Oligonucleotides containing disulfide-tethered cytosine and specific base lesions were synthesized using the phosphoramidite derivative of O⁴-triazoyl-dU (Glen Research), and functionalized with cystamine. The DNA oligomers were synthesized on a Biosystems Expedite Nucleic Acid Synthesis System using standard reagents. Following synthesis, the resin was dried under vacuum and then incubated with 200 µl of solution containing 50% cystamine in water at room temperature for 16 h. The aqueous solution was collected and the resin was washed twice with 200 µl water. The washes were combined and neutralized to pH 6.0–6.5 with glacial acetic acid on ice and desalted by a NAP-10 column (GE healthcare). The aqueous solution was then lyophilized and finally purified by reverse-phase HPLC. The dsDNA used in the crystallization of N¹-meA and N⁶-meA is DNA1: 5′-CTGTATC*AT(^meA)GCG-3′ paired with 5′-TCGCTATAATACA-3′. The dsDNA used for N³-meC is DNA 2: 5′-CTGT(^meC)TC*ATTGCG-3′ paired with 5′-TCGCAATAAGACA-3′. Masses of the synthesized oligonucleotides were verified by MALDI-TOF mass spectrometry.

The abh2-ΔN55 gene was cloned between the NdeI and HindIII sites of a pET28a vector (Novagen). ABH2-ΔN55 E175C mutants were generated by using QuikChange Site-Directed Mutagenesis Kit (Stratagene) and this plasmid was transformed into E. coli BL21(DE3) cells (Stratagene) for protein overexpression. The protein was purified using HisTrap HP column (GE Healthcare) with an elution buffer of 50 mM sodium phosphate (pH 8.0), 300 mM NaCl and 400 mM imidazole. The N-terminal His-tag was removed by an overnight thrombin (MP Bio) digestion at 4°C and the tag-free protein was purified the next day with another round of HisTtrap chromatography. About 4–6 mg of protein could be obtained from 1 l of bacterial culture.

DNA duplexes were annealed by mixing 1 mM thiol-tether containing strand with the corresponding complimentary strand in buffer containing 10 mM Tris (pH 7.4) and 100 mM NaCl, incubating at 75°C for 10 min and cooling to 4°C by a step gradient of −1°C/min. The cross-linked complexes of protein ABH2-ΔN55 E175C with synthetic oligonucleotides were achieved by incubating dsDNA (1 mM, 50 µl) with protein (1.2 eq.) in 5 ml of buffer [100 mM NaCl, 10 mM Tris–HCl (pH 7.4)] at 14°C for 16 h. The covalently linked ABH2–DNA complexes were purified using MonoQ anion exchange chromatography (GE healthcare), then buffer exchanged to a solution containing 100 mM NaCl, 10 mM Tris–HCl (pH 8.0) (Supplementary Figure S3). Finally, the complexes were concentrated to 5 mg/ml for use in crystallization experiments.

Protein–dsDNA complex crystals were grown by hanging drop vapor diffusion crystallization at 4°C in drops containing 1 µl of complex solution and 1 µL of reservoir solution of 100 mM NaCl, 50 mM MgCl₂, 100 mM cacodylate (pH 6.5) and 12% 5 K PEG. Hexagonal rod-shaped crystals grew in 1–2 weeks at 4°C. Subsequently, crystals were transferred to a cryoprotectant solution composed of 80% reservoir solution and 20% glycerol and frozen in liquid nitrogen before data collection. An X-ray data set (diffracted to 2.0 Å) for the N¹-meA-containing complex (ABH2–N¹-meA) crystals and an X-ray data set (diffracted to 1.8 Å) for the N⁶-meA-containing complex (ABH2–N⁶-meA) were collected at beamline 23ID-B (General Medicine and Cancer Institutes Collaborative Access Team [GM/CA-CAT]) of the Advanced Photon Source at Argonne National Laboratory. The N³-meC-containing complex (ABH2–N³-meC) gave the best crystals that diffracted to 2.0 Å (Table 1). Data were integrated and processed with the HKL2000 package. Table 1.

Data collection and refinement statistics

The ABH2–dsDNA complex structures were phased by molecular replacement [using Phaser (35)], using the previously published ABH2–dsDNA structure as a search model [PDB 3BTX (34)]. The model was built by using COOT and refinement was carried out with the program REFMAC5 from the CCP4 suite. Atomic coordinates for the three structures reported have been deposited in Protein Data Bank under accession number 3H8O (ABH2–N¹-meA), 3H8R (ABH2–N⁶-meA) and 3H8X (ABH2–N³-meC).

In an attempt to crystallize duplex DNAs containing the methylated base N¹-meA and N³-meC, we employed the Dickerson–Drew dodecamer sequence (9–12), which failed to yield crystals. Then, we decided to use the cross-linked ABH2-dsDNA complex (34) as a host–guest-like system for crystallization since the covalent cross-linking stabilizes the protein–dsDNA complexes. The lesion-containing dsDNA is cross-linked to the ABH2 protein with an engineered Cys at position 175 (ABH2-ΔN55 E175C) (Supplementary Figure S2). Out of multiple DNA sequences screened, a 13-mer dsDNA1 (Figure 1B), with a central C*:A base pair (C* represents a disulfide-tethered cytosine), yielded the best quality crystals of the ABH2–dsDNA complexes containing N¹-meA (Figure 1A) or N⁶-meA (Figure 1E). Another sequence, dsDNA2 (Figure 1D), with N³-meC at a different location, gave the best crystals for ABH2–dsDNA complex with N³-meC (Figure 1C). All of these complexes bearing modified nucleotides are crystallized in the hexagonal space group P6₅22, and the crystals diffract to 1.8 Å for the N⁶-meA-containing complex (ABH2–N⁶-meA), and 2.0 Å for both N¹-meA- and N³-meC-containing complexes (ABH2–N¹-meA and ABH2–N³-meC, Table 1). Figure 1.

Crystal structure of the ABH2–dsDNA complex. (A) Cartoon of the ABH2–dsDNA1 complex with N¹-meA. The protein is shown in green, DNA in orange, N¹-meA in magenta and complementary T in blue. (B) Schematic diagram showing the interaction between ABH2 and dsDNA 1 containing N¹-meA. Solid arrows stand for direct interactions and dashed arrows for water mediated contacts. (C) Cartoon trace of the ABH2–dsDNA2 complex with N³-meC. The same color coding in A is used with N³-meC in magenta and the opposite G in blue. (D) Schematic diagram showing the interaction between ABH2 and dsDNA 2 containing N³-meC. (E) Cartoon of the ABH2–dsDNA1 complex with N⁶-meA. The same color coding in (A) is used with N⁶-meA in magenta and complementary T in blue. (F) Schematic diagram showing the interaction between ABH2 and dsDNA 1 containing N⁶-meA.

Besides minor shift and rotation, the packing of the dsDNA helix in ABH2–N¹-meA is similar to the unlesioned DNA in 3BTX (Figure 3). The N¹-meA-containing dsDNA shows two slightly different backbone conformations, which do not seem to cause noticeable changes in the base-pairing mode of the DNA structure (Supplementary Figure S6). As N¹-meA can be converted to N⁶-meA via Dimroth rearrangement (36), we also crystallized ABH2–N⁶-meA for comparison. The DNA structure with N⁶-meA and its alignment to the unlesioned DNA in 3BTX are shown in Supplementary Figure S7. Besides an ∼15° bending of the DNA caused by the end packing of DNA, the overall DNA conformation in these two structure remains basically B-form with an average rise per base pair along axis of around 3.2 Å, a slide of −0.5 Å and a roll of 3.6°. Because N⁶-meA forms Watson–Crick base pair with the complementary T (Figure 4E), this N⁶-meA-containing dsDNA basically resembles an unmodified B-form-like duplex structure. Figure 3.

Diagram of the dsDNA1 structure containing N¹-meA:T. (A) Structure of dsDNA1 with N¹-meA:T. N¹-meA10 in magenta, opposite T5′ in cyan. dsDNA 1: 5′-CTGTATC*AT(^1meA)GCG-3′ paired with 5′-TCGCTATAATACA-3′. (B) Structure of dsDNA1 with N¹-meA:T aligned with the unlesioned dsDNA in structure 3BTX. DNA containing N¹-meA in red, unlesioned DNA in green. dsDNA in 3BTX: 5′-CTGTATC*ATT GCG-3′ paired with 5′-TCGCAATAATACA-3′.

Due to the lack of hydrogen bonding between N³-meC and the opposite G, the crystal structure of ABH2–N³-meC shows a distorted backbone compared to the complex without the base lesion. The DNA ends are packed differently in two structures (Figure 7) that helps to release the bending of the dsDNA caused by DNA end packing. The absence of any interaction between N³-meC to the opposite G leads to a shift of N³-meC towards the major groove. This base has a relatively poor electron density indicating its high flexibility. Instead of completely flipping out, the N³-meC base stays partially intrahelical (Figure 8A), with a tilt angle around 18° (0.6° for B-form DNA) and a twist angle of 49° (36° for B-form DNA). The distances of P_T4 … P_N³-meC5 and P_N³-meC5 … P_T6 are stretched to 7.2 Å and 6.9 Å, respectively. In contrast, the complementary guanine of N³-meC remains intrahelical, stabilized by the stacking to A11′ and A9′, its two flanking bases. Water-mediated hydrogen bonds of N3_G10′… O1′_A11′ and O1′_G10′… N3_A9′ also contribute to the stabilization of the guanine. In addition, another water-mediated interaction between G10′ and T4, N1_G10′ to water (2.8 Å) and water to N1_T4 (3.1 Å), might also help to ‘lock’ the unpaired G10′ intrahelix. Although the neighboring base pair, T4:A11′, is not apparently influenced by this lesion, a marked change occurred to the base pair T6:A9′. This base pair is disrupted such that T6 adopts a syn conformation to form a Wobble base pair with A8′ (N3_T6 … N1_A8′, 2.8 Å; O2_T6 … N6_A8′, 3.1 Å), which is the base opposite the flipped out C7*. The disruption of the two base pairs around the N³-meA lesion site creates a pocket that is partly occupied by the unpaired N³-meC5 and A9′. Benefitting from potential stacking with the adjacent A8′ and G10′, nucleotide A9′ stays in this pocket in the absence of base pairing. These observed distortions may suggest that the introduction of N³-meC lesion to dsDNA results in a significant alternation to the duplex structure. Figure 7.

Diagram of the DNA structure containing N³-meC:G. (A) Structure of dsDNA2 with N³-meC:G. N³-meC5 in magenta, opposite G10’ in blue. dsDNA 2: 5′-CTGT(^3meC)TC*ATTGCG-3′ paired with 5′-TCGCAATAAGACA-3′. (B) Structure of dsDNA2 with N³-meC:G aligned with the unlesioned dsDNA in structure 3BTX. DNA- containing N³-meC in red, unlesioned DNA in green. dsDNA in 3BTX: 5′-CTGTATC*ATT GCG-3′ paired with 5′-TCGCAATAATACA-3′. Different packing pattern at the end of the dsDNA is observed.