Structure determination of DNA methylation lesions N¹-meA and N³-meC in duplex DNA using a cross-linked protein–DNA system

Abstract

N¹-meA and N³-meC are cytotoxic DNA base methylation lesions that can accumulate in the genomes of various organisms in the presence of S_N2 type methylating agents. We report here the structural characterization of these base lesions in duplex DNA using a cross-linked protein–DNA crystallization system. The crystal structure of N¹-meA:T pair shows an unambiguous Hoogsteen base pair with a syn conformation adopted by N¹-meA, which exhibits significant changes in the opening, roll and twist angles as compared to the normal A:T base pair. Unlike N¹-meA, N³-meC does not establish any interaction with the opposite G, but remains partially intrahelical. Also, structurally characterized is the N⁶-meA base modification that forms a normal base pair with the opposite T in duplex DNA. Structural characterization of these base methylation modifications provides molecular level information on how they affect the overall structure of duplex DNA. In addition, the base pairs containing N¹-meA or N³-meC do not share any specific characteristic properties except that both lesions create thermodynamically unstable regions in a duplex DNA, a property that may be explored by the repair proteins to locate these lesions.

INTRODUCTION

Genomic DNA is constantly subjected to modifications caused by exogenous environmental chemicals and cellular metabolites. Among various DNA damages, nucleobase damage is a common type that can be induced by oxidation, hydrolysis and alkylation (1,2). If left unrepaired, these lesions can cause deleterious effects on nearly all aspects of cellular functions (3,4). Therefore, guarding genetic integrity by repairing DNA damage is one of the most fundamental processes of life (5,6). Most DNA base damage is repaired through four biochemically and mechanistically distinct pathways: direct reversal repair (DRR), base excision repair (BER), nucleotide excision repair (NER) and mismatch repair (MMR) (7).

Structural characterization of base lesions in double-stranded DNA (dsDNA) is important to understand the origin of their mutagenic or cytotoxic effects. The structural information may also help to reveal how the lesions are detected and repaired by DNA repair proteins. Despite considerable advances in DNA structure characterization (8), high-quality crystals of naked dsDNA containing base lesions are hard to obtain. Even with the Dickerson–Drew dodecamer sequence (9–12), base lesions can introduce unstable base pairs, which prevent the formation of well-ordered 3D crystal lattice. In order to aid DNA crystallization, host–guest systems employing various DNA-binding proteins have been used to obtain high-quality crystals (13–17). Although this method has helped to solve the structures of some important lesions (16,17), many more remain to be elucidated. The host–guest approach has also been shown to be effective for RNA crystallization (18–22).

Alkylated nucleobases are produced mostly by endogenous and environmental alkylation agents (23). N1-methyladenine (N¹-meA) and N3-methylcytosine (N³-meC) are two forms of base methylations generated usually in single-stranded DNA (ssDNA). These lesions are unable to form normal Watson–Crick base pairs, and consequently block DNA replication, resulting in cytotoxicity to cells. Escherichia coli alkB is an inducible gene of the adaptive response to alkylating agents. The alkB gene encodes a DNA-repair protein that counteracts toxic alkylation damages, such as N¹-meA, N³-meC and 1,N⁶-ethenoadenine (24–28). Human homologues of AlkB can process similar base modifications and exhibits profound functional roles in human cells (29–31). Solving the structures of these base modifications will be important to visualize how these lesions affect the local dsDNA structure, thereby shedding light on the origin of the cytotoxic effects of the damage. The structure of 1, N⁶-ethenoadenine:T pair was recently characterized using the DNA glycosylase AlkA bound to the ends of dsDNA (17). The crystal structures of the other two lesions, N¹-meA and N³-meC in dsDNA, remain unknown. We have solved the structure of hABH2, the human protein that repairs N¹-meA and N³-meC in ssDNA and dsDNA (32,33), cross-linked to dsDNA (34). As most of the DNA bases have no direct contact with the protein residues in the ABH2–dsDNA structure (PDB 3BTX, Supplementary Figure 1), this complex seems to be a suitable host–guest-like system for characterizing lesion-containing dsDNA structures. In our design, the protein–dsDNA complex is stabilized through covalent cross-linking in the protein active site, whereas the lesioned base is installed to the locations without any direct contact to the protein (Supplementary Figure S2). A noticeable advantage of this approach versus other host–guest-like systems is its robustness since the complex is stitched together with a covalent bond. The complex cannot be crystallized in the absence of this cross-linking.

This host–guest-like system is validated through the structural characterization of the methylation base modifications N¹-meA, N³-meC and N⁶-meA in dsDNA cross-linked to ABH2. In these structures, N¹-meA forms an unambiguous anti-Hoogsteen base pair to the opposite T; N³-meC does not form any hydrogen-bonding interaction with the opposite G, but stays partially intrahelical inside the duplex DNA.

MATERIALS AND METHODS

Oligonucleotide synthesis and purification

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.

Cloning, expression and purification of ABH2-ΔN55 E175C

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.

Preparing cross-linking, crystallization and data collection of the ABH2–DNA complexes

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

Crystal	ABH2–N1-meA:T	ABH2–N6-meA:T	ABH2–N3-meC:G
Data quality
Resolution (Å)	20–2.0 (2.05–2.0)	20–1.8 (1.77–1.8)	20–2.0 (1.95–2.0)
Unique reflections	29687	38371	31492
Completenessa (%)	99.9 (99.9)	99.9 (100)	98.3 (78.6)
Redundancy	20.0	20.3	21.1
<I/σ>a	24.3 (2.6)	32.6 (1.6)	36.2 (2.2)
Crystal parameters
Space group	P6522	P6522	P6522
Cell constants (Å)	a = 79.11	a = 78.08	a = 79.11
	b = 79.11	b = 78.08	b = 79.11
	c = 242.27	c = 228.71	c = 242.30
α, β, γ (°)	α = 90	α = 90	α = 90
	β = 90	β = 90	β = 90
	γ = 120	γ = 120	γ = 120
Refinement
Resolution (Å)	20–2.0	20–1.77	20–1.95
Rworkb (%)	20.2	21.0	20.5
Rfreec (%)	23.7	23.6	23.6
Model quality
R.m.s. deviation bond (Å)	0.015	0.010	0.0107
R.m.s. deviation angle (°)	1.58	1.53	1.57
Average B factor	27.7	29.2	36.5
Average B factor of Protein	27.1	27.9	34.6
Average B factor of DNA	23.1	29.0	39.6
Average B factor of water	39.4	42.5	45.3
Model content
Protein residues	56–258	56–258	56–258
Nucleotides	26	26	26
Water atoms	197	222	179
PDB accession code	3H8O	3H8R	3H8X

^aValues in parentheses refer to the highest resolution bins.

^bR_work = Σ |F_o – F_c|/Σ F_o, where F_o and F_c are observed and calculated structure factor amplitudes, respectively.

^cR_free was calculated based on a percentage of data (5%) randomly selected and omitted through the structure refinement procedure.

Structure determination and refinement

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).

RESULTS

Overall structures

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.

The structures of ABH2–N¹-meA and ABH2–N⁶-meA overlap well with the original structure 3BTX (Supplementary Figure S5), with a root mean square deviation (RMSD) of 0.49 Å and 0.21 Å, respectively. Except for the site of the flipped-out C*7, the DNA adopts a right-handed conformation and its interactions with the protein are mainly through the phosphodeoxyribose backbones, which are mostly located in the middle of the lesion-containing strand and 5′-end of the complementary strand. Interestingly, the structure of ABH2–N¹-meA complex shows reduced contacts between protein residues and the backbone of dsDNA at the 5′-end of the complementary strand. For instance, the backbone region around T5′ (which pairs with N¹-meA) and its flanking nucleotides, A6′ and C4′ (Figure 2), do not form hydrogen bonds with the protein residues Arg198, Gly204, Lys205 and Arg215; such hydrogen-bonding contacts exist in the original structure of 3BTX. Besides Phe102, the finger residue that helps to flip C*7 out of the duplex DNA, no other protein residue affects base pairing in both structures of ABH2–N¹-meA and ABH2–N⁶-meA as well as in 3BTX (34).

Figure 2.

Interaction of ABH2 protein to the dsDNA backbone. (A) Cartoon diagram of the ABH2–DNA1 complex with N¹-meA. The protein is shown in green, DNA in gold, Arg198 and Arg215 in yellow and two residues on the loop (Gly204, Lys205) in magenta. (B) Same structure as in (A), rotated 90° to the right to show the side view of the residues and dsDNA. (C) Overall superposition of ABH2–N¹-meA complex with 3BTX to show the backbone distortion caused by N¹-meA. ABH2–N¹-meA complex in green with backbone of T10' in yellow and A9′ in blue. 3BTX in magenta. (D) Overall superposition of ABH2–N⁶-meA complex with 3BTX to show the backbone distortion caused by N⁶-meA. The color coding is same as in (C). (E) Overall superposition of ABH2–N³-meC complex with 3BTX to show the backbone distortion caused by N³-meC. Same color coding in (C) is used.

A superposition of the structure of ABH2–N³-meC to 3BTX shows an apparent DNA backbone shift toward the 3′-end of the complementary strand (Figure 2E), which is most likely caused by different packing of the DNA ends. The original structure (3BTX) has the terminal bases twisted out of the duplex DNA to interact with the neighboring oligonucleotide; the terminal bases in ABH2–N³-meC remain intrahelical and form coaxial stacking with the neighboring duplex. As a result, the DNA bending induced by DNA end packing is slightly released when compared to the original structure (3BTX). Similar to ABH2–N¹-meA, a loss of contact between ABH2 and the backbone region around the nucleotides C4′-A6′ was observed in the structure of ABH2–N³-meC.

Structures of ABH2–N¹-meA and ABH2–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′.

Figure 4.

Structure of DNA1 with N¹-meA or N⁶-meA lesion opposite with T. (A) Side view of N¹-meA:T lesion and neighboring base pairs. Atoms are colored as follows: carbon, green; nitrogen, blue; oxygen, red; phosphate, orange. (B) F_o – F_c simulated annealing omit electron density map of N¹-meA:T pair. The F_o – F_c simulated annealing omit electron density map is represented as an orange mesh contoured at 3σ and the 2F_o – F_c map in gray mesh. The color scheme is same as in (A). Black dotted lines represent hydrogen bonds. (C) The N¹-meA:T base pair as viewed from the major groove, illustrating an overall dihedral around 30°. Same color coding in (A) is applied. (D) Side view of N⁶-meA:T and neighboring base pairs. (E and F) F_o – F_c simulated annealing omit electron density map of N⁶-meA:T pair from top and major groove, respectively. Color scheme and dotted lines are same as in (B).

To perform a detailed analysis of the distortion of the duplex structure caused by the N¹-meA lesion, the structure of the N¹-meA-containing oligonucleotide was superimposed onto that of the N⁶-meA-containing DNA (Supplementary Figure S8). The base pair parameters of these two DNA structures were calculated using the program 3DNA v1.5 (37) and the changes for propeller twist, opening, buckle, roll and twist are shown in Figure 5. Major changes occur around the lesion site, where the positions of N¹-meA10, complementary T5′, and the flanking base pair G11:C4′ shift relative to their positions in the structure of the N⁶-meA-containing duplex. These differences are results from the N1-methylation of A10, which blocks the normal Watson–Crick base pairing (Figures 3 and 4).

Figure 5.

Helix parameters of the dsDNA with N¹-meA:T pair (light gray) and the one with N⁶-meA:T (dark grey), as calculated by program 3DNA v1.5. The base pairs are numbered from G3·C12′ to G13·C2′ in the absence of the C*7·A8′ mismatch and the steps are numbered from G3·C12′/T4·A11′ to C12·G3′/G13·C2′ without two steps neighbored to the C*7·A8′ mismatch. Most of the remarkable changes are at the local position of the N¹-meA:T pair.

A Hoogsteen base pair between N¹-meA10 and the opposite base T5′ is clearly observed with two hydrogen bonds formed between these two bases (N2_ N¹-meA10 … O4_T5′, 3.1 Å; N7_ N¹-meA10 … N3_ T5′, 3.2 Å; Figure 4B). Although the Hoogsteen base-pair mode has been reported by the NMR study of the N¹-meA:T pair (38), the distances we observed here are much shorter than what have been assigned previously (around 4.3 Å). To accommodate both the methylated adenine and the pyrimidine ring of T within the duplex, the purine ring of N¹-meA adopts a syn conformation with respect to the N9 atom around the N-glycosidic bond. In addition, a distinct dihedral angle of about 30° is found between the plane of N¹-meA10 and the opposite T5′ (Figures 4C and 6), which is significantly larger than that in the ABH2–N⁶-meA structure (Figure 4D). Interestingly, despite dramatic changes occurring on the N¹-meA10:T5′ base pair, the neighboring base pair T9:A6′ (5′ to the lesion) is almost unaffected, but the base pair G11:C4′ (3′ to N¹-meA), shifts slightly towards the 3′ direction of the lesion-containing strand (Figure 6A). Compared to the N⁶-meA structure, the rise of the N¹-meA:T pair to its flanking base pairs was decreased, which was caused by the inclination of the N¹-meA base. In spite of this, the overall distance from T9:A6′ base pair to G11:C4′ is similar to that in a typical B-form DNA, which may indicate that the influence of the lesion is mainly limited to the local structure, but not the overall duplex.

Figure 6.

Structure comparison of N¹-meA:T lesion with N⁶-meA:T base pair. (A) Side view structure of N¹-meA10:T5′ with neighboring base pairs aligned with N⁶-meA10:T5’. N¹-meA10:T5′ with neighboring base pairs appears in red, and N⁶-meA10:T5′ with neighboring base pairs in green. The figure illustrates a significant dihedral between N¹-meA10 and N⁶-meA10 when two bases are overlaid together. (B) Overlay of N¹-meA10 and adjacent T9 (in red) with N⁶-meA10 and neighbor T9 (in green) as viewed from top. The purine ring of N¹-meA10 is in syn conformation, and the N⁶-meA10 base is in anti-conformation.

Since the protein has no direct contact to the base pairing region around N¹-meA10 or N⁶-meA10, an alignment of the two duplex DNAs in this region (Figure 6) allows a clear comparison of the properties of N¹-meA with those of N⁶-meA, and reveals impacts on the local DNA structure by these modifications. It can be seen that N¹-meA adopts a syn conformation to form the Hoogsteen base pair instead of the anti conformation adopted by N⁶-meA in a Watson–Crick base pair. A comparison of N¹-meA:T pair with the N⁶-meA:T (Table 2) reveals that the shear of N¹-meA:T pair is bigger (0.7 Å compared to −0.05 Å), and so is the stretch value (−3.8 Å in N¹-meA:T versus −0.2 Å in N⁶-meA:T). The stagger of N¹-meA:T amounts to −1.1 Å (0.01 Å in N⁶-meA:T), which is mainly caused by a shift of the N1-methylated adenine into the major groove by ∼3.0 Å and a shift of the opposite T by ∼1.7 Å. The propeller twist of the methylated base pair is varied from −9.2° in N⁶-meA:T pair to 7.7° in N¹-meA:T, while the two flanking base pairs are almost unaffected. The dramatically increased base opening of N¹-meA:T pair is observed (73.6° compared to 0.3° in N⁶-meA:T, Figure 5B), as well as the greater buckle angle (29.1° versus 1.5°). However, the buckle value of G11:C4′ in the N¹-meA structure, 3′ to the lesion pair, is decreased to −3.6°, which is 6.3° for the corresponding base pair in the N⁶-meA structure. Apparent changes are also observed in the N¹-meA structure compared to the N⁶-meA and the unlesioned ones when calculating the base pair step parameters as shown in Table 2. The tilt and twist values of the N¹-meA:T pair to both flanking base pairs show significant changes. With respect to the roll angles, only the value of the N¹-meA:T pair to 3′ adjacent G11:C4′ pair is altered dramatically to −101.2° (7.0° in N⁶-meA structure), but that of the 5′ neighbored T9:A9′ pair is only slightly affected. Furthermore, the C1′–C1′ distance between N¹-meA and the opposite T decrease to 9.0 Å (10.5 Å in N⁶-meA structure), and the distances of RN9–YN1 and RC8–YC6 are also shortened by 1.7 Å and 3.1 Å, respectively, compared to those in the N⁶-meA:T pair. These changes might be induced by the rotation of N¹-meA base around the glycosidic bond, which ‘drags’ the complementary thymine further to the center of the helix (Supplementary Figure 9). A slight DNA distortion of the backbone is indicated by an elongation of the P_T9 … P_A′10 from 7.2 Å in the N⁶-meA-containing structure to 7.4 Å in the case of N¹-meA, which probably explains the lack of interaction between the local backbone and the protein.

Table 2.

Comparison of base pair parameters of N¹-meA:T, N⁶-meA:T and T:A^a

Local base-pair parametersb
		Shear	Stretch	Stagger	Buckle	Propeller	Opening
N1-meA10:T10		0.7	−3.8	−1.1	29.1	7.7	73.6
N6-meA10:T10		−0.05	−0.2	0.01	1.5	−9.2	0.3
T10:A10		−0.05	−0.2	−0.2	1.4	−21.0	−0.5
Local base-pair step parametersc
		Shift	Slide	Rise	Tilt	Roll	Twist
N1-meA	T9A/ N1-meA10T	−1.7	−3.8	2.4	−175.7	1.1	−47.6
	N1-meA10T/G11C	−0.8	−3.5	−1.7	135.6	−101.2	132.3
N6-meA	T9A/ N6-meA10T	−0.2	−0.2	3.4	−0.9	−3.3	38.9
	N6-meA10T/G11C	0.4	−0.02	3.2	−2.8	7.0	31.3
3BTX	T9A/T10A	−0.3	−0.04	3.2	2.9	−3.7	40.2
	T10A/G11C	0.8	0.9	3.6	0.4	6.0	37.3

^aAll data are calculated by Program 3DNA (v1.5) (ref. 37).

^bParameters for Shear, Stretch and Stagger are distances (Å). Parameters for Buckle, Propeller and Opening are angles (°).

^cParameters for Shift, Slide and Rise are distances (Å). Parameters for Tilt, Roll and Twist are angles (°).

A comparison of the structure of N¹-meA:T to that of the Hoogstein base-paired A:T reveals that the same syn conformation of N¹-meA or adenine is observed in both structures. A similar shortening of the C1′-C1′ distance is also observed (9.0 Å for N¹-meA:T pair, 8.5 Å for A:T Hoogsteen base pair, as opposed to 10.5 Å for A:T Watson–Crick base pair) (39,40). However, the distinct dihedral angle of ∼30° between the plane of N¹-meA10 and the opposite T5′ as well as the major changes in twist angles in this region, are only found in the N¹-meA:T structure, which could be explained by the steric effect of the N1-methylation.

Structure of dsDNA-containing N³-meC

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.

Figure 8.

Structure of dsDNA2 with N³-meC lesion opposite with G and its flanking base pairs. (A) Side view of N³-meC5 lesion complementary with G10′ and neighboring base pairs. Atoms are colored as follows: carbon, green; nitrogen, blue; oxygen, red; phosphate, orange. A Wobble base pairing between T6 and A8′ is shown in place of the Watson–Crick base-pairing of T6 with A9′. Black dotted lines represent hydrogen bonds. (B) Interactions between G10′ with the neighboring nucleotides help unpaired G10′ and A9′ stay intrahelical. Atoms are in same color coding as (A) with water in purple.

DISCUSSION

Cytotoxic damages such as N¹-meA and N³-meC can block Watson–Crick base pairing in duplex DNA, which leads to cytotoxic consequences to living cells. These base lesions are efficiently detected and repaired by the AlkB family proteins (24–31). To structurally characterize these methylation bases in duplex DNA, a disulfide cross-linked ABH2–dsDNA complex was employed as a host–guest-like system for DNA crystallization. To minimize potential influence from the bound protein, we placed the methylated bases away from the interface between the protein and DNA.

This protein–DNA complex, stabilized by covalent cross-linking between the protein and DNA, affords the first crystal structures of N¹-meA and N³-meC in duplex DNA. We show that N¹-meA forms a Hoogsteen base pair with the opposite T with the purine ring of N¹-meA adopting a syn conformation, which is consistent with the previous NMR result (38). This crystal structure also provides detailed parameters on the N¹-meA:T pair and its impact on the local DNA duplex structure. Due to steric clashes caused by the N1-methylated adenine, we observed a 3 Å shift of N¹-meA base into the major groove in the structure, as well as significant changes in the values of the opening, tilt and twist angles. A shortening of C1′–C1′, RN9–YN1 and RC8–YC6 distances between N¹-meA and the opposite T is observed, probably induced by the displacement of the N¹-meA to the major groove and the movement of the complementary thymine toward the center of the helix. As a result of these conformational changes, the neighboring base pair G11:C4’ shifts to the 3′ direction, while the 5′ neighboring base pair T9:A6′ is left almost unaffected. A detailed investigation of ABH2–N¹-meA also reveals that the protein residues, which normally establish contacts with the backbone in the middle of the lesion-containing strand and 5′-end of the complementary strand, lose most of the hydrogen-bonding interactions with the backbone of the complementary strand (Figure 1B). This observation suggests that: (i) these interactions to the complementary strand might be weak, perhaps non-existent when the ABH2 protein is scanning through the duplex DNA to detect potential base lesions; (ii) the covalent cross-linking is important for holding the complex together, thus providing necessary robustness to this host–guest-like system with a low binding affinity in the absence of cross-linking.

We also report the structure of the ABH2–dsDNA complex containing the N³-meC lesion. The electron density of N³-meC is poor due to the absence of any hydrogen-bonding interaction to the opposing G. However, N³-meC remains intrahelical with a movement of ∼2 Å to the major groove. The opposite guanine is also intrahelical and stacks to the neighboring A11′ and A9′. The disruption of the N³-meC5:G10′ base pair also yields an elongated back-bone P … P distance between C5 to T4 and T6. Furthermore, the flanking base pair between T6 and A9′ is broken, which leads to a Wobble base pairing between T6 and A8′. The melting of the DNA structure around the lesion site generates low thermodynamic stability and produces the distortion of the DNA backbone, an effect that may lead to the loss of interactions between the protein and the backbone around the A6′–C4′ stretch on the complementary strand. These changes in DNA structure may explain why this lesioned base cannot be crystallized through typical DNA-crystallization strategies.

Both N¹-meA and N³-meC are efficiently detected and repaired by the AlkB family proteins, yet they show distinct structural features in duplex DNA. The only common property of the base pairs containing these two lesions is the low thermodynamic stability of the local duplex DNA structure induced by the methylation. Thus, the repair proteins could detect unstable regions in the genome to locate these base damages.

SUPPLEMENTARY DATA

Supplementary data are available at NAR Online.

FUNDING

National Institutes of Health (GM071440 to C.H.); Beamline 23ID-B (General Medicine and Cancer Institutes Collaborative Access Team ([GM/CA-CAT]) at the Advanced Photon Source at Argonne National Laboratory; National Institutes of Health and the United States Department of Energy. Funding for open access charge: GM071440.

Conflict of interest statement. None declared.