Phenanthrene-enriched extract from Eulophia macrobulbon using subcritical dimethyl ether for phosphodiesterase-5A1 inhibition

Eulophia macrobulbon (E.C.Parish & Rchb.f.) Hook.f. contains a natural PDE5A1 inhibitor, phenanthrene, 1-(4'-hydroxybenzyl)-4,8- dimethoxyphenanthrene-2,7-diol (HDP), a potential agent for the treatment of erectile dysfunction. The aim of this study was to improve the extraction efficiency of HDP from E. macrobulbon by using a more environmentally friendly extraction method, subcritical liquid dimethyl ether extraction (sDME), instead of classical solvent extraction (CSE) and ultrasound-assisted extraction (UAE). The efficiency and quality of the extracts obtained were evaluated using the following criteria: %process yield; solvent amount; extraction time; temperature; %HDP content by LC–MS, bioactivity as inhibition of phosphodiesterase-5A1 (PDE5A1) by radio-enzymatic assay; and chemical profiles by LC-QTOF-MS. sDME provided the highest content of HDP in the extract at 4.47%, much higher than the use of ethanol (0.4–0.5%), ethyl acetate (1.2–1.7%), or dichloromethane (0.7–1.4%). The process yield for sDME (1.5–2.7%) was similar to or lower than the other solvents (0.9–17%), but as long as the process yield is not prohibitively low, the concentration is a more important measure for clinical use. The optimal conditions for sDME extraction were: Extraction time, 40 min; 200% water as co-solvent; sample-to-solvent ratio of 1:8; temperature, 35 °C. Phenanthrene aglycone and glycoside derivatives were the major constituents of the sDME extracts and lesser amounts of phenolic compounds and sugars. The inhibition of PDE5A1 by sDME (IC50 0.67 ± 0.22 µg/ml) was tenfold more potent than ethanolic extract and other extraction methods, suggesting a high probability of clinical efficacy. Thus, sDME was a more efficient, faster, solvent-saving and environmentally friendly extraction method and more selective for phenanthrene when extracted from E. macrobulbon.

. Extraction of E. macrobulbon root by classical solvent extraction (cDCM, cEtOAc, and cEtOH), ultrasound-assistance (uDCM, uEtOAc, and uEtOH), and subcritical dimethyl ether (sDME). Significant values are in bold. Time (t) is the duration of extraction. Process yield (Y) is %weight of extract in dried root powder. % HDP content in extract (B) was measured by LC-MS. All extractions and analyses were done in triplicate. The data are represented as means, ± SD. Extraction efficiencies were calculated by divided %yield and % HDP content in extract with time and solvent volume (v). www.nature.com/scientificreports/ changed little across the three time points. Increasing the amount of all three solvents also increased the process yield (Y) at all time points. The percentage of bioactive compound or HDP content (B) in the crude extract was similarly increased. However, both the Y and B parameters for cEtOH at 72 h appeared unaffected by the increase in solvent. This indicated that the extraction was nearly complete under these two conditions (72 h and a 1:20 sample to solvent ratio). In contrast, 72 h extraction with cDCM and cEtOAc showed further process yields (Y) and %HDP contents. Nevertheless, higher volumes of all three solvents were associated with lower extraction efficiencies Y/v and B/v ( Table 1). The extractable HDP amount in mg from one kilogram of dried plants using different solvents was then compared (Table1). The result showed that the overall the extractable HDP amount was greatest for cEtOH (~1000 mg/kg), slightly less for cEtOAc (~400 mg/kg) but miserable for cDCM (~200 mg/ kg). However, the extractable mass or crude extract is further used as an ingredient in nutraceutical, cosmetic, pharmaceutical and food industries, so higher bioactive content in crude extract means higher therapeutic efficacy. The extractable mass from EtOH gave very high %process yield but negligible %HDP content in the extract compared to other solvents (Table1). This is attributed to the fact that EtOH was non-specific phytochemical extraction for E. macrobulbon while EtOAc and DCM gave better selective HDP extraction. Ultrasonic assisted extraction produced both process yield (Y) and HDP content (B) equivalent to classical extraction using the appropriate solvent and the most extreme protocol conditions, but within only 40 min and a 1:10 sample-to-solvent ratio. Thus, ultrasonics significantly increases the extraction efficiency and the extractable amount of HDP from dried plant ( Table 1).
In both conventional and ultrasonic extraction protocols, DCM and EtOAc were the most selective solvents for extraction of the target HDP compound. However, DCM is toxic and reactive in the atmosphere, a property that misaligns with the idea of herbal medicines being natural and healthier than synthetic medicines.
Optimization of subcritical fluid dimethyl ether extraction. Subcritical liquid dimethyl ether extraction: Liquid dimethyl ether (DME) is becoming increasingly popular for plant extractions. Here, we explored DME as an alternative solvent to maximize the HDP content and bioactivity of E. macrobulbon. We started with a temperature of 35 °C and 30 min as used by others, e.g, 26,31 and then systematically varied the sample-to-solvent ratio, extraction period, extraction temperature, and adding co-solvents, water, or EtOAc ( Fig. 1). The optimum extraction values were selected for each variable and adopted as a fixed value for the next series of determinations.
For each extraction in all protocols using DME, the HDP content was consistently higher than the classical and ultrasonic extraction methods using DCM, EtOAc and EtOH, with increases of ~9-fold, 5-fold and 4-fold respectively. The process yield of the extract with DME was similar to that with EtOAc and DCM, but EtOH extracted a larger bulk (Table1). Fig. 1a showed the effect of sample-to-solvent ratio on the process yield and HDP content. Using DME twice of the dried plant (w/w) (sample-to-solvent ratio 1:2) was not sufficient for the extraction. The sample-to-solvent ratio that provided the optimal content of HDP (~2.8%) was 1:8. Larger solvent volumes or prolonged extraction decreased the apparent content of bioactive HDP as seen elsewhere 34 . This is due to the fact that the overall process yield of the process is increased, while at the same time the risk of undesirable ingredients increases.
The extraction periods were varied from 20 to 120 min and we found that the constant plateau of %yield and %HDP were reached at 40 min (Fig. 1b). The extraction period of 40 min was then selected for further experiments to minimize the time consumption during extraction. The extraction temperatures of 30 to 50 °C were studied (Fig. 1c). It was found that the %yield and %HDP reached the maximum plateau at 35 °C. Therefore, the temperature of 35 °C was selected for further experiments to minimize energy consumption during extraction.
In the classical extraction protocol, EtOAc was an effective solvent for the extraction of HDP and it was classified as a green solvent. So, EtOAc was used as co-solvent in the DME extraction protocol. The result showed that addition of DME with up to ~ 40% EtOAc increased the %HDP in the extract, but further increase in EtOAc content resulted in a decrease in %HDP (Fig. 1d). At 500% EtOAc, the solvent yields an extract with similar properties to one without DME.
Water is commonly used as co-solvent in DME because it is partially miscible in DME solvent and has low cost. Initially, 0.5-10 g of water was added to 5 g of powder, resulting in a sticky mass that increased the process yield of the extract and the HDP yield (Fig. 1e). 10% water was likely absorbed by DME at a pressure and temperature in the extraction chamber 29 and consumed by hydration of the plant root powder constituents 35 . To our knowledge, the most favorable extraction was observed at 200% water, where most of the water could hydrate or swell the plant matrix, so that DME could easily penetrate to break and extract the plant matrix under pressure 36 . Moreover, the presence of water could lead to higher overall solvent polarity, which ultimately improved the extraction process 37 .
The low extractant concentration in the aqueous phase then provides a steep diffusion or unbinding gradient between the hydrated particles. In our experiments, the mixture of both phases was collected and dehydrated, with the DME forming depository for moderately non-polar compounds. Despite this mechanistic uncertainty, DME with 200% added water was 3-5 folds more efficient for HDP content than the next best extraction protocol, cDCM, cEtOAc or uDCM, uEtOAc (Table 1). In addition, the method is fast and requires a fairly economical amount of solvent. Interestingly, the extraction efficiency parameters with B/t and B/v of sDME were about threefold higher than the best classical and ultrasonic assisted protocol ( Table 1). The extractable HDP amount or recovery from the dried plant using sDME reached a peak value of ~ 1000 mg/kg which was equivalent to that of cEtOH (~ 1000 mg/kg) and uEtOH (~ 1000 mg/kg). This indicates that sDME could acheive the maximum extraction of HDP from the dried plant. Moreover, the crude extract from sDME exhibited the highest HDP content among the classical and ultrasonic assisted extractions using EtOH, DCM and EtOAc.  Fig. 2. Extraction of saccharides (retention time, 1-2 min) was prominent for the more polar solvents (EtOH and EtOAc), whereas DCM extracted only compounds that eluted after ~ 6 min (Fig. 2B). In contrast, EtOH extracted material that eluted mostly before 10 min. For DME extraction, 23 compounds were identified. The compounds of potential pharmacological interest were polyphenols and glycosides (eluted at 3.0-7.5 min) and of current interest, phenanthrenes as glycosides (7.0-9.1 min) and less polar phenanthrene aglycones (9.5-14.5 min) ( Table 2). The phenanthrene derivatives were found in the same range with the identifiable peak area in percentage for all extracts, cEtOH, cEtOAc, cDCM and sDME were 62.71, 68.26, 64.76 and 62.68%, respectively. The more polar phenanthrene glycosides were predominantly existed in cEtOH, cEtOAc and sDME with 48.62, 33.55 and 23.77%, respectively. The major phenanthrene glycoside compounds in those extracts were compounds 5, 8 and 9, which possess core aglycone mass of 284 [M] + , which is the same mass as the aglycone of compound 14. Only 1.35% of phenanthrene glycosides were found in DCM. This was due to the polarity indices   19 . Worth noting, a peak of the natural PDE5A1 inhibitor, HDP (compound 21) was predominant in sDME with 13.19% of the total identifiable peaks, EtOAC was almost as high with 7.80%. Some phenolic compounds such as 4-hydroxybenzaldehyde and methyl arbutin were found in the extracts ( Table 2).
Inhibition of PDE5-1A. The favorable extracts from the different extraction methods were evaluated for their PDE5A1 inhibitory activity by enzymatic and [ 3 H]cGMP radioassay, and the result is shown in Table 3. We found that the PDE5A1 inhibition of all extracts was related to the %HDP content in the extract. More potent PDE5A1 inhibitory activity was observed at higher %HDP content (Table 3). This result supports that HDP is a suitable biomarker for the PDE5A1 inhibitory activity of this plant 17 . Moreover, PDE5A1 inhibitory activities of the extracts of the classical method were slightly stronger than those of the ultrasonic-assisted method in all solvents. This is due to the fact that the extracts of the ultrasonic-assisted method contain more undesirable compounds than others. Both extracts of EtOAc and DCM showed more potent PDE5A1 inhibition than EtOH, indicating that DCM and EtOAc with polarity indices of 3.1 and 4.4 could selectively extract PDE5A1 inhibitors than EtOH with polarity indices of 5.5. The extract with DME/200% water showed the most potent PDE5A1 inhibition (Table 3, Fig. S1.) compared to DCM and EtOAc (twofold lower) and EtOH (~ tenfold lower). This confirms that DME (with/without water) is the most selective solvent for PDE5A1 inhibitors. The differences of bioactivity are roughly consistent with the differences in HDP content (Table 1). In addition, other phenanthrenes in E. macrobulbon root that are known PDE5A1 inhibitors 14,19 might play roles.  Table 2. Extraction protocols were: (A) 10 g water added 5 g powdered E.
Macrobulbon root and extracted with 40 g DME (method of sample no. 31,

Conclusions
This study investigated the potential of phenanthrene enrichment extraction using a more environmentally friendly and safer technique: extraction with liquefied dimethyl ether from E. marcobulbon. We found that an optimized sDME protocol with an extraction time of 40 min, addition of 200% water to sDME (%w/w), a sample to solvent ratio of 1:8, and a temperature of 35 °C gave a process yield of 1.55% with an HDP concentration of 4.47% in the resulting extract. The process yield was comparable or in some cases lower than the optimal protocols using cDCM, cEtOAc and cEtOH. However, HDP concentration was dramatically higher using sDME than the best non-DME protocol (cEtOAc gave a maximum HDP concentration of 1.75%), CSE and UAE in all solvents. A high HDP concentration is critical for clinical applications, as higher compound purity is likely to lead to more predictable and effective results. Indeed, we found that the extract obtained with our optimized sDME protocol exhibited approximately tenfold higher efficacy in inhibiting PDE5A1 compared to the uEtOH extract, suggesting a promising clinical application. In addition to a high HDP concentration and promising PDE5A1 inhibition, sDME is a more environmentally friendly and safer solvent than other organic solvents such as DCM, chloroform, petroleum ether, benzene and the others used here. The chemical fingerprint profile of the sDME extract was identified using LC-QTOF /MS and could be classified into 4 main classes: sugars, phenolic compounds, phenanthrene glycosides and phenanthrene aglycones. The main constituent of the extract was phenanthrene derivatives. Thus, the use of sDME is a promising technique for selective enrichment of phenanthrene extract from E. macrobulbon.

Materials and methods
General materials. Dimethyl    www.nature.com/scientificreports/ ground into fine powder (4 kg) and sieved (150-170 µm) and stored in a desiccator at room temperature until use.

Isolation of the main bioactive compound from E. macrobulbon. The isolation of HDP followed
previous reports with some modifications 14 . In brief, dried powders of E. macrobulbon (4 kg) were macerated two times with 95% EtOH (28L), then filtered and the solvent was removed under reduced pressure to provide 450 g of crude extract (11.2% yield). The extract (384.4 g) was dissolved in 100% MeOH and partitioned twice with hexane. The hexane part was discarded and the MeOH part was diluted with DI water to give 20% MeOH and partitioned twice with DCM. The DCM portion was dried under reduced pressure to yield 19.9 g of crude extract. The DCM residue was mixed with silica gel and loaded on to a silica gel chromatography column (i.d. 103 × 40 mm). The mobile phase for gradient elution was 100%DCM to 0.5-4% MeOH in DCM. Eighteen fractions were collected (EMD-1-18). The target compound was monitored to reference standard of HDP by TLC using DCM:MeOH (9.5:0.5 %v/v) as the mobile phase (the Rf value was around 0.3). The fraction of EMD-14 was obtained 0.49 g and chosen for further isolation. EMD-14 (0.24 g) was dissolved in methanol and subjected in a Sephadex LH-20 column (i.d. 1.5 × 200 cm) eluting with 100% MeOH to yield 19 fractions. Three fractions (EMDLH14-10 to EMDLH14-12) were pooled and evaporated and recrystallized with MeOH/DCM to give 0.19 g of crystalline bioactive compound (HDP). The spectroscopic data of 1 H-NMR and MS were in agreement with those reported in the literature 14 . The purity and spectroscopic data of HDP are described in supplementary materials, Figs. S2, S3 and S4. The isolated HDP was used as a reference standard to quantitatively control the quality of the extracts using LC-MS.
The sample-to-solvent ratio (w/w) was varied from low to high (1:6.25, 1:10 and 1:20), each maceration period was either 24, 48 or 72 h. Ultrasound-assisted extraction; the fine powder (10 g) was macerated with different organic solvents, EtOH, EtOAc and DCM at a fixed sample-to-solvent ratio of 1:10 at 40 °C for 40 min. The ultrasound frequency was set at low to high intensity (100 kHz to 1 MHz) (Transonic, Themo Fisher Scientific, Göteborg-Sweden). Whenever the extraction process reached the time course, the extraction samples were filtered (Whatman paper 2 µm) and then dried under reduced pressure to provide the crude extract. The extract was then dried over a desiccant for 48 h and weighed.
Subcritical fluid dimethyl ether extraction; the dried powder (5 g) was mixed with the required volume of water or co-solvent and the mixture was placed in cellulose thimble (30 × 100 mm) along with a magnetic bar of 15.9 × 8 mm (length × diameter). The DME extractor was applied for this work and the apparatus was schematically presented in reference of 38 . The thimble was then placed in an extractor (100 ml total volume of stainless-steel with a closed system). Liquefied DME was filled into the extractor at the required solvent to solid weight ratio. The extraction was carried out at a controlled temperature and stirring speed of 500 rpm required time (see below). After extraction, DME and the liquid sample were passed through a stainless steel filter (5 µm pore diameter, Swagelok). The chamber was inverted to a 75 ml Erlenmeyer flask. The remaining liquid sample was then dried over desiccant for 48 h, the amount weighed and the yield determined.
The PDE5A1 inhibitory bioactivity, %HDP content and chemical profile were determined for all extracts. The extraction efficiency was also evaluated using the following parameters; (1) Y; Percentage of process yield (%w/w) (2) B; Percentage of HDP content in the extract (%w/w) (3) t; Extraction period (min) (4) v; Solvent amount (5) Y/v and Y/t (6) B/v and B/t (7) Extractable HDP amount to dried plant (mg/kg) These parameters were determined for all extraction methods, classical solvent extraction, ultrasound-assisted extraction, and subcritical fluid dimethyl ether extraction. During the analysis, the stability of the LC-MS system was checked by using QC1 (concentration of 1.5 µg/ ml) before starting each experimental batch. In addition, QC1 was added for the injections at the beginning, middle and end of the experiment to evaluate the LC-MS system and the stability of HDP throughout the analysis of the sample batch.

Qualitative analysis of E. macrobulbon extracts by LC-ESI-QTOF-MS.
Conditions for LC-MS to measure secondary metabolites in E. macrobulbon samples were determined using a Zorbax Eclipse Plus C18 (4.6 × 100 mm, 3.5µm) column and gradient elution with 0.1% formic acid in water (A) and 0.1% formic acid in ACN (B). The elution program ran for 0 min, 5%B; 0-6 min, 35%B; 6-10 min, 50%B; and 10-18 min, 20% B with a follow-up time of 2 min (post-run). The flow rate was 0.6 ml/min, the injection volume was 10µl, and the column temperature was maintained at 35 °C.
The Samples from a suitable extraction condition were prepared at 5 mg/ml in 100% MeOH and diluted to 50 µg/ml. They were then filtered through nylon syringe filters with a pore size of 0.45 μm before injection into the LC system.

Preparation of phosphodiesterase-5 (PDE5-A1). HEK293 cells were grown in DMEM supplemented
with 10% FBS, in 75 mm flasks at 37 °C in a humidified 5%CO 2 . A human PDE5A1 plasmid, a gift from Professor Dr Joseph A. Beavo, University of Washington, Seattle, WA, USA, were sub-cloned into a pcDNA3 vector containing an ampicillin resistant gene. The human PDE5-A1 plasmid was scaled up and purified using Hipure plasmid Maxiprep kit (Invitrogen-PureLink). HEK293 cells were transfected with human PDE5A1 plasmid using Lipofectamine-2000 following the company protocol. After 2 days of transfection, PDE5-A1 expression was induced by a selective antibiotic (Geneticin (G418, Gibco), 1 mg/ml) for 7 days. The surviving cells were sub-cultured in DMEM, supplemented with 10% FBS in 175 mm flasks at 37 °C in a humidified 5% CO 2 atmosphere, and the cells further cultured until they reached 90-100% confluence. The cells were then harvested using a scraper and lysed by sonication in 1 ml of Tris buffer [50 mM Tris pH 7.5, 2 mM EDTA, 1mM dithiothreitol (DTT) and 1:100 of 100 mM PMSF]. The homogenate was centrifuged at 4 °C for 20 min and the supernatant was used as a source of PDE5A1. A PDE5 inhibitor, sildenafil, was used to confirm the presence of PDE5A1 enzymatic activity.
Measurement of PDE5-A1 enzyme activity. To assess PDE5A1 inhibition, a reaction mixture comprising 20 µl of reagent A (100 mM TrisHCl (pH 7.5), 100 mM imidazole, 15 mM MgCl 2 , 1.0 mg/ml BSA and 2.5 mg/ml snake venom), 20 µl of 10 mM EGTA, 20 µl of PDE5A1 solution, and either 20 µl of test sample or solvent (5% DMSO) only as a control. The reaction was started by adding substrate 20 µl of 5 µM [ 3 H]cGMP (~50,000 cpm) and performed at 30 ºC for 40 min. Then, 100 µl of 50% DEAE resin was added to the reaction. After shaking for 10 min, the resin was allowed to settle (20 min), the supernatant was treated with a second cycle of 50% DEAE resin. This supernatant (100 µl) was shaken with 200 µL of MicroScint-20 and tritium counted on a Top-Count NXT scintillation counter (PerkinElmer, USA) for 2 h. The PDE5A1-hydrolyzed <25% of the substrate. Each was performed in duplicate in 96-well plates 27,28 .
In preliminary screening, samples of plant extract and pure compound were tested at final conc of 50 µg/ml and 10 µM respectively. All samples were dissolved in DMSO and diluted with water. DMSO was limited to 1% in the final assay medium. When PDE5A1 inhibition was >80%, samples were further diluted and re-analyzed. IC 50 s were calculated using Prism software (Graph Pad Inc., San Diego, CA). Sildenafil was used as the positive control.