Anti-Allergic and Antioxidant Potential of Polyphenol-Enriched Fractions from Cyclopia subternata (Honeybush) Produced by a Scalable Process

: Anti-allergic activity was previously demonstrated for extracts of Cyclopia subternata Vogel plant material, containing substantial amounts of xanthones, release assay. In vivo testing will be required to determine whether the increased activity of fractions is worth the effort and expense of fractionation.


Introduction
Over the past few decades, research elevated polyphenols from mostly unwanted food constituents to natural products with commercial value, due to a plethora of beneficial bioactivities. These include anti-allergic activities [1], which are very relevant, given the increase in the prevalence of food allergies [2] and other allergies such as asthma, especially in children [3]. Allergies not only negatively affect the quality of life, but they can also be fatal [3].
Polyphenols exert anti-allergic activity via many mechanisms [1,4]. Interaction with allergic effector cells such as mast cells can inhibit the release of mediators. As antioxidants, they may also limit the extent of cellular injury caused by free radicals during

Static Adsorption and Desorption
The freeze-dried C. subternata extract was reconstituted with deionized water. All experiments were performed in triplicate in polypropylene 24-well deep-well microplates (Axygen Scientific, Union City, CA, USA) according to the general procedure described by Miller et al. [13]. Briefly, except for the control wells, the wet resin (250 mg equivalent dry weight based on measured moisture content) was weighed into each well of the microplate. An aliquot (5 mL) of the reconstituted extract solution was then added to each well. The wells were sealed with aluminum film and the plate was shaken at 450 rpm (Eppendorf Mixmate, Hamburg, Germany) for different durations depending on the experiment. On completion of the treatment, the supernatant in each well was sampled and analyzed by HPLC with diode-array detection (DAD).
The effect of initial sample concentration on the adsorption of target compounds onto the resin was determined at room temperature (23 • C) for both XAD 1180N and HP20. This involved shaking of the resin and aqueous C. subternata extract solutions varying in concentration (1, 2, 3, 5, 7.5 and 10 mg/mL) for 24 h at room temperature. Control wells contained the extract solutions without resin. Subsequently, the effect of contact time on the adsorption of target compounds onto XAD 1180N was determined at room temperature, using an extract solution of 3 mg/mL. Separate wells of the microplate represented individual contact times (20,40,60,90, 120 and 180 min). Control wells containing extract solutions without resin were prepared for each time point. The wells were sampled at the aforementioned time points from 0-180 min.
A range of EtOH-water concentrations (0, 5, 10, 15, 20, 30, 40, 50, 75, 100%; v/v) was investigated to determine an appropriate EtOH concentration for desorption of the target compounds from the loaded resin. A batch of loaded resin was prepared by mixing the aqueous C. subternata extract solution (3 mg/mL) and resin (5:1; v/v ratio) and shaking the mixture at room temperature for 90 min. After removal of the supernatant by vacuum filtration, portions of the loaded resin, equaling 250 mg of dry resin, were weighed into the wells of the microplate. Aliquots (5 mL) of the EtOH-water mixtures were added to the wells containing loaded resin and shaken for 4 h at room temperature. Control wells, without added resin, but with extract or the supernatant after resin loading were included for each EtOH concentration.
For each experiment, the adsorption capacity (Q e ; mg/g), adsorption ratio (AR, %) and/or desorption ratio (DR, %) were calculated as follows [14]: Separations 2022, 9, 278 4 of 16 where C 0 is the initial concentration of the target compound in solution (mg/mL); C e is the equilibrium concentration of the target compound in solution (mg/mL); V 0 is the volume of extract solution added to the well (mL); W is the dry mass of the resin (g); C d is the concentration of the target compound in the desorption solution at equilibrium (mg/mL); V d is the volume of the desorption solution (mL).

Small-Column Dynamic Adsorption and Desorption
A glass column wet-packed with XAD 1180N macroporous resin (Omnifit Labware, Diba Industries Inc., Danbury, CT, USA; ID 25 mm, height 700 mm, resin bed height 195 mm, bed volume (BV) 960 mL) was used for small-scale dynamic adsorption and desorption experiments. A Gilson Minipuls 3 peristaltic pump (Gilson Inc., Middleton, WI, USA) was fitted at the bottom of the column to control the eluent flow rate by suction. The bed height-to-diameter ratio was 7.8:1. Before sample loading, the column was conditioned by flushing it with six BVs of deionized water. All dynamic sorption experiments were performed in triplicate at room temperature (23 • C).
To determine the breakthrough volume, 10 BVs of aqueous C. subternata extract solution (10 mg/mL) were loaded onto the column at a flow rate of 2 BV/h. Each BV of eluate (n = 10) was collected and analyzed by HPLC-DAD. Thereafter, a dynamic desorption assay was performed. Seven BVs of aqueous C. subternata extract solution (10 mg/mL) were loaded onto the column, followed by step-wise gradient elution using 3 BVs each of 12, 20, 30 and 50% EtOH at a flow rate of 2 BV/h. The successive BVs of eluate were collected and analyzed by HPLC-DAD.

Large-Column Fractionation
Fractionation of the aqueous C. subternata extract solution was performed on a larger scale to produce four fractions enriched in different phenolic compounds. The protocol described in Section 2.4 was up-scaled by a factor of 22 while maintaining a bed heightto-diameter ratio of 7.8:1. Fractionation was carried out in a 70-mm internal diameter glass column. A peristaltic pump (Model 505U, Watson-Marlow Ltd., Falmouth, England) fitted to the bottom of the column was used to control the eluate flow rate (2 BV/h). The column was wet-packed with resin up to a bed height of 550 mm, equivalent to a BV of 2.1 L. Following column loading with 7 BVs of the extract solution (10 mg/mL; total mass loaded = 147 g), step-wise gradient elution was performed using 12, 20, 30 and 50% aqueous EtOH (3 BVs each). Each BV of the eluate was collected separately and analyzed by HPLC-DAD, whereafter they were pooled according to phenolic composition to produce four fractions: fraction 1 (12% EtOH BV 1 and 2), fraction 2 (12% EtOH BV 3 and 20% EtOH BV 1, 2 and 3), fraction 3 (30% EtOH BV 1, 2 and 3) and fraction 4 (50% EtOH BV 1, 2 and 3). The EtOH was removed by vacuum rotary evaporation at 40 • C (Rotavapor R-215, Buchi, Flawil, Switzerland). The aqueous residue was freeze-dried and analyzed using HPLC-DAD. The freeze-dried fractions were stored in screw cap glass jars sealed with Parafilm under desiccation.
Compound identity and peak purity were confirmed using high-resolution-mass spectrometry (HR-MS). The same HPLC method was used for separation as for quantification using a Waters Acquity UPLC and a Synapt G2 Q-ToF mass spectrometer (Waters, Milford, MA, USA). The electrospray ionization (ESI) source operated in negative ionization mode. A sodium formate solution was used for mass calibration and leucine enkephalin served as the lock spray solution. The eluent was split in a 1:1 ratio before entering the ionization chamber. MassLynx v.4.1 (Waters) software was used for data acquisition and analysis. A mass range of 150 to 1500 amu was scanned. Other MS parameters were: capillary voltage, −2.5 kV; sampling cone voltage, 15 V; source temperature, 120 • C; desolvation temperature, 275 • C; desolvation nitrogen (N 2 ) gas flow, 650 L/h; cone gas flow (N 2 ), 50 L/h. MS E was performed by ramping the trap collision energy from 20.0 to 60.0 V.

Superoxide Anion Radical (O •−
2 ) Scavenging Assay The O •− 2 scavenging activity of the extract and fractions (prepared in 10% DMSO) was determined using a method adapted from Chisté et al. [15]. The scavenging activity of the samples was determined spectrophotometrically in 96-well, flat-bottom microplates (Greiner Bio-One, Kremsmünster, Austria) using a BioTek SynergyHT microplate reader (BioTek Instruments, Winooski, VT, USA). The half-maximal inhibitory concentration (IC 50 ) of a sample was determined, using a 12-point concentration range. The analysis was done in triplicate and the assay was repeated three times to obtain triplicate IC 50 values.
Potassium phosphate buffer (100 mM; pH 7.4) was used to prepare the reagents (phenazine methosulphate [PMS], nicotinamide adenine dinucleotide [NADH] and nitro blue tetrazolium [NBT]). The reaction mixture containing equal volumes (60 µL) of 664 µM NADH, 1 mM NBT and the sample was incubated at 37 • C for 2 min. Control wells contained 10% DMSO instead of the sample. Then, 60 µL of 108 µM PMS was dispensed into each of the control and sample wells (final reaction volume = 240 µL). Each sample well had a corresponding blank well, containing 60 µL buffer instead of PMS. Absorbance at 560 nm was measured at 0 and 2 min for all wells. The O •− 2 scavenging activity of the samples was calculated as follows: where A S , A B and A C refer to the net absorbance (difference in absorbance measured at 2 and 0 min) of the sample, blank and control, respectively.

2,2-Diphenyl-1-picrylhydrazyl Radical (DPPH • ) Scavenging Assay
The DPPH • scavenging activity of the extract and fractions (prepared in 10% DMSO) was determined in 96-well microplate format [16] using a BioTek SynergyHT microplate reader (BioTek Instruments), and a 12-point concentration range of each sample was used for the determination of its IC 50 value. All sample concentrations were analyzed in triplicate and the assay was repeated three times to determine triplicate IC 50 values.

Oxygen Radical Absorbance Capacity Assay (ORAC)
The ORAC values of the extract and fractions (prepared in 10% DMSO) were determined in triplicate according to the method of Huang et al. [17]. Fluorescence was Separations 2022, 9, 278 6 of 16 measured using a BioTek SynergyHT microplate reader. The total antioxidant capacity (TAC ORAC ) of the extract and fractions was expressed in µmoles Trolox equivalents/g.

Xanthine Oxidase (XO) Inhibition Assay
Inhibition of XO by the extract and fractions (prepared in 10% DMSO) was determined in triplicate at three concentrations according to a method adapted from Leiro et al. [18]. Absorbance measurements of the samples were determined using 96-well, flat-bottom microplates (Greiner Bio-One) and a BioTek SynergyHT microplate reader. The reaction mixture consisted of equal volumes (60 µL) of 1 mM EDTA disodium salt dihydrate in 100 mM potassium phosphate buffer (pH 7.4), XO solution (0.025 U/mL) and sample or allopurinol (positive control; 1.36 µg/mL in 10% DMSO) or 10% DMSO (negative control). The microplate was incubated at 37 • C for 3 min, whereafter the reaction was initiated by adding 60 µL of 4 mM xanthine to each of the negative control, sample and allopurinol wells (final reaction volume = 240 µL). Each sample well and allopurinol well had a corresponding blank well containing 60 µL buffer instead of the xanthine solution. Absorbance measurements at 295 nm were recorded every 30 s for 8 min. The XO inhibition activity of the test samples and allopurinol was calculated as follows: where A S , A SB and A C refer to the net absorbance (difference in absorbance measured at 8 and 0 min) of the sample, sample blank and negative control, respectively.

β-Hexosaminidase Release Assay (Anti-Allergy Potential)
RBL-2H3 cells, purchased from JCRB (Osaka, Japan), were cultured in 96-well plates (Corning, NY, USA) at 5 × 10 4 cells/well using Minimal Essential Medium (MEM) supplemented with 100 U/mL of penicillin, 100 µg/mL of streptomycin and 10% fetal bovine serum. Anti-DNP-IgE antibody (Sigma) was added to the wells at 500 ng/mL. After culturing for 24 h, the medium was removed and the cells were washed twice with MEM. The samples dissolved in MEM were added at specific concentrations (62.5, 125 and 250 µg/mL); wortmannin (positive control) was added at 100 nM. MEM was added to the blank wells. After incubation for 30 min, 2,4-dinitrophenyl-human serum albumin was added to all wells at 5 ng/mL. After 1 h of incubation, the supernatants were collected from the wells and the lysis buffer (1% Triton X-100 in 50 mM Tris and 20 mM EDTA; pH 7.5) was added to each well. The plates were sonicated, and the cell lysates were collected from the wells. The same procedure was carried out for the wells without samples and cells to serve as cell lysate blank. The β-hexosaminidase activity of each supernatant and cell lysate was measured as follows: The same volume of substrate solution (1.3 mg/mL of 4-nitrophenyl-N-acetyl-β-D-glucosaminide in 0.1 M citrate buffer) was added to each supernatant and lysate, and stop solution (0.1 M Na 2 CO 3 + 0.1 M NaHCO 3 ) was added after incubating for 90 min. The absorbance of each sample was measured at 405 nm. The release rate was calculated as follows: where A S , A SB , A L and A BL refer to the absorbance of the sample, sample blank, cell lysate and cell lysate blank, respectively.

Data Analysis
Univariate analysis of variance (ANOVA), using the General Linear Models procedure (SAS Version 9.4; SAS Institute Inc., Cary, NC, USA), was performed to compare treatments for the data obtained from the static sorption tests. Fisher's least significant difference Separations 2022, 9, 278 7 of 16 was determined at the 5% level to compare treatment means where significant differences (p < 0.05) were found.
The half-maximal inhibitory concentration (IC 50 ) for O •− 2 and DPPH • scavenging activity was determined by non-linear regression analysis of the concentration-effect data (log concentration vs. % scavenging) using GraphPad Prism (Version 8.2.1; GraphPad Software, San Diego, CA, USA). The four-parameter variable slope regression model was used, with the bottom value for % scavenging constrained to a constant value of 0. Triplicate IC 50 , TAC ORAC and XO inhibition values were also subjected to ANOVA to compare the values for the extract and fractions.
The statistical analysis of the β-hexosaminidase release assay data was performed by one-way ANOVA followed by the Dunnett test with p < 0.05 considered significant.

Optimization of MARC Parameters
Static sorption assays were performed to determine the optimal sample concentration, adsorption time and desorption solvent for the separation of the target compounds.
The sample concentration affected the adsorption capacities (Table S1) and ARs (Table S2) of the phenolic compounds to the resins. At the highest sample concentration, XAD 1180N had a significantly higher (p < 0.05) adsorption capacity for IDG, IMG, HPDG, isomangiferin, vicenin-2 and p-coumaric acid than HP20. However, at this concentration, HP20 had a significantly higher (p < 0.05) adsorption capacity for PDG, eriocitrin, hesperidin, mangiferin and scolymoside than XAD 1180N. Even at the highest sample concentration, the flavanones (eriocitrin and hesperidin), dihydrochalcones (PDG and HPDG) and flavones (scolymoside and vicenin-2) had ARs of 100% for both resins (data not shown). Mangiferin and isomangiferin had ARs > 95%, irrespective of loading concentration and resin. The AR for IDG was the limiting factor as it decreased from 88.5% to 72.2% and 87.6% to 62.5% for XAD 1180N and HP20, respectively, when the sample concentration was increased from 1 to 10 mg/mL. For adequate adsorption of all the target compounds onto the macroporous resin, 3 mg/mL was selected. Furthermore, the XAD 1180N resin was selected for further experiments due to its higher ARs compared with that of HP20. Many of the compounds reached maximum adsorption at 20 min, whereas some increased their adsorption marginally up to 60 min ( Figure S1). To ensure maximum adsorption of all compounds, 90 min was selected.
EtOH concentration affected the desorption of the compounds, resulting in large variations in the DR of a compound ( Figure 1). IDG, p-coumaric acid and IMG already started to desorb in pure water (0% EtOH) with DRs equaling 14.9, 6.9 and 4.9%, respectively. On the other hand, mangiferin and isomangiferin had extremely low DR values (≤1%) and none of the dihydrochalcones, flavanones and flavones was detected in the supernatant. Except for hesperidin and p-coumaric acid, all compounds reached 50% desorption at 20-30% EtOH. For these two compounds, an EtOH concentration > 30% was necessary to achieve 50% desorption. Hesperidin was completely desorbed at 75% EtOH.
Small-column dynamic experiments were conducted at a flow rate of 2 BV/h to determine the final protocol before up-scaling of the separation. Initial experiments using 15 BVs of aqueous C. subternata extract solution (3 mg/mL) showed breakthrough of IDG after loading 12 BVs of the extract solution (data not shown). The sample concentration was, therefore, increased to 10 mg/mL to reduce the breakthrough volume, as well as the time required to load the column. The breakthrough point was set at 10% to ensure minimal loss of IDG.
Breakthrough curves for IDG and IMG ( Figure 2) show a gradual increase in the leakage of IDG and IMG from the column as the loading volume increased to seven BVs. Further loading of the column resulted in a rapid increase in the IDG concentration in the eluent, surpassing the 10% breakthrough set point; therefore, 7 BVs (672 mL; 6.72 g extract) were determined as the maximum sample loading volume of the column. Small-column dynamic experiments were conducted at a flow rate of 2 BV/h to determine the final protocol before up-scaling of the separation. Initial experiments using 15 BVs of aqueous C. subternata extract solution (3 mg/mL) showed breakthrough of IDG after loading 12 BVs of the extract solution (data not shown). The sample concentration was, therefore, increased to 10 mg/mL to reduce the breakthrough volume, as well as the time required to load the column. The breakthrough point was set at 10% to ensure minimal loss of IDG.
Breakthrough curves for IDG and IMG ( Figure 2) show a gradual increase in the leakage of IDG and IMG from the column as the loading volume increased to seven BVs. Further loading of the column resulted in a rapid increase in the IDG concentration in the eluent, surpassing the 10% breakthrough set point; therefore, 7 BVs (672 mL; 6.72 g extract) were determined as the maximum sample loading volume of the column.
A step-wise solvent gradient consisting of 12, 20, 30 and 50% aqueous EtOH (3 BVs of each) was employed for the dynamic desorption of the target compounds from the loaded resin column at a flow rate of 2 BV/h. Complete desorption of IDG was achieved after 2 BVs of 12% EtOH (Figure 3). Most of the IMG (64.1%) also eluted after 2 BVs of 12% EtOH, with the remaining IMG eluting in the third BV of 12% EtOH. Desorption of mangiferin, isomangiferin and vicenin-2 occurred gradually with successive BVs of 12% EtOH and the addition of 1 BV of 20% EtOH resulted in desorption maxima for mangiferin, isomangiferin and vicenin-2. Maximum desorption of HPDG occurred at 2 BVs of 20% EtOH, along with the co-elution of PDG, scolymoside and eriocitrin. A sharp increase in the desorption of PDG, scolymoside, eriocitrin and hesperidin was observed with successive BVs of 30% EtOH. Hesperidin, accompanied by the co-elution of small quantities of PDG, p-coumaric acid and scolymoside, was completely desorbed after 3 BVs of 50% EtOH.  resin column at a flow rate of 2 BV/h. Complete desorption of IDG was achieved after 2 BVs of 12% EtOH (Figure 3). Most of the IMG (64.1%) also eluted after 2 BVs of 12% EtOH, with the remaining IMG eluting in the third BV of 12% EtOH. Desorption of mangiferin, isomangiferin and vicenin-2 occurred gradually with successive BVs of 12% EtOH and the addition of 1 BV of 20% EtOH resulted in desorption maxima for mangiferin, isomangiferin and vicenin-2. Maximum desorption of HPDG occurred at 2 BVs of 20% EtOH, along with the co-elution of PDG, scolymoside and eriocitrin. A sharp increase in the desorption of PDG, scolymoside, eriocitrin and hesperidin was observed with successive BVs of 30% EtOH. Hesperidin, accompanied by the co-elution of small quantities of PDG, p-coumaric acid and scolymoside, was completely desorbed after 3 BVs of 50% EtOH. PDG, p-coumaric acid and scolymoside, was completely desorbed after 3 BVs of 50% EtOH.

Fractionation of C. subternata Extract Using MARC
The small-scale separation was scaled-up 22-fold by keeping the bed height-to-column diameter constant. To obtain four fractions enriched in different phenolic compounds, fractions were pooled based on the data from the dynamic desorption curves (Figure 3). Content values of the compounds for the extract and fractions are summarized in Table 1. The HPLC-DAD chromatograms are depicted in Supplementary information ( Figure S2). LC-HR-MS data are provided in Table S3. Of the 147 g of extract loaded onto the resin column, a total of 97.8 g was recovered. The recovery of the individual phenolic compounds varied from 67.7% (hesperidin) to 100% (isomangiferin) (average 87.7%).  Fraction 1 was enriched 7.4-and 3.6-fold with IDG and IMG, respectively, compared with the extract (Table 1). Small amounts of HPDG, mangiferin, isomangiferin and vicenin-2 were also present along with trace quantities of the minor benzophenone, MMG, two tetrahydroxyxanthone-di-O,C-hexose isomers, two phenolic acid glycosides and an eriodictyol-O-(hexose-O-deoxyhexose) as detected by MS (Table S3). The main compounds in fraction 2 were mangiferin and isomangiferin. Their content and that of HPDG and vicenin-2 were increased 4.5-, 4.6-, 3.6-and 4.5-fold, respectively. All the other major compounds, except hesperidin, were also present in fraction 2, but in smaller quantities than the xanthones. MS analysis of fraction 2 showed trace quantities of MMG, a tetrahydroxyxanthone-di-O,C-hexose, a phenolic acid glycoside and a pentahydroxyxanthone-C-hexose. Eriocitrin, PDG and scolymoside were increased by two-fold in fraction 3. In addition to these compounds, fraction 3 contained all the major compounds and traces of two pentahydroxyxanthone-C-hexose isomers, an eriodictyol-O-hexose and isorhoifolin. Fraction 4 consisted predominantly of hesperidin (5.3-fold enrichment) together with small amounts of eriocitrin, PDG, mangiferin, p-coumaric acid and scolymoside and traces of a naringenin glycoside and isorhoifolin.

Bioactivity of C. subternata Extract and Fractions
The radical scavenging activity of the extract and fractions was evaluated using three assays, i.e., the O •− 2 scavenging, DPPH • scavenging and ORAC assays ( Table 2). The samples scavenged O •− 2 in the descending order of activity: fraction 3 ≈ fraction 2 ≈ extract > fraction 4 > fraction 1. Considering scavenging of DPPH • , fractions 2, 3 and 4 were more effective (p < 0.05) than the extract with the IC 50 value of fraction 2 (most effective) ca. 1.7-fold lower than that of the extract. In the ORAC assay, fractions 1 and 2 were the most effective as antioxidants with TAC ORAC values significantly higher (p < 0.05) than that of the extract.
The extract and fractions showed relatively poor XO inhibitory activity (Figure 4) even at 400 µg/mL with their inhibitory activity varying in descending order: fraction 4 (33.8%) > fraction 2 (25.2%) ≈ fraction 3 (24.9%) > fraction 1 (15.7%) ≈ extract (15.0%). All samples were less potent than the positive control allopurinol, which inhibited 52.5% of the XO activity at a concentration of 1.36 µg/mL. Only the extract (p < 0.05) and fraction 1 (p < 0.01) were effective to inhibit mast cell degranulation as indicated by the decrease in the release of β-hexosaminidase in the RBL-2H3 cell model ( Figure 5). Fraction 1 dose-dependently inhibited β-hexosaminidase release, but the extract was only significantly effective (p < 0.05) at 250 µg/mL. Data are given as mean ± standard deviation (n = 3). Different letters above bars indicate significant differences (p < 0.05) at 400 µg/mL. * Value not available due to background absorbance of the fraction at this concentration.
Only the extract (p < 0.05) and fraction 1 (p < 0.01) were effective to inhibit mast cell degranulation as indicated by the decrease in the release of β-hexosaminidase in the RBL-2H3 cell model ( Figure 5). Fraction 1 dose-dependently inhibited β-hexosaminidase release, but the extract was only significantly effective (p < 0.05) at 250 µg/mL.

Production of Polyphenol-Enriched Fractions of C. subternata
Several factors determine the adsorption capacity of a system so that it cannot be predicted. These factors are the characteristics of both the adsorbent (e.g., surface area and particle size) and the adsorbate (e.g., structure, water solubility, polarity, size, and molec-

Production of Polyphenol-Enriched Fractions of C. subternata
Several factors determine the adsorption capacity of a system so that it cannot be predicted. These factors are the characteristics of both the adsorbent (e.g., surface area and particle size) and the adsorbate (e.g., structure, water solubility, polarity, size, and molecular weight), as well as that of the solvent (e.g., polarity, pH, and temperature) [11]. Therefore, static and dynamic adsorption experiments were used to determine the most suitable experimental conditions to produce polyphenol-enriched fractions from a hot water extract of C. subternata. The static adsorption experiments indicated that the XAD 1180N resin was more suitable than the HP20 resin due to its higher adsorption ratios for some of the target compounds. A sample concentration of 3 mg/mL and a contact time of 90 min were required to obtain good adsorption of these phenolic compounds onto the resin. The desorption profile of the compounds indicated that a step-wise gradient from 12-50% EtOH should be suitable.
Although the static experiments provided information on the effectiveness of the adsorbent, the results are not directly transferable to a column system [11] and subsequently, dynamic adsorption and desorption studies were conducted using a fixed-bed column for scalability. In dynamic adsorption, column loading is performed until the concentration of the target compound in the eluate exceeds a predefined threshold level (%) referred to as the breakthrough point [19]. This breakthrough percentage value is defined by the user, and reported values ranged from 3% [13] to 10% [20], depending on the purpose of the separation. In the present study, the maximum sample loading (7 BVs of 10 mg/mL solution; 6.72 g extract) was determined based on a 10% breakthrough value for IDG, the compound with the least affinity for the resin.
The step-wise desorption gradient, consisting of 3 BV each of 12%, 20%, 30% and 50% EtOH, was found suitable for crude fractionation of the extract into four fractions enriched in different phenolic compounds. Linear upscaling of the fractionation column (22 times) was very successful as also previously reported by Miller et al. [13] for a C. genistoides extract. Linear scalability plays a key role in the development of industrial-scale processes [21]. Besides the column height-to-diameter ratio, other factors such as pressure drop, the compressibility of the resin and wall friction also need to be considered in such a case [22]. The phenolic compounds in the four fractions were identified by comparing their UV-vis spectral data, accurate mass and MS fragmentation patterns to data previously reported in the literature [7,[23][24][25]. The main phenolic compounds of the respective fractions were IDG and IMG (fraction 1), mangiferin, isomangiferin, HPDG and vicenin-2 (fraction 2), PDG, eriocitrin and scolymoside (fraction 3) and hesperidin and p-coumaric acid (fraction 4).

Bioactivity of C. subternata Extract and Fractions
The differences in bioactivity between the extract and fractions provided some insight into the compounds responsible for the bioactivity. Fraction 4 enriched in hesperidin was less effective than the extract in scavenging O •− 2 . Orallo et al. [26] reported that  [27,28]. Their weak activity could be attributed to the replacement of the C-7 OH-group with a glucopyranosyl moiety, which sterically prevents the remaining free OH-groups from accessing the radical site [29]. The high activity of fraction 2 is attributed to its high mangiferin and isomangiferin content. Both compounds are wellknown radical scavengers [28,30]. DPPH • scavenging activity of fraction 3 is attributed to its high PDG content. The free hydroxyl groups at the 2 and 6 positions of the A ring of the dihydrochalcone are essential for effective radical scavenging activity [31,32]. The major phenolic compound in fraction 4, namely hesperidin, is also known for good DPPH • scavenging activity [33].
The C. subternata extract and fractions showed poor XO inhibitory activity (≤40% at 500 µg/mL). According to previous studies, a planar flavonoid structure created by the double bond between C-2 and C-3 is required for strong XO inhibitory action [34,35]. A large complement of C. subternata flavonoids, however, lacks this double bond, explaining the relatively low inhibitory activity. Furthermore, free hydroxyl groups at C-5 and C-7 are also deemed essential for XO inhibition [36,37]. Whilst vicenin-2, a flavone, fulfills all these criteria, only small amounts are present in the extract and fractions.
Only the extract and fraction 1 inhibited IgE-mediated degranulation of RBL-2H3 cells to some extent, but fractionation did not improve this activity despite enrichment in specific polyphenols. This is contrary to what would be expected as many polyphenols, especially planar flavonoids such as flavones and flavonols, have shown activity in this model [12]. Glycosylation had a severe negative effect on the activity of flavones and flavonols in the study by Mastuda et al. [12]. However, Suntivich et al. [38] showed ca. 50% inhibition of mast cell degranulation for the glycosylated flavone luteoloside at 25 µM. This is relevant for the planar flavone glycosides, vicenin-2 and scolymoside, in fractions 2 and 3, respectively. Flavanones were shown to have poor activity [12] as observed for fraction 4, enriched in hesperidin. It is thus likely that IDG and IMG (both glycosides) will also not be effective inhibitors of IgE-mediated degranulation of RBL-2H3 cells. Further investigation of fraction 1 is needed to identify active compounds.
The relative activity of the extract and fractions in the different assays demonstrated that radical scavenging, irrespective of the radical, is not a good proxy for inhibition of mast cell degranulation, despite the similar structural features of flavonoids that determine both types of activity [12,43].

Conclusions
A MARC protocol for fractionation of a C. subternata hot water extract was successfully developed and implemented to obtain four fractions with differing phenolic compositions. Despite enrichment of the fractions in specific polyphenols and the increased DPPH • , ORAC and XO inhibitory activity of some fractions, the inhibition of mast cell degranulation by the extract was only slightly enhanced by fractionation. The antioxidant and XO inhibitory assays were, therefore, not good proxies for screening the anti-allergy potential of C. subternata extract and fractions, despite the similar structural features of flavonoids that determine both types of activity. Fraction 1 exhibited anti-allergy potential and needs to be further investigated to determine the compounds responsible for the activity.

Conflicts of Interest:
The authors declare no conflict of interest.