DNA Sequence Modulates Geometrical Isomerism of the trans-8,9- Dihydro-8-(2,6-diamino-4-oxo-3,4-dihydropyrimid-5-yl-formamido)- 9-hydroxy Aflatoxin B1 Adduct.

Aflatoxin B(1) (AFB(1)), a mycotoxin produced by Aspergillus flavus, is oxidized by cytochrome P450 enzymes to aflatoxin B(1)-8,9-epoxide, which alkylates DNA at N7-dG. Under basic conditions, this N7-dG adduct rearranges to yield the trans-8,9-dihydro-8-(2,6-diamino-4-oxo-3,4-dihydropyrimid-5-yl-formamido)-9-hydroxy aflatoxin B(1) (AFB(1)−FAPY) adduct. The AFB(1)−FAPY adduct exhibits geometrical isomerism involving the formamide moiety. NMR analyses of duplex oligodeoxynucleotides containing the 5′-XA-3′, 5′-XC-3′, 5′-XT-3′, and 5′-XY-3′ sequences (X = AFB(1)−FAPY; Y = 7-deaza-dG)demonstrate that the equilibrium between E and Z isomers is controlled by major groove hydrogen bonding interactions.Structural analysis of the adduct in the 5′-XA-3′ sequence indicates the preference of the E isomer of the formamide group,attributed to formation of a hydrogen bond between the formyl oxygen and the N(6) exocyclic amino group of the 3′-neighboradenine. While the 5′-XA-3′ sequence exhibits the E isomer, the 5′-XC-3′ sequence exhibits a 7:3 E:Z ratio at equilibrium at 283K. The E isomer is favored by a hydrogen bond between the formyl oxygen and the N(4)-dC exocyclic amino group of the 3′-neighbor cytosine. The 5′-XT-3′ and 5′-XY-3′ sequences cannot form such a hydrogen bond between the formyl oxygen and the 3′-neighbor T or Y, respectively, and in these sequence contexts the Z isomer is favored. Additional equilibria between α and β anomers and the potential to exhibit atropisomers about the C5−N(5) bond do not depend upon sequence. In each of the four DNA sequences, the AFB(1)−FAPY adduct maintains the β deoxyribose configuration. Each of these four sequences feature the atropisomer of the AFB(1) moiety that is intercalated above the 5′-face of the damaged guanine. This enforces the Ra axialc onformation for the C5−N(5) bond.

Introduction

Aflatoxin B1 (AFB1, 1, Scheme 1) is a mycotoxin that is isolated from Aspergillus flavus and which contaminates agricultural products.[1−4] AFB1 is a mutagen in bacteria[2,5−7] and in mammalian cells.[8] It is a carcinogen in fish[9] and rodents.[10−12] Dietary exposures to AFB1 are high in areas of Asia[13,14] and sub-Saharan Africa.[15,16] Epidemiological studies suggest that chronic exposure to AFB1 is a contributing factor in the etiology of hepatitis B virus (HBV) associated hepatocellular carcinomas (HCC).[4,14,17−19] Effective biomarkers allowing quantitation of human dietary exposures to AFB1 have been identified,[14,20,21] and there have been efforts to develop chemopreventive interventions to chronic exposures.[14,22−25]

Scheme 1

Formation of the trans-8,9-Dihydro-8-(2,6-diamino-4-oxo-3,4-dihydropyrimid-5-yl-formamido)-9-hydroxy Aflatoxin B1 Adduct (AFB1–FAPY Adduct) and Numbering of the AFB1 Protons

AFB1 is metabolized by cytochrome P450 3A4[26−29] to yield AFB1-exo-8,9-epoxide (Scheme 1).[30] The efficiency of adduction at N7-dG by this epoxide[31] is attributed to its intercalation on the 5′-face of guanine.[32−35] This facilitates the formation of trans-8,9-dihydro-8-(N7-guanyl)-9-hydroxy aflatoxin B1.[36−38] This cationic adduct depurinates to release AFB1–guanine,[36] or alternatively, undergoes base-catalyzed opening of the imidazole ring forming the trans-8,9-dihydro-8-(2,6-diamino-4-oxo-3,4-dihydropyrimid-5-yl-formamido)-9-hydroxy aflatoxin B1 (AFB1–FAPY) adduct.[39−41]

The AFB1–FAPY adduct[41] (Scheme 1) is of interest because the genotoxicity of AFB1 is ascribed primarily to this adduct. It induces G → T transitions[42] associated with AFB1 mutagenesis in bacteria[6,42] and mammalian cells.[8] The cationic N7-dG AFB1 adduct exhibits a similar spectrum of mutations but at lower levels.[43,44] Moreover, the AFB1–FAPY adduct is persistent in vivo.[39,45,46] It may be linked to G → T transversions in the tumor suppression gene p53[47−54] and ras proto-oncogenes.[55] Smela et al.[42] have demonstrated that the AFB1–FAPY adduct equilibrates between two species, one of which is mutagenic whereas the other blocks DNA replication. These are α and β deoxyribose anomers.[41] In duplex DNA, the β anomer is favored, but in single-strand DNA, the equilibrium shifts to favor the α anomer.[41] At the nucleoside level, the AFB1–FAPY adduct exhibits atropisomerism about the C5–N5 bond.[41] This has not been observed in duplex DNA, probably because both deoxyribose anomers of the adduct intercalate with the AFB1 moiety on the 5′ face of the adducted nucleotide,[56−58] similar to the N7-dG cationic adduct.[34,57,59,60] This enforces the Ra axial conformation for the C5–N5 bond.[56−58] Additionally, the AFB1–FAPY adduct exhibits geometrical isomerism involving the formamide moiety.[41] In the 5′-XA-3′ sequence,[56−58] the potential for a hydrogen bond between the formyl oxygen and the N6 exocyclic amino group of the 3′-neighbor adenine has been recognized, and this has been proposed to explain the preference of the E isomer as opposed to the Z isomer of the formamide moiety.[40,41] This suggests that the orientation of the formamide moiety is controlled by differential hydrogen bonding opportunities in the DNA major groove, and the equilibrium between E and Z isomers of the formamide moiety is sequence dependent.

To characterize how DNA sequence modulates the conformation of the AFB1–FAPY formamide moiety, four oligodeoxynucleotide duplexes containing the 5′-XA-3′, 5′-XC-3′, 5′-XT-3′, and 5′-XY-3′ sequences (X-AFB1–FAPY; Y = 7-deaza-dG) (Chart 1) have been constructed, in which the 3′-neighbor base with respect to the adducted guanine has been changed from adenine to thymine, cytosine, and 7-deaza-dG. NMR analyses demonstrate that the equilibrium between E and Z isomers of the formamide moiety is controlled by major groove hydrogen bonding interactions. When the 3′ neighbor is adenine, the E isomer is favored. A mixture of E and Z isomers coexists when the 3′ neighbor is cytosine. The major isomer is the E isomer. In contrast, when the 3′ neighbor is either thymine or 7-deaza-dG, the Z isomer predominates. Structures of each of these duplexes have been refined from NMR data using molecular dynamics calculations restrained by NOE distances.

Chart 1

Four Oligodeoxynucleotide Duplexes Used in This Studya

(A) The 5′-XA-3′ duplex. (B) The 5′-XT-3′ duplex. (C) The 5′-XC-3′ duplex. (D) The 5′-XY-3′ duplex. In all cases, X = AFB1–FAPY; Y = 7-deaza-dG. In each instance, the 3′-neighbor base pair is shown in red.

Materials and Methods

Materials

Unadducted oligodeoxynucleotides were purchased from the Midland Certified Reagent Co. (Midland, TX). AFB1 was purchased from Aldrich Chemical Co. (Milwaukee, WI). AFB1-exo-8,9-epoxide was prepared by oxidizing AFB1 in the presence of dimethyldioxirane,[61−63] as described.[30]

Caution:

Crystalline AFB1is hazardous due to its electrostatic nature and should be handled using appropriate containment procedures and respiratory mask to prevent inhalation. AFB1can be destroyed by treatment with NaOCl. It should be assumed that AFB1-exo-8,9-epoxide is toxic and carcinogenic. Manipulations should be carried out in a well-ventilated hood with suitable containment procedures.

Sample Preparation

The oligodeoxynucleotide 5′-d(CTAAGATTCA)-3′, containing the targeted N7-dG alkylation site (underlined) was annealed with 5′-d(ATCTT)-3′ to form a partially double-stranded scaffold in 200 μL of 100 mM Na2HPO4 (pH 6.5). AFB1-exo-8,9-epoxide was added in 200 μL of CH2Cl2. The biphasic mixture was stirred for 15 min at 5 °C. The aqueous phase was dissolved in 100 mM Na2CO3 (pH 10) for 2 h at room temperature to form the AFB1–FAPY adduct. The oligodeoxynucleotide containing the AFB1–FAPY adduct 5′-d(CTAAXATTCA)-3′ (X = AFB1–FAPY) was recovered from the aqueous phase using reverse-phase HPLC (Gemini C18 250 mm × 10 mm column, Phenomenix, Inc., Torrance, CA) at a flow rate of 2 mL/min, with a linear 30 min gradient of 5–30% CH3CN in 0.1 M ammonium formate (pH 6.5). The eluent was monitored by UV absorbance at 260 and 360 nm. The adducted oligodexoynucleotide was lyophilized and characterized by MALDI-TOF mass spectrometry. The oligodeoxynucleotides 5′-d(CTAAXCTTCA)-3′, 5′-d(CTAAXTTTCA)-3′, and 5′-d(CTAAXYTTCA)-3′ (Y = 7-deazaG) were synthesized and characterized using the same procedures. Each purified AFB1–FAPY modified oligodeoxynucleotide was annealed at room temperature with the complementary strand in buffer consisting of 10 mM NaH2PO4, 0.1 M NaCl, and 50 μM Na2EDTA (pH 7.0), and the annealed duplex oligodeoxynucleotide was eluted from DNA grade Biogel hydroxylapatite using a gradient of 10–200 mM NaH2PO4 (pH 7.0). The modified duplexes were desalted by gel filtration chromatography over Sephadex G-25.

Thermal Melting Experiments

The melting temperatures were measured in 10 mM Na2HPO4, 0.1 M NaCl, and 50 μM Na2EDTA (pH 7.0). The strand concentration was 1.6 μM. The temperature was increased at a rate of 1 °C/min from 10 to 80 °C. Absorbance was measured at 260 nm on a Varian Cary 4E spectrometer. The melting points (Tm values) of the unmodified and modified oligodeoxynucleotides were obtained by determining the inflection points of the absorbance vs temperature curves from the first-order derivatives.

NMR

Spectra were recorded at 1H frequencies of 600, 800, and 900 MHz using cryogenic probes (Bruker Biospin, Inc., Billerica, MA). Samples were prepared in D2O, containing 0.1 M NaCl, 10 mM NaH2PO4, and 50 μM Na2EDTA (pH 7.0). The program TOPSPIN (Bruker Biospin, Inc., Billerica, MA) was used for data collection and processing. Chemical shifts were referenced to the chemical shift of the water resonance at the corresponding temperature, with respect to 4,4-dimethyl-4-silapentane-1-sulfonic acid (DSS). HMQC spectra[64,65] were recorded with 96 real data in the t1 dimension and 1024 real data in the t2 dimension. COSY and NOESY spectra were recorded with 512 real data in the t1 dimension and 2048 real data in the t2 dimension. In the NOESY experiment for exchangeable protons, the samples were prepared in 9:1 H2O:D2O and the water signal was suppressed by using the WATERGATE pulse sequence.[66] Resonance assignment and peak integration were performed using the program SPARKY.[67]

Experimental Restraints

Footprints were drawn around the NOE cross-peaks obtained at NOESY mixing times of 150, 200, and 250 ms using the program SPARKY. Cross-peak intensities were determined by volume integrations. The intensities of the cross-peaks were combined with the intensities generated from complete relaxation matrix analysis of a starting DNA structure to generate a hybrid intensity matrix.[68,69] The program MARDIGRAS,[70−72] using the RANDMARDI[72,73] algorithm, was used to refine the hybrid matrix by iteration between the calculated and experimental NOE intensities. The calculations were initiated using isotropic correlation times of 2, 3, and 4 ns. Analysis of this data yielded experimental distance restraints used in molecular dynamics calculations.

Molecular Dynamics Calculations

Restrained molecular dynamics (rMD) calculations for the oligodeoxynucleotide duplexes utilized a simulated annealing approach.[74] The AFB1–FAPY modified oligonucleotides were constructed using the BUILDER module of INSIGHT II (Accelrys, Inc., San Diego, CA). The partial charges on the AFB1–FAPY nucleotide were obtained from density function theory (DFT) calculations using a neutral total charge, utilizing the B3LYP/6-31G* basis set and the program GAUSSIAN.[75] To obtain the starting structures used for rMD calculations, the AFB1 FAPY-modified duplex was energy minimized using 200 iterations with the conjugate gradients algorithm. The rMD calculations were conducted with AMBER[76] using the parm99[77] force field. The generalized Born (GB)[78] model with parameters developed by Tsui and Case[79] was used for implicit water simulation. The program CORMA[68,69] was utilized to calculate the NOE intensities from the structures emergent from rMD calculations. Molecular dynamics simulations in explicit water were performed using the AMBER force field. The average structure converged from the series of simulated annealing rMD calculations was used as the starting structure. This was placed in an truncated octahedral TIP3P water box with periodic boundaries at a distance of 8.0 Å from the solute.[80] The necessary Na+ ions were added to neutralize the duplex using restraints having a lower bound of 3.0 Å and an upper bound of 8.0 Å. The system was subjected to 1000 iterations of potential energy minimization using steepest descents. The solvent was brought to thermal equilibrium by a rMD simulation at constant volume for 10000 iterations with an integrator time of 1 fs, at 300 K. The experimental distance and torsion angle restraints and empirical restraints were increased linearly during the heating. rMD calculations were performed at constant pressure for 1 ns with an integrator time of 1 fs. Bond lengths involving hydrogens were fixed with the SHAKE algorithim.[81] The particle mesh Ewald (PME) method was used to approximate nonbonded interactions.[82,83] The cutoff radius for nonbonded interactions was 8.0 Å. The PTRAJ program[76] from the AMBER package was used to analyze the rMD trajectories.

Results

AFB1–FAPY Modified Duplexes

The 10-mers 5′-d(CTAAGATTCA)-3′, 5′-d(CTAAGCTTCA)-3′, 5′-d(CTAAGTTTCA)-3′, and 5′-d(CTAAGYTTCA)-3′ were designed to contain a single guanine, the targeted N7-dG alkylation site (underlined). These were individually annealed with 5-mers, 5′-d(ATCTT)-3′, 5′-d(AACTT)-3′, containing a C:A mismatch, 5′-d(AACTT)-3′, or 5′-d(ACCTT)-3′, respectively, to form double-stranded scaffolds extending two base pairs in either direction from the targeted N7-dG alkylation site. The scaffold facilitated the reaction with AFB1-exo-epoxide, by allowing intercalation above the 5′-face of the targeted guanine, and directing alkylation at the N7-dG position.[37,38] The resulting 10-mers containing trans-8,9-dihydro-8-(N7-guanyl)-9-hudroxy-AFB1 adducts[36] were separated from the scaffolds using reverse-phase HPLC. Under basic conditions, over a period of several hours, the 10-mers spontaneously rearranged to the corresponding AFB1–FAPY adducts.[39−41] The utilization of 7-deaza-dG in the 5′-XY-3′ sequence directed the alkylation to the single dG; it also facilitated the subsequent isolation of the AFB1–FAPY modified oligodeoxynucleotide using reverse-phase HPLC. The 7-deaza-dG:C base pair was not anticipated to disrupt the helical structure of the duplex.[84] The 10-mers containing the AFB1–FAPY adducts were characterized by MALDI-TOF mass spectrometry: for 5′-XA-3′, calcd 3357.1, found 3356.5; for 5′-XC-3′, calcd 3333.1, found 3333.2; for 5′-XT-3′, calcd 3348.1, found 3348.0; for 5′-XY-3′, calcd 3372.1, found 3372.8.

NMR of the AFB1–FAPY Adducts

AFB1 Moiety

For each of the four duplexes, one set of resonances of the X5 AFB1 moiety was observed. Figure 1 shows the data for the 5′-XC-3′ duplex. Figures S1, S2, and S3 in the Supporting Information show the corresponding spectra for the 5′-XA-3′, 5′-XT-3′, and 5′-XY-3′ duplexes. The AFB1 H5, H6a, H8, H9, H9a, and −OCH3 resonances were assigned similar to previous studies, consistent with the notion that in each instance, the AFB1 moiety was intercalated above the 5′-face of the modified guanine and the each adduct maintained a similar Ra axial conformation for the C5–N5 bond linking the pyrimidine ring to the formamido nitrogen.[41,56,85] The assignments of the AFB1 protons are tabulated in Tables S1–S4 of the Supporting Information.

Figure 1

NOESY spectrum showing the assignments of the AFB1 protons of the 5′-XC-3′ duplex. The cross peaks are assigned as (a) X5 AFB1 OCH3 → X5 AFB1 H5; (b) X5 AFB1 H9a → X5 AFB1 CHO; (c) X5 AFB1 H9a → X5 AFB1 H6a; (d) X5 AFB1 H9a → X5 AFB1 H9; (e) X5 AFB1 H9 → X5 AFB1 CHO; (f) X5 AFB1 H9 → X5 AFB1 H6a; (g) X5 AFB1 H8 → X5 AFB1 CHO, (h) X5 AFB1 H8 → X5 AFB1 H6a, (i) X5 AFB1 H8 → X5 AFB1 H9, (j) X5 AFB1 H5 → X5 AFB1 H6a, (k) X5 AFB1 H6a → X5 AFB1 CHO. The 900 MHz spectrum with a mixing time of 250 ms was collected at 283 K. Corresponding spectra for the 5′-XA-3′, 5′-XT-3′, and 5′-XY-3′ duplexes may be found in Figures S1, S2, and S3 of the Supporting Information.

AFB1–FAPY Formyl Group

A series of HMQC spectra[64,65] allowed the characterization of the E:Z equilibrium of the AFB1–FAPY formyl group in the 5′-XA-3′, 5′-XC-3′, 5′-XT-3′, and 5′-XY-3′ duplexes (Figure 2). For the 5′-XA-3′ duplex, the carbonyl 13C resonance was observed at 171 ppm, coupled to a proton resonance at 8.43 ppm. This resonance was assigned to the E geometrical isomer of the formyl group. For the 5′-XC-3′ duplex, two cross-peaks were observed in this region of the HMQC spectrum, suggesting that for this duplex, two species were present in equilibrium. One 13C carbonyl resonance at 167 ppm was coupled to a proton resonance at 8.33 ppm. The other 13C carbonyl resonance at 168 ppm was coupled to a proton resonance at 7.47 ppm, assigned as the Z isomer of the formyl group. For the 5′-XT-3′ duplex, the carbonyl 13C resonance was observed at 168 ppm, coupled to a proton resonance at 7.71 ppm. For the 5′-XY-3′ duplex, the carbonyl 13C resonance was observed at 169 ppm, coupled to a proton resonance at 7.71 ppm. For each of the four duplexes, the assignments of the formyl proton resonances were supported by NOEs to the deoxyribose H1′ and AFB1 H6a protons. Additionally, different NOEs were observed between the formyl protons and AFB1 H8. For the 5′-XA-3′ and the major isomer of the 5′-XC-3′ duplexes, strong NOEs were observed between the X5 formyl protons and the AFB1 H8 proton, which were observed at δ 8.3 ppm and δ 5.7 ppm, respectively. For the 5′-XT-3′ and 5′-XY-3′ duplexes, weak NOEs were observed between the X5 formyl protons and the AFB1 H8 protons, which were observed at δ 7.7 ppm and δ 6.1 ppm, respectively.

Figure 2

NMR analysis of the formyl group proton resonances for the four AFB1–FAPY modified duplexes. The spectra show cross-peaks between the carbonyl 13C resonance and the aldhyde proton resonances, for each of the four sequences. (A) The 5′-XA-3′ duplex. (B) The 5′-XC-3′ duplex. (C) The 5′-XT-3′ duplex. (D) The 5′-XY-3′ duplex. The HMQC spectra were collected at 283 K at a 1H frequency of 600 MHz.

Anomeric Configurations

NMR spectra of the nonexchangeable protons of the AFB1–FAPY modified 5′-XA-3′, 5′-XC-3′, 5′-XT-3′, and 5′-XY-3′ duplexes were assigned using the sequential NOE connectivity of the base proton H6 or H8 dipolar couplings with H1′ deoxyribose protons.[86,87] For each of the duplexes, the deoxyribose spin systems were assigned from the assignments of the H1′ protons. The assignments of the deoxyribose H2′ and H2″ resonances were determined from the relative cross peak intensities in the NOESY spectrum between the H2′ and H2″ resonances and the H3′ resonances. These measurements were obtained at a NOE mixing time of 60 ms, which minimized spin diffusion. The anomeric configurations at X5 C1′ were determined by analyzing the NOEs between the H2′ and H2″ protons and the H1′ protons. Figure 3 shows the results for the E and Z geometrical isomers of 5′-XC-3′duplex. In each instance, the intensity of the X5 H1′ to X5 H2′ NOE was less than the X5 H1′ to X5 H2″ NOE, confirming the β configurations of the glycosyl bonds. Figures S4 and S5 of the Supporting Information show the corresponding data for the 5′-XA-3′-and 5′-XT-3′-duplexes.

Figure 3

Comparison of NOE intensities for the cross-peaks arising between the X5 H1′ and H3′ deoxyribose protons to the X5 H2′ and H2″ deoxyribose protons for the E and Z geometrical isomers of the 5′-XC-3′ duplex. (A) The E isomer. (B) The Z isomer. The 900 MHz NOESY spectrum with a mixing time of 60 ms was collected at 283 K. Corresponding spectra of the 5′-XA-3′ and 5′-XT-3′ duplexes may be found in Figures S4 and S5 of the Supporting Information.

Watson–Crick Base Pairing

For each of the 5′-XA-3′, 5′-XC-3′, 5′-XT-3′, and 5′-XY-3′ duplexes, the resonances of the nucleobase imino protons were assigned on the basis of sequential connectivity between adjacent base pairs in NOESY spectra. These assignments were supported by NOEs to the amino protons of Watson–Crick base pairs.[88] In each duplex, there was an interruption of the sequential imino-to-imino proton NOEs between the T17 N3H proton at base pair A4:T17 and the X5 N1H proton at base pair X5:C16, consistent with the 5′-intercalation of the AFB1 moiety. In each duplex, at the X5:T16 base pair, the observation of strong interstrand cross peaks from X5 N1H to the C16N4 H1 and C16N4 H2 amino protons indicated that Watson–Crick hydrogen bonding between X5 and C16 was intact (Figure 4). For the X5 N1H imino proton in the 5′-XC-3′ duplex, two sets of resonances were observed at 278 K. The ratio between them was 1:2.8. Compared to the unmodified duplex, the C6N4 H1 non-hydrogen bonded exocyclic amine proton shifted downfield 0.21 ppm at 278 K (Figure 4). The assignments of the exchangeable protons are tabulated in Tables S5–S8 of the Supporting Information.

Figure 4

Comparison of the A6N6 H1 and N6 H2 amino proton resonances of the unmodified 5′-GA-3′ duplex with those of the 5′-XA-3′ duplex. (A) The 5′-GA-3′ duplex. The cross peaks are assigned as (a) A6N6 H1 → T15 N3H; (b) A6N6 H2 → T15 N3H; (c) A6 H2 → T15 N3H; (d) C16 H5 → G5 N1H; (e) C16N4 H1 → G5 N1H; (f) C16N4 H2 → G5 N1H. (B) 5′-XA-3′ duplex. The cross-peaks are assigned as (g) A6N6 H1 → T15 N3H; (h) A6N6 H2 → T15 N3H; (i) A6 H2 → T15 N3H; (j) C16 H5 → X5 N1H; (k) C16N4 H1 → X5 N1H; (l) C16N4 H2 → X5 N1H. Comparison of the C6N4 H1 and N4H2 amino proton resonances of the unmodified 5′-GC-3′ duplex with those of the 5′-XC-3′ duplex. (C) Unmodified 5′-GC-3′ duplex. The cross-peaks are assigned as (a) C16N4 H1 → G5 N1H; (b) C16N4 H2 → G5 N1H; (c) C6N4 H1 → G15 N1H; (d) C6N4 H2 → G15 N1H. (D) 5′-XC-3′ duplex. The cross-peaks are assigned as (e) C6 H5 → G15 N1H; (f) C6N4 H1 → G15 N1H; (g) C6N4 H2 → G15 N1H; (h) C16 H5 → X5 N1H; (i) C16N4 H1 → X5 N1H; (j) C16N4 H2 → X5 N1H. The NOESY spectra were collected at 800 MHz with a mixing time of 250 ms. The temperature was 278 K.

DNA Duplex Structure

For each of the 5′-XA-3′, 5′-XC-3′, 5′-XT-3′, and 5′-XY-3′ duplexes, in the modified strand, the sequential NOE connectivity between nucleobase purine H8 or pyrimidine H6 protons and deoxyribose H1′ protons was observed from C1 to A4. In each instance, the sequential connectivity exhibited an interruption between A4 H1′ and X5 due to the loss of the guanine H8 proton because of the opening of the guanine imidazole ring (Figure 5). The connectivity resumed starting from an intranucleotide NOE between X5 H1′ and the formyl proton of the FAPY base and continued to the 3′-terminus. For the complementary strand, in all instances, an interruption of the sequential NOEs was observed between C16 H1′ and T17 H8. For the 5′-XC-3′ duplex, two sets of resonances for several protons were observed for the base pairs X5:C16, C6:G15, and T7:A14, in both COSY and NOESY experiments, indicating the existence of two species. The assignments of the nonexchangeable protons are tabulated in Tables S9–S12 of the Supporting Information.

Figure 5

Sequential NOE connectivity of base H8 or H6 protons with deoxyribose H1′ protons for the four AFB1–FAPY modified duplexes. (A) The 5′-XA-3′ duplex. (B) The 5′-XC-3′ duplex. (C) The 5′-XT-3′ duplex. (D) The 5′-XY-3′ duplex. X = AFB1–FAPY, Y = 7-deaza-dG. The arrows indicate the NOEs between the formyl protons and the AFB1 H8 protons. The 900 MHz NOESY spectra with a mixing time of 250 ms were collected at 283 K.

Thermal Stabilities of the Modified Duplexes

For the 5′-GA-3′, 5′-GC-3′, 5′-GT-3′, and 5′-GY-3′ unmodified duplexes, the Tm values, as monitored by UV absorbance at 260 nm, were 31, 37, 32, and 35 °C, respectively. The higher values of Tm for the 5′-GC-3′ and 5′-GY-3′ duplexes were attributed to the presence of the 3′-neighbor C:G or 7-deaza-dG:C base pairs vs 3′-neighbor A:T or T:A base pairs, respectively. As anticipated,[41,56,85] the presence of the AFB1–FAPY adducts increased the thermal stabilities of each of the four duplexes. The melting points for the 5′-XA-3′, 5′-XC-3′, 5′-XT-3′, and 5′-XY-3′ duplexes were 41, 45, 41, and 45 °C, respectively. The higher values of Tm for the 5′-XC-3′ and 5′-XY-3′ duplexes were also attributed to the presence of the 3′-neighbor C:G or 7-deaza-dG:C base pairs, respectively.

The thermal melting of the 5′-XA-3′, 5′-XC-3′, 5′-XT-3′, and 5′-XY-3′ duplexes was also examined by monitoring NMR spectra of the imino protons as a function of temperature; the imino proton resonances were assigned using standard methods.[88] For each duplex, at the X5:C16 base pair the X5 N1H imino proton resonance and the 5′-neighbor A4:T17 base pair T17 N3H imino proton resonance were observed at 40 °C (Figure 6). Increased stabilization of the 3′-neighbor base pairs was also observed. For the 5′-XA-3′ duplex the T15 N3H imino proton resonance was observed. For the 5′-XC-3′ duplex, the G15 N1H imino proton resonance was observed. At 308 K, the ratio between the two isomers changed to 1:2.2 (Figure 6). This was also observed in COSY experiments at both temperatures. The additional imino resonances at 308 K were also observed for the G15 N1H resonance (Figure 6). For the 5′-XY-3′ duplex, the Y6 N1H imino proton resonance was observed. For the 5′-XY-3′ duplex, a single resonance was observed at both 278 and 308 K for the X5 N1H imino proton (Figure 6). For the 5′-XC-3′ duplex, the presence of two species in equilibrium was evident, with doubling of the imino proton resonances being observed for the X5:C16 and 3′-neighboring C6:G15 base pairs. The imino proton resonance for the 5′-neighbor A4:T17 base pair exhibited line broadening. The ratio of the two species was approximately 3:1 at 10 °C, decreasing to 2:1 at 40 °C. Both species exhibited similar overall Tm values.

Figure 6

NMR spectra showing the Watson–Crick base paired imino proton region for the four AFB1–FAPY modified duplexes as a function of temperature. (A) The 5′-XA-3′ duplex. (B) The 5′-XC-3′ duplex. (C) The 5′-XT-3′ duplex. (D) The 5′-XY-3′ duplex. The arrows indicate imino proton resonances assigned to the Z geometrical isomers in equilibrium with the E isomers for the 5′-XA-3′ and 5′-XC-3′ duplexes. X = AFB1–FAPY, Y = 7-deaza-dG. The spectra were collected at a1H frequency of 800 MHz.

Structural Refinement of the Modified Duplexes

NOEs Between AFB1 and DNA

For each of the 5′-XA-3′, 5′-XC-3′, 5′-XT-3′, and 5′-XY-3′ duplexes, the protons of the two AFB1-fused furan rings showed NOEs to major groove and imino protons of the DNA; most of these were to the 5′ neighboring base-pair A4:T17. Thus, H6a and H9a, which are located on the same face of the AFB1 moiety, both exhibited NOEs to A4 H8. A weaker NOE was observed for AFB1 H9. The AFB1 H5 and −OCH3 protons exhibited NOEs with minor groove and imino DNA protons. These were primarily to base A4:T17 in the 5′ direction and to the modified nucleotide X5. These included NOEs between AFB1 −OCH3 and A4 H1′, A4 H2′, A4 H2″, A4 H2, T17 N3H, X5 H1′, and X5 N1H. The cyclopentenone ring H2α and H2β produced NOEs with C16 H1′, H2′, and H2″, and T17 H1′, and H3′ of the complementary strand.

A series of rMD calculations, using a simulated annealing protocol, were performed from initial A- and B-form DNA starting structures for each of the AFB1–FAPY modified duplexes. For the 5′-XC-3′ duplex, only the structure of the E isomer was refined; the Z isomer did not exhibit sufficient NOE cross-peaks to allow structural refinement. Emergent structures, five each from A- and B-DNA starting structures, with lowest energies, were obtained for each duplex and subjected to potential energy minimization. For each duplex, all structures converged as indicated by pairwise rmsd comparisons (Table 1, and Figures S6, S7, S8, and S9 in the Supporting Information). The accuracies of the emergent structures were evaluated by comparison of theoretical NOE intensities calculated by CORMA[70] for the refined structure to the experimental NOE intensities to yield sixth root residuals (R1).[69] The R1 values for overall residuals, as well as the residuals for intra- or internucleotide NOEs, were consistently less than 0.1 (Table 1) and for each nucleotide were less than 0.15, suggesting that the refined structures for each duplex were in good agreement with the NOESY data. Figures S10, S11, S12, and S13 in the Supporting Information show the R1 values for individual nucleotides.

Table 1

Distribution of Restraints Applied to Structural Refinements and Statistical Analyses for the 5′-XA-3′, 5′-XC-3′, 5′-XT-3′, and 5′-XY-3′ Duplexes

	5′-XT-3′	5′-XY-3′	5′-XA-3′	5′-XC-3′
experimental NOE distance restraints	244	264	256	250
intranucleotide NOE restraints	152	161	160	159
internucleotide NOE restraints	92	103	96	91
NOEs of FAPY	41	39	34	34
empirical base pairing restraints	40	40	40	40
empirical backbone torsion restraints	50	50	50	50
empirical pseudorotation restraints	50	50	50	50
total restraints for rMD calculation	384	404	396	390
Structure Statistics
NMR R-factor (R*) (×10–2)	6.88	8.28	8.09	7.47
intranucleotide NOEs	7.03	7.64	8.48	6.68
internucleotide NOEs	6.60	9.40	7.35	8.82
rmsd deviation of refined structures	0.48	0.63	0.48	0.50

Expanded views of the structures are shown in Figure 7. In each instance, the AFB1 moiety intercalated above the 5′-face of the modified nucleotide X5 and between base pairs A4:T17 and X5:C16, causing the rise between these base pairs to increase to 7 Å. The adduct-induced unwinding was localized to the adducted base pairs X5:C16 and their 5′- and 3′-neighbor base pairs A4:T17 and A6:T15. The modified duplexes were each unwound approximately 15° at the adducted sites. For the 5′-XA-3′ duplex, the A6N6 H1 non-Watson–Crick hydrogen bonded exocyclic amine proton was within hydrogen bonding distance of the X5 formyl oxygen; this positioned the formamide moiety in the E configuration. For the 5′-XC-3′ duplex, only the structure of the E isomer was refined; the Z isomer did not exhibit sufficient NOE cross-peaks to allow structural refinement. The C6N4 H1 non-Watson–Crick hydrogen bonded exocyclic amine proton was within hydrogen bonding distance of the X5 formyl oxygen; this positioned the formamide in the E configuration. The structure of the 5′-XT-3′ duplex positioned the formamide in the Z configuration. The structure suggested that the X5 formyl oxygen was within hydrogen bonding distance of the X5 N9H exocyclic amine proton.

Figure 7

Refined structures of the four AFB1–FAPY modified duplexes. (A) The 5′-XA-3′ duplex. (B) The 5′-XC-3′ duplex. (C) The 5′-XT-3′ duplex. (D) The 5′-XY-3′ duplex. X = AFB1-β-FAPY, Y = 7-deaza-dG.

Molecular Dynamics Simulations

For each of the 5′-XA-3′, 5′-XC-3′, 5′-XT-3′, and 5′-XY-3′ duplexes, 1 ns of equilibrium rMD calculation was performed in explicit water at constant pressure at 300 K to examine the dynamics of the refined structure and to analyze hydrogen bond occupancies involving the formyl oxygen atom of the AFB1–FAPY adduct. The 1 ns rMD trajectory was analyzed for occupancies of hydrogen bonding motifs. Hydrogen bond occupancies were calculated using a distance cutoff of 3.5 Å and an angle cutoff of 120°. Using these criteria, for the 5′-XA-3′ duplex, the hydrogen bond between the formyl oxygen and the A6 N6-dA exocyclic amino group was satisfied for >95% of the trajectory. Using the same criteria, for the E configuration of the 5′-XC-3′ duplex, this hydrogen bond was satisfied for 93% of the trajectory. In contrast, using these criteria, for the 5′-XT-3′, and 5′-XY-3′ duplexes the rMD trajectories in explicit solvent suggested that the X5 formyl oxygen was within hydrogen bonding distance of the X5 N9H exocyclic amine proton; this positioned the formamide in the Z configuration. This orientation was satisfied for approximately 75% of the trajectory for both the 5′-XT-3′, and 5′-XY-3′ duplexes.

Deposition of Structural Coordinates

The structural coordinates for the 5′-XA-3′ duplex, the E isomer of the 5′-XC-3′ duplex, the 5′-XT-3′ duplex, and the 5′-XY-3′ duplex have been deposited in the Protein Data Bank. PDB ID codes: 2MMF for the 5′-XA-3′ duplex, 2MMR for the E isomer of the 5′-XC-3′ duplex, 2MMQ for the 5′-XT-3′ duplex, and 2MMS for the 5′-XY-3′ duplex.

Discussion

The structures of AFB1–FAPY adducts in DNA[56−58] and in template:primer complexes with DNA damage bypass polymerases[89] are of interest due to their potential to equilibrate between α and β deoxyribose anomers, atropisomers of the C5–N5 bond linking the pyrimidine ring to the formamide nitrogen, and geometrical isomers of the formamide moiety.[41] In duplex DNA, in differing sequence contexts, these equilibria are anticipated to modulate the biological processing and genotoxicity of AFB1–FAPY adducts. The AFB1–FAPY adduct[41,42] may be linked to site-specific G → T transversions in the tumor suppression gene p53[47−54] and ras proto-oncogenes.[55]

Sequence-Specific Hydrogen Bonding in DNA Modulates Geometrical Isomerism of the AFB1–FAPY Formamide Group

In the 5′-XA-3′ duplex, the E configuration of the formamide group (Figure 2) allows formation of a hydrogen bond between the formyl oxygen of the AFB1–FAPY adduct and the N6-dA non-Watson–Crick hydrogen bonded exocyclic amino proton of the 3′-neighbor A6 (Figure 7). Several features of the NMR spectrum provide evidence for the existence of this hydrogen bond, which had been predicted from studies by Mao et al.[56] and Brown et al.[58] The downfield chemical shifts of the formyl proton (Figure 2) and the A6 non-Watson–Crick exocyclic amino proton (Figure 4) are consistent with the electronic deshielding of both protons. The appearance of an additional NMR resonance at 308 K, assigned to the Z configuration (Figure 6), suggests that the Z configuration is favored in the absence of a thermally stable DNA duplex, as is observed at the nucleoside level.[58] The slow interconversion of E and Z isomers (Figure 6) is consistent with a significant lifetime for this hydrogen bond on the NMR time scale. It suggests that the interconversion between the two configurations within the major groove is sterically hindered in duplex DNA. In support of this conclusion, the rMD trajectories calculated in the presence of explicit solvent over a 1 ns time period predict a significant occupancy of this hydrogen bond.

In the 5′-XC-3′ duplex, a mixture of E and Z isomers is observed (Figure 2). The intensity of the NOE between the formyl proton and AFB1 H8 proton establishes the major isomer as the E configuration (Figure 5). This allows for the formation of a hydrogen bond between the formyl oxygen and the C6N4-dC non-Watson–Crick amino proton (Figure 7). In the refined structure, the distance between the formyl oxygen and C6N4 H1 is 2.7 Å, and the calculated angle O···H―N is 110° (Figure 7). As compared to the 5′-XA-3′ duplex, the C6N4-dC non-Watson–Crick amino proton is less deshielded in the 5′-XC-3′ duplex (Figure 4). This suggests a weaker hydrogen bond, as compared to the 5′-XA-3′ duplex. The coexistence of α and β anomers is ruled out by comparison of the NOEs intensities between the X5 H1′ to X5 H2′ and H2″, which indicates that both E and Z geometrical isomers maintain the β configuration of the deoxyribose (Figure 3). Moreover, only one set of NMR resonances is observed for the X5 AFB1 moiety (Figure 1) and the coexistence of both α and β anomers would have been predicted to result in distinguishable subspectra for each isomer.[58] Similar to the 5-XA-3′ duplex, as the temperature is increased from 278 to 308 K, the equilibrium shifts toward the Z isomer (Figure 6). This again suggests that the Z configuration is favored in single-strand DNA,[58] but that in duplex DNA, the major groove hydrogen bonding favors the E configuration. As for the 5-XA-3′ duplex, slow interconversion of E and Z isomers (Figure 6) suggests a significant lifetime for this hydrogen bond and that interconversion between the two geometrical isomers within the major groove is hindered in duplex DNA. The rMD trajectories calculated in the presence of explicit solvent over a 1 ns time period are consistent with this conclusion and predict significant occupancy of this hydrogen bond.

In the 5′-XT-3′ and 5′-XY-3′ duplexes, there is no 3′-neighbor exocyclic amino proton in the major groove for either T or 7-deaza-dG, so a major groove hydrogen bond cannot form. Consequently, the formamide moiety favors the Z geometrical isomer (Figure 2), as is observed at the nucleoside level.[58] For both duplexes, the preference of the Z isomer is confirmed by the less intense NOE between the formyl proton and AFB1 H8 proton, as compared to the 5′-XA-3′ duplex, indicating a greater distance between the two protons in the Z isomer (Figure 5). For the 5′-XT-3′ duplex, the refined structure predicts that the distance between the formyl oxygen and X5 N9H is 3.1 Å, and the calculated angle O···H―N is 138° (Figure 7). For the 5′-XY-3′ duplex, the corresponding predicted distance is 2.7 Å and 147° (Figure 7). These results suggest that the Z isomer could be stabilized by a hydrogen bond between the formyl oxygen of the AFB1–FAPY adduct and the X5 N9H exocyclic amine proton. However, the rMD trajectories calculated in explicit solvent are equivocal. They suggest 76% and 81% occupancies for this hydrogen bond in the 5′-XT-3′ and 5′-XY-3′ duplexes, respectively. The observation of a single NMR resonance for the X5 N1H imino proton at both 278 and 308 K for both the 5′-XT-3′ and 5′-XY-3′ duplexes (Figure 6) is consistent with the notion that the Z isomer is favored at all temperatures.[58] Likewise, a single resonance is observed for the imino protons of the 3′-neighbors, T6 N3H and 7-deaza-dG6 N1H in the 5′-XT-3′ and 5′-XY-3′ duplexes, respectively, at 308 K (Figure 6).

Structures of the AFB1–FAPY Adducts

The significant structural differences between the four AFB1–FAPY adducts studied here, under equilibrium conditions, are with regard to the geometrical isomerization of the formamide moiety in the major groove (Figure 7). For each of the four duplexes, only one set of NMR resonances is observed for the X5 AFB1 moiety (Figure 1), which indicates that the equilibrium between E and Z isomers induces a localized perturbation to the duplex. Spectroscopic differences between the E and Z isomers are confined to base pairs X5:C16, C6:G15, and T7:A14, and the modest chemical shift perturbations of corresponding nucleobase and deoxyribose H1′ protons of the E and Z isomers become smaller proceeding from the modified base pair X5:C16 and proceeding in the 3′-direction toward base pairs C6:G15, and T7:A14 (Figure 5). The equilibria between α and β anomers[41] and the atropisomerism about the C5–N5 bond[41] do not exhibit sequence dependence. For all four duplexes, the AFB1 moiety intercalates above the 5′-face of the damaged base, the Ra axial conformation about the C5–N5 bond is maintained, and the β anomer of the deoxyribose is favored, similar to studies of the β anomer of the AFB1–FAPY adduct.[56,57] A study of the α anomer of the AFB1–FAPY adduct[58] concluded that the AFB1 moiety also intercalates on the 5′ face of the damaged base, also maintaining the Ra axial conformation about the C5–N5 bond. However, perturbations of the ε and ζ backbone torsion angles were observed, and the base stacking of the duplex was perturbed, which correlates with the observation that the α anomer of the AFB1–FAPY adduct blocks DNA replication.[42]

Thermodynamic Considerations

One notable feature of the β anomer of the AFB1–FAPY adduct is its thermal stabilization of the DNA duplex.[56,57] It has been proposed that hydrogen bonding between the formyl oxygen of the AFB1–FAPY adduct and the N6 non-Watson–Crick hydrogen bonded exocyclic amino proton of the 3′-neighbor A6 provides a potential contribution to this stabilization,[56,57] but the present results suggests that this is not the major contributing factor. Although in the 5′-XA-3′ and 5′-XT-3′ duplexes, the formamide is held in E vs Z isomers, the 5′-intercalation of the AFB1 moiety stabilizes both 5′-XA-3′ and 5′-XT-3′ duplexes by a similar 10 °C increase in Tm, as compared to the unadducted duplexes. A similar increase in Tm is observed in the 5′-XY-3′ and 5′-XC-3′ duplexes. Therefore, it seems that the hydrogen bond between the AFB1 formamide moiety and the 3′- neighbor base, an intrastrand hydrogen bond, does not contribute to the stability of the AFB1–FAPY modified DNA duplex. Instead, the thermal stability of the β anomer of the AFB1–FAPY adduct in duplex DNA is likely attributable to favorable base stacking interactions.[56,57]

Biological Implications

Sequence-dependent hydrogen bonding patterns in the DNA major groove modulate E vs Z geometrical isomerism of the formamide moiety of the AFB1–FAPY adduct. The differential accommodation of these AFB1–FAPY adducts within the active site may, in part, modulate lesion bypass. It will be of interest to determine if sequence-dependent structural perturbations, observed here for DNA duplexes in vitro and at equilibrium, are translated into differential outcomes during error-prone replication bypass of the AFB1–FAPY adduct. Banerjee et al.[89] have obtained structural data for the error-prone bypass of the AFB1–FAPY adduct by the Sulfolobus solfataricus P2 DNA polymerase IV (Dpo4) in the 5′-XA-3′ sequence context. The Dpo4 polymerase conducts error-prone replication past the AFB1–FAPY adduct, including misinsertion of dATP, consistent with G → T transversions.[8,42−44] In the 5′-XA-3′ sequence, the structure of a ternary (Dpo4–DNA–dATP) AFB1–FAPY adducted template reveals that the oxygen atom of the FAPY formyl group participates in a water-mediated hydrogen bond with polymerase residue Arg332.[89] In this case, the AFB1–FAPY formamide group exists as the Z geometrical isomer.[89] Thus, geometrical isomerism of the AFB1–FAPY formamide group is modulated not only by DNA sequence but also by specific interactions within the active sites of specific error-prone bypass polymerases. It is also possible that geometrical isomerism of the AFB1–FAPY formamide group modulates the recognition and repair of these adducts in a sequence-specific manner. FAPY-type adducts are often substrates for base excision repair (BER),[90] but the AFB1–FAPY adduct is a substrate for nucleotide excision repair (NER) in bacterial,[91] yeast,[92] and mammalian cells.[93−95] These reports are consistent with the notion that NER targets bulky DNA lesions. On the other hand, the AFB1–FAPY lesion thermally stabilizes the DNA duplex.[56,85] Such thermal stabilization has been associated with resistance to NER.[96,97] Additionally, the data suggest differences in damage recognition between bacterial and mammalian repair systems. Croy and Wogan[46] noted that the AFB1–FAPY lesion is removed less efficiently in mammalian cells than is the initially formed AFB1–N7-dG cationic adduct, whereas both lesions are repaired with equal efficiencies in bacteria. More recently, Alekseyov et al.[91] confirmed that in Escherichia coli AFB1–FAPY lesions are efficiently repaired by NER. While Takahashi[98] had reported that XPA-deficient mice exhibited enhanced AFB1-induced liver tumorigenesis, Mulder et al.[99] did not observe increased repair of AFB1–FAPY adducts in lung or liver extracts of NER-compromised heterozygous p53 knockout mice and they also did not observe changes in XPA or XPB protein levels in the p53 knockout mice. Consequently, the precise factors modulating NER recognition and repair of AFB1–FAPY lesions remain incompletely understood. There is precedent for the idea that sequence-specific differences in DNA adduct structure may modulate NER efficiencies. Geacintov, Broyde, and co-workers[100−102] have investigated the sequence dependence of NER and have noted the potential for sequence specific differences in steric factors involving the minor groove-aligned PAH adducts and nearby guanine amino groups; they conclude that the NER apparatus binds productively or unproductively to the damaged DNA duplex depending upon the structural and stereochemical properties of specific DNA adducts.[102] Thus, it will be of interest to examine DNA sequence modulation of NER for AFB1–FAPY adducts.