Identification of a Reduction Product of Aristolochic Acid: Implications for the Metabolic Activation of Carcinogenic Aristolochic Acid

Abstract

Aristolochic acids are nephrotoxic and carcinogenic natural products that have been implicated both in the endemic nephropathy in the Balkans region and ailments caused by ingestion of herbal remedies. Aristolochic acids are metabolized to active intermediates that bind to DNA. In this study, reduction of aristolochic acid I with zinc/acetic acid afforded a new product which was characterized as 9-methoxy-7-methyl-2H-1,3-oxazolo[5′,4′-10,9]phenanthro[3,4-d]1,3-dioxolane-5-carboxylic acid, designated as aristoxazol, along with the expected aristolactam I. This new compound is a condensation product of aristolochic acid and acetic acid that may be related to the aristolochic acid-DNA-adducts. The proposed mechanism of formation of aristoxazol involves nucleophilic attack of acetic acid on the nitrenium ion of aristolochic acid I. On the basis of these studies a route to the metabolic activation of aristolochic acids and formation of adducts with DNA in in vitro systems is proposed and discussed.

Aristolochic acids are carcinogenic and nephrotoxic nitroarenes that occur in Aristolochia species, with aristolochic acids I and II (AAI and AAII, 1 and 2, respectively; for numbering see structure 1) usually being the most abundant of these compounds. Ingestion of AAs is implicated in the type of renal fibrosis known as “Chinese herbs nephropathy” or “Balkan endemic nephropathy”.^1-9 A report on the carcinogen aristolochic acid was recently issued by the U.S. Department of Health and Human Services.¹⁰

The toxicity of AAs is due to the formation of active intermediates during the detoxification process. The crucial step in the generation of DNA-reactive and mutagenic metabolites is the reduction of the nitro group of AAs. In rats and other mammals, the major metabolic pathway involves reduction of AAI (1) and AAII (2) to aristolactams (5, 5a; Scheme 1). In addition, O-demethylation of AAI (1) and aristolactam I (5), as well as oxidation at C-8 of AAII (2) are observed. All of these metabolites are detected in urine and feces in either their original form or as conjugates of glucuronic or sulfuric acids. Furthermore, covalent adducts of DNA such as (deoxyguanosin-N²-yl)aristolactam I (dG-AAI) (7), (deoxyadenosin-N⁶-yl)aristolactam I (dA-AAI) (8), (deoxyguanosin-N²-yl)aristolactam-II (dG-AAII) (7a) and deoxyadenosin-N⁶-yl)aristolactam-II (dA-AAII) (8a) (Scheme 1) were identified in the renal tissue of patients that ingested herbal preparations containing AAI and AAII.^5-7 Interestingly, only adducts formed at C-9 (C-7 in an alternative numbering)^5,7,11 of the aristolochic acids are known. There are no reports of adduct formation at the C-10 nitrogen as usually occurs with nitroaromatic compounds.

Scheme 1.

Metabolic activation and DNA adduct formation of AAI (1) and AAII (2) (1, 3, 4, 5, 6, 7 and 8, R = OCH₃; 2, 3a, 4a, 5a, 6a, 7a and 8a, R = H).

The accepted mechanism for the reductive transformations of AAs into DNA adducts is shown in Scheme 1. ^5,7,11 It was proposed that the aristolactam-nitrenium ion 4 is the ultimate carcinogenic species which binds to the amine group of purine nucleotides (7, 7a, 8, 8a), or is hydrolyzed to the corresponding 9-hydroxyaristolactam I (6). Note that in previous reports a different numbering system was used in which this compound was named 7-hydroxyaristolactam.¹¹

In view of the importance of the reduction of AAs on genotoxicity, we reinvestigated the reduction of AAI using zinc in acetic acid. As a result, we report the identification of a new reduction product from AAI, namely 9-methoxy-7-methyl-2H-1,3-oxazolo[5′,4′-10,9]phenanthro[3,4-d]1,3-dioxolane-5-carboxylic acid, which is named aristoxazol (15). We discuss its mechanism of formation and the possibility that there is a nitrenium ion-based pathway leading to adduct formation, which is complementary to that proposed by Pfau et al.¹¹

Results and Discussion

1. Identification of Aristoxazol

Previous studies have established that AAs underwent reduction to either the corresponding aristolactam or a complex mixture of uncharacterized products.¹² In this study, AAI was reduced with zinc powder because this is a well-established route for the reduction of nitro groups. The zinc reduction of AAI (1) in boiling acetic acid yielded the expected aristolactam I (5) as the major product. However HPLC analyses revealed the presence of a minor reduction product, subsequently identified as aristoxazol (15) [for simplicity the numbering of the atoms matches that of AAI (1)]. When AAI was reduced with Zn/acetic acid at temperatures between 60 °C and 118 °C aristolactam I (5) and aristoxazol (15) were obtained in a ca. 3:1 ratio. The reaction proceeds at room temperature (25 °C), but the aristolactam I: aristoxazol ratio changes to 6:1. Aristoxazol (15), MW 351, C₁₉H₁₃NO₆, is a stable compound. It cannot be reduced further when treated under the same conditions.

The molecular formula indicates that two carbon atoms were added to AAI during the formation of aristoxazol. These were supplied by acetic acid since the reduction was carried out in this solvent. The UV spectrum shows a strong absorption at 251 nm typical of the phenanthrene nucleus. The possibility that aristoxazol was an N-acetyl derivative related to aristolactam I was discounted for several reasons. Evidence of an intact carboxylic acid group was provided by chemical behavior, the detection of a fragment ion at m/z 306 (M-44) under electron impact, the IR spectrum (carbonyl stretch at 1668 cm^-1), and formation of a methyl ester with diazomethane. The ¹H NMR spectrum (Table 1) provided limited data because the compound contains only a small number of hydrogens. The presence of a broad singlet at ca. δ_H 13, which exchanged with D₂O gave further confirmation of the carboxylic acid moiety. The doublet at δ_H 8.74, triplet at δ_H 7.62, singlet at δ_H 7.49 and doublet at δ_H 7.34 were assigned as the aryl hydrogens at C-5, C-6, C-2 and C-7, respectively, on the basis of the splitting patterns and comparison of the chemical shifts with those of AAI.¹³ The expected singlet for H-9 (ca. δ_H 7.20) was absent, as well as the signal for a probable NH lactam proton (ca. δ_H 10.70). The singlets at δ_H 6.38 and δ_H 4.04 were assigned to the methylenedioxy and methoxy groups. An extra methyl group was observed at δ_H 2.69. The above results and structural considerations suggested the phenanthroxazol structure 15 for the unknown compound, the methyl group being attributed to the C-12 methyl substituent.

Table 1.

¹H NMR and ¹³C NMR Data for Aristoxazol (15) and Aristoxazol Methyl Ester in DMSO-d₆ (δ in ppm, J in Parenthesis)

pos.	Aristoxazol (15)		15 Me ester
	δH	δCa	δH
1		127.7
2	7.50 s (6.5)	109.1	7.57 s
3		144.4
4		144.3
4a		114.1
4b		125.0
5	8.74 d (8.4)	119.4	8.80 d (8.4)
6	7.62 t (8.4)	127.1	7.70 t (8.4)
7	7.34 d (8.0)	108.9	7.42 d (8.0)
8		153.5
8a		111.8
9		142.6
10		133.2
10a		117.3
12		161.6
1-COOH	13.00 s (broad)	170.1
3,4-CH2O2	6.38 s	101.8	6.43 s
8-OMe	4.04 s	56.0	4.08 s
12-Me	2.69 s	14.3	2.73 s
1-COOMe			3.94 s

^aAssignments of signals with similar chemical shifts may be reversed.

The proton-decoupled ¹³C NMR spectrum (Table 1) supported structure 15. The spectrum was composed of 19 signals, seven of which appeared above δ_C 130. Four of these signals were assigned by comparison with those of AAI¹⁴ as follows: δ_C 170.1 (carboxyl group), δ_C 153.4 (C-8), δ_C 144.3, and 144.2 (C-3 and C-4 or vice versa). The remaining three downfield signals at δ_C 161.6, δ_C 142.6, and δ_C 133.2 were assigned as C-12, C-9, and C-10 of the benzoxazole unit based on a comparison with the corresponding carbons of isosalviamine C (16) which have chemical shifts of δ_C 162.3, δ_C 143.0, and δ_C 133.0, respectively.¹⁵ Further support for the oxazole ring is provided by the signal at δ_C 14.3, whose chemical shift matches very well with that of the corresponding methyl substituent in 2-methylbenzoxazol (δ_C 14.4) and natural products containing this moiety (δ_C 14.9-15.1).¹⁵ Equally important, this signal is much further upfield than the methyl carbons of acetyl groups on aryl substrates such as acetanilide (δ_C 24.1), phenyl acetate (δ_C 21.1), and acetophenone (δ_C 25.7).

The UV spectrum of aristoxazol shows a remarkable resemblance to that of aristolic acid (19, Figure 1) (λ_max at 254.5, 296.2, 319sh, 327.1, 355.0, and 374.5 nm), which indicates that these compounds have similar chromophores. The two minor bands of aristoxazol at 363 and 382 nm are about 8 nm bathochromically shifted from those of the desnitro AAI, a shift that can be attributed, in part, to the presence of the oxazole ring at C-9 and C-10. Based on the comparison of the λ_max values for benzene (254 nm) and 2-methylbenzoxazole (271 and 277 nm),¹⁶ a bathochromic shift of about 20 nm would have been expected. The smaller bathochromic shift observed with aristoxazol is attributed to steric compression by crowding of the substituents at the peri 8, 9, 10 and 1-positions.¹⁷ Molecular modeling of the aristoxazol revealed distortion within the phenanthrene ring (Figure 1), presumably as a means of reducing steric compression. Support for this proposal was obtained from the modeling of aristolic acid, which lacks the oxazole ring. The data revealed that the phenanthrene ring of aristolic acid was essentially planar (Figure 1). In summary, chemical and spectroscopic data support the proposed structure 15 for aristoxazol in which the oxazole unit is fused to the C-9–C-10 bond of the phenanthrene moiety.

Figure 1.

Calculated ball and wire structures of aristolic acid and aristoxazole

2. Considerations about the Formation of Aristoxazol

Reduction of AA-I with Zn/acetic acid produces aristolactam I (5) accompanied by smaller amounts of aristoxazol. Overall, the formation of aristoxazol from AAI involves the reduction of the nitro group and the insertion of an acetoxy group at C-9, which will generate the oxazole unit after several steps. Thus, aristoxazol is a derivative of an intermediate reduction product of AAI. In contrast, aristolactam I (5) represents the final product from the complete reduction of the nitro group of AAI.

Reduction of nitro arenes to amines using Zn occurs via the nitroso and hydroxylamine intermediates. Each reduction step occurs by a series of single electron transfers (SET) in which an electron from the Zn and a proton are transferred to the species being reduced.¹⁸ Thus, the formation of the N-hydroxyarylamine 9, and the aryl amine 10, which are probable intermediates in the formation of aristoxazol (15) and aristolactam I (5), are assumed to occur via SET reactions (Scheme 2). Both amines 9 and 10 can readily cyclize to form the more stable aristolactam, 5, either directly or via the postulated N-hydroxylactam, 3. Thus, formation of aristolactam I from the reduction process is entirely expected. Furthermore, the stability of the lactam makes it an unlikely precursor to aristoxazol indicating that the pathway must diverge prior to the reduction that leads to aristolactam.

Scheme 2.

Proposed mechanism for formation of aristoxazol (15).

With the exception of structures 1, 5 and 15, none of the structures in this scheme have been isolated. Thus, it is not possible to determine their bioactivities. The bioactivities of structures 1 and 5 are well established.⁵,⁷,¹¹,²⁵

More than 100 years ago Bamberger reported that N-phenylhydroxylamine rearranged in sulfuric acid solution to give 4-aminophenol.¹⁹ The key features of this reaction are the O-protonation of the NOH group followed by N-O bond cleavage and loss of water to form an intermediate nitrenium ion.^20,21 The above studies suggest that formation of aristoxazol (15) during reduction of AAI with Zn/acetic acid involves formation of the corresponding N-hydroxylamine and a nitrenium ion as the reactive intermediate species.

It was important to consider whether the aristoxazol was formed by a variation of the reactions of the nitrenium ion, 4, of Scheme 1, which was proposed for the activation of AAs in biological systems. However, we were unable to account for the formation of aristoxazol via this pathway. The stumbling block is that hydrolysis of the lactam structure, 3, in Scheme 1, which is required to generate the N-hydroxylamine 9, the expected initial precursor of aristoxazol, would not be possible under the reaction conditions. This topic is discussed further in Section 7.

In light of these difficulties, it was necessary to propose an alternative pathway, which is shown in Scheme 2. Before discussing the most reasonable pathway to aristoxazol, it is necessary to examine the possibility that aristoxazol formation occurred by direct heterolysis of the N-O bond of the N-hydroxylamine 9 to give the nitrenium ion 12 prior to the formation of the lactam 3 or reduction (Scheme 2). This conversion of the N-hydroxylamine 9 into the nitrenium ion 12 seems unlikely because the acetic acid medium is insufficiently strong to drive the transformation of 9 into 12. Even if it occurs, it is probable that it would be very slow, which would mean that spontaneous formation of the lactam 3 or reduction to 10 would be by far the predominant processes. Although the 9-to-12 conversion cannot be ruled out, an alternative and more reasonable mechanism is shown in Scheme 2. In this mechanism, formation of aristoxazol is proposed to occur via the formation of an oxazinone 11.

It is assumed that the 9-hydroxylamine, 9, can be stabilized by spontaneous cyclization. The carboxyl group of 9 can undergo an addition-elimination reaction with the hydroxylamine unit to form either the lactam 3 by reaction at the nitrogen or the oxazinone 11 by reaction at the oxygen (pathways a and b in Scheme 2). Also, the nucleophilic addition-elimination reaction of a hydroxylamino oxygen atom with a carboxylic acid has been documented.^22,23 It is proposed that part of the N-hydroxylamine 9 is converted into the oxazinone 11 instead of being transformed into either the lactam 3 or the amine 10 by further reduction. In contrast to the N-hydroxylamine 9, the oxazinone 11 could readily ionize to give the nitrenium ion 12 which could be attacked by the available nucleophiles such as acetic acid and water. The 9-acetoxy adduct 13 could then undergo aromatization and dehydration to give aristoxazol (15). Like the nitroso derivative of AA-I (not shown), N-hydroxyarylamine, 9, arylamine, 10, and the oxazinone, 11, would be too reactive to be either isolated or detected among the reaction products. Similarly, it is proposed that transformation of 13 into 15 via the more stable isomer, 14 (Table 2) may occur rapidly and synchronously because formation of the lactam by reaction of the C-1-carboxyl group and the C-10 amino group in 14 was not observed.

Table 2.

Density Function Calculations^a of Likely Intermediates to Aristoxazol and Related Compounds

Compound	Energy (a.u.)
1	-1235.9836
3	-1085.5347
5	-1010.3526
9	-1161.9774
10	-1086.8029
11	-1085.5324
13	-1314.7125
14	-1314.7325
15	-1238.2741

^aCalculations were carried out at the B3LYP 6-311+G** level.

The N-hydroxyaristolactam, 3, was not detected directly in this synthesis, which is not altogether surprising because this compound has also not been isolated as a product of other chemical or enzymatic reductions of AAI. Intriguingly, the analogous N-hydroxyaristolactam of aristolochic acid II is a stable compound that has been characterized in plants (see Section 11).

Support for the formation of 11 is provided by the known cyclization of the oximes of 2-acylbenzoic acids to form 2,3-benzoxazin-1-ones.²² A similar reaction also occurs during the reduction of o-nitrobenzoic acid with tin or zinc leading to the formation of 2,1-benzisoxazolin-3-one following cyclization of the reduction intermediate, o-hydroxylaminobenzoic acid.²³ Additionally, the final condensation between the amino and acetoxy groups in adduct 14 to give the oxazole ring resembles the reaction of 2-aminophenols with carboxylic acids, which is usually employed to prepare oxazole derivatives.²⁴

Formation of aristoxazol is discussed further in connection with problems concerning the mechanisms of aristolochic acid activation and formation of adducts (see Section 7).

3. Molecular Mechanics Calculations

In view of the fact that most of the compounds that are postulated in Scheme 2 were not isolated, we carried out molecular mechanics Density Function calculations at the B3LYP 6311+ G** level. The energies of the calculated structures are summarized in Table 2. The resultant model of aristoxazol shows a distortion within the phenanthrene skeleton (Figure 1) which would explain the smaller than expected bathochromic shifts in its UV spectrum (Section 1). The calculations indicated that compound 14 was more stable than its isomer, 13, as would be expected. Calculations also showed that the N-hydroxyaristolactam, 3, was only slightly more stable than the isomeric oxazinone, 11. This result was unexpected because the 8-demethoxy analogue, 3a, of the N-hydroxylactam, 3, is a stable compound that was isolated from plants (see Section 11). We repeated the calculations at different levels of sophistication, but in all cases, the energies of the two compounds were similar. Nevertheless, we have included the data (Table 2) since it is the first time that such information has been reported for aristolochic acid and related compounds. The bond lengths, bond angles, and molecular dipole moments for the compounds that are listed in Table 2 are given in the supporting information.

4. Mechanism of Aristolochic Acid Activation via the N-hydroxyaristolactam and Formation of DNA-adducts

The mechanism for metabolic activation of AAs and DNA-adduct formation proposed by Pfau et al. is shown in Scheme 1 in which the key features are the formation of the N-hydroxyaristolactam 3 and the aristolactam-nitrenium ion 4.^5,7,11 The cyclic nitrenium ion 4 with its delocalized positive charge was suggested to be the ultimate carcinogen. AA-adduct formation occurs by reaction of the electrophilic C-9 of the nitrenium ion with the nucleophilic exocyclic amino group of purine nucleotides in DNA. While adducts of AA-reduction intermediates with DNA (7, 7a, 8, 8a) and the lactams 5 and 5a have been characterized, the reactive species that covalently binds to DNA and their precursors remain unknown. Attempts to isolate the N-hydroxyaristolactam (3) were unsuccessful.¹¹ However, support for the proposed mechanism was obtained by the generation of DNA-adducts from both aristolactam I (5) and the N-chloroaristolactam II.^25,26 Isolation of 9-hydroxyaristolactam I (6) from enzymatic reduction of AAI also provided indirect evidence for the existence of the cyclic hydroxamic acid intermediate 3.¹¹

Although the proposed mechanism explains much of what is observed, some factors in the formation of DNA adducts remain unclear. For example, in vitro experiments revealed that AAs readily form DNA-adducts under conditions in which it would be impossible to activate the hydroxylamine 3 as O-sulphate or O-acyl ester. Similarly, the regioselectivity of DNA-adduct formation at C-9 is somewhat surprising because nitroarenes usually afford both N and C adducts. Another intriguing question is why medicinal and edible plants containing significant concentrations of N-hydroxyaristolactams are non-toxic. These issues are discussed below in relation to the reduction product aristoxazol.

5. Activation of Nitroarenes: Nitroreduction and O-esterification of the N-hydroxylarylamine

Polycyclic nitroaromatic compounds are known to be metabolized into potent carcinogens in laboratory animals. They undergo reduction of the nitro group to give reactive nitrenium ions that bind to DNA with N-hydroxyarylamine intermediates being the immediate precursors of the nitrenium ions. Mutagenicity and carcinogenicity are dependent upon two activation reactions, namely nitroreduction and O-esterification of the N-hydroxylamine. Specific nitroreductases and acyltransferases are involved in these processes.^27,28 The -hydroxylamines can undergo N-O bond cleavage under mildly acidic conditions, however for this reaction to be biologically relevant a previous activation step involving formation of reactive esters is necessary to render the N-O bond labile enough for cleavage. There is strong evidence that the sulfuric or carboxylic acid esters of the resulting N-hydroxylamines are among the more important carcinogenic metabolites because in aqueous solution they undergo heterolysis of the N-O bond to yield the arylnitrenium ions.^28-32 It is likely that the N -hydroxylamine 3 is activated as sulfate or acyl ester prior to the formation of the nitrenium ion 4 since it was demonstrated that expression of the human sulphotransferase SULT1A1 in bacterial and mammalian target cells enhances the mutagenicity of AAs.³³

O-esterification of N-hydroxylamine intermediates of nitroarene carcinogens is required to exert biological effects. However, in the case of AAs there is also evidence of formation of DNA-adducts under reduction conditions that are not conducive to such esterification processes. This subject is discussed below.

6. “In vitro” Experiments. Adduct Formation by Aristolochic Acids without Activation of the N-hydroxylamine by O-sulfation or O-esterification

AAs can efficiently form DNA-adducts under in vitro conditions where the activation of the N-hydroxylamine is not possible. For example DNA-adducts were formed by reduction of AAs with either zinc in potassium phosphate buffer, pH 5.8, at 37 °C or zinc in aqueous 1 % acetic acid at 37 °C in the presence of DNA.^34-36 Likewise, enzymatic reduction of AAI by xanthine oxidase in the presence of hypoxanthine and calf thymus DNA also yielded DNA-adducts.¹¹ The nitroreduction products of most nitroarenes are barely mutagenic unless they are activated by esterification^27-32; however, a few nitroarenes seem to be mutagenic in bacteria after nitroreduction without esterification.^37-40

As was the case in the above reactions and the enzymatic generation of the 9-hydroxyaristolactam 6,¹¹ aristoxazol was obtained under conditions that were not conducive to activation of the hydroxy group of the N-hydroxylamine by O-sulfation or O-acylation. Formation of the oxazinone 11 would provide an explanation for the apparent lack of activation with aristolochic acids during in vitro reductions, as will be discussed in Section 8.

7. Probable Routes for Aristoxazol Formation

In principle, it could be possible to account for the formation of aristoxazol by reduction of AAI with Zn/acetic acid using the mechanism of Scheme 1 in the following way. Under acidic conditions, the N-hydroxyaristolactam 3 could ionize to the nitrenium ion 4 which would be then trapped by acetic acid. The resulting 9-acetoxy adduct could undergo hydrolysis of the lactam moiety to yield the acyclic form (10-amino-1-carboxyl derivative). Then, condensation of the 10-amino group with the 9-acetoxy group would generate aristoxazol. Thus, the nitrenium ion 4 could be a precursor of both the 9-hydroxyaristolactam-I (6) and aristoxazol (15). However, this possibility fails to take into account several factors.

The ionization of N-hydroxylamines is moderate under weakly acidic conditions at pH close to 6. Such cleavage should be even slower with the N-hydroxyaristolactam 3 due to “intramolecular acylation”. The N-O heterolysis is energetically less favorable in arylamides than in the parent N-hydroxyarylamines. For example, studies with model esters of N-arylhydroxylamines established that the reaction rate of the N-O heterolysis to give the nitrenium ion would be 10⁶-fold slower when the NH group is replaced by the N-acetyl group.⁴¹ Consistently, computational studies show that N-acylation is deactivating (as opposed to O-acylations which are activating).^42,43 Consequently, the N–O bond cleavage in the N-hydroxyaristolactam 3 would be strongly suppressed and the 3-to-4 ionization without activation via O-esterification would be extremely slow for reactions such as the in vitro reductions described above. The same problem is likely to occur in the formation of aristoxazol from the N-hydroxyaristolactam 3. In order to form aristoxazol from 3, it is necessary that the cleavage of the N–O bond (3 to 4) be as fast as reduction (3 to 5 or 9 to 10) which does not seem to be the case. Thus, the participation of the N-hydroxyaristolactam 3 in the formation of aristoxazol seems unlikely.

The strongest evidence against the formation of aristoxazol from either the N-hydroxyaristolactam-I (3) or the putative 9-acetoxy derivative of 5 is that these species would have to undergo hydrolysis of the stable lactam ring to form the acyclic 10-amino-1-carboxylic acid. In that way the 9-acetoxy group can react with the C-10 amine. Ring opening of the lactam is difficult, at best. For example, the amide units of aristolactams and related compounds (e.g. benzanilide) are not cleaved when treated with boiling acetic acid. Furthermore, the opening of the lactam unit of aristolactams probably generates substantial strain from interactions of the substituents at C-9, C-10, and C-1. Clearly, opening the lactam ring is an energetically unfavorable process, which is very unlikely to occur under the reaction conditions that lead to the formation of aristoxazol. An alternative pathway for aristoxazol formation is proposed below.

8. Proposed Mechanism for the Formation of Aristoxazol via the Oxazinone 11

The 9-hydroxyaristolactam I (6) and aristoxazol (15) are significant in vitro reduction products of AAI, accompanying the major product aristolactam I (5). It seems likely that these compounds were not formed by the simple dissociation of the N-hydroxyaristolactam 3 (without activation) for the reasons discussed above. If the reduction products 9-hydroxyaristolactam I and aristoxazol are congeners to the known AA-DNA-adducts, both of them may be derived from a common nitrenium ion precursor. The high reactivity of AA-intermediates during in vitro reductions suggests that aristoxazol is formed via the pathway shown in Scheme 2 in which the oxazinone 11, which would be the activated form of the N-hydroxylamine 9, forms the nitrenium ion. This pathway also avoids the need for the unfavorable ring opening of the lactam ring, which is a key step in the route based on the N-hydroxyaristolactam 3 (Scheme 1).

9. Considerations about the Pattern of Adducts Formed. Only One Type of Aristolochic Acid Adduct is Identified Whereas Nitroaromatic Compounds Usually Generate Two or More Types of Adducts

Oxidation of aromatic amines or amides or reduction of aromatic nitro compounds produces nitrenium ions with reactive sites which are targets for nucleophilic reagents. Although nitroarenes may produce several types of adducts,^44,45 the two that are most often encountered involve the N atom and/or the ortho ring carbon which can bind to C-8 or the amino group of the purine base, respectively.^44,45 For example, 6-nitrochrysene (17), which has a similar aromatic skeleton to AAs, forms adducts at the N and at the adjacent carbon (see arrows in formula 17).^45,46 The carcinogenic N-acetyl-2-aminofluorene (18) also forms two adducts after being metabolically activated by N-hydroxylation followed by O-esterification (see arrows in formula 18). Structure 18 resembles the N-hydroxyaristolactam 3 in that it too is an N-acylated compound.

A priori predictions of the base/carcinogen type of adduct formed are difficult because multiple factors are involved, which include, for example, DNA repair mechanisms and the stability of the adducts during isolation.^44,47-49 Nevertheless, the relative preferences of carcinogenic aromatic nitrenium ions for adduct formation with different DNA base sites have been predicted by computational studies.^49-51 All these theoretical studies point toward the specificity of the N-site of the nitrenium ion toward C-8 of purine bases as well as the specificity of the ortho-C-site of the nitrenium ion toward the adenine and guanine amino group.

In contrast to typical nitroaromatic compounds, only one type of adduct is commonly encountered with AAs in which C-9 is bound to either guanine or adenine amino group (7, 7a, 8, 8a). The apparently preferred formation of C-9-adducts by AAs, as opposed to N-6-adducts formed by 6-nitrochrysene, suggests that the AAI-carboxyl group plays a role in controlling the regiospecificity of the reaction and the pattern of adducts formed. Possible explanations are given below.

10. Tentative Explanation for the Formation of C-9 Adducts by Aristolochic Acids Based on the Oxazinone Pathway

There are two possible mechanisms that account for the regiospecificity. The reaction might occur via an S_N2’ mechanism in which the intermediate 11 undergoes a concerted addition-elimination reaction upon being attacked by the nucleophile (H₂O or CH₃CO₂H). Alternatively, 11 could generate the nitrenium ion 12 as an “ion pair” in which the ring carbon will be the more reactive site for nucleophilic attack due to the influence of the carboxylic acid group.

10.1 The S_N2′ Mechanism. Concerted Addition-elimination Without Formation of a Nitrenium Ion

In this mechanism, there is direct attack of acetic acid, or other nucleophile, on C-9 of the oxazinone 11 (see arrows in structure 11) to give the imine 13. Subsequently, tautomerization and condensation between the substituents at C-9 and C-10 would afford aristoxazol (15). This pathway would seem to be the less likely of the two on the basis of studies with small model molecules, which indicate that solvolysis of N-O bond cleavage occurs through an ionic pathway (S_N1′ mechanism) involving a nitrenium ion.⁵²

10.2 Regiospecicific Formation of the C-9 Adducts via a Nitrenium Ion-pair in Which the Nitrogen is Blocked from Attack

“Ion pairs” or “electrostatic complexes” are implicated in the formation of nitrenium ions by carcinogens.^43,52,53 The nitrenium ions from AAs have the intriguing possibility of forming a “tight ion-pair” by intramolecular interaction of the cation with the carboxylate ion. A “tight ion-pair” consists of a cation and an anion held strongly together by electrostatic attraction, and further stabilized by hydrogen-bond interaction between the ions, which prevent the solvent from forming solvent-separated ions.^43,52 Consequently, tight ion pairs are insensitive to the presence of trapping agents (H₂O, Cl⁻, reducing agents) in the medium. The nitrenium ion 12 would be an ion pair in which the cation and the anion (the carboxylate group) occur in the same molecule. The regiospecificity of DNA base attack at C-9 may be due to the occurrence of this intramolecular ion pair that fixes the anion at one side of the reacting centers. The ion pair prevents the participation of the nitrenium nitrogen atom in reactions with nucleophiles, leaving the electron-deficient C-9 position as the only site for nucleophilic attack. In summary, formation of an intramolecular nitrenium ion pair could account for the generation of the dominant C-9-adducts with DNA. In contrast, other nitroarenes lacking the carboxyl group can generate adducts at both the nitrogen and ortho carbon. Although computational studies are needed to test this hypothesis, the proposed oxazinone 11 intermediate provides a plausible explanation not only for the unusual pattern of adducts from AAs, but also for the activation processes needed to generate nitrenium ions.

11. Are N-hydroxyaristolactams Toxic?

N-hydroxyaristolactams such as 3 and 3a seem to be implicated in nitrenium ion formation and toxicity, but there is evidence that questions the suspected toxicity of N-hydroxyaristolactams. N-hydroxyaristolactams have not been documented in the Aristolochiaceae, however N-methoxy and N-hydroxyaristolactams are stable compounds that have been isolated from Piper spp. (Piperaceae).^54-56 Of particular interest is P. umbellatum, a medicinal plant widely used in Cameroon traditional medicine.⁵⁶ It contains significant amounts of N-hydroxyaristolactams (ca. 100 mg per kg of air-dried branches).⁵⁶ Leaves, stems, inflorescences and fruits of P. umbellatum are also used as food in many parts of Africa.⁵⁷ There have been no reports that this plant is toxic. Since the N-hydroxyaristolactams (3 and 3a) are the immediate precursors of the nitrenium ions (4a and 4b), they are expected be as toxic as the AAs themselves since they can be activated in vivo by sulphotransferases and acyltransferases. The lack of toxicity of N-hydroxyaristolactam containing plants and a number of additional questions concerning AAs-activation and toxicity remain unanswered.

The major reduction product of AAI in both chemical and biological reductions is the aristolactam I (5). The AA-DNA-adducts (7, 7a, 8, 8a) and the 9-hydroxyaristolactam I (6) are byproducts from reactive intermediates generated during reduction of AAI. The new reduction product, aristoxazol (15), may also belong to the same family of adducts. The formation of aristoxazol probably occurs via a nitrenium ion pathway in which an acetic acid molecule attacks at C-9 of the nitrenium ion-pair 12. Although the formation of aristoxazol may proceed via the N-hydroxyaristolactam 3 (Scheme 1), it seems more likely that the compound is formed from the oxazinone 11 (Scheme 2). The oxazinone 11, which can be considered an activated form of the N-hydroxylamine, could also provide a rational explanation for adduct formation in in vitro systems that lack other means to activate the N-hydroxylamine. In addition to the aristolochic acid-DNA-adducts generated via the N-hydroxyaristolactam pathway (Scheme 1), the oxazinone 11 and the nitrenium ion-pair 12 could also serve as important electrophiles involved in DNA-adduct formation in vivo.

Experimental Section

General Experimental Procedures

UV spectra were obtained by an Agilent 8453 UV-vis spectrophotometer. IR spectra were recorded on a Perkin Elmer Spectrum 2000 with KBr pellets. NMR spectra were recorded on a Bruker 400 MHz FT-NMR Spectrometer in DMSO with TMS as internal standard. ESI and APCI mass spectra were acquired on a Thermo Scientific LCQ Deca XP MAX instrument and Thermo Scientific DSQ Mass Spectrometer, respectively. HPLC analysis was carried out with a Thermo-Finnigan chromatograph (Thermo Electron Corporation, San Jose, California). The chromatograph consisted of a SpectraSystem SMC1000 solvent delivery system, vacuum membrane degasser, P4000 gradient pumps and AS3000 autosampler. Column effluent was monitored at 254 nm with a SpectraSystem UV6000LP variable wavelength PDA detector and ChromQuest 4.1 software. Analytical separations were performed using a C18 RP Hypersil GOLD column (RP5, 250 × 4.6 mm, pore size 5 μm, Thermo Electron Corporation). HPLC solvents were employed without further purification. They were filtered through a 0.22 μm Millipore membrane. The water used was deionized and filtered through a nylon membrane of 0.45 μm. The following eluting systems were used: System 1, A. 0.1 % TFA in MeCN, B. 0.1 % TFA in H₂O, linear gradient 10 to 100 % A in 120 min; System 2, A. MeCN, B. 0.1 M NH₄OAc buffer pH 7.5, linear gradient 10 to 100 % A in 120 min; System 3, A. 0.1 % TFA in MeCN, B. 0.1 % TFA in H₂O; linear gradient 10 to 100 % A in 30 min. Flow rate 1.0 mL/min at room temperature. In systems 1, 2 and 3 aristolactam I showed Rt 48.4, 47.6 and 19.0 min., respectively.

Molecular mechanics calculations were carried out using Spartan ’06 for windows. (Wavefunction Inc.). The calculations were done at the B3LYP 6311 + G ** level.

Aristolochic acid I was obtained from Aristolochia argentina as previously described and purified by recrystallization from dioxane.¹³

9-Methoxy-7-methyl-2H-1,3-oxazolo[5′,4′-10,9]phenanthro[3,4-d]1,3-dioxolane-5-carboxylic acid. Aristoxazol (15)

Aristolochic acid I (1) (15.3 mg) was refluxed for 50 minutes with zinc powder (70 mg) and glacial HOAc (1 mL) with magnetic stirring. The reaction mixture was treated with H₂O (5 mL) and EtOAc (5 mL), shaken and centrifuged. The upper phase was removed, washed with H₂O and extracted with aqueous 5 % NaHCO₃. Evaporation to dryness of the EtOAc phase gave a yellow residue of aristolactam I (5) (9.1 mg). The NaHCO₃ solution containing aristoxazol was acidified to pH 3 with dil. HCl and extracted with EtOAc. Removal of the solvent from the organic phase under vacuum yielded a white residue of aristoxazol (3.3 mg). Colorless needles (isoPrOH), HPLC, Rt 46.60 min (system 1), 24.1 min (system 2), 18.2 min (system 3); UV/PDA λ_max: 251, 300, 326, 363, 383 nm; UV (H₂O-NaOH) λ_max: 253,0, 299.5, 325.1, 363.8, 382.0 nm; IR (KBr) ν_max 1668, 1587, 1459, 1310, 1104, 1018 cm^-1; APCI/MS: m/z 380.1101 [M+C₂H₅]⁺ (calcd for C₂₁H₁₈NO₆, 380.1134), 352.0824 [M+H]⁺ (calcd for C₁₉H₁₄NO₆, 352.0821), 334 [M+H-H₂O]⁺, 308 [M+H- CO₂]⁺; ESI/MS (MeOH) positive mode: 352.0 [M+H]⁺, 406.0 [M+MeOH+Na]⁺, 703.0 [2M+H]⁺, 725.0 [2M+Na]⁺. EIMS m/z (rel. int.): 350.9 [M]⁺ (98), 306.9 [M-CO₂]⁺ (66), 291.9 [M-CO₂- CH₃]⁺ (100), 263.9 [M-CO₂-CH₃-CO]⁺ (15).

Aristoxazol methyl ester

Aristoxazol was treated with ethereal diazomethane as usual to give aristoxazol methyl ester, Rt 57.3 min. (system 1); UV/PDA λ_max 254, 300, 324, 363, 384 nm; APCI/MS, m/z 394.1298 [M+C₂H₅]⁺ (calcd for C₂₂H₂₀NO₆, 394.1291), 366.0952 [M+H]⁺ (calcd for C₂₀H₁₆NO₆, 366.0978), 365.0872 [M]⁺ (calcd for C₂₀H₁₅NO₆, 365.0899), 350.0864 [M-MeO]^+..

Reduction at lower temperatures

Aristoxazol can also be obtained by heating the AAI/Zn/HOAc mixture at lower temperatures. HPLC analyses of reaction mixtures and yield of products (5 and 15) revealed that the ratio of aristolactam I to aristoxazol on heating at 60 °C, 90 °C and at reflux (118 °C) ranged from 2.2-3 to 1. In contrast the ratio of aristolactam I to aristoxazol was 6:1 when the reaction was carried out at room temperature (25°C).

Formulas