Introduction

The genus Melicope is a member of the Rutaceae (Citrus family) and contains more than 200 species distributed in the Malagasy, Indo-Himalayan, South-East Asian and Pacific regions. More than 50 endemic species of Melicope are found on the Hawaiian Islands and belong to the section Pelea. Melicope species are proven to produce many interesting secondary metabolites including poly-methoxylated flavonoids, furocoumarins, acetophenones and quinolone alkaloids. Moreover, several Melicope species are used in traditional and modern medicine. Some of the constituents possess antibacterial, antidiabetic, cytotoxic and antiproliferative activities in human cancer lines .

Five years ago, we described new acetophenones and chromenes isolated from a dichloromethane extract of leaves of Melicope barbigera A. Gray, a species endemic to the island of Kaua’i, Hawaiian Islands . In addition, small amounts of two tetracyclic citrans were identified as a 30:70%-mixture of isomeric melifoliones A (1), and melifolione B (2) , both formerly found in Melicope latifolia (syn. Euodia latifolia .

We now report on the identification and structure elucidation of a new natural compound 4 in the dichloromethane extract of leaves of Melicope barbigera, which was characterized as an oxidation product of melifolione B (2) by means of high resolution electrospray ionization mass spectrometry (HRESIMS) and NMR spectra. However, the isomeric compound 3 could not be detected in the extract (Figure 1).

Figure 1: Structures of isomeric melifoliones and corresponding oxidation products.

Since 4 could be an artefact, built by oxidation of 2 during working up of the extract, and to finally confirm the structure, the isomeric melifoliones should be synthesized first, to get enough quantity for oxidation to the corresponding quinols in a second experiment. Using the method of Wang and Lee , 1 could be prepared regio-specifically in good yield by heating phloroacetophenone and citral in DMF under catalysis of ethylenediammonium diacetate (EDDA). However, synthesis of 2 was difficult. Despite numerous attempts, we only were able to get mixtures of both melifoliones with minor amounts of 2. Since all attempts to separate both isomers on a preparative scale failed, we were only able to record NMR spectra from a very small quantity of 2, received via fractional crystallization. However, the oxidation procedures were performed with mixtures of both isomers with the aim to get the quinols 3 and 4.

Results and Discussion Structure of compound 4

Compound 4 was isolated from a dichloromethane extract of the leaves of M. barbigera as a colorless and amorphous substance. The molecular formula was determined as C18H22O5 by high resolution electrospray ionization mass spectrometry (HRESIMS), thus containing one more oxygen atom than the melifoliones. The UV spectrum showed a maximum at 250 nm, typical for an α,β-unsaturated carbonyl structure.

The 1H NMR spectrum of 4 showed signals for 22 protons (Table 1) with a certain similarity to the melifoliones, but with an exchangeable signal found at δ 2.88 ppm, typical for a tertiary carbinol, instead of a phenolic OH-group at about δ 13 ppm, found in the spectra of the melifoliones. All 18 signals detected in the 13C NMR spectrum of 4 (Table 2) could be assigned by careful analysis of the 2D-NMR spectra (HSQC and HMBC). The position of the acetyl group was confirmed by a cross peak between C-2 and the protons of the acetyl-methyl group in the HMBC spectrum and by interactions between the acetyl-methyl group and the methyl group at C-9 as well as between H-4 and one of the geminal methyl groups at position 6 in the NOESY-spectrum. This allowed us to clarify the structure of 4 as an oxidation product of melifolione B (2). We suggest the following name: 2-acetyl-6a,7,8,9,10,10a-hexahydro-10b-hydroxy-6,6,9-trimethyl-1,9-epoxy-6H-dibenzo[b,d]pyran-3(10bH)-one and also propose the common name: melifoliox B for this new natural compound.

Table 1: 1H-NMR data of melifolione A (1), melifolione B (2) and compound 4 in CDCl3 (600 MHz, δ in ppm).

Position 1 2 4 H-2 6.04 (s) – – H-4 – 6.01 (s) 5.77 (s) H-6a 2.09 (ddd)a 2.02 (ddd)a 2.74 (t) H2-7 0.86/1.31 (m) 0.87/1.30 (m) 1.69/1.79 (m) H2-8 1.45/1.80 (m) 1.45/1.80 (m) 1.55/1.68 (m) H-10ax 1.86 (dd)b 1.86 (dd)b 1.26 (m) H-10eq 2.19 (ddd)c 2.19 (ddd)c 1.36 (m) H-10a 2.73 (br. s) 2.83 (br. s) 2.34 (dt) 6-CH3 ax 1.13 (s) 1.07 (s) 1.46 (s) 6-CH3 eq 1.59 (s) 1.51 (s) 1.54 (s) 9-CH3 1.40 (s) 1.45 (s) 1.24 (s) CH3CO 2.62 (s) 2.62 (s) 2.61 (s) 3-OH 13.3 (s) 13.6 (s) – 10b-OH – – 2.88 (s)

J in Hz: a11.7/5.3/2.7; b13.3/1.6; c13.3/4.5/3.5

Table 2: 13C NMR data of compounds 1, 2 and 4 in CDCl3 (150 MHz, δ in ppm).

Position 1 2 4 HMBC correlations of 4 C-1 162.8 158.9 195.2 C-1/H-10a C-2 97.1 106.4 107.5 C-2/H-4; C-2/CH3-CO C-3 164.2 165.3 190.5 – C-4 107.4 98.5 106.3 – C-4a 159.4 163.4 171.5 C-4a/H-10a C-6 86.8 85.2 88.5 C-6/6-CH3 ax; C-6/6-CH3 eq; C-6/H-6a C-6a 46.4 46.1 35.8 C-6a/6-CH3 ax; C-6a/6-CH3 eq; C-6a/H-10a C-7 22.0 22.1 24.7 C-7/H-6a C-8 37.6 37.8 36.2 C-8/H-6a; C-8/9-CH3; C-8/H2-7 C-9 76.2 76.7 70.6 C-9/9-CH3; C-9/H2-7; C-9/H2-10 C-10 35.0 34.9 37.0 C-10/H-10a; C-10/9-CH3 C-10a 27.7 27.5 44.1 C-10a/H-6a; C-10a/H-7eq; C-10a/H2-10 C-10b 107.5 106.6 72.5 C-10b/H-4 6-CH3 ax 24.6 24.3 28.3 6-CH3 ax/H-6a; 6-CH3 ax/6-CH3 eq 6-CH3 eq 30.1 29.7 31.7 6-CH3 eq/H-6a; 6-CH3 eq/6-CH3 ax 9-CH3 28.8 29.0 26.5 9-CH3/H2-8; 9-CH3/H2-10 CH3-CO 32.2 32.6 28.2 – CH3-CO 202.3 202.9 201.5 CH3-CO/CH3-CO

The stereochemistry of 4 is characterized by the three centers of asymmetry of the isomeric melifoliones at position 6a, 9, and 10a. Since the two hydrogenated pyran rings of 4 can only be linked cis, the two possible enantiomers of 4 have the following configuration: (6aR,9S,10aS,10bR) or (6aS,9R,10aR,10bS). Due to the finding, that the isolated compound showed no optical rotation, it exists as a racemate.

A stereo model of 4 provided important information for a better understanding of the NMR spectra. In contrast to the planar benzene ring of melifoliones, the para quinol ring in 4 adopts a flat but rigid boat conformation. However, the sp3-hybridized C-10b leads to greater flexibility of the whole molecule. Therefore, it is easier for the cyclohexane ring to deviate from a chair to a twisted boat conformation by rotating around the C-7/C-8-axis. As a result, the orientation of the geminal protons at these positions can change more easily between a pseudo-axial (ax) and a pseudo-equatorial (eq) position, and the difference in chemical shift values of these protons decreases compared to melifolione B (2). However, all protons in a pseudo-axial position show the expected high field shift compared to the pseudo-equatorial positions (Table 1).

The biggest change in chemical shift values of 4 compared to 2 is observed in the geminal protons at C-7 and C-10. While the protons at C-7 in 2 are strongly shielded by the aromatic ring current effect and therefore appear at the highest field, this effect is absent in 4. The protons at C-10 of 4 come to resonance at the highest field. With the greater flexibility of the cyclohexane ring of 4, the geminal methyl groups in the pyran ring can fold more easily to avoid the steric hindrance caused by the ethane bridge above the pyran rings. Consequently, these methyl protons resonate at similar field strength. The high field shift of H-10a in 4 is a result of the loss of its benzylic position (Table 1). The chemical shift values of C-6a and C-10a in the 13C NMR spectrum of 4 compared to 2 can be explained by the effect of the new OH-group at C-10b. The γ-position of C-6a leads to a high field shift (from 46.1 to 35.8 ppm) while C-10a shows the opposite effect (from 27.5 to 44.1 ppm), due to its β-position in regard to the OH-group at C-10b (Table 2).

Synthesis of melifoliones

Melifolione A (1) was prepared according to the method of Wang and Lee by heating phloroacetophenone and citral with ethylene-diammonium diacetate (EDDA) in DMF. However, reaction of the educts in pyridine at 60 °C led to chromene 5, which is a strong phenol. It is therefore easily separable as a nearly pure compound from the complex reaction mixture by extraction of the ether solution with NaOH. According to the literature, 5 should be converted into melifolione B (2) under acidic conditions .

Numerous attempts were made to cyclize 5 into melifolione B (2) with the following results (Scheme 1)

Heating of 5 in DMF at 100 °C yielded melifolione A (1) with traces of melifolione B (2). When 5 was heated in acetic acid at 80 °C, only benzoxocin 6 could be isolated. Reaction of 5 with catalytic amounts of p-toluenesulfonic acid at 80 °C in toluene afforded benzoxocin (7) as the only product. On the other hand, melifolione A (1) was completely decomposed under these conditions. Irradiation of 5 in methanol with UV-light (300 nm) for 3 days at room temperature gave melifolione B (2) with traces of melifolione A (1), but the yield was less than 5% and therefore of no preparative interest. The best results to get melifolione B (2) were achieved by briefly heating 5 in a closed microwave apparatus without solvent at 130–140 °C after addition of catalytic amounts of acetic acid. By this procedure, the proportion of 2 in the mixture of melifoliones was increased to a maximum of about 10–15%.

Scheme 1: Synthesis of melifoliones and by-products: a) pyridine/60 °C/8 h/51%, b) microwave/140 °C/20 min/70%, c) acetic acid/80 °C/1 h/57%, d) p-toluenesulfonic acid/toluene/80 °C/1 h/68%.

Since it was not possible to distinguish or even to separate the isomeric melifoliones using chromatographic methods, their quantitative ratio after cyclization of 5 could only be determined using the 1H NMR spectra. After careful analysis of the 2D NMR spectra (HSQC, HMBC) all carbon signals of both isomers could be unambiguously assigned to the corresponding atoms (Table 2). This clearly shows that the data given in literature must be corrected.

Oxidation of melifoliones

Mixtures of melifolione A (1) and melifolione B (2) (ratio ca. 10:1) were used for the following oxidation procedures:

After oxidation with alkaline ferricyanide or using the Fenton reaction with iron(II) sulfate/hydrogen peroxide in acidic solution, only the unchanged melifoliones could be recovered in both cases. Oxidation with potassium dichromate in acidic solution or with potassium permanganate in acidic or alkaline solution led to complete decomposition of the melifoliones. A solution of the melifoliones in methanol with addition of methylene blue decomposed completely into undefined products within 24 hours when exposed to sunlight and atmospheric oxygen. Oxidation with potassium peroxodisulfate (Oxone®) according to a method by Carreño et al. led to a complex mixture of compounds in which the quinols 3 and 4 could be excluded. While oxidation with alkaline hypochlorite solution gave a mixture of not identified chlorinated products, reaction with alkaline Lugol’s solution yielded the mono-iodized isomers of both melifolione A (8) and melifolione B (9), identified by their NMR spectra (Figure 2). Reaction with iodosobenzene diacetate – a typical reagent to form para-quinols from phenols – yielded small amounts of a substance, that could be identified as iodized phenyl ether of melifolione A (10), based on its HRESIMS and 1H NMR spectrum. The quinols 3 or 4 could not be detected.

Figure 2: Structure of iodized melifoliones.

Phenols with additional electronegative substituents (e.g., nitro or acetyl) and at least one ortho-positioned hydrogen atom initially react with iodosobenzene diacetate to form iodonium ylides or betaines, which, however, are unstable and react to form stable iodized phenolic ethers (Scheme 2). In the 1H NMR spectrum of compound 10, the signals for the phenolic OH and for the aromatic proton are missing compared to melifolione A (1). In contrast, the aromatic protons of the added phenyl ring appear as a typical coupling pattern in the expected range for phenyl ethers. The acetyl-methyl group is shifted by 0.3 ppm to a higher field and the methyl group in position 9 resonates at the same field strength as in the case of 2-iodomelifolione A (8).

Scheme 2: Reaction of melifolione A (1) with iodosobenzene diacetate.

Since all attempts to prepare the quinols 3 or 4 via oxidation of melifoliones had failed, the combined use of ferricyanide and hydrogen peroxide in alkaline solution was finally used. Under these conditions, superoxide and hydroxy radicals or even singlet oxygen can be formed . After addition of an alkaline solution of ferricyanide to a hydrogen peroxide containing solution of the melifoliones (1:2 ratio ca. 10:1) in acetonitrile, the color turns deep purple and later orange. After consumption of educts (TLC), extraction with ether yielded small amounts of a new substance as glistening crystals. The high-resolution mass spectrum gave a protonated mol peak at 279.1226 indicating the formula C15H19O5. Compared to melifoliones this means a loss of 3 C-atoms. Based on the NMR spectra (1H, 2D-COSY, 13C, HSQC, HMBC) the structure of the oxidation product was determined as the spirofuranone 11.

The 1H NMR spectrum of 11 (Table 3), compared to melifoliones, is characterized by the loss of the signals for the phenolic OH- and the acetyl-methyl-group and it is complex in the range of 1.3 to 1.7 ppm. Only the evaluation of the HSQC spectrum revealed that the singlet for the equatorial 1-methyl group and the center of the multiplet for the axial H-5 lie exactly on the top of each other at 1.61 ppm. The same applies to the signals of both axial H-7 and H-8 at 1.57 ppm. The signal for H-8a appears as a nearly symmetrical quintet, resulting from the trans diaxial coupling with H-8ax (about 10 Hz) and two gauche couplings with H-4a and H-8eq (each about 5 Hz). In addition to the typical vicinal couplings, two 4J-long-range couplings can be recognized in the 2D-COSY spectrum of 11. Both pairs of equatorial protons at C-5 and C-7 and at C-4a and C-8 are in a so-called W-arrangement, typical for a rigid chair conformation of the cyclohexane ring.

Table 3: NMR data of 11 in CDCl3 (1H = 600 MHz, 13C = 150 MHz, δ ppm).

Position 1H 13C HMBC correlations 1 – 83.5 C-1/1-CH3 (2x), C-1/H2-8 3 – 165.2 C-3/H-4’ (4J), C-3/1-CH3 (2x 4J) 4 = 2’ – 80.4 C-4/H-4’, C-4/H-4a, C-4/H-5ax, C-4/H-8a 4a 3.12 33.4 C-4a/H-8a, C-4a/H-8eq 5 1.61a/1.92 35.0 C-5/6-CH3, C-5/H2-7 6 – 86.0 C-6/H-4a, C-6/H2-5, C-6/H2-7, C-6/H2-8, C-6/6-CH3 7 1.57a/2.10 34.9 C-7/6-CH3 8 1.57a/1.99 19.1 C-8/H-4a, C-8/H-7ax 8a 1.86 39.4 C-8a/H-5eq, C-8a/H-7eq, C-8a/1-CH3 (2x) 1-CH3 ax 1.42 26.7 1-CH3 ax/1-CH3 eq 1-CH3 eq 1.61a 27.6 1-CH3 eq/1-CH3 ax 6-CH3 1.40 28.4 6-CH3/H-5ax, 6-CH3/H-7eq 3’ – 179.4 C-3’/H-4’, C-3’/6-CH3 (4J) 4’ 5.52 101.4 - 5’ – 172.2 C-5’/H-4’

aSignals are not separated and only detectable in the HSQC-spectrum.

The 13C NMR spectrum of 11 (Table 3) shows 15 signals that could be clearly assigned after evaluating the HSQC and HMBC spectra. In the HMBC spectrum, cross peaks occur for C-3 with H-4’ and with the protons of both geminal CH3 groups at C-1. Furthermore, a cross peak was found for the interaction of C-3’ with the protons of the 6-CH3 group. These observations suggest unusual (4J)-long-range C/H-couplings in the rigid spirocyclic 4-ring system.

The formation of 11 can be suggested by the following hypothetical pathway (Scheme 3): In a first step, the combination of hydrogen peroxide with ferricyanide reacts with the phenol in a dioxygenase like reaction to give a dioxetane a, followed by opening of the aromatic ring to form the oxonol anion b in alkaline solution, thus explaining the purple color. Oxidation of b leads to the delocalized radical c. Further oxidation of the hydrate d results in the spirofuran e . After ferricyanide oxidation of e, the 3-C dicarbonyl residue splits off as pyruvate , leaving the spirofuranone 11 as a stable product.

Scheme 3: Hypothetical pathway for the oxidative ring contraction of melifolione A (1) with ferricyanide/hydrogen peroxide.

Cristallographic investigation

A problem arose during the X-ray crystallography of 11. In addition to 11 DMF and an isomeric compound 12 were identified in the unit cell (Figure 3). Single-crystal X-ray diffraction data were collected for all samples using a Synergy diffractometer (Rigaku Oxford Diffraction) equipped with a photon-counting detector system . Despite careful selection and mounting of the crystals, all specimens exhibited weak and diffuse diffraction patterns, which significantly limited data quality. This behavior is attributed to poor crystallinity and partial disorder within the crystal, likely exacerbated by the gradual release of solvent molecules during sample handling and data acquisition. Data collection was therefore applied with seriously elongated exposure time.

Figure 3: Left Part: 11 (main fraction); Right part: 12 (side product).

The crystal structure was successfully solved using methods as implemented in the SHELXT program , followed by refinement with SHELXL . The refined asymmetric unit was found to contain about 2.75 molecules of compound 11 and 0.25 molecules of compound 12, indicating partial occupancy between the two components in terms of a solid solution. This effect is caused by the similarity of the shape of the two molecules (see Figure 3). In addition, one molecule of DMF, presumably incorporated during crystallization, was identified within solvent-accessible voids in the crystal lattice. Due to its high level of disorder, the DMF molecule was removed computationally using the SQUEEZE routine implemented in the PLATON software package .

The final structural model clearly establishes the connectivity and constitution of both compounds within the crystal, despite the challenges posed by disorder and weak diffraction data. Structural parameters are consistent with the proposed molecular structures and support the presence of both components within a single crystalline phase.

Crystallographic data for the structural analysis have been deposited with the Cambridge Crystallographic Data Centre (CCDC) under deposition number 2477299. Copies of the data can be obtained free of charge via https://www.ccdc.cam.ac.uk/data_request/cif.

Compound 12 is formed by oxidative contraction of the acetylated phenol ring of melifolione B (2) which was present in the starting material (Scheme 4). Indeed, upon close inspection of the 13C NMR-spectrum of 11, low intensity satellites were detected next to the high-field signals, while the deep-field signals were completely uniform. This fact was initially surprising, but can be explained based on the X-ray data. Both structures 11 and 12 can be precisely superimposed in the region of the furanone and pyranone rings (Figure 3). Therefore, the corresponding C-atoms have the same magnetic environment and the signals >100 ppm have no satellites. In contrast, the two cyclohexane rings of 11 and 12 are in a staggered position and their C-atoms are not magnetically equivalent. Due to the small amount of 12, its C-signals appear as small satellites next to the intense signals of 11 in the NMR spectrum (see Supporting Information File 1, Figure S 23). In contrast, no clear evidence of signals from the isomer 12 could be detected in the 1H NMR spectrum of 11.

Scheme 4: Oxidative ring contraction of melifolione B (2).

Oxidative ring contraction of melifolione B (2) can be explained in a similar way as in the case of melifolione A (1) (Scheme 4). Alternatively, the routes via intermediately formed pyranone epoxides 13 and 14 seems also to be possible (Scheme 5).

Scheme 5: Alternative routes to the spirofuranones 11 and 12 via hypothetical pyranone epoxides 13 and 14.

Compounds 11 and 12 are the first representatives of two new heterocyclic ring systems consisting of two pyran rings, one cyclohexane and one furan ring.