16-nor Limonoids from Harrisonia perforata as promising selective 11β-HSD1 inhibitors

Two new 16-nor limonoids, harperspinoids A and B (1 and 2), with a unique 7/5/5/6/5 ring system, have been isolated from the plant Harrisonia perforate together with a known one, Harperforin G (3). Their structures were elucidated by NMR spectroscopy, X-ray diffraction analysis and computational modelling. Compound 1 exists as polymorphic crystals. Conformations of 1 in solution were further discussed based on the computational results. These compounds exhibited notable inhibitory activity against the 11β-HSD1 enzyme. Compound 3 had potencies for the inhibition of human 11β-HSD1 with high selectivity against 11β-HSD2 (IC50 0.58 μM, SI > 174). Molecular docking and quantitative structure-activity relationship studies revealed a mixed regulatory mechanism.

The relative configuration assigned for 1 was deduced by the analysis of ROESY data. Examination of a Dreiding molecular model of 1, suggested that 1 adopts a conformation in which both of the five-member rings, B1 and B2, is orthogonal to each other. ROESY correlations of H-30β/H-9/H-12β/H-17 indicated that H-30β, H-9, and H-17 are all on the same face of the octhydroisobenzo-furan moiety, thus, orienting them toward oxygen atom of ring B1. ROESY interactions of H-30α with H 2 -6, H-6α with H-5, and H-5 with Me-19 placed the corresponding substituents together on the opposite face of the octhydroisobenzo-furan ring, and fixing the relative configuration at C-7 as shown.
Needle and prism shaped crystals of 1 were obtained simultaneously from MeOH/H 2 O (9:1) via slow evaporation. The single-crystal X-ray diffraction analysis of each sample showed that 1 exists in two crystalline forms, termed HA and HB, respectively (deposition no. CDDD 999635 and 999636, https://www.ccdc.cam.ac.uk/services/structure_deposit/) (Fig. 3). Form HA crystallizes in the monoclinic P2 1 space group. Rings A and D adopt a half-chair conformation; rings B, C, and E take the envelope conformation; and the β-furan ring is almost planar and adopts a cis-conformation with a dihedral angle of 24.5°. In contrast to HA, form HB crystallizes in the triclinic, P2 1 2 1 2 1 space group. Ring A takes a boat conformation, and the β-furan ring is arranged in a trans-conformation with a dihedral angle of − 125.8°. To the best of our knowledge, this is the first case of limonoids with polymorphism. Moreover, the final refinement of HA on the Cu Kα data resulted in a Flack parameter of 0.2(2), which gave an unambiguous assignment of the absolute configuration of 1 as (5 S, 7 S, 9 R, 10 R, 13 R and 17 S) 21 .
As an extensive exploration of all the conformations of 1 in solution, a computational modelling study was conducted using Gaussian-03 program at the B3LYP/6-31 G* level (Gaussian-03, revision D.01, Gaussian Inc., Pittsburgh). The calculations showed four low-energy conformations of 1, which were roughly distinguished as boat, trans (BT), half-chair, cis (HCC), half-chair, trans (HCT), and boat, cis (BC) according to the conformational difference of ring A and the orientation of the β-furan ring (Fig. 4). Then, we compared bond lengths, bond angles, and dihedral angels of HCC and BC with those of the crystal structure in the forms HA and HB, and their RMS value are calculated to be 0.4985. All these data indicate that calculated HCC and BC are in good agreement with HA and HB, respectively. Moreover, the calculation also showed two transition states (TS1 and TS2) (Fig. 5), which corresponded to the conformational conversion of ring A and the rotation of the β-furan ring; their free energies against the most stable conformer HCC were 17.7 and 4.9 kcal/mol, respectively. Therefore, the occurrence of rapid interconversion of the four conformers of 1 is a logical process in solution at room temperature [22][23][24] (for details, see the supporting information, SI). Furthermore, the stability of the two polymorphs was calculated using a molecular mechanics method performed using SYBYL 8.1. Molecular energy of the two crystal polymorphs of 1 is 260.1 kcal/mol (HA) and 258.1 kcal/mol (HB), respectively. The small energy difference between them indicated that the two conformers of 1 could simultaneously assemble into different crystal at room temperature 23 .
Harperspinoid B (2) had the molecular formula C 25 H 28 O 7 based on HRESIMS, which is the same as that of 1. The NMR features of 2 (Table 1) closely resembled those of 1 except for the resonances near C-9. The data from the 1 H-1 H COSY, HSQC and HMBC spectra indicated that compound 2 shared the same planar structure as that of 1. The ROE correlation of Me-18 (δ H 0.92)/H-9 (δ H 4.39) suggested that H-9 is α-orientated. In addition, the relative configuration of the remaining chiral centres of 2 would be analogous to those of 1 based on 13 C NMR shifts and NOE data (Fig. 6). On the basis of biogenetic considerations, the absolute configuration of 2 is tentatively assigned as 5 S, 7 S, 9 S, 10 R, 13 R and 17 S. Thus, the structure of 2 was eventually established as shown in Fig. 1.
Biogenetically, compounds 1 and 2 might be derived from Citriolide A, which may be converted to the key hemiketal intermediate A via a free radical mechanism. Subsequently, the intermediate A may undergo oxidation, cyclization, and double-bond migration in turn to generate 1 and 2 (Fig. 7).
The inhibitory activity of compounds 1-3 on murine and human 11β-HSD1 was evaluated using the scintillation proximity assay (SPA) 25 . In intact CHOP cells transfected with murine HSD11B1, only compound 1 showed inhibitory effects with an IC 50 value of 0.60 μ M. Moreover, compound 1 were highly selective against murine 11β-HSD2 activity since it did not inhibit the enzyme at all at 1 mM (SI > 1661). We further used the intact cell  To better understand the structure-activity relationship of the compounds, a molecular docking simulation was performed using co-crystal structures of the 11β-HSD1 enzyme (4K1L for human). Despite the structural variety of the different inhibitors that have previously been investigated, the crystal structures of the NADP(H)-dependent 11β-HSD1 proteins are comparatively similar 26 . AutoDock 4 was employed to quantify the parameters that are crucial for high affinity ligand binding. According to the three dimensional images,      for the above three substrates. LigPlot+ was then used for further analysis of the complex between compound 3 and the enzyme. In the docking model, 3 adopted a V shape to fit well into the hydrophobic pocket of the receptor (Fig. 8). In addition to the hydrophobic interaction due to the nature of the polycyclic aliphatic skeleton bearing a furan unit, the hydrogen bonding interactions induced by Arg66, Ser43, Lys44 and Thr220, together with the adjacent O-containing functional groups surrounding 3 could significantly enhance the affinity.
In contrast to the general 11β-HSD1 inhibitors, which form the key hydrogen bonding interactions through Tyr183 and Ser170 within the active site 26 , compound 3 not only lodges in the usual anchoring position and participates in interactions with unusual catalytic residues but also encroaches partially on the cofactor site. NADP(H) specificity in 11β-HSD1 is achieved through the packing interaction with Lys44, a hydrogen bond between its 3′ -OH and Ser43, and an electrostatic interaction of the ribose 2′ -phosphate with guanidinium N atoms in Arg66; at the same time, there is additional contact to this 2′ -phosphate by the backbone amide of Arg66 27 . The competitive interactions of compound 3 with Arg66, Ser43, and Lys44, which comprise crucial residues for electrostatic interaction and H-bond formation involved with NADP(H) specific localizations in 11β-HSD1,would obviously attenuate or reduce their corresponding interactions with NADP(H). However, the reimbursements were most likely provided by the emerging interactions within the optimum approach distance  between NADP(H) and the invasive ligand 3. Thus, the unitary binding affinity seems to be retained. In this special manner with synergistic effects, the high inhibitory activity of incorporating compound 3 could be explained more reasonably.
In summary, we have isolated and identified three 16-nor limonoids, including two new ones from the aerial parts of H. perforata. Two polymorphic forms of harperspinoid A (1) were discovered and unambiguously characterized. Its conformers and their interconversion process in solution were further discussed based on computational modelling. Compound 3, the most potent one, had an IC 50 of 0.58 μM in a whole cell assay. As illustrated through the docking simulation of compounds 1-3 with 11β-HSD1 (4K1L for human), the structural analogues 1-3 probably inhibit 11β-HSD1 in a mixed manipulating pattern. They might occupy the common locating pocket and compete for the catalytic residues that affect NADP(H) binding while generating compensatory interactions via the invasion of the active sites in this cofactor. The unexpected dual modulation of compounds 1-3 on both the substrate and NADP(H) bindings is worthy of further investigation, which might be an interesting objective for future exploration. Our present discovery has demonstrated the versatility and elegance of regulating mechanisms relating to traditional 11β-HSD1 accompanied by its cofactor and supplied valuable information for the design of novel alternative inhibitors. was extracted with MeOH three times, followed by combination, concentration, and suspension in water. It was subsequently partitioned successively with PE (petroleum ether), EtOAc, and nBuOH. The EtOAc part (560 g) was chromatographed on a silica gel column eluted with PE/acetone (from 1:0 to 0:1) to give 6 fractions (A1-A6). A3 (PE/acetone5:1-3:1, 17 g) was fractionated via an MCI gel column eluted with gradient 80% MeOH/H 2 O and further separated by Sephadex LH-20 (MeOH) recrystallization in methanol to afford 3 37 mg. A4 (PE/acetone 3:1, 150 g) was fractionated via an MCI gel column eluted with gradient MeOH/H 2 O from 5:5 to 9:1 to obtain five fractions (B1-B5). Fraction B2 (28 g) was subjected to Sephadex LH-20 (MeOH) to afford (C1-C4) four elutes. Fraction C2 (6.1 g) was subjected to CC with C18 reversed-phase silica gel (MeOH/H 2 O = 30:70-100:0) followed by extensive CC over columns of LH-20 and silica gel yield a mixture of 1 and 2 (40 mg),which was further separated by semi preparative HPLC (MeOH/H 2 O, 55:35, 3 ml/min) to yield 1 (26 mg) and 2 (5.9 mg). Calculation Methodology. All calculations were performed using the Gaussian 03 program package.

Methods
Geometries were fully optimized with the density functional theory methods of B3LYP at the 6-31 G* level. Only one negative eigenvalue and one imaginary frequency were obtained for TSs in computations. Intrinsic reaction coordinates (IRC) were also calculated to authenticate the transition state. The free energy magnitudes were used throughout the theoretical studies.
11β-HSD Enzyme Activity Assay. The inhibitory activities of the compounds on human or mouse 11β-HSD1 and 11β-HSD2 were determined using the scintillation proximity assay (SPA). The full-length cDNAs of human or murine11β-HSD1 and 11β-HSD2 were isolated from the cDNA libraries provided by NIH Mammalian Gene Collection. The cDNAs were cloned into pcDNA3 expression vectors. HEK-293 cells were transfected with the pcDNA3-derived expression plasmid and selected by cultivation in the presence of 700 μg/ml of G418. The microsomal fraction overexpressing 11β-HSD1 or 11β-HSD2 was prepared from the HEK-293 cells, which were stably transfected with 11β-HSD1 or 11β-HSD2. The fraction was then used as the enzyme source for SPA. Microsomes containing human or mouse 11β-HSD1 were incubated with NADPH and [ 3 H] cortisone. The product, [ 3 H] cortisol, was specifically captured by a monoclonal antibody coupled to protein A-coated SPA beads. The 11β-HSD2 screening was performed by incubating 11β-HSD2 microsomes with [ 3 H] cortisol and NAD + and monitoring substrate disappearance. All tests were done twice with glycyrrhizinic acid as a positive control. IC 50 (X + SD, n = 2) values were calculated by using Prism Version 4 (GraphPad Software, SanDiego, CA).
11β-HSD Enzyme Docking Assay. We searched for a possible binding sites for the compounds (1-3) on 11β -HSD enzyme using the AutoDock4 docking program and the structure of 11β-HSD enzyme (Protein Data Bank 4K26 and 4K1L) as the receptor molecules. The docking studies of these compounds with human (4K1L) and murine (4K26)11ß-HSD1 enzymes have been performed respectively. Further analyses using LigPlot+ program revealed the Interactions of the ligand and the enzymes.