Late-stage skeleton transformation (LSST) of natural products, especially those with biological activities, is an effective strategy for constructing valuable derivatives compared to stepwise synthetic approaches. In recent years, selective C-H activation and bio-catalysis strategies have developed rapidly, enabling efficient late-stage editing of natural products [1–7]. Such skeleton transformation often undergoes the breaking and reorganization of bonds, sometimes involving complex cascade reactions [8]. It is worth mentioning that although the LSST strategies for natural products are fascinating organic transformation process, the relevant strategies are limited [9].

Photochemical reaction can offer unique opportunities to achieve fast and efficient skeleton transformation of natural products by unlocking site-specific reactivities under generally mild reaction conditions [10]. Undoubtedly, the use of photochemical reaction is a powerful strategy that allows the transformation between two highly different natural product skeletons in a single step. To our knowledge, although some meaningful derivatives have been obtained by photochemical reaction of natural products in recent years [9, [11], [12]], there are still some issues need to be addressed on the photochemical study of natural products. This is mainly due to the complex structure of natural products, which leads to the multiple possible photoreaction sites [9,13–15]. Therefore, it is important to understand the reaction mechanism for further prediction of possible products.

Polycyclic furanobutenolide-derived norcembrane diterpenoids are a small class of marine natural products with multiple remoted stereogenic centers in their macrocyclic ring and a wide range of biological activities, mainly isolated from soft corals of the genus Sinularia [16–19]. These bioactive metabolites have attracted extensive interests for total syntheses in recent years [20–22]. In our previous investigation of the chemical composition of Hainan soft coral Sinularia sp., norcembrane diterpenoid yonarolide (1) was found to be easily converted into yonarolide A (2), featuring an unprecedented 5/6/4/4/7 pentacyclic ring skeleton, through a [2+2] photocycloaddition reaction under sunlight (Fig. 1) [23]. As a continuation of our previous work, we further investigated the chemical composition of Hainan soft coral Sinularia sp., and five polycyclic furanobutenolide-derived norditerpenoids were obtained, including 5-epi-sinuleptolide (3) [24], sinuleptolide (4) [24], 10-epi-gyrosanolide E (5) [23], norcembrene 5 (6) [23], and sinulin D (7) (Fig. 2) [23,25].

Herein, inspired by our previous study of the photoreaction of 1 via a [2+2] photocycloaddition towards 2 (Fig. 1), we report the LSST of 3 to 4 by a Norrish type I reaction, along with the acquisition of a novel product 8 with new 4,14-dioxabicyclo[10.2.1]pentadecane scaffold (Fig. 2), the reaction mechanism study by deuterium experiment and detailed density functional theory (DFT) calculation, as well as the substrate scope examination leading to the late-stage modification of these intriguing natural products and the configuration revision of sinulin D (7).

To explore potantial new photo-products of polycyclic furanobutenolide-containing norditerpenoids, we first irradiated 3 in MeOH under 250 W long arc mercury lamp for 16 h, and a total of five products were obtained, with the yields shown in Table 1. Their structures were determined by 1D and 2D nuclear magnetic resonance (NMR) (the structure of 9 was determined by comparing ¹H NMR with 8 and the structure of 11 was determined by comparing ¹H NMR with 10). It is worth mentioning that 8 and 9 featuring a new 4,14-dioxabicyclo[10.2.1] pentadecane scaffold, which is a nice example of using the LSST strategy to obtain new natural products derivatives. In addition, we consider that in all photo-products, the Δ¹² E/Z isomerization is due to the photoexcitation of Δ¹², on the other hand, the photoexcitation of the C-6 carbonyl group lead to the C-5 chiral epimerization (products 4 and 11) and tetrahydrofuran ring opening (products 8 and 9 in methanol), which was consistent with the Norrish type I reaction. In fact, Norish type I cleavage occurs at the α-position of the excited carbonyl group, which generates acyl and alkyl radicals, followed by a series of reactions [26,27]. We next performed trapping experiment with 2,2,6,6-tetramethylpiperidinooxy (TEMPO) (Table S2 in Supporting information) and confirmed that our reaction involves radical intermediates, which is consistent with the feature of the Norrish type I reaction. Subsequently we investigated the effect of solvent on the reaction, dichloromethane (DCM) was shown to be more likely than MeCN to generate the Δ¹² E/Z isomerization product 10, which was further confirmed by examining the product when the reaction underwent for 2 h (Fig. S1 in Supporting information). We also periodically monitored the ¹H NMR spectra of the reaction when methanol and acetonitrile was used as the solvent (Figs. S2 and S3 in Supporting information). The gradual appearance process of each product within 16 h was visualized.

To further investigate the generation mechanism of 8, we performed a deuterium experiment, and obtained a mixture of the deuterated products 12 and 13. By comparing the ¹H NMR spetra of 12 and 13 with 8 (Fig. 3), the disappearance of H₃–20 at 3.71 ppm (the chemical shift of H₃–20 in 8) in ¹H NMR confirmed that the methoxy group (CH₃–20) in 8 was derived from the solvent methanol. In addition, the integral ratio of the H₂–7 (chemical shift H_a-7: 2.46 ppm and H_b-7: 2.62 ppm) decreased from the original 2 to 1, which confirmed that the proton of methanol was transferred to one of the H₂–7. Meanwhile, the integral ratio of H_b-7 and H_a-7 is 0.67 to 0.32 (2.32 includes the other two hydrogens), which means that the ratio of 12 to 13 is approximately 2 to 1. Furthermore, the 4 mass units more of 12 and 13 (407.1980 [M + Na]⁺) than that of 8 (381.1908 [M + H]⁺) in high resolution electrospray ionization mass spectroscopy (HR-ESI-MS) spectra confirmed that the H₃–20 and H-7_a (or H-7_b) were involved in the conversion process. Therefore, we confirmed the formation of product 8 through a ketene intermediate [28].

Based on the aforementioned experimental result, the isomerization process of 3 under irradiation was proposed (Fig. 4). In detail, the C-6 carbonyl group of 3 was excited under irradiation, and intersystem crossing (ISC) from ¹n−π* to ³n−π*, followed by conformational relaxation to the stable excited state ³Int_1. Then ³Int_1 underwent α-cleavage to form acyl radicals and alkyl radicals by Norrish type I reaction to obtain diradical intermediate ³Int_2, with a Δ¹² E/Z isomerization under irradiation at the same time. Subsequently, ³Int_1 could spontaneously release energy to form 3 and 10. ³Int_2 could undergo C-C rotation to become ³Int_3' easily, then ³Int_3' could be transformed into 4 and 11 through a diradical combination. Furthermore, diradical ³Int_2 underwent an internal hydrogen transfer to produce a ketene intermediate ¹Int_4. ¹Int_4 is with high reactivity, which could easily undergo an addition reaction with methanol, and then, the addition product of enol type would undergo keto-enol tautomerization to yield more stable esters 8 and 9.

To gain mechanistic insights into LSST of natural product 3, DFT and time-dependent DFT (TD-DFT) calculations were performed (Fig. 5) [29]. The TD-DFT calculations performed by using different functionals (M06–2X, B3LYP and CAM-B3LYP), and the excitation types of S₁, T₁ and T₂ were by analyzed electron-hole analysis (Fig. 5A and Fig. S11 in Supporting information) [30–32]. The results showed that Δ¹² and C-6 carbonyl group can be easily excited, which means that they are the first or second triplet excited states, more likely to be obtained from S₁ by ISC [33,34]. We consider that the excitation of Δ¹² is responsible for the isomerization of the Δ¹² of the photo-products, e.g., 9, 10 and 11, and we further investigated the condition after the C-6 carbonyl group excitation (Fig. 5B). The carbonyl group of ³Int_1 underwent α-cleavage to give ³Int_2, this process is controlled by transition structure ³TS_1 with free energy 2.7 kcal/mol above ³Int_1, in the transition state ³TS_1, the spin density was majorly shared by C5, C6, and O21. ³Int_3' and ³Int_2 can be converted quickly by ³Int_2', with activation barrier 6.7 kcal/mol. The radical of ³Int_3' attacks from the Si face to give ³Int_4'. When the two electrons in ³Int_4' spin in reverse to form open-shell singlet specie, the two radicals could directly combine and form 4. For the formation of product 8, the α hydrogen of the acyl radical in ³Int_2 was captured by the alkyl radical to form ³Int_3, the transition state ³TS_2 of this process is with a free energy of 18.7 kcal/mol higher than ³Int_2. In the transition state ³TS_2, the spin density was majorly shared by C5, C6, C7 and O21. Then, the formation of the ketene intermediate ¹Int_4 was accomplished by the release of energy in ³Int_3. Finally, with the aid of methanol, ¹Int_4 could be transformed into the thermodynamic product 8.

In addition, we noted that the Δ¹² excitation of compound 3 may also generate [2+2] photo-products similar to compound 2, for which we also calculated the [2+2] photoreaction of compound 1 to 2 as well as compound 3 (Figs. S5–S9 in Supporting information). The results showed that for compound 1 to form 2, the energy barrier is only 11.1 kcal/mol. However, for compound 3 to form [2+2] photo-product, the only feasible way need to overcome a high energy barrier of 23.5 kcal/mol (see Fig. S9 for detail). On the other hand, the calculations also showed that for compound 3, the formation of [2+2] photo-product has a higher energy barrier compared to the Norrish type I products.

With the clear mechanism of the reaction in hand, we expected to increase the chemical diversity of photochemical products 8 and 9, which are rare macrocyclic ethers. Therefore, the other four natural norditerpenoids 4, 5, 6, and 7 were employed in an attempt to explore the universality of this reaction (Fig. 6A). Compounds 4–6 yield the photochemical ethers 8, 14 and 15. The structures of 8, 14 and 15 were elucidated by 2D NMR spectroscopic data analysis. However, instead of the expected 16, 7 produced 15, whose structure was determined by X-ray diffraction analysis, indicating that the C-8 configuration of the original structure of 7 should be wrong. The X-ray diffraction analysis confirmed that the C-8 absolute configuration of 15 was R, which implied that 6 and 7 should be C-5 epimers. To further verify our hypothesis, quantum NMR calculation of 7 (revised) and 7 (original) was subjected to the GIAO method, at the mPW1PW91/6–31+G(d, p) level, with the PCM model in chloroform (Fig. 6B) [35,36]. The result indicated that the calculated NMR data of (1R*, 5R*, 8R*)-7 (revised) were consistent with the experimental data, with a correlation coefficient R² = 0.9993 and DP4+ probability of 100% [37–39]. Therefore, we proposed that the correct absolute configuration of sinulin D (7) should be 1R, 5R, 8R.

Glucagon-like peptide 1 (GLP-1) is a peptide hormone produced and secreted by intestinal L-cells, which lowering postprandial blood glucose by binding the GLP-1 receptor to stimulate the insulin secretion from beta cells. It is a currently popular target for the obesity and type diabetes related drug discovery [40–43]. In this study, the stimulation of GLP-1 secretion by the compounds 1, 2, 10, 14, and 15 at 20 µmol/L was evaluated and compared to that of INT777, a Takeda G-protein-coupled receptor 5 (TGR5, a GPCR receptor) agonist. Among these five compounds 10 and 15 could stimulate GLP-1 secretion significantly with no cell toxicity, and dose-dependently (Fig. 7 and Fig. S13 in Supporting information). In the mechanism study, compounds 10 and 15 exhibited significant stimulation of GLP-1 secretion by increasing intracellular calcium concentration dependent on the protein kinase A (PKA) pathway (the details bioactivity results are described in Supporting information 1.7.3).

In summary, inspired by our previous study of the photoreaction of 1 via a [2+2] photocycloaddition towards 2, we conducted a photo-induced late-stage skeleton transformation of the co-occurring compound 3 towards compounds 8 and 9 with new 4,14-dioxabicyclo[10.2.1]pentadecane skeleton. The biomimetic conversion of 3 to 4 was achieved meanwhile. A Norrish type I reaction mechanism was proposed based on the deuterium experiments and further confirmed by the DFT calculations. Subsequently, a substrate scope examination produced new skeleton derivatives 14 and 15, and further confirmed that the correct absolute configuration of 7 is 1R, 5R, 8R. All these findings not only enlarged the chemical space, but also gave us a clue for obtaining more new derivatives of bioactive natural products. Intriguingly, compounds 10 and 15 were found to stimulate the intracellular calcium and promote GLP-1 secretion in the PKA-dependent pathway, which could inspire the discovery of new metabolic disease-related drug leads from marine sources. The present work forms the first paradigm for efficient construction photo-induced late-stage skeleton transformation strategy of complex natural products through combination of quataum chemistry calculation and experimental validation, which will give a perspective for efficient LSST of interesting natural products towards new bioactive compounds.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

CRediT authorship contribution statement

Quan Xu: Writing – original draft. Ye-Qing Du: Validation, Data curation. Pan-Pan Chen: Validation. Yili Sun: Validation. Ze-Nan Yang: Investigation. Hui Zhang: Data curation. Bencan Tang: Software. Hong Wang: Methodology. Jia Li: Funding acquisition. Yue-Wei Guo: Funding acquisition. Xu-Wen Li: Funding acquisition.

Acknowledgments

We sincerely appreciate Prof. K. N. Houk at University of California and Dr. Yike Zou at Lawrence Livermore National Laboratory for the helpful suggestions and comments on our manuscript. We are grateful of the National Key Research and Development Program of China (Nos. 2021YFF0502400, 2022YFC2804100), the Natural Science Foundation of China (Nos. 82022069, 81991521, 42076099, 22171153, 81903682), Shandong Laboratory Program (No. SYS202205), Ningbo Natural Science Foundation Programme (No. 2022J171), the CAS Youth Interdisciplinary Team, and Taishan Scholars Program (Nos. tstp0648, tsqn202312302).

Supplementary materials

Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.cclet.2024.110141.