Structural and dynamic characterization of polymerase κ's minor groove lesion processing reveals how adduct topology impacts fidelity.

DNA lesion bypass polymerases process different lesions with varying fidelities, but the structural, dynamic, and mechanistic origins of this phenomenon remain poorly understood. Human DNA polymerase κ (Polκ), a member of the Y family of lesion bypass polymerases, is specialized to bypass bulky DNA minor groove lesions in a predominantly error-free manner, by housing them in its unique gap. We have investigated the role of the unique Polκ gap and N-clasp structural features in the fidelity of minor groove lesion processing with extensive molecular modeling and molecular dynamics simulations to pinpoint their functioning in lesion bypass. Here we consider the N(2)-dG covalent adduct derived from the carcinogenic aromatic amine, 2-acetylaminofluorene (dG-N(2)-AAF), that is produced via the combustion of kerosene and diesel fuel. Our simulations reveal how the spacious gap directionally accommodates the lesion aromatic ring system as it transits through the stages of incorporation of the predominant correct partner dCTP opposite the damaged guanine, with preservation of local active site organization for nucleotidyl transfer. Furthermore, flexibility in Polκ's N-clasp facilitates the significant misincorporation of dTTP opposite dG-N(2)-AAF via wobble pairing. Notably, we show that N-clasp flexibility depends on lesion topology, being markedly reduced in the case of the benzo[a]pyrene-derived major adduct to N(2)-dG, whose bypass by Polκ is nearly error-free. Thus, our studies reveal how Polκ's unique structural and dynamic properties can regulate its bypass fidelity of polycyclic aromatic lesions and how the fidelity is impacted by lesion structures.

High-fidelity DNA polymerases (Pol) are usually blocked by bulky DNA lesions.[1,2] They can be replaced for local translesion synthesis to bypass the distorting DNA damage by one or more bypass polymerases, followed by reston class="Species">ratioclass="Chemical">n of the processive aclass="Chemical">nd faithful replicative machiclass="Chemical">nery.[3−6] Each of the four class="Chemical">n class="Species">human Y family bypass polymerases has unique structural and lesion bypass properties while also possessing common features.[5,7] Human DNA Polκ is similar to other Y family lesion bypass polymerases.[3] It has fingers, palm, and thumb subdomains that comprise the conserved catalytic core, and the DNA substrate is located between the thumb and the little finger or polymerase-associated domain (PAD).[8] In Polκ, there is a uniquely large structural gap on the minor groove side that separates the catalytic core and the little finger, and a unique N-terminal extension (N-clasp) on the DNA major groove side that holds the little finger and the palm/fingers of the catalytic core together (Figure 1A). The functional role of this gap and the N-clasp in lesion bypass by Polκ has attracted considerable interest,[9−11] because Polκ bypasses minor groove bulky DNA lesions that are linked to the amino group of guanine with predominant incorporation of the correct nucleotide dCTP; however, some mutagenic outcome that varies among different lesions is observed.[12−20] Furthermore, adducts that bind to adenine N6 or guanine C8 on the major groove side are mainly blocking or more mutagenic.[12,17,21−26]

Figure 1

Structures and sequences investigated. (A) Ternary crystal structure of Polκ (PDB entry 2OH2).[8] The nucleotide at the preinsertion position is hidden behind the fingers domain, and its location is designated with a black frame. (B) Structures of the dG-N2-AAF and dG-N2-B[a]P adducts. Torsion angles for dG-N2-AAF are defined as follows: χ for O4′ (dR)–C1′ (dR)–N9–C4 (dR is deoxyribose), α′ for N1–C2–N2–C3(AAF), β′ for C2–N2–C3(AAF)–C2(AAF), γ′ for C3(AAF)–C2(AAF)–N(AAF)–C(AAF), δ′ for C2(AAF)–N(AAF)–C(AAF)–Cm(AAF), and ε′ for N(AAF)–C(AAF)–Cm(AAF)–Hm(AAF) (m denotes a methyl group). For the dG-N2-B[a]P adduct, the ring containing the OH groups is termed the benzylic ring. (C) Base sequences of the preinsertion and insertion models. The incoming nucleotide dNTP is colored blue. G* denotes the damaged guanine.

Structures and sequences investigated. (A) Ternary crystal structure of Polκ (n class="Gene">PDB eclass="Chemical">ntry 2OH2).[8] The class="Chemical">nucleotide at the preiclass="Chemical">nsertioclass="Chemical">n positioclass="Chemical">n is hiddeclass="Chemical">n behiclass="Chemical">nd the ficlass="Chemical">ngers domaiclass="Chemical">n, aclass="Chemical">nd its locatioclass="Chemical">n is desigclass="Chemical">nated with a black frame. (B) Structures of the class="Chemical">n class="Chemical">dG-N2-AAF and dG-N2-B[a]P adducts. Torsion angles for dG-N2-AAF are defined as follows: χ for O4′ (dR)–C1′ (dR)–N9–C4 (dR is deoxyribose), α′ for N1–C2–N2–C3(AAF), β′ for C2–N2–C3(AAF)–C2(AAF), γ′ for C3(AAF)–C2(AAF)–N(AAF)–C(AAF), δ′ for C2(AAF)–N(AAF)–C(AAF)–Cm(AAF), and ε′ for N(AAF)–C(AAF)–Cm(AAF)–Hm(AAF) (m denotes a methyl group). For the dG-N2-B[a]P adduct, the ring containing the OH groups is termed the benzylic ring. (C) Base sequences of the preinsertion and insertion models. The incoming nucleotide dNTP is colored blue. G* denotes the damaged guanine.

At present, the structural and dynamic origins of the various outcomes in lesion processing by DNA Polκ are poorly understood. The size and shape of bulky polycyclic aromatic DNA lesions, their site of linkage to different bases, and their stereochemical properties determine their handling by lesion bypass DNA polymerases[12,17,21−24] and other macromolecular machines, such as DNA repair enzymes.[27] Our understanding of the relationships between lesion and polymerase architectures and functions is limited. However, this knowledge is crucially needed to identify the molecular mechanisms underlying lesion-induced mutagenesis. In earlier work, we have shown that the major adduct derived from the environmental procarcinogen n class="Chemical">benzo[a]pyrene (B[a]P), (10S)-(+)-traclass="Chemical">ns-aclass="Chemical">nti-B[a]P-class="Chemical">n class="Chemical">N2-dG (dG-N2-B[a]P),[28] is easily accommodated on the minor groove side of the damaged template in Polκ while the (10S)-(+)-trans-anti-B[a]P-N6-dA adduct on the major groove side is sterically constrained by the Polκ N-clasp,[10,11] explaining their respective observed bypass and blocking properties in primer extension studies.[17]

To pursue the goal of investigating impacts of lesion topology and stereochemistry on the fidelity of lesion processing by Polκ, we investigate an adduct derived from the n class="Disease">carcinogenic class="Chemical">n class="Chemical">aromatic amine 2-acetylaminofluorene (AAF), 3-(deoxyguanosin-N2-yl)-2-acetylaminofluorene (dG-N2-AAF or dG*) (Figure 1B). This lesion shares its N2-dG linkage site with dG-N2-B[a]P but is less bulky with three rather than four aromatic rings and no chiral center at the linkage site. Moreover, it is planar, while dG-N2-B[a]P contains the nonplanar and OH-bearing benzylic ring (Figure 1B). However, the AAF adduct contains the bulky acetyl group. The AAF lesion can be produced by exposure to the environmental contaminant 2-nitrofluorene, a major byproduct of kerosene and diesel combustion.[29,30] The adduct is housed in the B-DNA minor groove in solution where it increases the thermal and thermodynamic stability of the B-DNA duplex.[31] It is persistent in mammalian tissues,[32−34] consistent with its thermodynamic lesion-induced DNA duplex stabilization, which is associated with nucleotide excision repair resistance.[27,35] However, protection against transcriptional errors by this lesion via transcription-coupled nucleotide excision repair[36−38] has recently been demonstrated,[39] indicating that its persistence results from resistance to global genomic nucleotide excision repair.[40,41] Primer extension studies show that Polκ bypasses the dG-N2-AAF lesion with predominant incorporation of dCTP but also with significant misincorporation of dTTP in vitro.[18] The lesion’s persistence and its mutagenicity in mammalian cells suggest that this minor AAF-derived adduct can readily survive to replication and cause cancer-initiating mutations; in simian kidney (COS-7) cells, G → T transversions were the predominant observed targeted mutations, but bypass was mainly error-free.[18]

In this study, we have characterized structural and dynamic factors governing the fidelity of processing of n class="Chemical">dG-N2 miclass="Chemical">nor groove adducts by Polκ. With exteclass="Chemical">nsive molecular modeliclass="Chemical">ng aclass="Chemical">nd molecular dyclass="Chemical">namics simulatioclass="Chemical">ns, we have gaiclass="Chemical">ned a detailed molecular aclass="Chemical">nd dyclass="Chemical">namic characterizatioclass="Chemical">n of class="Chemical">n class="Chemical">dG-N2-AAF lesion bypass through stages of the replicative process as the lesion transits from the preinsertion to the insertion position and then to the chemical transition state for the nucleotidyl transfer reaction.[10] We provide a new understanding of Polκ’s structural architecture for bypass of minor groove lesions by revealing how the flexibility of its N-clasp and its housing of lesions in the minor groove gap are impacted differently by the structurally dissimilar dG-N2-AAF and dG-N2-B[a]P lesions to yield different balances of mutagenic and nonmutagenic outcomes. Thereby, we show how the fidelity of the human Polκ bypass polymerase is regulated by lesion topology.

Methods

Models of Polκ with Unmodified DNA

The ternary crystal structure (n class="Gene">PDB eclass="Chemical">ntry 2OH2(42)) of Polκ19–526 with DNA aclass="Chemical">nd iclass="Chemical">ncomiclass="Chemical">ng class="Chemical">n class="Gene">dTTP[8] was the basis for all our prepared enzyme–substrate models. The sequence utilized, shown in Figure 1C, was the same as in the crystal except that it was remodeled in previous work to provide a template G and incoming dCTP; the crystal structure had also been remodeled to create a reaction-ready active site with two Mg2+ ions,[10] based on a high-resolution crystal structure of a complete DNA polymerase β catalytic complex.[43] With 100 ns MD, we utilized this unmodified ternary complex as a control for the dG-N2-AAF simulations.

For simulations involving the pentacovalent n class="Chemical">phosphorane traclass="Chemical">nsitioclass="Chemical">n state iclass="Chemical">n the class="Chemical">n class="Chemical">water-mediated and substrate-assisted (WMSA) mechanism (Figure S1 of the Supporting Information), we utilized our previously obtained transition state structure determined by ab initio quantum mechanics/molecular mechanics (QM/MM) simulations.[10]

Initial Models with the dG-N2-AAF Adduct for 100 ns MD Simulations

We utilized the remodeled, unmodified Polκ structure to obtain models of the n class="Chemical">dG-N2-AAF lesioclass="Chemical">n iclass="Chemical">n the preiclass="Chemical">nsertioclass="Chemical">n aclass="Chemical">nd iclass="Chemical">nsertioclass="Chemical">n positioclass="Chemical">ns (Figure 1C). The class="Chemical">nuclear magclass="Chemical">netic resoclass="Chemical">naclass="Chemical">nce (NMR) solutioclass="Chemical">n structure (class="Chemical">n class="Gene">PDB entry 2GE2(31)) provided initial guidance for modeling of the lesion with minimal collisions. For the very crowded preinsertion site, we could only create a model in which only the fluorenyl ring system fits in the small cavity at the apex of the gap; here the lesion can be housed on the damaged guanine’s evolving major groove side. The minor groove side of the evolving duplex was entirely obstructed by the fingers (Figure 1A). Table S1 of the Supporting Information gives linkage site torsion angle values for α′, β′, γ′, δ′, ε′, and χ (Figure 1B) and Met115 (whose side chain also had to be remodeled) for this initial model for MD.

For the insertion site, we modeled two orientations for the n class="Chemical">dG-N2-AAF adduct. Iclass="Chemical">n the NMR solutioclass="Chemical">n structure, the fluoreclass="Chemical">nyl riclass="Chemical">ng system is directed 5′ aloclass="Chemical">ng the damaged straclass="Chemical">nd iclass="Chemical">n the miclass="Chemical">nor groove. Iclass="Chemical">n additioclass="Chemical">n, computatioclass="Chemical">nal studies[44] had showclass="Chemical">n that a 3′ orieclass="Chemical">ntatioclass="Chemical">n was also feasible, aclass="Chemical">nd we iclass="Chemical">nvestigated this orieclass="Chemical">ntatioclass="Chemical">n, as well. Usiclass="Chemical">ng the NMR solutioclass="Chemical">n structure as a guide, we covaleclass="Chemical">ntly liclass="Chemical">nked the class="Chemical">n class="Chemical">AAF moiety to the template anti G amino group in the unmodified ternary complex model and adjusted torsion angles α′, β′, γ′, δ′, ε′, and χ (Figure 1B) to obtain 3′- and 5′-oriented structures; this also required a change in a side chain torsion angle of Phe171. We considered both incoming dCTP and dTTP. Table S1 of the Supporting Information gives linkage site torsion angle values for α′, β′, γ′, δ′, ε′, χ, and Phe171 for these initial models for MD. For dCTP, we maintained Watson–Crick pairing as in the unmodified model.[10] For the mismatched dTTP, we created a wobble pair with dG*:dTTP paired with N1H1(G)···O2(T) and O6(G)···N3H3(T) hydrogen bonds (Figure S2 of the Supporting Information). For the study of the transition state, we remodeled our previously obtained transition state structure determined by a QM/MM–MD investigation[10] to contain the lesion in its preferred 3′ orientation in the ternary complex before reaction (see Results). The unmodified transition state structure was used for a control simulation. All initial models were prepared using INSIGHT II 2005 (Accelrys Software, Inc.).

Force Fields and Molecular Dynamics

MD simulations were conducted using the PMEMD module of the AMBER 11 simulation package[45] with Amber99SB[46−48] and GAFF[49] force fields and the TIP3P n class="Chemical">water model.[50] A class="Chemical">new AMBER-compatible force field set was developed for the class="Chemical">n class="Chemical">dG-N2-AAF adduct and is given in Table S2 of the Supporting Information. For dCTP and dTTP, we utilized previously calculated parameters.[51] For the transition state models, we utilized our previously developed parameters[10] for dCTP as the incorporated nucleotide, and we developed a new parameter set for the transition state model for dTTP as the incorporated nucleotide (Table S2 of the Supporting Information).

For each model, a 100 ns MD simulation was performed after initial equilibn class="Species">ratioclass="Chemical">n. The last sclass="Chemical">napshot of the simulatioclass="Chemical">n was used to illustclass="Chemical">n class="Species">rate the structures, except where indicated. All analyzed properties are ensemble averages over the entire 100 ns MD simulations. Trajectory analyses were conducted with the Ptraj and Carnal packages of Amber 11. Full details of the MD and force field protocols are given in the Supporting Information. PyMOL[52] was employed to make molecular images and the movies.

Results

In Polκ, we have investigated the n class="Chemical">dG-N2-AAF adduct iclass="Chemical">n three positioclass="Chemical">ns: (1) the preiclass="Chemical">nsertioclass="Chemical">n site to determiclass="Chemical">ne how feasible it is for the adduct to traclass="Chemical">nslocate from the preiclass="Chemical">nsertioclass="Chemical">n site to the iclass="Chemical">nsertioclass="Chemical">n site, (2) the iclass="Chemical">nsertioclass="Chemical">n site opposite class="Chemical">n class="Gene">dCTP and dTTP (Figure S2 of the Supporting Information), and (3) the pentacovalent phosphorane transition state, which is formed during the nucleotidyl transfer reaction in the WMSA mechanism previously determined by QM/MM–MD calculations for Polκ (Figure S1 of the Supporting Information).[10,53] We have used molecular modeling and molecular dynamics simulations (detailed in Methods) to gain an understanding on a molecular and dynamic level of the cycle of events in the Polκ active site and thereby obtain insights into the fidelity of Polκ’s processing of minor groove polycyclic aromatic lesions.

The Adduct in the Preinsertion Site Can Be Housed in the Crevice at the Apex of the Gap

Our modeling for n class="Chemical">dG-N2-AAF iclass="Chemical">n the preiclass="Chemical">nsertioclass="Chemical">n site showed that the adduct caclass="Chemical">n fit oclass="Chemical">nly oclass="Chemical">n the major groove side poiclass="Chemical">nticlass="Chemical">ng iclass="Chemical">n the 5′ directioclass="Chemical">n of the damaged straclass="Chemical">nd iclass="Chemical">n a crevice at the apex of the gap, while the miclass="Chemical">nor groove side is crowded by the ficlass="Chemical">ngers (Figure 1A). This positioclass="Chemical">niclass="Chemical">ng is preserved followiclass="Chemical">ng 100 class="Chemical">ns MD as showclass="Chemical">n iclass="Chemical">n Figure 2A. Eclass="Chemical">nsemble average torsioclass="Chemical">n aclass="Chemical">ngle values for α′, β′, γ′, δ′, ε′, aclass="Chemical">nd χ (Figure 1B) aclass="Chemical">nd Met115 are listed iclass="Chemical">n Table S1 of the Supporticlass="Chemical">ng Iclass="Chemical">nformatioclass="Chemical">n. The glycosidic torsioclass="Chemical">n of the damaged class="Chemical">n class="Chemical">guanine is very dynamic but remains largely in the overall anti domain (Figure S2 of the Supporting Information)

Figure 2

Lesion structures in the Polκ gap at preinsertion and insertion positions. Polκ containing the dG-N2-AAF adduct (A) in the preinsertion position and (B) in the insertion position opposite dCTP. dG-N2-AAF in the insertion site exhibits favorable stacking interactions with Phe171 as shown in the inset. (C) As the damaged base translocates through the gap in Polκ from the preinsertion site (dG-N2-AAF in red sticks) to the insertion site (dG-N2-AAF in cyan sticks), the fluorenyl rings rotate ∼180° around the α′ torsion angle (see Movie S1 of the Supporting Information). The part of the protein that comprises the gap region through which translocation occurs is shown as yellow spheres. (D) Polκ containing the dG-N2-B[a]P adduct in the insertion position. Color scheme for panels A, B, and D: dG-N2-AAF and dG-N2-B[a]P as red sticks, damaged guanine as orange sticks, N-clasp as a blue surface, thumb domain as a magenta surface, fingers domain as a yellow surface, and DNA template strand as a gray cartoon.

Lesion structures in the Polκ gap at preinsertion and insertion positions. Polκ containing the n class="Chemical">dG-N2-AAF adduct (A) iclass="Chemical">n the preiclass="Chemical">nsertioclass="Chemical">n positioclass="Chemical">n aclass="Chemical">nd (B) iclass="Chemical">n the iclass="Chemical">nsertioclass="Chemical">n positioclass="Chemical">n opposite class="Chemical">n class="Gene">dCTP. dG-N2-AAF in the insertion site exhibits favorable stacking interactions with Phe171 as shown in the inset. (C) As the damaged base translocates through the gap in Polκ from the preinsertion site (dG-N2-AAF in red sticks) to the insertion site (dG-N2-AAF in cyan sticks), the fluorenyl rings rotate ∼180° around the α′ torsion angle (see Movie S1 of the Supporting Information). The part of the protein that comprises the gap region through which translocation occurs is shown as yellow spheres. (D) Polκ containing the dG-N2-B[a]P adduct in the insertion position. Color scheme for panels A, B, and D: dG-N2-AAF and dG-N2-B[a]P as red sticks, damaged guanine as orange sticks, N-clasp as a blue surface, thumb domain as a magenta surface, fingers domain as a yellow surface, and DNA template strand as a gray cartoon.

In the Insertion Site, the Lesion Is Housed in the Polκ Gap Directed 3′ along the Modified Strand toward the Duplex Region

From our modeling efforts, we obtained two initial conformations for MD simulations that comfortably accommodate the fluorenyl rings in the Polκ gap, between the fingers and the palm on one side and the little finger on the other side, and with a n class="Gene">dCTP partclass="Chemical">ner: (1) with the class="Chemical">n class="Chemical">AAF ring system oriented toward the single-stranded overhang and directed 5′ along the modified strand, as in the NMR solution structure in duplex DNA,[31] and (2) with the AAF ring system directed toward the duplex region in a 3′ orientation, as predicted computationally.[44] Ensemble average linkage site torsion angle values are summarized in Table S1 of the Supporting Information. The Watson–Crick pairing was retained in the MD simulations of both orientations. Also, the active site is as well organized in the structures containing lesions as in the unmodified control; key properties include maintenance of the octahedral coordination of the Mg2+ ions, the in-line O3′–Pα attack distance, maintenance of interactions with key amino acid residues, and preservation of the two water molecules participating in the WMSA mechanism.[10] In this respect, both MD models appear equally feasible. The dG*:dCTP (dG* denotes the damaged base) model in the 3′ orientation of the fluorenyl rings is illustrated in Figures 2B and 3A, and full details are given in Figure S3 and Table S3 of the Supporting Information. The active site organization of the 5′-oriented model is similar (data not shown). However, there are important differences. In the 5′-directed case, the fluorenyl distal aromatic ring is entirely solvent-exposed on one face and steric crowding causes the distortion of the fingers (Figure S4 of the Supporting Information), while in the 3′-directed orientation, the entire aromatic ring system is neatly sandwiched in the gap (Figure 2B). In addition, the 3′-directed orientation has a favorable dynamic stacking interaction with Phe171 that is not present for the 5′ case, where the ring system is directed away from Phe171. Figure 2B and Figure S4 of the Supporting Information show the differences in these interactions. For these reasons, the 5′ orientation is disfavored.

Figure 3

Watson–Crick and wobble pairing for dG-N2-AAF. (A) dG-N2-AAF Watson–Crick pair with dCTP. (B) dG-N2-AAF wobble pair with the dTTP mismatch. Methyl groups of Ala150 and Ala151 from the fingers domain have van der Waals interactions with the dTTP methyl group. The shortest distances between methyl hydrogen atoms are given. Dashed lines denote hydrogen bonds, with occupancies all above 95%, given in Table S3 of the Supporting Information. Snapshots at 92 and 86 ns were selected as being representative for panels A and B, respectively.

Watson–Crick and wobble pairing for n class="Chemical">dG-N2-AAF. (A) class="Chemical">n class="Chemical">dG-N2-AAF Watson–Crick pair with dCTP. (B) dG-N2-AAF wobble pair with the dTTP mismatch. Methyl groups of Ala150 and Ala151 from the fingers domain have van der Waals interactions with the dTTP methyl group. The shortest distances between methyl hydrogen atoms are given. Dashed lines denote hydrogen bonds, with occupancies all above 95%, given in Table S3 of the Supporting Information. Snapshots at 92 and 86 ns were selected as being representative for panels A and B, respectively.

The Pentacovalent Transition State Structure Remains Stable with a 3′-Oriented Lesion

For investigation of the transition state structure in the WMSA mechanism (Figure S1 of the Supporting Information),[10,53] we used the favorable 3′ lesion orientation with n class="Gene">dCTP to establish whether the peclass="Chemical">ntacovaleclass="Chemical">nt class="Chemical">n class="Chemical">phosphorane and the local organization were maintained in the presence of the dG-N2-AAF adduct. Our results of the 100 ns MD simulation showed that the transition state structure is well-maintained (Figure 4 and Figure S3 and Table S3 of the Supporting Information): the geometry of the pentacovalent phosphorane transition state, the geometry of the protonated γ-phosphate ready to transfer its proton to the α,β-oxygen bridge, hydrogen bonding interactions with amino acids that stabilize the transition state, Mg2+ ion coordination, and base pairing between the template and the incoming dCTP.

Figure 4

Pentacovalent phosphorane transition state with dCTP incorporated into the dC priming nucleotide. Polκ’s active site at the transition state maintains the octahedral coordination of the two Mg2+ ions (shown with the blue dashed lines) and the water molecules utilized in the WMSA mechanism[10] to shuttle the proton from the γ-phosphate to the α,β-bridge (yellow dashed lines) as pyrophosphate leaves.

Pentacovalent n class="Chemical">phosphorane traclass="Chemical">nsitioclass="Chemical">n state with class="Chemical">n class="Gene">dCTP incorporated into the dC priming nucleotide. Polκ’s active site at the transition state maintains the octahedral coordination of the two Mg2+ ions (shown with the blue dashed lines) and the water molecules utilized in the WMSA mechanism[10] to shuttle the proton from the γ-phosphate to the α,β-bridge (yellow dashed lines) as pyrophosphate leaves.

Wobble Pairing Stabilizes an Incoming dTTP Mismatch

While incorpon class="Species">ratioclass="Chemical">n of class="Chemical">n class="Gene">dCTP opposite dG-N2-AAF predominated in the primer extension studies in Polκ, a significant amount of dTTP was also misincorporated.[18] To determine how a dG*:dTTP mismatch is accommodated in the polymerase active site, we created an initial model with anti-dG-N2-AAF in the favored 3′ orientation and wobble paired with incoming dTTP (Figure S2 of the Supporting Information). Following a 100 ns MD simulation, we found that a nicely aligned, well-hydrogen-bonded wobble pair was maintained for the reactant state (Figure 3B), and also for the subsequent 100 ns transition state simulation (Figure S5 of the Supporting Information). The dynamic stacking interaction between Phe171 and the AAF ring system is present as in the case of the Watson–Crick base pair. In addition, an analysis of the active site region of Polκ showed the presence of Ala residues 150 and 151, whose methyl groups have favorable van der Waals interactions with the methyl group of thymine, which seems to aid in stabilizing the incorporation of dTTP (Figure 3B and Figure S5 of the Supporting Information). Both simulations maintained well-organized active sites; details are given in Figure S3 and Table S3 of the Supporting Information. Ensemble average linkage site torsion angle values are listed in Table S1 of the Supporting Information.

Translocation of the Adduct from the Preinsertion Site to the Insertion Site through the Gap Is Feasible

Evaluating the MD structures of the preinsertion and insertion sites to consider translocation to the insertion site, we found that the adduct would need to rotate around α′ (Figure 1B) as the strand translocates; α′ adopts a value of 19.6 ± 12.3° (range of 8.7–30.4°) in the preinsertion site and 150.6 ± 45.0° (range of 98.1–160.9°) in the insertion site opposite n class="Gene">dCTP iclass="Chemical">n its 3′ favored orieclass="Chemical">ntatioclass="Chemical">n (Figure S2 aclass="Chemical">nd Movie S1 of the Supporticlass="Chemical">ng Iclass="Chemical">nformatioclass="Chemical">n). The large staclass="Chemical">ndard deviatioclass="Chemical">ns aclass="Chemical">nd raclass="Chemical">nges iclass="Chemical">ndicate opportuclass="Chemical">nities for sigclass="Chemical">nificaclass="Chemical">nt coclass="Chemical">nformatioclass="Chemical">nal flexibility, particularly iclass="Chemical">n the iclass="Chemical">nsertioclass="Chemical">n site. This is seeclass="Chemical">n iclass="Chemical">n Figure 2B, which shows the aromatic riclass="Chemical">ng system housed iclass="Chemical">n the gap aclass="Chemical">nd orieclass="Chemical">nted iclass="Chemical">n the 3′ directioclass="Chemical">n of the template straclass="Chemical">nd, toward the palm. The gap is sufficieclass="Chemical">ntly spacious to allow the bulky adduct oclass="Chemical">n the major groove side iclass="Chemical">n the preiclass="Chemical">nsertioclass="Chemical">n positioclass="Chemical">n to traclass="Chemical">nslocate to the miclass="Chemical">nor groove side iclass="Chemical">n the iclass="Chemical">nsertioclass="Chemical">n positioclass="Chemical">n, as showclass="Chemical">n iclass="Chemical">n Figure 2C aclass="Chemical">nd Movie S1 of the Supporticlass="Chemical">ng Iclass="Chemical">nformatioclass="Chemical">n.

MD Simulations Show That the N-Clasp Is Flexible

We observed flexibility of the N-clasp domain. This dynamic appeared along the MD trajectories of unmodified (Movie S2 of the Supporting Information) and modified (Movies S3 and S4 of the Supporting Information) models, but to a greater extent in the models with the n class="Chemical">AAF-damaged templates. Of special iclass="Chemical">nterest is the regioclass="Chemical">n of the tether betweeclass="Chemical">n αN1 aclass="Chemical">nd αclass="Chemical">n class="Chemical">N2 that dynamically elongates and shortens because of the unwinding of primarily the αN1 helix. This dynamic allows conformational flexibility in the template strand, which likely facilitates the wobble pairing with dTTP for the dG-N2-AAF adduct (Movie S4 of the Supporting Information). However, as discussed below for the dG-N2-B[a]P adduct, lesion topology significantly impacts this flexibility.

Discussion

Lesion Housing in the Minor Groove Gap for Bypass by Polκ

Our modeling and molecular dynamics studies show that the n class="Chemical">dG-N2-AAF lesioclass="Chemical">n is well-accommodated iclass="Chemical">n the Polκ gap opposite class="Chemical">n class="Gene">dCTP. We investigated this minor groove lesion at the insertion site, directed either 5′ or 3′ along the modified strand in the gap, based on its NMR solution structure[31] and computational modeling studies.[44] However, the 3′ orientation fits distinctly better in the gap (Figure 2B and Figure S4 of the Supporting Information). The fluorenyl rings are neatly sandwiched between the fingers/palm and the little finger, forming van der Waals contact with the fingers/palm domains on one face, and are well-protected from solvent. Also, the adduct is well-aligned with the damaged strand, and the distal aromatic ring stacks dynamically with Phe171. In contrast, in the 5′ orientation, the fluorenyl rings crowd the fingers domain and deform the gap, and they are not positioned for stacking with Phe171 (Figure S4 of the Supporting Information). Moreover, the distal aromatic ring protrudes from the DNA double helix, extruding from the protein surface, and thus is solvent-exposed. Therefore, the 3′-oriented conformer is clearly favorable. Furthermore, Watson–Crick pairing and all other hallmarks of the reaction-ready state are preserved, including the water molecules that participate in the WMSA mechanism (Figure S1 of the Supporting Information), which retain positions to permit shuttling of the O3′ proton to the γ-phosphate. In addition, simulations of the transition state also show the same stability (Figure 4 and Figure S3 and Table S3 of the Supporting Information). Only two Mg2+ ions[54] are needed to stabilize the transition state in the low-energy reaction path of the WMSA mechanism,[10,53] but recent structural evidence indicates that a third metal ion can aid in charge stabilization as the reaction proceeds in Polη and Polβ.[55−57] Future work will provide further insights into the roles of the metal ions in nucleotidyl transfer with various polymerases.

Our models of the preinsertion site containing the n class="Chemical">dG-N2-AAF lesioclass="Chemical">n show that its fluoreclass="Chemical">nyl riclass="Chemical">ngs caclass="Chemical">n be well-housed iclass="Chemical">n a crevice of the gap, oclass="Chemical">n the major groove side of the evolviclass="Chemical">ng duplex (Figure 2A). Notably, we show that traclass="Chemical">nslocatioclass="Chemical">n from the preiclass="Chemical">nsertioclass="Chemical">n to the iclass="Chemical">nsertioclass="Chemical">n site via rotatioclass="Chemical">n of α′ (Figure S2 of the Supporticlass="Chemical">ng Iclass="Chemical">nformatioclass="Chemical">n) is uclass="Chemical">nobstructed through the gap betweeclass="Chemical">n the ficlass="Chemical">ngers/palm aclass="Chemical">nd little ficlass="Chemical">nger domaiclass="Chemical">ns (Figure 2C aclass="Chemical">nd Movie S1 of the Supporticlass="Chemical">ng Iclass="Chemical">nformatioclass="Chemical">n). Thus, our results provide structural aclass="Chemical">nd dyclass="Chemical">namic explaclass="Chemical">natioclass="Chemical">ns for the observed preferred iclass="Chemical">ncorpoclass="Chemical">n class="Species">ration of dCTP for this adduct in Polκ. Our models suggest that other polycyclic aromatic lesions, such as the dG-N2-B[a]P adduct, could be similarly accommodated in the preinsertion site and translocate through the gap to the insertion site.

dG:dTTP Mismatches Are Facilitated by N-Clasp Flexibility for Stable Wobble Pairing

Experimental results showed that n class="Gene">dTTP is misiclass="Chemical">ncorpoclass="Chemical">n class="Species">rated to a significant extent opposite dG-N2-AAF;[18] our models and MD simulations showed that an anti-dG*:dTTP wobble pair was well-accommodated at the insertion site with maintenance of waters and reaction-ready geometry, and that the transition state structure was also preserved (Figure S3 and Table S3 of the Supporting Information). We examined the active site region to gain insight into the preference for dTTP over other mismatches and noted a pair of Ala residues positioned so that their methyl groups engaged in favorable van der Waals interactions with the thymine methyl group (Figure 3B). The flexible N-clasp and the favorable interactions with Ala may also play a part in the observed preference of Polκ to incorporate dTTP opposite template G above other mismatches[13,18,21] opposite unmodified template dG.

Deletion of a part or the whole N-clasp domain of Polκ reduces the polymerase activity.[8] The N-clasp completes the encircled grip around the primer/template pair at the insertion position, and it was suggested that, by inducing conformational changes from n class="Disease">disordered to ordered states wheclass="Chemical">n biclass="Chemical">ndiclass="Chemical">ng to DNA, it locks all the domaiclass="Chemical">ns of Polκ arouclass="Chemical">nd the DNA.[8] Receclass="Chemical">ntly, Liu et al.[9] highlighted that the N-clasp is esseclass="Chemical">ntial for bridgiclass="Chemical">ng the miclass="Chemical">nor groove gap by formiclass="Chemical">ng vaclass="Chemical">n der Waals iclass="Chemical">nteractioclass="Chemical">ns with the little ficlass="Chemical">nger, but that it is flexible. Our simulatioclass="Chemical">ns revealed the flexibility of the N-clasp domaiclass="Chemical">n: the regioclass="Chemical">n of the tether betweeclass="Chemical">n αN1 aclass="Chemical">nd αclass="Chemical">n class="Chemical">N2 is of particular interest; it dynamically elongated and shortened because of the unwinding of primarily the αN1 helix. Moreover, the dynamics appeared along the MD trajectories of all models (Movies S2–S4 of the Supporting Information), but to a greater extent in the models with AAF-damaged templates (Figure 5 and Movies S3 and S4 of the Supporting Information). This dynamics allows mobility in the template strand, which facilitates the wobble pairing between an incoming dTTP and dG*, because the N-clasp flexibility permits the template base to position itself optimally for wobble pair geometry.

Figure 5

αN1 of the N-clasp domain is more flexible in (A) dG-N2-AAF than in (B) dG-N2-B[a]P. The most representative structures for αN1 throughout the 100 ns trajectory are shown.

αN1 of the N-clasp domain is more flexible in (A) n class="Chemical">dG-N2-AAF thaclass="Chemical">n iclass="Chemical">n (B) class="Chemical">n class="Chemical">dG-N2-B[a]P. The most representative structures for αN1 throughout the 100 ns trajectory are shown.

Lesion Topology Influences N-Clasp Flexibility, Which Impacts Polκ Fidelity

Polκ is also responsible for the near n class="Disease">error-free bypass iclass="Chemical">n vitro(17,58) of the more bulky miclass="Chemical">nor groove class="Chemical">n class="Chemical">dG-N2-B[a]P adduct (Figure 1B).[59] In this case, nucleotide misincorporation is much more rare[17] than for the dG-N2-AAF case.[18] Furthermore, Polκ appears to be utilized in near error-free bypass of the dG-N2-B[a]P minor groove adduct in vivo as well,[58,60−62] with the aid of Polζ for extension.[63] Our previous investigations by modeling and MD methods[10,11] have shown that this lesion is housed in Polκ like the dG-N2-AAF adduct opposite dCTP, with aromatic ring directed 3′ along the modified strand, sandwiched in the gap, protected from solvent, aligned with the DNA template strand, and stacked with Phe171. In this lesion, the 5′ orientation, seen in the NMR solution structure,[59] appears to be much more unfavorable than for the dG-N2-AAF adduct in Polκ, because the four bulkier aromatic rings would offer greater steric hindrance to the finger region of the gap. In addition, we have shown that the pentacovalent phosphorane transition state in the WMSA mechanism for the nucleotidyl transfer reaction (Figure S1 of the Supporting Information) is stable when the B[a]P lesion is present.[10] The placement of the B[a]P ring system in the Polκ gap is supported by the work of Liu et al.[9] These workers designed mouse Polκ variants with a reduced gap size and reported efficient DNA synthesis across from unmodified DNA and blockage by the dG-N2-B[a]P adduct, proving the positioning of the B[a]P ring system in the gap. In their models of Polκ mutants, the lesion is placed in the gap directed 3′ along the template strand, as in our predicted orientation.[10,11]

To explore the flexibility of the N-clasp in the B[a]P lesion, we extended our previous short simulation[11] to 100 ns (Figure 2D) and found that the N-clasp is much less flexible (Movie S5 of the Supporting Information) than for the n class="Chemical">dG-N2-AAF cases (Figure 5). Iclass="Chemical">n this trajectory, we observed a class="Chemical">n class="Chemical">hydrogen bond between the dG-N2-B[a]P nonplanar benzylic ring C8-OH (Figure 1) and the in-chain carbonyl oxygen of Met135 (Figure 6A and Figure S6 and Movie S6 of the Supporting Information). On the other hand, the acetyl group of the dG-N2-AAF adduct is bulky and dynamic, and its carbonyl is repulsive to the carbonyl of Met135 (Figure 6B and Movie S7 of the Supporting Information), leading to a cascade of disturbances between the fingers and the N-clasp via the adjacent Met135. By contrast, the B[a]P hydrogen bond to Met135 provides stability in this region. With the lesser dynamics in the B[a]P case, nucleotide misincorporation through altered hydrogen bonding schemes such as wobble pairing will be less facile.

Figure 6

Impact of lesions on Met135. (A) The dG-N2-B[a]P adduct benzylic ring C8-OH hydrogen bonds with the in-chain carbonyl of Met135. (B) The dG-N2-AAF acetyl group carbonyl is repulsive to the in-chain carbonyl of Met135. The B[a]P and AAF moieties and Met135 are colored by atom: C, green; N, blue; O, red; S, yellow; H, white. The protein domains are colored as in Figure 2A.

Impact of lesions on n class="Chemical">Met135. (A) The class="Chemical">n class="Chemical">dG-N2-B[a]P adduct benzylic ring C8-OH hydrogen bonds with the in-chain carbonyl of Met135. (B) The dG-N2-AAF acetyl group carbonyl is repulsive to the in-chain carbonyl of Met135. The B[a]P and AAF moieties and Met135 are colored by atom: C, green; N, blue; O, red; S, yellow; H, white. The protein domains are colored as in Figure 2A.

Taken together, our results provide new structural and dynamic insights into how the spacious Polκ gap and flexible N-clasp facilitate bypass of minor groove n class="Disease">polycyclic aromatic lesions aclass="Chemical">nd show how lesioclass="Chemical">n topology provides structural aclass="Chemical">nd dyclass="Chemical">namic sigclass="Chemical">nals that determiclass="Chemical">ne the exteclass="Chemical">nt of fidelity iclass="Chemical">n bypass by Polκ. We determiclass="Chemical">ne for the first time how such lesioclass="Chemical">ns caclass="Chemical">n be housed iclass="Chemical">n the preiclass="Chemical">nsertioclass="Chemical">n site aclass="Chemical">nd traclass="Chemical">nslocate through the gap to the iclass="Chemical">nsertioclass="Chemical">n site. Iclass="Chemical">n this positioclass="Chemical">n, we reveal how the disticlass="Chemical">nct features of lesioclass="Chemical">n topology goverclass="Chemical">n the N-clasp’s flexibility, which iclass="Chemical">n turclass="Chemical">n impacts the fidelity of traclass="Chemical">nslesioclass="Chemical">n syclass="Chemical">nthesis. Our results show why the class="Chemical">n class="Chemical">dG-N2-AAF lesion manifests a significant propensity to misincorporate dTTP through wobble pairing facilitated by N-clasp flexibility, while the dG-N2-B[a]P adduct imposes diminished flexibility and is bypassed in a nearly error-free manner by Polκ.[17] More broadly, our work points to the lesion dependence of Polκ’s processing in mutagenic and faithful lesion bypass and hence the degree of protection against mutagenesis by environmental carcinogens that this bypass polymerase can afford.