Regiospecificity of Peroxyl Radical Addition to (E)-Retinoic Acid

Victor M. Samokyszyn*, Mary Ann Freyaldenhoven, Hebron C. Chang, ... and of Biopharmaceutical Sciences, University of Arkansas for Medical Sciences, L...
0 downloads 0 Views 327KB Size
Download PDF
Chem. Res. Toxicol. 1997, 10, 795-801

795

Regiospecificity of Peroxyl Radical Addition to (E)-Retinoic Acid Victor M. Samokyszyn,*,§ Mary Ann Freyaldenhoven,§ Hebron C. Chang,§ James P. Freeman,† and R. Lilia Compadre‡,§ Departments of Pharmacology & Toxicology (Division of Toxicology) and of Biopharmaceutical Sciences, University of Arkansas for Medical Sciences, Little Rock, Arkansas 72205, and Mass Spectrometry Branch, Chemistry Division, National Center for Toxicological Research, Jefferson, Arkansas 72079 Received March 20, 1997X

The regiochemistry of peroxyl radical addition to (E)-retinoic acid (RA) was investigated. Peroxyl radicals, generated by reaction of 13-hydroperoxy-(9Z,11E)-octadecadienoic acid with hydroxo(porphyrinato)iron(III) in Tween 20 micelles, were reacted with RA. The major, and virtually exclusive, RA oxidation product was 5,6-epoxy-RA which was identified on the basis of cochromatography with the synthetic oxirane (in a reverse phase HPLC system), electronic absorption spectroscopy, high-field 1H-NMR, and EI mass spectrometry. These results suggest that peroxyl radicals react with RA by regioselective addition to either C5 or C6 yielding an endocyclic tertiary allylic or tertiary carbon-centered radical adduct, respectively. Subsequent β-elimination of an alkoxyl radical yields the oxirane. Computational studies were carried out in order to gain mechanistic insights into the observed regiospecificity of the peroxyl radicaldependent epoxidation reaction; molecular mechanics and semiempirical quantum mechanical calculations were carried out using Tripos force field parameters and AM1, respectively. The results suggest that the regiospecific epoxidation may be influenced by the 5,6-olefinic function behaving as a partially-isolated double bond as well as inherent allylic A1,2 strain in the substituted cyclohexene ring as a consequence of substitutions at C1 and C6. In addition, calculated heats of formation indicated preferential peroxyl radical addition to C5 versus C6; this may reflect differences in the geometries of sp2-orbitals containing the radical densities rather than resonance contributions by the highly conjugated polyene system.

Introduction

Chart 1. Structures of RA and 5,6-Epoxy-RA

1

(E)-Retinoic acid (RA) (Chart 1) and (13Z)-retinoic acid ((13Z)-RA) represent major in vivo and in vitro metabolites of retinol (vitamin A), and metabolism of retinol to RA (or a metabolite of RA) is required for both its pharmacological activity in skin and its teratological effects (1, 2). RA is used topically in the treatment of acne and photodamaged skin (3, 4). 5,6-Epoxy-RA (Chart 1) has been identified as a major in vivo metabolite of RA (as well as retinol and retinyl acetate) in rodents (2, 5-7), we have recently identified the oxirane as a major in vivo metabolite of (13Z)-RA in mouse skin,2 and we have detected the glucuronide of 5,6-epoxy-RA in human bile.2 5,6-Epoxy-RA also exhibits potent biological activity in various biological assays. For example, the oxirane was more effective than retinyl acetate at exhibiting potent growth effects in vitamin A-deficient rats (8). In * Author to whom correspondence should be addressed at Division of Toxicology, University of Arkansas for Medical Sciences (Slot 638), 4301 W. Markham, Little Rock, AR 72205. Phone: (501) 686-5766. Fax: (501) 686-8970. E-mail: [email protected]. § Department of Pharmacology & Toxicology. ‡ Department of Biopharmaceutical Sciences. † NCTR. X Abstract published in Advance ACS Abstracts, July 1, 1997. 1 Abbreviations: RA, (E)-all-trans-retinoic acid (3,7-dimethyl-9(2,6,6-trimethyl-1-cyclohexen-1-yl)-2,4,6,8-nonatetraenoic acid); (13Z)RA, (13Z)-retinoic acid; TPA, 12-O-tetradecanoylphorbol 13-acetate; MM, molecular mechanics; MO, molecular orbital; EI, electron ionization (mass spectrometry, MS); 13-OOH-18:2, 13-hydroperoxy-(9Z,11E)octadecadienoic acid; 13-OH-18:2, 13-hydroxy-(9Z,11E)-octadecadienoic acid; hematin, hydroxo(porphyrinato)iron(III); rms, root-mean-square; HOMO, highest occupied molecular orbital; UHF, unrestricted Hartree-Fock. 2 V. M. Samokyszyn et al., unpublished observations.

S0893-228x(97)00045-3 CCC: $14.00

addition, the oxirane exhibits similar inhibitory effects as RA on 12-O-tetradecanoylphorbol 13-acetate (TPA)dependent tumor promotion in the two-stage (initiationpromotion) mouse skin carcinogenesis assay (9). Interestingly, the oxirane is actually more potent than RA in opposing the effects of TPA on induction of tumor promotional markers in bovine lymphocytes (10). Similarly, Kensler and Trush have demonstrated that the oxirane is more effective than RA in the inhibition of TPA-stimulated chemiluminescence (an index of both the generation and reactions mediated by superoxide and singlet oxygen) in human polymorphonuclear leukocytes (11). Evidence from the latter two investigations suggests that the retinoid response may actually involve the metabolic activation of RA to the 5,6-epoxide with the oxirane representing the pharmacologically-active agent. Epoxidation of olefins in vivo occurs by either cytochrome P450-catalyzed mechanisms or peroxyl radical© 1997 American Chemical Society

796 Chem. Res. Toxicol., Vol. 10, No. 7, 1997

dependent mechanisms. Sietsema and DeLuca have investigated RA epoxidation in vitro in rat tissue homogenates and reported indirect evidence of peroxyl radical involvement (12, 13). The investigators demonstrated that RA epoxidation was NADPH-dependent, unaffected by cytochrome P450 inhibitors, stimulated by exogenously-added iron, and inhibited by the antioxidant N,N ′-diphenyl-p-phenylenediamine which suggests the involvement of NADPH cytochrome P450 reductasedependent, iron-catalyzed initiation of lipid peroxidation. We have reported the epoxidation of RA and (13Z)-RA to the corresponding 5,6-oxiranes during hydroperoxideor arachidonic acid-dependent cooxidation by microsomal and purified prostaglandin H synthase (14-16). We reported evidence for a mechanism involving retinoid (C4)-derived peroxyl radical intermediates. In addition, we have demonstrated peroxyl radical-dependent oxidation of RA2 and (13Z)-RA (17, 18) to the corresponding 5,6-epoxides during 2,2′-azobis(2-amidinopropane)-initiated linoleic acid autoxidation in SDS micelles and during microsomal lipid peroxidation, respectively. However, several other retinoid oxidation products were also detected, but the nature of these structures was not determined; this was primarily because of limited mass recoveries and the absence of synthetic HPLC standards. Similarly, Kennedy and Liebler have demonstrated peroxyl radical-dependent epoxidation of β-carotene across the 5,6- and 5′,6′-double bonds (19). Collectively, these results suggest that peroxyl radical-dependent oxidation of RA and (13Z)-RA may represent a major oxidative pathway in vivo giving rise to a biologically-active metabolite, the 5,6-epoxide. However, whether peroxyl radicals may react with RA at other olefinic sites has not been investigated. RA, (13Z)-RA, and other retinoids are effective inhibitors of lipid peroxidation in liposomal and microsomal systems (17, 18, 20). In contrast, RA functions as a prooxidant in micellar2 and liquid phase systems (21). Because peroxyl radicals are involved in free radical propagation reactions characterizing hydrocarbon autoxidation mechanisms, we have undertaken an investigation of the fate of RA in a peroxyl radical-generating system. Specifically, we have investigated the regiochemistry of the reaction of RA with chemically-generated peroxyl radicals. We report that peroxyl radicals react regiospecifically resulting in epoxidation across the 5,6-double bond. Molecular mechanics (MM) and semiempirical molecular orbital (MO) calculations suggest that the regiospecificity is not dictated by vinylic electron densities. Instead, the 5,6-double bond, unlike the remainder of the polyene, may function as a partiallyisolated olefin, thus conferring regiospecificity. In addition, the regiospecificity may be influenced by the inherent A1,2 steric strain in the C6-substituted 1,1,5-trimethylcyclohex-5-enyl ring. Our results also suggest that peroxyl radical addition to C5 is thermochemically preferred over addition to C6 which may reflect slight differences in geometries of sp2-orbitals containing the unpaired electron densities.

Materials and Methods Chemicals. RA and phthalic anhydride were purchased from Acros Organics (Pittsburgh, PA). RA was stored at -80 °C under argon, and stock solutions were prepared fresh in Me2SO purchased from Aldrich (Milwaukee, WI). All experiments were carried out in the dark or under yellow light to avoid photochemical isomerization of the retinoid. Hematin (hydroxo-

Samokyszyn et al. (porphyrinato)iron(III)), Tween 20, linoleic acid, sodium borohydride, 1-methyl-3-nitro-1-nitrosoguanidine, and Diazald (Nmethyl-N-nitroso-p-toluenesulfonamide) were also purchased from Aldrich. 13-Hydroperoxy-(9Z,11E)-octadecadienoic acid (13-OOH-18:2) was prepared from linoleic acid employing soybean lipoxidase (Sigma, St. Louis, MO) (22). The corresponding alcohol (13-hydroxy-(9Z,11E)-octadecadienoic acid, 13OH-18:2) was synthesized by sodium borohydride reduction of the hydroperoxide. Monoperphthalic acid was prepared as described elsewhere (23). 5,6-Epoxy-RA was synthesized by reaction of RA with monoperphthalic acid as described by Wertz et al. (10): EI-MS m/z 330 (M•+), 315 (M•+ - CH3), 299 (M•+ OCH3), 271 (M•+ - CO2CH3); 1H-NMR (300 MHz, Me2SO-d6) δ 6.97 (dd, J ) 15.0, 15.3 Hz, 1H), 6.42 (d, J ) 15.1 Hz, 1H), 6.25 (d, J ) 6.5, 11.1 Hz, 2H), 6.04 (d, J ) 15.7 Hz, 1H), 5.78 (s, 1H), 2.26 (s, 3H), 1.95 (s, 3H), 1.73 (brs, 2H), 1.37 (brs, 2H), 1.09 (s, 3H), 1.08 (s, 3H), 0.85 (s, 3H). All other chemicals were obtained through commercial sources. Analysis of Peroxyl Radical Oxidation Product(s) of RA. RA (1.5 × 10-4 M) was mixed with hematin (5.0 × 10-7 M) in 0.1 M phosphate buffer (pH 7.8) containing 2.0 × 10-4 M Tween 20 for 1 min at 25 °C followed by the addition of 13OOH-18:2 or 13-OH-18:2 (2.0 × 10-4 M). After 15 min, samples were saturated with NaCl and extracted with HPLC-grade ethyl acetate. The ethyl acetate extracts were pooled and dried over anhydrous Na2SO4, the solvent was removed in vacuo, and the residue was dissolved in HPLC-grade methanol and filtered through 0.45 µm Nylon-66 filters (Scientific Resources, Inc., Eatontown, NJ). Analytical reverse phase HPLC was carried out using a Waters Novapak C18 (125 Å, 4 µm, 3.9 × 300 mm) HPLC column and a Waters HPLC system consisting of a Waters 600E pumping system and Waters 994M photodiode array detector run on Waters Millennium software, version 2.1, using a Gateway 2000 P5-90 computer. The solvent system consisted of 20% solvent B in solvent A at t ) 0 followed by a linear gradient to 90% solvent B over 30 min followed by isochratic conditions for 25 min. Solvent A contained 0.01 M ammonium acetate in 50:50 methanol/water (pH 6.65 ( 0.05), and solvent B contained 0.01 M ammonium acetate in 90:10 methanol/water (pH 6.65 ( 0.05). The flow rate was 1.0 mL/ min, and eluents were detected by absorbance at 330 nm. The major oxidation product, for mass spectrometric and 1H-NMR spectroscopic analysis, was obtained from reaction mixtures which were scaled up 100-fold. The product was purified by semipreparative reverse phase HPLC using a Waters µBondapak C18 (125 Å, 10 µm, 7.8 × 300 mm) HPLC column. The major oxidation product, 5,6-epoxy-RA, was identified on the basis of cochromatography with synthetic standard, UV spectroscopy, 1H-NMR, and mass spectrometry. Mass Spectrometry. The methyl ester of the peroxyl radical-dependent oxidation product was prepared by reaction of the purified material with ethereal diazomethane, generated by reaction of 1-methyl-3-nitro-1-nitrosoguanidine with KOH at 4 °C or by using a large-scale diazomethane-generating system utilizing Diazald (24). Caution: Diazomethane is a potent electrophile and is carcinogenic. In addition, diazomethane is potentially explosive. Thus, all procedures should be carried out in a fume hood, behind protective shielding, and a lab coat, gloves, and goggles should be worn. In addition, sharp jagged edges should be avoided. The sample was analyzed by electron ionization mass spectrometry (EI-MS) at 70 eV energy from a direct exposure probe. The mass spectrometer was a model 4000 upgraded to 4500 capabilities (Finnigan MAT, San Jose, CA). The quadrupole analyzer was scanned from 45 to 645 Da in 1 s. The ion source was set to 150 °C (uncorrected). The probe consisted of a Finnigan direct exposure probe with a rhenium wire tip and a Finnigan MAT current controller. The sample was applied to the rhenium wire tip via a methanol solution, and the solvent was allowed to partially evaporate in a hood before insertion into the probe vacuum lock for final solvent evaporation before

Retinoic Acid Epoxidation

Chem. Res. Toxicol., Vol. 10, No. 7, 1997 797

Scheme 1

MS analysis. The probe current was ramped linearly from 0 to 650 mA at 5 mA‚s-1 as the data were collected. NMR Spectroscopy. 1H-NMR spectra of the isolated, diazomethane-derivatized product were recorded at 500.13 MHz and at approximately 28 °C on a Bruker AM-500 spectrometer (Bruker Instruments, Billerica, MA). Preliminary 13C-NMR measurements were also carried out. The sample was dissolved in “100 atom %” methanol-d4. Chemical shifts were reported in ppm downfield from TMS by assigning the residual proton signal of the solvent to 3.30 ppm. Measurements were firstorder, and the data point resolution was 0.215 Hz/point after zero-filling. Molecular Modeling. Molecular models for the ground states were built and optimized using SYBYL force field parameters. The resulting geometries were subject to full energy optimization using restricted Hartree-Fock (RHF) SCFMO as implemented in AM1 (25) from MOPAC 6.0 (QCPE). The geometries of the C5 and C6 radicals were optimized using unrestricted Hartree-Fock (UHF) calculations. All computations were performed with the PRECISE key word. Comparison of our minimized RA structure with published X-ray crystallographic coordinates was achieved using Chem 3D Plus (Cambridge Scientific, Cambridge).

Results and Discussion To investigate the regiospecificity of peroxyl radicaldependent oxidation of RA, we have used a peroxyl radical-generating system, reported by Dix and Marnett (26, 27), involving the reaction of 13-hydroperoxy-(9Z,11E)-octadecadienoic acid (13-OOH-18:2) with hydroxo(porphyrinato)iron(III) (hematin) in Tween 20 micelles at ambient dioxygen tension. The mechanism involves the hematin-dependent homolytic reduction of 13-OOH18:2 (in the presence of dioxygen), resulting in peroxyl radical generation (Scheme 1). One-electron reduction of the hydroperoxide by hematin yields a hydroperoxidederived alkoxyl radical that cyclizes yielding epoxyallylic carbon-centered radicals. These carbon radicals can diffuse from the solvent cage and couple with dioxygen yielding epoxyallylic peroxyl radicals. This method gives high yields of peroxyl radicals. Reverse phase HPLC analysis of RA oxidation products, resulting from the reaction of RA with peroxyl radicals generated in the 13-OOH-18:2/hematin system, demonstrated the formation of 5,6-epoxy-RA as the major oxidation product (Figure 1, top). In fact, other than the oxirane, virtually no other products were detected. The RA oxidation product exhibited an identical retention time (tR ) 31.9 min in our reverse phase HPLC system) as synthetic 5,6-epoxy-RA, and both exhibited identical electronic absorption spectra (λmax ) 330 nm) (Figure 2, top). In addition, the methyl esters of both the isolated oxidation product and synthetic oxirane exhibited identi-

Figure 1. Reverse phase HPLC profile of RA oxidation products generated by reaction of RA with 13-OOH-18:2-derived peroxyl radicals: (top) complete system (RA/hematin/13-OOH18:2); (bottom) same as above except 13-OOH-18:2 was substituted with the corresponding alcohol (13-OH-18:2). Reaction conditions and analytical methods are described in the Materials and Methods section.

cal 1H-NMR and EI mass spectra. The metabolite isolated from hematin/13-OOH-18:2/RA reaction mixtures was derivatized with ethereal diazomethane to prepare the corresponding methyl ester. Results from mass spectrometry analysis in the EI mode (Figure 2, bottom) demonstrated a molecular ion of m/z 330 which corresponds to the oxirane methyl ester. In addition, the fragmentation pattern is consistent with the structure of methyl 5,6-epoxyretinoate which we and others have previously reported (5, 7, 16). For example, the ions observed at m/z 315 (M•+ - CH3), 299 (M•+ - OCH3), and 271 (M•+ - CO2CH3) are indicative of a methylated retinoic acid derivative undergoing loss of C15 terminal carboxymethyl fragments. The absence of m/z 312 (M•+ - H2O) confirms that the additional oxygen atom is not present as a hydroxyl group. The identification of the oxirane (methyl ester) was further confirmed by high-field 1H-NMR in methanol-d4 (Table 1). Resonance assignments were determined from homonuclear decoupling, NOE measurements, integration, and heteronuclear chemical shift correlation experiments. The data are consistent with formation of a 5,6oxirane with s-trans orientation of the side chain. NMR

798 Chem. Res. Toxicol., Vol. 10, No. 7, 1997

Figure 2. Electronic absorption spectrum of the major RA oxidation product (tR ) 31.9 min) (top) and EI mass spectrum of the corresponding methyl ester (bottom). Table 1. 1H-NMR Assignments and Spectral Parameters of 5,6-Epoxyretinoic Acid Methyl Estera assignment

chemical shift

2 3 4 7 8 10 11 12 14 16c 17c 18 19 20 methoxy

1.11, 1.43b 1.43b 1.83b 6.07 6.27 6.21 7.06 6.39 5.82 1.12d 0.92 1.13d 1.97 2.32 3.68

a The HPLC-purified RA oxidation product was dissolved in methanol-d4. Chemical shifts are in ppm. Coupling constants (in Hz) are as follows: J7-8 ) 15.7, J8-10 ) 0.4, J10-11 ) 11.3, J10-19 ) 1.3, J11-12 ) 14.9, and J14-20 ) 1.3. Numerous unresolved longrange coupling constants were apparent. b Center of multiplet.c No stereochemical distinction between methyl groups is implied. d May be reversed.

data of RA methyl ester and several geometric isomers have previously been reported (ref 28 and references cited within). Resonance assignments for the side chain of 5,6epoxy-RA methyl ester are in agreement with the prior work on RA methyl ester except for the chemical shifts of H7 and H8. The chemical shifts are within e0.3 ppm of those previously reported for 5,6-epoxy-RA (i.e., free

Samokyszyn et al.

acid) in methanol-d4 (29), with the greatest deviation being of H11 and H13. Possibly, the state of ionization of the carboxyl function affects the conformation of the side chain. This in turn would be expected to affect the chemical shifts. For the ester reported here, we have added the results for all methylene protons and reversed resonance assignments for H7 and H8 as well as for two of the ring methyl groups (see Table 1). In contrast to the 13-OOH-18:2/hematin system, substitution of 13-OOH-18:2 with the corresponding alcohol (13-OH-18:2) virtually abolished any detectable RA oxidation products (Figure 1, bottom). Our results differ from those of Iwahashi et al. (29) who demonstrated the epoxidation of RA by hematin alone. We have previously noted that great care must be taken during handling and storage of retinoids; otherwise, the retinoids autoxidize yielding retinoid-derived hydroperoxides.2 Thus, the results of Iwahashi et al. may reflect trace amounts of hydroperoxides in their system. It should also be noted that 4-hydroxy-RA (tR ) 25.3 min) was tentatively detected in the complete system as a very minor component (Figure 1, top). This is consistent with the results of Mayo et al. demonstrating that the reaction of peroxyl radicals with conjugated allylic functions occurs preferentially by addition rather than H atom abstraction mechanisms (30). Collectively, these results indicate that peroxyl radicaldependent oxidation of RA is regiospecific resulting in epoxidation across the 5,6-double bond. Peroxyl radicals have been shown to be π-radicals with large dipole moments in the 2.3-2.6 D range (31). We investigated whether the regiospecificity of peroxyl radical addition to RA is dictated by the electron densities at the RA vinylic carbons, with the peroxyl radical reacting with the vinylic carbon exhibiting the highest electron density. Molecular mechanics (MM), using Tripos force field parameters, and semiempirical molecular orbital (MO) calculations, using the AM1 Hamiltonian, were carried out on RA as described in the experimental section. The minimized structure, obtained by MM and optimized using the AM1 algorithm, was nearly superimposable with the published X-ray crystallographic structure (32). This is reflected in an overlay of our calculated-minimized coordinates with the X-ray crystallographic coordinates which indicated root-mean-square (rms) deviations in the heavy atom positions of less than 0.07 Å. The relative electron densities at the RA carbons were quantitated by calculation of partial atomic charges (Table 2). Our results indicate that the electron densities at C5 and C6 are significantly lower compared to other vinylic carbons (e.g., compare C5 with C14). These results suggest that electron density does not play an apparent role in the regiospecificity of peroxyl radical addition to RA. Alternatively, preferential epoxidation of the 5,6-double bond may occur if this position exhibits significantly impaired p-orbital overlap with the polyene, imposed by geometric constraints, and thus functions as a partiallyisolated double bond. This is supported by our molecular mechanics calculations (and optimized using the AM1 algorithm) which demonstrated that the C5-C6-C7C8 torsional angle was -49.3°, whereas most double bonds in the remainder of the polyene are completely staggered (Table 3). This preferred geometry is probably a result of the regional substitution pattern, particularly dimethyl substitution at C1 and polyene substitution at C6. Thus, epoxidation across the 5,6-position would be energetically more favorable because epoxidation else-

Retinoic Acid Epoxidation

Chem. Res. Toxicol., Vol. 10, No. 7, 1997 799

Table 2. Partial Atomic Charges of Carbon and Oxygen Molecular Orbitals in RAa atom

partial charges

C1 C2 C3 C4 C5 C6 C7 C8 C9 C10 C11 C12 C13 C14 C15 C16 C17 C18 C19 C20 O1 O2

-0.01 -0.15 -0.16 -0.13 -0.08 -0.09 -0.10 -0.15 -0.04 -0.16 -0.08 -0.17 0.04 -0.22 -0.35 -0.20 -0.20 -0.19 -0.19 -0.22 -0.38 0.32

Scheme 2

Table 4. Heats of Formation of RA C5- versus C6-Methylperoxyl Adductsa

a Partial charges were calculated using the semiempirical quantum mechanics program AM1, employing the minimized structure of RA.

Table 3. Torsional Bond Angles in the Minimized Structure of RAa bonds

torsional angle (deg)

C1-C2-C3-C4 C2-C3-C4-C5 C3-C4-C5-C6 C4-C5-C6-C1 C5-C6-C1-C2 C6-C1-C2-C3 C16-C1-C2-C3 C17-C1-C6-C5 C2-C1-C6-C7 C4-C5-C6-C7 C1-C6-C7-C8 C5-C6-C7-C8 C6-C7-C8-C9 C7-C8-C9-C10 C8-C9-C10-C11 C9-C10-C11-C12 C10-C11-C12-C13 C11-C12-C13-C14 C12-C13-C14-C15

-60.7 47.9 -18.9 -0.1 -10.7 41.1 162.5 110.6 169.6 179.6 130.4 -49.3 -178.1 162.1 -178.6 -176.3 -179.8 -179.2 -0.1

a The minimized structure of RA was obtained via MM calculations and optimized using the AM1 algorithm.

where would result in disruption of p-orbital overlap in a highly-conjugated polyene system. Furthermore, an additional exothermic contribution promoting epoxidation across the 5,6-position may arise by diminishing the inherent A1,2 allylic strain in the C6-polyene-substituted 1,1,5-trimethylcyclohex-5-enyl ring, arising as a consequence of the substitutions at C1 and C6 (33). However, the extent of this thermochemical contribution is unclear, and we are attempting to obtain quantitative insights based on ab initio calculations involving heavy atomlimited cyclic and acyclic analogs of the substituted cyclohexenyl structure. These results are consistent with peroxyl radicaldependent epoxidation of RA to 5,6-epoxy-RA occurring by peroxyl radical addition to C5 or C6 followed by β-scission of the peroxide moiety (with elimination of an alkoxyl radical) and ring closure yielding the oxirane (Scheme 2). The peroxyl radicals are expected to add preferentially to C5 because this position is less sterically

a Computational studies were carried out as described in the experimental section utilizing unrestricted Hartree-Fock calculations (UHF).

hindered compared with the C6 position. To investigate potential thermochemical differences in reactivity at the C5 versus C6 endocyclic vinylic carbons, we carried out MM and MO calculations on C5- versus C6-peroxyl adducts using methylperoxyl as a molecular probe (R′ ) CH3 in Scheme 2). Calculated heats of formation indicate that the C5 adduct (with unpaired electron density at C6) is 3.5 kcal‚mol-1 lower compared with the C6 adduct (with unpaired electron density on C5) (Table 4). These results also indirectly suggest that addition to C5 may be kinetically preferred because heats of formation, characterizing free radical formation, are proportional to the enthalpy of the transition states. To determine if this ∆∆Hf difference may be a result of resonance contributions from the polyene, reflecting a more stable free radical in the C5 adduct, we plotted the relative HOMO distribution in the minimized structure of the C5-peroxyl radical adduct and the C6-peroxyl radical adduct (Figure 3). Our results suggest that the C6 radical orbital (in the C5-peroxyl radical adduct) is isolated from the polyene because the radical-containing orbital appears to be orthogonal with respect to the p-orbital at C7, negating any p-orbital overlap. The HOMO distribution involving C7-C8, C7-proton, and C8-proton bonding is composed of σ (rather that π) orbitals based on symmetry considerations (Figure 3, top). Thus, the carbon-centered radical at C6, in the minimized C5peroxyl radical adduct, does not appear to be resonance stabilized via contributions by the polyene π system. Therefore, what accounts for the thermochemical preference of peroxyl radical addition to C5 versus C6? One possibility is that the regiospecificity may reflect stereochemical factors which affect the geometry of the sp2carbon center containing the unpaired electron density.

800 Chem. Res. Toxicol., Vol. 10, No. 7, 1997

Samokyszyn et al.

Figure 3. Distribution of HOMOs characterizing the C5-methylperoxyl adduct (top) and C6-methylperoxyl adduct (bottom) of RA.

Our calculations indicate that the C6 radical contains bond angles which are optimal for a trigonal planar structure (C1-C6-C5 ) 119.74°, C1-C6-C7 ) 119.70°, C5-C6-C7 ) 120.35), whereas the C5 radical exhibits slightly less-than-optimal geometry characterized by bond angles for C4-C5-C6, C4-C5-C18, and C6-C5C18 of 117.05°, 120.52°, and 122.39°, respectively. These geometric differences may ultimately contribute to the thermodynamic preference for addition to C5. In summary, reaction of peroxyl radicals with RA involves the regiospecific epoxidation at the 5,6-olefinic position. This mechanism may involve preferential peroxyl addition to the C5 position. Peroxyl radical-dependent regiospecific epoxidation across the 5,6-position was confirmed by reverse phase HPLC analysis of RA oxidation products generated in an established peroxyl radicalgenerating system (Figure 1, Scheme 1). The identity of the oxirane was confirmed by cochromatography with a synthetic standard, electronic absorption spectroscopy, mass spectrometry, and 1H-NMR (Figure 2, Table 1). Computational studies suggest that the regiospecificity of the reaction is not conferred by vinylic electron densities (Table 2). Instead, this regiospecificity may result as a consequence of the 5,6-double bond, in the parent retinoid, functioning as a partially-isolated olefinic

function. The latter is evidenced from the analysis of the torsional angle characterizing C5-C6-C7-C8 in the minimized structure which, as described, exhibits significant deviation from the polyene geometries which are staggered (Table 3). In addition, regiospecificity may be conferred as a consequence of reducing the inherent A1,2 allylic strain that arises from 1,1-dimethyl substitution at C1 and polyene substitution at C6. Computational studies, employing methylperoxyl as a molecular probe, indicate that peroxyl radical addition to C5, versus C6, is thermodynamically favored (Table 4). This may involve differences in geometries of the carbon-centered radical sp2-orbitals. However, we cannot rule out the possibility of limited p-orbital overlap, involving the polyene substituent and the C6 carbon radical orbital (in the liquid phase at 298 K), which may arise as a consequence of rotational degrees of freedom. This cannot be reflected in vacuum calculations characterizing the minimized structure. Obviously, if the latter contributions are present they are minimal because we would otherwise expect significant distribution of unpaired electron density in the polyene and associated generation of other oxidation products (which were not observed), resulting from the reaction of polyene-derived carboncentered radicals with dioxygen. In terms of various in

Retinoic Acid Epoxidation

vitro and in vivo investigations, RA may represent an excellent and convenient peroxyl radical probe because RA is exclusively oxidized by peroxyl radicals to the 5,6oxirane. In contrast, the oxirane does not appear to be generated by mechanisms involving cytochrome P450 catalysis (12, 13).

Acknowledgment. We are grateful to Dr. Fred E. Evans (National Center for Toxicological Research, Jefferson, AR) for carrying out the NMR studies described in this manuscript. In addition, we are grateful to Dr. Tom R. Cundari (Department of Chemistry, University of Memphis, Memphis, TN) for the structural comparison of our minimized RA structure with the published X-ray crystallographic structure. This study was supported by a grant from the NIH (5 R29 ES06765-03). Dr. Freyaldenhoven is the recipient of the Kappa Epsilon-Nellie Wakeman-American Foundation for Pharmaceutical Education First-Year Graduate Scholarship.

References (1) Hill, D. L., and Sani, B. P. (1991) Metabolic disposition and development of new chemopreventive retinoids. Drug Metab. Rev. 23, 413-438. (2) Blaner, W. S., and Olson, J. A. (1994) Retinol and retinoic acid metabolism. In The Retinoids. Biology, Chemistry, and Medicine, 2nd ed. (Sporn, M. B., Roberts, A. B., and Goodman, D. S., Eds.) pp 229-255, Raven Press, New York. (3) Kelly, A. P. (1990) Acne and related disorders. In Principles and Practice of Dermatology (Sams, W. M., Jr., and Lynch, R. J., Eds.) pp 781-796, Churchill Livingstone, New York. (4) Weiss, J. S., Ellis, C. N., Headington, J. T., Tincoff, T., Hamilton, T. A., and Voorhees, J. J. (1988) Topical tretinoin improves photoaged skin. A double-blind vehicle-controlled study. J. Am. Med. Assoc. 259, 527-532. (5) McCormick, A. M., Napoli, J. L., Schnoes, H. K., and DeLuca, H. F. (1978) Isolation and identification of 5,6-epoxyretinoic acid: A biologically active metabolite of retinoic acid. Biochemistry 17, 4085-4090. (6) McCormick, A. M., Napoli, J. L., Yoshizawa, S., and DeLuca, H. F. (1980) 5,6-Epoxyretinoic acid is a physiological metabolite of retinoic acid in the rat. Biochem. J. 186, 475-481. (7) McCormick, A. M., and Napoli, J. L. (1982) Identification of 5,6epoxyretinoic acid as an endogenous retinol metabolite. J. Biol. Chem. 257, 1730-1735. (8) John, K. V., Lakshmanan, M. R., and Cama, H. R. (1967) Preparation, properties and metabolism of 5,6-monoepoxyretinoic acid. Biochem. J. 103, 539-543. (9) Verma, A. K., Slaga, T. J., Wertz, P. W., Mueller, G. C., and Boutwell, R. K. (1980) Inhibition of skin tumor promotion by retinoic acid and its metabolite 5,6-epoxyretinoic acid. Cancer Res. 40, 2367-2371. (10) Wertz, P. W., Kensler, T. W., Mueller, G. C., Verma, A. K., and Boutwell, R. K. (1979) 5,6-Epoxyretinoic acid opposes the effects of 12-O-tetradecanoylphorbol-13-acetate in bovine lymphocytes. Nature 277, 227-229. (11) Kensler, T. W., and Trush, M. A. (1981) Inhibition of phorbol esterstimulated chemiluminescence in human polymorphonuclear leukocytes by retinoic acid and 5,6-epoxyretinoic acid. Cancer Res. 41, 216-222. (12) Sietsema, W. K., and DeLuca, H. F. (1979) In vitro epoxidation of all-trans-retinoic acid in rat tissue homogenates. Biochem. Biophys. Res. Commun. 90, 1091-1097.

Chem. Res. Toxicol., Vol. 10, No. 7, 1997 801 (13) Sietsema, W. K., and DeLuca, H. F. (1982) Retinoic acid 5,6epoxidase. Properties and biological significance. J. Biol. Chem. 257, 4265-4270. (14) Samokyszyn, V. M., Sloane, B. F., Honn, K. V., and Marnett, L. J. (1984) Cooxidation of 13-cis-retinoic acid by prostaglandin H synthase. Biochem. Biophys. Res. Commun. 124, 430-436. (15) Samokyszyn, V. M., and Marnett, L. J. (1987) Hydroperoxidedependent cooxidation of 13-cis-retinoic acid by prostaglandin H synthase. J. Biol. Chem. 262, 14119-14133. (16) Samokyszyn, V. M., Chen, T., Maddipati, K. R., Franz, T. J., Lehman, P. A., and Lloyd, R. V. (1995) Free radical oxidation of (E)-retinoic acid by prostaglandin H synthase. Chem. Res. Toxicol. 8, 807-815. (17) Samokyszyn, V. M., and Marnett, L. J. (1990) Inhibition of microsomal lipid peroxidation by 13-cis-retinoic acid. Free Radical Biol. Med. 8, 491-496. (18) Samokyszyn, V. M., and Marnett, L. J. (1990) Inhibition of microsomal lipid peroxidation by 13-cis-retinoic acid. Methods Enzymol. 190, 281-288. (19) Kennedy, T. A., and Liebler, D. C. (1991) Peroxyl radical oxidation of β-carotene: Formation of β-carotene epoxides. Chem. Res. Toxicol. 4, 290-295. (20) Tesoriere, L., Ciaccio, M., Bongiorno, A., Riccio, A., Pintaudi, A. M., and Livrea, M. A. (1993) Antioxidant activity of all-transretinol in homogeneous solution and in phosphatidylcholine liposomes. Arch. Biochem. Biophys. 307, 217-223. (21) Finkelshtein, E. I., Rubashinskaya, Y. F., and Kozlov, Y. I. (1981) Kinetics and mechanism of the autoxidation of retinoic acid and its methyl ester. Zh. Org. Khim. 17, 936-943 (in Russian). (22) Funk, M. O., Isaac, R., and Porter, N. A. (1976) Preparation and purification of lipid hydroperoxides from arachidonic and γ-linolenic acids. Lipids 11, 113-117. (23) Vogel, A. I. (1972) Monoperphthalic acid. In A Textbook of Practical Organic Chemistry, 3rd ed., p 810, Longman Group Ltd., London. (24) Black, T. H. (1983) The preparation and reactions of diazomethane. Aldrichim. Acta 16, 3-10. (25) Dewar, M. J. S., Zoebisch, E. G., Healy, E. F., and Stewart, J. J. P. (1985) AM1: A new general purpose quantum mechanical molecular model. J. Am. Chem. Soc. 107, 3902-3909. (26) Dix, T. A., and Marnett, L. J. (1985) Conversion of linoleic acid hydroperoxide to hydroxy, keto, epoxyhydroxy, and trihydroxy fatty acids by hematin. J. Biol. Chem. 260, 5351-5357. (27) Dix, T. A., and Marnett, L. J. (1981) Free radical epoxidation of 7,8-dihydroxy-7,8-dihydrobenzo[a]pyrene by hematin and polyunsaturated fatty acid hydroperoxides. J. Am. Chem. Soc. 103, 6744-6746. (28) Curley, R. W., and Fowble, W. (1989) Assignment of 1H and 13C NMR spectra of retinoic acid ester isomers observed in the longwavelength UV-induced photostationary state. Magn. Reson. Chem. 27, 707-709. (29) Iwahashi, H., Ikeda, A., Negoro, Y., and Kido, R. (1986) Retinoic acid 5,6-epoxidation by hemoproteins. J. Biochem. (Tokyo) 99, 6371. (30) Mayo, F. R., Miller, A. A., and Russell, G. A. (1958) The oxidation of unsaturated compounds. IX. The effects of structure on the rates and products of oxidation of unsaturated compounds. J. Am. Chem. Soc. 80, 2500-2507. (31) Boyd, S. L., Boyd, R. J., and Barclay, L. R. C. (1990) A theoretical investigation of the structures and properties of peroxyl radicals. J. Am. Chem. Soc. 112, 5724-5730. (32) Stam, C. H. (1972) The crystal structure of a monoclinic modification and the refinement of a triclinic modification of vitamin A acid (retinoic acid), C20H28O2. Acta Crystallogr. B28, 2936-2945. (33) Johnson, F. (1968) Allylic strain in six-membered rings. Chem. Rev. 68, 375-413.

TX970045M