Introduction

Alkylating agents are common reactive chemicals in the environment (e.g. tobacco smoke)1,2,3 and in cells (e.g. S-adenosylmethionine) that can modify the structures of biological macromolecules by transferring alkyl carbon groups4. DNA bases can be alkylated at the ring nitrogen and extracyclic oxygen to generate a variety of adducts5. N1-methyl-deoxyadenosine (1-MeA) is a mutagenic adduct formed by methylation of deoxyadenosine at N1 (Fig. 1). 1-MeA is particularly pernicious because the N1 atom in adenosine is engaged in Watson-Crick (W-C) base pairing with thymine and its modification by a methyl group impairs W-C base pairing and presents a strong block to normal DNA replication.

Figure 1
figure 1

Chemical structure of adensoine (left) and N1-methyl-deoxyadenosine (right).

Full size image

Cells have evolved a variety of mechanisms to repair alkylated DNA bases6,7,8. This includes the classical multi-step pathways invoking base excision repair (BER), mismatch repair (MMR), and nucleotide excision repair (NER), as well as specific enzymes that can directly dealkylate the bases. Amongst the latter, AlkB in E. coli9,10 and ABH2 in mammals11,12,13 use a mononuclear iron (II) center and cofactors such as 2-ketoglutarate and dioxygen to demethylate the 1-MeA adduct directly8,14. Accordingly, mouse embryonic fibroblast lines derived from ABH2 null mice are found to be highly defective in the repair of 1-MeA adducts15. However, not all 1-MeA are repaired and will be encountered by the replication machinery.

The Y-family of DNA polymerases allow for the continuity of the replication fork by allowing replication through lesions that impede the replicative polymerases16. Humans have four Y-family polymerases – Polι, Polη, Polκ, and Rev1 – each with a unique DNA damage bypass and fidelity profile. Amongst these, Polι stands out in that it does not rely on W-C base pairing between the template base and incoming nucleotide for catalysis. Instead, the active site cleft of Polι is much narrower than in other DNA polymerases, favoring Hoogsteen base pairing17,18. As such, Polι would appear to be well suited to bypass 1-MeA, which has an altered W-C edge but an intact Hoogsteen edge19. Indeed, recent genetic studies in human cells show that translesion synthesis (TLS) across 1-MeA is mediated by three pathways, one of which is dependent on Polι20.

We show here by steady state kinetic analysis that Polι exhibits an ~100 fold higher catalytic efficiency for insertion of the correct nucleotide T relative to the incorrect C opposite 1-MeA. We also present ternary structures of Polι bound to template 1-MeA and incoming dTTP or dCTP. We show that template 1-MeA adopts the syn conformation in both structures, though with significant differences. dTTP and dCTP insert differently opposite template 1-MeA with dTTP participating in Hoogsteen base pairing, while dCTP is largely disordered, consistent with multiple conformations. Together, our kinetic and structural studies show that Polι can not only accommodate lesions such as 1-MeA with impaired W-C edges, but that it can maintain discrimination between correct and incorrect incoming nucleotides opposite the lesion.

Results

Kinetic Analysis

We carried out steady state kinetic analyses to determine the catalytic efficiency (kcat/Km) and fidelity of Polι for nucleotide insertion opposite 1-MeA (Table 1). Polι inserts T opposite 1-MeA with an ~5-fold higher catalytic efficiency than opposite undamaged A; however, it also inserts incorrect nucleotides opposite 1-MeA more efficiently than opposite undamaged A. For example, whereas no insertion of C was detected opposite undamaged A, Polι inserts C opposite 1-MeA with a kcat/Km of 0.1 min−1 μM−1. Importantly, although Polι inserts a C opposite 1-MeA, it does so with a 100 fold lower efficiency than correct T. To understand the ability of Polι to discriminate between T and C opposite 1-MeA, we determined the crystal structures of human Polι bound to a template-primer duplex with 1-MeA as the templating base and dTTP or dCTP as the incoming nucleotide.

Table 1 Steady-state kinetic parameters of nucleotide incorporation opposite templates deoxyadenosine (A) and N1-methyl-deoxyadenosine (1-MeA) by human Polι.
Full size table

Structure Determination

To crystallize Polι with 1-MeA, we used an 18-nt template-primer duplex designed to have two identical replicative ends (Fig. 2) and 1-MeA as the templating base (see Methods). Cocrystals with incoming dTTP or dCTP grow under the same conditions (from PEG solutions) and belong to space group P6522, with nearly identical cell dimensions of a = 98.0 Å, b = 98.0 Å, c = 202.5 Å or 202.2 Å (for incoming dTTP and dCTP respectively) and α = β = 90°, γ = 120° (Table 2). The Polι1-MeA.dTTP and Polι1-MeA.dCTP structures were solved by molecular replacement (MR) using the structure of the PolιA.dTTP complex as a search model (PDB ID: 2FLL)17 with coordinates of the template A and incoming dTTP omitted. Clear electron density was visible for both the template 1-MeA and the incoming dTTP in initial Fo-Fc and simulated annealing Fo-Fc omit maps for Polι1-MeA.dTTP (Fig. 3). The Polι1-MeA.dTTP ternary complex (Rfree of 24.9%, Rcryst of 20.5%) was refined to 2.6 Å resolution (Table 2) and contains Polι residues 26–350, 356–371, 378–397 and 403–414, DNA nucleotides 3–11, incoming dTTP, 1 Mg2+ ion, 1 Cl− ion, and 87 water molecules.

Figure 2
figure 2

Overall structure of (a) Polι1-MeA.dTTP and (b) Polι1-MeA.dCTP ternary complexes. In both complexes, a molecule of Polι is bound to each end of the template-primer duplex. Palm, fingers, and thumb domains, and PAD are shown in cyan, yellow, orange, and green respectively. DNA is shown in tan; template 1-MeA is shown in red. For Polι1-MeA.dTTP , incoming dTTP is shown in red and the Mg2+ ion at site B is shown as a dark blue sphere.

Full size image
Table 2 Data Collection and Refinement Statistics.
Full size table
Figure 3: Comparison of active site in the Polι1-MeA.dTTP and Polι1-MeA.dCTP ternary complexes.
figure 3

(Panels a,d) Close-up view of the active site regions in Polι1-MeA.dTTP and Polι1-MeA.dTCP respectively. The catalytic residues (D34, D126, and E127), residues apposed close to the template base (Q59, K60, L62, V64, L78, S307, and R347), and those near the incoming nucleotide (Y39, T65, Y68, R71, and K214) are highlighted and labeled. Template 1-MeA and incoming dTTP are shown in red. (Panels b,e) Simulated annealing Fo-Fc omit maps (contoured at 3σ) around the templating base and the incoming nucleotide in the structures of Polι1-MeA.dTTP (b) and Polι1-MeA.dTCP (e) respectively. For the incoming dCTP, electron density is very weak with clear density visible only for its γ-phosphate (modeled in orange stick) and poor density visible for the base. (Panel c) 1-MeA.dTTP base pairing in the active site of Polι1-MeA.dTTP ternary complexes. (Panel f) Comparison of 1-MeA…dTTP (green) base pairing in the active site of Polι1-MeA.dTTP with 1-MeA (magenta) in the structure of Polι1-MeA.dCTP. Template 1-MeA in the incoming dCTP structure protrudes into the dNTP binding pocket by 1 Å relative to that in the incoming dTTP structure. In Polι1-MeA.dTTP , D126 interacts with the backbone of L35 and with the Mg2+ ion at site B. In contrast, in the structure of Polι1-MeA.dCTP , D126 remains in the “binary” like conformation.

Full size image

Crystals of Polι1-MeA.dCTP diffracted to 2.0 Å and clear electron density was visible for the templating 1-MeA in the initial Fo-Fc and simulated annealing Fo-Fc omit maps. However, the electron density for the incoming dCTP was not as well-defined as for the dTTP in the Polι1-MeA.dTTP structure, despite the substantially higher resolution (2.0 Å vs. 2.6 Å) of the Polι1-MeA.dCTP structure (Fig. 3e). Strong Fo-Fc density (above 3σ) was visible only for the γ-phosphate of dCTP and partial electron density was observed for what would be considered as the base. To help improve the density, we performed iterative rounds of refinement and water picking with Phenix21 and Coot22. However, no significant improvement in the electron density for the incoming dCTP was observed.

We rationalized that the absence of well defined electron density was suggestive of multiple conformations for the dCTP, with only the γ-phosphate ordered and held in place by interaction with positively charged amino acids from the fingers domain (see below). The Polι1-MeA.dCTP ternary complex was refined to 2.0 Å resolution (Rfree 23.9%; Rcryst of 21.6%) and contains Polι residues 26–350, 356–371, 378–397 and 403–414, DNA nucleotides 4–11, 1 Cl− ion and 291 water molecules.

Overall Arrangement

In both the Polι1-MeA.dTTP and Polι1-MeA.dCTP complexes, a Polι molecule binds to each replicative end of the double-ended template-primer (Fig. 2). The two molecules are related by a crystallographic two-fold axis and thus make identical contacts with the template-primer. Polι has the familiar right-handed architecture with palm (residues 25–37, 99–224), fingers (38–98), and thumb (225–288) domains, and the PAD (polymerase associated domain; residues 298–414) unique to Y-family polymerases17,18,23,24. The palm domain forms the floor of the DNA binding cavity and contains the active site residues (Asp34, Asp126 and Glu127) that catalyze the nucleotidyl transfer reaction, whereas the fingers domain drapes over the template 1-MeA in both structures (and over dTTP in the Polι1-MeA.dTTP structure). The thumb domain and the PAD are connected by a long linker that spans the width of the DNA. The thumb skims the minor groove on one side of the DNA duplex whereas the PAD occupies the major groove on the other side. The majority of Polι-DNA interactions are mediated by the PAD, wherein the main chain amides on “outer” β-strands of the PAD β-sheet make a series of hydrogen bonds with the template and primer strands.

The Polι1-MeA.dTTP ternary complex

The structure reveals how the 1-MeA.dTTP nascent base pair is accommodated in the active site of Polι (Fig. 3a). In previous structures of Polι with template purines (A or G), the steric restraints imposed by the narrow active site of the polymerase are overcome by the template being pushed from the anti into the syn conformation by the incoming dNTP17,18,23. Rotation of the template 1, N6-etheneodeoxyadenosine (εdA) to the syn conformation has also been observed in the structures of Polι with template εdA and incoming dTTP or dCTP19. Template 1-MeA is similarly observed in the syn conformation, presenting its Hoogsteen edge for hydrogen bonding with dTTP which remains in the anti conformation (Fig. 3b and c). The 1-MeA and dTTP bases are almost coplanar and two putative hydrogen bonds are established between the N6 and N7 atoms of 1-MeA with the O4 and N3 atoms of T (2.8 Å and 3.2 Å respectively). The 1-MeA.T base pair is isomorphic with the A.T and εdA.T base pairs in the structures of PolιdA.dTTP and PolιεdA.dTTP, respectively. Superimposition of the Polι1-MeA.dTTP structure with that of PolιdA.dTTP and PolιεdA.dTTP reveals almost perfect overlap between the common N6 and N7 atom of the templating bases and the O4 and N3 atoms of the incoming dTTP.

Incoming dTTP is anchored at one end of the dNTP binding cavity by hydrogen bonding interactions between its γ-phosphate and the side chains of Tyr68 and Arg71 from the fingers domain and Lys214 from the palm domain (Fig. 3a). At the other end, Hoogsteen base pairing with 1-MeA secures the base of dTTP in the binding pocket. The α- and β-phosphates are fixed by interactions with the side chains of Asp126 and Thr65 and with the backbone atoms of Leu35 and Phe38. The dTTP sugar packs against the aromatic ring of Tyr39, and makes a hydrogen bond between its 3′OH and the main chain amide of the Tyr39. A single Mg2+ ion (metal B) is coordinated by the triphosphate moiety of dTTP, as well as the active site residues Asp34 and Asp126. Overall, Polι1-MeA.dTTP is well poised for catalysis with a 3′-OH modeled at the primer terminus located ~3.1 Å from the dTTP α-phosphate and aligned more or less linearly with respect to the scissile Pα-O3′ bond (~162°).

The Polι1-MeA.dCTP ternary complex

Overall, as in the Polι1-MeA.dTTP structure, template 1-MeA is rotated about its glycosidic bond to the syn conformation and presents its Hoogsteen edge to the dNTP binding pocket (Fig. 3d and e). However, relative to the template in the incoming dTTP complex, template 1-MeA in Polι1-MeA.dCTP is inclined towards the DNA helical axis by ~10° and protrudes into the minor groove, partially occluding the dNTP binding site (Fig. 3f). As a result of this inclination, the N6 atom of 1-MeA moves by ~1.0 Å into the dNTP binding cavity relative to the incoming dTTP ternary complex.

Another notable difference between the structure of Polι1-MeA.dCTP and Polι1-MeA.dTTP is in the conformation of the catalytic residue Asp126 (Fig. 3f). In the binary complexes of Polι with template purines, Asp126 is involved in hydrogen bonding interactions with solvent molecules that occupy the vacant dNTP binding pocket, as well as a putative hydrogen bond with 3′OH group modeled at the primer terminus23. In Polι1-MeA.dTTP (and other ternary structures with purine templates), the side chain of Asp126 undergoes an ~30° rotation to establish new interactions with the backbone carbonyl of Leu35, the metal ion at site B, and with the α-phosphate of incoming dNTP (Fig. 3f)17,18,19,23. By contrast, in Polι1-MeA.dCTP, Asp126 remains in the same conformation as in the binary structures.

The electron density for incoming dCTP is very weak with only the density for γ-phosphate clearly visible. The “base” of dCTP is poorly defined (Fig. 3e). Attempts to model dCTP using the conformation of dTTP as a guide leads to steric overlap between its N4 amino group and the N6 amino group of 1-MeA. The dCTP C5′ sugar atom and the α-phosphate also clash with the side chain of Asp126. Efforts to relieve these steric clashes by small movements of the base, sugar and phosphate groups of dCTP lead instead to steric clashes with Val64, Thr65 and Tyr39 from the fingers domain or with the template 1-MeA. These steric clashes are aggravated by the narrow active site of Polι17,18,19,23,25 and protrusion of the template 1-MeA into the dNTP binding pocket by ~1.0 Å. Taken together, the weak electron density for incoming dCTP and the binary like conformation of Asp126 suggest a disordered dCTP with only its γ-phosphate anchored in place while the sugar and base sample a range of conformations.

Discussion

Alkylating agents modify DNA by adding alkyl groups to both the ring nitrogens and the exocyclic oxygen atoms, generating adducts that have cytotoxic effects5,6,8. 1-MeA is a common adduct generated by the transfer of methyl group to the N1 nitrogen atom of deoxyadenosine. If left unrepaired, 1-MeA presents a strong block to replicative polymerases due to its inability to participate in W-C base pairing. We show here by steady state kinetics that Polι, a Y-family polymerase, is capable of TLS across 1-MeA, and that it incorporates the correct T with an ~100 fold higher efficiency than the incorrect C. We also derive a structural framework for the ability of Polι to accommodate the 1-MeA adduct, and a basis for the selection of correct from incorrect incoming nucleotide.

In all previous ternary structures of Polι with a template purine, the template is in syn and dNTP is in anti conformation17,18,19,23. The Polι active site cleft is narrower than in other polymerases, which effectively pushes the template purine into a syn conformation when the incoming dNTP binds. The resulting C1′-C1′ distance across the nascent base pair reduces to <9 Å, favorable for Hoogsteen base pairing. Thus, given the constraints of the Polι active site cleft, it is not surprising that 1-MeA is also pushed in to the syn conformation for Hoogsteen base pairing with incoming dTTP (which remains in the anti conformation). Importantly, the complex is competent for catalysis with the scissile Pα-O3′ bond of incoming dTTP aligned favorably with respect to a 3′OH modeled at the primer terminus.

By contrast, although the dCTP γ-phosphate occupies the same position as the γ-phosphate of dTTP in the Polι1-MeA.dTTP structure, the rest of the molecule is disordered (Fig. 3e). The ~100 fold lower efficiency of Polι in inserting C relative to T opposite 1-MeA can be rationalized by the inability of dCTP to gain a firm “foothold” opposite 1-MeA. Incoming dCTP does not offer the same hydrogen bonding opportunities opposite 1-MeA as dTTP, and its N4 amino group sterically impinges on the N6 group of 1-MeA. Accordingly, it seems to adopt a range of conformations that is not conducive to catalysis. Also, Asp126 remains in a binary-like conformation and prevents the α-phosphate of dCTP from aligning properly with respect to the primer terminus for catalysis.

Although Polι is inefficient at incorporating the incorrect C (and incorrect A) opposite 1-MeA, it does so more efficiently than opposite undamaged template A (Table 1). We suspect that this is because the methyl group at N1 favors the imino tautomer of 1-MeA26. In its imino tautometric form, the N6 imino group of 1-MeA would be in a position to establish a putative N6(1-MeA)…N4(dCTP) or N6(1-MeA)…N6(dATP) hydrogen bond, enabling Polι to incorporate C or A opposite 1-MeA more readily than opposite A.

In conclusion, we present here the first kinetic and structural analysis of the ability of Polι to replicate through the 1-MeA adduct. 1-MeA is highly cytotoxic because a methyl group at N1 atom impairs W-C base pairing and presents a strong block to normal DNA replication. By pushing 1-MeA in to the syn conformation, Polι can carry out effective TLS opposite 1-MeA via Hoogsteen base pairing with correct incoming T.

Methods

Crystallization

The GST-Polι (residues 1–420) fusion protein was expressed and purified as described previously27. A self-complementary 18-mer oligonucleotide was synthesized containing dideoxycytosine at its 3′ end (5′-TCT-1-MeA-GGGTCCTAGG ACCCdd-3′, 1-MeA: N1-methyl-deoxyadenosine). Prior to crystallization, the oligonucleotide was annealed with itself to give a “double-ended” template-primer with two replicative ends18. For crystallization of the Polι1-MeA.dTTP and Polι1-MeA.dCTP ternary complexes, Polι and DNA were mixed in the ratio of 1:1.2, followed by the addition of dTTP or dCTP and MgCl2 to final concentrations of 20 mM and 10 mM respectively. The ternary complexes were crystallized from solutions containing 15–20% PEG 5000 MME and 0.2 –0.4 M (NH4)2SO4 in 0.1 M MES buffer (pH = 6.0). Crystals belong to space group P6522 with cell dimensions of a = 98.0 Å, b = 98.0 Å, c = 202.5/202.2 Å and α = β = 90°, γ = 120°. For data collection, the crystals were cryoprotected by soaks for 5 minutes in mother liquor solution containing 5%, 10%, 15% and 20% and 25% glycerol, respectively, and then flash frozen in liquid nitrogen.

Structure Determination and Refinement

X-ray data on cryocooled crystals were measured at Brookhaven National Laboratory (BNL beamline X-25) and Advanced Photon Source (APS, beamline 24-ID-E). Data sets were indexed and integrated using HKL200028. The Polι1-MeA.dTTP and Polι1-MeA.dCTP structures were solved by molecular replacement (MR), using the PolιA.dTTP complex as a search model (2FLL, with template and incoming dNTP omitted). The first round of refinement and map calculation was carried out without the template and the incoming nucleotide. Initial electron density map showed unambiguous density for the template 1-MeA in both the structures, which was then included in the model for subsequent refinement. Iterative rounds of refinement and water picking were performed with Phenix21 and model building with program Coot22. All models have good stereochemistry, as shown by MolProbity29,30 with >99% of the residues in the most favored regions of the Ramachandran plot and 0.8% in the disallowed regions. Figures were prepared using PyMol31.

DNA Polymerase Assay

DNA substrates consisted of a radiolabeled oligonucleotide primer annealed to a 75nt oligonucleotide DNA template by heating a mixture of primer/template at a 1:1.5 molar ratio to 95 °C and allowing it to cool to room temperature for several hours. The template 75-mer oligonucleotide contained the sequence 5′AGC AAG TCA CCA ATG TCT AAG AGT TCG TAT AAT GCC TAC ACT GGA GTA CCG GAG CAT CGT CGT GAC TGG GAA AAC-3′ and was either undamaged A or harbored a 1-MeA at the underlined position. For steady-state kinetic analyses of nucleotide insertion opposite the undamaged A or 1-MeA, a 44 mer primer 5′ GTT TTC CCA GTC ACG ACG ATG CTC CGG TAC TCC AGT GTA GGC AT-3′ was used annealed to the above mentioned 75 mer templates.

The standard DNA polymerase reaction (5 μl) contained 25 mM Tris·HCl (pH 7.5), 5 mM MgCl2, 1 mM dithiolthreitol, 100 μg/ml BSA, 10% glycerol, and 10 nM DNA substrate, and Polι (0.02–0.2 nM).

Steady-State Kinetic Analysis

Steady-state kinetic analyses for deoxynucleotide incorporation were performed as described32. Polι (0.02–0.2 nM) was incubated with primer:template DNA substrate (10 nM) and increasing concentration of dNTPs for 10 min, at 37 °C. Gel band intensities of the substrate and products of the deoxynucleotide incorporation reactions were quantified by using a PhosphorImager and the IMAGEQUANT software (Molecular Dynamics). The observed rate of deoxynucleotide incorporation, vobs was determined by dividing the amount of product formed by the reaction time and protein concentration. The vobs was graphed as a function of the deoxynucleotide concentration, and the data were fit to the Michaelis-Menten equation describing a hyperbola: vobs = (kcat[E] × [dNTP])/(Km + [dNTP]). From the best fit curve, the apparent Km and kcat steady-state kinetics parameter were obtained for the incorporation of dNTP by the Polι and the efficiencies of nucleotide incorporation (kcat/Km) were determined.

Additional Information

How to cite this article: Jain, R. et al. Mechanism of error-free DNA synthesis across N1-methyl-deoxyadenosine by human DNA polymerase-ι. Sci. Rep. 7, 43904; doi: 10.1038/srep43904 (2017).

Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.