Acute metabolic actions of the major polyphenols in chamomile: an in vitro mechanistic study on their potential to attenuate postprandial hyperglycaemia

Transient hyperglycaemia is a risk factor for type 2 diabetes and endothelial dysfunction, especially in subjects with impaired glucose tolerance. Nutritional interventions and strategies for controlling postprandial overshoot of blood sugars are considered key in preventing progress to the disease state. We have identified apigenin-7-O-glucoside, apigenin, and (Z) and (E)−2-hydroxy-4-methoxycinnamic acid glucosides as the active (poly)phenols in Chamomile (Matricaria recutita) able to modulate carbohydrate digestion and absorption in vitro as assessed by inhibition of α-amylase and maltase activities. The latter two compounds previously mistakenly identified as ferulic acid hexosides were purified and characterised and studied for their contribution to the overall bioactivity of chamomile. Molecular docking studies revealed that apigenin and cinnamic acids present totally different poses in the active site of human α-amylase. In differentiated Caco-2/TC7 cell monolayers, apigenin-7-O-glucoside and apigenin strongly inhibited D-[U-14C]-glucose and D-[U-14C]-sucrose transport, and less effectively D-[U-14C]-fructose transport. Inhibition of D-[U-14C]-glucose transport by apigenin was stronger under Na+-depleted conditions, suggesting interaction with the GLUT2 transporter. Competitive binding studies with molecular probes indicate apigenin interacts primarily at the exofacial-binding site of GLUT2. Taken together, the individual components of Chamomile are promising agents for regulating carbohydrate digestion and sugar absorption at the site of the gastrointestinal tract.


Results
Analytical and semi-preparative HPLC-DAD analysis of chamomile. Initial detection of target (poly)phenols in ChE was based on our previous report 3 where we identified by mass spectrometry apigenin, apigenin-7-O-glucoside and two compounds with corresponding [M-H]ion masses at m/z 355 and 711 (compounds 1 and 2) as the most abundant constituents. In this study, apigenin-7-O-glucoside (compound 3) and apigenin (compound 4) ( Fig. 1A; Table 1) were detected based on the comparison of their retention time with those of authentic standards. The UV absorption spectra of compounds 1 and 2 ( Fig. 1B) were characteristic of methoxycinnamic acid derivatives. The compounds were separated, isolated and concentrated by semi-preparative column chromatography (supplementary S1) yielding enough material to perform in vitro experiments. The purity of the collected fractions, pooled after freeze-drying, was >80% for compound 1 and >99% for compound 2 as assessed by liquid chromatography-mass spectrometry (LC-MS) based on the total ion count chromatograms (TIC).
Chemical characterisation of the hydroxycinnamic acid derivatives. The identities of the hydroxycinnamic acid derivatives were assigned based on HPLC-DAD/QTOFMS, 1 H and 13 C NMR, NOESY and chemical synthesis. The full-scan TOF spectra of both compounds showed prominent ions at m/z 355.1035 and   Table 1. Identification and quantification of targeted (poly)phenols in ChE. *The quantification was conducted with the isolated standard with a purity ≥80%. An isomerisation reaction took place where (Z)-MCAG (compound 1) changed to the most thermodynamically stable form (E)-MCAG (compound 2) which accounts for the impurity detected. The identification of compound 1 and 2 is described in section 3.2.
SCIentIfIC REPORTS | (2018) 8 (Table 2). This assignment was further confirmed by the appearance of the [M+Cl]ion at m/z 391.0801 in the spectrum of compound 2. The (neutral) elemental formula of the two hydroxycinnamic acid derivatives was established as C 16 11 based on the predominant ion at m/z 711 for compound 2. Covalently bound 'dimeric ferulic acids' are well characterised compounds, which act like polysaccharide cross-linking agents in cell walls 12,13 , and therefore their (di)glycoside structures can also be theoretically expected in plant     11,14,15 Acid hydrolysis of compound 2 did not result in the release of ferulic acid as would be expected if the compounds were ferulic acid hexosides ( Fig. 2A-C). However, a compound with m/z (-) 193 (supplementary S2) with UV spectra characteristic of hydroxycinnamates was detected at 13.4 min (Fig. 2C). We also observed a peak at 15.3 min with a UV spectra distinctive of coumarins 16 , later confirmed as 7-methoxycoumarin by comparison with the authentic standard (Fig. 2C). The presence of 7-methoxycoumarin was indicative of the release of a O-hydroxycinnamate which, under the acidic conditions 16 employed for hydrolysis, underwent intramolecular cyclization.
The 1 H-NMR data displayed spectroscopic signals characteristic of a free carboxylic group, which was identified from detection of the signal at δ 12.00. The presence of a (Z) and (E)-alkene group in the structure of compound 1 and 2 were identified from the signals at δ 7.19 and 5.78 (J = 12.7 Hz), and δ H 7.82 and 6.39 (J = 16.2 Hz), respectively. The spectrum also indicated the presence of three aromatic protons in the ABD substitution pattern characteristic of the coupling constants of J = 8.5-8.6 (ortho) and J = 2.4-2.5 (meta) 17 . Acid hydrolysis of compounds 1 and 2 resulted in the release of D-glucose (Fig. 2D). The 1 H chemical shift in the range 4.85 (Z-isomer) and 4.98 (E-isomer) with coupling constants of J = 7.5 Hz from the anomeric protons indicate a β-glycosidic linkage and the location of the D-glucose moiety was assigned at the aromatic ring by the observed NOE enhancements (Fig. 3A). The NOE enhancements gave an indication of the position of the other functional groups around the ring. The 13 C NMR data showed 16 carbon resonances for the phenolic moiety in isolated compound 2, suggesting that the compounds are (Z) and (E)−2-β-D-glucopyranosyloxy-4-methoxycinnamic acid (Z and-E-MCAG). Further confirmation of the identity of the cinnamate moiety was obtained by chemical synthesis, and comparison with the released aglycone after acid hydrolysis (Fig. 2C).
Inhibition of human α-amylase and rat α-glucosidase activities by ChE (poly)phenols. In vitro assays. As we previously found that ChE could inhibit human α-amylase and rat maltase activity 3 and further confirmed in this study (Fig. 4A), we tested the inhibitory activity of its individual component (poly)phenols. Apigenin and its precursor apigenin 7-O-glucoside exhibited the most potent inhibition against α-amylase as evidenced by the dose-dependent response (Fig. 4B). (Z)-MCAG and (E)-MCAG were also active, although they showed only modest inhibition with maximum values ~25 and ~20% respectively (p < 0.05) (Fig. 4B). When the four (poly)phenols were combined at concentrations equivalent to those quantified in the ChE preparation,  apigenin 7-O-glucoside showed the highest contribution (p < 0.05) (Fig. 4C). Moreover, their combination resulted in a ~20% increase of the inhibitory activity when compared to ChE.
While ChE weakly inhibited rat intestinal maltase, no inhibitory effect was observed on rat sucrase and isomaltase (Fig. 5A); even at the highest tested concentration only a modest ~33% (p < 0.05) inhibition was evident. When individual constituents were tested at concentrations equivalent to those present in ChE and higher, rat maltase was similarly only weakly inhibited by the individual (poly)phenols and apigenin 7-O-glucoside was the principal contributor (p < 0.05) (Fig. 5B,C). Nevertheless, a dose response was obtained for inhibition with ChE  and the four individual compounds. Although apigenin was found to be the most potent inhibitor of the four constituent (poly)phenols, its limited solubility prohibited testing at concentrations >50 µM (Fig. 5B). None of the individual (poly)phenols or their combination achieved 50% inhibition. (E)-MCAG was resistant to digestive hydrolytic enzymes, indicating that 2-(E)-MCA would not be produced in significant amounts in the small intestine, and therefore 2-(E)-MCA was not assessed for inhibition of carbohydrate-digesting enzymes (supplementary S3).
Molecular docking studies. Molecular docking was performed to investigate the structural basis of the observed inhibition on human α-amylase by individual (poly)phenols. Based on the strong inhibition observed for apigenin, three important features were identified: (1) the formation of H-bonds between the OH group of the B ring and the catalytic residue Asp 197 , (b) the formation π-π stacking interactions with Trp 59 and Tyr 62 and (c) the stacking towards the binding subsites −1 and −2. As shown in supplementary S4, these observations were extended to other strong (poly)phenol inhibitors of human α-amylase (arbitrarily defined as molecules with IC 50 <100 µM) including EGCG, quercetin and luteolin 18 . Apigenin 7-O-glucoside is able to form π-π stacking interactions with Trp 59 and Tyr 62 , however it does not form the H-bond with the catalytic residue Asp 197 resulting in a ~10 fold lower inhibition in agreement with the in vitro results (Fig. 4J).
(Z) and (E)-MCAG bind in a completely different pose (Fig. 4F,G) and are able to form H-bonds with the catalytic residues, but, unlike apigenin, the H-bonds are formed mainly with the OH groups of the glucose moiety and would be possibly weaker judging by the distance of H-bonds formed. In the case of (Z)-MCAG, an H-bond is formed between the 6-OH group of the glucose unit with Asp 197 while (E)-MCAG presents a more complex interaction. In (E)-MCAG, 2 covalent bonds are formed between the carbon at position 2 and Asp 197 (indicated as carbon-hydrogen bond in Fig. 4). Furthermore, H-bond interactions occur between the 6-OH group and Glu 233 and Asp 300 while an additional H-bond is formed between the Glu 233 and the OH-present in the carboxylic group of the phenolic skeleton. These additional interactions would explain the slightly higher inhibition observed by the (E)-isomer (Fig. 4C) in vitro, and could perhaps account for its stability in the binding pocket given the lack of π-π stacking interactions. In order to elucidate which (poly)phenols in ChE were responsible for the effect, each was tested separately on D-[U-14 C]-monosaccharide transport across Caco-2/TC7 cell monolayers. Among individual (poly)phenols, apigenin 7-O-glucoside was a very effective inhibitor of D-[U-14 C]-glucose and D-[U-14 C]-sucrose transport, and IC 50 values were obtained ( Fig. 6B-1). Although an IC 50 value could not be obtained for apigenin at the highest soluble concentration tested (Fig. 6B-2), it can be deduced that it is a more potent inhibitor based on the substantial 3-fold higher effect when compared with apigenin 7-O-glucoside at equivalent concentrations. Both  (Fig. 6B-4).
To dissect the mechanism of inhibition of glucose transport, we conducted D-[U-14 C]-glucose transport experiments under Na + -free conditions where only GLUT2, and not SGLT1, is active. An apparent increase (~50%) in the inhibitory activity was observed for ChE (p < 0.05), apigenin 7-O-glucoside and apigenin as evidenced by the decrease in the inhibition constants, suggesting that the main inhibitory effect is on GLUT2 (Fig. 6C-1 and 2). When tested in combination under Na + -free conditions, apigenin 7-O-glucoside and apigenin (Fig. 6C-3) were responsible for ~80% of the inhibitory activity observed by the whole ChE (p < 0.05) (Fig. 6C-4).
Since the main inhibition we found was on GLUT2, we investigated the binding affinities of apigenin and apigenin 7-O-glucoside. Immunostaining of GLUT2 showed that GLUT2 protein was present on both apical and basolateral sides of differentiated Caco-2/TC7 cells (Fig. 7A). This was also confirmed by functional assays where the rate of D-[U-14 C]-glucose transport under Na + -free conditions was the same from apical to basolateral compartments and vice versa (Fig. 7B). Inhibition constants for the exofacial glucose transporter inhibitor 4,6-O-ethylidene glucose and the endofacial glucose transporter inhibitor cytochalasin B on basolateral to apical D-[U-14 C]-glucose transport were compared to those for apigenin and apigenin 7-O-glucoside (Fig. 7C-E). The results suggest that the two (poly)phenols bind at the exofacial side of the protein, since a 3-fold lower inhibition was observed when the D-[U-14 C]-glucose transport was assessed from basolateral to apical side, which was also conditions. (C-3) Chemical structures of apigenin and A7G and their respective IC 50 values under Na + -free conditions. (D) Combined inhibition of apigenin and A7G under Na + -free conditions; apigenin, 12 µM; A7G, 148 µM. Each data point represents the mean ± SEM of three independent experiments with six technical replicates. Results are expressed as % compared to the transported sugars in control incubations (no inhibitor; normalised to 100%). When not visible, the error bars are smaller than the data point. In Fig (A2), (B1), (C1 and C2) statistical difference is indicated against the incremental concentrations of each inhibitor. *p < 0.05, **p < 0.05.

Discussion
Repeated high postprandial glucose spikes and fructose related metabolic burden have been proposed as independent risk factors for developing insulin resistance and progression to type 2 diabetes [21][22][23][24] . Delaying the absorption of sugars in the intestine is considered a clinically relevant approach for reducing type 2 diabetes risk 22,23 . The present study was designed to identify the (poly)phenols involved in the inhibition of carbohydrate digestion and absorption previously demonstrated by ChE 3 and further explore their mechanism of action.
In the work presented here we have provided conclusive evidence of the presence of (Z) and (E)−2-β-D-glucopyranosyloxy-4-methoxycinnamic acids in ChE, in agreement with the tentative identification of Weber et al. 17 . These compounds were previously misidentified as ferulic acid hexosides and 'dimer' forms by several authors 11,14,15,25 . When they were individually tested, they showed modest inhibitory activity on human α-amylase; however, when assessed in combination with apigenin and apigenin 7-O-glucoside, it was evident that they significantly contribute to the overall bioactivity of ChE.
In vitro and in silico studies revealed that apigenin and apigenin 7-O-glucoside are primarily responsible for the inhibition of human α-amylase by ChE, and this arises in part from their ability to occupy the binding pocket and stack between subsite "−1 and −2" through π-π interactions. The side chain residue of Asp 197 is believed to act as nucleophile during the catalytic reaction of α-amylase and its substitution dramatically decreased (10 6 -fold) its catalytic activity 26 . Unlike apigenin 7-O-glucoside, apigenin showed additional interaction with Asp 197 through an H-bond and this was reflected in a ~10-fold increase in its inhibitory activity. Thus, H-bonds between OH groups of the B ring and catalytic residues, and π-π stacking interactions appear to be requirements for an optimal inhibition of human α-amylase by (poly)phenols acting via a competitive mechanism. Indeed, in a previous study, Lo Piparo et al. 27 observed similar features; (poly)phenols showing IC 50 <100 µM were anchored in the active site through π-π interactions with Trp 59 , and formed H-bonds with residues Asp 197 and Glu 233 through OH groups in the A and B rings. This mechanism is further supported by X-ray structural analyses of the (poly) phenols myricetin and montbretin A with human pancreatic α-amylase 28,29 .
The final step of starch digestion is catalysed by α-glucosidases, maltase-glucoamylase and sucrase-isomaltase. Their combined catalytic activities hydrolyse linear and branched malto-oligosaccharides after the action of pancreatic α-amylase 30 while sucrase-isomaltase additionally hydrolyses sucrose, producing glucose and fructose. Inhibition of α-glucosidases is complementary in strategies aiming at blunting blood sugar spikes stemming from consumption of readily absorbable sugars which increase the prospect of type 2 diabetes. 31,32 Our results indicate only a modest inhibition of maltase-glucoamylase by the four (poly)phenols, even when tested in combination, consistent with the inhibition observed for ChE. In contrast, Kato et al. 5 found a stronger inhibition of ChE against maltase-glucoamylase (IC 50 = 2.6 mg/mL) and sucrase-isomaltase (IC 50 = 0.9 mg/mL) using brush border membranes prepared from rat small intestine. The authors suggested that this effect was driven by the presence of esculetin (maltase IC 50 = 534 μM; sucrase IC 50 = 72 μM) and quercetin aglycone (maltase IC 50 = 216 μM; sucrase IC 50 = 71 μM). However, the phytochemical composition and content of the plant material used in the study was not reported. Data from other studies on chamomile composition indicate that these compounds, if present at all, are unlikely to be effective since they are several orders of magnitude below the concentrations necessary for inhibition 11,33,34 .
At low concentrations, the uptake of glucose and fructose relies on SGLT1 and the facilitated transporter GLUT5, respectively. However, after consumption of a meal high in sugars, glucose concentration may reach 50-300 mM in the intestinal lumen prompting the facilitated GLUT2 transporter to translocate to the apical membrane. GLUT2 accounts for ~75 and ~60% of total glucose and fructose uptake respectively. 35,36 Our data show that apigenin and apigenin 7-O-glucoside are potent and selective inhibitors of GLUT2 as these compounds account for up to ~80% of transport inhibition under Na + -depleted conditions. They bind at the exofacial site of the protein as demonstrated by comparing binding affinities with specific inhibitors of the intra-and extracellular cavities of the transport channel. In contrast, apigenin and apigenin 7-O-glucoside showed a modest inhibition by apigenin and A7G. Dotted line represents the inhibition values obtained at the maximum concentration tested when D-[U-14 C]-glucose transport was assessed from apical (a) to basolateral (b) (data replotted from Fig. 6C-2). Data is expressed as mean ± SEM of three independent experiments with at least three technical replicates. In panel (B,C) results are expressed as % compared to the transported glucose in control incubations (no inhibitor; normalised to 100%). When not visible, the error bars are smaller than the data point. In panel (B) and (C), statistical difference is indicated against the incremental concentration of cytochalasin B and 4,6-ethylidene glucose respectively. **p < 0.05.
SCIentIfIC REPORTS | (2018) 8:5471 | DOI:10.1038/s41598-018-23736-1 of fructose transport and only a small fraction of the transport of D-[U-14 C]-fructose was inhibited following hydrolysis of D-[U-14 C]-sucrose by brush-border α-glucosidase. However, a higher inhibition of fructose transport in a physiological context can be expected since the apical expression of GLUT5 in the Caco-2 clone is comparable to the one noted in an adult small intestine 37 and Caco-2/TC7 cell line presents a 2-fold increase in apical GLUT5 compared to Caco-2 cells 3 .
Like other (poly)phenols, apigenin and apigenin 7-O-glucoside are bioavailable in humans but are rapidly metabolised and excreted, while they are subject to extensive colonic catabolism 38,39 . Currently, there are no pharmacokinetic data on (Z)-or (E)-MCAG but it is anticipated that they can reach the colon and are subject to catabolic reactions as observed for other hydroxycinnamic acids 40 , especially given that (E)-MCAG was not a substrate for rat intestinal enzymes. Based on information from other related compounds 41,42, no harmful effects are predicted after dietary exposure of these (poly)phenols through consumption.
Several papers have reported that apigenin in various herbal and plant extracts can inhibit α-glucosidase activity. For example, apigenin-7-O-glucoside inhibited rat intestinal α-glucosidase activity, but the study used p-nitrophenyl-glucopyranoside, rather than the physiological substrate (maltose) 43 . Many studies have used yeast as a source of α-glucosidase activity, which has no relevance to the mammalian enzymes 44 . Hence apigenin did not inhibit yeast α-glucosidase action on the non-physiological substrate p-nitrophenyl-glucopyranoside in one report 45 , but did inhibit in other reports with an IC 50 value of 22 μM in 46 and of 25.5 μg/ml in 47 . In the latter report, apigenin-7-O-glucoside also inhibited with an IC 50 value of 29.7 μg/ml 47 . Apigenin lowered GLUT1 expression in a nude mouse model of laryngeal carcinoma 48 and in prostate cancer cells 49 .
Based on our data, we propose that ChE, a natural source that combines (Z) and (E)-MCAG, apigenin and apigenin 7-O-glucoside, could pose a good alternative to modulate postprandial sugar excursions. Unlike the antidiabetic drug acarbose, the combination of these (poly)phenols can mildly attenuate carbohydrate digestion and sugar absorption, avoiding the gastrointestinal side effects of acarbose. Furthermore, apigenin and apigenin 7-O-glucoside could be employed either alone or in combination with other inhibitors for an enhanced formulation, for example in a population that has already developed a pathophysiological phenotype.

Materials and Methods
Preparation of ChE. The ChE used in this study and in enzymic in vitro assays has been previously fully described. 3

Isolation of hydroxycinnamic acid derivatives by semi-preparative column chromatography.
Two hydroxycinnamic acid derivatives detected based on previously published multiple reaction monitoring pairs by LC-MS were separated and isolated using an ÄKTA Purifier System (GE Healthcare Life Sciences, UK) equipped with an Gemini C6 phenyl column (250 × 10 mm I.D., 5 μm; Phenomenex, Cheshire, UK). The ChE extract (150 mg/mL) was loaded manually (1 mL) using a syringe through a sample loop of 1 mL and eluted using water containing 0.1% TFA (solvent A) and methanol (Solvent B) at a flow rate of 2.3 mL/min as follows: 0-12.8 min linear gradient to 24% B; 12.8-25.6 min isocratic at 24% B; 25.6-111 min linear gradient to 100% B; 111-123.8 min isocratic at 100%B, 123.8-125 min linear gradient to 5% B: 125-150.6 min isocratic at 5%B. The elution was followed at 320 nm and fractions containing the separated compounds (supplementary S1) were collected and analysed for purity by LC-MS (supplementary S6). Multiple semi-preparative separation cycles were undertaken and the fractions for each peak combined, freeze-dried, and stored at −20 °C.
Chemical characterisation of hydroxycinnamic acid derivatives. HPLC/QTOFMS. Analysis of the purified hydroxycinnamic acid derivatives was carried out using an Agilent 1200 series HPLC system coupled to an Agilent 6530 Q/TOF mass spectrometer (Agilent Technologies, USA) equipped with a dual spray ESI source operating in negative mode. Followed by an isocratic separation with reversed phase HPLC, elemental formulae of compounds were determined with the aid of MassHunter software based on high-resolution (>20,000 FWHM) accurate mass (<5 ppm) data completed with evaluation of isotope abundance matching and isotope spacing. Data was obtained from full-scan TOF-only peak spectra acquired in the mass range of m/z 50-1100. Next, manually selected ion peaks of the TOF-only spectra were subjected to Q/TOFMS analysis and elemental formulae of fragments were determined based on accurate mass (<20 ppm) data obtained from product ion peak spectra.
Acid hydrolysis. An aliquot of the purified compound 2 ( Fig. 1 and supplementary S1) was dissolved in HCl (2 M) to a concentration of 500 µM and subjected to hydrolysis for 2 h at 60 °C. The reaction mixture was concentrated to dryness under vacuum, reconstituted in water (1 mL) and analysed by HPLC-DAD on an Agilent 1200 series system (Agilent Technologies, USA) equipped with a Kinetex C18 analytical column (150 × 2.1 mm I.D., 2.6 μm; Phenomenex, Cheshire, UK) maintained at 35 °C. 10 µL was injected and separated using a 41 min gradient of premixed 5% acetonitrile in water (5:95, v/v) (A) and premixed 5% water in acetonitrile (5:95, v/v) (B), both modified with 0.1% formic acid at 0.25 mL/min. The gradient started at 0% solvent B and increased to 10% (5 min), 25% (10 min), 35% (20 min), 50% (25 min), and was held at a plateau up to 30 min. The gradient was increased 100% at 30.5 min and returned to 0% solvent B over 5.5 min before initial starting conditions were resumed for a 6 min column re-equilibration. Online detection was carried out at 320 nm.
NMR. Purified samples were dissolved in d 6 DMSO and transferred to 5 mm OD NMR tubes. All NMR data were recorded on a Bruker Avance 500 spectrometer ( 1 H frequency 500.573 MHz, 13 C frequency 125.878 MHz) temperature controlled at 299.3 K. Chemical shifts are expressed in ppm and are reported with reference to the residual solvent peak. Multiplicities are reported with coupling constants and are given to the nearest 0.1 Hz. Where needed, two-dimensional correlation spectroscopy (2D-COSY), heteronuclear single quantum coherence spectroscopy (HSQC) and heteronuclear multiple bond correlation spectroscopy (HMBC) and nuclear Overhauser effect spectroscopy (NOESY) was performed.
(Z)-2-D-Glucopyranosyloxy-4-methoxycinnamic acid present as a mixture of E and Z in a 1:3 ratio, E peaks eliminated and Z peaks picked at: 1    Enzyme assays. The human α-amylase and rat α-glucosidase inhibition assays were performed under optimised conditions as previously reported 3,18,52 . The ChE stock solution was dissolved in DMSO while the individual (poly)phenols were prepared in a water-EtOH solution (50:50, v/v). The maximum concentration of DMSO and EtOH in the assay was 2 and 1% respectively, which did not affect the activity of the enzyme as assessed in control incubations.

Molecular Modelling.
We investigated the interactions of (poly)phenols with human α-amylase using the three-dimensional crystal structure 1MFV 53 . The α-amylase structure was prepared for docking by removing from the crystallographic structure water molecules and co-crystallised ligands. No constraints were introduced and the default parameters of the dock module of the CLC Drug Discovery 3.0.1 (CLC Bio-Qiagen) were used. Validation of the docking parameters was performed by docking the co-crystal modified structure of acarbose into the active site. A root mean square deviation of 1.36 Å was obtained between the original and the docked conformations (supplementary S10). The number of docking iterations for evaluation of the preferred binding pose was set at 500. Ten best docked poses per docking calculation were analysed based on their docking score and the top ranked pose for each ligand was selected for further analysis. Ligand-receptor interactions were visualised with CLC workbench and Discovery Studio (BIOVIA, Client version 2016).
Transport assay of monosaccharides in Caco-2/TC7 cells. Caco-2/TC7 cells were cultured and transport experiments performed as fully described before 3 .

Estimation of conversion of apigenin 7-O-glucoside to apigenin by differentiated Caco-2/TC7 cells.
Cells were seeded at 2.4 × 10 5 cells/well on 6-well plates (Corning, Appleton Woods, UK) and cultured for 21-23 d at 37 °C as described above. Cells were exposed to 150 μM of apigenin 7-O-glucoside for 25 and 60 min. After the treatment the medium was collected and analysed by HPLC as detailed above.

Statistical analysis. Statistical analysis was performed by one-way ANOVA using the Number Cruncher
Statistical System version 6.0 software (NCSS, LLC). Significant differences were assessed with Tukey-Kramer multiple comparison test (p < 0.05). Independent samples t-test was used to compare means of the contribution of apigenin 7-O-glucoside and apigenin to the total inhibition of sugar transport. Data are expressed as the mean ± standard error of mean (SEM).
Data availability statement. Data available on the Open Science Framework (ID: osf.io./fs32u).