Polyacetylenes from Oplopanax horridus and Panax ginseng: Relationship between Structure and PPARγ Activation.

Oplopanax horridus and Panax ginseng are members of the plant family Araliaceae, which is rich in structurally diverse polyacetylenes. In this work, we isolated and determined structures of 23 aliphatic C17 and C18 polyacetylenes, of which five are new compounds. Polyacetylenes have a suitable scaffold for binding to PPARγ, a ligand-activated transcription factor involved in metabolic regulation. Using a reporter gene assay, their potential was investigated to activate PPARγ. The majority of the polyacetylenes showed at least some PPARγ activity, among which oplopantriol B 18-acetate (1) and oplopantriol B (2) were the most potent partial PPARγ activators. By employing in silico molecular docking and comparing the activities of structural analogues, features are described that are involved in PPARγ activation, as well as in cytotoxicity. It was found that the type of C-1 to C-2 bond, the polarity of the terminal alkyl chain, and the backbone flexibility can impact bioactivity of polyacetylenes, while diol structures with a C-1 to C-2 double bond showed enhanced cytotoxicity. Since PPARγ activators have antidiabetic and anti-inflammatory properties, the present results may help explain some of the beneficial effects observed in the traditional use of O. horridus extracts. Additionally, they might guide the polyacetylene-based design of future PPARγ partial agonists.

P olyacetylenes are a group of secondary metabolites produced commonly by members of the Apiaceae, Araliaceae, and Asteraceae families. Bioactive polyacetylenes (= polyynes) produced by these plants typically consist of aliphatic chains with several C−C triple bonds and serve plants for pathogen defense. Therefore, polyacetylenes have mostly been investigated due to their inhibitory properties against fungi and bacteria, as well as against various types of cancer cells in vitro and in vivo. 1,2 Additionally, polyacetylenes were found to have allergenic, neurotoxic, and anti-inflammatory properties. 1 PPARγ is a member of the nuclear receptor family of transcription factors that acts as regulators of lipid homeostasis, adipogenesis, and inflammation. 3 In order to activate the transcription of their target genes, nuclear receptors require binding of their respective ligands. Ligands that activate PPARγ are beneficial for improving insulin sensitivity in conditions such as type II diabetes and for reducing inflammation. 4 Since long-chain polyunsaturated fatty acids are effective PPARγ binders, and aliphatic polyacetylenes resemble their structure, 5 our group investigated previously as to whether they could act as PPARγ activators. Indeed, the aliphatic C 17 polyacetylene falcarindiol from Notopterygium incisum was shown to act as a partial PPARγ agonist. 6 In order to use a more systematic approach for the search for polyyne-derived PPARγ activators, a collection of polyacetylene structural analogues was purified from Devil's club (Oplopanax horridus (Sm.) Miq., Araliaceae) and Asian ginseng (Panax ginseng C.A. Mey., Araliaceae). Polyacetylenes are major bioactive constituents of the Devil's club, with a variety of different structures. 7 O. horridus is indigenous to the North American Pacific area and has been used traditionally by the indigenous peoples for treating respiratory infections and inflammatory conditions such as rheumatism and arthritis. 8 Panax ginseng is a popular medicinal plant from East Asia, which, in addition to ginsenosides, also produces polyacetylenes. 9 In the present work, we isolated and determined the structures of 23 aliphatic C 17 and C 18 polyacetylenes, of which five (5−7, 14, and 15) were identified for the first time. Using a reporter gene assay, their PPARγ activation potential was investigated. By employing molecular docking and comparing structural analogues, structural features were identified that are involved in receptor binding and activation. In addition, structural characteristics correlating with their cytotoxicity are described.
The petroleum ether extract of P. ginseng was subjected to successive column chromatography, yielding three native polyacetylenes. Their structures were determined by spectroscopic methods as (−)-falcarinol (21), 23 (9R,10S)-epoxyhep-tadecan-4,6-diyn-3-one (22), 24,25 and (9R,10S)-epoxy-16heptadecene-4,6-diyn-3-one (23). 24 Compound 5 was isolated as colorless solid. The HRESIMS data allowed for the assignment of a molecular formula of C 20 H 30 O 5 . The observed UV absorption maxima at 232, 245, and 258 nm and the four quaternary carbon signals (δ 81.8, 80.7, 69.1, and 68.9) in the 13 C NMR spectrum were characteristic of a bisacetylene/diyne skeleton. 14 The assignments of all proton and carbon resonances were achieved through interpretation of the 2D NMR data. When compared with the 1 H NMR spectrum of oplopantriol B 1-acetate (1), the triplet of a terminal methyl group was replaced by a signal of a hydroxy methylene group at δ H 3.69 (m) in 5. Inductive effects were therefore observed for the protons of its vicinal methylene (H-2; Δδ H −0.19) and the hydroxy methine on its β-position (H-3; Δδ H −0.25), with both sharing an HMBC correlation with C-4 of the diyne system. As also revealed by the HMBC and COSY spectra, the diyne system was connected through a hydroxy methine to a cis-configured olefinic bond, which, in turn, was connected to the C-11 of the aliphatic chain. On comparison of their 13 C NMR spectra, the remaining signals belonging to an aliphatic chain of compounds 5 and 1 were highly superimposable. In representatives of 1−4, 9, and 10, possessing an oplopantriol A/B-type skeleton, in which the two stereogenic centers at C-3 and C-8 are separated by two triple bonds, these stereogenic centers were found to have S-configuration, and all these compounds have shown large positive specific rotations. 10,13 Considering the similar motif of molecular constitution and the large positive value of the specific rotation of 5, the same configuration at C-3 and C-8 was assumed for compounds 5, as well as for 6 and 7, which possess the same skeleton (see also Figure 1). Therefore, the structure of 5 was elucidated as (3S,8S,9Z)-octadec-9-ene-4,6-diyne-1,3,8,18-tetrol 18-acetate. It is assigned here the trivial name 1hydroxyoplopantriol B 18-acetate. Deacetylation occurred when the methanol solution of compound 5 was kept at room temperature. The deacetylation product 6 was obtained and given the trivial name 1-hydroxyoplopantriol B.
Compound 7 was obtained as a colorless solid. The HRESIMS data allowed the deduction of a molecular formula of C 17 H 26 O 3 . The UV absorption maxima at 231, 244, and 258 nm and four quaternary carbon signals (δ 82.2, 79.9, 69.3, and 68.9) in the 13 C NMR spectrum were characteristic of a diacetylene skeleton. 14 Interpretation of the 2D NMR data led to the assignment of all proton and carbon resonances. The 1 H NMR signals at δ 0.99 (t, J = 7.4 Hz, H-1), 1.68 (m, H-2), and 4.30 (t, J = 6.6 Hz, H-3) were assigned to a hydroxy propyl group, for which the connection to the polyacetylene skeleton was supported by HMBC correlations observed between H-3 and the acetylenic C-4 and C-5. Although highly overlapping 1 H NMR signals at δ 5.49 (m, H-9) and 5.47 (m, H-10) provided little coupling constant information, their corresponding carbon signals at δ 136.1 (C-10) and 130.2 (C-9) indicated clearly the presence of a cis-configured olefinic group.
Both olefinic carbons showed HMBC correlations with a doublet at δ 5.20 (H-8), which, in turn, also had HMBC correlations with the acetylenic C-6 and C-7. These aforementioned data, together with aliphatic chain signals in the higher field region of the NMR spectrum of compound 7, resembled those of oplopandiol (9). 10 The major difference between 7 and 9 is the hydroxy group functionalization at the C-11 position in 7, as revealed by the HMBC correlations between δ 5.49 (H-9), 5.47 (H-10) and δ 68.4 (C-11). This led to the downfield shift of the adjacent C-10 (Δδ C −2.2 ppm) and C-12 (Δδ C −8.0 ppm) resonances. With each of these having a hydroxy group at the respective α position, the two alkenyl protons, H-10 and H-9, experienced similar deshielding effects and hence displayed highly overlapping NMR signals. By analogy to compound 5, compound 7 was assigned as having the same 3S,8S-configuration. The absolute configuration of C-11, however, remained undetermined. Therefore, compound 7 was elucidated as (3S,8S,11ξ,9Z)heptadec-1,9-diene-4,6-diyne-3,8,11-triol and was given the name 11-hydroxyoplopandiol.
Compound 14 was obtained as a white, colorless oil. Its HRESIMS data allowed the assignment of a molecular formula of C 20 H 28 O 4 . The observed UV absorption maxima at 215, 242, 254, 268, and 283 nm and the four quaternary carbon signals (δ 82.2, 78.0, 74.0, and 70.6) of the 13 C NMR spectrum were typical for a conjugated ene-yne-yne chromophore. 17 The assignments of all proton and carbon resonances were achieved through interpretation of the 2D NMR data. The lower field signals of the 1 H NMR spectrum of 14 resembled those of panaxydiol (17). However, instead of a characteristic methyl triplet, one group of signals, including a methylene triplet at 4.05 (J = 6.7 Hz, H-18), a methyl singlet at δ 2.02 (H-20), and   -19), indicating acetylation at a primary alcohol position. The third one was assigned as belonging to the methylene group vicinal to the acetylation position, as evidenced by the HMBC correlation with the carbon signal at δ 65.7 (C-18). Its HMBC correlation with another methylene carbon at δ 30.0 (C-16) supported the acetyl functionalization at the end of the aliphatic chain. Comparison of their 13 C NMR spectra suggested compound 14 has a longer aliphatic chain, with one more methylene group, than compound 17. A second cluster of known polyacetylenes (12,13,16,17,20) isolated from O. horridus exhibited stereogenic centers at C-3 and C-10, which are separated by two triple bonds and an additional trans double bond; these compounds are derived from (3R,10S,8E)-8-octadecene-4,6-diyne-3,10,18-triol (13). Representatives of this type have been found to possess an R-configuration at C-3 and an S-configuration at C-10, 26 and, with the exception of the epoxide (20), all of these compounds exhibit a small negative specific rotation ( Figure 1). Due to their similar specific rotation values, compounds 14 and 15 were assumed to have the same 3R,10S configuration as the known compounds of this cluster. Therefore, the structure of compound 14 was determined as (3R,10S,8E)-octadec-1,8diene-4,6-diyne-3,10,18-triol 18-acetate, and it was assigned the trivial name oplopantriol C 18-acetate. After being kept at room temperature, a methanol solution of compound 14 yielded its deacetylation product (15), which was given the trivial name oplopantriol C. Biological Evaluation. Using a luciferase reporter gene assay, altogether 22 compounds were tested for their potential to activate PPARγ (compound 6 was not tested due to the insufficient amount available). Mean fold activation is reported in Table 2 and Figure S1 (Supporting Information), arranged according to activity, from highly to least active compounds. The most potent activators were oplopantriol B 18-acetate (1) (mean ± SD: 4.2 ± 0.7), oplopantriol B (2) (4.1 ± 0.7), followed by oplopandiol (9) (2.5 ± 1.5), and (3R,8E,10S)-8octadecene-4,6-diyne-3,10−18-triol (13) (2.5 ± 0.5), although only the effects of 1 and 2 reached statistical significance. Concentration−response characteristics of compounds 1 and 2 in comparison to the full agonist pioglitazone classify them as partial agonists ( Figure S2, Supporting Information). All four compounds were previously examined in various cell lines for their cytostatic effects, 16,29,30 and compound 9 was also shown to have antimycobacterial properties. 10 To the best of our knowledge, no other specific biological functions have been assigned to these four polyacetylenes. Extracts of O. horridus have been reported to have antidiabetic properties, albeit with inconclusive results. 7,8 However, the ability of polyacetylenes to activate PPARγ may explain some of the previously obtained results and should warrant further research. Additionally, ligand binding to PPARγ inhibits inflammatory responses, which partially may contribute to the anti-inflammatory properties of polyacetylenes.
Interestingly, in the present assay falcarindiol (10) did not show any PPARγ activation, while a previous report on this compound isolated from N. incisum showed activation of PPARγ with an EC 50 of 3.2 μM and E max = 3.26-fold. 6 The most likely reason for this is that the previous study used falcarindiol in the 3R,8S configuration, while the one isolated from O. horridus was the 3S,8S diastereomer, highlighting the importance of the absolute configuration.
Molecular docking was employed to investigate the polyacetylene binding mode to the receptor. Polyacetylenes are anchored in the PPARγ ligand binding pocket via interactions with residues of all three subpockets. A more detailed description of the binding pocket and the docking experiments is available in the Supporting Information ( Figure  S3). Hydrogen bond predictions for the most active (1,2,9,13,11) and least active (14,10,17,8) polyacetylenes are reported in Table 2. The anchoring of the two most active polyacetylenes, 1 and 2, is shown in Figure 2. Both polyacetylenes were positioned similarly with the polyyne chain within arm I and the alkyl tail in arm II. These compounds each formed hydrogen bonds between their C-3 hydroxy group and Ser289 and Cys285 within arm I and between the C-8 hydroxy group and Leu340 in the entrance.
Next, the structure−activity relationships were explored between polyacetylenes and PPARγ activation. Within the compound collection used, three structural categories were identified that may influence PPARγ activation: the C-1 to C-2 bond, the terminal group, and the backbone flexibility ( Figure  3). Polyacetylenes were grouped based on the type of C-1 to C-2 bond, which could be either single or double. The terminal 1 Multifactorial ANOVA rejected the contribution of interactions between the structural categories on PPARγ activation; as such, each category was assessed independently. The strongest impact on activation referred to the type of C-1 to C-2 bond (p = 0.001 67), followed by the type of terminal group at the end of the alkyl chain (p = 0.033 99). The backbone flexibility showed only a moderate tendency to impact PPARγ activation (p = 0.072 801 6).
Within each category, a single bond between C-1 and C-2 strongly favored PPARγ activation over a double bond (mean ± SD: 2.3 ± 0.9 vs 1.3 ± 0.6, p = 0.001 67). Compounds with a terminal hydroxy group had a higher mean fold activation in comparison to those with acetoxy and methyl groups (2.4 ± 1.0 vs 1.9 ± 1.2 vs 1.5 ± 0.8, respectively), but the difference was significant only between hydroxy and methyl groups (p = 0.029 03). Although not significant, flexible structures showed a tendency toward higher PPARγ activation (2.0 ± 1 vs 1.5 ± 0.7).
The polyacetylene collection utilized contained structural analogues, with pairs of polyacetylenes having identical structures, except for the examined feature. Therefore, in order to verify the observations made, the structural counterparts were compared further individually ( Figure S4, Supporting Information). Although pairwise comparisons were not always statistically significant, the differences in activity were consistent and followed the trend observed in Figure 3.
Molecular docking was employed also to compare the structural pairs of each category in their binding to the PPARγ ligand binding pocket. A comparison of the docking positions of the C-1 to C-2 single:double bond pair 1:3 is shown in Figure 4. Compound 1 was able to form two hydrogen bonds between the C-3 hydroxy group and the amino acids Ser289 and Cys285, while in compound 3, the C-3 hydroxy group was positioned differently, allowing the formation of a hydrogen bond with Ser289 only. Otherwise, the two compounds were positioned similarly in the binding pocket and both formed a hydrogen bond with Leu340.
Molecular docking did not reveal any structural basis for the observed differences in polyacetylene activity based on their terminal group, possibly because the differences in activity were too small and due to high flexibility of the alkyl chain in arm II. However, the increased PPARγ activation by the more polar polyacetylenes may reflect their increased bioavailability.  were categorized according to their type of C-1 to C-2 bond, terminal group, or backbone flexibility. Each category included all tested polyacetylenes that were divided further into groups. Categories were assessed for association with PPARγ activation by three-way ANOVA, and groups were compared with a TukeyHSD post hoc test; ***p < 0.001, *p < 0.05, ns, not significant. Journal of Natural Products pubs.acs.org/jnp Article Structural analogues that were identified between polyacetylenes of the 3S,8S orientation and 3R,10S orientation differed in the presence of an extra C-8 to C-9 double bond after the polyyne chain, thus shifting the C-8 hydroxy group of the 3S,8S polyacetylenes to the C-10 position in 3R,10S polyacetylenes. The extended and rigid spacer between the C-3 and C-10 hydroxy groups was unfavorable for the interaction in arm I, and, since the distance cannot be shortened by bending, the compounds were flipping in orientation within the PPARγ ligand binding pocket. Figure 5 shows compound 2 (flexible) with its polyyne part in arm I, whereas compound 13 (rigid) had its alkyl chain in arm I and the polyyne chain in arm II. However, the amino acids involved in the interactions stayed the same in both groups.
Overall, the present observations made suggest that PPARγ activation by aliphatic polyacetylenes is supported by the presence of a single C-1 to C-2 bond, polar terminal groups at the end of the alkyl chain, and a flexible backbone that can bend.
Next, a group comparison was used to discern whether the structural features that impact PPARγ activation may also impact viability and thus confound the present results. As a measure for cell viability during the luciferase assay, EGFP expression (fluorescence) in comparison to vehicle control (0.1% DMSO) was quantified. No statistically significant association was obtained using the multiway ANOVA and post hoc TukeyHSD tests ( Figure 6), but the trend was observable particularly for the C-1 to C-2 bond. The mean relative viability values were higher for polyacetylenes with a single C-1 to C-2 bond in comparison to those with the presence of a double bond (mean ± SD: 0.7021 ± 0.3014 vs 0.4392 ± 0.3873, respectively), which was in agreement with previously published observations on polyacetylenes from O. horridus. 29−31 However, this did not hold true for the compounds 18−21, which all possess a double bond in this position, but showed viability higher than 70% ( Figure S5, Supporting Information) and were not diols. Interestingly, falcarinol (21) was not toxic in our assay, while falcarindiol (10) was cytotoxic, contradictory to previous reports. 32,33 Although direct comparisons are difficult due to different cell lines and procedures used, it is important to consider the configuration, as previous reports compared (−)-falcarinol to (3R,8S)-falcarindiol, typical for Apiaceae plants. Therefore, the most toxic polyacetylenes in the present group were diols with two hydroxy groups at each side of the polyyine chain and a C-1 to C-2 double bond.
Since all of the structural pairs of C-1 to C-2 bond type contained the second hydroxy group, the C-1 to C-2 doublebond analogues were shown to be more toxic by direct comparison to their single-bond counterparts ( Figure S6A, Supporting Information). The present observations were verified with an independent resazurin cell viability assay. Since in this assay cells are not stressed with the transfection solution, polyacetylenes were not as cytotoxic even at 10 μM and after 24 h of incubation ( Figure S7, Supporting Information). However, the relative toxicity between C-1 to C-2 single:double bond pairs remained the same ( Figure S6B, Supporting Information). Even though the luminescence values used to estimate the PPARγ activation were normalized to fluorescence values, it cannot be excluded that the PPARγ activation data are not at least partly influenced by the cytotoxicity of these compounds.
In summary, 23 aliphatic C 17 and C 18 polyacetylenes were isolated from O. horridus and P. ginseng, with five of these being new compounds. Oplopantriol B 18-acetate (1) and oplopantriol B (2) were identified as nontoxic, partial PPARγ activators, thus corroborating the antidiabetic and anti-inflammatory potential of O. horridus extracts. In addition, the relationship was explored between the structural features of polyacetylenes and their potential to activate PPARγ, pointing to the significance of the type of bond between C-1 and C-2, the terminal group polarity, and backbone flexibility. Additionally, diols with a double bond between C-1 and C-2 and two hydroxy groups at each side of the polyyne chain appear to enhance polyacetylene cytotoxicity. The present results may help to guide a polyacetylene scaffold-based design for future PPARγ partial agonists with reduced toxicity.

■ EXPERIMENTAL SECTION
General Experimental Procedures. Optical rotations were measured on a JASCO P-2000 multioption polarimeter. 1 H, 13 C, and 2D NMR spectra (COSY, HSQC, and HMBC) were recorded at  Extraction and Isolation. The dried root bark of O. horridus (1.75 kg) was ground manually and extracted successively with dichloromethane and methanol. The dichloromethane extract (145 g) was subjected to silica gel chromatography and eluted with a hexane− ethyl acetate gradient to afford 117 fractions (OD 1−117, each 1.5 L). Fractions with a characteristic polyacetylene UV pattern were further subjected to chromatography using reversed-phase C 18 material and Sephadex LH-20, with aqueous methanol as mobile phase. All compounds were finally purified with semipreparative HPLC. The petroleum ether extract of P. ginseng (provided by Prof. Qipin Gao, Changchun University of Chinese Medicine) was subjected to successive silica gel chromatography and finally purified with semipreparative HPLC. Further details of isolation procedures are summarized in Tables S2 and S3 (Supporting Information).
Modified Liquid Loading Method for Open Column Reversed-Phase Chromatography (At-Column Homogenization). To ensure efficient sample loading and robust chromatography for the airsensitive and lipophilic polyacetylenes with reversed-phase materials, a new method, based on conventional "liquid loading", has been employed. After dissolution of the sample in methanol it was loaded to a column conditioned with a methanol−water solution (methanol concentration ∼20% lower than the desired starting condition of the gradient). The loading region was flushed with the same conditioning solvent and homogenized until the new front appeared (detailed description of the procedure: Supporting Information). This loading method allowed the separation of closely eluting polyacetylene isomers on Sephadex LH-20 ( Figure S8, Supporting Information) and facilitated the follow-up purification on reversed-phase C 18 material. To our knowledge, this is the first report of isolation of polyacetylenes on Sephadex LH-20 eluted with an aqueous mobile phase. An animation illustrating this process can be found in the Supporting Information.
Biological Evaluation. For PPARγ activation experiments, luciferase reporter assays were performed in HEK 293 cells. Cells were cultivated under standard conditions (37°C, 5% CO 2 , passage every 3 days) in DMEM supplemented with 10% FBS, 2 mM glutamine, 100 U/mL benzylpenicillin, and 100 μg/mL streptomycin (DMEM complete). A total of 6 × 10 6 cells were seeded in 15 cm Petri dishes and incubated overnight. Cells were transfected using the calcium phosphate method with the following plasmids: 6 μg of PPARγ expression plasmid, 6 μg of a luciferase reporter plasmid containing PPAR response elements, and 3 μg of pEGFP-N1 for internal transfection efficiency control. After 6 h, cells were washed with PBS and the medium was replaced with DMEM containing 5% charcoal-stripped serum (supplemented with glutamine and antibiotics as DMEM complete). Then, 4 × 10 5 cells per well were reseeded in a 96-well plate and treated with 0.1% DMSO as vehicle control or the 3 μM test compound and incubated for 18 h. Cells were lysed using the Promega luciferase cell culture lysis 5× reagent (Mannheim, Germany). Luminescence and fluorescence values were measured on a Tecan Spark instrument (Mannedorf, Switzerland). Luminescence values were normalized to the EGFP fluorescence and expressed as fold changes relative to the vehicle control. Relative viability was calculated from fluorescence values relative to those of the vehicle controls.
For the resazurin assay, HEK293 cells (5 × 10 5 cells per well) were seeded in a 96-well plate and incubated overnight. On the following day, old medium was removed and cells were treated with 0.4% DMSO as vehicle control or 10 μM test compound diluted in stripped DMEM. After 24 h of incubation, wells were washed with PBS, and 10 μg/mL resazurin solution in PBS was added to the cells for 4 h. Resazurin conversion was measured with an excitation/emission wavelength 535 nm/580 m. Viability was expressed relative to the vehicle control.
Statistical Analysis. All experiments were performed at least three times, each in four technical replicates. Data were analyzed using the GraphPad Prism 6 software (La Jolla, CA, USA) and R version 3.4.4. 34 Docking. For the docking, the X-ray crystal structure 3R5N 35 with a resolution of 2.0 Å and a good electron density distribution was employed. In this structure, PPARγ was crystallized with two molecules of magnolol, a partial agonist on the PPARγ. After calculation of the starting conformations in Omega 2.5.1.4, 36,37 the docking was performed using the software GOLD 5.2. 38, 39 The protonation state of His323 was set to NE2H, and a water molecule (HOH35) was extracted and set to "toggle and spin" in the advanced options. Both of the cocrystallized ligands were extracted, and the binding site was defined by one or more ligands. The two ligands were chosen, and a cavity file was generated. Per compound, up to 15 binding poses per ligand (GA run) were reported, and the setting "allow early termination" was disabled. The slow and most accurate GA search option was employed, and ChemPLP was used as scoring function. Finally, the docking poses were analyzed in LigandScout 3.12, 28 and the most probable poses were selected with regard to structure−activity relationship (see Table 1). In order to achieve a distinct structure−activity analysis, only the comparatively highly active and the inactive compounds were considered. With regard to the flexibility of the receptor−ligand system, the interaction pattern of the docking poses was investigated before and after minimization of MMFF94 energy in LigandScout. Due to this minimization process, some of the molecules were able to form additional interactions with essential amino acids.