Stereospecific Formation of Interstrand Carbinolamine DNA Crosslinks by Crotonaldehyde- and Acetaldehyde-Derived α-CH₃-γ-OH-1,N²-Propano-2’-deoxyguanosine Adducts in the 5′-CpG-3′ Sequence

Abstract

The crotonaldehyde- and acetaldehyde-derived R- and S-α-CH₃-γ-OH-1,N²-propanodeoxyguanosine adducts were monitored in single-stranded and duplex oligodeoxynucleotides using NMR spectroscopy. In both instances the cis and trans diastereomers of the α-CH₃ and γ-OH groups underwent slow exchange, with the trans diastereomers being favored. In single-stranded oligodeoxynucleotides, the aldehyde intermediates were not detected spectroscopically, but their presence was revealed through the formation of N-terminal conjugates with the tetrapeptide KWKK. When annealed into 5′-d(GCTAGCXAGTCC)-3′•5′-d(GGACTCYCTAGC)-3′ containing the 5′-CpG-3′ sequence context (X=R- or S-α-CH₃-γ-¹³C-OH-PdG; Y=¹⁵N²-dG), at pH 7, partial opening of the R- or S-α-CH₃-γ-¹³C-OH-PdG adducts to the corresponding N²-(3-oxo-1-methyl-propyl)-dG aldehydes was observed at temperatures below the T_m of the duplexes. These aldehydes equilibrated with their geminal diol hydrates; higher temperatures favored the aldehydes. When annealed opposite to T, the S-α-CH₃-γ-¹³C-OH-PdG adduct was stable. At 37 °C, an interstrand DNA crosslink was observed spectroscopically only for the R-α-CH₃-γ-OH-PdG adduct. Molecular modeling predicted that the interstrand crosslink formed by the R-α-CH₃-γ-OH-PdG adduct introduced less disruption into the duplex structure than did the crosslink arising from the S-α-CH₃-γ-OH-PdG adduct, due to differing orientations of the R- and S-CH₃ groups. Modeling also predicted that the α-methyl group of the aldehyde arising from the R-α-CH₃-γ-OH-PdG adduct oriented in the 3′ direction in the minor groove, facilitating crosslinking. In contrast, the α-methyl group of the aldehyde arising from the S-α-CH₃-γ-OH-PdG adduct oriented in the 5′ direction within the minor groove potentially hindering crosslinking. NMR revealed that for the R-α-CH₃-γ-OH-PdG adduct, the carbinolamine form of the crosslink was favored in duplex DNA, in situ, with the imine or Schiff base form of the crosslink remaining below the level of spectroscopic detection. Molecular modeling predicted that the carbinolamine linkage maintained Watson-Crick hydrogen bonding at both of the tandem C•G base pairs. Dehydration of the carbinolamine crosslink to an imine, or cyclization of the latter to form a pyrimidopurinone crosslink, required disruption of Watson-Crick hydrogen bonding at one or both of the crosslinked base pairs.

Introduction

Crotonaldehyde 1, an α,β-unsaturated aldehyde, is genotoxic and mutagenic in human lymphoblasts (1). It induces liver tumors in rodents (2). Michael addition of crotonaldehyde to deoxyguanosine yields the enantiomeric R- and S-α-CH₃-γ-OH-1,N²-propano-2’-deoxyguanosine adducts 2a and 2b (Scheme 1) (3–6); a second DNA adduction pathway yields paraldol-releasing DNA adducts (7), primarily N²-(3-hydroxybutylidene)-dG (8). Adducts 2a and 2b and are also formed through the reaction of acetaldehyde, a mutagen and potential human carcinogen (9), with deoxyguanosine (6, 10). The R- and S-α-CH₃-γ-OH-PdG adducts 2a and 2b were detected in human and rodent tissues (11, 12). In humans, these probably result from various endogenous and exogenous exposures, including lipid peroxidation (11, 13, 14), exposure to tobacco smoke (15, 16), and exposure to N-nitrosopyrrolidine (17, 18).

Scheme 1.

Equilibrium Chemistry of the R- and S-α-CH₃-γ-OH-PdG Adducts in the 5′-CpG-3′ Sequence in Duplex DNA.

Adducts 2a and 2b are structurally analogous to the major adduct produced by the reaction of acrolein with DNA, γ-OH-PdG (11, 13, 14, 19, 20). When placed opposite dC, opening of the γ-OH-PdG adduct resulted in formation of the N²-(3-oxopropyl)-dG aldehyde, and its hydrate, which oriented into the minor groove of the DNA (21). The presence of the N²-(3-oxopropyl)-dG aldehyde in the minor groove facilitated DNA interstrand crosslinking. In the 5′-CpG-3′ sequence context γ-OH-PdG formed crosslinks that were isolated, following NaCNBH₃ reduction, as saturated three-carbon interstrand N²,N²-dG linkages (22, 23). NMR analysis of the crosslinks, in duplex DNA, in situ, identified them as carbinolamines (24, 25). The γ-OH-PdG adduct also formed crosslinks with the N-terminal amine of the peptide KWKK (26).

The site-specific synthesis of oligodeoxynucleotides containing cyclic adducts 2a or 2b (22, 27, 28) enabled their chemistry and biology to be examined in duplex DNA. In duplex DNA at neutral pH, adducts 2a or 2b exhibited chemistry similar to that of the γ-OH-PdG adduct. When adducts 2a or 2b were placed into oligodeoxynucleotide duplexes at 5′-CpG-3′ sequences, Kozekov et al. (23) trapped interstrand saturated three carbon crosslinks by treatment with NaCNBH₃, corroborating the reports of Wang et al. (6, 8). Significantly, crosslinking was dependent upon the stereochemistry of the adduct, favoring the R-α-CH₃-γ-OH-PdG adduct 2a, as opposed to the S-α-CH₃-γ-OH-PdG adduct 2b. However, both the R-α-CH₃-γ-OH-PdG adduct 2a, and the S-α-CH₃-γ-OH-PdG adduct 2b readily formed N-terminal conjugates with the peptide KWKK (26). This implied that both the adducts 2a and 2b underwent opening to the corresponding aldehydes 3a or 3b when placed opposite dC in duplex DNA, but only aldehyde 3a efficiently formed interstrand DNA crosslinks in the 5′-CpG-3′ sequence.

Nevertheless, the chemistry of adducts 2a and 2b in DNA, including the identity of the interstrand crosslink formed in the 5′-CpG-3′ sequence by adduct 2a, remained elusive. The ability to trap a saturated three-carbon interstrand crosslink by treatment of the duplex with NaCNBH₃ implied the intermediacy of crosslinked imine 6a (23). However, imine 6a was anticipated to exist in equilibrium with carbinolamine 5a and, potentially, pyrimidopurinone 7a (23). Consistent with this expectation, ESI-Q-TOF mass spectrometry yielded signals consistent with the presence of carbinolamine 5a, and imine 6a and/or pyrimidopurinone 7a, with the latter signal predominating (10). On this basis, Lao and Hecht suggested that the imine 6a was the predominant species in duplex DNA, existing in equilibrium with the carbinolamine 5a and the pyrimidopurinone 7a. Consequently, it was of interest to examine the structures of crotonaldehyde- and acetaldehyde-derived adducts 2a and 2b in duplex DNA spectroscopically, and those of interstrand crosslinks in the 5′-CpG-3′ sequence.

In the present work, site-specific introduction of ¹³C at the aldehydic C_γ of crotonaldehyde (Scheme 2 and Scheme 3) and of ¹⁵N at the exocyclic amine of the targeted dG in the complementary strand of 5′-d(GCTAGCXAGTCC)-3′•5′-d(GGACTGYCTAGC)-3′ (X=R- or S- α-CH₃-γ-¹³C-OH-PdG; Y=¹⁵N²-dG) (Scheme 4), enabled the chemistry of α-CH₃-γ-¹³C-OH-PdG adducts 2a and 2b to be monitored. The chemistry of adducts 2a and 2b in duplex DNA, while similar to that of the acrolein-derived γ-OH-PdG adduct, differed in that the propensity for these 1,N²-dG adducts to open to the corresponding N²-(3-oxo-1-methyl-propyl)-dG aldehydes in duplex DNA was reduced. Moreover, their ability to form interstrand crosslinks in the 5′-CpG-3′ sequence was dependent upon stereochemistry of the methyl group at the α-carbon. Similar to the interstrand crosslink induced by the acrolein-derived γ-OH-PdG in the 5′-CpG-3′ sequence, crotonaldehyde-derived adduct 2a yielded carbinolamine crosslink 5a, with the corresponding imine 6a remaining below the level of detection. Molecular modeling suggested that interstrand crosslinks arising from the R-α-CH₃-γ-OH-PdG adduct introduced less disruption into the duplex than did those arising from the S-α-CH₃-γ-OH-PdG adduct, due to differing orientations of the R- and S-CH₃ groups. Moreover, for the R-α-CH₃-γ-OH-PdG adduct 2a, the methyl group was predicted to orient in the 3′ direction in the minor groove, favoring formation of carbinolamine crosslink 5a. In contrast, the lower level of interstrand crosslinking observed for the S-α-CH₃-γ-OH-PdG adduct 2b in the 5′-CpG-3′ sequence might be attributed, in part, to the unfavorable orientation of the methyl group in the 5′ direction within the minor groove, hindering reaction between aldehyde 3b and the targeted dG in the complementary strand. Modeling studies also predicted that the carbinolamine crosslink observed for the R-α-CH₃-γ-OH-PdG adduct maintained Watson-Crick hydrogen bonding at both of the tandem C•G base pairs. Dehydration of the carbinolamine crosslink to an imine, or cyclization of the latter to form a pyrimidopurinone crosslink, required disruption of Watson-Crick hydrogen bonding at one or both of the crosslinked base pairs.

Scheme 2.

Preparation the ¹³C-labeled amino diols used for site-specific synthesis of adducts 2a and 2b in oligodeoxynucleotides. The S diastereomer is shown. Reagents: (a) (Boc)₂O, 1 M NaOH, overnight, 81.5%, (b) MsCl, Et₃N, CH₂Cl₂, rt, 2 hr; K¹³CN, DMSO, 40 °C, 15 hr, 69% over 2 steps, (c) DIBAH, CH₂Cl₂, −78 °C, 32%, (d) Me₃PCH₂Cl,t-BuOK, THF, 70%, (e) OsO₄, NMP, THF/t-BuOH/H₂O, 76%, (f) Amberlist-H, CH₂Cl₂/CH₃OH; 4 M NH₃ in CH₃OH, 91%. The antipodal 4R-enantiomer 14a was prepared by an identical sequence starting from commercially available (R)-2-amino-1-propanol.

Scheme 3.

Site-specific synthesis of the ¹³C-labeled oligodeoxynucleotides 17, 17a containing adducts 2a and 2b.

Scheme 4.

The duplex oligodeoxynucleotide containing the 5′-CpG-3′ sequence context. X = α-CH₃-γ-¹³C-OH-PdG; Y = ¹⁵N²-dG.

Materials and Methods

Oligodeoxynucleotide synthesis

The preparation of non-isotopically labeled oligodeoxynucleotides containing site-specific γ-OH-PdG and α-CH₃-γ-OH-PdG adducts has been described (22, 27, 28). All commercially obtained chemicals were used as received. Methylene chloride was freshly distilled from calcium hydride. Anhydrous THF was freshly distilled from a sodium/benzophenone ketyl. All reactions were performed under argon atmosphere. Glassware was oven dried and cooled under argon. Concentrations of the single-stranded oligodeoxynucleotides were determined from calculated extinction coefficients at 260 nm (29). The purities of the modified oligodeoxynucleotides were analyzed using a PACE 5500 capillary electrophoresis (Beckman Instruments, Inc., Fullerton, CA) instrument. Electrophoresis was conducted using an eCAP ssDNA 100-R kit applying 12,000 V for 30 min. The electropherogram was monitored at 254 nm. MALDI-TOF mass spectra were measured on a Voyager-DE (PerSeptive Biosystems, Inc., Foster City, CA) instrument in negative reflector mode. The matrix contained 0.5 M 3-hydroxypicolinic acid and 0.1 M ammonium citrate.

Stereospecific synthesis of aminodiols 4-(R)- and 4(S)-aminopentane-1,2-¹³C-diol (14a, 14b)

These were prepared using identical methods starting with (R)- or (S)-2-amino-1-propanols respectively; the synthesis of 4(S)-aminopentane-1,2-(¹³C)-diol 14b is provided below.

tert-Butoxy (2-hydroxy-1(S)-methylethyl)carbamate (9b)

To a stirred solution of (S)-2-amino-1-propanol (8b, 0.38 g, 5 mmol) and 1N NaOH (5 mL) cooled to 0° C, was added dropwise di-tert-butyl dicarbonate (1.2 g, 5.5 mmol) in CH₂Cl₂ (10 mL). The mixture was stirred at room temperature overnight. The layers were separated and the organic phase was sequentially washed with 0.1 N HCl, 5% NaHCO₃ and brine, and dried over MgSO₄ and concentrated. Purification by flash chromatography on silica, eluting with 2–3% CH₃OH in CHCl₃, gave 9b, (0.714 g, 81.5 %): ¹H NMR (CDCl₃) δ 4.72 (br, 1H), 3.76 (m, 1H), 3.63 (m, 1H), 3.50 (m, 1H), 1.44 (s, 9H), 1.14 (m, 3H).

tert-Butoxy (2-(¹³C-cyano)-1(S)-methylethyl)carbamate (10b)

To a stirred solution of 9b (395 mg, 2.25 mmol) and trimethylamine (0.50 mL, 9.41 mmol, 3.6 mmol) in anhydrous CH₂Cl₂ (5 mL) cooled to 0° C, was added dropwise a solution of methanesulfonyl chloride (310 mg, 2.64 mmol) in anhydrous CH₂Cl₂ (2 mL). The reaction was stirred at room temperature for 2 h. The solution was then washed with saturated NaHCO₃ solution. The organic layer was separated, dried over K₂CO₃ and concentrated in vacuo. The crude mesylate was dissolved in DMSO (10 mL ) and K¹³CN (237 mg, 3.6 mmol ), which had been dried in the oven and ground into a powder, was added. The reaction was stirred at 40° C for 15 hr. After cooling at room temperature, water (20 mL) was added and the mixture was extracted with ether (3 × 10 mL). The combined organic phases were washed with a saturated brine (3 × 10 mL), dried over MgSO₄, filtered, and concentrated in vacuo. Flash chromatography eluting with 12–15% ethyl acetate in hexane gave 10b (286 mg, 69.0%): ¹H NMR (CDCl₃) δ 4.62 (br, 1H), 3.96 (m, 1H), 2.80 (m, 1H), 2.56 (m, 1H), 1.45 (s, 9H), 1.32 (d, 3H, J = 6.8 Hz).

tert-Butoxy (1(S)-methyl-3-¹³C-oxopropyl)carbamate (11b)

A solution of 10b (258 mg, 1.4 mmol) in anhydrous CH₂Cl₂ (3 mL) was cooled to −78° C, then DIBAL-H (1.0 M in CH₂Cl₂ , 2.1 mL) was added dropwise over 20 min. The reaction was stirred below −70° C for 10 min, then quenched by the addition of acetone (1 mL) followed by saturated NH₄Cl (10 mL). The resulting mixture was allowed to warm to room temperature and stirred for 40 min. The mixture was filtered through a pad of Celite and the filtrate was washed with brine, dried over MgSO₄ and concentrated in vacuo. Purification by flash chromatography on silica gel, eluting with 20–25% ethyl acetate in hexane, gave 11b (88 mg, 32 %)²: ¹H NMR (CDCl₃) δ 10.05 and 9.47 (d × m, 1H, J_C-H= 175.5 Hz), 4.99 (br, 1H), 4.14 (m, 1H), 2.64 (m, 2H), 1.43 (s, 9H), 1.23 (m, 3H).

tert-Butoxy (1(S)-methylprop-3-¹³C-enyl)carbamate (12b)

To a stirred suspension of methyltriphenylphosphonium bromide (214 mg, 0.6 mmol) in THF (5 mL) was added a solution of potassium tert-butoxide (1M in THF; 56 µL, 0.56 mmol)] and the resulting solution stirred for 30 min. The solution was cooled to 0 °C, then a solution of 11b (88 mg, 0.47 mmol) in THF (2 mL) was added dropwise. After the addition was complete, the reaction was warmed to room temperature and stirred for 1 h. The reaction was then quenched by the addition of saturated aqueous NH₄Cl and the layers separated. The aqueous phase was extracted with ether (3×20 mL), dried over Na₂SO₄, and concentrated in vacuo. Purification by flash chromatography on silica gel, eluting with 3% ethyl acetate in hexane, gave 12b (61 mg, 70.0%) as a colorless oil: [α]²⁰_D –19.3° (c 1.10, CHCl₃); ¹H NMR (CDCl₃) δ 5.96 and 5.58 (d×m, 1H, J_C-H = 144 Hz), 5.01 (m, 2H), 4.38 (br, 1H), 3.73 (m, 1H), 2.20 (m, 2H), 1.44 (s, 9H), 1.12 (d, 3H, J = 6.6 Hz).

tert-Butoxy (1(S)-methyl-[3,4-dihydroxy-3-¹³C-propyl]) (13b)

To a stirred solution of 12b (42 mg, 0.225 mmol), N-methylmorpholine N-oxide (30 mg, 0.248 mmol), THF (0.8 mL), t- BuOH (0.32 mL), and water (0.16 mL) was added osmium tetroxide (10 µL, 0.1 M solution in benzene, 1.0 µmol). The mixture was stirred for 12 h and then a second portion of osmium tetroxide (6 µL, 0.1 M solution in benzene, 0.6 µmol) was added and stirring continued for an additional 8 h. The reaction was then quenched with 5% aqueous NaHSO₃ (5 mL) with vigorously stirring for 15 min then poured into water (10 mL) and extracted with methylene chloride (3 _ 10 mL). The combined organic extracts were dried over Na₂SO₄, filtered and concentrated in vacuo. Purification by flash chromatography on silica, eluting with 1–2 % methanol in chloroform, gave 13b (37.5 mg, 76.0 %) as an inseparable mixture of diastereomers: ¹H NMR (major, CDCl₃) δ 4.55-4.50 (m, 2H, NH, OH), 3.95 (m, 1H), 3.93 and 3.47 (d×m, 1H, J_C-H = 138 Hz), 3.50 (m, 2H), 2.48 (br, 1H, OH), 1.53 (m, 1H), 1.45 (s, 9H), 1.26 (m, 1H), 1.19 (d, 3H, J = 6.7 Hz); (minor, CDCl₃) δ 4.61 (m, 1H), 4.04 & 3.57 (m and m, 1H, J = 142 Hz), 3.79 (m, 1H), 3.64 (m, 1H), 3.47 (m, 1H), 3.22 (br, 1H), 2.40 (br, 1H), 1.60 (m, 2H), 1.44 (s, 9H), 1.19 (d, 3H, J = 6.6 Hz).

4(S)-amino-1,2-dihydroxy-2-¹³C-pentane (14b)

To a solution of 13b (30 mg, 0.136 mmol) in CH₂Cl₂ (1.5 mL) and methanol (0.5 mL) was added H-Amberlyst 15 resin (0.3 g). The mixture was gently shaken over 14 h until TLC analysis showed the complete disappearance of 13b. The resin was removed by filtration and successively washed with THF and methanol. This amine-bound resin was transferred to a solution of 4 M ammonia in methanol and gently shaken for 50 min. The resin was then removed by filtration and washed with methanol (3 × 5 mL). The combined filtrates were evaporated in vacuo to give 14b (15 mg, 91%) which was used without further purification: ¹H NMR (CD₃OD) δ 3.94 & 3.58 (m and m, 1H, J = 144 Hz), 3.45 (m, 2H), 3.24 (m, 1H), 1.56 (m, 2H), 1.18 (d, 3H, J = 6.5 Hz).

Synthesis of stereoisomeric R-and S-α-CH₃-γ-¹³C-OH-PdG oligodeoxynucleotides

The O⁶-TMSE-2-fluoroinosine-modified oligodeoxynucleotide 15 (125 A₂₆₀ units) (30, 31) was mixed in a plastic test tube with diisopropylethylamine (150 µL), DMSO (300 µL), and either 4(R)- or 4(S)-aminopentane-1,2-(¹³C)-diol (14a, 14b) (5 mg). The reaction mixtures were stirred at 55 °C for 24 h. The solvents were evaporated in vacuo using a centrifugal evaporator. The residues were dissolved in 5% acetic acid (500 µL) and stirred for 2 h at room temperature to remove the O⁶-TMSE group. The mixtures were neutralized with 1 M NaOH and purified by HPLC using gradient A (see below) to give the corresponding stereoisomeric N²-(1-methyl-3,4-dihydroxybutyl)guanine-modified oligodeoxynucleotides 16a and b. ¹³C-labeled R-stereoisomer: 54.0 A₂₆₀ units (43%). MALDI-TOF MS: calcd for [M − H]⁻ 3747.7, found 3747.4. ¹³C-labeled S-stereoisomer: 65.0 A₂₆₀ units (52%); MALDI-TOF MS: calcd for [M − H]⁻ 3747.7, found 3747.1

An aqueous solution of NaIO₄ (500 µL, 20 mM) was added to solutions of N²-(1-methyl-3,4-dihydroxybutyl)guanine-modified oligodeoxynucleotides 16, 16b (54.0 A₂₆₀ units) in 0.05 M, pH 7.0 phosphate buffer, (500 µL) and the reaction mixtures were stirred at room temperature for 10 min. The mixtures were purified by HPLC using gradient B (see below) to give 8-hydroxy-6(S or R)-methyl-5,6,7,8-tetrahydropyrimido[1,2-a]purin-10(3H)-one) adducts 17a and b. ¹³C-labeled R-stereoisomer: 41.8 A₂₆₀ units (77%); MALDI-TOF MS: calcd for [M − H]⁻ 3715.7, found 3715.7. ¹³C-labeled S-stereoisomer: 61.5 A₂₆₀ units (90%) from 65 A₂₆₀ units MALDI-TOF MS: calcd for [M − H]⁻ 3715.7, found 3716.6.

Synthesis of ¹⁵N²-dG-modified complementary oligodeoxynucleotide 5′-GGACTCGCTAGC-3′

The O⁶-TMSE-2-fluoroinosine-modified oligodeoxynucleotide (31) was deprotected using 6 M ¹⁵NH₄OH, desilyated with 5 % acetic acid, and purified by C8 HPLC in 0.1 M ammonium formate (pH 6.5), yielding the ¹⁵N²-dG-modified oligodeoxynucleotide. Negative ion MALDI-TOF mass spectrometry yielded m/z 3645.9 (calcd for [M−H]⁻ 3645.6).

HPLC separations

The purifications of oligodeoxynucleotides 16a, 16b and 17a, 17b were performed on a Beckman HPLC system (32 Karat software version 3.1, pump module 125) with a diode array UV detector (module 168) monitoring at 260 nm using Phenomenex Luna 5µ C8 column (250 mm × 10 mm i.d., 3 mL/min for purification) with 0.1 M aqueous ammonium formate and CH₃CN for oligonucleotides. HPLC gradients: A) 1–7 % acetonitrile over 25 min, hold for 3 min, 7–99% acetonitrile over 2 min, hold for 5 min, hold for 2 min, and then to 1% acetonitrile over 2 min, hold for 5 min; B) 1–6.7 % acetonitrile over 25 min, hold for 10 min, and then to 1% acetonitrile over 2 min, hold for 5 min.

Trapping of Covalent DNA—Peptide Complexes Using NaCNBH₃

The peptide KWKK was prepared by the National Institute of Environmental Health Sciences Center Protein Chemistry Laboratory at the University of Texas Medical Branch, Galveston, TX. Following initial synthesis, the peptide was analyzed by mass spectrometry, purified by preparative HPLC, and resuspended in a solution of 20:80 acetonitrile/water. The concentration of KWKK in solution was determined by monitoring Trp absorbance at 280 nm. The concentration was calculated using 5500 M⁻¹ cm⁻¹ as the Trp molar extinction coefficient. Adducted oligodeoxynucleotides (non-¹³C or ¹⁵N-isotopically labeled) were γ-³²P-labeled on the 5′ end with T4 polynucleotide kinase following standard procedures. For trapping reactions, adducted DNA (75 nM) was incubated with peptide in 100 mM HEPES (pH 7.0) and 100 mM NaCl at 4 °C. The concentrations of KWKK and NaCNBH₃ in trapping reactions were 1.0 mM and 50 mM, respectively, and reactions were quenched by the addition of 100 mM NaBH₄. (Aqueous solutions of both NaCNBH₃ and NaBH₄ were prepared fresh on the day of use.) Each reaction mixture was subsequently diluted 5-fold by the addition of 1.25× loading buffer (59 % v/v formamide, 12.5 mM EDTA, 0.012 % w/v bromophenol blue, 0.012 % xylene cyanol) and heated at 90°C for 3 min. The products of each reaction were separated on a 15% denaturing polyacrylamide gel (8.3 M urea) in sequencing buffer (134 mM Tris base, 44 mM boric acid, 10 mM EDTA) for 5 h at 1500 V. Results were visualized from wet gels by phosphorimager analysis, and product bands were quantitated using ImageQuant (5.0) software.

NMR

The isotopically labeled α-CH₃-γ-OH-PdG-modified oligodeoxynucleotide duplexes 5′-d(GCTAGCXAGTCC)-3′•5′-d(GGACTCYCTAGC)-3′ (X = α-CH₃-γ-OH-PdG; Y = dC, T, or dA) (Scheme 4) were annealed in a buffer consisting of 10 mM NaH₂PO₄, 0.1 M NaCl, and 50 µM Na₂EDTA (pH 7.0). The duplex oligodeoxynucleotides were eluted from DNA Grade Biogel hydroxylapatite (Bio-Rad Laboratories, Hercules, CA) with a gradient from 10 to 200 mM NaH₂PO₄ (pH 7.0). They were desalted using Sephadex G-25. The duplex oligodeoxynucleotides were prepared at a concentration of 1 mM in 0.3 mL of 9:1 H₂O:D₂O containing 10 mM NaH₂PO₄, 0.1 M NaCl, and 50 µM Na₂EDTA (pH 7.0). They were placed in micro-NMR tubes (Shigemi Glass, Inc., Allison Park, PA). NMR experiments were carried out at ¹H frequencies of 500.13 or 600.13 MHz (¹³C frequencies of 125 or 150 MHz; ¹⁵N frequencies of 50.66 or 60.79 MHz). One-dimensional ¹³C NMR was conducted using a probe with inner-coil ¹³C geometry using inverse-gated ¹H WALTZ16 decoupling. Typical acquisition parameters were 16 K total data points, with a digital resolution of 1.3 Hz/pt, 12K scans, and a relaxation delay of 8 s. ¹³C HSQC experiments were performed using standard ¹H-detected pulse programs with States-TPPI phase cycling and watergate water suppression (32). Typical experimental parameters were 512 FIDs, each of 2K points. The ¹³C sweep width was varied from 20 to 180 ppm. ¹⁵N-HSQC spectra (33) were recorded with 8/180 scans per increment, using States-TPPI phase cycling, a delay time 1/2 ¹J_N-H of 5.56 ms, 1536 complex data points for 10,000 Hz in the acquisition dimension and 256 points in the indirect dimension, covering 10,136.8 Hz, centered at 100 ppm. A relaxation delay of 1.5 sec was used. Fully coupled and ¹⁵N decoupled spectra were recording using the same parameters, leaving the ¹⁵N decoupling off during the acquisition time of the coupled spectrum. TOCSY-HSQC experiments (34) were recorded applying States phase cycling, 60 ms isotropic mixing time applied with a 10 KHz dipsi spin lock pulse sequence optimized for a 90 Hz ¹J_N-H coupling. 1536 complex data points for 10000.0 Hz in the acquisition dimension and 128 points in the indirect dimension, covering 1000.0 Hz centered around 106 ppm were measured. A relaxation delay of 1.2 sec was used and ¹⁵N was fully decoupled during the acquisition time. NOESY-HSQC (34, 35) experiments were recorded applying States phase cycling, 150 ms mixing time and were optimized for a 90 Hz ¹J_N-H coupling. 1536 complex data points for 10000.0 Hz in the acquisition dimension and 128 points in the indirect dimension, covering 1000.0 Hz centered at 106 ppm were measured. A relaxation delay of 1.5 sec was used and ¹⁵N was fully decoupled during the acquisition time. ¹H chemical shifts were referenced to water. ¹³C and ¹⁵N chemical shifts were referenced indirectly (36–38). NMR data were processed on Silicon Graphics Octane workstations using the programs FELIX2000 (Accelrys, Inc., San Diego, CA) or NMRPipe (39).

Molecular modeling

Modeling was performed on Silicon Graphics Octane workstations using the program AMBER 8.0 (40). Classical B-DNA was used as a reference structure to create starting structures for potential energy minimization (41). Diastereomers at the γ-carbon of carbinolamine crosslink 5a and pyrimidopurinone crosslink 7a were constructed using the BUILDER module of INSIGHT II (Accelrys, Inc., San Diego, CA). The program ANTECHAMBER was used and the atom types were based on AMBER atom types for parameterization. RESP atomic charges were calculated using GAUSSIAN98 (42) and the Hartree-Fock 6-31G* basis set. The generalized Born model for solvent (43, 44) was utilized for potential energy minimization. Potential energy minimization used the AMBER 8.0 force field.

Results

Site-Specific Synthesis of ¹³C-Labeled R- and S-α-CH₃-γ-¹³C-OH-PdG Oligodeoxynucleotides

Stereospecific synthesis of the ¹³C-labeled adducted oligodeoxynucleotides was accomplished using the post-oligomerization strategy previously employed for related modified oligodeoxynucleotides (22, 27, 28). This involved the incorporation of an electrophilic base, 2-fluoro-O⁶-(2-trimethylsilylethyl)-2′-deoxyinosine (31), into an oligodeoxynucleotide via the corresponding phosphoramidite, followed by displacement of the fluoro group by an amine analogue of the mutagen via a nucleophilic aromatic substitution reaction (Scheme 3). A vicinal diol unit was used as a surrogate for the aldehyde group (Scheme 2), which was cleaved with sodium periodate after the adduction reaction to give the desired modified oligodeoxynucleotide. A significant advantage of this strategy was that access to both stereoisomers 2a and 2b in the resulting adducted oligodeoxynucleotides 17a, 17b was obtained by individually reacting the (R) and (S)-stereoisomers of the amines 14a, 14b with the same oligodeoxynucleotide containing the 2-fluoroinosine base (15, Scheme 3).

The synthesis of the ¹³C-labeled amino diols (14a, 14b) is shown in Scheme 2. Commercially available (S)-2-amino-1-propanol 8b was N-protected as the corresponding Boc derivative 9b. The hydroxyl group was then converted to the mesylate and displaced with ¹³C-labeled potassium cyanide to give nitrile 10b. Reduction of the nitrile to the aldehyde (11b) was followed by Wittig methylenation to olefin 12b in acceptable overall yield. Treatment of the olefin with osmium tetroxide gave diol 13b as a mixture of stereoisomers. Since the diol was eventually cleaved to the aldehyde, the stereochemistry of the diol was of no consequence. Deprotection gave 4S-amino-pentane-1,2-diol (14b). The antipodal 4R-enantiomer 14a was prepared by an identical sequence starting from commercially available (R)-2-amino-1-propanol.

Epimerization of R- and S-α-CH₃-γ-OH-PdG Oligodeoxynucleotide Adducts

The single stranded 5′-d(GCTAGCXAGTCC)-3′, modified with either the R- or S-α-CH₃-γ-OH-PdG adducts, was examined using ¹³C HSQC NMR (Figure 1). At pH 7 and 15 °C, a single crosspeak was observed for R-α-CH₃-γ-OH-PdG adduct 2a, at a ¹³C chemical shift of 71.9 ppm, and a ¹H chemical shift of 5.96 ppm. This crosspeak was assigned as the epimer in which the α-CH₃ and γ-OH groups were in the trans configuration. For S-α-CH₃-γ-OH-PdG adduct 2b, two crosspeaks were observed at pH 7 and 15 °C. The major crosspeak was observed at a ¹³C chemical shift of 71.5 ppm, and a ¹H chemical shift of 5.89 ppm, also assigned as the epimer in which the α-CH₃ and γ-OH groups were in the trans configuration. The minor crosspeak was observed at a ¹³C chemical shift of 71.6 ppm, and a ¹H chemical shift of 6.12 ppm. It was assigned as the hydroxyl epimer at the γ-carbon in which the α-CH₃ and γ-OH groups were in the cis configuration. In both instances, as temperature was increased to 37 °C, increasing amounts of the cis epimers appeared in the ¹³C HSQC spectra, as measured by comparative volume integrations of the two resonances as a function of temperature. For R-α-CH₃-γ-OH-PdG adduct 2a, the cis epimer crosspeak was observed at a ¹³C chemical shift of 71.3 ppm, and a ¹H chemical shift of 6.10 ppm. No resonances for γ-¹³C aldehydes 3a or 3b, or hydrated aldehydes 4a or 4b were observed, suggesting that at equilibrium, the levels of these ring-opened species remained below the spectroscopic limit of detection.

Figure 1.

¹³C HSQC spectra of R- and S-α-CH₃-γ-OH-PdG adducts 2a and 2b in the oligodeoxynucleotide 5′-d(GCTAGCXAGTCC)-3′ at 15° and 37° C. A. R-α-CH₃-γ-OH-PdG, adduct 2a. B. S-α-CH₃-γ-OH-PdG, adduct 2b.

To detect the transient presence of aldehydes 3a or 3b, a series of peptide trapping experiments were performed. The single-stranded oligodeoxynucleotide 5′-d(GCTAGCXAGTCC)-3′ containing either a R- or S-α-CH₃-γ-OH-PdG adduct was treated with the peptide KWKK for 0–120 min in the presence of NaCNBH₃. Reaction mixtures were quenched at the designated time points by adding NaBH₄ to reduce the aldehyde substrate. A gel-shifted complex was observed by denaturing PAGE analysis and designated as DNA—peptide crosslink (Figure 2, 12-mer + KWKK), consistent with the transient presence of aldehydes 3a or 3b in single-stranded DNA. The accumulation of this product band was monitored over a 2 hr time course (Figure 2), at 4, 15, 37 or 50 °C. Higher temperatures facilitated faster formation of the peptide-DNA conjugate. At 4 °C, there was little complex accumulation over the 2 hr time course, whereas at 50 °C a substantial amount of complex accumulated over this time period. These results were consistent with the NMR data, in which the rate of epimerization of the R- or S-α-CH₃-γ-OH-PdG adducts increased at higher temperatures in single-stranded DNA.

Figure 2.

DNA-peptide crosslinking involving R- and S-α-CH₃-γ-OH-PdG adducts. A. For trapping reactions, single-stranded crotonaldehyde-adducted oligodeoxynucleotides (75 nM) were incubated with 1.0 mM KWKK the presence of 50 mM NaCNBH₃ at 4, 15, 37 or 50 °C. Reactions were carried out in 100 mM HEPES (pH 7.0) and 100 mM NaCl and were incubated for 0, 15, 30, 60, 90 or 120 min. Reactions were quenched at the end of the incubation period by the addition of 100 mM NaBH₄. Labels indicate the positions of the substrate 12-mer DNAs and the major reduced Schiff base conjugates (12-mer + peptide) following denaturing PAGE analysis. B. Kinetics of trapped conjugate formation are plotted over the 2 h time course at 4 °C [R-α-CH₃-γ-OH-PdG, π; S-α-CH₃-γ-OH-PdG, ρ], 15 °C [R-α-CH₃-γ-OH-PdG, ´; S-α-CH₃-γ-OH-PdG, ≤], 37 °C [R-α-CH₃-γ-OH-PdG, ; S-α-CH₃-γ-OH-PdG, ], and 50 °C [R-α-CH₃-γ-OH-PdG, θ; S-α-CH₃-γ-OH-PdG, σ].

On the ms time scale of the NMR experiments, the two epimers of the γ-hydroxyl groups of the of 1,N²-dG adducts 2a and 2b were in slow exchange. A series of ¹³C HSQC spectra collected as a function of temperature enabled van’t Hoff analysis (Figure 3). These studies revealed that for the R adduct 2a, the value of ΔH for the cis to trans interconversion was −14 kcal/mol and the value of ΔS for the interconversion was −42 cal/mol K. For the S adduct 2b, the value of ΔH for the cis to trans interconversion was −10 kcal/mol and the value of ΔS for the interconversion was −29 cal/mol K.

Figure 3.

van't Hoff analysis of the epimerization of R- and S-α-CH₃-γ-OH-PdG adducts 2a and 2b in the oligodeoxynucleotide 5′-d(GCTAGCXAGTCC)-3′. A. R-α-CH₃-γ-OH-PdG, adduct 2a. ΔH_cis→trans = −14 kcal/mol; ΔS_cis→trans = −42 cal/mol K. B. S-α-CH₃-γ-OH-PdG, adduct 2b. ΔH_cis→trans = −10 kcal/mol; ΔS_cis→trans = −29 cal/mol K.

Equilibrium Chemistry of the R-α-CH₃-γ-OH-PdG Adduct in Duplex DNA

The R- α-CH₃-γ-¹³C-OH-PdG adduct 2a was placed opposite dC in 5′-d(GCTAGCXAGTCC)-3′•5′-d(GGACTCGCTAGC)-3′ at pH 7, and the sample was allowed to equilibrate at 37 °C. An inverse-gated ¹³C spectrum was obtained immediately upon annealing the duplex. The γ-¹³C resonances from aldehyde 3a and hydrated aldehyde 4a were detected, indicating that opening of adduct 2a occurred before the ¹³C spectrum could be collected. At pH 7, opening of the crotonaldehyde-derived 1,N²-dG adduct 2a to aldehyde 3a and hydrated aldehyde 4a was incomplete. After 20 days, no further spectroscopic changes were observed. At equilibrium, the γ-¹³C resonance appeared as a mixture of four species. Furthest downfield, at 208 ppm, was a resonance assigned as γ-¹³C aldehyde 3a. A second γ-¹³C resonance, assigned as hydrated aldehyde 4a, was observed at 90 ppm. The third resonance, identified as carbinolamine crosslink 5a, was observed at 73 ppm. This resonance increased in intensity over a period of 20 days at 37 °C. The two γ-hydroxyl diastereomers of crosslink 5a were not resolvable in the ¹³C spectrum, but were resolved using ¹H and ¹⁵N NMR, as will be discussed below. The failure to observe a γ-¹³C resonance in the 140–160 ppm spectral region, the range in which a resonance arising from γ-¹³C imine 6a would be anticipated, indicated that the amount of imine 6a in equilibrium with carbinolamine 5a was below the level of detection by ¹³C NMR. This placed an upper limit on the amount of 5′-CpG-3′ imine crosslink 6a in equilibrium with carbinolamine crosslink 5a, estimated to be ≤ 5%. A fourth resonance, assigned as cyclic adduct 2a, was observed at 72 ppm. The ~1 ppm ¹³C chemical shift difference of adduct 5a as compared to adduct 2a was consistent with the expectation that the γ-¹³C nuclei in adducts 2a and 5a, both of which were bonded to hydroxyl groups, should exhibit similar chemical shifts.

Confirmation of the assignment of carbinolamine crosslink 5a came from a series of ¹⁵N-HSQC and ¹⁵N-edited HSQC and NOESY experiments (Figure 5). The ¹⁵N HSQC experiment revealed crosspeaks corresponding to the anticipated diastereomers of carbinolamine crosslink 5a. The stronger of these two crosspeaks exhibited a ¹⁵N chemical shift of 106 ppm and a ¹H chemical shift of 8.4 ppm. This crosspeak exhibited a 90 Hz coupling constant. The weaker of the two crosspeaks was observed at a ¹⁵N chemical shift of 96 ppm and a ¹H chemical shift of 8.7 ppm. Two additional weaker crosspeaks in the ¹⁵N-HSQC spectrum were assigned as arising from non-crosslinked oligodeoxynucleotide, in which ¹⁵N²-dG-labeled base pair C⁶•Y¹⁹ retained Watson-Crick hydrogen bonding. The crosspeak at 8.0 in the ¹H dimension was assigned as arising from the hydrogen bonded amino proton, whereas that at 6.5 in the ¹H dimension arose from the non-hydrogen bonded amino proton. An additional minor ¹⁵N-HSQC, labeled as peak e in Figure 5a, remained unidentified.

Figure 5.

A. ¹⁵N HSQC spectrum of R-α-CH₃-γ-OH-PdG in the oligodeoxynucleotide 5′-d(GCTAGCXAGTCC)-3′•5′-d(GGACTCYCTAGC)-3′. Crosspeaks a, major stereoisomer of the carbinolamine crosslink 5a; b, minor stereoisomer of the carbinolamine crosslink 5a; c, d, hydrogen- and non-hydrogen-bonded ¹⁵N²H protons of non-crosslinked base pair C⁶•Y¹⁹; e, unidentified crosspeak. B. ¹⁵N NOESY HSQC spectrum. Crosspeaks a, Y^{19 15}N²H→Y¹⁹ N1H; b, Y^{19 15}N²H→X⁷ N1H; c,Y^{19 15}N²H autocorrelation; d, Y^{19 15}N²H→X⁷ N²H; e, Y^{19 15}N²H→H_γ; f, Y^{19 15}N²H→H_α; g, Y^{19 15}N²H→H_β′β″; h, Y^{19 15}N²H→α-CH₃; i, j, hydrogen-and non-hydrogen-bonded ¹⁵N²H protons of non-crosslinked pair C⁶•Y¹⁹. C. ¹⁵N TOCSY HSQC spectrum. Crosspeaks a, autocorrelation peak for major stereoisomer of carbinolamine crosslink 5a; b, coupling to H_γ; c, couplings to H_β′,β″.

The ¹⁵N-HSQC-filtered TOCSY experiment (Figure 5) established that the ¹H signal at 8.35 ppm, assigned as Y¹⁹ N²H in carbinolamine crosslink 5a, exibited scalar coupling to protons of the crosslink crotonaldehyde moiety. The signal observed at 5.79 ppm indicated coupling to H_γ of the crotonaldehyde crosslink. Crosspeaks at 1.5 and 2.2 ppm were observed to the H_β′,β″ crotonaldehyde protons. No crosspeaks were observed for the minor diastereomer of the carbinolamine crosslink, presumably due to its low abundance.

A ¹⁵N HSQC-filtered NOESY experiment (Figure 5) revealed that for the major diastereomer of crosslink 5a, the Y¹⁹ N²H→Y¹⁹ N1H NOE was observed at 12.8 ppm, in the expected chemical shift range for this imino proton involved in Watson-Crick hydrogen bonding. For the major diastereomer, NOEs were observed from Y¹⁹ N²H to H_α, H_β′,β″, H_γ, and the methyl protons of the crotonaldehyde crosslink. For the minor diastereomer of crosslink 5a, the Y¹⁹ N²H→Y¹⁹ N1H NOE was observed at 13.0 ppm, also in the expected chemical shift range for a Watson-Crick hydrogen bonded imino proton.

The assignment of carbinolamine crosslink 5a was corroborated by a triple resonance HNC experiment, in which the complementary strand of the duplex was site-specifically labeled with ¹⁵N²-dG at the crosslinked dG residue (Figure 6). This experiment exploited the fact that crosslink formation resulted in bonding between the ¹⁵N²-dG and γ-¹³C isotopes. A correlation was observed between the 73 ppm γ-¹³C resonance, and a ¹⁵N resonance at 106 ppm, establishing that these resonances arose from the same chemical species observed in ¹⁵N HSQC experiments, and assigned as carbinolamine 5a. No correlation was observed between the signal arising from the minor diastereomer observed at 96 ppm in the ¹⁵N HSQC spectrum, and ¹³C, presumably because of the low abundance of the minor diastereomer of crosslink 5a observed in the ¹⁵NHSQC experiments.

Figure 6.

Triple resonance ¹H¹⁵N¹³C spectrum of R-α-CH₃-γ-OH-PdG in the oligodeoxynucleotide 5′-d(GCTAGCXAGTCC)-3′•5′-d(GGACTCYCTAGC)-3′, confirming the presence of crosslinked carbinolamine 5a.

Equilibrium Chemistry of the S-α-CH₃-γ-OH-PdG Adduct in Duplex DNA

The S-α-CH₃-γ-¹³C-OH-PdG adduct differed from the R-α-CH₃-γ-¹³C-OH-PdG adduct. At equilibrium, only low levels of interstrand crosslinks were observed in the 5′-CpG-3′ sequence, as demonstrated by reductive trapping with NaCNBH₄ (23). The S-α-CH₃-γ-¹³C-OH-PdG adduct 2b was placed opposite dC in 5′-d(GCTAGCXAGTCC)-3′•5′-d(GGACTCGCTAGC)-3′ at pH 7 and 37°C, and an inverse-gated ¹³C spectrum was obtained immediately upon annealing the duplex at 37 °C. Similar to the R-α-CH₃-γ-¹³C-OH-PdG adduct, the γ-¹³C resonances from aldehyde 3b and hydrated aldehyde 4b were detected. Thus, opening of cyclic adduct 2b occurred before the ¹³C spectrum could be collected. Also similar to the R-α-CH₃-γ-¹³C-OH-PdG adduct, at pH 7, opening of crotonaldehyde-derived 1,N²-dG adduct 2b to aldehyde 3b and hydrated aldehyde 4b was incomplete. After 20 days, no further spectroscopic changes were observed. Both the R-α-CH₃-γ-¹³C-OH-PdG and S-α-CH₃-γ-¹³C-OH-PdG exhibited similar quantities of the 1,N²-dG cyclic adducts 2a or 2b in equilibrium with aldehydes 3a or 3b and hydrated aldehydes 4a or 4b, suggesting that the positions of the equilibria involving the 1,N²-dG adducts and their ring-opened rearrangement products were independent of stereochemistry at C_α of the crotonaldehyde moiety. Significantly, however, and corroborating the reductive trapping experiments (23), ¹³C NMR failed to detect crosslink 5b, confirming that formation of the interstrand crosslink in the 5′-CpG-3′ sequence was dependent upon stereochemistry at C_α of the crotonaldehyde moiety.

Figure 7 shows the ¹³C NMR spectrum of the S-α-CH₃-γ-¹³C-OH-PdG adduct 2b when placed opposite dC in 5′-d(GCTAGCXAGTCC)-3′•5′-d(GGACTCGCTAGC)-3′ at 37 °C, at pH values of 4.7, 9.3, and 10.7. At pH 4.7, the equilibrium between 1,N²-dG adduct 2b and N²-(3-oxopropyl)-dG aldehyde 3b and its hydrate 4b favored cyclic adduct 2b. Increasing the pH to 9.3 favored formation of N²-(3-oxopropyl)-dG aldehyde 3b and its hydrate 4b. At pH 10.7, denaturation of the oligodeoxynucleotide duplex occurred, and only cyclic adduct 2b was observed.

Figure 7.

Chemical species arising from S-α-CH₃-γ-OH-PdG in the oligodeoxynucleotide 5′-d(GCTAGCXAGTCC)-3′•5′-d(GGACTCYCTAGC)-3′ as a function of pH. A. pH 9.3. B. pH 10.7. C. pH 4.7. Crosspeaks a, aldehyde 3b; b, hydrated aldehyde 4b; c, cyclic adduct 2b.

Mispairing of T Opposite the γ-OH-PdG Adduct

Figure 8 shows the ¹³C NMR spectrum of the S-α-CH₃-γ-¹³C-OH-PdG adduct 2b when placed opposite T in 5′-d(GCTAGCXAGTCC)-3′•5′-d(GGACTTGCTAGC)-3′ at 37 °C, at pH 7. Under these conditions, cyclic adduct 2a was favored, with N²-(3-oxopropyl)-dG aldehyde 3b and its hydrate 4b remaining below the level of detection by ¹³C NMR.

Figure 8.

¹³C NMR of S-α-CH₃-γ-OH-PdG in the oligodeoxynucleotide 5′-d(GCTAGCXAGTCC)-3′•5′-d(GGACTTGCTAGC)-3′. Only cyclic adduct 2b is observed.

Molecular Modeling

The two carbinolamine crosslinks 5a and 5b arising from interstrand crosslinking by adducts 2a or 2b, respectively, were modeled in 5′-d(GCTAGCXAGTCC)-3′•5′-d(GGACTCGCTAGC)-3′. The model structures were subjected to potential energy minimization using the conjugate gradients algorithm in AMBER 8.0 (Figure 9). The potential energy minimization predicted that for crosslink 5a, arising from R-α-CH₃-γ-OH-PdG, the methyl group projected into the minor groove, without disruption of duplex DNA structure. In contrast, the calculations predicted that for crosslink 5b, arising from S-α-CH₃-γ-OH-PdG, the methyl group interfered with the 3′-neighbor base pair A⁸•T¹⁷, presumably reducing the stability of the crosslinked duplex.

Figure 9.

Molecular modeling of carbinolamine interstrand crosslinks formed in the 5′-CpG-3′ sequence by the R- and S-α-CH₃-γ-OH-PdG adducts 2a and 2b. A. Crosslink formed by R adduct 2a. B. Crosslink formed by S adduct 2b.

Additionally, the two N²-(3-oxopropyl)-dG aldehydes 3a and 3b were modeled in 5′-d(GCTAGCXAGTCC)-3′•5′-d(GGACTCGCTAGC)-3′ and compared to the corresponding acrolein-derived N²-(3-oxopropyl)-dG aldehyde, lacking the stereocenter at C_α of the N²-(3-oxopropyl)-dG (Figure 10). The model structures were subjected to potential energy minimization using the conjugate gradients algorithm in AMBER 8.0. The potential energy minimization predicted that both N²-(3-oxopropyl)-dG aldehydes 3a and 3b maintained Watson-Crick hydrogen bonding at both base pairs C⁶•Y¹⁹ and X⁷•C¹⁸ involved in the interstrand 5′-CpG-3′ crosslinks. The modeling studies suggested that for the R-stereoisomer of N²-(3-oxopropyl)-dG aldehyde 3a, the C_α methyl group oriented within the minor groove in the 3′-direction from the adducted nucleotide X⁷. This oriented the reactive aldehyde group in the 5′-direction, placing it proximate to the crosslinking target N²-dG in base pair C⁶•Y¹⁹. The favored orientation of the corresponding acrolein-derived N²-(3-oxopropyl)-dG aldehyde was similar, placing the aldehyde group in the 5′-direction, proximate to the crosslinking target N²-dG in base pair C⁶•Y¹⁹. In contrast, the modeling studies suggested that the for the S-stereoisomer of N²-(3-oxopropyl)-dG aldehyde 3b, the C_α methyl group oriented within the minor groove in the 5′-direction from the adducted nucleotide X⁷. This oriented the aldehyde group in the 3′-direction, placing it distal to the crosslinking target N²-dG in base pair C⁶•Y¹⁹.

Figure 10.

Molecular modeling of aldehydes 3a and 3b formed in duplex DNA when adducts 2a and 2b are placed opposite dC in the complementary strand. A. Adduct 3b arising from S-α-CH₃-γ-OH-PdG adduct 2b, illustrating the 5′-minor groove orientation of the α -carbon methyl group. B. The N²-(3-oxo-propyl)-dG aldehyde formed by the acrolein-derived γ-OH-PdG adduct in duplex DNA. C. Adduct 3a arising from R-α-CH₃-γ-OH-PdG adduct 2a, illustrating the 3′-minor groove orientation of the α-carbon methyl group.

The two γ-hydroxyl diastereomers of the 5′-CpG-3′ carbinolamine crosslink 5a arising from R-α-CH₃-γ-OH-PdG adduct 2a were modeled and compared to the corresponding unmodified oligodeoxynucleotide sequence. The model structures were subjected to potential energy minimization using the conjugate gradients algorithm in AMBER 8.0 (Figure 11). The potential energy minimization predicted that both diastereomers of carbinolamine crosslink maintained Watson-Crick hydrogen bonding at both of the tandem C•G base pairs involved in the interstrand crosslinks. The modeling studies suggested that the sp³ hybridization at the γ-carbon allowed the crosslinks to form without substantial perturbation of duplex structure. For the S-diastereomer of the carbinolamine crosslink, the molecular modeling predicted an additional hydrogen bond between the carbinolamine hydroxyl and N3-dG of the 5′ C•G base pair of the crosslink. In contrast, the imine in crosslink 6a mandated sp² hybridization of the amino group, which would require breaking the Watson-Crick hydrogen bond between the amine proton of N²-dG and O²-dC of the 5′ C•G base pair in the crosslink. Formation of either diastereomer of pyrimidopurinone crosslink 7a prevented Watson-Crick hydrogen bonding at the 3′ G•C base pair of the crosslink. It also disrupted Watson-Crick hydrogen bonding at the 5′ C•G base pair of the crosslink. The parameterization of the carbinolamine and the pyrimidopurinone crosslinks, for the AMBER 8.0 forcefield, is provided in Figure S1 of the Supporting Information.

Figure 11.

Molecular modeling of diastereomeric carbinolamine and pyrimidopurinone crosslinks formed by aldehyde 3a, arising from R-α-CH₃-γ-OH-PdG adduct 2a. A. The R-diastereomer at the α-carbon of the carbinolamine crosslink. B. The R-diastereomer at the α-carbon of the pyrimidopurinone crosslink. C. The S-diasteromer at the α-carbon of the carbinolamine crosslink. D. The S-diastereomer at the α-carbon of the pyrimidopurinone crosslink.

Discussion

Epimerization of the Stereoisomeric R- and S-α-CH₃-γ-OH-PdG Adducts in the 5′-CpG-3′ Sequence

At equilibrium in single-stranded DNA, the R- and S-α-CH₃-γ-OH-PdG adducts 2a and 2b existed as mixtures of epimers of the 1,N²-dG adduct at C_γ, in slow exchange on the NMR time scale. In both instances, the trans orientation of the α-CH₃ and γ-OH groups predominated. This observation was consistent with previous observations in the NMR spectra of the nucleosides (45, 46). This differed from the acrolein-derived γ-OH-PdG adduct lacking the CH₃ group at C_α, which exhibited equal amounts of both epimers.

The failure to observe a γ-¹³C resonance corresponding to ring-opened aldehydes 3a or 3b in single-stranded DNA was consistent with the observation that at pH 7, cyclic adducts 2a or 2b were favored as compared to ring-opened aldehydes 3a or 3b. The NMR data suggested that in single-stranded DNA, adducts 2a and 2b spontaneously epimerized, but slowly on the NMR time scale, without accumulation of aldehydes 3a or 3b. Nevertheless, the peptide trapping data revealed the transient presence of the aldehydes 3a or 3b (Figure 2).

Ring Opening of the R- and S-α-CH₃-γ-OH-PdG Adducts in Duplex DNA in the 5′-CpG-3′ Sequence

For both the R- and S-α-CH₃-γ-OH-PdG adducts the presence of duplex DNA was required for the stable formation of the aldehydes 3a and 3b, as observed for the acrolein-derived γ-HOPdG adduct (21). In duplex DNA, the aldehydes 3a or 3b, and their hydrates 4a or 4b, similar to the acrolein γ-OH-PdG adduct (21) and the malondialdehyde M₁dG adduct (47, 48), are accommodated in the minor groove of DNA, enabling maintenance of Watson-Crick hydrogen bonding at the adducted base pair. Significantly, in comparing the chemistry of the R- and S-α-CH₃-γ-OH-PdG adducts 2a and 2b when placed into duplex DNA opposite dC at pH 7 with the corresponding acrolein-derived γ-OH-PdG adduct (21), ring-opening of the R- and S-α-CH₃-γ-OH-PdG adducts 2a and 2b to the aldehydes 3a and 3b and the hydrated aldehydes 4a and 4b was incomplete. When the acrolein γ-OH-PdG adduct was placed opposite dC in duplex DNA, equilibrium favored the ring-opened N²-(3-oxopropyl)-dG aldehyde and its hydrate, to the extent that the 1,N²-dG adduct was no longer detected (21). The incomplete opening of the R- and S-α-CH₃-γ-OH-PdG adducts 2a and 2b in duplex DNA, when placed opposite dC, might be explained by the fact that cyclic adducts 2a and 2b position the CH₃ groups to avoid steric clash with N3 of the guanine. This becomes an issue upon formation of the N²-(3-oxopropyl)-dG aldehydes 3a and 3b. The stabilization of cyclic adducts 2a and 2b might also arise from the Thorpe-Ingold effect. For the R- and S-α-CH₃-γ-OH-PdG adducts 2a and 2b in duplex DNA, the degree to which the 1,N²-dG adducts opened to the N²-(3-oxopropyl)-dG aldehydes 3a and 3b increased significantly at pH 9.3 (Figure 7). In duplex DNA at pH 7, crosslinks formed between the N-terminal peptide amine and the aldehyde rearrangement products 3a and 3b of the R- and S-α-CH₃-γ-OH-PdG adducts at 4 °C (26).

One factor with regard to the ring opening of adduct 2b to aldehyde 3b in duplex DNA was the identity of the nucleotide opposite the 1,N²-dG adduct. When placed opposite to T in duplex DNA, the S-α-CH₃-γ-OH-PdG adduct did not undergo ring-opening to aldehyde 3b (Figure 8), whereas the acrolein-derived γ-OH-PdG adduct when mispaired with T, existed as a mixture of the 1,N²-dG adduct and the ring-opened aldehyde (25). When mispaired with T, the M₁dG adduct remained as the 1,N²-dG adduct (47). Similar to the acrolein-derived γ-OH-PdG adduct, the ratio of aldehyde to hydrated aldehyde (3a:4a or 3b:4b) for the crotonaldehyde-derived adducts increased with temperature.

Role of Stereochemistry in Interstrand Crosslinking in the 5′-CpG-3′ Sequence Context

The presence of aldehydes 3a or 3b in the minor groove at pH 7 and 37 °C was significant with regard to their potential for forming interstrand crosslinks 5a or 5b under physiological conditions in the 5′-CpG-3′ sequence. The present NMR studies corroborate the stereospecific preference for DNA interstrand crosslinking by the R-α-CH₃-γ-OH-PdG adduct, as opposed to the S-α-CH₃-γ-OH-PdG adduct (23). The time required for crosslink 5a to reach equilibrium at pH 7 and 37 °C was approximately 20 days, with approximately 26% crosslinking observed.

To examine why interstrand crosslinking was much more extensive for the R-α-CH₃-γ-¹³C-OH-PdG adduct 2a than for the S-α-CH₃-γ-¹³C-OH-PdG adduct 2b in the 5′-CpG-3′ sequence context, a molecular modeling approach was employed (Figure 9). The modeling studies suggested that the low levels of crosslinking observed for the S-α-CH₃-γ-¹³C-OH-PdG adduct 2b in the 5′-CpG-3′ sequence might be attributed to the fact that the carbinolamine crosslink 5b was of lower stability than that of crosslink 5a, presumably due to the differential orientation of the CH₃ group at the α-carbon of the crosslink. Anecdotally, Lao and Hecht reported conducting molecular dynamics studies of pyrimidopurinone crosslinks 7a and 7b, formed from adducts 2a and 2b, respectively, and reaching a similar conclusion, i.e., that the pyrimidopurinone crosslink arising from the R-α-CH₃-γ-¹³C-OH-PdG adduct 2a was of greater stability, due to a more favorable orientation of the α-carbon methyl group within the minor groove (10). The modeling studies also suggested that aldehyde 3b, arising from adduct 2b in duplex DNA, would not be oriented favorably for reaction with the crosslinking target N²-dG in base pair C⁶•Y¹⁹; thus the rate of formation of the crosslink arising from aldehyde 3b would be expected to be slower (Figure 10). NMR studies designed to examine these structural hypotheses are in progress.

Formation of an Interstrand Carbinolamine Crosslink by the R-α-CH₃-γ-OH-PdG Adduct

At equilibrium in duplex DNA, the interstrand R-α-CH₃-γ-OH-PdG crosslink is comprised of a mixture of carbinolamine 5a, imine 6a, and pyrimidopurinone 7a. The presence of some of imine 6a was inferred since the crosslink was reductively trapped in the presence of NaCNBH₃ (23). Lao and Hecht, using negative ion mode ESI-Q-TOF-MS analysis of a crosslinked oligodeoxynucleotide, observed m/z values corresponding to carbinolamine 5a and either imine 6a or pyrimidopurinone 7a, with the lower m/z signal corresponding to imine 6a or pyrimidopurinone 7a predominating (10). The present NMR studies suggested that the predominant form of the R-α-CH₃-γ-OH-PdG crosslink in situ, is not imine 6a. The amount of imine 6a remained below the level of spectroscopic detection in duplex DNA. The reduction of crosslinked imine 6a was slow in the presence of NaCNBH₃ (23), consistent with the notion that conversion of carbinolamine 5a to the reducible imine 6a was rate limiting in reductively trapping the crosslink in duplex DNA.

Similar to the acrolein-derived γ-OH-PdG interstrand crosslink (25), molecular modeling revealed that the carbinolamine linkage of crosslink 5a maintained Watson-Crick hydrogen bonding at both of the tandem C•G base pairs (Figure 11). In contrast, dehydration of the carbinolamine crosslink to imine (Schiff base) crosslink 6a, or cyclization of the latter to form pyrimidopurinone crosslink 7a, required disruption of Watson-Crick hydrogen bonding at one or both of the tandem crosslinked C•G base pairs. The NMR studies supported this conclusion, suggesting intact Watson-Crick base pairing at the crosslinked X⁷•C¹⁸ (Figure 5). In contrast, enzymatic digestion of duplex DNA containing crosslink 5a afforded a bis-deoxyguanosine conjugate, characterized by a combination of mass spectrometry and NMR as pyrimidopurinone 7a arising from annelation of imine 6a with N1-dG in the 5′-CpG-3′ sequence (23). The likely explanation is that the equilibrium between carbinolamine 5a, imine 6a, and pyrimidopurinone 7a depends upon the conformational state of the DNA. Enzymatic degradation of duplex DNA favors collapse of the carbinolamine crosslink 5a to the pyrimidopurinone bis-nucleoside crosslink 7a.

Structure-Biological Activity Relationships

Site specific mutagenesis in COS-7 mammalian cells using the single-stranded pMS2 shuttle vector (49) indicated that both the R- and S-α-CH₃-γ-OH-PdG adducts yielded mutations at a 5–6% frequency. These were predominantly G→T transversions, corroborating experiments utilizing a randomly modified shuttle vector and replicated in human cells (50). In the same mammalian site-specific mutagenesis system, the acrolein-derived γ-OH-PdG adduct showed a greater frequency of mutations also predominantly G→T mutations (51). The propensity of cyclic adducts 2a or 2b to undergo ring opening to the N²-(3-oxopropyl)-dG aldehydes 3a or 3b may facilitate lesion bypass, as reported for the acrolein-derived γ-OH-PdG adduct (52–56). On the other hand, in duplex DNA, incomplete conversion of crotonaldehyde-derived adducts 2a and 2b to the aldehydes 3a and 3b or hydrated aldehydes 4a and 4b may result in a more efficient block to DNA replication, possibly reducing their mutagenicity. The R- and S-α-CH₃-γ-OH-PdG adducts were reported to block trans-lesion synthesis by the Klenow (exo-) fragment of DNA polymerase I and DNA polymerase ε (57). The enzymes responsible for trans-lesion synthesis of the R- and S-α-CH₃-γ-OH-PdG adducts remain to be identified. However, Washington et al. (58, 59) showed that the Y-polymerase pol ι in conjunction with pol κ or Rev 1 in combination with pol ζ could efficiently bypass the acrolein-derived γ-OH-PdG adduct. Minko et al. showed that pol η could bypass the γ-OH-PdG to a lesser extent (60). It seems plausible that these error-prone polymerases might also bypass the R- and S-α-CH₃-γ-OH-PdG adducts.

Summary

The chemistry of the R- and S-α-CH₃-γ-OH-PdG adducts in DNA was similar to that of the acrolein-derived γ-OH-PdG adducts, but the positions of chemical equilibria differed. At pH 7 in single-stranded DNA only the 1,N²-dG adducts 2a and 2b were observed. The epimers having trans configuration of the α-CH₃ and γ-OH groups were favored. The transient presence of ring-opened aldehydes 3a and 3b was established by peptide trapping experiments. When placed opposite dC in duplex DNA at pH 7, the R- and S-α-CH₃-γ-OH-PdG adducts 2a and 2b did not completely convert to ring-opened aldehydes. When the S-α-CH₃-γ-OH-PdG adduct 2b was placed opposite T, only the 1, N²-dG adduct was observed. Interstrand crosslinking in the 5′-CpG-3′ sequence was observed spectroscopically only for the R-α-CH₃-γ-OH-PdG adduct, corroborating results of reductive trapping experiments (23). Molecular modeling predicted that carbinolamine 5a arising from adduct 2a should be favored as compared to carbinolamine 5b arising from adduct 2b. It also predicted that the aldehyde rearrangement product 3a of the R-α-CH₃-γ-OH-PdG adduct 2a was oriented in the minor groove with the aldehyde moiety in the 5′-direction, potentially facilitating crosslinking with N²-dG in the complementary strand of the 5′-CpX-3′ sequence, whereas for the S-α-CH₃-γ-OH-PdG adduct 2b, the aldehyde moiety oriented in the minor groove in the 3′-direction, hindering crosslinking. Similar to the γ-OH-PdG crosslink, the interstrand crosslink formed in the 5′-CpG-3′ sequence existed predominantly as the carbinolamine 5a.

Supplementary Material

si20051026_070. Supporting Information Available.

The Supporting Information contains Figure S1, showing the parameterization of the carbinolamine and the pyrimidopurinone crosslinks for the AMBER 8.0 forcefield. This material is available free of charge via the Internet at http://pubs.acs.org.

Figure 4.

¹³C NMR spectrum of R-α-CH₃-γ-OH-PdG in the oligodeoxynucleotide 5′-d(GCTAGCXAGTCC)-3′•5′-d(GGACTCYCTAGC)-3′. Resonances a, aldehyde 3a; b, hydrated aldehyde 4a; c, carbinolamine crosslink 5a; d, cyclic adduct 2a.