Synthesis of endoperoxides by domino reactions of ketones and molecular oxygen

Innovative Centre of the Faculty of Chemis Department of Chemistry, University of Nov Faculty of Chemistry, University of Belgr Belgrade 118, Serbia. E-mail: rsaicic@chem ICTM Center for Chemistry, Njegoseva 12, † Dedicated to the memory of our col (1953–2015). ‡ Electronic supplementary information NMR spectra for all compounds, CIF le deposited. CCDC 1411164–1411167. For E other electronic format see DOI: 10.1039/ Cite this: RSC Adv., 2015, 5, 99577


Introduction
Endoperoxides are important pharmacophores in a number of natural and synthetic, biologically active compounds. Although best known as potent antimalarials, 1 cyclic peroxides also exhibit a range of activities which encompasses antitripanosomal, 2 antifungal, antiviral and anticancer activity. 3,4 Whereas tetraoxanes and trioxanes are important units of synthetic bioactive peroxides, 1,2-dioxanes and dioxenes are the most frequent constituents of naturally occuring endoperoxides. Therefore, synthetic chemists have invested considerable efforts in developing synthetic approaches to this class of compounds. 5

Results and discussion
During the course of a synthetic study of platensimycin, we exposed methyl cyclopentyl ketone 1 to the enolization conditions under a presumed argon atmosphere, aiming to perform a Michael addition with the thermodynamic enolate. Surprisingly, spectral data of the isolated compound did not match the expected product, but indicated a dimeric structure with a higher oxidation level. Detailed analysis of spectra indicated the product structure 2 (Scheme 1). Stereochemical assignement of the compound 2, as deduced from NOESY spectrum, was confirmed by the results of X-ray diffraction structural analysis, as represented in Figure 1. As it was clear that the product arised from the reaction of ketone enolate with molecular oxygen, the reaction was performed under an oxygen atmosphere, in order to improve the yield; this time, surprisingly, endoperoxide 2 was not observed. The optimization of the reaction conditions involved variations of the base, solvent, atmosphere and the reaction temperature. We found that the highest yield of endoperoxide 2 could be obtained when the reaction was performed in THF, using a mixture of KOt-Bu and KOH as a base, under an argon atmosphere where oxygen was present only in low concentration; under these conditions, peroxide 2 could be isolated in 48% yield as a crystalline, analytically pure compound (Scheme 1, example 1). Cyclopent-3-enyl methyl ketone 3 behaved similarly, affording endoperoxide 4 in 43% yield (Scheme 1, example 2).
Scheme 1: Formation of 1,2-dioxanes 3 and 4 in the basecatalyzed reactions of ketones 1 and 2 with oxygen. In subsequent experiments, attempts were made to accomplish cross-cyclotrimerization of two different ketones with oxygen. Indeed, the reaction of methyl cyclopentyl ketone and acetone afforded peroxide 6a in 64% yield (Table  1, entry 1). Several other ketones (5) were also submitted to the cross-cyclotrimerization sequence with methyl cyclopentyl ketone 1 and oxygen, to give the desired peroxide products (mostly as crystalline compounds) in modest to fair yields. In some cases the products (6) contained variable amounts of the side product of homo-trimerization (2) (entries 2, 3, 4 and 5). However, the preponderance of the cross-products (6) is of note, given the fact that ketones were used in equimolar amounts. The results of these experiments are represented in Table 1. All products were obtained as single stereoisomers, with both hydroxy groups assuming axial positions ( Figure 2). A literature search revealed that Barton and collaborators encountered this reaction while inspecting more closely autooxidation of isobutyrophenone. 6 However, no subsequent report on this reaction ensued, although "the phenomenon certainly deserves further study". 7 The proposed reaction mechanism is represented in Scheme 2. It involves the oxidation of methyl cyclopentyl ketone enolate 7 with molecular oxygen to give hydroperoxide 8, which then acts as the electrophile in subsequent aldol addition of enolate 9. Selectivity in cross-cyclotrimerization can be explained by the well known propensity of methyl cyclopentyl ketone for the formation of a thermodynamic enolate 7, 8 which undergoes the reaction with molecular oxygen. 9 This reaction could proceed either as a direct reaction of the enolate with triplet oxygen, or via a cage radical pair mechanism; the intermediacy of free radical species can be ruled out, as the addition of BHT did not inhibit the reaction. The hydroperoxide thus formed would then act as the acceptor (probably activated by intramolecular hydrogen bond) in the cross aldol addition. Surprisingly, this mechanistic pathway was not corroborated with experimental evidence. Hydroperoxide 8 (represented in Scheme 2 in the form of the corresponding potassium salt) could be separately prepared in 71% yield, when the reaction of methyl cyclopentyl ketone with KOt-Bu was performed under an oxygen atmosphere (this raction is also unaffected by the presence of BHT). 10 However, a mixture of hydroperoxide 8 and methyl cyclopentyl ketone 1 did not produce observable amounts of endoperoxide 2 when exposed to the action of KOt-Bu. To clarify the relationship between enolate stability and the course of cross cyclotrimerization, and to explain the specific reactivity of methyl cyclopent(en)yl ketone(s) 1 and 3, dispersion corrected density functional theory calculations (DFT) were performed. Free energy changes for the t-BuOmediated formation of more substituted enolates in THF solution have been calculated at BP86-D3/TZP level of theory (COSMO) with two methods. (The first one is based on direct calculation of thermodynamic properties in THF within the COSMO model, 11 while in the second, indirect method, the gas-phase free energies of reactants and products are corrected by solvation energies and energies due to the geometry relaxation of a molecule in the solvent, 12 see Computational Details section for details). The results of both, direct and indirect method, are consistent, and Gibbs free energies for the formation of more substituted enolates, corresponding to all aliphatic ketones used in cross trimerization reactions (1, 3, 5a-e) are listed in Table 2. 14 It can be seen that methyl cyclopentyl ketone 1 and methyl cyclopentenyl ketone 3 have the most stable thermodynamic enolates. This indicates that methyl cyclopentenyl ketone 3 should also act as the acceptor in the cross cyclotrimerization reaction. Indeed, when 3 was submitted to the cyclotrimerization conditions in the presence of methyl isobutyl ketone 5c, endoperoxide 10 was obtained in 37% yield (Scheme 3, example 1). With these results in hand, we Scheme 3: Reactions of ketones 3 and 11, as predicted by calculations searched for some other ketone structures that would give rise to stable thermodynamic enolates and hence be good partners for the cross cyclotrimerization. Calculations showed that indanyl methyl ketone 11 should be such a compound (Table  2), and indeed, its' cyclotrimerization with acetone provided endoperoxide 12 in 77% yield (Scheme 3, example 2). Some ketones did not participate in the crosscyclotrimerization; these include progesterone, α-ionone, and 2-acetylnorbornene, inter alia. Therefore, we searched for a method to introduce the endoperoxide structural unit into various structures, not directly obtainable by the reaction. To this aim, cross-cyclotrimerisation was performed with acetylstyrene 13. Ozonolysis of the product 14 provided aldehyde 15, which could be submitted to further synthetic transformations, as examplified by the organometallc addition represented in Scheme 4, which gave peroxy-triol 16. In this way, various structural entities could be linked to the peroxide moiety, via nucleophilic addition.

Conclusions
The new method for the synthesis of endoperoxides (1,2dioxanes) is described, which relies on the base-catalyzed cyclotrimerization of two ketones with molecular oxygen. The domino reaction involves the oxidation of 1cyclopentylethanone with molecular oxygen, cross-aldol addition and cyclization. The product of the reaction with acetylstyrene is amenable to further synthetic transformations, which allows for the linkage of the endoperoxide structural unit with other molecules.

General experimental
All chromatographic separations were performed on Silica, 10-18, 60A, ICN Biomedicals. Standard techniques were used for the purification of reagents and solvents. NMR spectra were recorded on a Bruker Avance III 500. Chemical shifts are expressed in ppm (δ) using tetramethylsilane as internal standard. IR spectra were recorded on a Nicolet 6700 FT instrument, and are expressed in cm -1 . Mass spectra were obtained on Agilent technologies 6210 TOF LC/MS instrument (LC: series 1200). Melting points were determined on an electrothermal apparatus and are corrected. Microanalyses were performed using the Vario EL III instrument CHNOS Elementar Analyzer, Elementar Analysensysteme GmbH, Hanau, Germany. Diffraction data were collected on an Oxford Diffraction KM4 four-circle goniometer equipped with Sapphire CCD detector.
General procedure for the synthesis of endoperoxides: 8-Cyclopentyl-10-methyl-6,7-dioxaspiro[4.5]decane-8,10-diol (2) KOt-Bu (55.2 mg; 0.493 mmol; 1.97 equiv) and KOH (4.8 mg, 0.085 mmol, 0.34 equiv) were added to a cold (0 °C) solution of 1-cyclopentylethanone (28 mg; 0.25 mmol) in THF (0.5 mL), with stirring, under an argon atmosphere. During the addition of the base, the cork was opened for 10 sec, so that the air could enter into the reaction flask. The reacting solution should occupy ¼ of the flask volume (in this experiment a 2 mL flask was used). The reaction mixture was stirred for 1 h, when TLC indicated the consumption of the starting material. The reaction was quenched by the addition of water and 10% citric acid to attain pH 4, followed by ethyl acetate extraction.

10
KOt-Bu (531 mg; 4.74 mmol; 1.77 equiv) was dissolved in a solvent mixture: t-BuOH (4.3 mL) and DME (4.3 mL) under an argon atmosphere, and the resulting yellow-orange solution was cooled to -30 °C. After that, oxygen was bubbled through the solution until the reaction was complete. 1-Cyclopentylethanone (300 mg; 2.68 mmol; 1 equiv) was added and the reaction mixture was stirred for the next 75 minutes, when TLC (petroleum-ether/ethyl acetate = 4:1, PAA) indicated that nearly all of the substrate was converted and the product was starting to decay. Ice-cold sat. NaHCO 3 (3 mL) was added, layers were separated, pH of the aqueous layer was adjusted to 3-4 using conc. H 3 PO 4 (~0.5 mL), diluted with water and extracted with diethyl ether (2 × 10 mL). The combined etheral extract was dried over anhydrous MgSO 4 , filtered and concentrated under reduced pressure to afford the crude product, which was purified by column chromatography (SiO 2 ; eluent: petroleum-ether/ ethyl acetate = 6:1). Pure hydroperoxyde 8 (275.8 mg, 71%) was isolated as a colourless liquid.

4-(8,10-Dihydroxy-10-methyl-6,7-dioxaspiro[4.5]decan-8yl)benzaldehyde (15)
Ozone was bubbled through a cold (-78 °C) solution of 10methyl-8-(4-vinylphenyl)-6,7-dioxaspiro[4.5]decane-8,10-diol (9; 40 mg, 0.138 mmol) in dichloromethane (10 mL). As soon as the blue color of dissolved ozone was detected, argon was bubbled through the reaction mixture for 3-5 min and dimethylsulfide (0.5 mL) was added. functional, with Becke's integration grid of good quality. 22 Analytical harmonic frequencies 23 were calculated at the same level of theory, in order to ascertain that all the optimized structures correspond to the minima on the potential energy surface. In addition, vibrational analysis have been used to evaluate zero point effects, entropic and thermal corrections to the Gibbs free energy. Gibbs free energy change for a reaction of a ketone with tert-butoxide forming corresponding enolate and tert-butanol in THF solution have been calculated with two methods. The two methods are employed because accurate description of reactions in solutions are still not at the same high levels as those in the gas phase, and there are debates in the literature concerning the free energy calculations with implicit solvation methods. 24 The first method is direct calculation of thermodynamic properties within the COSMO model. COSMO model takes effectively into account cavitation, internal energy and entropy effects of the solvent, and therefore can yield an estimate of the Gibbs free energies. 25 The second method is indirect, and is based on the thermodynamic cycle for proton exchange reactions, commonly employed for the calculations of pKa values. 12 With the indirect method, the gas-phase free energies of reactants and products are corrected by solvation energies and energies due to the geometry relaxation in the solvent. These corrections are straightforwardly acquired from energy differences between the gaseous and solvated states, since