The stereochemistry of trans-4-hydroxynonenal-derived exocyclic 1,N2-2'-deoxyguanosine adducts modulates formation of interstrand cross-links in the 5'-CpG-3' sequence.

The trans-4-hydroxynonenal (HNE)-derived exocyclic 1, N(2)-dG adduct with (6S,8R,11S) stereochemistry forms interstrand N(2)-dG-N(2)-dG cross-links in the 5'-CpG-3' DNA sequence context, but the corresponding adduct possessing (6R,8S,11R) stereochemistry does not. Both exist primarily as diastereomeric cyclic hemiacetals when placed into duplex DNA [Huang, H., Wang, H., Qi, N., Kozekova, A., Rizzo, C. J., and Stone, M. P. (2008) J. Am. Chem. Soc. 130, 10898-10906]. To explore the structural basis for this difference, the HNE-derived diastereomeric (6S,8R,11S) and (6R,8S,11R) cyclic hemiacetals were examined with respect to conformation when incorporated into 5'-d(GCTAGC XAGTCC)-3' x 5'-d(GGACTCGCTAGC)-3', containing the 5'-CpX-3' sequence [X = (6S,8R,11S)- or (6R,8S,11R)-HNE-dG]. At neutral pH, both adducts exhibited minimal structural perturbations to the DNA duplex that were localized to the site of the adduction at X(7) x C(18) and its neighboring base pair, A(8) x T(17). Both the (6S,8R,11S) and (6R,8S,11R) cyclic hemiacetals were located within the minor groove of the duplex. However, the respective orientations of the two cyclic hemiacetals within the minor groove were dependent upon (6S) versus (6R) stereochemistry. The (6S,8R,11S) cyclic hemiacetal was oriented in the 5'-direction, while the (6R,8S,11R) cyclic hemiacetal was oriented in the 3'-direction. These cyclic hemiacetals effectively mask the reactive aldehydes necessary for initiation of interstrand cross-link formation. From the refined structures of the two cyclic hemiacetals, the conformations of the corresponding diastereomeric aldehydes were predicted, using molecular mechanics calculations. Potential energy minimizations of the duplexes containing the two diastereomeric aldehydes predicted that the (6S,8R,11S) aldehyde was oriented in the 5'-direction while the (6R,8S,11R) aldehyde was oriented in the 3'-direction. These stereochemical differences in orientation suggest a kinetic basis that explains, in part, why the (6S,8R,11S) stereoisomer forms interchain cross-links in the 5'-CpG-3' sequence whereas the (6R,8S,11R) stereoisomer does not.

trans-4-Hydroxynonenal (1, HNE)1 is produced from the metabolism of membrane lipids(1), and it is the major in vivo peroxidation product of ω-6 polyunsaturated fatty acids[2,3]. Several routes for the formation of HNE from ω-6 polyunsaturated fatty acids have been described[4-6]. HNE exhibits a range of biological effects, from alteration in gene expression and cell signaling to cell proliferation and apoptosis[7-13]. Human exposures to HNE have been implicated in the etiologies of a number of diseases associated with oxidative stress, including Alzheimer’s disease(14), Parkinson’s disease(15), arteriosclerosis(16), and hepatic ischemia reperfusion injury(17).

Abbreviations: HNE, trans-4-hydroxynonenal; HNE−dG, trans-4-hydroxynonenal-derived 2′-deoxyguanosine adduct; NOE, nuclear Overhauser effect; NOESY, nuclear Overhauser effect spectroscopy; COSY, correlation spectroscopy; DQF-COSY, double-quantum-filtered correlation spectroscopy; TOCSY, total correlation spectroscopy; rMD, restrained molecular dynamics; rmsd, root-mean-square deviation.

With regard to genotoxicity, HNE induces the SOS response in Escherichia coli(18). Chromosomal aberrations were observed upon exposures to HNE in rodent[19,20], mammalian[21,22], and human(23) cells. In mammalian cells, the genotoxicity of HNE depends upon glutathione levels that modulate the formation of HNE−DNA adducts[24-26]. Michael addition of the N2-amino group of 2′-deoxyguanosine to HNE gives four diastereomeric exocyclic 1,N2-dG adducts 2−5(27–29) that have been detected in cellular DNA[30-36]. Alternatively, oxidation of HNE to 2,3-epoxy-4-hydroxynonanal and further reaction with nucleobases afford etheno adducts[37-41].

The mutational spectrum induced by HNE−dDNA adducts in the lacZ gene of the single-stranded M13 phage transfected into wild-type E. coli revealed recombination events, C → T transitions, followed by G → C and A → C transversions, and frameshift mutations(29). HNE is mutagenic(42) and carcinogenic in rodent cells(43). Hussain et al.(44) reported that HNE caused G·C → T·A transversions at codon 249 of wild-type p53 in lymphoblastoid cells. Hu et al.(45) further reported that HNE−DNA adducts were preferentially formed with guanine at the third base of codon 249 in the p53 gene. The mutational spectrum induced by HNE−dDNA adducts in the supF gene of shuttle vector pSP189 replicated in human cells showed that HNE induced primarily G → T transversions, accompanied by lower levels of G → A transitions(46). Fernandes et al.(47) conducted site-specific mutagenesis studies and observed that in the 5′-CpG-3′ duplex of interest in this work, only stereoisomers 2 and 3 of the HNE-induced exocyclic 1,N2-dG adduct were mutagenic, inducing low levels of G → T transversions and G → A transitions(47). Evidence that the nucleotide excision repair pathway is involved in the repair of HNE−dG lesions has been obtained[46,48,49].

Wang et al.[50,51] synthesized the four stereoisomers of the exocyclic 1,N2-dG adduct (2−5) and incorporated them into the 5′-CpG-3′ sequence context of 5′-d(GCTAGCXAGTCC)-3′·5′-d(GGACTCGCTAGC)-3′, in which X denotes the HNE−dG adduct. The related 2′-deoxyguanosine adducts of acrolein[52-55] and crotonaldehyde(56) formed reversible interchain cross-links in this sequence context(57). DNA interstrand cross-links block DNA replication and transcription, resulting in cell death if the lesion is not repaired(58). At equilibrium, the (6R) stereoisomer of the crotonaldehyde adduct reached cross-linking levels of 38% as compared to only 5% for the (6S) stereoisomer(53). Similarly, of the four HNE−dG adducts, only stereoisomer 3 possessing (6S,8R,11S) stereochemistry resulted in the formation of DNA reversible interchain cross-links. DNA cross-linking by (6S,8R,11S) stereoisomer 3 represented the predominant species present (>85%) when chemical equilibrium was attained, suggesting that this cross-link is particularly stable in duplex DNA(51). Significantly, this HNE isomer possessed the same relative stereochemistry as the (6R) crotonaldehyde adduct. However, cross-link formation was slow, with chemical equilibrium being attained only after several months at room temperature(51). The discovery that diastereomeric HNE−dG adducts 2 and 3 exist primarily as cyclic hemiacetals 8−11 when placed into duplex DNA provided a rationale for the slow rate of interstrand cross-link formation by stereoisomer 3(59). This has also been observed for the Michael addition of protein nucleophiles to HNE(60). The presence of the hemiacetal effectively masks the reactive aldehyde species necessary for cross-link formation. Kurtz and Lloyd(61) demonstrated that HNE adduct 3 formed conjugates with the tetrapeptide KWKK more rapidly than did the other three stereoisomeric HNE adducts, 2, 4, and 5.

In this work, the cyclic hemiacetal rearrangement products 8 and 10 of the HNE-derived exocyclic 1,N2-dG adducts with (6R,8S,11R) and (6S,8R,11S) stereochemistry(59) were examined with respect to the structure in 5′-d(GCTAGCXAGTCC)-3′·5′-d(GGACTCGCTAGC)-3′, containing the 5′-CpX-3′ sequence (X = HNE−dG), using high field NMR. Both of these stereoisomeric hemiacetals exhibited minimal structural perturbation of the DNA duplex, which in both instances was localized to the site of the adduction at X7·C18 and its neighboring base pair, A8·T17. Both stereoisomeric cyclic hemiacetals were oriented in the minor groove of the DNA duplex. However, the orientations of the cyclic hemiacetals were different for the two stereoisomers. The cyclic hemiacetal 10 derived from the (6S,8R,11S) HNE adduct 3 was oriented in the 5′-direction, while the cyclic hemiacetal 8 derived from the (6R,8S,11R) HNE adduct 2 was oriented in the 3′-direction. Cyclic hemiacetals 8 and 10 represent surrogates for reactive aldehydes 6 and 7 that mediate DNA interstrand cross-linking. From the refined structures of the two cyclic hemiacetals 8 and 10, the conformations of the corresponding diastereomeric aldehydes 6 and 7 were predicted, based upon molecular mechanics calculations. Potential energy minimizations of the DNA duplexes containing the two diastereomeric aldehydes 6 and 7 were consistent with the prediction that the (6R) aldehyde 6 was oriented in the 3′-direction while the (6S) aldehyde 7 was oriented in the 5′-direction. These differences in minor groove orientation suggest a kinetic basis for explaining, in part, the relative abilities of the (6S,8R,11S) and (6R,8S,11R) diastereomeric adducts 2 and 3 to form interchain cross-links in the 5′-CpG-3′ sequence(51).

Materials and Methods

Materials

The oligodeoxynucleotide 5′-d(GGACTCGCTAGC)-3′ was synthesized and purified by anion-exchange chromatography by the Midland Certified Reagent Co. (Midland, TX). The HNE-derived (6R,8S,11R) and (6S,8R,11S) exocyclic 1,N-2′-deoxyguanosine adducts 2 and 3 [which rearrange predominately to 8 and 10 in duplex DNA, respectively(59)] were incorporated into 5′-d(GCTAGCXAGTCC)-3′ (X = HNE−dG) as reported previously[50,51]. The oligodeoxynucleotides were characterized by MALDI-TOF mass spectrometry. Capillary gel electrophoresis and C-18 HPLC were utilized to assess their purities. The oligodeoxynucleotides were desalted by chromatography on Sephadex G-25 (Sigma-Aldrich, St. Louis, MO). The concentrations of the oligodeoxynucleotides were determined by UV absorption at 260 nm, and the extinction coefficients of both sequences were calculated to be 1.12 × 105 L mol−1 cm−1(62). The strands were annealed at a 1:1 stoichiometry 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 then cooled to room temperature. The duplex DNA was purified using DNA grade hydroxylapatite chromatography, 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. The duplexes were also characterized by MALDI-TOF mass spectrometry.

NMR Experiments

NMR experiments were performed at 1H frequencies of 600 and 800 MHz; the data at 800 MHz were collected using a cryogenic probe. Samples were at a strand concentration of 1.0 mM. Samples for the nonexchangeable protons were dissolved in 10 mM NaH2PO4, 100 mM NaCl, and 50 μM Na2EDTA (pH 7.0) to a volume of 280 μL. They were exchanged with D2O and suspended in 280 μL of 99.996% D2O. The pH was adjusted using dilute DCl or NaOD. The temperature was 25 °C. Samples for the observation of exchangeable protons were dissolved in 280 μL of the same buffer containing a 9:1 H2O/D2O mixture (v/v). The temperature was 5 °C. The 1H chemical shifts were referenced to water. Data were processed using FELIX 2000 (Accelrys Inc., San Diego, CA) on LINUX workstations (Dell Inc., Austin, TX). For all experiments, a relaxation delay of 1.5 s was used. The NOESY spectra were recorded with 512 real data in the t2 dimension and 2048 real data in the t1 dimension. The spectra were zero-filled during processing to create a matrix of 1024 × 1024 real points. The TOCSY mixing time was 80 ms for stereoisomer 8 and 100 ms for stereoisomer 10. TOCSY spectra were zero-filled to create a matrix of 1024 × 512 real points. For assignment of exchangeable protons, NOESY experiments used the Watergate solvent suppression scheme(63). The mixing time was 250 ms. For assignment of nonexchangeable protons and the derivation of distance restraints, NOESY experiments used TPPI quadrature detection, and mixing times of 60, 150, 200, and 250 ms were used. The DQF-COSY experiments were performed with TPPI quadrature detection and presaturation of the residual water during the relaxation delay. 1H−31P HMBC spectra[64,65] were recorded at 30 °C. The data matrix consisted of 96 (t1) × 1024 (t2) complex points. The data were Fourier-transformed after zero filling in the t1 dimension, resulting in a matrix size of 128 (D1) × 512 (D2) real points. The 31P chemical shifts were not calibrated.

Experimental Distance and Torsion Angle Restraints

Footprints were drawn around cross-peaks obtained with a mixing time of 250 ms using FELIX 2000. Identical footprints were transferred and fit to the corresponding cross-peaks obtained at the other two mixing times. Cross-peak intensities were determined by volume integrations. These were combined as necessary with intensities generated from complete relaxation matrix analysis of a starting structure to generate a hybrid intensity matrix[66,67]. MARDIGRAS[68-70] iteratively refined the hybrid intensity matrix and optimized agreement between calculated and experimental NOE intensities. The RANDMARDI algorithm carried out 50 iterations for each set of data, randomizing peak volumes within limits specified by the input noise level(70). Calculations were initiated using isotropic correlation times of 2, 3, and 4 ns, and with both A-form and B-form starting structures and the three mixing times, yielding 18 sets of distances. Analysis of these data yielded experimental distance restraints used in subsequent rMD calculations, and the corresponding standard deviations for the distance restraints.

The 2-deoxyribose pseudorotational angles (P) were estimated by examining the 3JHH values of sugar protons(71). J1′−2′ and J1′−2′′ were measured from ECOSY spectra, while the intensities of H2′′−H3′ and H3′−H4′ cross-peaks were determined from DQF-COSY spectra. The data were fit to curves relating the coupling constants to the 2-deoxyribose pseudorotation (P), the sugar pucker amplitude (ϕ), and the percentage of S-type conformation. The pseudorotation and amplitude ranges were converted to the five dihedral angles ν0−ν4. Coupling constants measured from 1H−31P HMBC spectra were applied[72,73] to the Karplus relationship(74) to determine the backbone 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′) backbone angles were calculated from the correlation between ε and ζ in B-DNA.

rMD Calculations

The HNE-adducted duplexes, in either A-form or B-form DNA helical coordinates, were constructed by bonding the stereospecific cyclic hemiacetal forms of the HNE adducts(59) to N2 of G7 using Insight II. The partial charges on the cyclic hemiacetal form of the HNE−dG adduct were obtained from density functional theory (DFT) calculations using a neutral total charge, utilizing the B3LYP/6-31G* basis set and GAUSSIAN(75). To obtain the A-form and B-form starting structures that were used for subsequent restrained molecular dynamics (rMD) calculations, these A-form or B-form modified duplexes were energy-minimized using 200 iterations with the conjugate gradient algorithm, in the absence of experimental restraints.

Distance restraints were divided into classes weighted according to the error assessed in their measurements. Class 1, class 2, class 3, class 4, and class 5 were calculated from completely resolved, somewhat overlapped, slightly overlapped, intermediately overlapped, or heavily overlapped cross-peaks, respectively, which were at least 0.5 ppm from the water resonance or the diagonal line of the spectrum. Class 5 also included all other cross-peaks. NOEs that did not have a distance calculated by MARDIGRAS were estimated by relative peak intensities. The spectroscopic data indicated that the duplexes conserved Watson−Crick base pairing, so empirical restraints preserving Watson−Crick hydrogen bonding and preventing propeller twisting between base pairs were used(76). Empirical backbone and 2-deoxyribose torsion angle restraints derived from B-DNA were used(77). The potential energy wells associated with the dihedral angle restraints were ±30°. The force constants of the restraints were scaled from 3.2 to 32 kcal mol−1 Å−2 during the first 10 ps and were maintained at 32 kcal mol−1 Å−2 for the remainder of the simulations.

Ten sets of randomly seeded rMD calculations (five from A-type and five from B-type DNA starting structures) were conducted using AMBER (version 7.0)(78) and the parm99 force field. The Hawkins, Cramer, Truhlar pairwise generalized Born (GB) model[79,80] was used to simulate implicit waters. The parameters developed by Tsui and Case(81) were used. The cutoff radius for nonbonding interactions was 18 Å. The restraint energy function contained terms describing distance and torsion angle restraints, both in the form of square well potentials. Bond lengths involving hydrogens were fixed with the SHAKE algorithm(82). A 1000-step energy minimization was performed with an integrator time of 1 fs without experimental restraints, followed by a 100000-iteration simulated annealing protocol with an integrator time step of 1 fs. The system was heated to 600 K in 5000 iterations, kept at 600 K for 5000 iterations, and then cooled to 100 K with a time constant of 4.0 ps over 80000 iterations. A final cooling was applied to relax the system to 0 K with a time constant of 1.0 ps over 10000 iterations.

Convergence was assessed for structures having the fewest deviations from the experimental distance and dihedral restraints, the lowest van der Vaals energies, and the lowest overall energies. Finally, the 10 refined structures were energy-minimized for 250 iterations without restraints to yield average structures. CORMA(67) was utilized to calculate the predicted NOE intensities from the structures refined from rMD calculations. Input volumes (intensities) were normalized from the intensities of protons with fixed intranuclear distances (i.e., cytosine H5−H6 and thymine CH3−H6 distances). Random noise was added to all intensities to simulate spectral noise. An isotropic correlation time (τc) of 3 ns was used. The rotation of thymidine CH3 groups was modeled using a three-jump site model(83). A sixth-root residual (R1) factor(84) was calculated for each structure. Helicoidal analysis was carried out with 3DNA(85).

Molecular Modeling

The starting structures were created from the refined structures of the duplexes containing cyclic hemiacetal 8 or 10, using INSIGHT II. The partial charges of aldehydes 6 and 7 arising from HNE−dG adducts 2 and 3 were obtained from density functional theory (DFT) calculations using a neutral total charge, utilizing the B3LYP/6-31G* basis set and GAUSSIAN(75). Potential energy minimization calculations were conducted with AMBER (version 7.0)(78) and the parm99 force field. The pairwise generalized Born (GB) model[79,80] was used to simulate implicit waters. The parameters developed by Tsui and Case(81) were used. The cutoff radius for nonbonding interactions was 18 Å. A 1000-iteration potential energy minimization was performed, using the conjugate gradient algorithm.

Results

NMR Characterization of HNE-Derived Exocyclic 1,N2-dG Adducts and

The Watson−Crick face of the exocyclic 1,N2-dG adducts arising from HNE is blocked, thereby preventing base pairing with dC. Analyses of NMR data(59) indicated that when either exocyclic adduct 2 or 3 (Charts 1 and 2) was placed into duplex DNA opposite cytosine, ring opening to aldehydes occurred. However, in contrast to the corresponding acrolein-derived(86) and crotonaldehyde-derived exocyclic 1,N2-dG(56) adducts, the major forms of the ring-opened species derived from HNE adduct 2 or 3 were not aldehydes when at equilibrium in duplex DNA. Mass spectrometric analyses of oligodeoxynucleotides containing 2 and 3 indicated that they were not hydrates of the aldehyde, but rather diastereomeric sets of cyclic hemiacetals 8 and 9 or 10 and 11 (Chart 3), arising from HNE adduct 2 or 3, respectively(59). In each instance, NMR indicated that the H6 and H8 HNE protons preferred the trans configuration, which in both cases, was presumably driven by steric repulsion from the large substituent groups. Thus, when one starts from adduct 2, cyclic hemiacetal stereoisomer 8 (6R,8S,11R) is the major species at equilibrium and stereoisomer 9 (6R,8R,11R) is the minor species. Likewise, when one starts from adduct 3, cyclic hemiacetal stereoisomer 10 (6S,8R,11S) is the major species and stereoisomer 11 (6S,8S,11S) is the minor species(59). Therefore, these results detail the conformational analyses of cyclic hemiacetal stereoisomers 8 and 10, representing the major species present in the samples of duplex DNA, at equilibrium.

Chart 1

Formation of Exocyclic 1,N2-dG Adducts by HNE

Chart 2

Ring-Opening Chemistry of the HNE-Derived Exocyclic 1,N2-dG Adducts When Placed Opposite dC in Duplex DNA

Chart 3

(A) Numbering Scheme of the 5′-CpG-3′ Duplexes Containing Stereospecific HNE−dG Adducts and (B) Numbering Scheme of the HNE−dG Adducts

Duplex Containing (6R,8S,11R) HNE-Derived Cyclic Hemiacetal Adduct

(a) Nonexchangeable Protons

The sequential NOE assignment was accomplished using standard protocols[87,88]. The sequential NOEs between the aromatic and anomeric protons are displayed in panels A and B of Figure 1. The spectral resolution of this modified duplex was challenging because of a large number of pyrimidine aromatic resonances, e.g., T3 H6, T17 H6, C18 H6, C20 H6, T21 H6, and C24 H6, resonating between 7.35 and 7.45 ppm. Nevertheless, for the modified strand, a complete sequential NOE connectivity was observed. For the complementary strand, the C18 H1′ and T17 H1′ resonances were superimposed, with C18 being the nucleotide complementary to X7. The geminal 2-deoxyribose proton resonances were assigned by utilizing a combination of DQF-COSY and NOESY spectra, based upon the expectation that the H2′′ protons were located further downfield[87,88]. With the exception of several of the H4′ protons, and the stereotopic assignments of the H5′ and H5′′ sugar protons, assignments were made unequivocally. In general, canonical B-DNA distances between the H4′, H5′, and H5′′ protons were used to tentatively assign the H5′ and H5′′ 2-deoxyribose protons. The assignments of the nonexchangeable protons are provided in Table S1 of the Supporting Information.

Figure 1

Expansion of the NOESY spectra for the oligodeoxynucleotide duplexes containing the 5′-CpX-3′ sequence, showing correlations of purine H8 and pyrimidine H6 protons with 2-deoxyribose H1′ protons. (A) Modified strand of the duplex adducted with cyclic hemiacetal 8. (B) Complementary strand of the duplex adducted with cyclic hemiacetal 8. (C) Modified strand of the duplex adducted with cyclic hemiacetal 10. (D) Complementary strand of the duplex adducted with cyclic hemiacetal 10.

(b) Exchangeable Protons

A plot of the region ranging from 12.0 to 14.4 ppm of the NOESY specrtum is shown in Figure 2A. The imino proton resonances were assigned following standard protocols(89). The X7 N9H → C18N4H(s) NOEs were observed, indicating the presence of the X7·C18 pair at the modification site (Figure 3A). A strong NOE cross-peak was observed between the guanine imino proton at the modified base, X7 N9H, and the guanine amino proton at the modified base, X7 N5H. This was also consistent with Watson−Crick base pairing at the adduct site. NOE correlations of some HNE protons with X7 N9H, X7 N5H, and G19 N1H were observed (Figure 3A). The X7 N9H → A8 H2 NOE was also observed, consistent with the intrahelical stacking of the modified nucleotide X7 (Figure 3A). A complete NOE connectivity was obtained, with the exceptions of terminal base pairs G1·C24 and C12·G13, the imino resonances of which were broadened by solvent exchange.

Figure 2

Expansions of NOESY spectra for oligodeoxynucleotide duplexes containing the 5′-CpX-3′ sequence, showing the sequential connectivity of the base imino protons. (A) Duplex adducted with cyclic hemiacetal 8. (B) Duplex adducted with cyclic hemiacetal 10. The T17 N3H resonance for the duplex containing cyclic hemiacetal 10 is broad. It exhibits a weak NOE cross-peak with X7 N9H, but the diagonal peak is missing.

Figure 3

Expansions of the NOESY spectra for oligodeoxynucleotide duplexes containing the 5′-CpX-3′ sequence, showing the conservation of Watson−Crick base pairing. (A) Duplex containing cyclic hemiacetal 8. The NOE cross-peaks were assigned as follows: (a) T17 N3H → A8 H2, (b) X7 N9H → X7 H12α, (c) X7 N9H → X7 H7α, (d) X7 N9H → C18N4H1, (e) X7 N9H → X7 N5H, (f) X7 N9H → A8 H2, (g) X7 N9H → C18N4H2, and (h) X7 N5H → X7 H12α. (B) Duplex containing cyclic hemiacetal 10. The NOE cross-peaks were assigned as follows: (a) T17 N3H → A8 H2, (b) X7 N9H → X7 H7, (c) X7 N9H → C18N4H1, (d) X7 N9H → A8 H2, (e) X7 N9H → X7 N5H, (f) X7 N9H → C18N4H2, and (g) X7 N5H → X7 H7. The two dashed lines indicated by an arrow at the top of the spectrum represent the X7 H12, H13, H14, and H15 resonances, which could not be assigned unequivocally. They exhibit NOE correlations with X7 N9H, X7 N5H, and A8 H2.

(c) HNE Protons

In the DQF-COSY spectrum, a resonance observed at 5.45 ppm exhibited both dipolar and scalar couplings to a resonance observed at 2.13 ppm (Figure 4A). This was assigned as a correlation between X7 H8 and the geminal X7 H7α proton. Another resonance, observed at 3.93 ppm, exhibited scalar coupling to a resonance observed at 2.15 ppm and was assigned as a correlation between X7 H6 and the X7 H7β proton. These assignments were corroborated by NOESY data obtained with a mixing time of 60 ms. The difference in the chemical shifts of the two geminal X7 H7 protons was <0.02 ppm. Both X7 H7 protons exhibited NOE cross-peaks with X7 H6 and X7 H8. The X7 H6 → X7 H11 and X7 H11 → X7 H12(s) correlations, observed in both NOESY and DQF-COSY spectra, were used to assign the resonances of the X7 H11 and X7 H12 protons. The resonances of X7 H12−H16 overlapped. Both H7 protons exhibited strong NOE cross-peaks with H8 and H6. The H6 → H7α correlation was stronger than the H6 → H7β correlation, while the H7α → H8 correlation was weaker than the H7β → H8 correlation. X7 H6 also exhibited a correlation with X7 H11 in both the NOESY and COSY spectra.

Figure 4

Expansions of DQF-COSY spectra of the oligodeoxynucleotide duplexes containing the 5′-CpX-3′ sequence, showing the H1′ → H2′(′) correlations. (A) Duplex containing cyclic hemiacetal 8. (B) Duplex containing cyclic hemiacetal 10. The peaks designated a and b were assigned to the X7 H8 → X7 H7 correlations.

The X7 H11 → X7 H12 correlations in the COSY and NOESY spectra were used to assign X7 H12 protons. The X7 H16 protons were the most upfield; they exhibited correlations with X7 H15 in the COSY and NOESY spectra. The chemical shifts of the geminal H12, H13, H14, and H15 protons were similar. The resonances and NOE cross-peaks of X7 H13 and H14 were assigned using an iterative strategy. NOE peaks associated with the H6, H7, H8, H11, and H16 protons were assigned and converted to distance restraints; rMD calculation then provided a preliminary structure, which was used to evaluate other unassigned NOEs. The final NOE assignments were made following several rounds of iteration. The chemical shifts of the HNE protons and the assigned NOEs are listed in Table 1.

Table 1

Chemical Shifts of HNE Protons of Stereoisomer 8 and Related NOE Cross-Peaks Used as rMD Distance Restraints

proton	δ (ppm)	NOEsa
H6	3.93	X7 H11 (s), X7 H12α (m), X7 H12β (m), X7 H13 (w), A8 H2 (w), A8 H8 (w), A8 H1′ (s)
H7α	2.13	X7 H6 (s), X7 H11 (m), X7 H12α (m), X7 H12β (w), A8 H2 (m), A8 H4′ (w)
H7β	2.15	X7 H6 (s), X7 H11 (m), X7 H12α (m), X7 H12β (w), X7 H13 (w), A8 H2 (m), A8 H4′ (w)
H8	5.45	H7α (s), H7β (s), X7 H6 (w), X7 H11 (w), X7 H12α (m), X7 H12β (w), X7 H13 (w), X7 H15 (w), C18 H1′ (w), C18 H2′′ (w), G19 H1′ (w), G19 H4′ (m), G19 H5′ (m), G19 H5′′ (w)
H11	4.26	X7 H12α (s), X7 H12β (s), X7 H13 (m), X7 H14 (m), X7 H15 (m), X7 H16 (w)
H12α	1.33	X7 H16 (m), X7 H1′ (w), X7 H5′ (w), A8 H4′ (w), A8 H5′ (m), G19 H1′ (w), C20 H1′ (m)
H12β	1.41	X7 H1′ (w), X7 H5′ (w), A8 H4′ (w), A8 H5′ (w), G19 H1′ (w), C20 H1′ (w)
H13	1.27	X7 H16 (m), X7 H1′ (w), X7 H5′ (w), A8 H4′ (w), A8 H5′ (w), C20 H1′ (w)
H14	1.15	X7 H16 (m), X7 H1′ (m), X7 H4′ (m), X7 H5′ (m), X7 H5′′ (w), A8 H3′ (w), A8 H4′ (w), A8 H5′ (w), C20 H1′ (w)
H15	1.20	X7 H16 (s), X7 H1′ (s), X7 H3′ (w), X7 H4′ (m), X7 H5′ (s), X7 H5′′ (m), A8 H3′ (w), A8 H4′ (m), A8 H5′ (s), A8 H5′′ (m), C20 H1′ (m)
H16	0.82	X7 H1′ (m), X7 H3′ (w), X7 H4′ (s), X7 H5′ (m), X7 H5′′ (m), A8 H3′ (w), A8 H4′ (w), A8 H5′ (m), A8 H5′′ (m), C20 H1′ (w)

Letters in parentheses indicate peak intensity: s, strong; m, medium; w, weak.

(d) Chemical Shift Perturbations

The chemical shift perturbations of the nonexchangeable pyrimidine H6, purine H8, and 2-deoxyribose H1′ protons, comparing the modified and the corresponding unmodified duplexes, are presented in panels A and B of Figure 5. Large differences were observed at the adducted base pair X7·C18. As compared to the X7 and C18 2-deoxyribose H1′ protons in the unmodified duplex, in the modified duplex, these H1′ resonances shifted downfield 0.32 and 0.41 ppm, respectively. Few chemical shift perturbations were observed for the pyrimidine H6 and purine H8 protons, indicating that the DNA duplex was minimally disturbed by the presence of the HNE-derived cyclic hemiacetal.

Figure 5

Proton chemical shift perturbations for the oligodeoxynucleotide duplexes containing the 5′-CpX-3′ sequence. (A) Modified strand of the duplex containing cyclic hemiacetal 8. (B) Complementary strand of the duplex containing cyclic hemiacetal 8. (C) Modified strand of the duplex containing cyclic hemiacetal 10. (D) Complementary strand of the duplex containing cyclic hemiacetal 10.

(e) NMR-Derived Distances

A total of 89 NOE cross-peaks associated with the protons of the cyclic hemiacetal moiety protons were assigned (Table 1). Figure 6A shows some of these correlations. Notably, NOE correlations were observed between the geminal X7 H7 as well as X7 H6 protons and A8 H2 and A8 H1′ protons that were in the 3′-direction. In contrast, the G19 H1′ and C20 H1′ protons that were in the 5′-direction exhibited NOE cross-peaks with the X7 H12−H15 protons. These findings suggested that the cyclic hemiacetal moiety was located in the minor groove with the tetrahydrofuran oriented in the 3′-direction and the aliphatic chain oriented in the 5′-direction.

Figure 6

NOE correlations associated with the HNE protons, showing the different orientations of the HNE moieties in the oligodeoxynucleotide duplexes containing the 5′-CpX-3′ sequence. (A) Duplex containing cyclic hemiacetal 8. (B) Duplex containing cyclic hemiacetal 10.

(f) 2-Deoxyribose and Backbone Angle Conformations

The 2-deoxyribose and backbone angle conformations were determined spectroscopically from DQF-COSY and 31P−H3′ HMBC correlations. Evaluation of the DQF-COSY spectrum revealed that the 2-deoxyribose pseudorotations for all nucleotides were either C1′-exo or C2′-endo.

(g) Structure Refinement

The structural refinement involved 527 distance restraints, including 296 intraresidue and 231 interresidue restraints, obtained from the intensities of NOE cross-peaks. In addition, 52 empirical distance restraints defining Watson−Crick base pairing were used to refine the structure of the duplex; their use was predicated upon inspection of the NMR data, which indicated that Watson−Crick base pairing was intact throughout the duplex. Finally, an additional 200 empirical backbone torsion angle restraints were also used for structure refinements; these were based upon inspection of the NMR data, which suggested that the adducted duplex maintained a B-type architecture (Table 2).

Table 2

rMD Restraints and Statistical Analysis of Structures Emergent from rMD Calculations Performed on the Oligodeoxynucleotide Duplex Site-Specifically Modified by Stereoisomer 8

total no. of restraints for rMD calculation	780
no. of experimental NOE distance restraintsa	528
no. of intraresidue NOE restraints	299
no. of interresidue NOE restraints	229
no. of restraints of the HNE unit	89
no. of empirical base pair restraints	52
no. of empirical torsion angle restraints	200
no. of backbone torsion angle restraints	100
no. of sugar torsion angle restraints	100
structure statistics
NMR R-factor (R1x) (×10−2)b	8.29
intraresidue NOEs	7.14
interresidue NOEs	10.0
rmsd of refined structures	0.42

The HNE unit was considered to be a single residue attached to G7 in the rMD calculations.

The 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 (non-zero) and calculated NOE cross-peaks, respectively.

The randomly seeded rMD calculations were performed starting with initial structures, which were created with either A- or B-form geometries. Pairwise rmsd analysis of emergent structures indicated that the calculations converged, irrespective of starting structure (Table 2). The accuracies of the emergent structures were evaluated by comparison of theoretical NOE intensities calculated by complete relaxation analysis for the refined structure to the experimental NOE intensities, which yielded sixth-root residuals (R1). This residual was less than 0.1 for the modified duplex (Table 2), and the inter- and intranucleotide residuals for individual nucleotides were less than 0.15 (Figure 7A,B), indicating that the refined structures provided an accurate depiction of the data.

Figure 7

Residue-by-residue sixth-root residuals (R1) of the oligodeoxynucleotide duplexes containing the 5′-CpX-3′ sequence, obtained from CORMA back calculation. (A) Modified strand of the duplex containing cyclic hemiacetal 8. (B) Complementary strand of the duplex containing cyclic hemiacetal 8. (C) Modified strand of the duplex containing cyclic hemiacetal 10. (D) Complementary strand of the duplex containing cyclic hemiacetal 10.

(h) Analysis of rMD Structures

The refined structures are overlaid in Figure 8A, and an expanded view of the adducted region of the average structure is shown in Figure 9A. The modified duplex maintained B-type DNA geometry. All nucleotides maintained the anti conformation about the glycosyl torsion angle. Few torsion angle differences were observed as compared to ideal B-DNA. The 2-deoxyribose pseudorotations were consistently either C1′-exo or C2′-endo. The HNE-derived cyclic hemiacetal moiety was folded in the minor groove. The tetrahydrofuran was oriented in the 3′-direction, and the aliphatic chain was oriented in the 5′-direction. Panels A and B of Figure 10 show the base stacking of the adduct region.

Figure 8

Refined structures obtained from rMD calculations for the oligodeoxynucleotide duplexes containing the 5′-CpX-3′ sequence. (A) Duplex containing cyclic hemiacetal 8. (B) Duplex containing cyclic hemiacetal 10. Blue sticks represent nucleotides and red sticks the HNE moiety.

Figure 9

Adducted regions of the oligodeoxynucleotide duplexes containing the 5′-CpX-3′ sequence, viewed from the minor grooves. (A) Average refined structure emergent from rMD calculations of the duplex containing cyclic hemiacetal 8. (B) Predicted structure, obtained by molecular mechanics calculations, of the duplex containing aldehyde 6. The dashed arrows indicate the spatial relationship between the reactive aldehyde carbon and the exocyclic amino nitrogen of cross-linking target G19 (7.1 Å). (C) Average refined structure emergent from rMD calculations of the duplex containing cyclic hemiacetal 10. (D) Predicted structure, obtained by molecular mechanics calculations, of the duplex containing aldehyde 7. The cyan sticks represent nucleotides. The blue sticks represent the two amino nitrogens of X7 and G19. The white, green, and red sticks represent hydrogens, carbons, and oxygens of the HNE moiety, respectively. The dashed arrows indicate the spatial relationship between the reactive aldehyde carbon and the exocyclic amino nitrogen of cross-linking target G19 (4.4 Å).

Figure 10

Base stacking of the adduct region for oligodeoxynucleotide duplexes containing the 5′-CpX-3′ sequence. (A) Duplex containing cyclic hemiacetal 8. Stacking of base pair C6·G19 above base pair X7·C18. (B) Duplex containing cyclic hemiacetal 8. Stacking of base pair X7·C18 above base pair A8·T17. (C) Duplex containing cyclic hemiacetal 10. Stacking of base pair C6·G19 above base pair X7·C18. (D) Duplex containing cyclic hemiacetal 10. Stacking of base pair X7·C18 above base pair A8·T17. For both duplexes containing either cyclic hemiacetal 8 or 10, base pairs C6·G19, X7·C18, and A8·T17 adopt Watson−Crick pairing.

Molecular Modeling of the Duplex Containing (6R,8S,11R) HNE-Derived Aldehydic Adduct

The refined structure of the cyclic hemiacetal 8 was then converted to the corresponding aldehyde 6 that represents the reactive species necessary for DNA cross-link formation. At equilibrium, this species was not present in sufficient quantity to enable detailed structural refinement. Consequently, a molecular mechanics approach, using potential energy minimization, was employed to predict the conformation of aldehyde 6 in the minor groove of the DNA duplex. Figure 9B shows the predicted structure of the duplex containing aldehyde 6, with the HNE moiety remaining in the minor groove, and the aldehyde group oriented in the 3′-direction.

Duplex Containing (6S,8R,11S) HNE-Derived Cyclic Hemiacetal Adduct

The resonances of the nonexchangeable protons were assigned using standard approaches[87,88]. The sequential NOEs between the aromatic and anomeric protons are displayed in panels C and D of Figure 1. Similar to that of the duplex containing adduct 8, a complete sequential NOE connectivity was observed. For the complementary strand, a complete sequential NOE connectivity was also observed. The 2-deoxyribose sugar proton resonances were assigned by utilizing a combination of DQF-COSY and NOESY spectra. Compared with the H2′ protons, the geminal H2′′ protons are located downfield[87,88]. With the exception of several of the H4′ protons, and the stereotopic assignments of the H5′ and H5′′ sugar protons, assignments were made unequivocally. In general, canonical B-DNA distances between the H4′, H5′, and H5′′ protons were used to tentatively assign the H5′ and H5′′ 2-deoxyribose protons. The assignments of the nonexchangeable protons are provided in Table S1 of the Supporting Information.

Figure 2B shows the region of the NOESY spectrum showing the NOEs between the imino protons. The T17 N3H imino proton appeared as a broad peak at 14.2 ppm; it exhibited a weak cross-peak with the X7 N9H imino proton. This assignment was supported by observation of an NOE cross-peak to A8 H2 (Figure 3B). NOE cross-peaks for base pairs C2·G23, T3·A22, A4·T21, G5·C20, C6·G19, X7·C20, A8·T17, G9·C16, T10·A15, and C11·G14 were observed. X7 N9H had NOE correlations with C18N4H(s) and A8 H2 (Figure 3B). Notably, a strong X7 N9H → X7 N5H correlation was observed, similar to that of the duplex containing adduct 8. NOE correlations of some HNE protons with X7 N9H and X7 N5H were observed (Figure 3B).

The HNE proton resonances were assigned on the basis of a combination of COSY, DQF-COSY, TOCSY, and NOESY (60 ms mixing time) spectra and rMD calculations. The resonances of two X7 H7 protons were not split. The X7 H8 → X7 H7 correlation was observed in the H1′ → H2′(′) correlation region in the DQF-COSY spectrum (Figure 4B). The X7 H6 → X7 H7 correlation appeared at 4.55/2.17 ppm. These assignments were evidenced by the NOESY spectrum (60 ms mixing time), in which the X7 H7 protons exhibited strong cross-peaks with both X7 H6 and X7 H8. X7 H6 also correlated with X7 H11 in both NOESY and COSY spectra, and the X7 H11 → X7 H12 correlations in the COSY and NOESY data were used to assign the X7 H12 protons. The X7 H16 protons were the most upfield and correlated strongly with X7 H15 in both COSY and NOESY spectra. The resonances of X7 H13 and X7 H14 were assigned using the same iterative strategy that was described for the assignment of the HNE protons in stereoisomer 8. Unfortunately, for stereoisomer 10, the chemical shifts of protons H12−H15 were less resolved than for stereoisomer 8. Thus, it was not possible to unambiguously assign all NOEs arising from these protons. The chemical shifts of the HNE protons and the assigned NOEs from stereoisomer 10 are listed in Table 3.

Table 3

Chemical Shifts of HNE Protons of Stereoisomer 10 and Related NOE Cross-Peaks Used as rMD Distance Restraints

proton	δ (ppm)	NOEsa
H6	4.55	X7 H11 (s), X7 H12α (w), X7 H12β (m), X7 6A (w), X7 H1′ (w), A8 H1′ (w), G19 H1′ (w)
H7	2.17	X7 H6 (s), X7 H11 (m), X7 H12α (s), X7 H12β (s), X7 H13 (s), X7 H1′ (w), C18 H1′ (w), G19 H1′ (s)
H8	5.43	X7 H7 (s), X7 H6 (m), X7 H11 (m), X7 H12α (w), X7 H12β (m), X7 H13 (m), C18 H1′ (w), G19 H1′ (w), G19 H4′ (m), G19 H5′ (w), G19 H5′′ (w), C20 H5′ (w)
H11	4.23	X7 H12α (s), X7 H12β (s), X7 H13 (s), A8 H2 (s)
H12α	1.34	A8 H2 (m), A8 H1′ (m), A8 H4′ (m), G9 H2′ (w), G9 H2′′ (w)
H12β	1.45	A8 H2 (m), A8 H1′ (m), A8 H4′ (m)
H13	1.36	X7 H16 (s), A8 H2 (m), A8 H1′ (m), A8 H4′ (m), G9 H1′ (m), G9 H4′ (m), T17 H1′ (m), C18 H1′ (m)
H14	1.45	X7 H16 (m), A8 H2 (m), A8 H4′ (m), G9 H1′ (m), G9 H2′ (w), G9 H2′′ (w), G9 H4′ (m), T17 H1′ (w), C18 H1′ (m)
H15	1.38	X7 H16 (s), A8 H2 (w), G9 H1′ (m), G9 H4′ (m), G9 H2′′ (w), T17 H1′ (w), C18 H1′ (w)
H16	0.96	G9 H1′ (w), G9 H5′ (w), G9 H4′ (w), G9 H5′′ (w), T10 H5′ (w), T10 H5′′ (w), C18 H1′ (w), C18 H4′ (m), C18 H5′ (w), C18 H5′′ (w)

Letters in parentheses indicate peak intensity: s, strong; m, medium; w, weak.

The chemical shift comparisons of the nonexchangeable pyrimidine H6, purine H8, and 2-deoxyribose H1′ protons versus those of the corresponding unmodified duplex are presented in panels C and D of Figure 5. Large perturbations were observed at the adducted base pair X7·C18. As compared to the G7 and C18 2-deoxyribose H1′ protons in the unmodified duplex, the X7 and C18 2-deoxyribose H1′ resonances are shifted downfield 0.42 and 0.31 ppm, respectively. Few chemical shift perturbations were observed for the pyrimidine H6 and purine H8 protons.

A total of 73 NOE cross-peaks associated with the HNE-derived cyclic hemiacetal protons were converted to distance restraints (Table 3). Figure 6B shows some of these distance restraints. Notably, NOE correlations were observed between X7 H12−H15 protons and A8 H2, as well as A8 H1′, G9 H1′, and T17 H1′, in the 3′-direction. On the other hand, X7 H7 exhibited a strong NOE with G19 H1′, in the 5′-direction. These NOEs suggested that the cyclic hemiacetal moiety was located in the minor groove, with the tetrahydrofuran oriented in the 5′-direction and the aliphatic chain oriented in the 3′-direction.

2-Deoxyribose and backbone angle conformations were determined from DQF-COSY and 31P−H3′ HMBC correlations. Evaluation of the DQF-COSY spectrum revealed that the pseudorotations of the 2-deoxyribose rings for all residues were either C1′-exo or C2′-endo.

The structural refinement employed 506 distance restraints, including 302 intraresidue and 204 interresidue restraints that were calculated from the intensities of the NOE cross-peaks. An additional 52 empirical distance restraints derived from Watson−Crick base pair interactions, as predicted by the NMR data, were used to refine the structure. Since the T17 N3H imino proton resonance was broad, a weak restraint was used for the A8·T17 Watson−Crick base pair. Finally, on the basis of analysis of the NMR data, 200 empirical backbone torsion angle restraints derived from B-DNA were also used (Table 4).

Table 4

rMD Restraints and Statistical Analysis of Structures Emergent from rMD Calculations Performed on the Oligodeoxynucleotide Duplex Site-Specifically Modified by Stereoisomer 10

total no. of restraints for rMD calculation	760
no. of experimental NOE distance restraintsa	508
no. of intraresidue NOE restraints	302
no. of interresidue NOE restraints	206
no. of restraints of the HNE unit	73
no. of empirical base pair restraints	52
no. of empirical torsion angle restraints	200
no. of backbone torsion angle restraints	100
no. of sugar torsion angle restraints	100
structure statistics
NMR R-factor (R1x) (×10−2)b	7.87
intraresidue NOEs	6.75
interresidue NOEs	9.74
rmsd of refined structures	0.53

The HNE unit was considered to be an single residue attached to G7 in the rMD calculations.

The 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 (non-zero) and calculated NOE cross-peaks, respectively.

(h) rMD Computation

The randomly seeded rMD calculations were performed, starting from both initial A- and B-form geometries. Pairwise rmsd comparisons of the emergent structures indicated that the calculations converged, irrespective of starting structure (Table 4). The accuracies of the emergent structures were evaluated by comparison of theoretical NOE intensities calculated by complete relaxation analysis to the experimental NOE intensities, to yield sixth-root residuals (R1). The residual for the duplex was less than 0.1 (Table 4), and the inter- and intranucleotide residuals for each nucleotide were less than 0.15 (Figure 7C,D), indicating that the structures provided an accurate depiction of the data.

(i) Analysis of rMD Structures

The refined structures are overlaid in Figure 8B, and an expanded view of the adduct region of the average structure is shown in Figure 9C. The modified duplex remained in the B-type geometry. All nucleotides maintained the anti conformation about the glycosyl bond. Only minor torsion angle differences were observed as compared to canonical B-DNA. The 2-deoxyribose pseudorotations were consistently either C1′-exo or C2′-endo. The HNE-derived cyclic hemiacetal was folded in the minor groove of the duplex. In agreement with the NMR data, the tetrahydrofuran moiety was oriented in the 5′-direction and the aliphatic chain was oriented in the 3′-direction. Panels C and D of Figure 10 show the base stacking of the adduct region. The base stacking interactions were comparable to those observed for the unmodified duplex, although base pair A8·T17 exhibited an increased twist.

Molecular Modeling of the Duplex Containing (6S,8R,11S) HNE-Derived Aldehydic Adduct

The refined structure of the cyclic hemiacetal 10 was then converted to the corresponding aldehyde 7, which represented the reactive species necessary for DNA cross-link formation. At equilibrium, this species was not present in sufficient quantity to enable detailed structural refinement. Consequently, a molecular mechanics approach, using potential energy minimization, was employed to predict the conformation of aldehyde 7 in the minor groove of the DNA duplex. Figure 9D shows the predicted structure of the duplex containing aldehyde 7. The HNE moiety remained in the minor groove. The aldehyde group was oriented in the 5′-direction.

Discussion

Interest in the mechanism of HNE-mediated DNA interstrand cross-linking is based upon the observations that HNE is cytotoxic[19-23] and mutagenic[29,42-46] and exocyclic 1,N2-dG adducts 2−5 have been detected in cellular DNA[30-36]. Moreover, as compared to the corresponding acrolein- and crotonaldehyde-induced exocyclic 1,N2-dG adducts[53,57], diastereomeric adduct 3 forms high levels of DNA interstrand cross-links at equilibrium, suggesting that the reversible cross-links associated with adduct 3, once formed, are stable in duplex DNA(51).

Conformations of the Duplexes Containing Stereoisomeric HNE-Derived Cyclic Hemiacetals and

The stereoisomeric HNE-derived cyclic hemiacetals 8 and 10 were each accommodated within the minor groove of the DNA duplex (Figure 8). This conclusion emerged from consideration of a number of lines of evidence obtained from NMR data. Thus, the sequential NOE connectivity of both duplexes was complete for both the modified and complementary strands (Figure 1). In both instances, large chemical shift perturbations were observed only for the minor groove X7 and C18 H1′ protons. The observation of NOE cross-peaks between HNE protons and A8 H2 and A8 H1′ minor groove protons indicated that the HNE-derived cyclic hemiacetals were positioned in the proximity of these protons (Tables 1 and 3). In contrast, there were no large chemical shift perturbations for the aromatic pyrimidine H6 or purine H8 protons, suggesting minimal changes to base stacking arrangements in the two modified duplexes (Figure 5). NOE correlations arising from Watson−Crick base pairs were observed for all nonterminal base pairs in both duplexes, also consistent with maintenance of Watson−Crick hydrogen bonding at the lesion sites (Figures 2 and 3). The exocyclic amine N5H protons of both stereoisomers 8 and 10 were observed as sharp resonances and showed strong NOEs to the imino N9H protons. At the adducted base pair, the base stacking of the duplex containing cyclic hemiacetal 8 was improved as compared to that of the duplex containing cyclic hemiacetal 10 (Figure 10). The A8·T17 base pair in the duplex containing cyclic hemiacetal 10 adopted distorted Watson−Crick base pairing. This may partially explain the 4 °C reduction in melting temperature for this duplex(51). The NOESY spectrum of the duplex with stereoisomer 10 exhibited broadening of the T17 N3H imino proton (Figure 2B), and there were few NOEs assigned to this proton. This suggested that distortion of the A8·T17 base pair resulted in a faster exchange of T17 N3H with solvent. On the other hand, the resonance of the T17 N3H imino proton in the duplex with cyclic hemiacetal 8 was sharp, and all anticipated NOEs arising from Watson−Crick base pairing were observed.

Role of Stereochemistry in DNA Interstrand Cross-Link Formation in the 5′-CpG-3′ Sequence

Among four stereoisomers of the exocyclic 1,N2-dG HNE adduct, only adduct 3, with (6S,8R,11S) stereochemistry, forms DNA interchain cross-linking in the 5′-CpG-3′ sequence(51). In duplex DNA, the cyclic hemiacetal stereoisomer 10 is the major species derived from exocyclic 1,N2-dG stereoisomers 3(59). Mechanistically, cross-linking requires adduct 7, the aldehyde form of adduct 10. Significantly, the orientations of the HNE-derived cyclic hemiacetals 8 and 10 differ (Figure 9). The NOE studies indicate that the tetrahydrofuran ring of stereoisomer 8 is directed toward the 3′-direction, while the tetrahydrofuran ring of stereoisomer 10 is directed toward the 5′-direction. The aliphatic chain of stereoisomer 8 exhibits NOEs with protons in the 5′-direction, and the tetrahydrofuran subunit correlates with protons in the 3′-direction (Figure 6). On the other hand, the aliphatic chain of stereoisomer 10 has NOEs with 3′-direction protons, and the tetrahydrofuran subunit has NOEs with 5′-direction protons (Figure 6). The presence of the long aliphatic chain within the minor groove suggests that the rotation of the HNE-derived cyclic hemiacetals around the X7 N5−X7 C6 bond is likely to be restrained. Consequently, to the extent that the cyclic hemiacetals open to unmask the corresponding aldehydes 6 and 7, the latter are anticipated to adopt orientations similar to those of the respective cyclic hemiacetals 8 and 10, from which they are derived. Therefore, the (6S,8R,11S) stereoisomer of the HNE-derived exocyclic 1,N2-dG adduct 3, which exists predominantly in duplex DNA as cyclic hemiacetal 10, is positioned to facilitate interstrand cross-linking in the 5′-CpG-3′ sequence. In contrast, the (6R,8S,11R) stereoisomer of the HNE-derived exocyclic 1,N2-dG adduct 2, which exists in duplex DNA predominantly as cyclic hemiacetal 8, is not positioned to facilitate interstrand cross-link formation. Molecular modeling of the respective aldehydes, 6 and 7, is consistent with this conclusion (Figure 9B,D). Aldehyde 6, which is in equilibrium with cyclic hemiacetals 8 and 9, is predicted to be oriented in the 3′-direction, whereas aldehyde 7, which is in equilibrium with cyclic hemiacetals 10 and 11, is predicted to be oriented in the 5′-direction. Thus, aldehyde 7, arising from exocyclic 1,N2-dG adduct 3, is predicted to be proximate to the C6·G19 base pair, facilitating formation of the interstrand cross-link in the 5′-CpG-3′ sequence context.

Comparison to the (6R)- and (6S)-Crotonaldehyde-Derived Exocyclic 1,N2-dG Adducts

Kozekov et al.(53) reported that the (6R) configuration of the crotonaldehyde-derived exocyclic 1,N2-dG adduct produced a greater percentage of DNA interstrand cross-links than did the (6S) configuration. Significantly, at the C6 position, the relative stereochemistry of the (6R) crotonaldehyde-derived exocyclic 1,N2-dG adduct corresponds to the (6S,8R,11S) stereochemistry of the HNE-derived exocyclic 1,N2-dG adduct 3. Thus, we conclude that the cyclic hemiacetal arising in duplex DNA from HNE-derived adduct 3 facilitates interstrand cross-linking for the same reason that the (6R)-crotonaldehyde-derived adduct does(56); it places the requisite aldehyde in the minor groove proximal to the cross-linking target in the 5′-CpG-3′ sequence context. In contrast, cyclic hemiacetal 8, arising from the (6R,8S,11R) stereochemistry of the HNE-derived exocyclic 1,N2-dG adduct 2, places the requisite aldehydic species in the minor groove distal to the cross-linking target in the 5′-CpG-3′ sequence context. Again, this is similar to what was observed for the (6S)-crotonaldehyde-derived adduct, in which the aldehyde oriented toward the A8·T17 base pair, distal to the targeted C·G base pair(90). In contrast to HNE, the crotonaldehyde alkyl chain is small with regard to the width of the minor groove, which allows the aldehydic form of the (6S)-crotonaldehyde adduct to transiently undergo conformational reorientation in which the aldehyde orients toward the cross-linking target C6·G19 base pair. This probably explains why a small amount of cross-link was observed (<5% cross-link) for the (6S)-crotonaldehyde-derived adduct(53). However, the reduced form of the (6S)-crotonaldehyde-derived cross-link was less stable than the favored (6R)-crotonaldehyde-derived cross-link(91), consistent with modeling studies(56). Provided the interchain cross-link was induced by cyclic hemiacetal 8, we anticipate that the long HNE aliphatic chain should induce a greater destabilization of the DNA duplex, as compared to the (6S)-crotonaldehyde-derived adduct. Consequently, structural analysis of the cross-linked species arising from the (6R,8S,11R) and (6S,8R,11S) HNE-derived adducts 2 and 3 will be of considerable interest.

Biological Implications

In light of the observation that the (6S,8R,11S) HNE-derived adduct 3 forms DNA interstrand cross-links in 5′-CpG-3′ DNA sequences in vitro(51), we anticipate that it will also form interstrand cross-links in vivo. It will now be of considerable interest to search for this reversible HNE-derived interstrand cross-link in cellular DNA. It occurs specifically at 5′-CpG-3′ sequences, and only for HNE-derived diastereomeric adduct 3, and is consequently expected to be present at low levels in vivo, challenging the limits of detection by mass spectrometry[92-96]. On the other hand, the potential biological consequences arising from low levels of this interstrand DNA cross-link may be of considerable genotoxic significance.

Conclusions

The solution structures of the stereoisomeric cyclic hemiacetals arising from the (6R,8S,11R) and (6S,8R,11S) HNE-derived exocyclic 1,N2-dG adducts 2 and 3 were obtained in a DNA duplex containing the 5′-CpG-3′ sequence motif in which adduct 3, but not adduct 2, forms interstrand cross-links. The orientations of the cyclic hemiacetal groups within the minor groove differ for the two diastereoisomers. The tetrahydrofuran ring of cyclic hemiacetal 10, arising from adduct 3 with (6S,8R,11S) stereochemistry, masking the aldehydic species necessary for cross-link formation, is oriented in the 5′-direction toward base pair C6·G19, while the tetrahydrofuran ring of cyclic hemiacetal 8, arising from adduct 2 with (6R,8S,11R) stereochemistry, masking the aldehydic species necessary for cross-link formation, is oriented in the 3′-direction toward base pair A8·T17. Thus, HNE-derived adduct 3 with (6S,8R,11S) stereochemistry facilitates formation of interstrand cross-links, whereas HNE-derived adduct 2 with (6R,8S,11R) stereochemistry does not form interstrand cross-links in DNA.