Synergism and Subadditivity of Verbascoside-Lignans and -Iridoids Binary Mixtures Isolated from Castilleja tenuiflora Benth. on NF-κB/AP-1 Inhibition Activity

Pharmacodynamic interactions between plant isolated compounds are important to understand the mode of action of an herbal extract to formulate or create better standardized extracts, phytomedicines, or phytopharmaceuticals. In this work, we propose binary mixtures using a leader compound to found pharmacodynamic interactions in inhibition of the NF-κB/AP-1 pathway using RAW-Blue™ cells. Eight compounds were isolated from Castilleja tenuiflora, four were new furofuran-type lignans for the species magnolin, eudesmin, sesamin, and kobusin. Magnolin (60.97%) was the most effective lignan inhibiting the NF-κB/AP-1 pathway, followed by eudesmin (56.82%), tenuifloroside (52.91%), sesamin (52.63%), and kobusin (45.45%). Verbascoside, a major compound contained in wild C. tenuiflora showed an inhibitory effect on NF-κB/AP-1. This polyphenol was chosen as a leader compound for binary mixtures. Verbacoside-aucubin and verbascoside-kobusin produced synergism, while verbascoside-tenuifloroside had subadditivity in all concentrations. Verbascoside-kobusin is a promising mixture to use on NF-κB/AP-1 related diseases and anti-inflammatory C. tenuiflora-based phytomedicines.


Isolation and Identification of Compounds
The chromatographic process of methanolic extract from aerial parts of C. tenuiflora allowed the isolation of eight different compounds ( Figure 1). According to UV spectra, retention time and mass spectra, peak 2 corresponded with geniposide, compound 3 with tenuifloroside and 4 with verbascoside. Compounds 1, 5, 6, 7, and 8 were identified using 1 H-and 13 C-NMR.
Compound 1: (Aucubin) 13 11.3, 9.1 Hz, 1H) (Figures S19 and S20). According to UV spectra, the absence of a signal greater than 160 ppm corresponding to C = O in 13 C-NMR was expected. This is corroborated by the 13 C-NMR spectrum. In C. tenuiflora, two iridoids have been reported with these characteristics, aucubin and bartsioside, the difference between them is the presence of an OH at the C6 in aucubin. The signal at 82.9 ppm belonged to CH-OH, confirming the presence of aucubin. Experimental data was compared with that reported by Ersoz et al. [28]. ESI [M + H] + : 347.48 m/z ( Figure S9).
Compound 7: (Eudesmin) 13 Figures S25 and S26). Eleven signals were observed in the 13 C-NMR spectrum. As sesamin, compound 7 was also symmetric. The R-CH 2 -O-CH 2 -R' signal was absent but O-CH 3 signals were observed at 55.9 ppm (3.9 (s) ppm) and 54.1 ppm (3.8 (s) ppm). Chemical shifts coincided for eudesmin [31]. ESI [M + Na] + : 410.87 m/z ( Figure S10).  (Figures S27 and S28). Compound 8 was an asymmetric lignan. A 102.8 ppm signal was found for R-CH 2 -O-CH 2 -R' but the absence of a signal in 6.0 ppm of 1 H-NMR showed that the methylene-di-oxy group was not present. There was also the presence of signals corresponding to methoxy groups (50-55 ppm). Compared to eudesmin, there were differences between 102.8 and 153.4 ppm, corresponding to the aromatic regions, indicating an extra substituent. An extra methoxy in C5 explains the changes in the aromatic region and the signal in 153.4 ppm. Experimental data match with magnolin, according to that reported by Miyazawa et al. [32]. ESI [M + Na] + : 440.13 m/z ( Figure S14).

EC 50 and E max of the Isolated Compounds
It was observed that none of the previously isolated compounds in the concentration range of 0.01-100 and 0.1-200 µM for verbascoside were toxic against RAW-Blue ™ cells ( Figure S29). Therefore, the ability of the compounds to inhibit NF-κB/AP-1 activity was evaluated in those concentration ranges. All isolated compounds inhibited NF-κB/AP-1 and their EC 50 (effective concentration 50) and E max (maximal effect) were calculated ( Table 1).

Pharmacodynamic Screening Using Binary Mixtures
Verbascoside-iridoids and verbascoside-lignans binary mixtures were evaluated looking for a combination with possible presence of pharmacodynamic interactions. So, first it was necessary to corroborate EC 50 determination. Solutions of each compound at its EC 50 were prepared and evaluated in RAW-Blue™ cells, where slight variations were observed, confirming that the determination was accurate (Table S1). Corroborated EC 50 were considered to calculate the expected NF-κB/AP-1 inhibition (%) of each binary mixture. Expected NF-κB/AP-1 inhibition was the inhibition activity calculated considering additivity between the corroborated inhibition at EC 50 of verbascoside and the corroborated inhibition at EC 50 of the different iridoids and lignans (Table S1). Binary mixtures were carried out combining each compound with verbascoside at its EC 50 . Verbascoside-iridoids mixtures had similar NF-κB/AP-1 inhibition and inhibition ratio with respect to expected data ( Table 2). Verbascoside-aucubin and verbascoside-geniposide mixtures inhibited NF-κB/AP-1 by 46.16% and 43.04%, respectively, with an inhibition ratio of 0.84 and 0.89, respectively.
Verbascoside-kobusin and verbascoside-tenuifloroside binary mixtures were selected to look for pharmacodynamic interactions among verbascoside-lignan mixtures because the inhibition ratio value suggested synergism and subadditivity; verbascoside-aucubin was chosen to aim for verbascoside-iridoids pharmacodynamic interactions, despite verbascoside-geniposide presenting a slightly better inhibition ratio, aucubin was easier to isolate than geniposide.

Discussion
Herbal extracts are very complex matrices of different PSM. Most of the time, pharmacological activity is attributed to the major PSM of the herbal extract or standardized fraction. However, the possibility that the PSM present in the herbal extract may work together at different parts and levels of a signaling pathway, in this case, the NF-κB/AP-1 pathway, must be considered. In this work, we tried to understand how PSM, present in the methanolic extract, may work together to inhibit NF-κB/AP-1. For this purpose, it was necessary to determine the presence of pharmacodynamic interactions between the main constituents of C. tenuiflora found in the methanolic extract.
Castilleja tenuiflora is characterized by the presence of glycoside iridoids, flavonoids, phenylethanoids, and lignans in its chemical profile. In this work, we isolated aucubin and geniposide as major iridoids [3,9], verbascoside as a major phenylethanoid glycoside [5], and tenuifloroside, eudesmin, magnolin, kobusin, and sesamin as major lignans. Until now, tenuifloroside was the only furofuran lignan reported for C. tenuiflora [4], so this is the first time that eudesmin, magnolin, kobusin, and sesamin are reported for this species.
According to the EC 50 results, methoxy and methylenedioxy groups were important in NF-κB/AP-1 inhibitory activity, which increased with the number of methoxy groups, as long as there were not methylenedioxy groups on the lignan structure. This result is similar to that found by Yang et al. [35] on RAW 264.7 macrophage cells, where they analyzed the role of methoxy groups on the anti-inflammatory effect of curcumin and realized that curcumin (Cur-OCH 3 ) potently inhibited inflammation in vitro, suppressing lipopolysaccharide (LPS)-induced phosphorylation of IκB kinase (IKK) and degradation of IκBα, more than its synthetic analogues (Cur-OH, Cur-Br > Cur-H, Cur-F, Cur-CH3, Cur-Cl, > Cur-NO 2 ).
Pharmacodynamic interactions were found in binary mixtures using verbascoside as a leader compound. Verbascoside was selected as a leader compound because it is a major constitutive PSM in wild C. tenuiflora methanolic extracts and has demonstrated NF-κB/AP-1 inhibitory activity [4,[36][37][38]. Verbascoside-aucubin and verbascoside-kobusin mixtures showed synergism and verbascoside-tenuifloroside displayed a subadditivity effect in the inhibition of NF-κB/AP-1. The synergistic effect could be due to a simultaneous inhibitory activity between verbascoside and aucubin or kobusin on the different receptors of the NF-κB/AP-1 signaling pathway. Conversely, verbascoside inhibits TAK-1 phosphorylation through Src-homology 2 domain-containing protein tyrosine phosphatases (SHP-1) activation [26]. TAK-1 is the key regulatory point of the bifurcation to AP-1 or NF-κB pathway, and an inhibition upstream to NF-κB or AP-1 nucleus translocation is desirable, aiming to reduce proinflammatory cytokine signaling and inflammatory gene expression. Aucubin blocks nuclear translocation of the p65 subunit of NF-κB and suppresses degradation of the κB inhibitor α (IκBα) [39]. Kobusin also inhibits nuclear translocation of the p65 subunit of NF-κB and IκBα phosphorylation [40]. Also, sesamin, a furofuran lignan, inhibits the expression of TLR-4 [21,41], which is the LPS receptor that initiates the signaling cascade for the activation of NF-kB and AP-1. Kobusin is a furofuran lignan with similar structure to sesamin and it is expected to have a similar mechanism of action. The upstream inhibition of TLR-4 by lignans could be the reason why verbascoside-kobusin presented a higher inhibition ratio than verbascoside-aucubin.
However, verbascoside-tenuifloroside binary mixture did not present results as expected; we thought that it was going to have the same behavior as the verbascoside-kobusin mixture (Figure 2), but only subadditivity was observed, possibly due to chemical interactions between verbascoside and tenuifloroside that might form an adduct with lower pharmacological effect. The adduct can be formed by sugar-sugar interactions between the glucopyranose moieties of the disaccharides in verbascoside and tenuifloroside through hydrogen bonds [42]; sugar-aromatic interactions [43,44] by the formation of CH/π bond between aromatic rings and the glucopyranose moieties presented in tenuifloroside and verbascoside; and/or hydrogen bonds between carbonyl and hydroxyl groups of both compounds.
expression. Aucubin blocks nuclear translocation of the p65 subunit of NF-κB and suppresses degradation of the κB inhibitor α (IκBα) [39]. Kobusin also inhibits nuclear translocation of the p65 subunit of NF-κB and IκBα phosphorylation [40]. Also, sesamin, a furofuran lignan, inhibits the expression of TLR-4 [21,41], which is the LPS receptor that initiates the signaling cascade for the activation of NF-kB and AP-1. Kobusin is a furofuran lignan with similar structure to sesamin and it is expected to have a similar mechanism of action. The upstream inhibition of TLR-4 by lignans could be the reason why verbascoside-kobusin presented a higher inhibition ratio than verbascoside-aucubin.
However, verbascoside-tenuifloroside binary mixture did not present results as expected; we thought that it was going to have the same behavior as the verbascoside-kobusin mixture (Figure 2), but only subadditivity was observed, possibly due to chemical interactions between verbascoside and tenuifloroside that might form an adduct with lower pharmacological effect. The adduct can be formed by sugar-sugar interactions between the glucopyranose moieties of the disaccharides in verbascoside and tenuifloroside through hydrogen bonds [42]; sugar-aromatic interactions [43,44] by the formation of CH/π bond between aromatic rings and the glucopyranose moieties presented in tenuifloroside and verbascoside; and/or hydrogen bonds between carbonyl and hydroxyl groups of both compounds. Moreover, the experimental strategy using a leader compound with known pharmacological activity (NF-κB/AP-1 inhibition) allowed us to identify pharmacodynamic interactions that in a conventional bio-guided separation would not be possible. In bio-guided separation, verbascoside (middle polarity) and non-glycosylated lignans as kobusin (slightly polar) would not be in the same fraction because of their difference in polarity, in other words, following a traditional bio-guided separation would be unlikely to find the verbascoside-kobusin pharmacodynamic interaction. Moreover, the experimental strategy using a leader compound with known pharmacological activity (NF-κB/AP-1 inhibition) allowed us to identify pharmacodynamic interactions that in a conventional bio-guided separation would not be possible. In bioguided separation, verbascoside (middle polarity) and non-glycosylated lignans as kobusin (slightly polar) would not be in the same fraction because of their difference in polarity, in other words, following a traditional bio-guided separation would be unlikely to find the verbascoside-kobusin pharmacodynamic interaction.

Plant Material
Ten plants of wild C. tenuiflora, free of potential host plants in 40

Extraction and Isolation of Chemical Compounds from Methanolic Extracts
The aerial part was separated from the roots and it was air-dried under dark conditions. This plant material (920 g) was ground and extracted with methanol by maceration at room temperature for 48 h. The liquid extract was filtered (No. 1 Whatman filter paper) and concentrated to dryness at 40 • C under low pressure (Büchi-490 rotary evaporator). Solid extract (172 g) was stored at 4 • C for chromatographic separation.
The methanolic extract was subjected to bipartition with an immiscible mixture of ethyl acetate/water. The aqueous fraction (Aqf) was used to separate polar compounds and the ethyl acetate fraction (EAf) was extracted with middle and less polar compounds.
According to their UV spectra, compound 1 corresponded to a glycosidic iridoid without the carboxyl group at C4 (λ max :195 nm). The presence of this carboxyl group at C-4 in compound 2 was established by UV absorption data (λ max :238 nm). In the case of compound 3, this furofuran lignan produced the characteristic UV absorptions (λ max : 207, 229, and 282 nm) and m/z (ESI [M + Na] + : 674.26 m/z) of tenuifloroside previously described by Herrera-Ruiz et al. [4]. Compound 2 UV spectra and retention time matched with the geniposide reported by López-Rodríguez et al. [9]. The presence of geniposide was confirmed by mass spectra analysis (ESI [M + H] + : 389.52 m/z). Compound 4 was a phenylpropanoid glycoside, according to the m/z (ESI [M + H] + : 625.18 m/z). The RT and that reported by Gómez-Aguirre et al. [5] correspond to verbascoside. Compounds 5, 6, 7, and 8 were colorless needle-shaped crystals with an UV spectrum similar to tenuifloroside.

TLC and HPLC Analysis
Analytical TLC was carried out on a precoated Merck silica gel 60F 254 or RP-18F 254 plates. Komarowski reagent was used to derivatize terpenes and lignans and natural products reagent to visualize phenylethanoid glycosides.

UPLC Analysis
This chromatographic analysis was done using an Acquity UPLC (Waters, Milford MA, USA) provided by quaternary pump, autosampler column oven, and a photodiode array-detector. Chromatographic separations were performed in a UPLC BEH 1.7 m-C18 column at a flow rate of 0.4 mL/min. The mobile phase consisted of 0.1% formic acid in water (A) and 0.1% formic acid in acetonitrile (B). The column was held at 100% of A for 1 min and subsequently ramped to 100% of B (curve 6) over 11 min, followed by a 4 min period at 100% of B before a rapid return to 100% of A, and an equilibration period of 2 min. The column was maintained at temperatures of 40 • C. The injection volume was 5 µL and absorbance was measured at a range of wavelength from 190-600 nm. Mass spectrometry analysis was performed and analyzed in a triple quadrupole mass spectrometer (Waters) through an electrospray Z-spray ion source in ESI positive mode. Source and desolvation temperatures were 150 and 400 • C, respectively. A combination of cone voltage of 20 V and capillary voltage of 2.5 kV was used. Nitrogen was employed both as a desolvation gas and cone gas. An MS scan was performed using argon gas as the collision gas [45].

1 H-and 13 C-NMR Experiments
One and two-dimensional Nuclear Magnetic Resonance (NMR) experiments (COSY, HSQC, HMBC) were performed on a Varian INOVA-400 instrument at 400 MHz CDCl 3 or CD 3 OD were used as solvents with tetramethylsilane (TMS) as an internal standard. Chemical shifts (δ) are reported in ppm values and coupling constants are in Hz.

Raw-Blue™ Cells Culture Experiments
Raw-blue™ cells were cultured at 37 • C using manufacturer recommendations (In-vivoGen) in a 24-well microplate until 70-80% confluence (1 × 10 5 cells/well) was reached. This cell line expressed the secreted embryonic alkaline phosphatase (SEAP) gene under the control of a promoter inducible by the transcription factors NF-κB and AP-1. The different treatments were added to each well, after 1 h of incubation with lipopolysaccharide (LPS) at 1 µg/mL. The microplate was incubated for 24 h at 37 • C in a CO 2 chamber at 5%. NF-κB and AP-1 activity was determined indirectly by quantifying SEAP activity in the supernatants. Fifty microliters of supernatant was collected in another 96-well microplate and 150 µL of the QUANTI-blue™ assay buffer (InvivoGen, San Diego, CA, USA) was added to measure the secreted alkaline phosphatase activity [46]. After one hour of incubation at 37 • C, absorbance was measured on a microplate reader (BIO-RAD iMark™) at 630 nm. Medium from cells without treatment was used as negative control. Cells + LPS were used as 100% of NF-κB/AP-1 activity. The inhibition activity (%) of NF-κB was calculated using the following equation: where IA: Inhibition activity (%), A: Absorbance of 100% activity, B: Absorbance of the treatment.
Indomethacin (1 µM) was used as a positive control for NF-κB/AP-1 inhibition. All experiments were performed four times with one repetition each.

Cell Viability on Raw-Blue ™ Cells Using the MTT Method
Cell culture was performed according to specifications made in the previously section. Once a 70-80% confluence was reached, five concentrations in log 10 of 100 µM (100, 10, 1, 0.1, and 0.01 µM) of all isolated compounds were evaluated. The different dilutions were prepared from a stock solution of 2000 µM from each compound. Compounds were diluted in ultrapure water using 0.02% tween 20. Ultrapure water with tween 20 (0.04%) was used as a control. The treated cells were incubated for 48 h at 37 • C in a humid atmosphere and 5% CO 2 . The test was done according to a modified method [47] by quadruplicate. After incubation, the medium was discarded, 80 µL of medium without fetal bovine serum and 20 µL of 3-[4,5-dimethylthiazol-2-yl]-2,5 diphenyl tetrazolium bromide (MTT) (5 mg/mL in phosphate buffered saline (PBS)) were added. Then, it was incubated for 24 h at 37 • C to allow formation of formazan crystals. The supernatant was removed and 100 µL isopropanol was added. Fifteen minutes of agitation were necessary to dissolve formazan crystals. The supernatant was taken and put in a new 96-well microplate. Absorbance was read in a microplate reader (Bio-rad iMark™) at 490 nm. The viability percentage was calculated by: where C: Absorbance of the control, T: Treatment absorbance. To compare and obtain pharmacodynamic interactions in the binary mixtures, it was necessary to determine EC 50 for each compound. Five concentrations in Log 10 from 100 µM (100, 10, 1, 0.1, and 0.01 µM) of iridoids glycosides and lignans and, 200, 100, 10, 1, and 0.1 µM for verbascoside were evaluated in Raw-blue™ cells. Each compound was dissolved in ultrapure water with tween 20 at 0.02%. Inhibitory activity (%) of NF-κB/AP-1 was determined as explained in the Raw-blue™ cells culture experiments section. The EC 50 was calculated by extrapolating the concentration in which 50% of the maximum effect was observed in a linearized effect-concentration model using the inverse.

Inhibition of NF-κB/AP-1 in LPS Stimulated Raw-blue ™ Cells by Binary Mixtures of Iridoid Glycosides and Lignans with Verbascoside
Binary mixtures were prepared with verbascoside as leader and considering the EC 50 of each compound. Treatments were 1.00AEC 50 /1.00BEC 50 , where A was verbascoside and B were aucubin, geniposide, sesamin, magnoline, kobusin, and tenuifloroside. Binary mixtures were prepared in a sterile conic tube, considering the EC 50 of each compound as the final concentration in each well.
After incubating the Raw-blue™ cells in a 24-well microplate, each treatment was added and the inhibition activity (%) of NF-κB/AP-1 was calculated as described in the Raw-blue™ cells culture experiments section.

Pharmacodynamic Interaction of Binary Mixtures of Iridoids and Lignans with Verbascoside in RAW Blue™ Cells
Isobolograms were performed to determine the type of pharmacodynamic interaction that presented the most active binary mixture of each group of compounds (lignans and iridoids glycoside with verbascoside). For this, the EC 50 of the isolated compounds was taken into account, of which subsequent dilutions of the EC 50 were evaluated in a fixed proportions scheme (1.00AEC 50 /0.00BEC 50 , 0.75AEC 50 /0.25EC 50 , 0.50AEC 50 /0.50BEC 50 , 0.25AEC 50 /0.75BEC 50 , 0.00AEC 50 /1.00BEC 50 , where A was verbascoside and B were aucubin, kobusin, or tenuifloroside). The inhibition ratio was calculated by: where OIA: Observed inhibition activity (%), TIA: Expected inhibition activity (%).

Statistical Analyses
EC 50 was determined by linear regression. All experiments were performed in quadruplicate. Statistical analyses were carried out using SPSS software (IBM, version 25.0) using one-way ANOVA followed by Duncan's multiple range test. p-values less than 0.05 were considered statistically significant.

Conclusions
All the compounds isolated from C. tenuiflora inhibited NF-κB/AP-1 on their own, magnolin was the most effective and aucubin was the least active. Methoxy groups are important for NF-κB/AP-1 inhibition activity of lignans. The inhibitory activity on NF-κB/AP-1 of the binary mixtures is explained by pharmacodynamic interactions; verbascosideaucubin showed synergism when verbascoside concentration was higher than aucubin and subadditivity when the opposite. Verbascoside-kobusin mixture showed synergism evaluated, principally at 0.50EC 50 /0.50EC 50 V/K. Verbascoside-tenuifloroside showed subadditivity in all concentrations, mainly 0.50EC 50 /0.50EC 50 V/T. Verbascoside-kobusin could be a promising mixture to be used on diseases associated with increased NF-κB/AP-1 activity. The proposed experimental strategy using binary mixtures and a leader compound was effective in the search of pharmacodynamic interactions. These results are essential in the development of clinical treatments from C. tenuiflora. Also, this is the first time that non-glycosylated furofuran lignans were reported for this species, as well as the NF-κB/AP-1 inhibition activity for tenuifloroside and pharmacodynamic interactions between verbascoside-aucubin, verbascoside-kobusin, and verbascoside-tenuifloroside. Anti-inflammatory C. tenuiflora-based phytomedicines formulas must consider verbascoside iridoids and verbascoside lignans interactions. This information also led to aim the biotechnological approach in order to obtain C. tenuiflora plantlets or cells in bioreactors that biosynthetize verbascoside and kobusin, avoiding tenuifloroside production.

Data Availability Statement:
The data presented in this study are available within the article or supplementary material.