A neoclerodane orthoester and other new neoclerodane diterpenoids from Teucrium yemense chemistry and effect on secretion of insulin

Teucrium yemense, a medicinal plant commonly grown in Saudi Arabia and Yemen, is traditionally used to treat infections, kidney diseases, rheumatism, and diabetes. Extraction of the dried aerial parts of the plant with methanol, followed by further extraction with butanol and chromatography, gave twenty novel neoclerodanes. Their structures, relative configurations and some conformations were determined by MS and 1-D and 2-D NMR techniques. Most were fairly conventional but one contained an unusual stable orthoester, one had its (C-16)–(C-13)–(C-14)–(C-15) (tetrahydro)furan unit present as a succinic anhydride and one had a rearranged carbon skeleton resulting from ring-contraction to give a central octahydroindene bicyclic core, rather than the usual decalin. Mechanisms are proposed for the biosynthetic formation of the orthoester and for the ring-contraction. Four novel neoclerodanes increased the glucose-triggered release of insulin from isolated murine pancreatic islets by more than the standard drug tolbutamide, showing that they are potential leads for the development of new anti-diabetic drugs.


Results and discussion
The dried aerial parts of the plant were defatted and extracted with methanol (MeOH). This solvent was evaporated and the residue was extracted with EtOAc, then extracted with BuOH. The BuOH extract was separated by column chromatography on silica gel. Radial chromatography and HPLC yielded twenty pure compounds (Fig. 1). Their structures were elucidated using 1D and 2D nuclear magnetic resonance (NMR) and high-resolution electrospray ionisation mass spectrometry (HRESIMS) data. Their absolute configurations cannot be confirmed from these data but are assumed on the basis of precedent for related compounds 9 (Table S1, Supplementary Information (SI)) showed 22 discrete resonances: 2 × CH 3 , 6 × CH 2 , 7 × CH, 7 × C q . The core structure was shown to be a decalin and related to the neoclerodane diterpenoids 9,15,16,25 . The infra-red (IR) spectrum showed an OH (3536 cm −1 ) and one γ-lactone carbonyl peak (1761 cm −1 ).
The NMR spectra showed, in addition to peaks for this major compound, a full set of peaks for a minor component, with similar chemical shifts and multiplicities. This minor set of peaks integrated for ca. 10% of the major compound present. As the sample gave only one pure peak on HPLC, we ascribed these peaks to a minor diastereoisomer in slow equilibrium with the major diastereoisomer. These are likely to be epimers at the lactol hemiacetal C-20. This assignment was supported by the largest differences in chemical shift between the epimers being for H-3 (Δδ 0.1 ppm), H-14 (Δδ 0.05 ppm), and H 2 -19 (Δδ 0.2), as these four protons are close in space to the epimeric C-20. We assign the structure 2 ( Fig. 1)  www.nature.com/scientificreports/ is trans-fused and that ring A is in the chair conformation. The conformation of ring B is less clear, owing to overlap of 1 H NMR signals but MM2-minimisation suggested that it may be a flattened boat. Acetal carbon C-6 was characterised by its chemical shift (δ 110.0) and by HMBC cross-peaks to both H-19. Further HMBC correlations were seen to the doublets at δ H 3.90 and δ H 4.45, showing that they were due to H 2 -18 and demonstrating the tetrahydrofuran ring. The acetal at C-6 was identified by observation of an OMe (δ H 3.40, δ C 48.9), linked by HMBC to C-6. These data characterise the structure of 3, fatimanol G ( Fig. 1). This structure is identical to that of teulepicephin 22, with the exception of the acetal (hemiacetal in 14) 15 (Table S2, SI) contained two very similar sets of signals, in ca. 1:1 ratio, suggesting diastereoisomers which interconverted slowly on the 1 H NMR timescale. This may indicate a cyclic hemiacetal (cf. 2). Detailed assignment of the signals was challenging, as many overlapped between the two stereoisomers. Taken together, the NMR data showed that 4 had a neoclerodane core. As for 1-3, the upper part was an aromatic furan, with 1 H NMR signals at δ 6.49 (H-14), δ 7.55 (H-15) and δ 7.61 (H-16). The signals for H-15 and H-16 were distinguished by a NOESY cross-peak from the former to H-14. The 13 C NMR signals for this ring were at δ 124.7 (C-13), δ 107.9 (C-14), δ 139.6 (C-15) and δ 144.3 (C-16), as identified by HSQC and HMBC. The spiro-lactone was initially identified by the chemical shift of H-12 (δ 5.59), corresponding to a benzylic ester. This showed an HSQC cross-peak to C-12 (δ 72.3) and HMBC cross-peaks to C-13, C-14, C-16, and C-11 (δ 40.2). HSQC then identified δ 2.73 dd as being due to one H-11 and the two signals at δ 2.56 and δ 2.57 (both dd, each integrating for 0.5 H) as due to the other H-11. The 13 C signal at δ 176.4 was shown to be the lactone carbonyl C-20 by HMBC to H-11 (δ 2.73), H-8 (δ 2.2), and H-10 (δ 2.85). H-3 (δ 4.41) and C-3 (δ 60.5) had the expected downfield chemical shifts arising from the OH. COSY then identified H-2 at δ 1.66 and δ 2.10, with HSQC showing C-2 (δ 21.6). Further COSY cross-peaks then showed the resonances for H-1 at δ 1.64 and δ 2.3. The lower fused butenolide became evident through HMBC cross-peaks from H-10 to the alkene C q peaks for C-4 (δ 128.5) and C-5 (δ 163.0). These were distinguished by their chemical shifts and by observation of HMBC from C-4 to H-2 (δ 2.01). The carbonyl C-18 was at δ 170.1. The lactone was completed by C-6 (δ 102.4), which shows HMBC correlations with both H 2 -7, H-8 and H-10. Interestingly, there is also a weak 4-bond HMBC cross-peak between H 3 C-17 and C-6. The overlap of many of the 1 H NMR signals for the two diastereoisomers precluded detailed assignments of the conformations of the decalins. We assign structure 4 ( Fig. 1) to this compound, fatimanol H. . The 13 C NMR spectrum contained signals for 21 discrete carbons: 2 × CH 3 , 4 × CH 2 , 7 × CH, 8 × C q . The IR showed absorbances for OH (3618 cm −1 ) and three carbonyls (1763, 1715, 1701 cm −1 ). The NMR spectra (Table S2, SI) were very similar to those for 4, with the exception of additional methyl signals at δ H 2.05 / δ C 20.9, an additional C = O signal at δ C 172.1 and a marked downfield change in the chemical shifts of H-3 (δ 5.59, Δδ 1.18 ppm) and C-3 (δ 64.65, Δδ 4.16 ppm). These indicate that 5 is the 3-O-acetate ester of 4. The 1 H NMR spectrum contained only one set of signals, showing that one of the possible hemiacetal diastereoisomers had significantly lower energy than the other but it was not possible to determine which from the spectroscopic data. The methyl protons (H-2′, δ 2.05) of the acetate showed a strong HMBC to the corresponding ester carbonyl (C-1′, δ 170). C-1′ also showed a cross-peak to H-3, confirming the location of the ester. We assign structure 5 ( Fig. 1) 13 C NMR spectrum showed discrete signals for 2 × CH 3 , 4 × CH 2 , 7 × CH and 7 × C q . The 1 H and 13 C NMR spectra (Table S2, SI) (with 2D spectra) were very similar to those for 4, with the addition of signals for OCH 3 (δ H 3.19, δ C 51.0). These protons showed a strong HMBC to the 13 C signal for C-6 (δ 150.8), identifying the methoxy group as being part of an acetal at this position. These data show that 6 (fatimanol J, Fig. 1) has the structure shown.  10 . The 13 C NMR spectrum showed discrete resonances for 20 carbons: 1 × CH 3 , 6 × CH 2 , 6 × CH, 7 × C q . The IR contained absorbances for two carbonyls (1769 cm −1 , 1737 cm −1 ) and OH (3433 cm −1 ). The NMR data (Table S2, SI) suggested a neoclerodane core. The upper part contained a hydroxyfuranone ring, as in fatimanol A 15 . This cyclic acetal was represented by H-14 (δ 7.17), which showed an HSQC cross-peak to C-14 (δ 146.6); the downfield shifts of these signals were due to the enone system. From H-14, 2-bond HMBC cross-peaks were seen to C-13 (δ 108.1) and to C-15 (δ 99.0) and 3-bond cross-peaks were evident to the carbonyl C-16 (δ 172.2) and (weakly owing to the adverse dihedral angle) to C-12 (δ 64.0). H-14 did not show a COSY cross-peak to the hemiacetal proton H-15 (δ 5.88), owing to the small coupling constant consequent to a dihedral angle ca. 90°. HSQC identified H-12 as part of a complex multiplet at δ ca. 4.6. A 2-bond HMBC from H-12 then led to C-11 (δ 41.3), from which HSQC showed the two signals for H-11 (δ 1.6, δ 2.25). From the downfield H-11 signal, HMBC identified the signals for C-9 (δ 50.2), C-10 (δ 43.9) and the www.nature.com/scientificreports/ lactone carbonyl C-20 (δ 174.6). This lactone was confirmed by HMBC cross-peaks from C-20 to H-19 (δ 4.58, δ 4.64) and the latter were linked on to C-4 (δ 85.6), C-5 (δ 49.7) and C-6 (δ 108.2). A 2-bond HMBC correlation from C-4 indicated H-3 (δ 3.83). The latter then gave an HMBC cross-peak to the methylene C-18 (δ 75.6) with H-18 (δ 3.97 d (J = 10.4 Hz), δ 4.35 d (J = 10.4 Hz)). The downfield H-19 also gave a 3-bond HMBC with C-10, completing the lower lactone ring. H-3 resonated as a dd (J = 11.6, 5.9 Hz), the larger of the two coupling constants indicating that this proton is axial. Completing the features of the lower part of the structure, H-18 (δ 4.35) showed a 3-bond HMBC with the hemiacetal carbon C-6. We assign the structure 7 ( Fig. 1)  , with its C-3 (δ 66.5).
The relative stereochemical configurations were largely determined by use of NOESY. An MM2-minimised structure suggested that the decalin would have both rings in chair conformations. Strong NOESY cross-peaks showed that H-17 and H-20 were cis on the lower face. Similarly, H 2 -19 were shown to be on the lower face by strong NOESY cross-peaks to H-20, with these methylenes being diaxial on ring B. Further strong cross-peaks from both H-19 to H-3 confirmed the latter as axial down on ring A and a cross-peak to H-1 (δ 2.1) also suggested that this was on the lower face. On the upper face, the downfield H-18 signal (δ 3.21) gave a cross-peak to H-6 (δ 3.78), which allowed differentiation of the two oxirane proton signals. The relative configuration at C-15 could not be determined. These data show the structure of 8 (Fig. 1) 11 . The 13 C NMR spectrum complied, with 25 discrete signals: 4 × CH 3 , 7 × CH 2 , 7 × CH, 7 × C q . The IR showed OH (3546 cm −1 ) and three carbonyls (1764, 1752, 1708 cm −1 ). The NMR spectra (Table S3, SI) showed considerable similarity to those for 8, except in the C-11 / C-12 region and the upper methoxyfuran. The structure of the trans-decalin and the lower appendages were identical to those of 8. HMBC from C-11 (δ 38.0) identified H-12 (δ 4.56); the downfield chemical shift of this peak indicated a hydroxy group. HSQC identified C-12 (δ 63.8). H-12 was also linked by COSY to both H-11 (δ 1.75, δ 1.80). A strong 3-bond HMBC from H-12 showed the signal at δ 144.8 to be due to C-14, with HSQC identifying H-14 (δ 7.13). Appropriate HMBC and HSQC correlations then identified C-13 (δ 144.2), C-15 (δ 104.2), H-15 (δ 5.88) and C-16 (δ 171.7). The methoxy group protons resonated as two singlets (δ 3.54, δ 3.55), each integrating for 1.5 H, suggesting that 9 was a mixture of epimers at C-15. The corresponding methoxy 13  The configuration at C-12 could not be determined. The structure is thus 9 (Fig. 1), fatimanol M. www.nature.com/scientificreports/ 1 × CH 3 , 7 × CH 2 , 5 × CH, 7 × C q . Three C q were carbonyls, with two coincident 13 C NMR signals at δ 174.5 and a singleton at δ 173.1, and IR bands at 1795, 1790, and 1689 cm −1 . The higher-frequency C = O absorptions suggested that they were likely to be a cyclic anhydride. As with other examples, the NMR data (Table S3, SI) suggested the neoclerodane skeleton for 10. However, the upper five-membered ring was unusual, in that it was a succinic anhydride. Carbonyl C-15 resonated at δ 174.5 and C-16 appeared at δ 173.  3) and C-3 (δ 73.2). Thus 11 has the novel polycyclic structure shown in Fig. 1, with many fused rings and bridges. Fortunately, these fusions and bridges make the structure fairly rigid and it was straightforward to assign the relative stereochemical configurations by use of coupling constants (Karplus relationship) and NOESY. Figure 2 shows the key NOESY interactions used in this assignment. Particularly useful was the NOE interaction between the methoxy protons and H-18 exo (δ 3.89), which confirms that the methoxy is on the lower face of the trans-decalin. The name fatimanol O is assigned to the novel compound 11 (Fig. 1  Completing ring A, the upfield H-1 gave a strong COSY with H-10. The structure of 12 is less rigid than 11 but it was still possible to assign its relative configurations by NOESY (Fig. 2). These data confirm the structure of 12, fatimanol P (Fig. 1).  1). HSQC then linked these to the 13 C peaks at δ 109.7 (C-14) and δ 143.5 (C-15, C-16). HMBC from H-14, H-15, and H-16 then identified the signal at δ C 132.4 as being due to C-13. HMBC from C-13, C-14 and C-16 to the 1 H signal at δ 4.68, along with HSQC from C-12, identified this multiplet signal as H-12. This benzylic proton gave COSY cross-peaks to 12-OH (δ 5.02, d) and both 11-H (δ 1.60, δ 2.01, both dd). HSQC from the latter then gave 11-C (δ 41.7). Linkage of this upper side-chain to the core bicycle was demonstrated by a HMBC correlation from H-12 to C-9 (δ 40.4, C q ) and by HMBC from both H-11 to C-8 (δ 35.6) and to C-10 (δ 51.5). The 3-bond HMBC cross-peak between H-11 (δ 1.60) and C-8 was weak, as to the dihedral angle (H-11)-(C-11)-(C-9)-(C-8) is close to 90°. H-8 (δ 1.47) also gave a HMBC with C-11. The other side-chain at C-9 was a hydroxymethyl (HOCH 2 -) unit, with C-20 (δ 53. . HSQC revealed C-7 (δ 37.9) and thence H eq -7 (δ 1.47). Both H-7 gave strong 2-bond HMBC correlations with C-6 (δ 69.4), from which HSQC showed H-6 at δ 3.36. The COSY cross-peak from H ax -7 to H-6 was strong, whereas that from H eq -7 to H-6 was much weaker, suggesting that the (H eq -7)-(C-7)-(C-6)-(H-6) dihedral angle was ca. 90° and thus that H-6 was axial. COSY linked H-6 to HO-6 (δ 4.46). Three-bond HMBC from both H-7 located quaternary C-5 (δ 59.8). Ring B was completed by observation of a 2-bond HMBC from C-5 to H-10 (δ 2.07). www.nature.com/scientificreports/ A 2-bond HMBC from H-10 identified C-1 (δ 24.6), from which HSQC showed that the H-1 signals were at δ 1.87 and δ 1.98. A strong 3-bond HMBC linked C-10 with one H-2 signal (δ 1.87), whereas the cross-peak to the other H-2 (δ 1.27) was weaker. A 3-bond HMBC was also seen linking H-1 (δ 1.27) to C-5; the 3-bond path between these two nuclei cannot pass through C-1 and C-10, thus ring A must be five-membered. Quaternary C-3 (δ 57.7) was identified through a 2-bond HMBC with H-2 (δ 1.27) and a 3-bond correlation with H-10, completing the cyclopentane. Three-bond HMBC from both H-2 showed the methylene C-4 (δ 62.5) as being attached at C-3 and the H-4 protons were observed as dd at δ 3.27 and δ 3.66. These H-4 signals were linked by COSY to each other and to HO-4 (δ 4.46). A 2-bond HMBC between H-4 (δ 3.27) and C-3 confirmed the attachment of the HOCH 2 -. 2-bond HMBC from C-3 and a 3-bond cross-peak from H-2 (δ 1.27) revealed the other substituent at C-3 by identifying C-18 (δ 102.0). The hemiacetal was confirmed by COSY from H-18 (δ 4.76) to HO-18 (δ 6.26). HMBC from C-5, C-6 and C-10 to the doublet signals at δ 3.80 and δ 3.87 identified both H-19. C-19 (δ 67.3) was shown to be a CH 2 by HSQC and 135DEPT. The chemical shifts of C-19 and both H-19 suggested the attachment of an oxygen but this was not an OH (no COSY cross-peak). However, HMBC linked C-19 with H-18 and C-18 with both H-19. The only structure consistent with these connectivities is the cyclic acetal shown. It was not possible to obtain a good NOESY spectrum, so the relative configuration shown in Fig. 1 is speculative, except where suggested by 3 J H-H coupling constants. The NMR spectra contained a second (smaller) set of peaks, which we ascribe to the presence of a minor diastereoisomer in slow equilibrium, probably the epimer at C-18. Thus 13 (fatimanol Q) has the structure in Fig. 1 13 C signals for C-13 (δ 132.8), C-14 (δ 109.6), C-15 (δ 143.6) and C-16 (δ 138.6) all duly linked by HSQC and HMBC. Strong 3-bond HMBC cross-peaks were evident from H-12 (δ 4.59) to C-14 and C-16. HSQC linked this proton to C-12 (δ 61.7), which also gave HMBC to H-14 and H-16. A 2-bond HMBC from H-12 identified C-11 (δ 39.7), from which both H-11 (δ 1.67, δ 1.86) were identified by HSQC. HO-12 (δ 4.90) was located through a 3-bond HMBC from C-11. A 3-bond HMBC from H-12 identified C-9 (δ 43.3), while a similar correlation from H-11 (δ 1.67) confirmed C-10 (δ 46.8) and thence H-10 (δ 1.96). From here, C-20 (δ 63.0) was located by HMBC to H-10; the H-20 protons were approximately co-incident at δ 3.28, with HMBC cross-peaks to C-9 (weak, 2-bond), C-10, and C-11. Two-bond HMBC linked C-10 to both H-1 (δ 1.75, δ 2.00), from which C-1 (δ 21.5) was identified by HSQC. The upfield H-1 signal gave a weak 2-bond HMBC with C-2 (δ 34.5), whereas the downfield signal correlated strongly with C-3 (δ 65.0). The signal at δ 1.17 was a dq (J = 4, 11 Hz), indicating that this was due to H-2 ax , while H-2 eq resonated as a narrow multiplet at δ 1.91. Thus the coupling to H-3 (δ 3.80) shows that this proton is axial and HO-3 is equatorial . Furthermore, both H-3 and H-10 are axial, showing that ring A is in the chair conformation. A 2-bond HMBC from H-3 identified C-4 (δ 70.0) and a 3-bond correlation revealed C-18 (δ 43.0). H 2 -18 resonated as doublets at δ 2.83 and δ 3.06; the chemical shifts suggested the spiro-oxirane ring. Strong HMBC from these protons were observed to C-4 and C-5 (δ 45.47). An AcOCH 2 -group was present at C-5, as demonstrated by HMBC from both H-19 (δ 4.40, δ 4.57) to C-4 and C-5. These protons also correlated with the ester carbonyl MeCO 2 -19 (δ 170.8), with the adjacent methyl group (MeCO 2 -19) resonating at δ H 2.01/ δ C 21.5. Three-bond HMBC cross-peaks were also seen from both H-19 to C-6 (δ 73.3) and HSQC then located the H-6 signal at δ 3.62 (brd, J = ca. 11 Hz). The H-6 signal was better resolved when the 1 H NMR spectrum was obtained on a solution in CDCl 3 , which showed it as δ 3.70 (dd, J = 10.0, 6.0 Hz). The larger axial-axial coupling shows that H-6 is axial on the trans-decalin. Both H-7 (δ 1.39, δ 1.52) were located both by HSQC with C-7 and by HMBC with C-6, whence C-7 (δ 34.7) was revealed by HSQC. H-7 also formed HMBC correlations with the methyl C-17 (δ 17.0), from which H-17 was identified at δ 0.82. Completing ring B, H-17 gave an HMBC cross-peak with C-9. These spectroscopic interpretations were aided, in part, by 1 H, COSY and HSQC spectra of a solution in CDCl 3 , which were better resolved, although paucity of sample precluded identification of the quaternary carbons (Table S4). The coupling constants confirmed of the trans-decalin structure and the relative configurations of most of the substituents, although the relative configuration at C-12 could not be established. We assign the structure shown in Fig. 1 to 14, fatimanol R.   The closure of the decalin system was confirmed by a weak HMBC from C-2 to H-10 (δ 1.80) and a strong peak from C-4 to H-10. The above spectroscopic assignments were aided by 1 H, COSY, HSQC, and HMBC spectra of a solution in CDCl 3 , which were better resolved, although shortage of sample precluded identification some quaternary carbons. The relative configurations corresponded to those of most of the neoclerodanes, particularly 15, as demonstrated largely by 1 H NMR coupling constants. The structure shown in Fig. 1 (Table S6, SI) allowed assignment of all the signals and confirmed that the overall structure was similar to a conventional neoclerodane. However, C-6 was quaternary but not a carbonyl (cf. Examination of a molecular model showed that this methyl group, if on the lower face, would be held in the anisotropic shielding zone of the C-20 carbonyl; thus it is likely to be located on the lower face. The configuration at the other acetal C-18 could not be determined, although it was clear that only one diastereoisomer was present. A 1 H NMR spectrum of 18 was also obtained in CDCl 3 , which was consistent with the structure determined above. The novel structure shown in Fig. 1 was assigned to 18 , and for C-5 (δ 46.9), reflecting changes in steric effects in that region and the absence of through-space effects from an ester carbonyl. Thus 20 (fatimanol X) was the desacetyl analogue of 19, as shown in Fig. 1.
The occurrence of orthoesters in plant natural products was reviewed by Liao et al. in 2008 28 . Orthoacetates have been reported in only a limited number of frameworks, principally the daphnane diterpenoids, phragmalin limonoids, bufadienolide and ergostanoid steroids. Several of these orthoacetates have potent biological activities, including systemic toxicity. In each case, the three oxygen atoms of the orthoacetate unit were situated appositely in space for formation of the orthoester and the rigidity of the framework of the diterpene contributed to the stability of this usually highly acid-labile functionality. We propose the mechanism shown in Fig. 3 for formation of the orthoacetate in 1. Proposed intermediate 23 is 19-acetylteulepicin, reported by Savona et al. 16 to be a secondary metabolite in T. buxifolium and the formal oxidation product of 2 at the lactol. It is therefore feasible that 23 is the true biosynthetic precursor of 1 in T. yemense. The carbonyl oxygen of the acetate attacks  Neoclerodanes with the A-ring contracted to a cyclopentane, as in 13, have been reported previously as plant natural products but with more complex substitution patterns. Examples with a cyclobutene fused to the cyclopentane were reported in the 1980s. 32,33 Ring-contracted neoclerodanes were later identified [34][35][36][37] from Pteronia, Conyza, and Microglossa species. In these neoclerodanes, there is extensive transannular bridging. Fatimanol Q 13 is the simplest ring-contracted neoclerodane identified to date. Bohlmann's group suggested that their ringcontracted compounds arose from rearrangements involving migration of C-2 to bond with C-4. Mechanisms proposed include protonation of a hydroxy group at C-4 to initiate the rearrangement, 34 trapping the aldehyde formed from C-3 with a hydroxy group at C-10 34,35 and formation of a C-1 = C-10 double bond. 34,36,37 In each case, the C-10 position is oxidised. However, in the present case, C-10 is not oxidised with either an oxygen function or an alkene. We propose the mechanism in Fig. 4 for the biosynthetic ring-contraction. In this mechanism, the rearrangement is triggered by protonation and ring-opening of the strained oxirane. The leaving group oxygen is on the lower face of the ring and thus almost antiperiplanar to the C-3-C-2 bond. The oxirane-opening and the migration of C-2 are probably concerted, given this conformation. C-2 will approach C-4 from the opposite side to the leaving group, resulting in stereochemical inversion at C-4. This places C-18 on the upper face and the newly generated carbocation (C-3) on the lower face, where it can readily form a hemiacetal with HO-19. This hemiacetal closes the second 5-membered ring such that the two 5-membered rings are cis-fused, an energetically favoured arrangement.
Acid anhydrides are relatively uncommon in natural products, owing to their potential for electrophilic reactivity and hydrolysis. However, we have firmly identified 10 as having a succinic anhydride as the upper ring; this is the first such neoclerodane to be reported. Only one natural product containing a simple (non-fused) succinic anhydride, tubogenic anhydride A, has previously been isolated, from Aspergillus tubingensis 38 .
Compounds 2,3,7-9,11,12,14-17,19,20 contain the conventional neoclerodane carbon framework with various differences in oxidation level, acetylation, bridging rings and ring-opening at the C-4/C-18 oxirane, whereas 4-6,18 lack C-19. The C-12-C-20 lactol in 2 has precedent in gnaphalin (isolated from T. gnaphalodes) 39 , although the latter lacks HO-3. The C-19-C-20 bridging lactone of 3,7,10 is present in teulepicephin 15 and many other neoclerodanes. Compound 11 contains a related bridging acetal, giving a rigid polycyclic structure. This polycycle is also present in teucrin P 1 (also from T. gnaphalodes) 40 , teupyrenone (from T. pyrenaicum) 41 , and teupolin III 42 , although the latter do not have the additional lower fused tetrahydrofuran to stiffen the structure further. The upper hydroxyfuranone hemiacetal in 7 is also present in the neoclerodane salvidivin (from Salvia divinorum) 22 , whereas the methoxyfuranone acetal moiety in 8,9 has precedent in the labdane 15-methoxyvelutine C (from Marrubium thessalum) 43 . Most neoclerodanes have the 12-S configuration, so the identification of 16 as a 12-R neoclerodane is noteworthy. Gács-Baitz et al. 42 used NOE NMR spectroscopy to determine configuration at C-12 in neoclerodanes featuring the upper aromatic furan and the spiro-lactone but did not have an exact epimeric pair for their study; 15 and 16 are exact epimers which facilitated their stereochemical identification. The lower hydroxyfuranone hemiacetal of 4,5 and the corresponding methoxyfuranone acetal feature of 6 have scant precedent, in teucvisin C 44 and cracroson B 45 , respectively, while the dimethoxydihydrofuran diacetal of 18 is completely novel in the series. The ajugamarins and related neoclerodanes have the upper furanone unit of 19 and 20 but all known ajugamarins are oxygenated at C-12 46 .
Compounds 1-12 were evaluated for their ability to enhance the glucose-triggered secretion of insulin by freshly isolated murine pancreatic islets, using our previous assay 47 . In negative-control islets, insulin secretion was 9.1 ± 0.3 ng islet −1 h −1 , triggered by glucose (16.7 mM) (Fig. 5). This release was increased 2.2-fold by www.nature.com/scientificreports/ the standard drug tolbutamide (20.2 ± 1.3 ng islet −1 h −1 ). The tested compounds showed a range of activities. Compounds 1,2,12 showed little or no effect on the secretion of insulin. Compounds 3-6 and 11 increased the glucose-triggered release of insulin by approximately the same extent as the positive control tolbutamide. Encouragingly, 7-10 showed strong enhancement of insulin secretion, by 3-4 × , although not as potently as the coumarins cluteolin D and clueolin J (from Clutia lanceolata) 48 .

Conclusions
We report the isolation and identification of twenty new neoclerodanes from the traditional medicinal plant T. yemense. Compound 1 contains an orthoacetate, which is previously unreported in naturally-occurring neoclerodanes. As shown (Fig. 3), the acetate, oxirane and ketone groups in proposed precursor 23 are appositely located to facilitate formation of the orthoester; precursor 23 is 19-acetylteulepicin, previously identified in T. buxifolium. The upper (tetrahydro)furan unit in 10 is a succinic anhydride, a reactive moiety not often found in plants but presumably stable in the arid climate in which T. yemense grows in nature. Compound 13 results from a relatively unusual ring-contracting skeletal rearrangement during biosynthesis. Interestingly, 7-10 were found to enhance the glucose-triggered release of insulin from isolated murine pancreatic islets to a greater extent than the standard anti-diabetic drug tolbutamide; these compounds represent new leads for the development of treatments for this widespread disease.