4.5 nM primer-template complexes, 100 M dNTPs, and 20 nM of the polymerase. Corresponding unmodified primer/template complexes were extended as controls. The reaction conditions for the unmodified DNA substrates were the same as those for the modified templates, except that the enzyme concentration was 0.025 nM. The reactions were stopped after preselected time intervals by the addition of a 7 l of stop solution (20 mM EDTA, 95% formamide, 0.05% bromphenol blue, 0.05% xylene cyanol). Characterization of dNTP incorporation preference opposite [BP]dG was performed using DNA duplexes identical to those used for crystallization. Corresponding unmodified primer-template complexes were extended as controls. The 15-mer BPmodified and unmodified templates were annealed with a 32 P-labeled 9-mer primer (Fig. 1C). Reaction conditions were identical to those described above for extension of the 19-mer primer except that 5 mM dCTP, dTTP, dATP, or dATP and 10 nM BF (for BP-modified duplex) or 200 M dNTP and 0.1 nM BF (for unmodified duplex) and 100 nM DNA duplex (both BP-modified and unmodified) were used. Reactions were performed at 25°C to prevent the duplex from unannealing and allowed to proceed for 10 min before quenching with the stop solution. All reactions were heated to 90°C (5 min) and chilled on ice (1 min). Reaction products were separated by gel electrophoresis (20% denaturing polyacrylamide gel containing 7 M urea), visualized by autoradiography, and quantitated by a Storm 840 PhosphorImager using Storm ImageQuant software.
Data Collection, Structure Determination, and Superposition-Crystals were transferred to cryoprotectant solution (60% saturated ammonium sulfate, 24% sucrose, 100 mM MES, pH 5.8) before being flashfrozen in liquid nitrogen. Data were collected at beamline X12-B at Brookhaven National Synchrotron Light Source. Data were processed using DENZO and SCALEPACK (48). The co-crystals belong to space group P2 1 2 1 2 1 and contain one molecule per asymmetric unit. Phases were calculated using molecular replacement (BF bound to unmodified DNA duplex, PDB entry 1L3S, as the probe), and models were refined using CNS (49). Force field and topology parameters allowing multiple sugar pucker conformations were used for the DNA and were modified to include parameters for the [BP]dG residue generated using PRODRG (50). The structures were visualized and modeled with the program O (51). The quality of the final models was assessed using PROCHECK (52) and REDUCE (53). DNA base pair parameters were determined using 3DNA (54). The figures were drawn using PyMOL (55). The structure of BF bound to [BP]dG was superimposed with the structure of BF bound to an unmodified DNA duplex (PDB entry 1L5U). A model of BP at the insertion site of BF in a ternary complex with DNA duplex and an incoming dNTP was constructed by superimposing the structure of BF bound to C:[BP]dG (primer-template) at the post-insertion site (BP2, PDB entry 1XC9) with the structure of BF bound to unmodified duplex DNA and an incoming dNTP (PDB entry 1LV5). All superpositions were done using the C␣ atoms of BF residues 646 -655, 823-838, and 863-869.

RESULTS
[BP]dG Blocks DNA Replication by BF in Solution-Primer extension assays using unmodified and BP-modified oligonucleotides as templates demonstrate the blocking potential of the [BP]dG adduct on replication by BF as shown in Fig. 1. Replication of the unmodified DNA templates is highly efficient at low enzyme concentration (0.025 nM) with full extension products observed within 2 min of reaction initiation. By contrast, replication of BP-modified templates is blocked either before (position 24, Fig. 1B) or after base incorporation opposite [BP]dG (position 25) despite a ϳ1000-fold higher enzyme concentration (20 nM). At this enzyme concentration, full primer extension does not exceed a few percent even after an incubation time of 30 min (Fig. 1B, right-hand panel). Insertion of dNTP opposite [BP]dG is notably more facile than synthesis past the lesion (Fig. 1B, right-hand panel) with dCTP being the preferred nucleotide incorporated opposite [BP]dG with ϳ11% of primers extended (Fig. 1C). These results demonstrate the potency of [BP]dG as a block to DNA replication, especially primer extension past the lesion.

Structures of [BP]dG at the Pre-insertion and Post-insertion
Sites-A co-crystal structure of BF bound to a DNA duplex where [BP]dG is positioned as the n template base was determined to 1.9 Å of resolution (complex BP1, Table I). This structure represents BF before nucleotide incorporation oppo-site [BP]dG. In unmodified DNA, the n template base occupies the pre-insertion site at this stage of the reaction cycle of BF (35). By contrast, the modified [BP]dG template base does not occupy the pre-insertion site when placed at the n template base but is instead disordered. In this structure all other parts of the DNA duplex and the polymerase, including the preinsertion site, post-insertion site, and catalytic site, are well ordered and resemble the structure of BF bound to unmodified DNA duplexes (34,35).
Two approaches were used to capture [BP]dG at the postinsertion site paired opposite the primer terminus base. First, enzymatic nucleotide incorporation in the crystal was attempted by soaking crystals of the BP1 complex in stabilization solutions containing dNTP. In the BF crystal, complete nucleotide incorporation opposite unmodified bases is typically observed within 1 h. However, nucleotide incorporation opposite [BP]dG was not observed even after 1 month of soaking; preand post-soak structures were identical, indicating that nucleotide incorporation had not occurred. Second, BF was co-crystallized with duplex DNA, resulting in a complex with [BP]dG paired opposite cytosine at the post-insertion site (BP2, Table  I). This co-crystallization strategy has been used successfully to capture DNA mismatches at the post-insertion site that are indistinguishable from structures of the same DNA mismatches determined after enzymatic synthesis in the crystal (38).
The co-crystal structure of BF bound to C:[BP]dG (primertemplate) at the post-insertion site (Fig. 2) reveals extensive distortions to both the DNA duplex and the polymerase. Distortions to the DNA duplex extend from the n-1 (post-insertion site) to the n-4 position (the DNA duplex region) (Fig. 3B). At each of these positions, Watson-Crick hydrogen bonds between the bases are retained, but the base pair parameters are distorted (Table II)  [BP]dG forces polymerase side chains to adopt alternate conformations such that the interactions between the protein and the C:[BP]dG base pair are lost (Fig. 3C). Rather than form direct hydrogen bonds with the minor groove of the newly formed base pair, protein residues Arg-615 and Gln-797 are confined to the floor of the active site, far removed from the primer and template strands, which lift away from the surface of the polymerase. This displacement of the primer strand results in loss of coordination between the primer 3Ј-OH and the universally conserved catalytic Asp-830. The hydrogen bond distance between the primer 3Ј-OH and Asp-830 is normally 2.5 Å; in the BP2 complex, this distance is 6.9 Å. The 3Ј-OH is, therefore, no longer in position for in-line attack during nucleotide addition.
The BP Moiety Prevents Binding of the Incoming dNTP-To assess the effect of [BP]dG on binding of an incoming dNTP, the structure of BF bound to C:[BP]dG at the post-insertion site was superimposed with the structure of BF bound to duplex DNA and an incoming dNTP (PDB entry 1LV5) (Fig. 3D). In this modeled structure, the adduct moiety overlaps with the bound dNTP and is, therefore, sterically incompatible with extension beyond the adducted base without significant structural distortion to the polymerase or DNA. With binding of the incoming dNTP prevented by the presence of the BP adduct, replication is, therefore, effectively blocked.

DISCUSSION
Structures of [BP]dG at the BF polymerase active site that represent the adducted base before and after incorporation of dCTP reveal structural distortions that provide a structural rationale for the block to replication exerted by [BP]dG.
[BP]dG induces distortions that promote stalling of the polymerase at either the pre-insertion or post-insertion site and dissociation of the polymerase from the duplex, thereby promoting subsequent excision or translesion synthesis by repair polymerases.
Upon encountering [BP]dG as the n template base, BF pauses, and replication is stalled (as shown by primer extension assays, Fig. 1). During replication of unmodified DNA, the n template base transitions from the pre-insertion site to the insertion site, where it pairs opposite the incoming dNTP. Because of the bulk of the BP moiety, [BP]dG cannot occupy the pre-insertion site and is, therefore, not poised for nucleotide incorporation even though the polymerase active site is not disrupted. The bulk of the BP moiety and the limited conformational degrees of freedom of [BP]dG relative to unmodified guanine may also impair the transition from pre-insertion site to insertion site, which involves a 90 o bend in the template strand. Nucleotide incorporation opposite [BP]dG is, therefore, significantly impaired.
Although limited nucleotide incorporation opposite [BP]dG is possible, extension past [BP]dG is essentially not observed. This inability to extend beyond the lesion in the template strand can be rationalized by the observation that [BP]dG adopts an anti conformation at the BF post-insertion site, where the BP residue positioned in the nascent minor groove points toward the 5Ј-end of the template strand.
[BP]dG also adopts anti conformations in structures of BP-modified DNA duplexes in the absence of protein as determined by NMR (42,56). However, comparison of these NMR structures with the present structure reveals significant differences. One component of these differences arises from distortion to the DNA duplex observed when the duplex is bound to polymerase. Deviation of the polymerase-bound DNA duplex from normal Bform DNA is observed even in the absence of [BP]dG as the duplex is partially unwound to an A-form-like conformation as   (20,58). The loss of interactions between the DNA duplex and polymerase are accompanied by extensive distortion to both the template and primer strands. Displacement of the primer strand results in lost coordination of the 3Ј-OH by Asp-830, an interaction that is central to the chemistry of the 3Ј to 5Ј nucleotidyl transfer reaction. In the rare instance that replication proceeds, [BP]dG might adopt an alternate conformation such as the syn conformation that has been shown in molecular modeling studies with the structurally homologous T7 DNA polymerase to be compatible with bypass of [BP]dG (59 -61). The nature and extent of [BP]dG-induced distortions have not previously been observed in structures of DNA polymerases bound to other types of DNA damage, such as DNA mismatches (38) or the bulky aminofluorene (37) and [(ϩ)-cis-[BP]-adenine adducts (43). Collectively these distortions to the polymerase and DNA duplex account for why DNA adducts that protrude into the minor groove block replication by high fidelity DNA polymerases.
The structural data presented here along with accompanying in vitro primer extension assays strongly suggest that upon encountering [BP]dG, a high fidelity polymerase dissociates from the BP-modified template in preference to replication past the lesion. By contrast, the error-prone Y family polymerases replicate past several types of DNA damage (62,63) and have been shown structurally to accommodate different types of DNA damage, including the (ϩ)-cis-[BP]-adenine adduct (43), an abasic lesion (64), and a cis-syn thymine dimer (65). In addition, molecular modeling studies suggest that the Y family DNA polymerase Dpo4 can accommodate [BP]dG in either an anti or syn conformation (61). The conformation of the template base is not regulated by Y family polymerases in the same manner as high fidelity polymerases because the active site of Y family polymerases is more open and capable of accommodating more than one base pair at a time (25). Bypass of minor groove adducts such as [BP]dG has been observed in solution for various Y-family polymerases (11-15, 62, 63, 67). In the case of one Y-family polymerase, polymerase , a template adenine is preferentially driven into a syn conformation, thereby promoting Hoogsteen base-pairing (66) and possibly replication past minor groove adducts (67). Although the template base has also been shown to adopt a syn conformation in T7 DNA polymerase and BF when the much smaller lesion 8oxoguanine pairs opposite adenine (36,57), the bulk of the [BP]dG and other steric constraints imposed by the high fidelity polymerase active site make translesion replication of [BP]dG less favorable for BF than for the error-prone polymerase .
In general, structures of different types of DNA damage in the context of polymerase active sites suggest that DNA lesions localized in the solvent-exposed major groove are better tolerated than lesions placed in the minor groove that interact with the polymerase surface. Structures of the bulky [BP]dG adduct placed in the minor groove at the BF active site reveal how disruption of the interface between the polymerase and the nascent minor groove of this DNA duplex results in distortions that block further DNA synthesis even in the rare instance that incorporation opposite [BP]dG does occur. Loss of protein-DNA interactions and steric block to dNTP incorporation at the insertion site reveal how steric constraints imposed by the active site of a high fidelity polymerase present an architecture that is clearly incompatible with DNA replication of certain types of damaged DNA.