Nonenzymatic epoxide ring-opening reaction of FST C (1)
The α/β hydroxylase Alp1U was previously shown to hydrolyze the epoxide of FST C (1) to afford FST C1 (2) and FST C2 (3) (Fig. 2a, traces i-ii)18,33. In contrast, the Alp1U homologue FlsH is unable to catalyze hydrolysis of the epoxide ring in 1 (Fig. 2a, trace iii)18. A possible explanation is that FlsH might require a protein partner or an exogenous cofactor to catalyze the epoxide hydrolysis. In order to test these hypotheses, several flavin reductases including Fre from E. coli34 were selected to examine their competence as putative protein partners for FlsH. It was found that incubation of FST C (1) with FlsH and Fre in the presence of NADH led to multiple products (Fig. 2a, trace vi). However, further experiments showed that formation of the same set of products was also observed in the absence of FlsH (Fig. 2a, trace vii), suggesting that Fre alone was able to mediate the same decomposition of FST C (1). Moreover, incubation of 1 with several other flavoenzymes including FlsO2 (prejadomycin oxidase)35, TiaM (tiacumicin halogenase)36,37, and XiaK (xiamycin N-hydroxylase)38 also led to the same result (Supplementary Fig. 2). Given that the activities of these flavoenzymes all rely on the reduction of FAD by NADH, the reduced flavin alone was proposed to serve as the actual catalyst of FST C (1) consumption. Indeed, similar products were observed upon incubation of FST C (1) with FAD and NADH in the absence of any enzymes, such that FAD and NADH are both sufficient (Fig. 2a, traces viii & xii) and necessary (Fig. 2a, traces ix‒xi) to facilitate the nonenzymatic decomposition of FST C (1).
Identification of the FST C decomposition products
Reaction products (3 and 7‒11) were isolated from a scaled up reaction of FST C (1) with FAD and NADH and structurally characterized as shown in Fig. 1c. The first product was identified to be FST C2 (3) (Supplementary Fig. 3 and Table 1), which is also a product of the Alp1U-catalyzed hydrolysis reaction33. The second reaction product 7 (Supplementary Fig. 4 and Table 2) was established to have the same planar structure as FST B, an FST naturally isolated from Streptomyces sp. TA-339112 and Streptomyces strain Acta 1383;13 however, the absolute configuration of FST B was not determined12. The (1R,2R,3S) absolute configuration of 7 was established by the solution TDDFT-ECD methodology (Supplementary Fig. 5) and confirmed by single crystal X-ray analysis (Fig. 2b, CCDC 2036399; Supplementary Table 3). The third product 8 (Supplementary Fig. 6 and Table 2) was determined to have the same planar structure as 7 and the (1R,2R,3R) absolute configuration of 8 was assigned by comparison of the experimental and calculated ECD spectra carried out on the (1R,2R,3R) stereoisomer (Supplementary Fig. 7)39. In order to distinguish the two stereoisomers of FST B, 7 and 8 were named FST B1 and FST B2 (Fig. 1), respectively. The product 9 (designated FST C3, Fig. 1) was isolated with the same retention time as that of 3 during in the HPLC analysis with a reversed-phase C18 column. Analysis of the assay using a polar HPLC column allowed separation of these two species (Fig. 2a, trace xii). Finally, 9 was identified to be a stereoisomer of 3 by HRESIMS and NMR data (Supplementary Fig. 8 and Table 4), and the (1R,2S,3S) absolute configuration of 9 was established by TDDFT-ECD methodology (Fig. 2c, Supplementary Fig. 9).
Two additional products 10 and 11 were also identified and structurally characterized (Fig.1c; Fig. 2a, traces vi-viii & xii). The product 10, designated FST C4 (Fig. 1), was found to possess a contracted five-membered A-ring by NMR spectroscopic analysis (Supplementary Fig. 10 and Table 5) and was determined to have the (1R,2R) absolute configuration by single crystal X-ray analysis (Fig. 2b; Supplementary Table 6; CCDC 2129081). The product 11, designated FST C5 (Fig. 1), was structurally elucidated to have a bridged and expanded seven-membered A-ring (Supplementary Fig. 11 and Table 5). Assignment of the (1R,2R,3R) absolute configuration to 11 was based on its origin from 1 but otherwise remains tentative, because a definitive experimental characterization was not possible on account of its instability. Currently, there are no reports of 9‒11 having been isolated from a natural source.
Labeling studies of the epoxide ring-opening reaction
To investigate the mechanism of the nonenzymatic epoxide ring opening reaction, FST C (1), FAD and NADH were coincubated in PBS buffer (pH 7.0) prepared with deuterium oxide (2H2O). Incorporation of a single deuteron into both 7 and 8 was indicated by the presence of a + 1 Da-shifted molecular ion peak at m/z 326.7 ([M ‒ H]‒) for both reduced products (Supplementary Fig. 12). In contrast, no changes in the molecular weights of 3, 9, 10 or 11 (Supplementary Fig. 12) were observed. Subsequently, 7–2H and 8–2H were isolated from scaled up reactions and analyzed by 1H and 13C NMR (Supplementary Figs. 13, 14 and Table 2). Comparison of the 1H NMR spectra of 7–2H and 7 revealed that the 3-H proton signal of 7 had vanished in 7–2H and meanwhile the 3-Me doublet of 7 had collapsed to a singlet in 7–2H (Fig. 3a), which located 2H at C-3 of 7–2H. In support of this assignment, the COSY correlation between the H-2 and H-3 protons observed in 7 was absent in 7–2H (Supplementary Figs. 4, 13). Likewise, a singlet was also observed for the 3-Me along with the absence of any COSY correlation between H-2 and H-3 in 8–2H (Fig. 3a; Supplementary Figs. 13, 14). Moreover, the 13C NMR spectra of 7–2H and 8–2H both showed broadening and reduced intensity of the signals from C-3, and 2H‒13C couplings at C-3 were also observed (Supplementary Figs. 13, 14)40. These observations demonstrate that a single solvent hydron is incorporated at C-3 during the reduction of 1 to both 7 and 8.
When FST C (1) was incubated with NADH and FAD under 18O2 (Fig. 3b, traces i−iii), LC-MS analysis of the reaction products revealed an increase of +2 Da in the molecular mass of the oxidized products yielding 3–18O, 9–18O, 10–18O and 11–18O (Fig. 3a; Supplementary Fig. 15). These data indicate that the conversion of 1 to 3, 9, 10, and 11 involves incorporation of one oxygen atom from O2. Furthermore, the relative production of 3/9 and 10/11 versus 7/8 was O2 dependent, because incubation of FST C (1) with FAD and NADH after saturating the reaction buffer with O2 led to a significant increase in the levels of the former (3, 9‒11) versus the latter (7 and 8); in contrast, saturating the reaction buffer with N2 led to 7 and 8 as the dominant products (Fig. 3b, traces iii and iv). To locate the exact position of 18O incorporation during the reaction (1 → 3/9), a scaled up reaction was performed by incubating 1, FAD, and NADH in 18O2-saturated PBS buffer (pH 7.0). Two products 3–18O ([M ‒ H]‒ m/z 343.0718) and 9–18O ([M ‒ H]‒ m/z 343.0722) (Fig. 3a; Supplementary Fig. 16) were isolated. NMR analysis showed that the 18O label was incorporated at C-3 of 9–18O (Supplementary Fig. 17). Specifically, comparison of the 13C NMR data for 9 and its 18O isotopolog 9–18O demonstrated a 3.8 Hz upfield shift of δC-3 in the latter relative to that of 9 (Supplementary Fig. 18 and Table 4), supporting the 18O label at C-3 (i.e., 9–18O)41,42.
Probing parameters affecting the epoxide ring opening reactions
The pH dependence of the FAD/NADH-dependent decomposition of 1 was subsequently investigated by performing the reaction in buffers over the pH range 3‒10. Following the incubation of 100 μM 1 with 100 μM FAD and 10 mM NADH for 2 h at 30 °C, 1 was more efficiently converted to the epoxide ring-opened products with buffers in the pH range 5‒7; however, the reaction was less efficient under more alkaline conditions at pHs 9–10 (Supplementary Fig. 19). FST B1 (7) was rather stable at pH 3–7; however, it readily converted to 8 at higher pH (Supplementary Fig. 20), where it also underwent oxidation to yield the previously reported natural product FST A (12) (Fig. 3b, trace v; Supplementary Fig. 20 and Table 1)12. FST B2 (8) was also stable at pH 3–6, and readily converted to 7 and 12 in more alkaline buffers (Supplementary Fig. 20). In contrast, FST C (1) was highly stable at pH 3‒10, even in the presence of FAD or NADH alone (Supplementary Fig. 21) again demonstrating that decomposition of the epoxide in 1 requires the presence of both FAD and NADH.
A time course analysis of 100 μM 1 with 100 μM FAD and 10 mM NADH demonstrated the transient formation of an intermediary species 13 that vanished after longer incubation (Supplementary Fig. 22). A fixed initial concentration of 1 (100 μM) was also incubated with varying concentrations of FAD (1, 10, and 100 μM) and NADH (2, 5, and 10 mM) for 30 min at 30 °C (Supplementary Fig. 22). The lifetime of the transient intermediate was significantly extended at decreased levels of FAD (Fig. 3b, trace vi). This allowed collection of the intermediate 13 followed by immediate reinjection on the HPLC demonstrating its conversion to 7 and 12 (Fig. 3b, trace vii). When the isolated intermediate 13 was immediately incubated with FAD under O2, the production of 7 and 12 was again observed (Fig. 3b, trace viii); however, if NADH was also included in the incubation, then 3 was instead observed as the major product with 12 as the minor product, and 7 was no longer observed (Fig. 3b, trace ix). The UV absorbance spectrum of the intermediate 13 was highly similar to those of 1, 3, and 7, and LC-MS analysis of the intermediate indicated an m/z 325.7 for the [M–H]– ion (Supplementary Fig. 23). Based on these measurements, the intermediate is tentatively proposed to be the enol 13 (Fig. 3b), which is essentially a tautomer of 7 and 8 and thus has the same oxidation state as these two products.
While several attempts were made to isolate the intermediate in order to confirm its structural assignment as 13, only 7 was obtained as the major product along with another minor compound 14. This is consistent with the poor stability of the intermediate and its facile conversion to 7 (Fig. 3b, trace vii, Supplementary Fig. 24). Nevertheless, 14 could be characterized and was thus identified as the C-3 peroxide designated hydroperoxyfluostatin (Fig. 3a; Supplementary Fig. 25 and Table 5); however, poor recovery of 14 hampered collection of sufficient spectra for a complete characterization of its absolute configuration at C-3. Hydroperoxyfluostatin (14) was stable in water, even in the presence of FAD (Fig. 3b, traces x and xi). While the coincubation of 14 with NADH alone led to multiple products 3, 9, 10 and 11, the inclusion of FAD with NADH did not change the product profile (Fig. 3b, traces xii and xiii; Supplementary Fig. 26). Given the observed conversion of 14 to both 3 and 9, compound 14 is likely a mixture of the (1S,2S,3R) and (1S,2S,3S) stereoisomers. The fact that isolation of 13 afforded a minor amount of 14 suggests that 14 is another intermediate that coelutes with 13 as opposed to being derived from 13. Furthermore, while 3/9 could be generated from both 13 and 14, 7/8 were only produced from 13, whereas 10/11 were exclusively derived from 14 (Fig. 3a).
When other flavin-related cofactors were considered, FMN and riboflavin in place of FAD were also found to efficiently mediate the decomposition of 1 in the presence of NADH (Fig. 3b, traces xiv-xvi). Isoalloxazine and NADH were also able to facilitate the formation of 3 from 1, albeit with a much lower efficiency (Fig. 3b, trace xvii). Only the products 3, 9‒11 were observed using the cofactor F420 in the presence of the reducing system constructed from glucose 6-phosphate and F420-dependent glucose-6-phosphate dehydrogenase (FGD)43,44,45, but notably no production of the reduced species 7 and 8 was observed (Fig. 3b, trace xviii). The observation of oxygen incorporation in the presence of F420 and NADH was unexpected, because reduced 5-deazaflavin has generally be reported to react very slowly with molecular oxygen46,47. Nevertheless, the results with flavin cofactor compatibility shown in Fig. 3b (traces xiv-xviii) were found to be reproducible (Supplementary Fig. 27). NADPH was also found to be an effective cofactor similar to NADH and thus support the epoxide ring opening reactions (Supplementary Fig. 19).
Generalization of the epoxide ring-opening reaction
To explore the generality of this reaction, an array of FST derivatives 15‒21 (Fig. 4) were tested as possible oxirane reactants. Although FST F (15) is inert to Alp1U18,33, it could be efficiently converted in the presence of FAD and NADH to 1-O-methyl-FST C2 (22), 1-O-methyl-FST B1 (23), 1-O-methyl-FST C3 (24), and 1-O-methyl-FST B2 (25) (Fig. 4 and Supplementary Fig. 28), which were structurally characterized by NMR and ECD spectroscopic analysis (Supplementary Tables 7, 8 and Figs. 29‒32), as well as single crystal X-ray diffraction analysis (e.g. 23, CCDCC 2036400, Supplementary Table 3).
Similarly, 7-O-methyl-FST C (16), 6-O-methyl-FST C (17), FST D (18), FST S (19), and difluostatin H (DiFST H, 20) could also undergo epoxide-ring opening reactions upon treatment with FAD/NADH to produce a set of multiple products related to 3, 7, 8, and 9 (Fig. 4, Supplementary Figs. 33‒37). The major products from 16 and 17 were structurally characterized by 1D and 2D NMR as well as ECD spectroscopic analysis as FST M (26)17, and 6-O-methyl-FST B1 (27), respectively (Fig. 4, Supplementary Table 9 and Figs. 38, 39). The structures of the other products from reactions with 18‒20 were deduced based on LC-MS analyses. Notably, FST Q (21), which differs from 16 in terms of a C-4 hydroxyl instead of a C-4 carbonyl, was unreactive with FAD/NADH (Supplementary Fig. 40) highlighting the essential role of the C-4 carbonyl in this chemistry.
The epoxides 28–37 (Fig. 4) were also tested as putative oxirane substrates. No activities were observed with 28–35 (Supplementary Figs. 41‒46)16,40. In contrast, treatment of auxarthrol H (36)48 with FAD/NADH led to multiple products (Supplementary Fig. 47), one of which was determined to be 38, which is a structural analog of auxarthrol F (Fig. 5; Supplementary Fig. 48 and Table 10)49, by comparison with the experimental and calculated ECD spectra of 36 and 38 (Supplementary Fig. 49). Reaction with menadione 2,3-epoxide (37) yielded three products (Supplementary Table 11 and Figs. 50‒53) including 3-hydroxy-3-methyl-2,3-dihydronaphthalene-1,4-dione (39), 2-hydroxy-3-methyl-1,4-naphthoquinone (40, phthiocol, characterized by X-ray diffraction, CCDC 2036401; Supplementary Table 12), and 3-methylnaphthalene-1,4-dione (41).
Two common structural features shared by all compounds that can react with FAD/NADH to undergo the epoxide ring opening reactions include (i) the presence of a carbonyl adjacent to the epoxide and (ii) the presence of an aromatic moiety adjacent to the carbonyl/epoxide pair (Fig. 4). The absence of a carbonyl group neighboring the epoxide ring in 21 and 28–31 may account for the observed lack of reactivity. While 36 and 37 underwent ring opening in the presence of FAD/NADH, no reaction was observed for 32–35 despite the presence of an epoxide/carbonyl pair in these compounds. This implies that conjugation of the carbonyl/epoxide pair to an aromatic system is also necessary for ring opening of the epoxide.
The commercial compound vitamin K1 2,3-epoxide (42) was also tested as a putative substrate, which features both an epoxide/carbonyl pair and an adjacent aromatic ring. However, incubation of 42 with FAD/NADH led to no changes in either its TLC retention profile or spectroscopic properties implying no reaction (Supplementary Fig. 54). A recent study demonstrated that the flavoenzyme RslO5 catalyzed a hydride-mediated reductive epoxide ring opening reaction to convert 43 to 44 during rishirilide biosynthesis (Fig. 4)11. The reaction proceeded in a manner that was highly similar to the conversion of 1 to 7. However, the production of 44 from 43 was confirmed to be strictly RslO5-dependent, and 43 was unreactive in the presence of FMN and NADH11. A possible explanation is that the presence of a long side chain next to the epoxide in both 42 and 43 might hinder both approach and orientation of the flavin thereby preventing suitable interactions for ring opening (Supplementary Fig. 55). Two additional commercial compounds trans-1,3-diphenyl-2,3-epoxypropan-1-one (45) and chalcone α,β-epoxide (46) were found to be unreactive with FAD/NADH; however, 46 but not 45 was accepted as a substrate by RlsO5 to yield the single product 47 (Supplementary Table 13 and Figs. 56, 57), which indicates that RslO5 is specific for only one of the two enantiomers. Deuterium incorporation into 47 was not observed from the RslO5 reaction in buffered 2H2O but was indeed observed when carrying out the RslO5 reaction in a coupled assay either with the glucose dehydrogenase from Bacillus megaterium DSM 2894 (BmGDH) to provide (S)-[4-2H]NADH50, or with the glucose dehydrogenase from Thermoplasma acidophilum ATCC 25905 (TaGDH) to provide (R)-[4-2H]NADH51, respectively, employing d-[1-2H]glucose as the deuterium donor (Supplementary Fig. 58)52. Previous studies demonstrated that the RslO5-H172N mutant had a deficiency in FMN binding and was inactive with the native substrate 4311, suggesting the critical role of FMN in the RslO5-catalyzed reaction. Taken together, a hydride from FMNH2 originating from NADH is proposed to be the effective nucleophile for RslO5-catalyzed epoxide opening reactions. In contrast, no deuterium incorporation into 7 was observed in reactions of 1 and FAD with either (S)-[4-2H]NADH or (R)-[4-2H]NADH generated as described above (Supplementary Figs. 58, 59). Therefore, nonenzymatic reduction of 1 to 7 in the presence of FAD/NADH differs from that of RslO5 catalysis with respect to the ultimate origin of the C-3 hydrogen.