Isolation and identification of two pairs of cytotoxic diterpene tautomers and their tautomerization mechanisms

Discovering anticancer drugs that do not have adverse side effects has been a developing research field worldwide in recent decades. In this work, four previously undescribed cytotoxic diterpenoids were isolated from the aerial parts of Isodon excisoides. Interestingly, these four diterpenoids were two pairs of tautomers that were first reported in plants. Their structures were further elucidated using various spectroscopic methods. The tautomerization phenomenon and mechanism for these two pairs of tautomers were emphatically described. The theoretical simulation results indicated that the diterpene tautomerization is greatly related to certain factors, including the existence of a transition state, the change of bond length and the level of conversion energy; the tautomerization for the two pairs of tautomers is mainly caused by proton transfer. For bioassays, the cytotoxicities of the tautomers against five human cancer cell lines were also investigated. The results indicated that each of the four diterpenoids showed significant cytotoxicity in at least three cell lines and could serve as potential anticancer agents for further investigation.

The relative configuration of the substituents was highlighted in a NOESY spectrum. The correlations of H-1 with H-5 and H-9, Me-18 with H-5, H-7 with H-5 and H-9, and H-13 with H-14 and H-16 indicated that H-1, H-5, H-7 and H-9 were positioned on the same side and that H-13, H-14 and H-16 were on the other side (Fig. 3).
To determine the absolute configuration, the electronic circular dichroism (ECD) spectrum of compound 1a was measured in MeOH and compared with the computed ECD spectra of 1a. The calculated curve matched well with that of the experimental curve (Fig. 4). According to the octant rule for saturated cyclopentanone 4 , the negative Cotton effect at 247.94 nm, based on the n-π* transition of the saturated cyclopentanone moiety, indicated that the D ring was β-oriented (Fig. 4). Finally, the structure of compound 1a was elucidated as 1α,7α-dihydroxy-14β,20-diacetoxy-ent-kaur-16-en-15-one (Fig. 1).
The same relative stereo-structure for 1a and 1b was deduced from their similar NOESY correlations (Fig. 3) and their almost identical 1 H-and 13 C-NMR data. In addition, compound 1b exhibited almost the same CD absorption as that of 1a. The calculated curve was in good agreement with that of the experimental curve (Fig. 4). Thus, the structure of 1a was determined to be 1α,14β-dihydroxy-7α,20-diacetoxy-ent-kaur-16-en-15-one ( Fig. 1).
Compound 2a was obtained as a white powder, and its molecular formula was determined to be C 24 H 34 O 7 by positive HRESIMS (m/z 457.21756 [M + Na], calcd C 24 H 34 O 7 Na + , m/z 457.21967). The UV spectrum of 2a showed an absorption maximum at 235 nm. The IR spectrum of 2a showed the presence of hydroxyl (3445 cm −1 ), carbonyl (1730 cm −1 ) and double bond (1648 cm −1 ) groups. Together with the NMR data of 2a, the results showed 2a was also an analogue of 1a.
The relative configuration of the substituents of 2a was determined with the NOESY spectrum. The correlations of H-1 with H-5 and H-9, Me-18 with H-5, H-7 with H-5 and H-9, and H-13 with H-14 and H-16 indicated that they were positioned on the same side and that H-14, H-13, and H-16 were on the other side (Fig. 3).
In addition, compound 2a exhibited almost the same CD absorption as that of 1a. The calculated curve matched well with that of the experimental curve (Fig. 4). Thus, the structure of 2a was determined to be 1α,14β-diacetoxy-7α,20-dihydroxy-ent-kaur-16-en-15-one (Fig. 1). www.nature.com/scientificreports www.nature.com/scientificreports/ Compound 2b was a white powder, and its molecular formula was determined to be C 24 H 34 O 7 by positive HRESIMS (m/z 457.21799 [M + Na] + , calcd for C 24 H 34 O 7 Na + , m/z 457.21967). The UV, IR and HR-ESI-MS spectra, together with the NMR data of 2b, showed that 2b was an isomer of 2a.

Tautomeric phenomenon and the dynamic equilibrium of two pairs of diterpene tautomers.
The UPLC-MS/MS and NMR data ( Fig. 5) (Figs. S1-4, Supporting information) showed that 1a and 1b and 2a and 2b were two pairs of tautomers. The interconversion phenomenon between each pair of diterpene tautomers was observed by HPLC.
As shown in Fig. 6, for a 1a sample solution maintained at 30 °C, the peak area (PA) of 1a initially accounted for 97.19% of the total chromatographic PA; the PA started to decrease after 24 h, and a corresponding increase in the PA of 1b was simultaneously detected. After 3 days, the PA of 1a significantly decreased, while the PA of 1b significantly increased. After 12 days, the reaction approached dynamic equilibrium, and the PA of 1a accounted for 68.20% of the total PA. After 16 days, the reaction reached dynamic equilibrium, and the PA ratio of 1a to 1b was 2:1. For a 1a solution incubated at 45 °C, the PA of 1a started to decrease after 7 h, and a corresponding increase in the PA of 1b was simultaneously detected. After 1 day, the PA of 1a significantly decreased, while that of 1b significantly increased. After 6 days, the reaction approached dynamic equilibrium, and the PA of 1b accounted for 63.94% of the total PA. After 9 days, the reaction reached dynamic equilibrium, and the PA ratio of 1a to 1b was also 2:1.
For a 1b sample solution stored at 30 °C, the PA of 1b started to decrease after 7 h, and a corresponding increase in the PA of 1a was simultaneously detected. After 2 days, the PA of 1b significantly reduced and that of 1a significantly increased. After 9 days, the reaction approached dynamic equilibrium. After 12 days, the reaction reached dynamic equilibrium, and the PA ratio of 1a to 1b was 2:1. For a 1b sample solution incubated at  www.nature.com/scientificreports www.nature.com/scientificreports/   www.nature.com/scientificreports www.nature.com/scientificreports/ 45 °C, the PA of 1b started to decrease after 1 h, and a corresponding increase in the PA of 1a was simultaneously detected. After 1 day, the PA of 1b was significantly reduced, while that of 1a significantly increased. After 5 days, the reaction approached dynamic equilibrium, and the PA of 1b accounted for 59.79% of the total PA. After 7 days, the reaction reached dynamic equilibrium, and the PA ratio of 1a to 1b was also 2:1.
All results showed that 1a was the preferential conformation in the pair of isomers that could interconvert, and the temperature affected both the stability and conversion speed of 1a to 1b and of 1b to 1a in a protic solvent, while their conversion rate was not affected. In addition, an interesting phenomenon was also found that under identical conditions, the conversion speed of 1b to 1a was faster than that of 1a to 1b in a protic solvent, and it increased with increasing temperature.
2a and 2b were another pair of tautomers, of which 2a was the preferred conformation. Similarly, the conversion speed from 2a to 2b was faster than that from 2b to 2a, and the conversion speed was faster at higher temperatures. As Fig. 3 and Fig. 4 show, under identical reaction conditions, the conversion speed to reach dynamic equilibrium for 1b⇌1a was slower than that for 2b⇌2a, which could be related to the spatial structure of the compounds (Fig. 6).
Theoretical studies on tautomerism of the two pairs of tautomeric diterpenoids. To describe this conversion reaction mechanism, the transition state, bond length, and activation energy of the two pairs of tautomers were determined using density functional theory.
Transition state analysis of the conversion reaction. Furthermore, the tautomerization mechanism of the two pairs of tautomeric diterpenoids was investigated, and the transition state calculations for the reactions were conducted using density functional theory. There is only a virtual frequency of −971.70 cm −1 for the transition state of the 1a⇌1b reaction (TS1), which suggests that the optimal transitional state was obtained. The corresponding normal coordinates of the imaginary vibration modes for TS1 are shown in Fig. 7A. There was an obvious stretching vibration of H and a wagging vibration of the C-C-O bonds. The vibration frequency directions of TS1 point to 1a and 1b, respectively.  www.nature.com/scientificreports www.nature.com/scientificreports/ The minimum energy path (MEP) was also obtained through intrinsic reaction coordinate (IRC) calculations, which indicated that the transition state was correlated with both tautomers and that the transition state was located on the right reaction path.
Similarly, the transition state of the 2a⇌2b reaction (TS2) was confirmed as well. There was only one virtual frequency of the transition state at −900.79 cm −1 . A similar stretching vibration mode of H and C-C-O in TS2 could also be found (Fig. 7B). www.nature.com/scientificreports www.nature.com/scientificreports/ Bond length analysis of the conversion reaction. As shown in Fig. 8A, the tautomerization between compound 1a and 1b was conducted mainly by a proton transfer reaction according to the change in bond lengths. TS1 was formed at the seventh point of the reaction coordinate, and the bond lengths of O21-H33, O17-H33, O21-C18 and O17-C18 of TS1 were 0.112, 0.133, 0.192 and 0.205 nm, respectively. The distance between O21 and H33 decreased gradually before TS1 formation and then remained constant after TS1 formation, suggesting the formation of the O21-H33 bond. The distance between O17 and H33 remained constant before TS1 formation and increased gradually after TS1 formation, suggesting the cleavage of O17-H33. At the same time, the increasing distance between O17 and C18 suggested the cleavage of O17-C18, and the shortening distance between O21 and C18 suggested the formation of O21-C18. The results showed that the breaking and formation of H-O bonds and C-O bonds were the key factors causing tautomerism in compounds 1a and 1b.
Activation energy difference analysis of the conversion reaction. The curve of the energy change along the reaction pathway is shown in Fig. 8B. The energy changes of 1a⇌1b also indicated that the conversion reaction had a low activation energy and that the conversion proceeded easily. The minimum energies for compounds 1a and 1b were −1461.612 a.u. and −1461.616 a.u., respectively. There was only a −0.004 a.u. (10.502 kcal·mol −1 ) energy difference between 1a and 1b. The tautomeric reaction of 1a to 1b was a forward reaction while that of 1b to 1a was the reverse reaction. The activation energies (Ea) of the forward reaction and reverse reaction were 252.0 kJ·mol −1 and 241.5 kJ·mol −1 , respectively. The results indicated that the tautomeric reaction had low activation energies and that their conversion proceeded easily.
It is worth mentioning that there were also similar structural functional groups on the other pair of tautomeric diterpenoids. However, no new tautomerization reaction between -CH 2 -OH on site C1 and -OAc on site C20 between 1a and 2a and between 1b and 2b was found in the experiment. To explain this observation, the theoretical transition states of the 1a⇌2a reaction (TS3) and 1b⇌2b reaction (TS4) were calculated as well. The structures of TS3 and TS4 are shown in Fig. 7C,D. The activation energies of 1a⇌2a and 1b⇌2b were obviously higher than those of 1a⇌1b and 2a⇌2b (Table 3), indicating that the tautomeric reactions of 1a⇌2a and 1b⇌2b did not proceed as easily as those of 1a⇌1b and 2a⇌2b. Therefore, only the tautomerization reactions 1a⇌1b and 2a⇌2b were observed, while 1a⇌2a and 1b⇌2b were not.
Proton transfer process analysis of the conversion reaction. The tautomerism between both pairs of tautomeric diterpenoids was mainly through a proton transfer reaction. The migration of acetate between C 14 and C 7 is shown in Fig. 9A. The carbonyl group of I was first protonated. After the attack of oxygen from a hydroxy group in II, the transient orthoester III was obtained. Then, the collapse of orthoester III gave the final product IV. In contrast, the acetates on C 1 and C 20 were relatively unable to form orthoester intermediates due to their configuration exhibited in V and VI.
The speed of the 1b⇌1a reaction is slower than that of the 2b⇌2a reaction. 1a, 1b, 2a, and 2b were different from each other due to different substituents at C1, C20, C7 and C14. Compared with the hydroxyl group on C20 of 2a, the acetate group on C 20 of 1a had one more chemical bond, so the acetate group with C 20 on the axial bond of 1a formed an intramolecular hydrogen bond with the hydroxy group on C 7 , which was not good for the transfer of the acetate group between C 7 and C 14 . However, in molecule 2a, the hydroxy group on C 20 did not easily form hydrogen bonds due to the distance between the hydroxy groups on C 20 and C 7 . Therefore, the transformation between 2a and 2b was much easier, as shown in Fig. 9B. Cytotoxicity assay. An MTT assay was performed to evaluate the cytotoxic effects of the tautomers against five human cancer cell lines, including HCT-116, A2780, NCI-H1650, BGC-823 and HepG2 (20170428, Beijing Bei Na Chuanglian Biotechnology Research Institute). The results are presented in Table 4.
The MTT test showed that the cytotoxicities of the four compounds against the five human cancer cell lines were very different. In the future, extensive studies should be conducted to reveal the structure-activity relationship of cytotoxic diterpenoids for the discovery of affective antitumour drugs.

Conclusions
In this paper, we isolated and identified four new 7,20-non-epoxy-ent-kaurane skeleton diterpenes as 1a, 1b, 2a and 2b. The interconversion experiments between 1a and 1b and between 2a and 2b in methanol solutions confirmed that they exists as two pairs of tautomers. Further analysis using density functional theory showed that the tautomeric reaction was closely related to the existence of a transition state, the change in bond length and the level of conversion energy. It is thought that proton transfer was the major tautomerization reaction mechanism between each pair of diterpene tautomers. In addition, the four diterpenes exhibited potent cytotoxicities against three or more tested human cancer cells. The results could serve as a valuable reference for the tautomerization mechanism of other tautomers and the study of the structure-activity relationship on cytotoxicities as anticancer drug precursors.

Methods
General experimental procedures. Optical rotation was measured using a SEPA-300 polarimeter (Horiba, Tokyo, Japan). NMR spectra were recorded on a Bruker Advance III spectrometer (Bruker, Billerica, Germany). Orientation separation and HRESIMS data were acquired using a UPLC-LTQ orbitrap (Thermo Fisher Scientific, Inc., Bremen, Germany). Semi-preparative HPLC was performed on a Waters 600/Waters 2487         Tautomeric phenomenon and dynamic equilibrium between each pair of diterpene tautomers monitored by HPLC. Appropriate amounts of compounds 1a, 1b, 2a and 2b were separately weighed, and each compound was divided into two equal portions, placed in amber sample vials, and dissolved in methanol to obtain sample solutions with concentrations of 1.0 mg/mL. The solutions were sealed and separately stored at 30 °C and 45 °C. Each sample solution was analysed by HPLC every hour until tautomerization occurred. Then, the analysis was performed every 24 h until equilibrium was achieved. Chromatographic analysis was performed on a Waters 2996/Waters 2487 (Waters, Milford, MA, USA) equipped with a YMC C 18 column (4.6 mm × 250 mm, 5 μm) with acetonitrile-water (23-70) implemented as the mobile phase for isocratic elution. The flow rate was 1.0 mL·min −1 , and the column temperature was 25 °C. The detection wavelength was 230 nm. The stability of 1a, 1b, 2a, 2b and the time taken to reach equilibrium under different temperatures were observed.
Theoretical studies on tautomerism of the two pairs of tautomeric diterpenoids. Furthermore, the tautomerization mechanism of the two pairs of tautomeric diterpenoids was investigated, and transition state calculations for the reactions were conducted using density functional theory. Gaussview 5.0 was used to build the molecular structures of the tautomers. The transition state was found at the B3LYP/6-31 + g (d) level and was confirmed by vibration frequency analysis. Then, the minimum energy path (MEP) was determined by using intrinsic reaction coordinate (IRC) calculations, and the imaginary vibrational mode of the transition state was studied. The stability of compounds 1a, 1b, 2a, and 2b and their transition states were calculated by means of the density functional theory (DFT) method at the 6-31 + g(d) level. The calculations were performed using the Gaussian 09 software package. Cytotoxicity assay. Five human cancer cell lines (colon carcinoma cell line HCT-116, hepatic cancer cell line HepG2, ovarian cancer cell line A2780, lung cancer cell line NCI-H1650 and gastric cancer cell line BGC-823) (20170428, Beijing Bei Na Chuanglian Biotechnology Research Institute) were used for pharmacological experiments. All cells were cultured in RPMI-1640 medium supplemented with 10% foetal bovine serum in a humidified atmosphere with 5% CO 2 at 37 °C. The cytotoxicity assay was performed according to the MTT method using 96-well microplates 12 .
In the test, each tumour cell was exposed to the test compound at concentrations of 1 × 10 −5 , 1 × 10 −6 , and 1 × 10 −7 mol/L. The inhibitory rate of the cell growth was calculated according to the following formula: inhibition rate (%) = (OD control − OD treated )/OD control × 100. Finally, IC 50 values were calculated using SPSS 16.0 statistical software.