Reductive Metabolism of Ellagitannins in the Young Leaves of Castanopsis sieboldii

The leaves of Castanopsis sieboldii (Fagaceae) contain characteristic hexahydroxydiphenoyl (HHDP) esters of 28-O-glucosyl 2α,3β,23,24-tetrahydroxyolean- and urs-12-en-28-oic acids. In this study, uncharacterized substances were detected in the young leaves, which are not observed in the mature leaves. Preliminary HPLC analyses indicated that the substances had dehydro-HHDP (DHHDP) ester groups; however, the esters were unstable and decomposed during extraction. Therefore, the compounds were isolated as their stable phenazine derivatives by extracting the young leaves with acidic aqueous EtOH containing o-phenylenediamine. The structures of the phenazine derivatives indicated that the unstable metabolites of the young leaves were 3,24-DHHDP esters of the abovementioned triterpenes. Extraction of the young leaves with 80% acetonitrile containing reducing agents, ascorbic acid or dithiothreitol afforded the corresponding HHDP esters. Furthermore, heating of the young leaves in 80% acetonitrile also yielded the same HHDP esters as the reduction products. The results suggested that the HHDP esters are reductively produced from DHHDP esters in the young leaves. In addition, the structures of five previously reported triterpene HHDP esters were revised.


Introduction
Ellagitannins are a group of hydrolyzable tannins having hexahydroxydiphenoyl (HHDP) ester moieties and related acyl groups [1][2][3]. The HHDP group is a dimer of the galloyl group, and enzymatic conversion of gallotannins to ellagitannins has been reported [4][5][6]; however, the details of the biosynthetic mechanisms are still ambiguous. In addition, there is no chemical or biological evidence for the production mechanism of other ellagitannin acyl groups, such as dehydro-HHDP (DHHDP) and chebuloyl [1]. To clarify the biogenetical relationship between ellagitannin acyl groups, the chemical reactivity of the acyl groups and change of tannin composition during leaf growth were examined in our laboratory. Previously, it was demonstrated that the major ellagitannin in the leaves of Camellia japonica was oxidatively degraded as the leaves matured, and the chemical mechanism was proposed [7]. In this study, change of tannin composition in the leaves of Castanopsis sieboldii (syn. C. cuspidata var. sieboldii, Fagaceae) was examined. C. sieboldii is an evergreen broadleaf tree widely distributed in western Japan. tree widely distributed in western Japan. The leaves contain galloyl shikimic acids [8] and ellagitannins, and the latter are mainly composed of a series of HHDP esters of 28-O-glucosyl 2α,3β,23,24-tetrahydroxyolean-and urs-12-en-28-oic acids (28-O-glucosyl castanopsigenins A and B) [9].

HPLC analysis of the leaves
In the present study, firstly the triterpene HHDP esters in the fresh leaves were compared at different growth stages in spring (April, 2015). HPLC analyses of the leaf buds and small young leaves at the top of the twig showed two broad peaks with similar UV absorptions to those of other triterpene DHHDP esters ( Figure 1A: 1′o,u; where o: oleanane type, u: ursane type). However, the substances were detected neither in the wintered leaves of the same twig ( Figure 1C) nor in summer mature leaves of the same tree collected in July. This observation suggests that the tannins detected as the broad peaks in the young leaves are key metabolites in the metabolism of the triterpene HHDP esters. On extraction of the young leaves with solvent containing o-phenylenediamine, the broad peaks were not observed and two peaks were detected at longer retention times ( Figure 1B, 1o, 1u, and 2o,u), suggesting that the compounds 1′o,u have DHHDP ester moieties [10]. In contrast, no change was observed on treatment of the mature leaves with o-phenylenediamine. This study aimed at characterization of the key metabolites 1′o,u in the young leaves.

Structures of Phenazine Derivatives
To determine the structures of the uncharacterized metabolites 1′o,u, the young fresh leaves were firstly extracted by a conventional method using aqueous acetone; however, the metabolites decomposed during concentration of the extract at 40-45 °C. The aforementioned preliminary experiments indicated that the unstable metabolites could be converted into stable derivatives by treatment with o-phenylenediamine ( Figure 1B); therefore, the young leaves were extracted with 30% AcOH in EtOH containing o-phenylenediamine. The extract was fractionated by Diaion HP20SS column chromatography and the fractions containing triterpene esters were further separated by d) 22

Structures of Phenazine Derivatives
To determine the structures of the uncharacterized metabolites 1 o,u, the young fresh leaves were firstly extracted by a conventional method using aqueous acetone; however, the metabolites decomposed during concentration of the extract at 40-45 • C. The aforementioned preliminary experiments indicated that the unstable metabolites could be converted into stable derivatives by treatment with o-phenylenediamine ( Figure 1B); therefore, the young leaves were extracted with 30% AcOH in EtOH containing o-phenylenediamine. The extract was fractionated by Diaion HP20SS column chromatography and the fractions containing triterpene esters were further separated by  [11], and the 1 H-and 13 C-NMR spectra (Table 1) showed signals arising from phenazine moieties, which were identical to those observed for the phenazine derivatives of dehydroellagitannins [12]. The 13 C-NMR spectra also exhibited signals arising from triterpene and β-glucopyranose moieties, the chemical shifts of which were very similar to those of triterpene HHDP esters reported in a previous study [9]. Comparison of the signals with those in the literature suggested that 1o and 1u were esters of β-d-glucopyranosyl 2α,3β,23,24-tetrahydroxyolean-12-en-28-oate [13] and β-d-glucopyranosyl 2α,3β,23,24-tetrahydroxyurs-12-en-28-oate [14], respectively. These triterpene units were originally named as castanopsigenins A and B by Ageta et al. [9]. The location of the glucopyranose moiety at the C-28 carboxyl group was confirmed by HMBC correlations between the anomeric protons and the carboxyl carbons ( Figure 2). The configurations of the A rings were determined by the appearance of NOESY correlations of the C-25 methyl protons with the H-2 and C-24 methylene protons. Furthermore, the locations of the phenazine ester moiety were determined based on the HMBC correlations of the ester carboxyl carbons with triterpene H-3 and H-24. The orientations of the acyl groups are also apparent from the HMBC correlations of the ester carbons and the aromatic protons. The atropisomerism of the biphenyl bond in 1o and 1u was determined to be in S configuration based on the appearance of positive and negative Cotton effects at 270 and 244 nm, respectively, in the CD spectrum ( Figure S43 in Supplementary Materials) [15]. The absolute configurations of glucose units were determined to be d-form by acid hydrolysis and HPLC comparison with the thiazoridine derivative prepared with l-cysteine methyl ester and o-tolylisothiocyanate [16]. a Only the data of oreanane isomer was shown. b-j Assignments may be interchanged.
The minor phenazine derivative 2o,u was an inseparable mixture of isomers. The 1 H-and 13 C-NMR spectra (Table 1) were very similar to those of 1o and 1u, and the HMBC correlations ( Figure 2) confirmed that the partial structure including the phenazine units and triterpene A ring was the same as those of 1o and 1u. The 13 C-NMR signals of the oleanan-and urusane-type triterpene moieties were also the same as those of 1o and 1u. Furthermore, the CD spectra displayed the same features as those of 1o and 1u ( Figure S44). However, the signals of glucose were not observed in the NMR spectra of 2o,u. Thus, 2o,u were characterized to be desglucosyl analogs of 1o and 1u as shown in Figure 2. as those of 1o and 1u ( Figure S44). However, the signals of glucose were not observed in the NMR spectra of 2o,u. Thus, 2o,u were characterized to be desglucosyl analogs of 1o and 1u as shown in Figure 2.

Reaction of the Unstable Metabolites
The production of these phenazine derivatives strongly suggested that the unstable compounds detected in the fresh young leaves were DHHDP esters 1′o and 1′u ( Figure 3). It is generally accepted that the configuration of the biphenyl bond of phenazine derivatives reflects the configuration of the methine carbon of DHHDP esters [17]. Therefore, the configurations of the DHHDP group of 1′o and 1′u were deduced to be S. Most DHHDP esters of ellagitannins reported so far are stable unless the compounds were heated under acidic or basic conditions; however, Foo reported an example [18]: the 3,6-DHHDP group of 1-O-galloyl-2,4;3,6-bis-(R)-DHHDP-β-D-glucose (amariin) spontaneously degraded in EtOH to yield a reduction product having a 3,6-(R)-HHDP group. The degradation of 1′o and 1′u in the young leaves of C. sieboldii may be the same phenomenon. To confirm this, the young leaves were first treated with a reducing agent, dithiothreitol, in 80% CH3CN at r.t. for 2 h, and HPLC analysis of the reaction mixture showed 4 as a major product along with a small amount of 3. Reduction with ascorbic acid also gave a similar result. The same products were also generated by merely heating the leaves at 70 °C in 80% CH3CN without any reducing agents. In this reaction, however, 3 and 4 were produced in an inverse ratio, that is 3 was the major and 4 was the minor product. Because the products 3 and 4 were detected in the extract of the mature leaves collected in October, the products were isolated from mature leaves to determine the structures.

Reaction of the Unstable Metabolites
The production of these phenazine derivatives strongly suggested that the unstable compounds detected in the fresh young leaves were DHHDP esters 1 o and 1 u ( Figure 3). It is generally accepted that the configuration of the biphenyl bond of phenazine derivatives reflects the configuration of the methine carbon of DHHDP esters [17]. Therefore, the configurations of the DHHDP group of 1 o and 1 u were deduced to be S. Most DHHDP esters of ellagitannins reported so far are stable unless the compounds were heated under acidic or basic conditions; however, Foo reported an example [18]: the 3,6-DHHDP group of 1-O-galloyl-2,4;3,6-bis-(R)-DHHDP-β-d-glucose (amariin) spontaneously degraded in EtOH to yield a reduction product having a 3,6-(R)-HHDP group. The degradation of 1 o and 1 u in the young leaves of C. sieboldii may be the same phenomenon. To confirm this, the young leaves were first treated with a reducing agent, dithiothreitol, in 80% CH 3 CN at r.t. for 2 h, and HPLC analysis of the reaction mixture showed 4 as a major product along with a small amount of 3. Reduction with ascorbic acid also gave a similar result. The same products were also generated by merely heating the leaves at 70 • C in 80% CH 3 CN without any reducing agents. In this reaction, however, 3 and 4 were produced in an inverse ratio, that is 3 was the major and 4 was the minor product. Because the products 3 and 4 were detected in the extract of the mature leaves collected in October, the products were isolated from mature leaves to determine the structures. The product 3 was identified as the oleanane (o), and ursane (u) forms of castanopsinin E with (S)-HHDP ester moieties [9], and the structure was confirmed by HMBC examination (Scheme 1). Comparison of the 13 C-NMR data of 4 with those in the literature suggested that this compound was castanopsinin C having an (R)-HHDP group at the triterpene 2,3-vicinal hydroxy groups [9]. This was also in agreement with the appearance of a negative Cotton effect at 231 nm. Furthermore, HPLC comparison with an authentic sample also supported that 4 was castanopsinin C. However, the HMBC spectrum of the isolated compound measured in methanol-d4 showed correlations of two ester carbonyl carbons with two doublet proton signals at δ 4.95 (d, J = 9.8 Hz, H-3) and δ 4.13 (d, J = 12.5 Hz). The coupling constant of the latter (12.5 Hz) was too large for the usual vicinal coupling between H-3 and H-2. In addition, the H-2 signal (δ 3.81, m) was not correlated with ester carbonyls in the HMBC spectrum. Therefore, it is more reasonable that the signal at δ 4.13 is attributable to geminally coupled H-23 or H-24, and one of the HHDP ester groups is attached to C-23 or C-24. Attempts to assign the signals of H-23 and H-24 by 1 H-1 H COSY, ROESY, HSQC, and HMBC measured in pyridine-d5 and CD3OD using different parameters all failed. Because of these unexpected difficulties in spectroscopic examination, DFT calculations were applied to two model compounds: 3,24-and 3,23-(R)-HHDP esters of methyl 2α,3β,24-trihydroxyolean-12-en-28-oate (the glucose moiety was replaced by a methyl group) ( Figure S45). After a conformational search, 1 H-and 13 C-NMR chemical shifts of the low-energy conformers with Boltzmann populations greater than 1% were calculated [19,20]. The experimental 1 H and 13 C-NMR chemical shifts were in agreement with calculated values for 3,24-(R)-HHDP esters (R 2 = 0.9802 for 1 H; R 2 = 0.9972 for 13 C) rather than 3,23-isomers (R 2 = 0.9744 for 1H; R 2 = 0.9952 for 13 C) ( Figure S46). Furthermore, DP4+ analysis gave 100.0% probability for 3,24-(R)-HHDP esters (Table S7) [21]. Based on these results, the structure of castanopsinin C was revised to be β-D-glucopyranosyl 3,24-(R)-HHDP-2α,3β,23,24-tetrahydroxyurs-and olea-12-en-28-oate (4o,u). The product 3 was identified as the oleanane (o), and ursane (u) forms of castanopsinin E with (S)-HHDP ester moieties [9], and the structure was confirmed by HMBC examination (Scheme 1). Comparison of the 13 C-NMR data of 4 with those in the literature suggested that this compound was castanopsinin C having an (R)-HHDP group at the triterpene 2,3-vicinal hydroxy groups [9]. This was also in agreement with the appearance of a negative Cotton effect at 231 nm. Furthermore, HPLC comparison with an authentic sample also supported that 4 was castanopsinin C. However, the HMBC spectrum of the isolated compound measured in methanol-d 4 showed correlations of two ester carbonyl carbons with two doublet proton signals at δ 4.95 (d, J = 9.8 Hz, H-3) and δ 4.13 (d, J = 12.5 Hz). The coupling constant of the latter (12.5 Hz) was too large for the usual vicinal coupling between H-3 and H-2. In addition, the H-2 signal (δ 3.81, m) was not correlated with ester carbonyls in the HMBC spectrum. Therefore, it is more reasonable that the signal at δ 4.13 is attributable to geminally coupled H-23 or H-24, and one of the HHDP ester groups is attached to C-23 or C-24. Attempts to assign the signals of H-23 and H-24 by 1 H-1 H COSY, ROESY, HSQC, and HMBC measured in pyridine-d 5 and CD 3 OD using different parameters all failed. Because of these unexpected difficulties in spectroscopic examination, DFT calculations were applied to two model compounds: 3,24-and 3,23-(R)-HHDP esters of methyl 2α,3β,24-trihydroxyolean-12-en-28-oate (the glucose moiety was replaced by a methyl group) ( Figure S45). After a conformational search, 1 Hand 13 C-NMR chemical shifts of the low-energy conformers with Boltzmann populations greater than 1% were calculated [19,20]. The experimental 1 H and 13 C-NMR chemical shifts were in agreement with calculated values for 3,24-(R)-HHDP esters (R 2 = 0.9802 for 1 H; R 2 = 0.9972 for 13 C) rather than 3,23-isomers (R 2 = 0.9744 for 1H; R 2 = 0.9952 for 13 C) ( Figure S46). Furthermore, DP4+ analysis gave 100.0% probability for 3,24-(R)-HHDP esters (Table S7) [21]. Based on these results, the structure of castanopsinin C was revised to be β-d-glucopyranosyl 3,24-(R)-HHDP-2α,3β,23,24-tetrahydroxyursand olea-12-en-28-oate (4o,u). On heating in 80% CH3CN, the postulated DHHDP esters 1′o,u were reduced to 3 and 4, similar to the behavior observed for 1-O-galloyl-2,4;3,6-bis-(R)-DHHDP-β-D-glucose [18]. The most likely explanation for this phenomenon is reduction-oxidation disproportionation. A related disproportionation reaction was previously observed in black tea polyphenol production during tea fermentation. Epigallocatechin-3-O-gallate, a major catechin of green tea leaves, is converted to dehydrotheasinensin A (5), which has a hydrated quinone structure closely related to the DHHDP group. The quinone dimer 5 is unstable and decomposes on heating to give the reduction products theasinensins A (6) and D (7), accompanied by a complex mixture of oxidation products including galloyl oolongtheanin (8) (Figure 4) [22]. In the case of 3,6-DHHDP esters of 1-O-galloyl-2,4;3,6-bis-(R)-DHHDP-β-D-glucose, only the reduction product has been characterized in previous studies and no oxidation products have been identified. This spontaneous reduction of the hydrated quinone dimer may play a role in the metabolism of plant polyphenols to some extent. On heating in 80% CH 3 CN, the postulated DHHDP esters 1 o,u were reduced to 3 and 4, similar to the behavior observed for 1-O-galloyl-2,4;3,6-bis-(R)-DHHDP-β-d-glucose [18]. The most likely explanation for this phenomenon is reduction-oxidation disproportionation. A related disproportionation reaction was previously observed in black tea polyphenol production during tea fermentation. Epigallocatechin-3-O-gallate, a major catechin of green tea leaves, is converted to dehydrotheasinensin A (5), which has a hydrated quinone structure closely related to the DHHDP group. The quinone dimer 5 is unstable and decomposes on heating to give the reduction products theasinensins A (6) and D (7), accompanied by a complex mixture of oxidation products including galloyl oolongtheanin (8) (Figure 4) [22]. In the case of 3,6-DHHDP esters of 1-O-galloyl-2,4;3,6-bis-(R)-DHHDP-β-d-glucose, only the reduction product has been characterized in previous studies and no oxidation products have been identified. This spontaneous reduction of the hydrated quinone dimer may play a role in the metabolism of plant polyphenols to some extent. On heating in 80% CH3CN, the postulated DHHDP esters 1′o,u were reduced to 3 and 4, similar to the behavior observed for 1-O-galloyl-2,4;3,6-bis-(R)-DHHDP-β-D-glucose [18]. The most likely explanation for this phenomenon is reduction-oxidation disproportionation. A related disproportionation reaction was previously observed in black tea polyphenol production during tea fermentation. Epigallocatechin-3-O-gallate, a major catechin of green tea leaves, is converted to dehydrotheasinensin A (5), which has a hydrated quinone structure closely related to the DHHDP group. The quinone dimer 5 is unstable and decomposes on heating to give the reduction products theasinensins A (6) and D (7), accompanied by a complex mixture of oxidation products including galloyl oolongtheanin (8) (Figure 4) [22]. In the case of 3,6-DHHDP esters of 1-O-galloyl-2,4;3,6-bis-(R)-DHHDP-β-D-glucose, only the reduction product has been characterized in previous studies and no oxidation products have been identified. This spontaneous reduction of the hydrated quinone dimer may play a role in the metabolism of plant polyphenols to some extent.

Revision of Structures of Triterpene HHDP Esters
In this study, the structure of castanopsinin C was revised to 4 based on spectroscopic and computational methods. In the original study conducted in 1988, 2-dimensional NMR analysis was not available and the structures were mainly determined based on low-resolution 1D NMR and chemical degradation and derivatization [9]. Therefore, the major constituents of the mature leaves were isolated and reinvestigated by HMBC experiments (Figure 5).

General Information
Optical rotations were measured on a JASCO P-1020 digital polarimeter (Jasco, Tokyo, Japan). IR spectra were measured on a JASCO FT/IR 410 spectrophotometer. Ultraviolet (UV) spectra were obtained on a JASCO V-560 UV/VIS spectrophotometer. ECD spectra were measured with a JASCO J-725N spectrophotometer. 1 H-and 13 C-NMR spectra were recorded on a Varian Unity plus 500 spectrometer (Agilent Technologies, Santa Clara, CA, USA) operating at 500 MHz and 126 MHz for the 1 H and 13 C nuclei, respectively. 1 H-and 13 C-NMR spectra were also recorded on a JEOL JNM-AL400 spectrometer (JEOL Ltd., Tokyo, Japan) operating at 400 and 100 MHz for the 1 H and 13 C nuclei, respectively. Coupling constants (J) were expressed in hertz and chemical shifts (δ) are reported in ppm with the solvent signal used as a standard (pyridine-d 5

Plant Material
Fresh leaves of C. sieboldii were collected at Nagasaki University in Japan in April 2015 and October 2016. A voucher specimen was deposited at the Nagasaki University Graduate School of Biomedical Sciences.

Extraction and HPLC Analysis
The crushed young leaves of C. sieboldii collected at the end of April (200 mg) were extracted with 2% trifluoroacetic acid (TFA) in 80% aqueous CH 3 CN (4 mL) in a screw-capped vial and shaken at r.t. for 2 h ( Figure 1A, Figure 3A). The crushed leaves (200 mg) were treated with 2% TFA in 80% aqueous CH 3 CN containing o-phenylenediamine (10 mg/4 mL) at r.t. for 2 h ( Figure 1B). The crushed leaves (200 mg) were treated with 80% aqueous CH 3 CN containing dithiothreitol (20 mg/4 mL) at r.t. for 2 h ( Figure 3B). The crushed leaves (200 mg) were heated at 70 • C for 1 h in 80% aqueous CH 3 CN (4 mL) ( Figure 3C). The crushed mature leaves collected at the end of April (200 mg) were extracted with 80% aqueous CH 3 CN (4 mL) at r.t. for 2 h ( Figure 1C, Figure 3D). After removal of plant debris by filtration with a membrane filter (0.45 µm), the filtrate was analyzed using reversed-phase HPLC.

Isolation of Phenazine Derivatives
Fresh young leaves (1.3 kg) of C. sieboldii collected at the end of April were crushed using a Waring blender with 30% AcOH in EtOH containing o-phenylenediamine (5.0 g/2.7 L) and left to stand at r.t. for 18 h. After filtration, the plant debris was extracted with 60% EtOH for 12 h. The extract was combined and concentrated by a rotary evaporator, and the resulting precipitate was removed by filtration. The filtrate was subjected to Diaion HP20SS column chromatography ( Table 1.

Acid Hydrolysis of Compound 1o,u
The absolute configuration of the glucose residue was determined according to the method previously reported with modifications [16]. Compound 1o,u (0.5 mg) was hydrolyzed with 0.5 M HCl (1 mL) in a screw-capped vial at 100 • C for 2 h and neutralized with Amberlite IR 400. After drying in vacuo, the residue was dissolved in pyridine (0.1 mL) containing l-cysteine methyl ester hydrochloride (0.5 mg) and heated at 60 • C for 1 h. To the mixture, a solution of o-tolylisothiocyanate (0.5 mg) in pyridine (0.1 mL) was added, and the solution was heated at 60 • C for 1 h. The reaction mixture was directly analyzed by reversed-phase HPLC. The peak at t R 16.95 min coincided with the derivative of d-glucose (d-glucose: 16.72 min; l-glucose: 15.37 min).

Computational Methods
A conformational search was performed using the Monte Carlo method and the MMFF94 force field with Spartan'14 (Wavefunction, Irvine, CA, USA). The obtained low-energy conformers within 6 kcal/mol of the most stable conformer were optimized at the B3LYP/6-31G(d,p) level in pyridine (PCM). The vibrational frequencies were also calculated at the same level to confirm the stability of the conformers, and no imaginary frequencies were found. The 1 H-and 13 C-NMR chemical shifts of the low-energy conformers with Boltzmann populations greater than 1% were calculated using the GIAO method at the mPW1PW91/6-311+G(2d,p) level in pyridine (PCM) [19,20]. The calculated NMR