Article pubs.acs.org/jnp
Standardized LC×LC-ELSD Fractionation Procedure for the Identification of Minor Bioactives via the Enzymatic Screening of Natural Extracts Paul Coulerie,†,‡ Yann Ratinaud,† Sofia Moco,† Loraine Merminod,† Martine Naranjo Pinta,† Julien Boccard,‡ Laurent Bultot,† Maria Deak,† Kei Sakamoto,† Emerson Ferreira Queiroz,‡ Jean-Luc Wolfender,‡ and Denis Barron*,† †
Nestle Institute of Health Sciences, EPFL Innovation Park, H, CH-1015, Lausanne, Switzerland School of Pharmaceutical Sciences, EPGL, University of Geneva, University of Lausanne, CMU, 1, Rue Michel Servet, 1211, Geneva 4, Switzerland
‡
S Supporting Information *
ABSTRACT: To identify natural bioactive compounds from complex mixtures such as plant extracts, efficient fractionation for biological screening is mandatory. In this context, a fully automated workflow based on two-dimensional liquid chromatography (2D-LC × LC) was developed, allowing for the production of hundreds of semipure fractions per extract. Moreover, the ELSD response was used for online sample weight estimation and automated concentration normalization for subsequent bioassays. To evaluate the efficiency of this protocol, an enzymatic assay was developed using AMP-activated protein kinase (AMPK). The activation of AMPK by nonactive extracts spiked with biochanin A, a known AMPK activator, was enhanced greatly when the fractionation workflow was applied compared to screening crude spiked extracts. The performance of the workflow was further evaluated on a red clover (Trifolium pratense) extract, which is a natural source of biochanin A. In this case, while the crude extract or 1D chromatography fractions failed to activate AMPK, semipure fractions containing biochanin A were readily localized when produced by the 2D-LC×LC-ELSD workflow. The automated fractionation methodology presented demonstrated high efficiency for the detection of bioactive compounds at low abundance in plant extracts for high-throughput screening. This procedure can be used routinely to populate natural product libraries for biological screening.
H
readout (fluorescent or quenching compounds) and promiscuous inhibitors15−17 that bind nonspecifically to proteins (e.g., tannins)18 or alter biological membranes.19 The residual complexity20 of the compounds purified from natural sources may also be a problem since minor contaminating species can significantly interfere with the biological activity of the major one.21,22 Another important issue while testing crude extracts is the high dynamic range among the concentrations of its constituents, making the active compounds too diluted to detect. All of these problems are often responsible for false positive or negative results during biological screening. Several strategies have been used to overcome the issues associated with extract complexity, as reviewed by Abel et al.23 and more recently by Potterat and Hamburger.24 These were based on solid-phase extraction (SPE) pretreatment on polyamide to remove tannins and semiautomatic prefractionation by liquid chromatography.25,26 Another prefractionation method based on SPE on diol was recently published.27 All of
istorically, natural product (NP) chemistry has been closely associated with the search for bioactive compounds of therapeutic interest for drug discovery. Bioactivity-guided fractionation has yielded many successful discoveries1−3 and the development of many industrial blockbuster drugs such as anti-infective and anticancer agents.4,5 Since they are considered privileged scaffolds that evolved specifically to interact with proteins,5 it is assumed that NP-based screening libraries yield higher hit rates than synthetically produced small-molecule libraries.6 This partly explains why Nature is still considered a promising reservoir of bioactives.7,8 Nevertheless, NP research is associated with a number of challenges, such as difficulties in “bioprospecting”, which are related to the necessity of adhering to biodiversity agreements,9−12 long processing times for active compound identification,13 difficult resupplies, and incompatibility with high-throughput screening (HTS) methodologies.14 HTS was developed to increase the chance of detecting and identify new bioactive compounds, but it is not well adapted to detect them when present in crude extracts. This is ascribed to the abundance of compounds that interfere with the assay © 2016 American Chemical Society and American Society of Pharmacognosy
Received: July 7, 2016 Published: October 28, 2016 2856
DOI: 10.1021/acs.jnatprod.6b00628 J. Nat. Prod. 2016, 79, 2856−2864
Journal of Natural Products
Article
Figure 1. Newly established procedure for the fractionation of a crude extract for bioactive compound identification during enzymatic highthroughput screening.
2D-LC×LC. The ability of the resulting fractions to elicit AMPK activation was compared between the two techniques.
these approaches have attempted to produce simpler fractions prior to biological assays. In this context, the semipreparative 2D-LC×LC system represents a promising alternative. This automated system involves several fractionation steps with different column selectivity and thus showed high chromatographic potential. The 2D-LC×LC technique has been occasionally applied in the past as a prefractionation step before rapid phytochemical characterization,28 sensory evaluation of the fractions,29 or deconvolution of active molecules such as inhibitors of Aβ production30 and α-glucosidase inhibitors.31 In the two last cases, the time gained in the fractionation process was the major identified added value for the 2D-LC×LC system. No attention has been paid to the issue of detecting minor and moderately bioactive components. In this study, an enzymatic HTS bioassay based on AMPactivated protein kinase (AMPK) was used to validate the workflow. AMPK is a central energy sensor and regulator of energy homeostasis.32 AMPK is activated when cells/tissues are exposed to metabolic stresses that lower cellular energy status by decreasing the catabolic generation of ATP or by promoting ATP consumption. In response, AMPK functions to restore energy homeostasis by slowing the rate of anabolic pathways and other ATP-consuming processes while accelerating ATPproducing catabolic pathways. AMPK is considered a major drug target to act against the growing epidemic of metabolic disorders because its activation by pharmacological and physiological interventions elicits metabolic responses, which are expected to counteract the metabolic abnormalities associated with obesity, insulin resistance, and type 2 diabetes.33 The aim of the present work was to include the 2D-LC×LC method in a new fully automated procedure connected to an inhouse HTS platform by generating prepurified fractions at a normalized concentration. Three steps are involved: (i) sample preparation, (ii) extract separation via 2D-LC×LC, and (iii) fraction processing into 96-well plates for HTS. The use of evaporative light scattering detection (ELSD) to quantify all of the collected fractions and normalize their concentrations before HTS was investigated. The efficacy of this procedure was assessed based on its ability to detect minor compounds of moderate bioactivity, and AMPK HTS was used as a case study. The detection of the activity of a previously reported natural AMPK activator, biochanin A,34−36 spiked into nonactive extracts was evaluated with and without fractionation. Finally, a red clover extract that naturally contains biochanin A was fractionated either by conventional 1D chromatography or by
■
RESULTS AND DISCUSSION The outline of our procedure is displayed in Figure 1. In the preliminary phase, four plants that were already well chemically characterized and that represent a wide variety of natural matrices and chemical compositions were selected as models for the evaluation of the procedure. These were selected to evaluate their matrix effects after spiking with a known AMPK activator because they are nonactive on AMPK. These four model biological matrices were the defatted methanolic extracts from olive leaves (Olea europaea L.; Oleaceae), bilberry fruits (Vaccinium myrtillus L.; Ericaceae), white willow bark (Salix alba L.; Salicaceae), and burdock roots (Arctium lappa L.; Asteraceae). In a second phase, an extract of a plant known to contain an AMPK activator (flowers of Trifolium pratense L.; Fabaceae) was processed with the same procedure to evaluate our ability to detect a minor and moderately active compound. All extracts were subjected to the fractionation process. The semipreparative automated 2D-LC×LC system is designed primarily for the fractionation of medium-polar compounds that may represent a low percentage for polar MeOH and aqueous extracts. Furthermore, the maximum extract loading capacity of the 2D system is 250 mg. Thus, medium-polar compounds may end up extensively diluted after the fractionation process when directly loading polar extracts. Since our focus was indeed on MeOH extracts, it was advantageous to enrich the medium-polarity components before loading the 2D-LC×LC system (Figure 1). Extract enrichment was performed using SPE C18 cartridges in which all four model methanolic extracts were loaded and eluted sequentially with water, MeOH, and i-PrOH. Three subfractions per extract were then collected and weighed (Figure S1, Supporting Information). For the four extracts, the polar subfraction (water) represented more than 50% (w/w) of the initial extract, while the medium-polar (MeOH) subfraction yielded 10−20% (w/w). Thus, this pretreatment allowed for the enrichment of the medium-polar compounds in the MeOH subfractions before fractionation. Then the SPE-enriched extracts could be submitted to a further 2D separation process. The separation of the SPE-enriched olive leaf MeOH extract yielded 176 fractions and a recovery of 46% for the 2D-LC×LC step. In comparison, the fractionation of the same extract without SPE pretreatment yielded 2 times fewer fractions and a much lower recovery for the 2D-LC×LC step (11%). This 2857
DOI: 10.1021/acs.jnatprod.6b00628 J. Nat. Prod. 2016, 79, 2856−2864
Journal of Natural Products
Article
Figure 2. Linear correlation between the ELSD signal after 2D-LC×LC fractionation and the fraction weight over the range of 0−2 mg for the following four plant extracts: olive leaves, bilberry fruits, white willow bark, and burdock roots.
libraries,38 profiling natural extracts,39 or determining the fraction collection threshold during the purification of combinatorial libraries.40 Parallel LC-MS-ELSD methods have been developed for the analysis of natural product libraries.25,41 However, the fractions produced from the extracts of interest were analyzed separately, which represents an additional step as compared to our procedure. To the best of our knowledge, this is the first time that the ELSD detector has been used directly online for generating fractions for screening at a standard concentration without the need of any additional analyses nor weighting steps. The instrumental linearity of the ELSD signal with the fraction weight was tested on the four SPE MeOH subfractions of the plant extracts (Figure 2). The 2D-LC×LC-ELSD fractionations were conducted using preweighed collection tubes. Accurate fraction weights were then obtained after freeze-drying. Using the optimized parameters, good linearity between the ELSD response and the fraction weight (r2 = 0.90) was observed for the four extracts fractionated into fractions of less than 2 mg (Figure 2). For fractions greater than 2 mg, the ELSD signal proved to be saturated, which led to underestimating the corresponding fraction weight. Consequently, this error could overestimate the biological activities of the fractions obtained in high amounts, which were quite rare in the four studied model extracts (i.e., fractions greater than 3 mg represented less than 4% of all collected fractions). However, the ELSD performance in terms of time and cost largely compensates for this approximation from the perspective of screening. Online quantification based on ELSD of the four tested extracts showed a standard deviation of 30%, which was found to be satisfactory for screening purposes. The third step of the fractionation workflow consisted of transferring the fractions from the 2D-LC×LC collection to the HTS platform (Figure 1). The fractions from the 2D-LC×LC fractionation were collected into 48-deep-well plates containing 6 mL tubes. On the basis of the ELSD signal for fraction weight evaluation, aliquots corresponding to 0.1 mg/fraction were transferred into 96-well plates, freeze-dried, and diluted with DMSO to achieve a concentration of 5 mg/mL. Two identical
illustrated the higher concentration of medium-polar compounds after using the SPE. The efficacy of the pretreatment was further shown by comparison of the ELSD chromatograms obtained during the first step of the 2D fractionation with or without pretreatment (Figure S2, Supporting Information). As a second step in the workflow, the enriched extracts were loaded and fractionated on the 2D-LC×LC system using an automated optimized protocol based on relative orthogonal separations on different reversed-phase columns, adapting the chromatographic conditions to the polarity (Figure 1). After 2D separation, the fraction collector was able to collect fractions in deep 48-well plates, which could be used for both analytical and biological screening purposes. Application of this approach yielded 100−200 fractions per extract. Analysis of the various fractions obtained by UHPLC-UV-ELSD/MS with several extracts demonstrated that many compounds were already obtained in a pure (i.e., a single peak) or semipure form (i.e., fractions that contain only two or three peaks with a predominant one representing more than 80% of the fraction), according to the ELSD chromatogram. Such samples were thus well-adapted to HTS screening with an enzymatic assay. An important characteristic of the procedure is the generation of fractions that can be directly submitted to HTS after aliquoting. Enzymatic HTS assays rely on testing samples with known concentrations for a molecular target. In the case of undefined mixtures, this implies that at least the mass of the fraction has to be correctly evaluated. This is performed traditionally by weighing all fractions, a process that is long and tedious, especially when hundreds of fractions have to be measured. It is also an important source of error due to the handling and the different evaporation efficiencies from the presence of different solvents, such as water, which can affect the fraction weight. Here, an online weighing procedure was designed using the ELSD of the 2D-LC×LC-ELSD system to estimate the fraction weight used for efficient aliquoting into 96-well plates. ELSD is a universal detector that was previously described for absolute quantification of pure natural products.37 ELSD has been used as a complementary method for detecting non-UV-absorbing compounds in generating natural product 2858
DOI: 10.1021/acs.jnatprod.6b00628 J. Nat. Prod. 2016, 79, 2856−2864
Journal of Natural Products
Article
Figure 3. EC50 of biochanin A on AMPK tested in 1% aqueous DMSO (A) and spiked in four extracts (B−E): B, biochanin A spiked into olive leaf extract; C, biochanin A spiked into bilberry fruit extract; D, biochanin A spiked into white willow bark extract; and E, biochanin A spiked into burdock root extract. In each experiment, a blank control (unshaded dots) was obtained without spiking biochanin A using 0.1% DMSO(aq) in A and the crude extracts in 0.1% DMSO in B−E. The dotted lines on each graph correspond to 20% enzyme activation, which is the limit over which an extract or compound is considered to be a “hit”.
batches of 96-well plates were generated (Figure 1), the first one for screening and the second one for UHPLC-MS analysis for later dereplication of the bioactive fractions.39 This automated procedure was then evaluated for a specific HTS aiming to detect AMPK bioactivators of plant origin. A previously reported natural AMPK activator, biochanin A, was used as a positive control.34−36 Biochanin A is not a potent and specific activator compared to synthetic AMPK activators such as A769662 (thienopyridone), 911 (benzimidazole derivative), and compound 2/13 (furan-2-phosphonic derivative).42 Nevertheless, biochanin A was chosen to evaluate how the developed automated procedure could highlight even weak AMPK natural activators in complex natural extracts. The EC50 of our analytical standard of this isoflavone was measured and was equal to 10 μM with an efficacy of approximately 40% (Figure 3A). To assess the matrix effects of natural extracts on the stimulation of AMPK activity, biochanin A was spiked into the four selected model plant extracts to achieve final concentrations of 0−100 μM in the extracts. The four spiked extracts were compared to the pure standard in terms of activity dose− response curves (Figures 3B−E). First, all extracts without biochanin A (see the blank controls in Figure 3B−E) resulted in a decrease in the signal corresponding to the basal activity of AMPK (approximately −15% for the bilberry fruit and burdock extracts, −55% for the olive leaf extract, and −75% for the white willow extract). This corresponds either to inhibition of the enzyme or to an interference effect on the AMPK HTRF assay. Given that the EC50 of biochanin A was not significantly affected by these different natural matrices, it was assumed that all extracts contained compounds that interfere with the assay readout, leading to a reduction of the observed signal. Second, the activity due to biochanin A was rarely detectable when associated with a complex natural extract. Under the screening conditions, an extract is considered positive when it generates an increase of at least 20% of the AMPK activity (highlighted
by a dotted line in Figure 3). In this case, the activity of biochanin A was detected in the burdock root extract (Figure 3E) only starting at 2 μM. This threshold for bioactivity detection, corresponding to 5% (w/w) concentration in the extract, indicated that only the activities of major compounds or highly potent minor compounds could be detected in the crude extracts. These results further highlighted the problems facing the detection of moderately active compounds and confirmed the poor efficiency of enzymatic screening when handling complex mixtures. To assess the extent to which the detection of minor bioactives could be improved, the 2D-LC×LC workflow was applied to a burdock root MeOH extract that is not known to contain biochanin A, spiked with this compound at different concentrations (0−10% w/w). All fractions collected were tested systematically in the AMPK assay at a standard concentration (Figure 4). The active fractions were dereplicated by UHPLC-UV-ELSD/MS to confirm the presence of biochanin A. As expected, no active fraction from the native burdock root extract was detected (Figure 4A). In contrast, after spiking biochanin A, at concentrations from 10% to 0.1% (w/w), bioactive fractions were identified in all cases after 2D-LC×LC fractionation (Figure 4B,C). However, bioactivity was no longer detected at 0.05% (w/w) biochanin A (data not shown). Therefore, a biochanin A concentration above 5% (w/w) in burdock root crude extract was required to stimulate AMPK activity when the crude mixture was screened. The automated fractionation workflow with an activity threshold of 0.1% (w/w) increased the sensitivity of detection by approximately 50 times and thus provided a valuable strategy for the identification of low-abundant bioactive NPs. Biochanin A is known to occur in Trifolium species.43,44 Thus, to verify if the approach could highlight the activity of naturally occurring biochanin A in crude plant extracts, Trifolium species were investigated. However, commercial crude extracts and fractions obtained from four Trifolium 2859
DOI: 10.1021/acs.jnatprod.6b00628 J. Nat. Prod. 2016, 79, 2856−2864
Journal of Natural Products
Article
AMPK activity at a standard concentration of 50 μg/mL (Figure 5). Two fractions (F64 and F65) were considered active in the screening conditions (>20% AMPK activation). UHPLC-UVELSD/MS analysis revealed that pure biochanin A was present in these two fractions (Figure S4, Supporting Information). Interestingly, a related compound with modest AMPK activity (12%) was identified as genistein in fraction F37. Using this automated 2D-LC×LC, the detection of minor active compounds from a crude extract was thus found to be efficient, even with extracts that initially appeared to test negative in rough 1D-LC fractionation. Additionally, good chromatographic performance was obtained by the 2D-LC×LC procedure, as seen in Figure 5. Indeed, more than 40 fractions were estimated to be partially pure after this separation. In this case, the ability to fractionate at this resolution was clearly correlated to the ability to detect active compounds. This fully automated method was developed to provide semipure fractions at a standardized concentration, based on a 2D-LC×LC-UV/ELSD fractionation system, given that the initial observation that directly screening crude extracts to identify AMPK activators proved inefficient. The proposed procedure is efficient, reproducible, and faster than previously described protocols.25,26 The 2D-LC×LC-UV/ELSD proved to be efficient for the collection of semipure fractions in a fully automated and standardized approach. Moreover, the direct weight estimation of the collected fractions using the ELSD detector avoids unnecessary and time-consuming steps and allows straightforward implementation in the HTS platform. The procedure developed in this study can be routinely applied for screening medium-polar natural products from complex matrices during enzymatic HTS. Further development will be required in the future to optimize the detection of highly polar bioactive compounds.
Figure 4. Activity on AMPK (%) of the 2D-LC×LC fractions obtained from the burdock root MeOH extract (A) and the same extract spiked with biochanin A at decreasing concentrations (B, 10%; C, 0.1% w/w). Orange bars: Fractions containing biochanin A.
■
species including T. pratense were proven to be negative against AMPK in this study. A MeOH extract of T. pratense flowers was specifically prepared for this study; it was also found to be inactive against AMPK, while the presence of biochanin A was proven using UHPLC-UV-MS, and its concentration was estimated to be 0.3% (w/w). According to our previous results at this concentration (equivalent to 0.1 μM in the screening conditions), the activity of biochanin A was indeed not expected to be detectable (Figure 3A). After pretreatment on C18-SPE, the biochanin A concentration in the methanolic subfraction of the red clover extract increased to 3% (equivalent to 1 μM), but it still proved to be inactive. To test if a simple 1D-LC chromatography separation could sufficiently deconvolute the mixture, this enriched SPE subfraction was then subjected to C18 flash chromatography fractionation, leading to 34 fractions. All fractions gave negative results against AMPK even though some were found to contain up to 20% biochanin A (equivalent to 7 μM) (Figure S3, Supporting Information). At this concentration, the bioactivity of biochanin A on AMPK should have been detectable (30% activation as shown in Figure 3A). It was possible that the presence of promiscuous inhibitors still masked the expected AMPK activation. Therefore, the flash chromatography fractions were still too complex and did not meet the enzymatic HTS requirements in this experiment. This result reinforced our previous findings and underlined the need for the production of more simple semipure compounds through additional orthogonal fractionation to detect their bioactivities during enzymatic screening. The red clover MeOH SPE subfraction was therefore submitted to the 2D-LC×LC procedure. In total, 142 fractions were collected and tested for
EXPERIMENTAL SECTION
General Experimental Procedures. 2D-LC×LC-ELSD fractionation was carried out on a Sepbox 2D-250 (Sepiatec, Berlin, Germany). This consists of a first-dimension LC coupled to online stainless steel SPE cartridges, which are eluted onto a series of seconddimension LC columns coupled to UV-DAD (Knauer, Berlin, Germany) and Sedex LT-ELSD 80 (Sedere, Alfortville France) detectors and an automatic fraction collector (Sepiatec). Flash chromatography was performed using a Puriflash 4100 (Interchim, Montluçon, France). A Freedom Evo robot (Tecan, Männedorf, Switzerland) coupled to a balance with an accuracy of 0.01 mg was used for automated fraction processing (aliquoting or weighting). Freeze-drying was performed with an HT-24 Genevac (StepBios, Muttenz, Switzerland). Fraction analysis was carried out on an Acquity UPLC (Waters, Milford, MA, USA), which consisted of a binary pump, a cooled autosampler, and a column oven, connected to three detectors, PDA, ELSD, and quadrupole-time-of-flight MS (Waters Xevo G2s QTof), equipped with an electrospray (ESI) source. Plant Material. All plant material was purchased from Dixa AG (St. Gallen, Switzerland), and specimens were stored at NIHS under the following references: powdered olive leaves (Olea europaea) (NI00043764; Dixa catalogue number 1115, batch 12001516), bilberry fruits (Vaccinium myrtillus) (NI00043685; Dixa catalogue number 1129, batch 136094), white willow bark (Salix alba) (NI00043770; Dixa catalogue number 1420, batch 130861), burdock roots (Arctium lappa) (NI00043610; Dixa catalogue number 227, batch 132870), and red clover flowers (Trifolium pratense) (NI00005445; Dixa catalogue number 1762, batch 144285). Plant Extracts. Commercial crude extracts and fractions from Trifolium spp. (T. alexandrium L., T. hybridum L., T. pratense L., and T. repens L.) were part of the screening collection of AnalytiCon 2860
DOI: 10.1021/acs.jnatprod.6b00628 J. Nat. Prod. 2016, 79, 2856−2864
Journal of Natural Products
Article
Figure 5. Illustration of the efficacy of the full 2D-LC×LC procedure for detecting active compounds from red clover (Trifolium pratense) flowers. The central 2D plot represents the MS feature intensity of the aligned UHPLC-MS chromatograms of each collected fraction submitted to the 2DLC×LC procedure. The x- and y-axes represent the fraction number according to the elution and retention times, respectively, of their components in the UHPLC-MS analysis. A UHPLC-MS chromatogram of the crude methanolic extract of red clover (T. pratense) flowers is given as a reference on the right side of the figure. The two bar charts on the bottom indicate the masses of the collected fractions and the percentage activation of AMPK, respectively, as measured by the bioassay. The active fractions, the structures of the corresponding compounds, and [M + H]+ are highlighted in orange (genistein) and red (biochanin A). Discovery (Potsdam, Germany). Extracts of olive leaves, bilberry fruits, white willow bark, burdock roots, and red clover aerial parts (T. pratense) were prepared in-house. The olive leaf powder (576 g) was first extracted three times with 2 L of CH2Cl2, and the filtered residue was further extracted three times with 2 L of MeOH upon stirring for 20−24 h, yielding 108 g (19% w/w recovery of extract/plant material) of CH2Cl2 dried extract and 138 g (24% w/w) of MeOH dry extract. The same procedure was applied to the remaining extracts: 425 g of bilberry powder was extracted with 1.6 L of CH2Cl2 and then 1.6 L of MeOH, yielding 37 g (9% w/w) of CH2Cl2 and 280 g (66% w/w) of MeOH dry extracts; 270 g of white willow bark powder was extracted with 1.5 L of CH2Cl2 and 1.5 L of MeOH, leading to 16 g (6% w/w) of CH2Cl2 and 8 g (3% w/w) of MeOH dry extracts; 380 g of burdock root powder was extracted with 2 L of CH2Cl2 and 2 L of MeOH, yielding 4 g (1% w/w) of CH2Cl2 and 123 g (32% w/w) of MeOH dry extracts; 280 g of red clover aerial part powder was extracted with 1.5 L of CH2Cl2 and 1.5 L of MeOH, yielding 39 g (14% w/w) of CH2Cl2 and 31 g (11% w/w) of MeOH dry extracts. Sample Pretreatment for Preparative-Scale Fractionation. The plant MeOH extracts prepared in-house were submitted to solidphase extraction on reversed-phase C18 material. To perform this step, 2 g of the dried methanolic extract was dissolved in MeOH and mixed with 6 g of C18 material (60 Å, 50 μm, Interchim) before being
evaporated and placed in an SPE cartridge over 2 g of pure C18 material. The SPE cartridges were connected to the Puriflash apparatus and successively eluted with 50 mL of water, 100 mL of MeOH, and 50 mL of i-PrOH at a flow rate of 10 mL/min. After evaporation of the solvents, the following subfractions were produced (% w/w recovery subfraction/dry MeOH extract): 1.12 g of water (56%), 0.38 g of MeOH (19%), and
