Structures of (5'S)-8,5'-Cyclo-2'-deoxyguanosine Mismatched with dA or dT.

Diastereomeric 8,5'-cyclopurine 2'-deoxynucleosides, containing a covalent bond between the deoxyribose and the purine base, are induced in DNA by ionizing radiation. They are suspected to play a role in the etiology of neurodegeneration in xeroderma pigmentosum patients. If not repaired, the S-8,5'-cyclo-2'-deoxyguanosine lesion (S-cdG) induces Pol V-dependent mutations at a frequency of 34% in Escherichia coli. Most are S-cdG → A transitions, suggesting mis-incorporation of dTTP opposite the lesion during replication bypass, although low levels of S-cdG → T transversions, arising from mis-incorporation of dATP, are also observed. We report the structures of 5'-d(GTGCXTGTTTGT)-3'·5'-d(ACAAACAYGCAC)-3', where X denotes S-cdG and Y denotes either dA or dT, corresponding to the situation following mis-insertion of either dTTP or dATP opposite the S-cdG lesion. The S-cdG·dT mismatch pair adopts a wobble base pairing. This provides a plausible rationale for the S-cdG → A transitions. The S-cdG·dA mismatch pair differs in conformation from the dG·dA mismatch pair. For the S-cdG·dA mismatch pair, both S-cdG and dA intercalate, but no hydrogen bonding is observed between S-cdG and dA. This is consistent with the lower levels of S-cdG → T transitions in E. coli.

Introduction

Hydroxyl radicals, generated by cellular oxidative stress and inflammation, damage nucleobases[1] or deoxyribose sugars,[2] or both in DNA.[3] At 2′-deoxyguanosines, hydroxyl radical-mediated hydrogen abstraction at the deoxyribose C5′-position followed by attack at the guanine C8 carbon forms an N7-centered radical, which may be oxidized to diastereomeric 8,5′-cyclo-2′-deoxyguanosines (cdG).[4−10] The 8,5′-cyclo-2′-deoxyadenosines (cdA) have also been characterized.[3,6,8−15] These 8,5′-cyclopurine-2′-deoxynucleosides have been detected at the nucleotide level,[4,11] in DNA,[4,16−18] and cells in vitro,[5] in human urine,[19] and in vivo.[18,20,21] They might contribute to neurologic disease in xeroderma pigmentosum complementation group C (XP-C) patients.[22] They are also believed to play roles in Cockayne syndrome,[18] breast and ovarian cancer,[20] and familial Mediterranean fever.[23]

It has been reported that S-cdG does not block primer elongation by Klenow DNA polymerases, and dATP is preferentially incorporated opposite the lesion in vitro.[25] However, in Escherichia coli, S-cdG blocks DNA replication and is refractory to repair.[24] Upon induction of the SOS response, it induces 34% mutations.[24] Most are S-cdG → A transitions, although S-cdG → T transversions and low levels of deletions of the 5′-neighbor dC are also observed.[24]

For both cdG and cdA, the diastereomeric ratio at the C5′-position depends on experimental conditions and DNA conformation.[4−6,13,19,26,27] Computational studies predicted that the incorporation of the cdA stereoisomers into DNA would result in helical distortions at the lesion site.[28−30] Both the R- and the S-diastereomers of the 8,5′-cA ribonucleoside have been crystallized, and both exhibit the anti conformation about the N-glycosidic bond with χO4′–C1′–N9–C8 = 30 or 27°, respectively.[31,32] The fused six-member ring C8–N9–C1′–O4′–C4′–C5′ adopts the half-chair conformation with the O4′ and C4′ out of plane. The ribose adopts the O4′-exo (0T1) pseudorotation with P = 289° and τm = 48°. Molecular mechanics calculations predicted that the cdA diasetereomers maintain the O4′-exo pseudorotation when placed opposite dT in DNA.[28] The NMR data and ab initio calculations suggest that incorporation of the S-cdA into di- or trinucleotides does not change the O4′-exo pseudorotation.[33]

Recently, we reported the structure of the S-cdG·dC pair in 5′-d(GTGCXTGTTTGT)-3′·5′-d(ACAAACACGCAC)-3′, containing the DNA sequence of p53 codons 272–275 (X = S-cdG, Scheme 1).[34] The S-cdG participates in Watson–Crick hydrogen bonding with the complementary dC. However, the S-cdG deoxyribose shifts to the O4′-exo pseudorotation with P = 280°. This altered backbone torsion angles γ from ∼50 to −67° and δ from ∼120 to 149°, as compared with canonical B-DNA. Additionally, the torsion angles β and χ are changed from ∼180 to −87° and from ∼−120 to −157°, respectively. The twist and base pair shift helicoidal parameters are perturbed at the C4·G21 and X5·C20 base pairs. The purine ring is anti about the N-glycosidic bond, and the fused six-membered ring adopts the half-chair conformation with O4′ and C4′ out of plane.

Scheme 1

Numbering Scheme of the Mismatched Oligodeoxynucleotide Duplexes Containing the S-cdG

Here, we report the structures of the duplexes 5′-d(GTGCXTGTTTGT)-3′·5′-d(ACAAACAYGCAC)-3′ (X denotes S-cdG; Y denotes either dA or dT, Scheme 1). These model the situation following mis-incorporation of dTTP opposite S-cdG, leading to S-cdG → A transitions, or following mis-incorporation of dATP opposite S-cdG, leading to S-cdG → T transversions in E. coli.[24] The S-cdG·dT mismatch pair adopts a wobble base pairing, providing a plausible rationale for the S-cdG → A transitions. For the S-cdG·dA mismatch pair, both S-cdG and dA intercalate, but no hydrogen bonding is observed between S-cdG and dA. This is consistent with the lower levels of S-cdG → T transitions in E. coli.[24]

Materials and Methods

Materials

The oligodeoxynucleotide 5′-d(GTGCXTGTTTGT)-3′ (X = S-cdG) was synthesized and characterized as reported.[34]The oligodeoxynucleotides 5′-d(GTGCGTGTTTGT)-3′ and 5′-d(ACAAACAYGCAC)-3′ (Y = dA or dT) were synthesized and purified by anion-exchange chromatography (Midland Certified Reagent Co., Midland, TX). Oligodeoxynucleotides were desalted using Sephadex G-25. The oligodeoxynucleotides 5′-d(GTGCGTGTTTGT)-3′ or 5′-d(GTGCXTGTTTGT)-3′ were annealed at 1:1 stoichiometry with the complementary oligodeoxynucleotide 5′-d(ACAAACAYGCAC)-3′ in 10 mM NaH2PO4, 100 mM NaCl, and 50 μM Na2EDTA (pH 7.0). The solutions were heated to 95 °C for 10 min and cooled to room temperature. The duplexes were isolated using DNA grade hydroxylapatite with a gradient from 10 to 200 mM NaH2PO4 in 100 mM NaCl and 50 μM Na2EDTA (pH 7.0) and desalted using Sephadex G-25.

Melting Temperature

Melting temperatures of the DNA duplexes were measured by UV/vis spectroscopy at 260 nm in 10 mM NaH2PO4, 100 mM NaCl, and 50 μM Na2EDTA (pH 7.0). The strand concentration was 10 μM. The thermal scans proceeded from 10 to 80 °C with an interval of 1 °C. The melting temperatures were calculated by differentiating the absorbance vs temperature profiles.

NMR

Samples were at 1.0 mM strand concentration. Samples for the nonexchangeable protons were dissolved in 500 μL in 10 mM NaH2PO4, 100 mM NaCl, and 50 μM Na2EDTA (pH 7.0). They were exchanged with D2O and suspended in 280 μL of 99.996% D2O, and the pH was adjusted with dilute DCl or NaOD. Experiments were performed at 800 MHz. The temperature was 25 °C. Magnitude correlated spectroscopy (COSY) spectra were recorded with 512 real data in the t1 dimension and 2048 real data in the t2 dimension. Total correlation spectroscopy (TOCSY) spectra were recorded with a mixing time of 80 ms. Chemical shifts were referenced to water. The exclusive correlation spectroscopy (E-COSY) spectra were recorded with 1024 real points in the t1 dimension and 4096 real points in the t2 dimension.[35] The spectra were zero-filled during processing to create a matrix of 2048 × 16384 points. The temperature was 30 °C. Nuclear Overhauser effect spectroscopy (NOESY) spectra were recorded with 512 real points in the t1 dimension and 2048 real points in the t2 dimension. NOESY spectra were zero-filled during processing to create a matrix of 1024 × 1024 real points. NOESY experiments used time-proportional phase increment (TPPI) quadrature detection[36] and mixing times of 60, 150, 200, and 250 ms. The relaxation delay was 1.5 s. Data were processed using the program TOPSPIN[37] and analyzed with the program SPARKY.[38] Samples for the observation of exchangeable protons were dissolved in 500 μL of 10 mM NaH2PO4 and 100 mM NaCl, 50 μM EDTA (pH 7.0) containing 9:1 H2O:D2O (v/v) (pH 7.0). NOESY experiments were performed at 500 MHz, at 5 °C. The Watergate sequence was used for water suppression,[39] with a nuclear Overhauser enhancement (NOE) mixing time of 250 ms. The 31P–H1 experiments were carried out at the 1H frequency of 600 MHz. 31P–H3′ 3J couplings were applied to determine the phosphodiester backbone conformation.[40]31P chemical shifts were referenced using indirect shift ratios.[41]

Distance and Dihedral Angle Restraints

Integration footprints were defined using NOE cross-peaks obtained at a mixing time of 250 ms. NOE intensities from data obtained at mixing times of 60, 150, 200, and 250 ms to check for the presence of spin diffusion effects were determined by volume integrations. For each mixing time, these were combined as necessary with intensities generated from complete relaxation matrix analysis of a starting structure to yield a hybrid NOE intensity matrix.[42,43] The program MARDIGRAS[44−46] iteratively refined the hybrid intensity matrix and optimized agreement between calculated and experimental NOE intensities. The RANDMARDI algorithm[45] carried out iterations, randomizing peak volumes within limits specified by the input noise level.[46] Calculations were initiated using isotropic correlation times of 2, 3, and 4 ns. Analysis of these data yielded distance restraints used in restrained molecular dynamics (rMD) calculations (Table S3 in the Supporting Information) and the corresponding standard deviations for the distance restraints.

The pseudorotations (P) were estimated by examining the 3JHH of deoxyribose protons.[47] The data were fit to curves relating the coupling constants to P, deoxyribose pucker amplitude (ϕ), and the percentage S type conformation. The P and ϕ ranges were converted to the dihedral angles ν0–ν4. To obtain backbone torsion angle restraints for the modified, flanking, and terminal base pairs, coupling constants measured from 1H–31P heternuclear multiple bond correlation spectroscopy (HMBC) spectra were applied[48,49] to the Karplus relationship[50] to determine the dihedral angle ε (C4′–C3′–O3′–P), related to the H3′–C3′–O3′–P angle by a 120° shift. The ζ (C3′–O3′–P–O5′) torsion angles were calculated from the correlation between ε and ζ in B-DNA.[40] At all other base pairs, backbone torsion angle restraints utilized canonical values derived from B-DNA.[51] Watson–Crick hydrogen-bonding restraints minimized propeller twisting between base pairs, except at the X5·T20 base pair in the S-cdG·dT mismatched duplex and at the X5·A20 and A6·T19 base pairs in the S-cdG·dA mismatched duplex.

Molecular Dynamics Calculations

The partial charges for the cdG nucleotide were obtained from density functional theory (DFT) calculations, utilizing the B3LYP/6-31G* basis set and the program GAUSSIAN.[52] The starting structures were generated from A and B type DNAs by constructing a bond between G5 C8 and G5 C5′ followed by 200 iterations of potential energy minimization using the conjugate gradients algorithm. The rMD calculations utilized a simulated annealing approach.[53] The calculations were conducted with the program AMBER[54] and the parm99 force field. The generalized Born (GB) model[55] with parameters developed by Tsui and Case[56] was used for implicit water simulation. The program complete relaxation matrix analysis (CORMA) was utilized to calculate the NOE intensities from the structures emergent from calculations. Helicoidal analyses were carried out with the programs CURVES[57] and 3DNA.[58]

Results

Thermal Stability

The duplex containing the S-cdG·dT base pair exhibited a melting temperature (Tm) of 38 ± 1 °C. The unmodified dG·dT mismatched duplex exhibited a Tm of 43 °C under the same conditions (Figure 1A). Thus, the incorporation of S-cdG reduced the Tm by 5 °C. The duplex containing the S-cdG·dA pair exhibited a Tm of 31 ± 1 °C. The unmodified dG·dA mismatched duplex exhibited a Tm of 39 ± 1 °C (Figure 1A). Thus, the incorporation of S-cdG reduced the Tm by 8 °C. 1H NMR spectra of the duplex containing the S-cdG·dT pair were compared with the unmodified mismatched duplex at different temperatures (Figure 1B). At the X5·T20 base pair, the T20 N3H resonance was not observed at 5 °C. For the modified duplex, the T6 N3H resonance broadened at lower temperature than did the other thymine imino resonances. 1H NMR spectra of the duplex containing the S-cdG·dA base pair at different temperatures are displayed in Figure 1C. The X5 and T6 imino resonances were not observed at 5 °C. The imino resonances for base pairs C4·G21, G5·A20, and T6·A19 of the unmodified duplex were not observed. Moreover, for the unmodified duplex containing the G5·A20 mismatch, the C4 and C18 H5 → H6 scalar couplings were not observed, and that of C22 was weak.

Figure 1

(A) UV melting profiles of the mismatched duplexes as compared with the corresponding unmodified duplexes: duplex containing dG·dA base pair (○), duplex containing S-cdG·dA base pair (●), duplex containing dG·dT base pair (□), and duplex containing S-cdG·dT base pair (■). (B) 1H NMR of the mismatched duplex containing the S-cdG·dT base pair at different temperatures. (C) 1H NMR of the mismatched duplex containing the S-cdG·dA base pair at different temperatures. The broad resonance at ∼13.1 ppm observed at 5–15 °C was unassignable; it might be assigned to X5 N1H or T6 N3H.

S-cdG Mismatched with dT

NMR Resonance Assignments

The nonexchangeable protons of the S-cdG-modified duplex were assigned based upon the NOE sequential connectivity of the base proton H6 or H8 dipolar couplings with H1′ deoxyribose protons (Figure 2A,B).[59,60] For the modified strand, the sequential connectivity was observed from G1 to C4. Because the S-cdG nucleotide lacked a proton at the C8 carbon, the sequential connectivity exhibited an interruption at X5. The X5 H1′ proton was identified at 6.11 ppm; it exhibited a weak X5 H1′ → T6 H6 NOE. The sequential connectivity resumed from T6 to T12. For the modified strand, all of the deoxyribose H1′ protons were observed within a narrow chemical shift window, between 5.8 and 6.3 ppm. The complete sequential connectivity was observed for the complementary strand.

Figure 2

NOE (250 ms) connectivities of base H8/H6 protons with deoxyribose H1′ protons of the S-cdG modified duplexes. (A) Modified strand for the duplex containing the S-cdG·dT base pair. (B) Complementary strand for the duplex containing the S-cdG·dT base pair. (C) Modified strand for the duplex containing the S-cdG·dA base pair. (D) Complementary strand for the duplex containing the S-cdG·dA base pair.

The assignments of X5 deoxyribose protons were made by analysis of scalar and dipolar couplings. Figure 3A displays a tile plot derived from a NOESY spectrum obtained at 60 ms mixing time. X5 H1′ exhibited dipolar couplings with H2′ and H2″; weak scalar couplings were also observed. H3′ exhibited dipolar couplings with H2′, H2″, and H4′, whereas the scalar couplings were not observed. H4′ exhibited both scalar and dipolar couplings with the H5′ proton. The geminal H2′ and H2″ protons were not resolved. For the remainder of the duplex, the H2′, H2″, H3′, and H4′ deoxyribose resonances were unequivocally assigned. The absolute configurations of the geminal H2′ and H2″ protons were assigned from their NOEs to H1′ and H3′. With the exception of the unresolved resonances for X5, G11, and T12, H2′ exhibited a weaker NOE with H1′ than did H2″, whereas it exhibited a stronger NOE with H3′ than did H2″. The resonance assignments of the nonexchangeable DNA protons are tabulated in Table S1 in the Supporting Information.

Figure 3

Expansions of NOESY spectra (60 ms) showing the assignment of S-cdG nonexchangeable protons. (A) Duplex containing S-cdG·dT base pair. (B) Duplex containing S-cdG·dA base pair.

The resonances of the base imino protons were assigned based on sequential connectivity in NOESY spectra, and the assignments were supported by NOEs to the amino protons of Watson–Crick base pairs (Figure 4A).[61] The NOE sequential connectivity was observed from T2 → G3 to G21 and from T6 → G7 → T8 → T9 → T10 to G11. At the 5′-neighbor base pair, G21 N1H exhibited NOEs with C4N4 H1 and N4 H2. At the 3′-neighbor base pair, the T6 N3H resonance exhibited NOEs with A19 H2 and A19N6 H1. With the exception of the terminal base pairs, the remaining NOEs arising from Watson–Crick hydrogen bonding were observed. The X5 N1H resonance was observed at ∼9.1 ppm. It exhibited an NOE with X5N2H, which also had an NOE with T6 N3H (Figure 4B). As compared to other guanine imino resonances, X5 N1H shifted upfield.

Figure 4

Assignment of the exchangeable protons of the S-cdG modified duplexes. (A) Duplex containing the S-cdG·dT base pair; NOE cross-peaks are assigned as follows: a, A19 H2 → G7 N1H; b, C4N4H2 → G21 N1H; c, C4N4H1 → G21 N1H; d, A19N6 H1 → T6 N3H; and e, A19 H2 → T6 N3H. (B) Assignment of X5 N1H in the duplex containing the S-cdG·dT base pair; NOE cross-peaks are assigned as follows: f, T6 N3H → X5N2H; and g, X5 N1H → X5N2H. (C) Duplex containing the S-cdG·dA base pair; NOE cross-peaks are assigned as follows: h, A19 H2 → G7 N1H; i, C4N4H2 → G21 N1H; and j, C4N4H1 → G21 N1H.

Deoxyribose Coupling Constants

The scalar couplings of the 2′-deoxyribose H1′ protons with the H2′ and H2″ protons were measured from an E-COSY spectrum (Figure S1 in the Supporting Information). The 3JH1′-H2′ and 3JH1′-H2″ values for X5 were 2.1 and 4.9 Hz, respectively. The 3JH4′-H5′ was 4.9 Hz, whereas the 3JH3′-H4′ was not measurable. For T20, 3JH1′-H2′ and 3JH1′-H2″ were 5.3 and 8.8 Hz, respectively. With the exception of the terminal nucleotides, the 3JH1′-H2′ for other nucleotides were 8–10 Hz, and the 3JH1′-H2″ were 5–7 Hz. The 3J coupling constants for the deoxyribose protons are tabulated in Table S2 in the Supporting Information.

Phosphodiester Backbone Conformation

The 31P resonances were assigned from a 31P–H3′ HMBC spectrum. With the exception of X5, each exhibited heteronuclear coupling with H3′ of the 5′-neighbor nucleotide. The 31P NMR spectrum of the S-cdG containing duplex was compared with the unmodified mismatched duplex (Figure S2 in the Supporting Information). At the S-cdG nucleotide, the 31P resonance shifted upfield. The other 31P resonances were clustered within a modest chemical shift range, centered in the spectral region characteristic of B-DNA.

Chemical Shift Perturbations

Chemical shifts of the nonexchangeable protons between the S-cdG-containing duplex andthe unmodified mismatched duplex were compared (Figure 5). Remarkable changes were observed at X5 and the 5′- and 3′-neighboring nucleotides of the modified strand. C4 H2″ shifted downfield by 0.72 ppm; X5 H2′ and H2″ shifted upfield by 0.21 and 0.51 ppm, respectively; and T6 H6, H1′, and H2′ shifted downfield by 0.32, 0.24, and 0.23 ppm, respectively. The chemical shift perturbations for the complementary strand were small, with the exception of A19 H8 and T20 H1′, which shifted upfield by 0.24 and 0.31 ppm, respectively.

Figure 5

Chemical shift perturbations of the duplex containing the S-cdG·dT base pair. (A) Base protons of the modified strand. (B) 2′-Deoxyribose protons of the modified strand. (C) Base protons of the complementary strand. (D) 2′-Deoxyribose protons of the complementary strand.

Structural Refinement

A total of 406 distance restraints, including 263 intranucleotide and 143 internucleotide restraints, were calculated from the intensities of NOE cross-peaks (Table S3 in the Supporting Information).[45] A total of 21 NOEs involving the S-cdG protons were used as restraints. A total of 47 empirical distance restraints arising from Watson–Crick base pairing were used, as were 160 empirical torsion angle restraints that were applied to the nonterminal nucleotides. These were justified based upon NMR data, which suggested that structural perturbations were localized at and adjacent to the lesion site. Weak wobble base pair restraints were used for the X5·T20 base pair, and no torsion angle restraints were used for the C4·G21, X5·T20, and T6·A19 base pairs. The restraints are summarized in Table 1.

Table 1

rMD Restraints and Statistical Analysis of rMD Converged Structures of the S-cdG Containing Duplexes

duplex	S-cdG·dT	S-cdG·dA
total restraints for rMD calculation	613	607
experimental NOE distance restraints	406	399
intranucleotide NOE restraints	263	260
internucleotide NOE restraints	143	139
S-cdG NOE restraints	21	30
empirical base pair restraints	47	43
empirical torsion angle restraints	160	165
backbone torsion angle restraints	90	95
deoxyribose torsion angle restraints	70	70
structure statisticsa
NMR R factor (R1x) (×10–2)	6.12	6.85
intranucleotide NOEs	5.33	6.45
internucleotide NOEs	7.76	7.67
rmsd deviation of refined structures	0.52	0.44

Mixing time used to calculate R1 was 250 ms. R1 = ∑|(a0)1/6 – (ac)1/6|/|(a0)1/6|, where a0 and ac are the intensities of observed (nonzero) and calculated NOE cross-peaks, respectively.

The rMD calculations for the S-cdG-containing duplex were performed from A and B form starting structures. Ten emergent structures, five each for A- and B-DNA starting structures, were obtained and minimized with respect to potential energy. All converged as indicated by pairwise rmsd comparisons (Table 1). The accuracies of the emergent structures were evaluated by comparison of theoretical NOE intensities for the refined structure calculated by the program CORMA[44] to the experimental NOE intensities, to yield sixth root residuals (R1).[42] These, as well as the residuals for intra- or internucleotide NOEs, were consistently <0.1 (Table 1). R1 values for each nucleotide were <0.15 (Figure S3 in the Supporting Information). Thus, the refined structures provided accurate depictions of the NOE data.

Structure of the S-cdG:dT Mispair

The X5·T20 pair adopted the wobble conformation (Figure 6). Figure 7A,B shows base stacking and base pairing at the lesion site. The X5·T20 pair exhibited a shift of −0.8 Å, displacing C4 toward the major groove. This pair exhibited a greater than normal opening of 16.8°. Typical B-DNA pairing and stacking interactions were maintained for the remaining base pairs (Table S4 in the Supporting Information). The S-cdG deoxyribose was in the O4′-exo, “west” pseudorotation (Figure 8A), with P = 280° and τm = 47°. The heavy atoms N9, O3′, and C5′ were axial about the deoxyribose ring. The complementary T20 was in the C4′-exo, “north” pseudorotation, with P = 65° and τm = 37°. Consequently, X5 H2″ was farther from the X5 purine ring as compared to the H2″ protons in B-DNA, and C4 H2″ was proximate to the X5 purine ring. With the exception of the terminal nucleotides, other pseudorotations were either C1′-exoor C2′-endo. The six-membered ring C8–N9–C1′–O4′–C4′–C5′ adopted the envelope (half boat) conformation. Helicoidal analysis of the backbone torsion angles (Figure S4 in the Supporting Information) showed that for S-cdG, the β angle shifted from 180 to −78°. The γ angle shifted from 50 to −57°. Perturbations of the δ and ζ torsion angles from 120 to 147° and from −90 to −58°, respectively, were also observed. There was also a change for the N-glycosidic torsion angle χ from −120 to −162°. For the complementary T20, a perturbation of the δ torsion angle from 120 to 83° was observed.

Figure 6

Expanded views of the refined structure of the S-cdG containing duplexes at the lesion site. (A) Duplex containing the S-cdG·dT base pair, viewed from the minor groove. (B) Duplex containing the S-cdG·dT base pair, viewed from the major groove. (C) Duplex containing the S-cdG·dA base pair, viewed from the minor groove. (D) Duplex containing the S-cdG·dA base pair, viewed from the major groove.

Figure 7

Base pairing and base stacking of the refined structures of the S-cdG containing duplex at the lesion site. The pink arrows indicate anticipated hydrogen-bonding interactions. (A) The C4·G21 and X5·T20 base pairs. (B) The X5·T20 and T6·A19 base pairs. (C) The C4·G21 and X5·A20 base pairs. (D) The X5·A20 and T6·A19 base pairs.

Figure 8

Deoxyribose conformations of the S-cdG in the refined structures. (A) Duplex containing the S-cdG·dT base pair. (B) Duplex containing the S-cdG·dA base pair. (C) O4′-exo pseudorotation of the deoxyribose.

S-cdG Mismatched with dA

The nonexchangeable protons were assigned based upon the sequential connectivity of the base proton H6 or H8 dipolar couplings with H1′ deoxyribose protons (Figure 2C,D).[59,60]For the modified strand, the NOE connectivity was observed from G1 to C4. The connectivity exhibited an interruption at X5 due to the lack of a proton at the C8 carbon. The X5 H1′ proton was identified at 6.12 ppm; it exhibited a weak X5 H1′ → T6 H6 NOE, suggesting that the distance between these two protons was greater than in B-DNA. The sequential connectivity resumed from T6 to T12. For the modified strand, the deoxyibose H1′ protons were observed within a narrow chemical shift window, between 5.8 and 6.3 ppm. The complete sequential connectivity was observed for the complementary strand.

The resonances of A19 H2 and A20 H2 appeared at 7.41 and 7.46 ppm, respectively. An NOE was observed between them. In addition, A19 H2 exhibited an NOE with G7 N1H (Figure 4C), suggesting both A19 and A20 were intercalated. As expected, both H2 protons exhibited NOEs with H1′ protons in the minor groove (Figure 9). Notably, A20 exhibited NOEs with both T6 H1′ and A19 H1′ of the 5′-flanking T6·A19 base pair but did not exhibit NOEs with C4 H1′ or G21 H1′ of the 3′-flanking C4·G21 base pair.

Figure 9

Expansion of the NOESY spectrum (250 ms) of the duplex containing the S-cdG·dA base pair showing the intercalation of A19 and A20. NOE cross-peaks are assigned as follows: a, T6 H6 → T6 H1′; b, T6 H6 → X5 H1′; c, A20 H2 → A20 H1′; d, A20 H2 → T6 H1′; e, A20 H2 → A19 H1′; f, A20 H2 → X5 H1′; g, A19 H2 → A20 H1′; and h, A19 H2 → A19 H1′.

The assignments of X5 deoxyribose protons were made by analysis of scalar and dipolar couplings. Figure 3B displays a tile plot derived from a NOESY spectrum at 60 ms mixing time. X5 H1′ exhibited dipolar couplings with H2′ and H2″; weak scalar couplings were also observed. H3′ exhibited dipolar couplings with H2′, H2″, and H4′, whereas the scalar couplings were not observed. H4′ exhibited both scalar and dipolar couplings with the H5′ proton. For the remainder of the duplex, the H2′, H2″, H3′, and H4′ resonances were assigned unequivocally. The absolute configurations of the geminal H2′ and H2″ protons were assigned from their NOEs to H1′ and H3′. With the exception of the unresolved resonances for G11, T12, and C24, H2′ exhibited a weaker NOE with H1′ than did H2″, whereas it exhibited a stronger NOE with H3′ than did H2″. The resonance assignments of the nonexchangeable DNA protons are tabulated in Table S5 in the Supporting Information.

The resonances of the base imino protons were assigned based on sequential connectivity in NOESY spectra and NOEs to the amino protons of Watson–Crick base pairs (Figure 4C).[61]The NOE connectivity was observed from T2 → G3 to G21 and from G7 → T8 → T9 → T10 to G11. The resonances of X5 N1H and T6 N3H were not assigned although a broad resonance was observed at ∼13.2 ppm at temperatures below 15 °C. The assignment failed due to a lack of NOE interactions. At the 5′-neighbor base pair, G21 N1H exhibited NOEs with C4N4 H1 and N4 H2. With the exception of the terminal base pairs, the remaining NOE cross-peaks arising from Watson–Crick hydrogen bonding were observed.

The scalar couplings of the 2′-deoxyribose H1′ protons with the H2′ and H2″ protons were measured from an E-COSY spectrum (Figure S1 in the Supporting Information). The 3JH1′-H2′ and 3JH1′-H2″ values for X5 were 2.7 and 7.2 Hz, respectively. The 3JH4′-H5′ was 6.7 Hz, whereas the 3JH3′-H4′ was not measurable. With the exception of the terminal nucleotides, the 3JH1′-H2′ for other nucleotides were 8–10 Hz, and the 3JH1′-H2″ were 5–7 Hz. The 3J coupling constants for the deoxyribose protons are tabulated in Table S6 in the Supporting Information.

The 31P resonances were assigned from a 31P–H3′ HMBC spectrum. With the exception of X5, each exhibited a heteronuclear coupling with H3′ of the 5′-neighbor nucleotide. The spectrum of the S-cdG-containing duplex was compared with the unmodified mismatched duplex (Figure S2 in the Supporting Information). At the modified nucleotide, the 31P resonance shifted upfield. The other 31P resonances were clustered within a modest chemical shift range, in the spectral region characteristic of B-DNA.

A total of 399 distance restraints, including 260 intranucleotide and 139 internucleotide restraints were calculated from the intensities of NOE cross-peaks (Table S7 in the Supporting Information).[45] A total of 30 NOEs involving the S-cdG protons were used as restraints. A total of 43 empirical distance restraints arising from Watson–Crick base pairing interactions were used, as were 165 empirical torsion angle restraints that were applied to refine the nonterminal nucleotides. These were justified based upon NMR data, which suggested that structural perturbations were localized at and adjacent to the lesion site. The data suggested that no base pairing existed at the X5·A20 and T6·A19 base pairs, so base pairing restraints were not used for these base pairs. No torsion angle restraints were used for the C4·G21, X5·A20, and T6·A19 base pairs. The restraints used for the structure refinement are summarized in Table 1.

Ten rMD calculations, five each for A- and B-DNA starting structures, were performed. The 10 emergent structures were minimized with respect to potential energy. All converged as indicated by pairwise rmsd comparisons (Table 1). The accuracies of the emergent structures were evaluated by comparison of theoretical NOE intensities calculated for the refined structure by the program CORMA[44] to the experimental NOE intensities to yield sixth root residuals (R1).[42] These, as well as the residuals for intra- or internucleotide NOEs, were consistently less than 0.1 (Table 1). R1 values for each nucleotide were less than 0.15 (Figure S3 in the Supporting Information). Thus, the refined structures provided accurate depictions of the NOE data.

Structure of the S-cdG:dA Mispair

Both X5 and A20 intercalated into the duplex (Figure 6). Consequently, the helical rise values from C4·T21 to X5·A20 and from X5·A20 to T6·A21 were greater than normal, 5.4 and 4.6 Å, respectively. Figure 7C,D shows the base stacking and base pairing at the lesion site. Significant perturbations in shift were observed from base pairs T2·A23 to G7·A18, centered at the X5·A20 base pair. The C4·T21 base pair exhibited a greater than normal base pair twist of 54°. The remaining base pairs exhibited normal base stacking (Table S8 in the Supporting Information). The S-cdG nucleotide was in the O4′-exo, “west” pseudorotation (Figure 8B), with P = 264° and τm = 47°. The heavy atoms N9, O3′, and C5′ were axial about the deoxyribose ring. Consequently, X5 H2″ was farther from the X5 purine ring as compared to the H2″ protons in B-DNA, while C4 H2″ was proximate to the X5 purine ring. With the exception of the terminal nucleotides, all other pseudorotations were either C1′-exo or C2′-endo. The six-membered ring C8–N9–C1′–O4′–C4′–C5′ adopted the envelope (half boat) conformation. Helicoidal analysis of the backbone torsion angles (Figure S4 in the Supporting Information) showed that at the lesion site, the β angle shifted from the characteristic 180 to −83°. The γ angle shifted from 50 to −59°. Perturbations of the δ and ζ torsion angles from 120 to +157° and from −90 to −75°, respectively, were also observed. There was also a change for the N-glycosidic torsion angle χ from −120 to −143°.

Discussion

This work extends an investigation of oligodeoxynucleotides containing the S-cdG lesion.[34] If not repaired, S-cdG blocks DNA replication in E. coli and is genotoxic.[24] In SOS-induced E. coli, a mutation frequency of 34% is observed. Most mutations are S-cdG → A transitions, although S-cdG → T transversions and a deletion of the 5′-neighbor C are also observed.[24] Accordingly, structures in which S-cdG is placed opposite dT or dA, representing intermediates leading to S-cdG → A transitions and S-cdG → T transversions, are of interest.

Structure of the S-cdG:dT Mispair

The dG:dT mismatch often exists as a wobble base pair with both bases in the anti conformation. One might predict that locking S-cdG into the anti conformation about the N-glycosidic bond would be consistent with the formation of a wobble S-cdG:dT mismatch pair, and this appears to be the case. The formation of a wobble pair is consistent with the upfield shift of the X5 N1H resonances, which was also observed for the G5 N1H of the corresponding unmodified duplex (Figure 1B).[62−66] However, the T20 N3H resonance was not observed, indicating enhanced solvent exchange at the S-cdG·dT wobble pair. The structural refinement suggests the potential formation of a three-point hydrogen bond among X5 N1H, X5N2H2, and T20O2 and a weak hydrogen bond between X5O6 and T20 N3H (Figure 7). However, the observation that the S-cdG·dT wobble pair exhibits a Tm 5 °C lower than the corresponding duplex containing a dG·dT mismatch pair (Figure 1A) suggests that the incorporation of S-cdG reduces the stability of dG·dT wobble pairing. The broadening of the T6 N3H and T20 N3H resonances as compared to the other thymine imino resonances (Figure 1B) suggests that the greatest destabilization occurs at the modified X5·T20 and 3′-neighboring T6·A19 base pairs. The 5′-neighboring C4·G21 base pair is mildly affected. The base pair shifts at the C4·G21 and X5·T20 base pairs are consistent with this conclusion. Similarly, for the S-cdG:dC pairing interaction, a 9 °C decrease in Tm was observed relative to the unmodified duplex.[34] The thermal destabilization of this duplex is likely associated with the shift of the S-cdG deoxyribose to the O4′-exo (west) pseudorotation, as opposed to the “south” pseudorotation (C2′-endo) observed in B-DNA or the “north” pseudorotation (C3′-endo) in A-DNA.[51,67] Moreover, the complementary T20 deoxyribose shifts to the C4′-exo (north) pseudorotation, as evidenced by the 3JH1′-H2′ and 3JH1′-H2″ of 5.3 and 8.8 Hz, respectively (Table S2 in the Supporting Information). The accommodation of the constrained S-cdG nucleotide necessitates helicoidal perturbation of the phosphodiester backbone torsion angles β, γ, δ, and ζ in the modified strand. Additionally, smaller perturbations of the T20 phosphodiester backbone torsion angle δ in the complementary strand (Figure S4 in the Supporting Information) may factor in the reduced stability of the S-cdG·dT vs dG·dT mispairing interaction.

S-cdG:dA Mispair

Both S-cdG and dA are inserted into the duplex, but they do not engage in hydrogen bonding. Instead, helicoidal perturbations of the modified strand allow both to intercalate, creating a gap at the mismatched region. The absence of the G21 N1H → A20 H2 NOE agrees with the gap between A20 and G21 caused by the intercalation of S-cdG (Figure 6C,D). This is also consistent with the observation that A20 H2 exhibits NOEs with both H1′ protons of the 3′-flanking T6·A19 pair, but not with the H1′ protons of the 5′-flanking C4·G21 pair, suggesting A20 was close to T6·A19 but further from C4·G21 (Figure 9). The observation of the A19 H2 → A20 H2 NOE suggests A20 remains intercalated. The 8 °C decrease of the Tm as compared to the duplex containing a dG·dA mispair is probably related to alterations of the S-cdG phosphodiester backbone torsion angles β, γ, δ, and ζ and to perturbations of the base pair shift parameters at the C4·G21 and X5·A20 base pairs, which are necessitated to accommodate the constrained S-cdG O4′-exo (west) pseudorotation. It is of interest to note that Malyshev et al.[68] obtained a similar structure for a “self-intercalating” non-natural, non-hydrogen-bonding base pair that demonstrates excellent polymerase activity, with slower rates of extension.

The S-cdG·dA mismatch is distinct from the dG·dA mismatch as dG·dA mismatch pairs are influenced by sequence and pH. The failure to observe the imino resonances for base pairs C4·G21, X5·A20, and T6·A19 of the unmodified duplex is consistent with this notion and suggests an increased rate of exchange of these protons with solvent, perhaps accompanied by structural disorder. A dG(anti)·dA(anti) pair was identified by Prive et al.[69,70] The dG(anti)·dA(syn) pair has been identified in the crystalline state at pH > 7,[71−73] while the protonated dG(syn)·dA+(anti) pair has been identified at pH 6.6.[74,75] In solution, the dG(anti)·A(anti) pair is observed at neutral or basic pH conditions.[76−80] The dG(syn)·dA+(anti) base pair is observed at acidic pH.[80,81] Carbonnaux et al.[80] observed that the latter is stabilized by bifurcated hydrogen bonds. Another type of dG(anti)·dA(anti) pairing is associated with tandem dG:dA mismatches.[78,82−84] This is referred to as a “sheared” or “type I1” G(anti)·A(anti) base pair and is accompanied by phosphodiester backbone perturbations. These differences between the S-cdG·dA and dG·dA mismatches are attributed to the fact that the S-cdG lesion is locked into the anti conformation about the N-glycosidic bond and the shift of the deoxyribose to the O4′-exo (west) pseudorotation. The dA(anti)·dG(anti) “face-to-face” conformation[69,70,76−78,80,82−87] would predict an NOE between G21 N1H and A20 H2, which is not observed (Figure 4C).

Structure–Activity Relationships

The wobble S-cdG·dT pair (Figure 6) is consistent with the site-specific mutagenesis studies in SOS-induced E. coli, showing a preponderance of S-cdG → dA transition mutations.[24] Because low levels of S-cd G → dT transversions are observed in E. coli,[24] we surmise that low levels of dATP are incorporated opposite S-cdG during trans-lesion synthesis. It has been suggested that Klenow DNA polymerases insert dATP opposite S-cdG.[25] Further studies of template·primers containing the S-cdG lesion complexed with error-prone polymerases will be of interest. The low levels of S-cdG → dT transversions might reflect the distortion of the S-cdG·dA mismatch, in which both S-cdG and dA are intercalated but do not hydrogen bond (Figure 6). Malyshev et al.[68] have reported that a non-natural, non-hydrogen-bonding base pair exhibiting a similar structure in duplex DNA demonstrates excellent polymerase activity, with slower rates of extension.

Conclusions

The structures of S-cdG placed opposite dA or dT were determined and compared with the structure when S-cdG placed opposite dC.[34] The S-cdG·dT base pair adopted a hydrogen-bonded wobble conformation, while the S-cdG·dA base pair differed from dG·dA mispairs—both S-cdG and dA were intercalated, but no hydrogen bonding was observed. In each instance, the S-cdG deoxyribose adopted the O4′-exo (west) pseudorotation and was accommodated by backbone and base-pairing helicoidal perturbations. Collectively, these perturbations may be important in understanding the mutagenicity and genotoxicity of S-cdG.