Deoxyguanosine Forms a Bis-adduct with E,E-Muconaldehyde, an Oxidative Metabolite of Benzene. Implications for the Carcinogenicity of Benzene

Abstract

Benzene is employed in large quantities in the chemical industry and is a ubiquitous contaminant in the environment. There is strong epidemiological evidence that benzene exposure induces hematopoietic malignancies, especially acute myeloid leukemia, in humans but the chemical mechanisms remain obscure. E,E-Muconaldehyde is one of the products of metabolic oxidation of benzene. This paper explores the proposition that E,E-muconaldehyde is capable of forming Gua-Gua cross-links. If formed in DNA, the replication and repair of such cross-links might introduce structural defects that could be the origin of the carcinogenicity. We have investigated the reaction of E,E-muconaldehyde with dGuo and found the reaction yields two pairs of interconverting diastereomers of a novel heptacyclic bis-adduct having a spiro ring system linking the two Gua residues. The structures of the four diastereomers have been established by NMR spectroscopy and their absolute configurations by comparison of CD spectra with those of model compounds having known configurations. The final two steps in formation of the bis-nucleoside (5-ring → 6-ring → 7-ring) have significant reversibility, which is the basis for the observed epimerization. The 6-ring precursor was trapped from the equilibrating mixture by reduction with NaBH₄. The anti relationship of the two Gua residues in the heptacyclic bis-adduct precludes it from being formed in B DNA but the 6-ring precursor could readily be accommodated as an interchain or intrachain cross-link. It should be possible to form similar cross-links of dCyt, dAdo, the ε-amino group of lysine, and N-termini of peptides with the dGuo-muconaldehyde monoadduct.

1. INTRODUCTION

Benzene is widely dispersed in the environment. It is a component of petroleum and is used extensively in the chemical industry as a solvent and as a building block in the synthesis of more complex organic compounds. It is also a product of incomplete combustion of fossil fuels and organic materials including tobacco. Benzene is a human carcinogen for which epidemiological studies have shown strong correlation of benzene exposure with induction of hematopoietic malignancies, especially acute myeloid leukemia.^1–4 No rodent model exists, although feeding studies with rodents have demonstrated induction of other types of tumors. On the basis of these observations, OSHA has restricted workplace exposure to levels less than 1 ppm. Nevertheless, there is evidence that even the 1-ppm threshold may be too high.^5–7

The mechanistic basis for the genotoxicity of benzene has been investigated extensively but remains elusive.^1,8,9 Many mechanisms have been proposed, all initiated by P450 oxidation to form benzene oxide. P450 2E1 has been shown by Valentine et al., via the use of transgenic P450 2E1 knockout mice, to be the major species involved in the metabolism of benzene.¹⁰ Both metabolism and toxicity were greatly reduced in the knockout mouse versus the wild-type. Residual activity in the knockout mouse was attributed to P450 2B1 for which benzene is not a particularly good substrate. Benzene oxide can undergo a variety of transformations including non-enzymatic rearrangement to phenol (NIH shift) and equilibration with oxepin, its valence tautomer. Phenol upon further oxidation yields hydroquinone, catechol, and their quinones. Redox cycling of the hydroquinone-benzoquinone couple will produce hydroxyl radicals, which can cause free radical and oxidative damage to DNA. Benzoquinone can also form adducts with DNA¹¹ and with topoisomerase II.¹² All of these processes may contribute to the carcinogenicity of benzene although animal and epidemiological studies provide no support for phenol being a carcinogen. Furthermore, blood and tissues of animals not exposed to benzene contain significant amounts of hydroquinone derived from dietary sources.¹³ Studies on protein adducts of quinone metabolites in the blood and bone marrow of rodents exposed to low doses of ¹⁴C/¹³C-benzene showed that labeled adduct levels were insignificant in comparison with those arising from dietary sources and thus unlikely to increase the risk of leukemia.^14,15 It should be noted in passing that hydroquinone and other hydroxylated benzenes that arise metabolically from phenol are structurally similar to the flavonoids in fruits and vegetables that are generally regarded as being beneficial antioxidants rather than carcinogens. In summary, although many workers have concluded that hydroquinone, catechol and/or their quinones are the primary initiators of acute myeloid leukemia,¹ there are significant problems with this scenario. In particular, why should benzene be genotoxic at low levels when background levels of the phenolic metabolites are high?

The possibility exists that some metabolite unique to benzene, perhaps formed in only small amounts, plays the critical role in the genotoxicity of benzene. Witz and Goldstein have suggested that muconaldehyde is the causative factor.^16–19 Muconaldehyde is readily formed from benzene oxide/oxepin by oxidation with a variety of mild oxidants as first shown by Davies and Whitham in 1977.²⁰ It has been proposed that the in vivo oxidizing agent is a cytochrome monooxygenase although the precise one has not been identified.²¹ Oxidation of oxepin gives the 2,3-epoxide which spontaneously isomerizes to Z,Z- and then E,Z-muconaldehyde (Scheme 1). In a reaction catalyzed by amines or glutathione,²² E,Z-muconaldehyde ultimately isomerizes to the E,E form.²² Whereas there are many natural sources of hydroxylated phenols and their quinones, metabolism of benzene is the only known source of E,E-muconaldehyde. E,E-muconaldehyde has been detected in liver microsomes incubated with ¹⁴C-benzene.²³

Scheme 1.

Metabolic activation of benzene.

E,E-Muconic acid was first shown to be a urinary metabolite of benzene by Parke and Williams in 1953.²⁴ Subsequently it has been found as a metabolite formed from E,E-muconaldehyde ²⁵ but not from phenol, catechol or hydroquinone.²⁶ The E,E isomer is the only one to be detected in biological systems^7,24 and has been used as a biomarker for benzene exposure, but a caveat is that it is not uniquely derived from benzene, being formed as well by enzymatic oxidation of sorbic acid, which is a common food preservative.²⁷

It has been argued that the muconaldehydes are too reactive to be able to make their way from the liver to the bone marrow, which is the site of malignancies in acute myeloid leukemia.³ Benzene oxide on the other hand is sufficiently stable that it has been directly identified in the blood of rats administered benzene.²⁸ The lifetime of benzene oxide would appear to be sufficient to permit in vivo transport from the liver to other organs via the bloodstream.²⁹ Turteltaub and Mani have postulated that a cytochrome P450 in bone marrow could perform the epoxidation of oxepin; analyses of marrow, liver and blood of rodents exposed to low levels of ¹³C₆-benzene revealed a low capacity, high affinity process in the marrow leading to levels of muconaldehydes (analyzed as E,E-muconic acid) in excess of hydroquinone.³⁰ Henderson et al. concluded from rodent studies of oral and inhalation exposure that high doses favor a low affinity, high capacity detoxification pathway, whereas low dose administration leads to toxic metabolites.^31,32 This observation has been confirmed by other groups.^25,33–37

The high cytotoxicity of E,E-muconaldehyde complicates evaluation of its mutagenicity and genotoxicity. It is weakly mutagenic in bacterial systems but more mutagenic in V79 Chinese hamster lung fibroblast cells.^38,39 The primary genetic damage associated with leukemia in benzene-exposed people and various tumors in benzene-exposed rodents involves large-scale damage, including micronuclei, chromosomal rearrangements and aneuploidy.^3,40,41 Few simple mutations are observed. One possible explanation for this is that the relevant genetic damage involves cross-links. The muconaldehydes, by virtue of their multiple electrophilic sites, should be capable of forming DNA-DNA and DNA-protein cross-links. Both types of cross-links could be the origin of structural alterations observed in the genomes of malignant cells. Indeed, Nakayama et al. in a study in human cells using supF shuttle vectors treated with hydroquinone or muconaldehyde found the E,E-muconaldehyde mutational spectrum was unique in its high proportion of tandem mutations, particularly at GG sites.⁴² Witz and coworkers have found evidence for E,E-muconaldehyde-mediated cross-links between DNA and proteins. They speculated that these cross-links may be the basis of the genotoxicity of benzene.^19,43,44 Such cross-links might be involved in the inhibition of human topoisomerase II by E,E-muconaldehyde reported by Frantz et al.⁴⁵

Examination of the literature reveals little information about the reaction of E,E-muconaldehyde with the constituents of nucleic acids or proteins. Golding and coworkers have studied the chemistry of muconaldehyde with nucleophiles. Their work has focused mainly on the Z,Z isomer.^21,46 A key finding was that reaction of the Z,Z isomer with Gua, Ade and their nucleosides under a variety of conditions including 0.1 M pH 7.0 buffer gave pyrrole-type monoadducts; Z,Z-muconaldehyde adduction of Guo is shown in Scheme 2. These authors did not report any studies of the reaction of E,E-muconaldehyde with nucleosides; they did observe, however, that E,E-muconaldehyde yielded a bis-imine upon reaction with 1-propylamine⁴⁶ in contrast to the pyrrole obtained with Z,Z-muconaldehyde. The trans double bonds in E,E-muconaldehyde promote formation of bis-adducts by protecting against formation of pyrroletype monoadducts. Golding and his colleagues have also studied the reaction of the muconaldehydes with glutathione.²² The E,E isomer reacts with GSH and other thiols to give both mono- and bis-adducts with the latter forming more slowly. The authors propose that the reversibility of GSH addition to the muconaldehydes could allow GSH to act as a transporter of E,E-muconaldehyde in biological systems. Latriano et al. demonstrated that the reaction of E,E-muconaldehyde with ¹⁴C-labeled dGuo-5´-phosphate in phosphate buffer, pH 7.4 yielded several radioactive products of unknown structure.¹⁷

Scheme 2.

Adducts of Z,Z- and E,E-muconaldehyde characterized by Golding’s group. ^21,46

In recent years we and others have carried out extensive studies of the reactions of simple enals, i.e., acrolein, crotonaldehyde and 4-hydroxynonenal, with nucleosides and DNA.^47,48 These enals are bis-electrophiles, being able to form Michael adducts by 1,4-addition and carbinolamines and/or imines by 1,2-addition of nucleophiles. They undergo two-fold reactions with DNA to form cyclic adducts of the bases plus intra- and interchain cross-links. They also form cross-links of DNA with peptides and proteins. The cross-links are less damaging than one might initially expect because they exist primarily as carbinolamines that readily dissociate.

The muconaldehydes have the potential to form bis-1,4-adducts with amines. Such bis-adducts are known to be less reversible^49–51 and, as a consequence, could be expected to be much more damaging to the genome. For this reason we decided to follow up on Golding’s studies and examine the reactions of nucleosides and nucleic acids with muconaldehydes, concentrating on the E,E isomer. Our hypothesis of bis-1,4-adducts was, in fact, naïve but in the course of these studies a more complex reaction process was discovered that leads to stable bis-adducts. This paper describes the reaction of dGuo with E,E-muconaldehyde to form novel heptacyclic bisnucleosides 1 (Figure 1). It is noteworthy that the condensation produces 4 new stereogenic centers and could potentially yield 16 diastereomers but structural and mechanistic constraints limited the number to four.

Figure 1.

Numbering and ring designations for the bis-adduct of dGuo with E,E-muconaldehyde.

The preferred IUPAC name and numbering system for the bis-base is 6,8,10,12,15,17,19,21,23,26-decaazaheptacyclo[14.10.0.0^1,5.0^7,15.0^9,13.0^18,26.0^20,24]hexacosa-7,9(13),11,18,20(24),22-hexaene-14,25-dione. In the present paper, rather than using this cumbersome numbering system, purine numbering will be used for the eastern (E) and western (W) Gua moieties and the E,E-muconaldehyde-derived fragment (M) will be numbered as shown in Figure 1. Subsequently in the paper E,E-muconaldehyde will simply be called muconaldehyde.

2. EXPERIMENTAL SECTION

2.1. Materials

Muconaldehyde was prepared from cis-1,2-dihydrocatechol (Aldrich) by the method of Golding et al.⁵² and stored at −20 °C. 2'-Deoxyguanosine (Aldrich) was obtained and used as the monohydrate.

2.2. Chromatography

Initial product isolation was carried out by solid phase extraction (SPE) on reverse phase cartridges (Strata-X, 33 µm, Phenomenex) with elution by water-methanol mixtures. HPLC experiments were performed on a Beckman Coulter system comprising a System Gold 126 solvent module and System Gold 168 Diode Array Detector; Beckman Coulter 32 Karat v.7.0 software was used for the acquisition, processing, and analysis of HPLC data. The diode array detector was configured to simultaneously monitor 254 and 360 nm wavelengths. Analysis and purification were performed using YMC ODS-AQ C-18 columns (250 mm × 4.6 mm i.d. for analysis and 250 mm × 10 mm i.d. for purification). The mobile phase consisted of water and either acetonitrile or methanol. Methods are described in the Supporting Information.

2.3. NMR and CD Spectra

NMR spectra were recorded in anhydrous DMSO-d₆ and/or DMSO-d₆/D₂O (~90:10, v/v) at 25 °C and referenced to residual DMSO-d₅ using 3 mm tubes on a 600 MHz Bruker spectrometer equipped with a cryoprobe. Small differences in chemical shifts of the non-exchangeable protons were observed between the two solvents. ¹³C chemical shift assignments (150 MHz) were derived from HSQC and HMBC spectra. Tabulations of ¹H and ¹³C chemical shifts of 1a–d, 2ac and 2bd, and 4a–d are in the Supporting Information. NOE measurements were made with the Bruker selective NOE program SELNOGP using solutions in DMSO-d₆ containing 8% D₂O at 25 °C. Under these conditions, the observed NOEs were negative, i.e., they had the same phase as the signal at the irradiation point.

CD spectra were obtained on a JASCO 720 spectropolarimeter at ambient temperature. Samples were dissolved in H₂O-methanol (1:1) to give an absorbance at 260 nm of 0.50–0.55. CD spectra were normalized to absorbance at 260 nm rather than molar absorbance because the absorbance of the compounds does not follow Beer’s Law; the molar extinction coefficient decreases with concentration indicating aggregation or stacking in the solvent mixture employed.

2.4. Reactions of Muconaldehyde with dGuo

A variety of reaction conditions were explored (see Results and Discussion) almost all of which led to the same mixture of products in yields ranging from 10–40%; the highest yields were obtained by using a 10:1 molar ratio of dGuo to muconaldehyde. Sonication of freshly mixed reaction mixtures for 10 min at room temperature also improved yields. A typical analytical-scale reaction was carried out with a mixture of 10 µmoles of dGuo and 5 µmoles of muconaldehyde in 2 mL of 50 mM phosphate buffer, pH 7.4, at room temperature or 37 °C. The course of the reaction was followed by HPLC (Method 1). Similarly, a typical preparative reaction employed 50 mg (0.175 mmol) of dGuo and 4.82 mg (0.044 mmol) of muconaldehyde in 30 mL of 50 mM phosphate buffer, pH 7.4. The reaction was sonicated for 10 minutes and allowed to stand at room temperature in the dark for 2 months. One-half of the reaction was further purified and the remainder frozen for future use. Crude purification was accomplished by applying 7.5 mL of the reaction mixture to each of two SPE cartridges (200 mg, prepared and eluted according to the manufacturer’s directions). Elution was carried out with methanol-water mixtures. dGuo eluted in the 0% and 5% methanol fractions. Compounds 1a–d eluted in the 15–25% methanol fractions. The fractions were analyzed by UV (260 nm) and HPLC (Method 2). The fractions containing 1a–d were combined and lyophilized after removal of most of the methanol by vacuum centrifugation to give 4.67 mg (34%) of crude 1a–d.

Further purification was performed by preparative HPLC (Method 2). dG-muc-dG isomers 1a and 1b could be obtained in >90% purity by repeated chromatography under the same conditions but isomers 1c and 1d could not be completely separated. Purified material for UV and mass spectroscopy of 1c was obtained by collecting only the leading edge of the peak; purified 1d was obtained from HPLC separation of an equilibrated sample of 1a (see below). 1a: Retention time 11.95 min; UV (H₂O-CH₃OH) λ_max 256 nm, 285 nm (sh); HRMS (ES+) calcd for C₂₆H₃₀N₁₀O₉ (M + Na⁺) 649.2095; found 649.2098. 1b: Retention time 12.40 min; UV (H₂O-CH₃OH) λ_max 256 nm, 285 nm (sh); HRMS (ES+) calcd for C₂₆H₃₀N₁₀O₉ (M + Na⁺) 649.2095; found 649.2072. 1c: Retention time 12.78 min; UV (H₂O-CH₃OH) λ_max 256 nm, 285 nm (sh); HRMS (ES+) calcd for C₂₆H₃₀N₁₀O₉ (M + Na⁺) 649.2095; found 649.2076. 1d: Retention time 13.02 min; UV (H₂O-CH₃OH) λ_max 256 nm, 285 nm (sh); HRMS (ES+) calcd for C₂₆H₃₀N₁₀O₉ (M + Na⁺) 649.2095; found 649.2100.

2.5. Depurination of dG-muc-dG Diastereomers 1a–d

Depurinations were performed by heating lyophilized samples of 1a–d in 0.1 M HCl at 70 °C for 1 hr followed by neutralization with 0.1 M NaOH. Analytical Scale: Reactions were followed by HPLC (Method 3). Preparative Scale: A crude mixture of dG-muc-dG diastereomers (59 A₂₆₀ units) isolated with an SPE cartridge was depurinated in 1.0 mL of 0.1 M HCl as above and purified by preparative HPLC (Method 4) to give bis–bases 2ac and 2bd. 2ac: Retention time 17.00 min; UV (H₂O-CH₃OH) λ_max 252 nm, 280 (sh); HRMS (ES+) calcd for C₁₆H₁₅N₁₀O₃ (M + H⁺) 395.1329; found 395.1331. 2bd: Retention time 18.10 min; UV (H₂O-CH₃OH) λ_max 252 nm, 280 (sh); HRMS (ES+) calcd for C₁₆H₁₅N₁₀O₃ (M + H⁺) 395.1329; found 395.1336.

2.6. Reduction of G-muc-G Diastereomers 2ac and 2bd

To a solution of 2ac in 0.5 mL of water-methanol (1:1) was added excess solid sodium borohydride. After standing for 12 h at ambient temperature, the reaction was neutralized with 5% acetic acid and an aliquot was analyzed by HPLC (Method 6); a single peak at 12.43 min was observed for 3ac. G-muc-G 2bd was treated similarly to give 3bd as a single peak at 13.35 min. The remainder of each sample was purified by preparative HPLC (Method 4). 3ac: UV (H₂O-CH₃OH) λ_max 246 nm, 275 (sh). HRMS (ES+): calcd for C₁₆H₁₇N₁₀O₃ (M + H⁺) 397.1479; found 397.1488. 3bd: UV (H₂O-CH₃OH) λ_max 243 nm, 278 (sh). HRMS (ES+): calcd for C₁₆H₁₅N₁₀O₃ (M + H⁺) 397.1479; found 397.1486.

2.7. Reduction of a crude mixture of dG-muc-dG isomers

Excess solid sodium borohydride (0.77 mg, 0.020 mmol) was added to a solution of crude dG-muc-dGs 1a–d (SPE purified, 59 A₂₆₀ units) in 3.0 mL of water-methanol (1:1). The mixture was stirred at ambient temperature for 10 h with monitoring by HPLC (Method 5), neutralized with 0.2 M HCl and purified by preparative HPLC (Method 2) to give four diastereomeric products: 4c: Retention time 10.50 min; UV (H₂O-CH₃OH) λ_max 252 nm, 274 nm (sh); HRMS (ES+) calcd for C₂₆H₃₃N₁₀O₉ (M + H⁺) 629.2426; found 629.2435. 4d: Retention time 11.17 min; UV (H₂O-CH₃OH) λ_max 252 nm, 275 nm (sh); HRMS (ES+) calcd for C₂₆H₃₃N₁₀O₉ (M + H⁺) 629.2426; found 629.2435. 4a: Retention time 12.20 min; UV (H₂O-CH₃OH) λ_max 252 nm, 276 nm (sh); HRMS (ES+) calcd for C₂₆H₃₃N₁₀O₉ (M + H⁺) 629.2426; found 629.2434. 4b: Retention time 13.32 min; UV (H₂O-CH₃OH) λ_max 252 nm, 274 nm (sh); HRMS (ES+) calcd for C₂₆H₃₃N₁₀O₉ (M + H⁺) 629.2426; found 629.2427. Individual isomers 1a–d were reduced on a small scale to confirm the correlation between reduced and unreduced species.

2.8. Effects of pH on the Reaction of dGuo with Muconaldehyde

A stock solution was prepared containing dGuo (10 mg, 0.035 mmol) and muconaldehyde (1.91 mg, 0.017 mmol) in 6 mL of water. The solution was distributed equally among 3 vials and lyophilized. To each vial was added 2 mL of 50 mM potassium phosphate buffer at a different pH values: 6.5, 7.5, and 8.5. The mixtures were stirred at ambient temperature for several days. At intervals, 100 µL aliquots were removed, filtered and analyzed by HPLC (Method 7).

2.9. Equilibration of dG-muc-dG Diastereomers 1a and 1b

Aliquots were removed from solutions of 1a and 1b that had been standing in DMSO-d₆/D₂O (~90:10, v/v) at ambient temperature for 5 months. The samples were lyophilized and then redissolved in 1.0 mL of potassium phosphate buffer (pH 7.4, 50 mM) and examined by HPLC (Method 8). Aliquots of the aqueous solutions were removed and analyzed periodically over 4 months at which point equilibration appeared essentially complete.

3. RESULTS

3.1. Reaction of dGuo with Muconaldehyde

As a first approach we explored the reaction under aqueous conditions near neutral pH, this being a reasonable approximation of biological milieu. The initial reaction was carried out with equimolar dGuo and muconaldehyde in phosphate buffer (50 mM, pH 7.4) at room temperature. The reactions were monitored by HPLC. Within a few hours several small peaks appeared and additional peaks appeared over the course of several days, several of which had UV spectra similar to those reported by Latriano et al.¹⁷ ESI-MS of the early mixtures showed the presence of species containing dGuo and muconaldehyde in a 1:1 ratio with and without loss of water; 2:1 species from which 0, 1 and 2 water molecules had been lost were also observed. An interpretation of these results is presented in the Discussion section. In addition, colored, late-eluting species were observed for which the mass spectra were consistent with two muconaldehyde molecules having reacted with one dGuo; these were more prominent when excess muconaldehyde was present.

Our attention quickly turned to exploring reactions in which dGuo was in excess because the products were more likely to have biological relevance. In spite of the mass spectra of these product mixtures being somewhat promising, the complexity of the mixtures held poor prospects for our being able to obtain significant yields of any adducts. As these experiments were underway we reexamined our original reaction mixture after it had been stored at room temperature for 7 weeks and found that the appearance of the chromatogram was much improved. Muconaldehyde had been depleted although dGuo was still present. Many of the initially formed products were diminished or had disappeared entirely. Four main products (1a–d) had appeared, which had not been visible in the early chromatograms (Figure 2). They eluted in a group much later than dGuo and muconaldehyde. Products 1a and 1b were well resolved but 1c co-eluted with 1d on most reverse-phase HPLC columns. Peaks for 1a and 1c were of similar intensity and about half again larger than peaks 1b and 1d. Molar ratios of dGuo:muconaldehyde varying from 10:1 to 1:2 gave similar results with chromatograms obtained shortly after the start of the experiments showing many peaks, the intensity of which varied somewhat depending on the starting ratio of dGuo and muconaldehyde. Reactions carried out at 37 °C were essentially complete after 4 days; in all cases the final chromatograms were dominated by the four product peaks. With a 10:1 ratio of dGuo to muconaldehyde, the combined yield of the four products approached 40%. In the absence of dGuo E,E-muconaldehyde itself is quite stable under the reaction conditions (pH 7.4, room temperature or 37 °C) diminishing < 5% over 4 days; over several months at room temperature, however, it gradually disappears without the appearance of any new UV-absorbing peaks. We conclude that most of the chromatographic peaks arising within hours/days of the start of the reactions with dGuo represent transient species that ultimately give the final bis-nucleosides.

Figure 2.

HPLC chromatogram showing the reaction of dGuo with muconaldehyde in phosphate buffer (0.05 M, pH 7.4) after 7 weeks at ambient temperature.

The effect of pH was examined. Reactions were carried out at pH 6.5, 7.5 and 8.5 in 50 mM phosphate buffer at ambient temperature with a 2:1 molar ratio of dGuo to muconaldehyde. The rate of disappearance of muconaldehyde was pH dependent. At pH 8.5, muconaldehyde was depleted within 24 h and after 72 h compounds 1a–d were the dominant species, whereas they were barely detectable after 72 h at pH 6.5 (Figure 3). Eventually all of the reactions reached the mixture of the four products shown in Figure 2. Based on the HPLC analyses, the yield of 1a–d was marginally better at pH 7.5.

Figure 3.

Effect of pH on the reaction of dGuo with muconaldehyde at ambient temperature. Reaction mixtures contained dGuo (5.8 µM) and muconaldehyde (2.9 µM) in potassium phosphate buffer (0.05 M) at pH 6.5, 7.5 and 8.5. Chromatogram shows reaction mixtures after 72 h.

3.2. Isolation of Bis-nucleosides 1a–d

For preparation of sufficient material for structural identification, reactions were run in phosphate buffer, pH 7.4, with a dGuo:muconaldehyde ratio of 4:1. A rough separation of the product mixture was made on SPE cartridges using H₂O/methanol with the mixture of 1a–d eluting in the 15–25% methanol fraction in a combined yield of 30–40%. The LC-ESI mass spectrum (positive-ion) of the 1a–d mixture gave an [M+H]⁺ signal at m/z 627 followed by sequential neutral losses of two deoxyribose fragments (m/z 627 - 116 → 511 and 511 - 116 → 395). The negative ion spectrum gave an [M-H]⁻ signal at m/z 625.

The mixture was partitioned by preparative HPLC. Compounds 1a and 1b were readily isolated in pure form and smaller amounts of 1c could be obtained in pure form by sacrificing the trailing side of the peak. Useful quantities of 1d could not be isolated even by repeated chromatography but eventually small samples of 1d were obtained by equilibration of 1a (see below). The quantity of 1d that could practically be obtained in this manner was only sufficient for UV and MS. Exact mass determinations on the four species were in accord with the empirical formula C₂₆H₃₀N₁₀O₉ (MW 626), establishing that they were, in fact, isomers resulting from the reaction of muconaldehyde with two dGuo residues with loss of one H₂O. They had indistinguishable UV spectra, which were similar to that of dGuo.

During the course of NMR studies (see below), it was discovered that, on long standing at ambient temperature, NMR samples of 1a and 1b in DMSO-d₆/D₂O became contaminated with 1d and 1c, respectively. HPLC examination of samples that had been standing for 5 months revealed that 1a had become an 89:11 mixture of 1a and 1d and 1b had become a 67:33 mixture of 1b and 1c. Equilibration of the isomers was faster in aqueous buffer, pH 7.4. After several weeks at room temperature, 1a had become a 61:39 mixture of 1a and 1d; 1b had become a 47:53 mixture of 1b and 1c.

By careful comparison of ¹H spectra of aged samples of 1a and 1b (2–4 months old) with spectra obtained when the samples were fresh, it was possible to deduce the spectra of 1c and 1d. No interconversion of 1a with 1b and 1c or of 1b with 1a and 1d was detected either by NMR or HPLC. From these observations we infer that the nucleosides 1a and 1c are enantiomeric with respect to their bis-base components and likewise 1b and 1d.

3.3. Structural Characterization of the Bis-Nucleosides 1a–d

Attention was focused on the more readily available bis-nucleosides 1a and 1b. Solutions in DMSO-d₆/D₂O gave high quality spectra. The spectra showed nine signals for non-exchangeable protons in the bis-bases plus an additional 14 derived from the pair of deoxyribose residues. Of the signals ascribed to the bis-bases, three were singlets and the other six multiplets were in a coupled network. Two downfield singlets (7.93 and 8.00 ppm in 1a and 7.95 and 7.99 ppm in 1b) were assigned on the basis of chemical shift to the protons derived from H8 of the constituent Gua moieties showing that they were in non-equivalent environments. The remaining singlet, 6.53 ppm in 1a and 6.21 ppm in 1b, had to be in the fragment derived from muconaldehyde. The coupled networks were assigned using COSY and HSQC spectra as CH₂-CH-CH₂-CH spin systems (Figure 4). Isomers 1a and 1b were distinguished by the fact that in 1a the geminal protons in each methylene group were separated by slightly more than 1 ppm whereas in 1b the separation was less than 0.2 ppm. Detailed analysis of the ¹H spectra of the bis-nucleosides was complicated by the fact that solvent and deoxyribose signals overlapped some of the CH₂-CH-CH₂-CH signals in the 1-D spectra but the location of the obscured signals could be deduced from COSY and HSQC spectra. The chemical shifts of protons on the muconaldehyde fragment of 1c and 1d were assigned using a mixture of 1c and 1d along with aged samples of 1a and 1b in which varying amounts of isomerization had occurred to 1d and 1c, respectively. Assignment of signals for 1c and 1d was facilitated by the fact that the ¹H spectra of 1a and 1c and those of 1b and 1d differed only minutely from each other (Δδ values <0.1 ppm). It can be inferred that 1a–d are configurational not structural isomers with the bis-base portions of 1a and 1c and of 1b and 1d being antipodal.

Figure 4.

¹H Chemical shift and coupled networks in compounds 1a–d. Spectra acquired in DMSO-d₆/D₂O (~90:10, v/v).

The NMR spectra of 1a–d obtained in anhydrous DMSO-d₆ were not analyzed due to the complexity resulting from the presence of two non-equivalent deoxyribose spin systems. Instead, attention was turned to the bis-bases, which retained the spin system derived from the muconaldehyde moiety and promised to yield spectra that would be easier to interpret.

3.4. Bis-Bases 2ac and 2bd

Deglycosylation of a mixture of bis-nucleosides 1a–d gave a 3:2 ratio of two well-separated HPLC peaks for mixtures of 2a with 2c and of 2b with 2d (Scheme 3). These mixtures are designated as 2ac and 2bd. Deglycosylation of the individual bis-nucleosides 1a and 1c gave the enantiomeric bis-bases 2a and 2c, respectively, which eluted at the position of the first peak. Likewise 1b and 1d gave 2b and 2d, which eluted at the position of the second peak. This confirmed the previous conclusion as to the stereochemical relationships among 1a–d. As with the bis-nucleosides, samples of each enantiomeric mixture gradually became contaminated with the other one showing that 2a was slowly equilibrating with 2d and 2b with 2c.

Scheme 3.

Deglycosylation of bis-nucleosides 1a–d to form the analogous bis-bases 2a–d and reduction of the bis-bases with NaBH₄ to form 3a–d.

NMR spectra of 2ac and 2bd were acquired in anhydrous DMSO-d₆. The spectra closely resembled those of the nucleosides but the absence of deoxyribose moieties provided considerable simplification. Five ¹H signals arising from OH or NH groups were well separated from the CH signals and were identified by exchange with D₂O. The CH₂-CH-CH₂-CH spin system, which had been the hallmark of the bis-nucleosides, was still present and closely resembled the pattern in the bis-nucleosides with the exception that the methines of the spin system were each coupled to exchangeable protons (Figure 5).

Figure 5.

¹H chemical shift and coupled networks in compounds 2ac and 2bd. Spectra recorded in anhydrous DMSO-d₆. Chemical shifts (δ) in ppm. Addition of D₂O caused disappearance of the exchangeable protons, which are indicated in red.

The central methine proton was coupled to a signal near 5 ppm, which can be assigned as a hydroxyl group on the basis of chemical shift. The terminal methine proton was coupled to a signal near 7.5 ppm which, can similarly be assigned as an NH group. Three additional exchangeable signals appearing below 7 ppm can be assigned as NH protons; they showed no coupling to other protons. Two of these protons appeared as a very broad signal at ~12.7 ppm and can be assigned to the N9 protons of the Gua residues. The remaining exchangeable proton was a singlet near 8.7 ppm.

The structural information gleaned from the ¹H spectra indicated that four of the six muconaldehyde carbons are present in the CH₂-CH-CH₂-CH fragment which is separated from the singlet methine by a tetrasubstituted carbon. It can be deduced that the four-carbon fragment and the tetrasubstituted carbon comprise a five-membered ring since (1) the mutually coupled terminal methylene protons are only coupled to the central methine and (2) the terminal methine is only coupled to the adjacent methylene group plus an exchangeable proton. The chemical shifts of the singlet at 6.52 ppm in 2ac and 6.25 ppm in 2bd suggest that these methine carbons are each linked to the tetrasubstituted carbon and to two nitrogen atoms.

HMBC and HSQC spectra permitted most of the carbon signals to be identified. The combined ¹³C and ¹H data made it possible to assign the heptacyclic structures shown in Figure 6. In spite of large differences between the ¹H spectra of the two isomers, the ¹³C spectra were virtually identical, providing corroboration that the two species were diastereomers not structural isomers. A pair of signals appearing at 135 and 140 ppm for 2ac and 136 and 138 ppm for 2bd can be assigned to C8 of the non-equivalent Gua moieties. The carbon signals associated with the muconaldehyde-derived isolated methine (M1, see Figure 1) and the CH₂-CH-CH₂-CH fragment (M3–M6) were identified using HSQC correlations. The quaternary carbon (M2) was assigned to signals at 72 ppm in both diastereomers on the basis of 2-bond couplings in the HMBC spectra.

Figure 6.

¹H spectra of 2ac and 2bd were recorded in anhydrous DMSO-d₆ ¹³C spectra in DMSO-d₆/D₂O. Chemical shifts (δ) in ppm n.o. = not observed.

With 2ac a 3-bond HMBC correlation was observed between the M1 proton and C2 of Gua_W. This correlation does not distinguish between regioisomers having M1 linked to N1 vs N² of Gua_W but on mechanistic grounds (see Discussion Section) the muconaldehyde linkage to Gua_W can be assigned as M2 to N1 and M1 to N² as shown in Figure 6. The environment of M1-H in 2ab is further defined by 3-bond HMBC correlations with the M3 carbon at 43 ppm. Neither H-H nor C-H coupling was observed between M1-H and the 2-NH of Gua_W, presumably due to NH exchange being too rapid. An HMBC correlation was observed between the M6 proton at 4.64 ppm and a carbon signal at 152 ppm, which is assigned to C2 of Gua_E. This correlation could have resulted from M6 being linked to either N1 or N² of Gua_E but the vicinal ¹H coupling between the M6 proton and the 2-NH of Gua_E observed in the COSY spectrum made it possible to assign the linkages of M6 to N² of Gua_E and, therefore, M1 to N1. The HMBC spectrum of 2bd had lower signal to noise and failed to show correlations of the muconaldehyde fragment with either Gua_E or Gua_W. However, as argued above, the two species (2ac and 2bd) are diastereomers not structural isomers.

3.5. Relative Configurations of 2ac and 2bd

For the parent 5-6-5 angularly fused carbocycle equivalent to rings A, C, and B, respectively, of 2, i.e., only derivatives of the cis-cis and trans-cis isomers have been reported^53,54 and it is not known whether the cis-trans and trans-trans would be accessible. To examine the possibility that 2ac and 2bd were epimers at M1, i.e., cis and trans fusions of rings A and C, they were reduced with excess NaBH₄ (methanol-water (1:1), 16 h, room temperature) to give 3ac and 3bd, respectively, in which ring C had opened (Scheme 3). Diastereomers 3ac and 3bd were well separated from each other on reverse phase HPLC (See chromatogram on page S8 in the Supporting Information) and eluted in the same order as 2ac and 2bd; each had a mass 2 units higher than its parent and the UV spectra were identical. Both reductions would have given the same product if the structural difference between 2ac and 2bd had been the configuration of M1.

The ¹³C NMR spectra of 2ac and 2bd were virtually identical which also suggests that epimerization is not occurring at a ring junction. In view of the fact that there are four asymmetric sites, three of which (M1, M2 and M6) involve ring junctions, it points to the site of epimerization being the remaining one, i.e., M4. Thus, it is reasonable to conclude that the epimers differ in the orientation of the M4 hydroxyl group.

There are obvious differences between the ¹H spectra of 2ac and 2bd. As in the spectra of 1a and b, the chemical shift differences between the geminal protons on M3 and M5 in 2bd are much smaller than in 2ac (Figures 6 and 7). The orientation of the M4 hydroxyl relative to the M6 amino group is the basis for this chemical shift difference in the protons on M5. Both functional groups shield proximal protons.^55–57 2ac can be assigned as having M6-NH on the same face as the M4-OH on the basis of the large chemical shift difference (1.12 ppm) for the geminal pair of protons on M5 reflecting the fact that the 1.33-ppm proton is shielded by both M4-OH and M6-NH. In 2bd the chemical shifts of the M5 geminal protons differ by only 0.20 ppm. Therefore the M4-OH and M6-NH can be assigned as being trans to each other because M4-OH shields one of the geminal protons and M6-NH the other. The average chemical shifts for the M5 protons are very similar for 2ac and 2bd, i.e., 1.89 and 1.79 ppm, respectively.

Figure 7.

Chemical shift differences between the M3 and M5 geminal protons of 2ac and 2bd. Spectra were recorded in anhydrous DMSO-d₆.

A similar analysis can be used to define the orientation of M4-OH relative to the carbonyl group of Gua_W and thereby the relative configuration of the spiro carbon. The Gua_W carbonyl group has a deshielding effect on proximal protons whereas M4-OH is shielding. The average chemical shifts for the M3 protons are 2.47 and 2.60 ppm, respectively, for 2ac and 2bd. The Gua_W carbonyl group and M4-OH of 2ac are on opposite faces of ring B leading to a large chemical shift difference (1.05 ppm) between the M3 geminal protons; they are on the same face in 2bd where, to a large extent, they cancel each other out. Thus in 2ac, M6-NH, the M5 proton at 1.33 ppm, M4-OH and the M3 proton at 1.96 ppm are on one face of ring B and the Gua_W carbonyl group on the other. In 2bd, M6-NH and M4-OH are on opposite faces of the ring. The difference in chemical shifts of the two protons on M3 is too small to make relative assignments but the small chemical shift difference between the two protons on M3 is indicative that the shielding/deshielding effects cancel and the M4-OH and the Gua_W carbonyl group are proximal to the same M3 proton. Consequently, in 2bd the carbonyl group of Gua_W is on the same face of ring B as M4-OH and opposite M6-NH. This analysis of the effect of the Gua_W carbonyl group on chemical shifts of the M3 protons permits the relative configuration of spiro carbon M2 to be assigned as shown in Figure 7 leading to a cis assignment for the A-C ring fusion.

Corroborative evidence for this assignment was obtained from NOE experiments carried out on 2ac and 2bd. The spectra were acquired at 25 °C on solutions in DMSO-d₆ containing 8% D₂O; the results are summarized in Figure 8. With 2ac, irradiation of the M1 proton (6.55 ppm) produced a strong NOE to M3-H_1.94 and a weak one to M3-H_3.01. An NOE was also observed to M5-H_1.34. NOEs to M4-H and M6-H were not observed. On this basis, M1-H, M3-H_1.94 and M5-H_1.34 can be assigned to one face of the ring, i.e., the β face as depicted in Figure 8. This establishes that the A:C and B:C ring fusions of 2ac are both cis. Irradiation of the M4 methine proton (4.40 ppm) produced NOEs to M3-H_3.01 and both M5 protons with the NOE to M5-H_1.34 being stronger than that to M5-H_2.39. An NOE was also observed to the M6 methine proton (4.64 ppm). Irradiation of the M6 methine proton (4.64 ppm) gave NOEs to M4-H and M5-H_2.39. Therefore, M4-H, M6-H, M3-H_3.01 and M5-H_2.39 can all be assigned to the α face of the molecule and the NOEs between M4-H and M6-H establish that M4-OH is β.

Figure 8.

NOE experiments with 2ac and 2bd conducted in DMSO-d₆/D₂O (92:8, v/v).

A similar analysis of the NOEs for 2bd makes it possible to assign M1-H_6.20, M4-H, M3-H_2.50 and M5-H_1.72 to the β face of the molecule and M6-H_5.00, M3-H_2.68, and M5-H_1.96 to the α face. It is of particular note that an NOE was observed between M1-H and M4-H establishes that the A:C and B:C ring fusions are both cis and the M4-OH is on the α face of the molecule.

In support of the A-cis-C-cis-B assignment of ring fusions, Ihara et al. have synthesized pentalenene, a triquinane-type sesquiterpene, by a route that passed through a tricyclo[7.3.0.0^1,5]dodecanedione intermediate, which is a carbocyclic analog of rings A, C and B. The method of synthesis of this intermediate led to formation of the thermodynamically most stable stereoisomer and the subsequent conversion to pentalenene established that the ring A:C and B:C ring fusions in the intermediate were both cis.⁵³ Furthermore, DFT energy calculations on 2 having cis and trans fusions of rings A and C (carried out without the M4 hydroxyl group) using the 6–31G* basis set with the B3LYP functional indicate that a cis fusion is significantly favored over trans.⁵⁸

3.6. Absolute Configurations of the Diastereomers of 1

The CD spectra of 1a and b were virtually identical indicating that the overall shapes and major chiral features were the same (See Supporting Information). The CD spectrum of a mixture (57:43) of 1c/1d was essentially a mirror image of those of 1a and 1b. This provides evidence that 1a and b differ only in the orientation of the M4 hydroxyl group since reversing the configuration of one of the ring junctures would be expected to influence the interaction of the Gua residues with each other. It is noteworthy that the configuration of the M4 hydroxyl group has negligible impact on the CD spectra of 1a and b. The application of the Octant Rule to configurational assignment of 1a and b was problematic due to the two Gua chromophores. Assignments of absolute configurations as shown in Scheme 4 were made on the basis of comparison with CD spectra of pyrimidopurinone-type dGuo adducts of 4-hydroxynonenal which had been synthesized by methodology that provided unambiguous assignment of absolute configuration.⁵⁹ Further support for the configurational assignments was obtained from comparison of the CD spectra of 1a and b with those of tetracyclic 1,N²-ethano-type dGuo adducts of 2,3-epoxy-4-hydroxynonanal which are excellent models for the Gua_W-A-C rings in 1.⁶⁰

Scheme 4.

Equilibria of bis-nucleoside 1a with 1d and of 1b with 1c, reduction of 1a–d to bis-nucleosides 4a–d with NaBH₄, and configurational assignments of 1a–d and 4a–d.

Reduction of a mixture of bis-nucleosides 1a–d with NaBH₄ gave dihydro products 4a–d (Scheme 4), which were readily separated by HPLC. Reduction of the diastereomers of 1 individually revealed that the order of HPLC elution of the reduced bis-nucleosides had changed to 4c < 4d < 4a < 4b. Reduction brought about replacement of the M1 singlets seen near 6 ppm in the bis-nucleosides by a pair of vicinal doublets in the region of 3.5 ppm establishing that reduction had cleaved the bond between carbon M1 and N1 of Gua_E. COSY and HSQC spectra showing the presence of the CH₂-CH-CH₂-CH network confirmed that the cyclopentane ring remained intact. Significant changes had occurred in the chemical shifts of the CH₂-CH-CH₂-CH protons such that all four reduced bis-nucleosides had unique spectra.

The reduction brought about a remarkable change in the ¹H spectrum of one deoxyribose in each diastereomer. The two anomeric protons in 1a–d had had unexceptional chemical shifts near 6.05 ppm. In the reduced species, one of the anomeric protons was strongly shielded giving upfield locations of 5.15, 5.14, 5.51, and 5.61 ppm, respectively, for 4a–d. The other anomeric proton and the H8 protons of the two dGuo rings had also been shifted upfield but only by 0.1–0.2 ppm. Bis-nucleosides 1a–d are rigid dihedral species with the two Gua residues extending away from one another. The reduced species have freedom of rotation around both C-N bonds of N² of Gua_E and increased capability for the cyclopentane ring to pucker permitting the Gua residues to stack. The anomeric proton of dGuo_E then lies near the face of Gua_W producing the shielding. Apparently, for 4a and 4b the anomeric protons of the dGuo_E deoxyribose residues lie in a more highly shielded location than for 4c and 4d, i.e., closer to Gua_W or the population of the folded conformers may be higher leading to larger upfield shifts.

CD spectra for 4a and 4b show negative ellipticity in the 280 nm region and positive ellipticity near 255 nm; 4c and 4d show positive ellipticity near 280 nm and negative near 255 nm (Figure 9). As with 1a and 1b, the Octant rule can not be reliably used for assignment of the configurations of 4a–d but the compounds are well-suited for application of Nakanishi’s exciton chirality method,^61,62 taking advantage of the relative orientations of the two chromophores. Molecular models reveal that having S and R configurations, respectively, for M2 and M6 in 4a and 4b leads to a left-handed helical relationship of the two Gua moieties (Figure 10) yielding the observed negative ellipticity for the longest wavelength CD band. Likewise having (R)-M2 and (S)-M6 in 4c and 4d leads to a right-handed relationship and positive ellipticity. This provides confirmation for the absolute configurations of 4a–d shown in Scheme 4 that had been deduced for 1a–d above. It can be seen in Scheme 4 that the configurational designations of M2 in 4a–d are reversed from 1a–d due to the change of Cahn-Ingold-Prelog priority rules.

Figure 9.

CD spectra of reduced bis-nucleosides 4a–d obtained at ambient temperature in water-methanol (1:1, v/v) mixtures.

Figure 10.

Depiction of the helical relationship of Gua residues in 4a and 4b versus 4c and 4d.

Comparison of the CD spectra of the reduced bis-nucleosides with those of B- and Z-DNA is illuminating. The spectra of 4a and 4b are similar to that of Z-DNA whereas the spectra of 4c and 4d are similar to that of B-DNA.⁶³ This provides additional confirmation of the configurational assignments since Z-DNA has left-handed helicity while B-DNA has right-handed.

4. DISCUSSION

The formation of bis-nucleosides 1a–d by the condensation of muconaldehyde with dGuo involves a cascade of reactions (Scheme 5). We believe the reaction sequence is initiated by addition of dGuo to a carbonyl group of muconaldehyde forming carbinolamine 5 and then cyclization by attack of N1 of dGuo_W on C2 on the muconaldehyde moiety to give 1,N²-ethano derivative 6 followed by imine formation with a second dGuo to give 8. We cannot exclude the possibility that the reverse sequence occurs in which 5 first reacts with the second dGuo to form imine 7 followed by cyclization to give 1,N² derivative 8, or that the two processes occur concurrently. Mass spectroscopic examination of early intermediates provided support for both mono- and bis-adducts. The formation of 8 is analogous to the reactions of dGuo with butene-1,4-dial,⁶⁴ 4-oxo-2-nonenal,⁶⁵ and with 2,3-epoxyaldehydes.^60,66

Scheme 5.

Proposed mechanism for formation of bis-nucleosides 1a–d.

A 1,4-addition of hydroxide ion to 8 yields imine 9. This could occur either by reaction with water or by an intramolecular attack of the hydroxyl group of 7 leading to tetrahydrofuran 10, followed by ring-opening of 10 to give 9. Imine 9 tautomerizes to 1,N²-etheno dGuo 11, which is in fact an enamine. Attack of the enamine on the imine of 11 creates the cyclopentane ring in 12. The final ring closure giving 1a–d occurs by attack of N1 of dGuo_E on the western imine of 12.

All steps in this reaction sequence are potentially reversible although we only detected reversibility for the final three steps. Specifically, we have found no evidence for the hydroxyl group at M4 ever epimerizing. This means that 9 does not revert to 8 under the reaction conditions we employed. Equilibrium between 1a and d and between 1b and c is evidence that the cyclopentane ring in 12 is opening and reclosing to invert the configuration at quaternary carbon M2. Furthermore, equilibration of 1 with precursor 12 is demonstrated by the reduction of 1 with sodium borohydride that is, in fact, a reduction of 12. Each diastereomer undergoes reduction with complete stereospecificity, i.e., the reduction of imine 12 by borohydride ion is much faster than the reversion of 12 to 11. Furthermore, we found no evidence for reduction products that would have arisen from 9 or 11. Also, we have not observed loss of the dGuo_E residue from 1, even when NMR samples in wet DMSO-d₆ were allowed to stand for more than 2 months at room temperature. Nevertheless, since 1 is in a slow equilibrium with 11, the possibility exists that Gua_E could be lost under appropriate conditions. Reversibility of the pathway by which diastereomers 1a–d are formed leads to them being the thermodynamically most favored forms. Cis fusions of ring C with rings A and B have been assigned for 1a–d. A strong preference for this stereochemistry explains why only four diastereomers are observed even though the presence of 4 stereogenic centers (M1, M2, M4, and M6) would have predicted the formation of as many as 16.

5. CONCLUSIONS

This paper demonstrates that muconaldehyde is able to form stable cross-links between dGuo residues. A central issue is whether muconaldehyde-based bis-dGuo adducts might form in DNA and become the chemical basis for the observed carcinogenicity of benzene. It may be problematic for bis-adducts 1 per se to be present as interchain cross-links in either a GpC or CpG sequence context in duplex DNA because the two Gua moieties of 1 are in an anti relationship to each other whereas the Gua residues in native DNA are syn. Acrolein and mitomycin C form Gua-Gua interchain cross-links in a CpG sequence context; structural studies have shown the guanines retain the relative orientations of native DNA in both cases.^48,67 Similarly, the two thymines involved in photochemically induced cross-linking by psoralen retain their syn relationship.^68,69 Intrachain cross-links involving 1 in a GpG sequence context will be precluded by geometrical constraints.

Nevertheless, the chemistry elucidated in the present study reveals that imine 12 and the cognate carbinolamine 13 will be in equilibrium with 1 (Scheme 6). Both would be promising candidates for interchain cross-link formation because they would be able to adopt a syn relationship between the Gua moieties needed for formation of relatively undistorted interchain cross-links. Moreover, 12 and 13 would have the flexibility to form intrachain cross-links. Our observation that the preferred conformation for 4 involves a partially stacked arrangement for the two Gua residues supports the possibility that an intrachain cross-link could form. The intrachain cross-link would be particularly attractive when the configurations of M2 and M6 are as in 4c and 4d so that the bis-adduct could conform to a right-hand helix. Muconaldehyde-induced tandem mutations reported by Nakayama et al. were unique to this compound; this observation lends credence to the possibility of muconaldehyde-mediated intrachain cross-links.⁴²

Scheme 6.

Reversible opening of ring C of bis-nucleosides 1 to form imines 12 and carbinolamines 13.

The most attractive type of muconaldehyde cross-links may involve those between DNA and proteins in which the free aldehyde of initially formed muconaldehyde mono-adducts 5 and/or 6 (Scheme 7) reacts with the ε-amino group of lysines to give 14 or analogously with the N-terminus of the protein. Evidence for DNA-protein cross-links has been reported for muconaldehyde-treated HL-60 cells by Schoenfeld et al., who have postulated that these cross-links may contribute to benzene-induced hematotoxicity and leukemogenesis.⁴³

Scheme 7.

Proposed reaction of the dGuo-muconaldehyde monoadduct with the ε-amino group of lysine to form DNA-protein cross-links.

ABBREVIATIONS

COSY

correlation spectroscopy

DFT

density function theory

dG-muc-dG

2:1 adducts 1a–d of deoxyguanosine with E,E-muconaldehyde

DMSO

dimethylsulfoxide

ES

electrospray

ESI-MS

elecrospray ionization mass spectrometry

G-muc-G

2:1 adducts 2a–d of guanine with E,E-muconaldehyde

HMBC

heteronuclear multiple bond correlation spectroscopy

HSQC

heteronuclear single-quantum correlation spectroscopy

MW

molecular weight

NOE

nuclear Overhauser effect

OSHA

U.S. Occupational Safety and Health Administration

SPE

solid phase extraction