C13 Megastigmane Derivatives From Epipremnum pinnatum: β-Damascenone Inhibits the Expression of Pro-Inflammatory Cytokines and Leukocyte Adhesion Molecules as Well as NF-κB Signaling

In order to identify active constituents and to gain some information regarding their mode of action, extracts from leaves of Epipremnum pinnatum were tested for their ability to inhibit inflammatory gene expression in endothelial- and monocyte-like cells (HUVECtert and THP-1, respectively). Bioactivity-guided fractionation using expression of PTGS2 (COX-2) mRNA as a readout resulted in the isolation of two C13 megastigmane glycosides, gusanlungionoside C (1) and citroside A (3), and the phenylalcohol glycoside phenylmethyl-2-O-(6-O-rhamnosyl)-ß-D-galactopyranoside (2). Further analysis identified six additional megastigmane glycosides and the aglycones β-damascenone (10), megastigmatrienone (11), 3-hydroxy-β-damascenone (12), and 3-oxo-7,8-dihydro-α-ionol (13). Pharmacological analysis demonstrated that 10 inhibits LPS-stimulated induction of mRNAs encoding for proinflammatory cytokines and leukocyte adhesion molecules, such as TNF-α, IL-1β, IL-8, COX-2, E-selectin, ICAM-1, and VCAM-1 in HUVECtert and THP-1 cells. 10 inhibited induction of inflammatory genes in HUVECtert and THP-1 cells treated with different agonists, such as TNF-α, IL-1β, and LPS. In addition to mRNA, also the upregulation of inflammatory proteins was inhibited by 10 as demonstrated by immune assays for cell surface E-selectin and secreted TNF-α. Finally, using a luciferase reporter construct, it was shown, that 10 inhibits NF-κB-dependent transcription. Therefore, we hypothesize that inhibition of NF-κB by β-damascenone (10) may represent one of the mechanisms underlying the in vitro anti-inflammatory activity of Epipremnum pinnatum extracts.

The aim of the present study was to identify compounds, which are contributing to the anti-inflammatory activity of E. pinnatum and to evaluate their inhibitory activity on the inflammatory response in THP-1 and human umbilical vein endothelial cells to gain information about the underlying mechanisms.

MATErIALS AND METhODS general Experimental Procedures
UV spectra were recorded on a Thermo Scientific Ultimate 3000 diode-array detector. 1D and 2D NMR spectra were measured in methanol-d4 (δ H 3.31/δ C 49.0) either on a Varian UnityInova 600 spectrometer ( 1 H: 400 MHz, 13 C: 100 MHz) or on a Brucker Avance III NMR spectrometer ( 1 H: 700.0 MHz, 13 C: 166.0 MHz) equipped with a cryo probe. HR-MS was conducted on a Thermo Scientific Q Exactive Hybrid Quadrupole-Orbitrap equipped with electron spray ionization. ESI-MS were obtained on a LTQ XL mass detector in positive and in negative mode. Analytical HPLC was conducted on an Agilent 1100 with an Agilent 1100 diode-array detector. A RP-18 Kinetex (2.6 µm, 100 × 2.6 mm; Phenomenex) column was used for analytical separations employed with a 40-min gradient separation of 0.1% HCOOH in water-MeCN (95:5-0:100 over 30 min, 0:100 held for 10 min). Semipreparative HPLC was performed on a Shimadzu LC-20AT with a Shimadzu DPS-M20A detector on a LUNA RP-18 column (10 µm, 250 × 10 mm, Phenomenex). GC-MS was performed on an Agilent 7890A GC equipped with an Agilent 5975 VL MSD on a HP-5MS capillary column (30 m × 250 µm × 0.25 µm, Agilent). GC-MS method consisted of a 66-min temperature program with an initial temperature of 60°C held for 1 min followed by 4°C/min increase to 280°C held for 10 min. The flow rate of the carrier gas helium was 1.2 ml/min performed in splitless mode. The GC-MS system was operated in EI mode at 70 eV. The identification of the C13 aglycones was performed by comparison of measured chromatogram with available reference compounds or with library softwares (HPCH2205, NIST 08, Wiley138). Column chromatography was performed on a silica gel 60 (0.04-0.063 mm, Merck), and on a sephadex LH 20 (163 µm, GE-Healthcare). Separations were monitored using TLC 60 F 254 (Merck) by staining with 5% sulfuric acid in EtOH and 1% vanilin in EtOH.
Plant Sciences, University of Graz, Austria). Voucher specimen (080514_EPI_Fol_tot_GZU, 080514_EPI_Fol_pulv_GZU) has been deposited at the herbarium of the Department of Pharmacognosy, University of Graz, Austria. The fresh plant material was airdried for 3 weeks at 30°C in an air-flow chamber.

Analysis of NF-κB-dependent genes
Analysis of NF-κB-dependent genes was performed in human umbilical vein endothelial cells (HUVECtert) and the human monocytic cell line THP-1. HUVECtert cells were seeded in a 12-well plate with M199 medium supplemented with 20% FCS, antibiotics (100 units/ml penicillin, 100 µg/ml streptomycine, 250 ng/ml fungizone; Lonza), 2 mM L-glutamine, and 12 µg/ml endothelial cell growth supplement/90 µg/ml heparin (PromoCell). Cells were grown to full confluence, before they were used for experiments. THP-1 cells were cultured as described above, seeded at density of 1x10 6 cells/well in a 12-well plate, and used for experiments without inducing differentiation into macrophages by phorbol ester treatment. For experiment, growing medium was replaced with medium containing 3% FCS and 20 mM HEPES. Cells were pre-incubated with sample compounds for 30 min. Afterward, LPS (30 ng/ml), TNF-α (0.3 ng/ml), or IL-1β (1 ng/ml) were added and incubated for 4 h at 37°C. Subsequently, cells were lysed in RNAzol (Molecular Research Center Inc.) and total RNA was isolated according to manufacturer's instructions. 900 ng RNA were reversely transcribed using murine leukemia virus reverse transcriptase (MuLV from Applied Biosystems ® ) and oligo (dT) 16 primers (Invitrogen). T100 ™ Thermal Cycler was set to following conditions: 25°C for 10 min, 42°C for 45 min, and 95°C for 5 min. Quantitative real time PCR was performed on a StepOnePlus instrument (Applied Biosystems ® ) using SYBRGreen master mix (PCR Biosystems Ltd). Primer sequences are presented in Table 1. mRNA expression was quantified using the ΔΔCt method. Amplification program was set as following: 30 s at 90°C, followed by 40 PCR cycles of 3 s at 95°C and 30 s at 60°C; melting point analysis in 0.1°C steps. The mRNA level of each target gene was normalized to expression of β2-microglobulin. Fold expression was defined as fold increase relative to β2-microglobulin-normalized level of expression in mock-stimulated treatment with vehicle) cells (0.1% DMSO).

TNF-α ELISA
THP-1 cells were seeded at a density of 1x10 5 cells per well in a 96-well plate with RPMI 1640 supplemented with 20 mM HEPES, 3% FCS and 100 units/ml penicillin (Gibco) and 100 µg/ml streptomycin. After 30 min pre-incubation with test compounds, cells were stimulated with 30 ng/ml LPS and incubated for another 4 h. Subsequently, TNF-α protein was quantified in medium supernatant by a DuoSet ELISA kit (R&D Systems).

Cell-based E-selectin ELISA
HUVECtert cells were seeded at density of 4000 cells per well in 96 well plates and incubated for 48 h before experiment. Afterward, medium was replaced with medium containing 3% FCS and 20 mM HEPES. Cells were pretreated with sample compounds for 30 min, followed by stimulation with LPS (30 ng/ml) for 4 h. Cells were fixed with 0.1% glutaraldehyde in PBS for 15 min at 4°C. Blocking was performed with 5% BSA in PBS for 1 h at 37°C. Afterward, cell-bound E-selectin was analyzed using a DuoSet ELISA development kit (R&D Systems).

NF-κB-Driven Luciferase reporter gene Transactivation
The stable cell line HEK 293/NF-κB-luc was obtained from Panomics (catalog RC0014). TNF-α stimulation of these cells induces the expression of the luciferase reporter gene that is regulated by multiple copies of NF-κB response elements. For the assay, cells were seeded in 96 well plates at a density of 4 × 10 4 cells per well in serum-free DMEM overnight. Cells were stained with 2 μM cell tracker green (CTG, Thermo Scientific) for 1 h, followed by subsequent treatment with respective compounds for 1 h and then stimulation with 2 ng/ml TNF-α for 4 h, which initiated NF-κB activation. Fluorescence of CTGstained cells and luminescence of the luciferase product were measured on a Tecan SPARK ® . The obtained luminescence values were normalized to CTG-derived fluorescence related to DMSO control. Normalized fluorescence units were used as indicator for cell number (Fakhrudin et al., 2014). To test whether β-damascenone inhibits NF-κB signaling via an electrophilic attack, 5 mM glutathione was added shortly before test compounds to all samples.

Statistical Analysis
Statistical analysis was calculated by one-way analysis of variance (ANOVA) with Bonferroni post hoc test using IBM SPSS Statistics 25 software. In graphs, data were expressed as mean values ± standard deviation (SD). Statistical significance was expressed as p values: p > 0.05 *, p > 0.01 **, p > 0.001 ***.

rESULTS AND DISCUSSION
Out of three successively prepared extracts (n-hexane, dichloromethane, and MeOH) from E. pinnatum leaves, the MeOH extract inhibited the LPS-induced COX-2 mRNA expression most potently (inhibition at 20 µg/ml: n-hexane: 10.8 ± 10.4%, DCM: no effect, MeOH: 54.3 ± 9.2%) and was, therefore, subjected to activity-guided isolation of the active compounds. First, the MeOH extract was fractionated by silica gel column chromatography using mixtures of n-hexane, ethyl acetate, and water. After TLC comparison and combining similar fractions (175 fractions in total), 20 fractions E1 to E20 were obtained. Subsequent pharmacological testing revealed two active fractions, E16 and E20 (inhibition of LPS-induced COX-2 mRNA expression at 20 µg/ml: E16: 32.7 ± 6.7%, E20: 31.3 ± 7.1%). E16 was further separated on a size exclusion column using Sephadex LH 20 to gain 11 fractions (S1-S11). Active fractions (S3-S6) were combined (inhibition at 20 µg/ml from 20.9 to 71.8%) and compounds of interest were separated on a RP18 column by semipreparative HPLC to yield gusanlungionoside C (1) and a mixture of phenylmethyl-2-O-(6-O-rhamnosyl)-ß-D-galactopyranoside (2) and citroside A (3 -210] − . The constitution of 1 was determined by complete assignment of 1H, DQF-COSY, HSQC, and HMBC spectra, which indicated the presence of a megastigmane derivative with two sugar units. The homonuclear coupling constant (H-1′, 7.8 Hz) of the glucose and the carbon chemical shift (69.5 ppm) of C-5 of rhamnose indicated the presence of a β-glucopyranose and a α-rhamnopyranose. HMBC correlation between the anomeric proton (H-1') of the glucose and the carbon at 75 ppm indicated that the glucose was attached to C-9 of the aglycon, while HMBC correlation between the anomeric proton of the rhamnose (H-1′′) and the carbon at 78.3 ppm pointed to the attachment of this sugar to C-2′ of the glucose. Hence, the constitution of the compound was determined as depicted in Figure 2. Three DBEs are provided by the ring fusions, two by the double bonds in the aglycon. The relative configurations of the stereogenic centers C-6 FIgUrE 1 | Extraction and isolation scheme of anti-inflammatory components from leaves of Epipremnum pinnatum. The plant material was successively extracted with n-hexane, dichloromethane and methanol. The active methanol extract was further fractionated on a silica gel column to gain E1-E20. Fraction E16 was fractionated on a Sephadex LH20 column to produce fractions S1-S11. The most active fraction (S2-S6) were combined to isolate 1 and a mixture of 2 and 3. Compounds 4-9 and 10-13 could be identified with ESI-MS, HR-MS and GC-MS. Active fractions are determined using qPCR analysis of COX-2 mRNA and are indicated with " +" in grey boxes. and C-9 of the aglycon were elucidated by comparison of the proton and carbon resonance values with literature data. The values fit well with compound 3 of the publication by Yu et al. (2011) with reported absolute configurations 6S, 9R, assigned to a compound named gusanlungionoside C.
A mixture of 2 and 3 was isolated as colorless powder. According to proton NMR, the sample comprised two major glycosidic compounds. As the aglycon parts of these components were chemically quite different, complete resonance assignments for the non-sugar portions of the compounds were possible. HR-MS measurement identified protonated molecules at m/z 417.1764 (2) and 387.2020 (3)  , indicating a cleavage of rhamnose and its aglycon with a remaining glucose molecule. This is in accordance with the NMR data. According to the observed resonances in proton NMR and HSQC, the disaccharide was formed by a hexopyranose in the β-form and a β-rhamnopyranose. The correlations in the HMBC spectrum between H-7 (4.94 ppm) and C-1' (101.6 ppm), and H-1' (4.42 ppm) and C-7 (71.5 ppm), respectively, indicated that the benzyl alcohol was attached to the anomeric carbon of the hexose. The three bond correlation between H-1'' (5.20 ppm) and 79.2 ppm indicated that the rhamnose was attached to C-2′. Due to severe signal overlap in the NMR spectra of the hexoses of 2 and 3, it was not possible to determine the relative configuration at C-4′, i.e. it was not possible to determine whether the β-hexose is a glucose or galactose. However, the relatively low carbon shift value of C-2' (79 ppm) points more to a galactose, because for glucose more than 82 ppm would be expected. Compound 2 has therefore been tentatively assigned as phenylmethyl-2-O-  that the aglycon was the so-called "grasshopper ketone"with a β-glucopyranose moiety attached to C-5. This compound is a cumulene with two different terminal substituents (H, acyl moiety) at C-8. This structural feature leads to a chiral axis and due to the stereogenic centers C-3 and C-5, respectively, to two possible diastereomers with different spatial orientation of the terminal groups, which are described as citroside A and citroside B in the literature. The NMR-resonance values of component A fit much better with those of citroside A (Umehara et al., 1988;Zhang et al., 2010).

hr-MS and ESI-MS Identification of Megastigmane glycosides
ESI-MS and HR-MS analyses of the initial combined fractions S3-S6 showed six additional compounds with characteristic fragmentations of megastigmane glycosides. Since the megastigmane aglycones can vary in the degrees of saturation and hydroxylation (Gribble, 1991;Hou et al., 2016), a structural determination could be made in comparison to the isolated compounds 1 and 3. Table 2 depicts the mass fragmentation of identified compounds. They were identified as gusanlungionoside A or B (4), actinidioionoside (5), roseoside (6), 7,8-dihydroroseoside (7) Otsuka et al. (2003).

gC-MS Identification of Megastigmane Aglycones
C13 megastigmane derivatives represent a large group of natural products important as aromatic components in fruits and plants. In order to identify the low molecular weight volatile megastigmane aglycones, capillary GC-MS has been employed. GC-MS of the fractions S3 to S6, as shown in Figure S5 (supporting information), allowed the identification of the four megastigmane aglycones β-damascenone (10), megastigmatrienone (11), 3-hydroxy-β-damascone (12), and 3-oxo-7,8-dihydro-α-ionol (13). The identification was performed by comparing the MS fragmentation patterns with library data (HPC2205, NIST08, Wiley138) and reference compounds. The mass spectra are dispicted in the supporting information. The other peaks were not identified because they were from a fraction, which did not show activity.

In Vitro Pharmacology of Megastigmane Derivatives
Epipremnum pinnatum is used in traditional medicine for treatment of various inflammation-associated conditions. We addressed the question whether megastigmane derivatives identified in this project can inhibit inflammatory reactions in cell types relevant to inflammation, such as endothelial cells and mononuclear leukocytes. To this end, LPS-induced upregulation of the pro-inflammatory PTGS2 (COX-2) mRNA in the monocyte-like THP-1 cell line was used as a readout. In addition to the isolated compounds, α-ionone (14), β-ionone (15), 7,8-dihydro-β-ionone (16), and damascone (17), which all have similar structures, were available as reference compounds. Figure 3 depicts the megastigmane aglycones 10-17. Analysis of the relative inhibitory activity showed that out of the tested compounds, only compound 10 was able to inhibit PTGS2 (COX-2) mRNA expression ( Table 3). The inhibitory effect was dosedependent (Figure 4). Since only 10 demonstrated significant activity, further studies were focused on the pharmacological analysis of this substance.
To characterize the anti-inflammatory properties of 10, endothelial cells (HUVECtert) were pretreated with this compound and then stimulated with LPS for 4 h. 10 inhibited the induction of typical NFκB-regulated pro-inflammatory genes, i.e. E-selectin, ICAM-1, and VCAM-1 (Figures 5A-C), as well as IL-8 (Figure 6). Similarly, in the monocyte-like cell line THP-1, 10 inhibited the induction of the NFκB-responsive genes TNF-α and IL-1β (Figures 5D, E). The effects were not due to cytotoxicity of 10, as illustrated by the lack of cytotoxic effect after 24 h (Figure 7). The concentration dependence of the anti-inflammatory action of 10 was analyzed using E-selectin mRNA expression in HUVECtert cells as readout. The inhibition of E-selectin mRNA expression was concentration-dependent showing a residual activity of 37.9% at a concentration of 5 µM (Figure 8). Similar effects were observed for the inhibition of E-selectin protein expression on the surface of HUVECtert cells (Figure 9) and the secretion of TNFα protein by LPS-stimulated THP-1 cells (Figure 9). The data of Figure 9 show that 10 not only reduced induction of mRNA of pro-inflammatory genes, but also blocked their expression at the protein level.
The experiments described in the previous paragraph were performed using cells stimulated by bacterial LPS. Further  experiments analyzed whether 10 can also inhibit effects induced by pro-inflammatory cytokines that bind to different receptors than LPS. To this end, endothelial cells were stimulated with TNF-α, IL-1β, or LPS in the absence or presence of 10.
All stimuli upregulated E-selectin mRNA in HUVECtert, and the induction was significantly inhibited by 10 (Figure 10). Similarly, 10 inhibited the induction of TNF-α mRNA in THP-1 cells treated by all three pro-inflammatory stimuli (Figure 10). Since these stimuli activate different receptors with different downstream signaling steps, we hypothesized that 10 inhibits the transduction of inflammatory signals at a post-receptor level. The signaling pathway, in which all three upstream pathways converge, is the NF-κB signaling pathway (Sun, 2017).
In order to directly test whether the anti-inflammatory activity of 10 may result from inhibiting the NF-κB signaling pathway, HEK 293 cells stably transfected with a NF-κB-driven luciferase gene were used. 10 inhibited the NF-κB-driven reporter gene transactivation concentration-dependently ( Figure 11A). The IC 50 value obtained was 21.3 µM. The effect was not due to cytotoxicity of 10 as documented by the Cell Tracker Green fluorescence ( Figure 11C). 10 is an active electrophile, and in contrast to other megastigmane FIgUrE 4 | Inhibition of LPS-induced (7.5 ng/ml) PTSG2 (COX-2) mRNA expression by 10. Cells were pretreated with β-damascenone (10) for 1 h followed by LPS stimulation for 3 h. Total RNA was isolated and reverse transcribed to cDNA. cDNA was amplified and expression of PTSG2 (COX-2) was normalized to the expression of GAPDH mRNA. Data are presented as mean ± SE (n = 3).
aglycones, 10 possesses two active sites. This characteristic feature makes 10 a reactive Michael acceptor able to react with thiol groups found in many signal-transducing proteins (Gerhauser et al., 2009). We, therefore, tested whether the inhibitory effect of 10 can be inhibited by adding glutathione acting as decoy for reactive Michael acceptors (Heilmann et al., 2001). Figure 11B shows that 5 mM glutathione largely blocked the NF-κB inhibitory activity of 10 as well as that of the positive control parthenolide, which is also a reactive Michael acceptor (Kwok et al., 2001).
Electrophiles have been shown to covalently modify IKKβ, thus leading to its inactivation (Kapahi et al., 2000;Kwok et al., 2001). This kinase phosphorylates the NF-κB inhibitor IκB and targets it for proteasomal degradation, finally leading to NF-κB activation. Thus, inactivation of IKKβ by electrophiles results in inhibition of NF-κB independently of the upstream mechanisms that induced activation of IKKβ.
In summary, the major finding of this investigation is that β-damascenone is a major active compound of Epipremnum pinnatum, and that it inhibits NF-κB signaling pathway in vitro in human cellular systems that had been activated with different inflammatory stimuli, and that this is most likely  mediated by the electrophilic property of the compound. Our results are in agreement with the data of Gerhauser et al., 2009), who showed that 10 was able to inhibit iNOS expression in LPS-stimulated murine macrophages and to activate the transcription factor Nrf2. Also, the activation of Nrf2 can be promoted by an electrophilic insult that covalently modifies Keap1, which finally leads to Nrf2 accumulation, nuclear translocation and transcriptional activation of respective target genes (Matzinger et al., 2018). Thus, 10 as an electrophilic compound can simultaneously activate Nrf2 and inhibit NF-κB, finally resulting in an antioxidant defense and suppression of pro-inflammatory target genes. Such a pharmacodynamic profile may be especially beneficial for treatment of acute inflammatory conditions, e.g., ischemiareperfusion injury, which is usually accompanied by severe oxidative stress. Therefore, our results show that megastigmane derivatives, in particular β-damascenone and its precursors, may be responsible for the anti-inflammatory effects of Epipremnum pinnatum via inhibition of the NF-κB pathway. It is most likely that its α,β-unsaturated carbonyl moiety is responsible for this effect. Since α,β-unsaturated carbonyl moieties interact with many signal-transduction proteins, it needs to be clarified whether β-damascenone expresses its effects only by targeting proteins within the NF-κB and Nrf2 signaling cascade, or whether it also affects other signaling pathways.

AUThOr CONTrIBUTIONS
Bio-activity guided isolation was performed by S-PP. Structure elucidation including LC-MS and GC-MS was performed by S-PP and OK. S-PP performed the COX-2 mRNA assay. Analysis of NF-κB dependent genes, TNF-α and E-selectin ELISA was performed by TP. NK and TP did cell viability assays. SH, SL, and JR performed the Luciferase reporter gene assays. Graphs and statistical analysis were made by contributors who performed the experiments. S-PP wrote the original draft of the manuscript. RB, VB, VD, and TP designed and conceived the study, contributed to writing and revisions of the manuscript. All authors contributed to writing and revisions of the manuscript. Results are normalized to β2-microglobulin. Data are presented as mean ± SE (n=4). P-values vs LPS-stimulated group are shown as ***<0.001.
FIgUrE 11 | β-Damascenone (10) inhibits TNF-α (2 ng/ml)-induced NF-κB-driven luciferase reporter gene transactivation concentration dependently; the inhibition is reversed in the presence of 5 mM glutathione (GSH). HEK293/NF-κB-luc cells were loaded with cell tracker green (CTG), a probe for vital staining. After 24 hours cells were pretreated for 1 hour with the indicated compounds, and in (B) in addition with 5 mM GSH (as indicated), and activated with TNF-α (2 ng/ml) for 4 hours. Then, (A and B) luciferase activity and (C and D) CTG fluorescence were measured. Luciferase activity is shown normalized to CTG fluorescence. Parthenolide (10 µM) served as positive control. Data are presented as mean ± SD (n = 4) and are normalized to the vehicle control DMSO.
FIgUrE 10 | Inhibition of (A) E-selectin mRNA expression in HUVECtert and (B) TNF-α mRNA expression in THP-1 cells stimulated with different agonists of NF-κB pathway. Cells were pretreated with β-damascenone (10) for 30 min followed by stimulation with LPS (30 ng/ml), TNF-α (0.3 ng/ml) or IL-1β (1 ng/ml) for 4 hours. Basal values refer to vehicle-stimulated cells. Isolation of total RNA, cDNA synthesis and real-time PCR were performed as described in materials and methods section. Results are normalized to β2-microglobulin. Data are presented as mean ± SE (n=4). P-Values are shown as **<0.01 ***<0.001