Formation of a N2-dG:N2-dG carbinolamine DNA cross-link by the trans-4-hydroxynonenal-derived (6S,8R,11S) 1,N2-dG adduct.

Michael addition of trans-4-hydroxynonenal (HNE) to deoxyguanosine yields diastereomeric 1,N(2)-dG adducts in DNA. When placed opposite dC in the 5'-CpG-3' sequence, the (6S,8R,11S) diastereomer forms a N(2)-dG:N(2)-dG interstrand cross-link [Wang, H.; Kozekov, I. D.; Harris, T. M.; Rizzo, C. J. J. Am. Chem. Soc.2003, 125, 5687-5700]. We refined its structure in 5'-d(G(1)C(2)T(3)A(4)G(5)C(6)X(7)A(8)G(9)T(10)C(11)C(12))-3'·5'-d(G(13)G(14)A(15)C(16)T(17)C(18)Y(19)C(20)T(21)A(22)G(23)C(24))-3' [X(7) is the dG adjacent to the C6 carbon of the cross-link or the α-carbon of the (6S,8R,11S) 1,N(2)-dG adduct, and Y(19) is the dG adjacent to the C8 carbon of the cross-link or the γ-carbon of the HNE-derived (6S,8R,11S) 1,N(2)-dG adduct; the cross-link is in the 5'-CpG-3' sequence]. Introduction of (13)C at the C8 carbon of the cross-link revealed one (13)C8→H8 correlation, indicating that the cross-link existed predominantly as a carbinolamine linkage. The H8 proton exhibited NOEs to Y(19) H1', C(20) H1', and C(20) H4', orienting it toward the complementary strand, consistent with the (6S,8R,11S) configuration. An NOE was also observed between the HNE H11 proton and Y(19) H1', orienting the former toward the complementary strand. Imine and pyrimidopurinone linkages were excluded by observation of the Y(19)N(2)H and X(7) N1H protons, respectively. A strong H8→H11 NOE and no (3)J((13)C→H) coupling for the (13)C8-O-C11-H11 eliminated the tetrahydrofuran species derived from the (6S,8R,11S) 1,N(2)-dG adduct. The (6S,8R,11S) carbinolamine linkage and the HNE side chain were located in the minor groove. The X(7)N(2) and Y(19)N(2) atoms were in the gauche conformation with respect to the linkage, maintaining Watson-Crick hydrogen bonds at the cross-linked base pairs. A solvated molecular dynamics simulation indicated that the anti conformation of the hydroxyl group with respect to C6 of the tether minimized steric interaction and predicted hydrogen bonds involving O8H with C(20)O(2) of the 5'-neighbor base pair G(5)·C(20) and O11H with C(18)O(2) of X(7)·C(18). These may, in part, explain the stability of this cross-link and the stereochemical preference for the (6S,8R,11S) configuration.

Introduction

trans-4-Hydroxynonenal (HNE) is produced from the metabolism of membrane lipids.(1) It is also the major peroxidation product of ω-6 polyunsaturated fatty acids in vivo.[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] HNE is 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)

HNE induces the SOS response in Escherichia coli, suggesting that it is also genotoxic.(18) Chromosomal aberrations have been observed upon exposures to HNE in mammalian, including human, cells.[19−23] In mammalian cells, the genotoxicity of HNE depends upon glutathione levels, which modulate levels of HNE-DNA adducts.[24−26] Michael addition of the N2-amino group of 2′-deoxyguanosine to HNE gives four diastereomeric 1,N2-dG adducts 1–4,[27−29] which have been detected in cells.[30−36] Alternatively, oxidation of HNE to 2,3-epoxy-4-hydroxynonanal, and further reaction with nucleobases, affords etheno adducts.[37−41]

Wang et al.[42,43] synthesized the four stereoisomers of the 1,N2-dG adduct (1–4, Chart 1) and incorporated them into 5′-d(GCTAGCZAGTCC)-3′·5′-d(GGACTCGCTAGC)-3′, in which Z denotes the HNE-dG adduct. Of the four diastereomeric adducts, only (6S,8R,11S) diastereomer 1 forms interstrand cross-links in the 5′-CpG-3′ sequence. At equilibrium, cross-linking by diastereomer 1 reached >85%.(43) However, it required several months to attain equilibrium at 37 °C.(43) The discovery that when placed into DNA diastereomeric HNE-dG adducts 1 and 3 rearrange to the cyclic hemiacetals provided a rationale for the slow rate of cross-link formation.(44) The hemiacetal effectively masks the reactive aldehyde necessary for cross-link formation.

Chart 1

The major cyclic hemiacetal rearrangement products from HNE-dG adducts 1 and 3 were subsequently examined as to structure in this duplex.(45) Both oriented in the minor groove of DNA. However, the cyclic hemiacetal rearranged from adduct 1 oriented in the 5′-direction, while the cyclic hemiacetal rearranged from adduct 3 oriented in the 3′-direction.[45,46] Molecular mechanics calculations predicted that the N2-dG aldehyde having 6R configuration oriented in the 3′-direction, while the N2-dG aldehyde having 6S configuration oriented in the 5′-direction.[45,46] These differences suggested a kinetic basis to explain, in part, the relative abilities of the (6S,8R,11S) and (6R,8S,11R) diastereomeric adducts 1 and 3 to form cross-links in the 5′-CpG-3′ sequence (Schemes 1 and 2).(43)

Scheme 1

Formation of the N2-dG:N2-dG Cross-link by the Oligodeoxynucleotide Duplex Containing the HNE Adduct 1 in the 5′-CpG-3′ Sequence

Scheme 2

(A) Numbering Scheme of the Duplex Containing (6S,8R,11S) 1,N2-dG Adduct 1;, and (B) Numbering Scheme of the (6S,8R,11S) Carbinolamine Cross-link 8,,

The 13C labeled carbon is in red.

Z7 represents the HNE adduct 1.

X7 and Y19 represent the N2-cross-linked deoxyguanosines; the atoms are numbered as for deoxyguanosine.

The atom numbers of the HNE linkage are consistent with those in the HNE adduct 1.

In the present work, the structure of this cross-link has been refined in 5′-d(G1C2T3A4G5C6X7A8G9T10C11C12)-3′·5′-d(G13G14A15C16T17C18Y19C20T21A22G23C24)-3′ [X7 is the dG adjacent to the C6 carbon of the cross-link (the α-carbon of the HNE-derived (6S,8R,11S) 1,N2-dG adduct 1), and Y19 is the dG adjacent to the C8 carbon of the cross-link (the γ-carbon of the HNE-derived (6S,8R,11S) 1,N2-dG adduct 1); the cross-link is in the 5′-CpG-3′ sequence]. 13C HSQC NMR reveals one 13C8→H8 correlation, indicating that the cross-link exists predominantly as a single species, identified as a (6S,8R,11S) carbinolamine linkage 8. NOE data indicate that the (6S,8R,11S) carbinolamine linkage 8 and the HNE moiety are oriented in the minor groove. The X7N2 and Y19N2 atoms are in the gauche-conformation with respect to the linkage, maintaining Watson–Crick hydrogen bonds at the cross-linked base pairs. Solvated molecular dynamics simulations indicate that the anti conformation of the hydroxyl group with respect to C6 of the tether minimizes steric interaction and predict a hydrogen bond between the hydroxyl group and C20O2 of the 5′-neighbor base pair G5·C20. As well, the C11 hydroxyl predominantly formed a hydrogen bond with C18O2. These hydrogen bonds may, in part, explain the stability of this cross-link and the preference for the (6S,8R,11S) configuration 8.

Results

Formation of the Cross-link

The oligodeoxynucleotide 5′-d(GCTAGCZAGTCC)-3′ [Z = (6S,8R,11S) HNE adduct 1] was annealed with the complementary strand 5′-d(GGACTCGCTAGC)-3′ at pH 7.0. The duplex was maintained at 37 °C for 3 months at pH 7.3. This afforded the N2-dG:N2-dG cross-link, as reported.(43) The presence of the cross-link was confirmed by reverse-phase HPLC analysis (Figure S1 in the Supporting Information). The cross-link was also characterized by MALDI-TOF mass spectrometry, which showed the cross-link at m/z of 7428.2 [calculated for imine cross-link 10 (M – 1): 7429.0] (Figure S2 in the Supporting Information). The intensities of the cytosine H5–H6 NMR scalar couplings were used to evaluate the extent of the DNA cross-linking reaction (Figure S3 in the Supporting Information). HPLC analysis suggested the duplex was >75% cross-linked.

NMR Resonance Assignments

The cross-linked sample was used for NMR experiments without further purification. The spectra were of high quality and suitable for structural analysis. The assignments of the nonexchangeable protons of the nucleotides were accomplished using standard protocols.[47,48] For the 5′-d(G1C2T3A4G5C6X7A8G9T10C11C12)-3′ strand, complete NOE connectivity was observed between the aromatic and anomeric protons (Figure S4 in the Supporting Information). A small cross peak was assigned to the C6 H1′→X7 H8 correlation. Complete NOE connectivity was also observed for the 5′-d(G13G14A15C16T17C18Y19C20T21A22G23C24)-3′ strand. With the exceptions of several of the H4′ protons, and the stereotopic assignments of the H5′ and H5″ sugar protons, all other 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″ deoxyribose protons. The chemical shifts of the nonexchangeable DNA protons are collected in Table S1 of the Supporting Information. As compared to the corresponding unmodified duplex, remarkable changes in chemical shifts were observed for protons located in the cross-linked region, suggesting perturbation of the cross-linked and the flanking base pairs (Figure S5 in the Supporting Information).

The imino proton resonances of Watson–Crick base paired guanines and thymines were also assigned following standard protocols.(49) The imino proton resonances were observed for all guanines and thymines, and a complete NOE connectivity was obtained, with the exceptions of the terminal guanines G1 and G13, the resonances of which were broadened by solvent exchange (Figure S6 in the Supporting Information). The X7 N1H→C18NH(s) and Y19 N1H→C6NH(s) NOEs were observed, indicating the presence of the C6·Y19 and X7·C18 pairs at the cross-linked site. Strong X7 N1H→X7N2H and Y19 N1H→Y19N2H NOEs were observed, also consistent with Watson–Crick base pairing at the cross-linked site. The X7 N1H→A8 H2 NOE was also observed, consistent with the intrahelical stacking of the modified nucleotide X7.

The assignments of HNE protons were made by a combination of 1H–1H NOESY (60 ms) (Figure 1), 1H–1H DQF-COSY, and 1H–1H TOCSY experiments. The HNE H6 and H8 protons were assigned by inspection of the H6→X7N2H and H8→Y19N2H NOEs. Both of these protons exhibited scalar and dipolar coupling with the geminal HNE H7 protons. H6 also exhibited scalar and dipolar coupling with H11. H11 exhibited scalar and dipolar coupling with the geminal H12 protons. The protons in the HNE side-chain were partially assigned sequentially based on the scalar and dipolar couplings from H12→H13→H14→H15→H16. The geminal H7 and H11 protons exhibited strong scalar and dipolar couplings. The absolute configurations of the geminal H7 protons were determined by their NOEs with H8. H7β, which was in the trans configuration with respect to H8, showed a smaller dipolar coupling with H8 than did H7α. The stereotopic assignments of the geminal H12 protons could not be unequivocally determined. The intensities of the H11→H12α and H11→H12β NOEs appeared equal, indicating H11 was in the gauche conformation with respect to both protons. A number of NOEs between these HNE protons and the DNA protons X7 N1H, X7N2H, Y19 N1H, and Y19N2H were observed. The chemical shifts of the HNE protons and the NOEs used for the rMD calculations are listed in Table 1.

Figure 1

Expansions of the 1H–1H NOESY spectrum obtained at a mixing time of 60 ms of the cross-link showing the assignments of some HNE protons. The strong H8→H11 NOE (peak a) implies the cross-link exists as carbinolamine 8.

Table 1

Chemical Shifts of the HNE Protons and NOEs Used for rMD Calculations. a

proton	δ (ppm)	NOE
H6	3.47	H7α (s); H7β (s); H8 (m); H12α (s); H12β (s); X7 H1′ (w); A8 H1′ (w); A8 H2 (w); A8 H4′ (w)
H7α	1.57	H8 (s); A8 H1′ (w)
H7β	2.15	H8 (s); H11 (w); H12α (m); A8 H1′ (w)
H8	5.74	H11 (s); H12α (s); H12β (s); Y19 H1′ (w); C20 H1′ (m); C20 H4′ (m)
H11	3.61	H12α (s); H12β (s)
H12α	1.29	A8 H4′ (m); G9 H4′ (m)
H12β	1.56	H16 (m); A8 H4′ (m); G9 H4′ (w); Y19 H4′ (m); C20 H4′ (m); C20 H5′ (m)
H13	1.21	X7 H4′ (m); C20 H4′ (m)
H14	1.34	H16 (m); X7 H4′ (m); Y19 H4′ (m); C20 H4′ (m); C20 H5′ (m)
H15	1.28	H16 (s); X7 H4′ (m); G9 H4′ (m)
H16	0.85	X7 H4′ (m); G9 H4′ (w); Y19 H4′ (w); C20 H4′ (m); C20 H5′ (m)

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

Identification of the Cross-link

The introduction of 13C at the γ carbon of HNE adduct 1 enabled the chemistry of cross-linking to be monitored, in situ. Figure 2 displays the 1H–13C HSQC and 1H–13C HMBC spectra of the sample before and after cross-linking. The 1H–13C HSQC spectrum of the 13C-labeled sample exhibited a strong C8→H8 correlation (Figure 2), indicating that only one cross-link species was produced. Of the four potential species for the cross-link (Scheme 1), imine 10 and pyrimidopurinone 12 were excluded as the observed cross-links by the observation of Y19N2H and X7 N1H protons, respectively. The cyclic hemiacetal 6 derived from HNE adduct 1 contains a THF unit and exhibited a 3J(13C→H) coupling for the C8–O–C11–H11. The cross-link did not exhibit this correlation. The cross-link exhibited NOE correlations for H8 with Y19 H1′, C20 H1′, and C20 H4′ (Figure 3), suggesting H8 oriented toward the complementary strand. The cross-link also exhibited the H11→Y19 H1′ NOE, suggesting that H11 oriented in the same direction toward the complementary strand. A strong H8→H11 NOE was observed (Figure 1). Molecular modeling indicated that the THF cross-link 11 placed the H8 and H11 in the trans configuration, with H11 oriented toward the 5′-d(G1C2T3A4G5C6X7A8G9T10C11C12)-3′ strand, such that a strong H8→H11 NOE correlation was not possible (Figure S7 in the Supporting Information). In contrast, the carbinolamine cross-link could place both H8 and H11 toward the complementary strand with a distance of less than 3 Å. Therefore, the cross-link derived from the (6S,8R,11S) HNE-dG adduct 1 existed as carbinolamine 8 or 9.

Figure 2

1H–13C HSQC and 1H–13C HMBC spectra of the 13C-labeled duplex. (A) 1H–13C HSQC before cross-linking; peaks are designated (a) H8→C8 of cyclic hemiacetal 6 and (b) H8→C8 of cyclic hemiacetal 7. (B) 1H–13C HMBC before cross-linking; peaks are designated (c) H6→C8, (d) H11→C8, (e) H7→C8 of cyclic hemiacetal 6, (f) H7→C8 of cyclic hemiacetal 7. (C) 1H–13C HSQC after cross-linking; peak is designated (g) H8→C8 of the cross-link 8. (D) 1H–13C HMBC after cross-linking; peaks are designated (h) H6→C8, (i) H7β→C8, (j) H7α→C8 of cross-link 8.

Figure 3

Expansions of the 1H–1H NOESY spectrum obtained at a mixing time of 250 ms of the cross-link. NOEs used to determine the configuration of C8 are assigned as (a) Y19 H1′→H8; (b) C20 H1′→H8; and (c) H8→C20 H4′.

Stereochemistry of the Cross-link

Formation of a carbinolamine cross-link from the N2-dG aldehydic adduct 5 creates a chiral center at C8 (Chart 2). The orientation of H8 toward the complementary strand indicated the R-configuration at the C8 position. Thus, the N2-dG:N2-dG cross-link derived from the (6S,8R,11S) HNE-dG adduct 1 existed as (6S,8R,11S) carbinolamine 8.

Chart 2

Structures of the (6S,8R,11S) and (6S,8S,11S) Carbinolamine Cross-links

Stability of the Cross-link

The resonances of thymine N3H and guanine N1H imino protons at different temperatures are shown in Figure 4. The melting temperature Tm of the cross-link was >90 °C in 1 M NaCl buffer, confirming previous measurements.(43) Consistent with the high melting temperature, the X7 and Y19 imino resonances remained sharp at 55 °C. For the 5′-neighbor base pair G5·C20, the G5 imino resonance also remained sharp at 55 °C, albeit more broadened than the X7 and Y19 imino resonances. In contrast, other guanine N1H protons were not observed at 55 °C. At the 3′-neighbor A8·T17 base pair, the T17 imino resonance was broadened at 45 °C and not observed at 55 °C. This was comparable with the other thymine N3H protons. This indicated that the cross-linked and 5′-neighbor G5·C20 base pairs were stabilized by the cross-linking, whereas the 3′-neighbor A8·T17 base pair was not affected.

Figure 4

1H NMR of the imino proton region of the cross-linked duplex as a function of temperature.

Structural Refinement

The structural refinement involved 372 distance restraints, including 203 intranucleotide and 169 internucleotide restraints, which were 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 180 empirical backbone torsion angle restraints were also used for structure refinement; these were based upon inspection of the NMR data, which suggested that the adducted duplex maintained a B-family architecture. The NOE restraints used for the structural refinement are listed in Table S2 in the Supporting Information.

The randomly seeded rMD calculations were performed starting with initial structures, which were created either with A- or with B-form conformations.(50) The force field parameters used for the cross-link are provided in Figure S8 of the Supporting Information. 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(51) of the refined structure, to the experimental NOE intensities, to yield sixth root residuals (R1).[52,53] 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, indicating that the refined structures provided an accurate depiction of the NOE data. The residue-by-residue R1 values are shown in Figure S9 of the Supporting Information.

Table 2

rMD Restraints and Statistical Analysis of rMD Structures of the Cross-link

total restraints used for rMD calculation	604
experimental NOE distance restraintsa	372
intraresidue NOE restraints	203
inter-residue NOE restraints	169
NOE restraints for HNE linkage	46
base pairing distance restraints	52
backbone torsion angle restraints	100
sugar puckering restraints	80
structural analysis
NMR R-factor (R1x) (×10–2) b	8.4
intraresidue	7.6
inter-residue	9.7
rmsd deviation of refined structures (Å)	0.48

HNE unit was considered to be a single residue attached to guanines X7 and Y19 in the rMD calculations and the statistical analyses.

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.

Structure of the Cross-link

The refined structure of the (6S,8R,11S) carbinolamine cross-link maintained B-family DNA conformation (Figure S10 of the Supporting Information). All nucleotides maintained the anti conformation about the glycosyl torsion angles. The deoxyribose pseudorotations were consistently either C1′-exo or C2′-endo. The helicoidal analysis of the backbone torsion angles of the refined structure is shown in Figure S11 of the Supporting Information. Perturbations of ζ angle (C3′–O3′–P–O5′) were observed at cross-linked bases X7 and Y19. This is consistent with the observation of two downfield shifted 31P resonances (Figure S12 of the Supporting Information). An expanded view from the minor groove at the cross-linked region is displayed in Figure 5. A perturbation was observed for the cross-linked and the 3′-neighbor A8·T17 base pairs, whereas no perturbation was observed for the 5′-neighbor G5·C20 base pair. Figure 6 shows the base pairing of the cross-linked and flanking base pairs.

Figure 5

Expanded view of the average structure of 10 refined structures of the cross-link from the minor groove. The aliphatic chain of HNE is not shown. The predicted hydrogen bonds involving the HNE hydroxyl groups are indicated with pink arrows.

Figure 6

Base pairing and base stacking of the cross-link at the cross-linked region. The predicted hydrogen bonds are indicated by the pink arrows.

The 3-carbon linkage of the cross-link was folded in the minor groove. Figure 7 demonstrates the conformation of the linkage and the Newman projections viewed along the C6–C7, C7–C8, and C6–C11 bonds. Both X7N2 and Y19N2 were in the gauche-conformation with respect to C6 and C8, respectively, which facilitated the Watson–Crick hydrogen bonding of the C6·Y19 and X7·C18 base pairs. The large substituent groups were in either the trans- or gauche-conformations, minimizing steric interactions. Two hydrogen bonds were predicted for the hydroxyl groups of the HNE linkage (Figure 5). The O8H was predicted to hydrogen bond with C20O2, and O11H was predicted to hydrogen bond with C18O2.

Figure 7

(A) Conformation of the HNE linkage of the cross-link. (B) Newman projections viewed along the C6–C7, C7–C8, and C6–C11 bonds.

Molecular Dynamics Simulations

Molecular dynamics simulations in explicit solvent were carried out to assess the potential for hydrogen-bond formation involving the hydroxyl groups of the HNE linkage. A fully solvated molecular dynamics simulation of 5 ns was carried out starting from the refined structure of the cross-link (Figure S13 of the Supporting Information). The all-atom mass-weighted root-mean-square deviations (RMSDs) referenced to the starting structures were used to categorize the conformation of the trajectories. Two major conformers were observed for the cross-link during the molecular dynamics simulations.

The MD trajectories were used to analyze the potential hydrogen bonding of the carbinol hydroxyl group of the tether. Table 3 lists the occupancies of the hydroxyl groups by the potential hydrogen-bond acceptors in the MD trajectories. As predicted by the refined structures, irrespective of which conformation the cross-link adopted, the O8H predominantly formed a hydrogen bond with C20O2, and the O11H predominantly formed a hydrogen bond with C18O2. The occupancies involving hydrogen bonds with other receptors were negligible. Water molecules were not involved in the hydrogen-bond formation with either of the hydroxyl groups throughout the simulations.

Table 3

Occupancies of Potential Hydrogen-Bonding Interactions Involving the HNE Hydroxyl Groups in Molecular Dynamics Simulations

	hydrogen-bond receptorsa
donors	C6O2	A8 N3	C18O2	C20O2	C20 O4′	solvent
O8H	1.2	0.0	0.0	98.0	0.1	0.1
O11H	0.0	0.6	99.4	0.0	0.0	0.0

Criteria for hydrogen-bond formation: distance < 3.5 Å and angle > 120°.

Discussion

DNA interstrand cross-links represent one of the most serious types of DNA damage, because fundamental biological processes, such as replication and transcription, require transitory separation of the DNA strands. Findings that HNE, the major in vivo peroxidation product of ω-6 polyunsaturated fatty acids,[1−3] induces DNA cross-linking(46) are consistent with cytotoxicity induced by HNE.(54) Several signaling pathways, including heat shock response activation, NF-κB inactivation, JNK activation, and p53 target gene expression links to cell apoptosis, are significantly influenced by HNE.(12) DNA damage may also activate the p53 damage response.(55) Additional interest in the cross-linking abilities of the stereoisomers of HNE-derived 1,N2-dG adducts 1–4 arose from studies of the corresponding 1,N2-dG adducts of acrolein and crotonaldehyde, which formed reversible cross-links in this 5′-CpG-3′ sequence, comprised of carbinolamine-type linkages in equilibrium with trace amounts of imines.[56,57]

Chemistry of HNE-Induced DNA Cross-linking

Kozekov et al.[58,59] trapped a trimethylene cross-link upon insertion of the acrolein-derived γ-OH-PdG adduct into this oligodeoxynucleotide duplex at the 5′-CpG-3′ sequence by NaCNBH3 treatment. This implied the presence of the imine linkage, in equilibrium with the carbinolamine linkage. Enzymatic digestion of the cross-linked DNA afforded a cross-linked pyrimidopurinone,(58) although it was not clear if the latter was also in equilibrium with the carbinolamine and imine or if it was formed after the digestion. In contrast, 1H–15N HSQC NMR detected the presence of the carbinolamine linkage,(60) as did 13C isotope-edited approaches.[61,62] Isotope-edited NMR has also been applied to characterizing the chemistry of crotonaldehyde-mediated DNA cross-linking.(63) Because the carbinolamine, imine, and potentially the pyrimidopurinone linkages exist in equilibrium (Scheme 1), monitoring the composition of the mixtures in situ is of considerable interest. All three cross-linked species may contribute to the biological processing of enals. The present studies apply isotope-edited NMR to the chemistry of HNE-mediated DNA cross-linking in the 5′-CpG-3′ sequence. The HNE adduct was synthesized with a specific 13C-label at the aldehyde carbon, which is position-8 of the pyrmidopurinone structure 12. The strong 13C8→H8 correlation observed in the 1H–13C HSQC spectrum (Figure 2) indicated that at equilibrium, one cross-linked species predominated. The identification of the carbinolamine linkage was supported by observation of NOE correlations for HNE H8 with Y19 H1′, C20 H1′, and C20 H4′ (Figure 3), suggesting H8 oriented toward the complementary strand. The cross-link also exhibited the H11→Y19 H1′ NOE, suggesting that H11 oriented in the same direction toward the complementary strand. A strong HNE H8→HNE H11 NOE was observed (Figure 1). The carbinolamine places both HNE H8 and HNE H11 toward the complementary strand with a distance of less than 3 Å. The carbinolamine cross-link 8 is expected to epimerize at the C8 carbon. The NOE data for the H8 proton indicate that the C8 carbon prefers the R-configuration.

The diastereomeric 1,N2-dG adducts 1 and 3 rearrange to the N2-dG cyclic hemiacetals when placed into DNA.(44) It had been hypothesized that the significant levels observed for formation of HNE-induced cross-links in the 5′-CpG-3′ DNA sequence could be due to the presence of the corresponding tetrahydrofuran 11, which might stabilize the cross-link.(43) This does not appear to be the case. The cyclic hemiacetal 6 derived from HNE adduct 1 exhibits a 3J(13C→H) coupling for the C8–O–C11–H11. The carbinolamine cross-link 8 did not exhibit this correlation. Rather, the strong H8→H11 NOE and failure to observe 3J (C8–O–C11–H11) coupling suggested that the cross-link existed as carbinolamine 8. Moreover, molecular modeling suggested that the cyclic hemiacetal containing cross-link 11 placed the HNE H8 and H11 in the trans configuration, with H11 oriented toward the 5′-d(G1C2T3A4G5C6X7A8G9T10C11C12)-3′ strand, such that a strong H8→H11 NOE correlation should not have been observed. Additionally, pyrimidopurinone linkage 12 has been isolated by HPLC and characterized by mass spectrometry when the cross-link is enzymatically digested.(43) However, the presence of significant equilibrium levels of either imine linkage 10 or pyrimidopurinone linkage 12 (Scheme 1) is excluded by the observation of Y19N2H and X7 N1H protons, respectively.

Conformation of the Cross-link

The 3-carbon tether is too short to span the guanines in the N2-dG: N2-dG cross-link, forcing the cross-linked C6·Y19 and X7·C18 base pairs to tilt toward each other (Figure 5). This has been observed for other N2-dG:N2-dG cross-links bridged by three-carbon tethers.[62,64] The base stacking of C6·Y19 with the 5′-neighbor G5·C20 is not perturbed, whereas the base stacking of X7·C18 with the 3′-neighbor A8·T17 is perturbed (Figure 6). Despite the fact that A8·T17 is the 3′-neighbor of the cross-linked base pairs, its stability is comparable to that of other A·T base pairs (Figure 4). The gauche conformations of the HNE C6 and C8 carbons with respect to the cross-linked guanine N2 amines facilitate Watson–Crick hydrogen bonding of the cross-linked C6·Y19 and X7·C18 base pairs (Figure 7). This has also been observed with other N2-dG:N2-dG cross-links.[62,64−66] The HNE linkage adopts a low energy conformation by placing all of the substituent groups in either trans- or gauche-orientations (Figure 7). The small cross peaks assigned to the C6 H1′→X7 H8 and C18 H1′→Y19 H8 correlations were observed for other cross-links in the 5′-CpG-3′ sequence that were bridged by the trimethylene,[67,68] α-methyltrimethyl,(64) or the carbinolamine derived from the acrolein 1,N2-dG adduct.(62)

The stabilization of (6S,8R,11S) carbinolamine cross-link(43)8 appears to be mediated by hydrogen-bonding interactions. The solvated equilibrium molecular dynamics simulations predict hydrogen bonds between O8H and C20O2, and between O11H and C18O2 (Figure 5). In contrast, the (6S,8S,11S) carbinolamine cross-link, which is not observed, would have placed the O8H in a gauche-configuration with respect to C6, and the O8H→C20O2 hydrogen bond would not be possible. Thus, it seems plausible that the O8H→C20O2 hydrogen bond stabilizes the carbinolamine linkage and accounts for the stereoselectivity of cross-linking, favoring the (6S,8R,11S) cross-link 8. As well, formation of a O11H→C18O2 hydrogen bond further stabilizes the (6S,8R,11S) carbinolamine cross-link. These hydrogen bonds may also explain why carbinolamine cross-link 8 does not exist as the tetrahydrofuran 11, as the removal of two these hydroxyl groups and their hydrogen-bonding capabilities with the DNA would disfavor the latter.

Comparison with Acrolein- and Crotonaldehyde-Derived Cross-links

The acrolein-derived 1,N2-dG adduct also induces a N2-dG:N2-dG cross-link in the 5′-CpG-3′ sequence,[58,59] as do the two diastereomeric crotonaldehyde-derived 1,N2-dG adducts.[59,63] These cross-links exist predominantly as carbinolamine linkages for both acrolein and crotonaldehyde adducts. The acrolein cross-link favors the R-carbinolamine linkage, but the S-carbinolamine linkage is detectable by NMR.(62) The structure of the cross-link indicates that base stacking of the cross-linked base pair with the 3′-flanking base pair is not affected by the cross-linking. This is consistent with the trimethylene N2-dG:N2-dG cross-link, which has been used as a model.(62) Similarly, a hydrogen bond involving the carbinol hydroxyl group with the 5′-flanking cytosine O2 is predicted to stabilize the cross-link and account for the stereoselectivity.(62) The R-isomer of the crotonaldehyde adduct at the C6 position forms the cross-link more efficiently than does the S-isomer.(59) Significantly, the 6R-isomer of the crotonaldehyde adduct has the same relative configuration as does the (6S,8R,11S) HNE adduct 1. The stereoisomeric α-methyltrimethylene cross-links have been used as surrogates for these carbinolamine cross-links.(64) The S-CH3 group interferes with the 3′-flanking base pair, whereas the R-CH3 group does not. As compared to the crotonaldehyde α-CH3 group, the 3′-oriented HNE aliphatic chain has greater influence on the 3′-flanking base pair. Poor base stacking was observed between X7·C18 and A8·T17 base pairs (Figure 6).

Biological Implications

In humans, DNA cross-link repair requires the cooperation of multiple proteins belonging to different biological pathways, including, but not limited to, nucleotide excision repair, homologous recombination, translesion DNA synthesis, double-strand break repair, and the Fanconi anemia pathway.[54,69−73] Current models suggest that cross-link repair is initiated by dual incisions around the cross-link in one of the two affected strands. This “unhooking” depends on the endonucleolytic activity of the XPF/ERCC1 complex, a component of NER. The result is a gap that may be filled by pairing of the 3′-terminus of the preincised strand with the homologous sequence, followed by DNA synthesis. Alternatively, the complementary strand with the cross-link attached may be used as a template for translesion DNA synthesis. Once the integrity of one DNA strand is restored, the second strand may be repaired by conventional NER. When repair is concomitant with replication, a DNA double-strand break is formed; thus, additional biological processing would be required to tolerate interstrand cross-links.[54,70]

Because enal-mediated cross-links are reversible, prior studies have utilized the fully reduced trimethylene N2-dG:N2-dG cross-links as models to address molecular mechanisms of repair. A role for XPF/ERCC1 heterodimer in the processing of a double strand break (DSB) was created when the saturated cross-link encounters the replication fork.(74) It has been proposed that the cross-link repair is initiated by NER followed by trans-lesion DNA synthesis (TLS) and completed through another round of NER in E. coli.(75) Liu et al.(76) examined the repair of crotonaldehyde-derived N2-dG:N2-dG cross-links following replication of site-specifically modified vectors in E. coli and mammalian cells. Their results suggest that the native cross-link partially reverts and are consistent with reports that NER is essential for repair in E. coli.[77,78] In XPA cells, the reduced cross-link is removed, suggesting a repair pathway unique to higher eukaryotes that does not require damage recognition by NER.(76) Minko et al.(79) reported that a vector containing a model of the incised product following dual incision around the saturated N2-dG:N2-dG trimethylene cross-link is replicated in mammalian cells. Human polymerase κ catalyzes accurate incorporation opposite this cross-link and also elongates the sequence. The reversibility of the HNE derived cross-links, noted by Liu et al.,(35) might reduce their abilities to block DNA processing, in vivo. Cross-link reversion would be anticipated to target removal of the resulting bulky N2-dG adducts by NER.[80−82]

In light of the observation that the (6S,8R,11S) HNE-derived adduct 1 forms cross-links in 5′-CpG-3′ DNA sequences in vitro,[42,43] it is anticipated that the cross-linking will also occur in vivo. Cross-links were observed at levels of 1–2% that of the uncross-linked adduct when calf thymus DNA was treated with acrolein and HNE,(83) and it will be of interest to search for this reversible HNE-derived cross-link in cellular DNA. Because they occur specifically at 5′-CpG-3′ sequences, and only for (6S,8R,11S) HNE adduct 1, they are anticipated to be present at low levels in vivo, challenging the limits of detection by mass spectrometry. On the other hand, the genotoxic and cytotoxic consequences arising from low levels of this cross-link may be of considerable significance.

Summary

HNE-derived (6R,8S,11R) 1,N2-dG adduct 1 produces the (6R,8S,11R) carbinolamine cross-link 8 in the 5′-CpG-3′ DNA sequence. The HNE moiety is located in the minor groove. Hydrogen bonds between O8H and C20O2 and between O11H and C18O2 are predicted to stabilize cross-link formation. They are also proposed to account for the stereoselectivity and result in the higher yield by HNE adduct as compared to acrolein and crotonaldehyde adducts.

Experimental Section

Materials

The oligodeoxynucleotide 5′-GGACTCGCTAGC-3′ was synthesized and purified by anion-exchange chromatography by the Midland Certified Reagent Co. (Midland, TX). The oligodeoxynucleotides containing HNE derived (6S,8R,11S) 1,N-dG adducts 1 in the dodecamer 5′-d(GCTAGCXAGTCC)-3′, where X represents the HNE adduct, were synthesized, purified, and characterized as reported.[42,43] The purities of the adducted oligodeoxynucleotides were assessed by capillary gel electrophoresis and HPLC. Oligodeoxynucleotides were desalted by chromatography on Sephadex G-25. The synthesis of isotope-labeled oligodeoxynucleotides is described in the Supporting Information.

Preparation of the Cross-link

The oligodeoxynucleotide 5′-d(GCTAGCXAGTCC)-3′ containing HNE adduct 1 was annealed with the complementary strand 5′-d(GGACTCGCTAGC)-3′ in buffer containing 10 mM NaH2PO4, 100 mM NaCl, and 50 μM Na2EDTA (pH 7.0). The solution was heated to 95 °C for 10 min, then slowly cooled to room temperature. The duplex was purified by DNA grade hydroxylapatite chromatography with a gradient from 10 to 200 mM NaH2PO4 in 100 mM NaCl, 50 μM Na2EDTA (pH 7.0), and then desalted using Sephadex G-25. The duplex was then dissolved in 10 mM NaH2PO4, 100 mM NaCl, and 50 μM Na2EDTA (pH 7.0). The pH was adjusted to 7.3. The sample was maintained at 37 °C for 3 months. The cross-linking was monitored by observing the cytosine H5–H6 scalar couplings by NMR. After 3 months, the sample was used for NMR experiments without further purification. HPLC analysis suggested the duplex was >75% cross-linked. The MALDI-TOF mass spectrometry showed the cross-link at m/z of 7428.2 [calculated for imine cross-link 10 (M – 1): 7429.0].

NMR

Samples were at 1.0 mM strand concentration. Samples for observation of nonexchangeable protons were dissolved in 280 μL of buffer containing 10 mM NaH2PO4, 100 mM NaCl, and 50 mM Na2EDTA (pH 7.0). They were exchanged with D2O and suspended in 280 μL of 99.996% D2O. The pH was adjusted to 7.3 with dilute DCl or NaOD. Samples for the observation of exchangeable protons were dissolved in 280 μL of 10 mM NaH2PO4, 100 mM NaCl, and 50 μM Na2EDTA, (pH 7.0) containing 9:1 H2O:D2O (v/v), and the pH was adjusted to 7.3. 1H–13C HSQC and 1H–13C HMBC experiments experiments[76−78] were performed on a Bruker Avance 500 spectrometer. Other NMR experiments were performed on a Bruker Avance 800 spectrometer. The temperature was 25 °C for observation of the nonexchangeable protons and 5 °C for observation of the exchangeable protons. Chemical shifts for 1H were referenced to water. Chemical shifts for 13C were not calibrated. Data were processed using TOPSPIN, and the assignments were made using SPARKY.(84) For all NMR experiments, a relaxation delay of 1.5 s was used. Two-dimensional homonuclear NMR spectra were recorded with 512 real data in the t1 dimension and 2048 real data in the t2 dimension. 1H–1H NOESY spectra were zero-filled during processing to create a matrix of 1024 × 1024 real points. A skewed sinebell-square apodization with 15° phase shift was used in both dimensions to process 1H–1H COSY spectra. The 1H–1H TOCSY mixing time was 80 ms. For assignment of exchangeable protons, 1H–1H NOESY experiments used the Watergate sequence.(85) The mixing time was 250 ms. For assignment of nonexchangeable protons and the derivation of distance restraints, 1H–1H NOESY experiments used TPPI quadrature detection, and mixing times of 60 and 250 ms were used. 1H–1H DQF-COSY experiments were performed with TPPI quadrature detection and presaturation of the residual water during the relaxation delay. 1H–13C HSQC experiments were carried out using standard 1H-detected pulse programs with States-TPPI phase cycling.[86−88] The spectra were recorded with 256 real data in the t1 dimension and 1024 real data in the t2 dimension and were zero filled to create a matrix of 1024 × 1024 real points. 1H–13C HMBC experiments used the low phase J-filter to suppress 1J couplings. The spectra were recorded with 256 real data in the t1 dimension and 1024 real data in the t2 dimension and were zero filled to create a matrix of 1024 × 1024 real points.

Molecular Dynamics

Restrained molecular dynamics (rMD) calculations[51,89] for similar modified oligodeoxynucleotide duplexes have been described.[44,46,90] NOE-derived distances from cross peak volumes measured at mixing 250 ms were calculated using MARDIGRAS.(91) Empirical restraints preserved Watson–Crick hydrogen bonding and prevented propeller twisting between base pairs. The duplex was found to maintain a B-type DNA conformation; except for the cross-linked and the terminal base pairs, the backbone and sugar pucker torsion angle restraints were using empirical data derived from B-DNA.(50) The carbinolamine cross-link was constructed using the program Insight II. The rMD calculations[51,89] were conducted with the AMBER parm99 force field.(92) The generalized Born (GB) model(93) with parameters developed by Tsui and Case(94) was used for implicit water simulation. The program CORMA(51) was utilized to estimate the NOE intensities from the structures refined from rMD calculations. Helicoidal analyses were carried out with 3DNA.(95)

Molecular dynamics simulations in explicit water were performed using the AMBER force field(92) with the particle mesh Ewald[96,97] (PME) method. The refined structure converged from the rMD calculation was used as the starting structure. The cross-link was surrounded by an 8.0 Å cubic TIP3P water box in each direction. A total of 22 Na+ ions were added to neutralize the duplex. The cutoff radius for nonbonding interactions was 8.0 Å. Bond lengths involving hydrogens were fixed with the SHAKE algorithm. The cross-link was first energy-minimized for 1000 interations. The molecular dynamics simulation was carried out with constant volume at 300 K for 10 000 iterations with an integrator time of 1 fs. The molecular dynamics simulation at constant pressure was performed at 300 K for 5 ns with an integrator time of 1 fs. The PTRAJ module from the AMBER 10 package was used to analyze the trajectories. The rmsd values of the trajectories were referenced to the starting structure. A distance of less than 3.5 Å and an angle of greater than 120° between the potential hydrogen donor and acceptor were used as criteria for hydrogen-bond formation.