Bioactivity and cell metabolism of in vitro digested sweet cherry (Prunus avium) phenolic compounds.

Abstract In this study, the bioaccessibility of phenolic compounds after in vitro gastrointestinal digestion of two cherry cultivars was assessed. The phenolic profile was modified during in vitro digestion, with a considerable decrease of total and individual phenolic compounds. Hydroxycinnamic acids and especially coumaroylquinic acids showed the highest bioaccessibility. Isomerisation of caffeoylquinic and coumaroylquinic acids was observed after in vitro digestion. Modification of the phenolic profile after digestion resulted in an increased or decreased scavenging activity depending on the assay. In vitro digested phenolic-rich fractions also showed antiproliferative activity against SW480 but no effect against Caco-2 cell lines. Both Caco-2 and SW480 cell lines were able to metabolise cherry phenolic compounds with remarkable differences. An accumulation of glycosylated flavonols was observed in SW480 medium. In conclusion, phenolic compounds from cherries and especially hydroxycinnamic acids were efficiently released and remained bioaccessible after in vitro digestion, resulting in antioxidant and antiproliferative activities.


Introduction
Sweet cherry (Prunus avium L., Rosacea) is a commercially important table fruit highly appreciated by consumers due to its taste, colour and sweetness, but also for its nutritional properties and beneficial effects (Kelley et al. 2018). Sweet cherry fruits contain different phenolic compounds, including anthocyanins, flavan-3-ols and hydroxycinnamic acids that have been related to the healthy properties of these fruits (Ferretti et al. 2010). Indeed, cherries were considered as the main dietary source of anthocyanins within the context of the Mediterranean Diet (Godos et al. 2017).
Recently, our research group thoroughly investigated the phenolic profile of different sweet cherry varieties, from pale yellow flesh and slightly reddish skin to dark red cultivars, using mass spectrometry (Martini et al. 2017). Among the 86 identified compounds, 3-coumaroylquinic acid, epicatechin and cyanidin-3-rutinoside were present at the highest concentrations (Martini et al. 2017). The amount of individual phenolic compounds in sweet cherries is variable and strongly dependent on the cultivars (Picariello et al. 2016;Martini et al. 2017). The dark red cultivars are rich in anthocyanins, whereas the slightly reddish cultivars are rich in hydroxycinnamic acids and/or flavan-3-ols (Picariello et al. 2016;Martini et al. 2017).
There are several human intervention studies suggesting that cherry consumption may have beneficial effects on markers of cardiovascular disease risk (Kelley et al. 2018). Short-term supplementation of cherries reduced both systolic and diastolic blood pressure in diabetic women or hypertensive subjects (Keane et al. 2016;Kent et al. 2016). The effect on blood pressure correlated in time with changes in plasma cyanidin-3-glucoside metabolites such as vanillic and protocatechuic acids (Keane et al. 2016). Furthermore, cherry intake ameliorated the lipid profile of overweight and obese subjects by reducing the level of very-low density lipoproteins and increasing the amount of high-density lipoproteins (Martin et al. 2011). Some mechanisms have been proposed to explain the protective effect of cherries such as antioxidant activity, increased expression of nitric oxide synthase and decreased expression of endothelin-1 (Kelley et al. 2013;Edwards et al. 2015).
Cherry intake also resulted in an anticancer activity at gastrointestinal tract level (Ferretti et al. 2010). Apc Min mice consuming cherries, anthocyanins or cyanidin had significantly fewer and smaller intestinal adenomas than Apc Min mice consuming the control diet (Kang et al. 2003). The protective effect of cherries against intestinal cancer has been attributed to the biological activities of anthocyanins. Extracts of sweet cherries as well as anthocyanins and cyanidin aglycone showed in vitro antiproliferative activity against human cancer cells from the colon (HT-29 and HCT-15) and stomach (MKN45) (Kang et al. 2003;Serra et al. 2011;Bastos et al. 2015). These in vitro studies with cell culture models, carried out with pure phenolic compounds or cherry extracts, did not take into account the bioaccessibility and gastrointestinal tract stability of the phenolic compounds. Some phenolic compounds, and especially anthocyanins, are unstable under the gastrointestinal tract conditions (Bouayed et al. 2012;Ju aniz et al. 2017;Yang et al. 2018). Indeed, cherry phenolic compounds are entrapped in a solid food matrix and only the compounds released from the food matrix may exert their beneficial effects in the gastrointestinal tract or at systemic level (Tagliazucchi et al. 2010). Furthermore, only few recent in vitro studies investigated the metabolism and/or the stability of phenolic compounds in cell cultures (Sala et al., 2015;Mele et al., 2016;Aragon es et al. 2017).
Therefore, this work aimed to investigate the effect of in vitro gastrointestinal digestion on the bioaccessibility and antioxidant properties of cherry phenolic compounds from two different cultivars (Celeste and Durone Nero I). In addition, the antiproliferative activity and the cell metabolism of in vitro digested cherry phenolic compounds against two models of human colon adenocarcinoma cell lines were assessed.

Cherry cultivars
Two sweet cherry (Prunus avium) cultivars (Celeste and Durone Nero I) were harvested at full maturity in Vignola (Modena province, Italy). For each variety, about 2 kg of cherries were randomly sampled from several trees and processed immediately or frozen within 1 h after harvesting and stored at -80 C until used. The two cherry cultivars were selected according to the different skin and flesh colours and phenolic composition (Martini et al. 2017). Celeste cultivar presented a pale red flesh and a bright red skin and predominantly contained hydroxycinnamic acids and flavan-3-ols. Durone Nero I instead was a dark red cultivar rich in anthocyanins.

In vitro gastrointestinal digestion of cherry cultivars and preparation of the chemical extracts
For the in vitro digestion, the protocol previously developed within the COST Action INFOGEST was followed (Minekus et al. 2014). The fluids (simulated salivary, gastric and intestinal fluids) were prepared according to Minekus et al. (2014). Salivary phase was simulated by homogenising with a laboratory blender five grams of each cherry cultivar in presence of 5 ml of the stock simulated salivary fluid solution. After homogenisation, 150 U/ml of porcine a-amylase were added and the samples were incubated for 5 min at 37 C in a rotating wheel. The gastric phase of the digestion was carried out by adding to the bolus 10 ml of simulated gastric fluid. After pH adjustment to 2.0 with 6 mol/litre HCl, 2000 U/ml of pepsin were added. After 2 h of incubation at 37 C, the final intestinal step was carried out by adding 15 ml of simulated intestinal fluid. Then, the pH was adjusted to 7.0, supplemented with pancreatin and the samples were incubated at 37 C for 2 h. All samples were immediately cooled on ice and frozen at -80 C for further analysis. The digestions were performed in triplicate.
In addition, phenolic compounds were extracted from each cherry cultivar (chemical extract) as reported in Martini et al. (2017). The extractions were performed in triplicate.

Preparation of the phenolic-rich fractions
Chemical extracts and samples collected at the end of the in vitro digestion were then passed through a SPE column preconditioned with 4 ml of acidified methanol (containing 0.1% of formic acid), followed by 5 ml of acidified water (containing 0.1% of formic acid). Elution was carried out with acidified water (6 ml) to eliminate the unbound material and phenolic compounds were then desorbed with 3 ml of acidified methanol. The obtained phenolic-rich extracts were then used for the subsequent analysis. Each sample was extracted in triplicate.
Identification and quantification of phenolic compounds by liquid chromatography electrospray ionisation ion trap mass spectrometry (LC-ESI-IT-MS) Phenolic-rich fractions from chemical extracts and in vitro digested samples were analysed on HPLC Agilent 1200 Series system equipped with a C18 column (HxSil C18 Reversed phase, 250 Â 4.6 mm, 5 lm particle size, Hamilton Company, Reno, NV) as reported in Martini et al. (2017). The mobile phase composition was (A) H 2 O/formic acid (99:1, v/v) and (B) acetonitrile/formic acid (99:1, v/v). After 0.5 min at 4% B, the gradient linearly ramped up to 30% B in 60 min. The mobile phase composition was raised up to 100% B in 1 min and maintained for 5 min to wash the column before returning to the initial condition. The flow rate was set at 1 ml/min. The samples were injected in the amount of 20 ll. After passing through the column, the eluate was split and 0.3 ml/min were directed to an Agilent 6300 ion trap mass spectrometer. Two MS experiments were performed, one in negative ion mode and one using positive ionisation (for anthocyanins), under the same chromatographic conditions. Identification of phenolic compounds in all samples was carried out using full scan, data-dependent MS 2 scanning from m/ z 100 to 800 and selected reaction monitoring. MS operating conditions, limits of detection (LOD) and limits of quantification (LOQ) for the different standards are reported in Martini et al. (2017).
Quantitative results were expressed as lmol of compounds per 100 g of cherry.

Bioaccessibility calculation
The percentage bioaccessibility of phenolic compounds after in vitro gastrointestinal digestion was calculated as follow: where CCd is the concentration of total or individual phenolic compounds in the phenolic-rich fraction from in vitro digested sample and CCe is the concentration in the phenolic-rich fraction from the chemical extract.

Antioxidant activity assays
The antioxidant properties of the phenolic-rich fractions obtained from chemical extracts and in vitro digested samples were detailed using four different assays. The ABTS (2,2'-azino-bis (3-ethylbenzothiazoline-6-sulfonic acid) and ferric reducing power (FRAP) assays were performed according to the protocols described by Re et al. (1999) and Benzie and Strain (1996), respectively. The capacity to scavenge hydroxyl and superoxide anion radicals was evaluated according to the methods reported by Martini et al. (2017).
The results were expressed as lmol of trolox equivalent per mmol of total phenolic compounds.

Cell cultures and antiproliferative activity of cherry phenolic-rich fractions
Human adenocarcinoma Caco-2 cells were cultured in Dulbecco's Modified Eagle's Medium (DMEM) supplemented with 10% foetal bovine serum (FBS), 1% antibiotic mix (streptomycin and penicillin) and 2 mmol/litre L-glutamine. Human adenocarcinoma SW480 cells were cultured in Leibowitz medium supplemented with 10% FBS, 1% antibiotic mix (streptomycin and penicillin) and 2 mmol/litre L-glutamine. Cells were maintained at 37 C in a humidified atmosphere of 5% CO 2 .
Cells were seeded at 5 Â 10 3 /100 ll and 10 Â 10 3 / 100 ll for Caco-2 and SW480, respectively, in 96-well plates for 24 h to allow cell adhesion to the bottom of the wells.
For the antiproliferative assays, a colorimetric method for the sensitive quantification of viable cells was performed, using MTS assay kit. 100 ll of digested and un-digested cherry phenolic-rich fractions diluted in cell culture media at different concentrations were added to the cell plates and incubated for 24 h. At the end of the treatments, media were refreshed with 180 ll of culture media and 20 ll of MTS reagent were added to each well. After 4 h of incubation at 37 C, the absorbance was measured at the wavelength of 490 nm using a microplate reader and results were expressed as IC 50 . IC 50 was defined as the concentration of phenolic compounds required to inhibit 50% cell proliferation and expressed as lmol of total phenolic compounds/ litre. The IC 50 values were determined using nonlinear regression analysis and fitting the data with the log (inhibitor) vs response model generated by GraphPad Prism 6.0 (GraphPad Software, San Diego, CA).

Liquid chromatography coupled to mass spectrometry (LC-MS/MS) analysis of cell media
Cell culture supernatants were collected at the end of the experiments and analysed by LC-MS/MS to determine the cell metabolism of the cherry phenolic compounds. Cell media were extracted according to Sala et al. (2015) and analysed using a HPLC Agilent 1200 Series system equipped with an Agilent 6300 ion trap mass spectrometer as reported above. Separations were performed using a C18 column (HxSil C18 Reversed phase, 250 Â 4.6 mm, 5 lm particle size, Hamilton Company), with an injection volume of 40 ll and elution flow rate of 1 ml/min. The mobile phase composition, the gradient and MS operating conditions are the same as reported above.

Statistic
All data are presented as mean ± SD for three replicates for each prepared sample. One-way analysis of variance (one-way ANOVA) with Tukey's post hoc test was applied using Graph Pad prism 6.0 (GraphPad software). The differences were considered significant with p < .05.

Results and discussion
In vitro bioaccessibility of phenolic compounds in cherry cultivars Table 1 shows data about the bioaccessibility of individual phenolic compounds in the two analysed cherry cultivars. The MS data of the individual phenolic compounds are reported in Martini et al. (2017).
A total of 57 and 56 phenolic compounds were identified and quantified in the chemical extracts of cherry cultivars Celeste and Durone Nero I, respectively (Table 1). In both the cultivars, 3-coumaroylquinic acid (found as cis and trans isomers) was the main phenolic compound accounting for 40.6% and 31.3% of the total compounds in Celeste and Durone Nero I, respectively. In general, hydroxycinnamic acids were the most representative class of phenolic compounds in both the cultivars (54.2% and 42.2% of the total phenolic compounds in Celeste and Durone Nero I, respectively). From a quantitative point of view, flavan-3-ols were the second class of phenolic compounds in both the cultivars (40.7% and 33.4% of the total phenolic compounds in Celeste and Durone Nero I, respectively). Anthocyanins accounted for the 15.4% of total compounds in the cultivar Durone Nero I whereas their incidence was about 3% in the cultivar Celeste. Finally, small amounts of others flavonoids were also detected in both the cultivars.
The phenolic compounds profile was significantly modified from a qualitative and quantitative point of view after in vitro gastrointestinal digestion ( Figure 1; Table 1). Only 25 and 39 of the original phenolic compounds were identified after simulated digestion of the cultivars Celeste and Durone Nero I, respectively. This meant that 56% and 30% of individual phenolic compounds were not bioaccessible.
Total phenolic compounds determined by LC-ESI-IT-MS/MS in the cherry cultivars chemical extracts were significantly higher (p < .05) than those found after in vitro gastrointestinal digestion (Figure 1(A); Table 1). The bioaccessibility of total phenolic compounds was 39.7% and 29.9% in the cultivars Celeste and Durone Nero I, respectively. Hydroxycinnamic acids were the most abundant phenolic compounds released from cherry matrices with a bioaccessibility of 71.7% and 46.5% in the cultivars Celeste and Durone Nero I, respectively (Figure 1(B); Table 1). At the end of the digestion, hydroxycinnamic acids represented the 98.0% and 65.5% of total phenolic compounds in the cultivars Celeste and Durone Nero I, respectively (Figure 1(H)). These compounds were particularly stable under gastrointestinal conditions and easily released from the cherry matrices. The stability of hydroxycinnamic acids under gastrointestinal conditions has been already reported. For example, Tagliazucchi et al. (2012) found that only the 30% of hydroxycinnamic acids were degraded during in vitro digestion of a coffee beverage whereas Monente et al. (2015) assessed a 14% decrease during in vitro digestion of a spent coffee water extract. Previous studies with solid food matrices confirmed the easy release of hydroxycinnamic acids during in vitro digestion. The bioaccessibility of hydroxycinnamic acids was 60-67% in cooked cardoon, 32-57% in apple and about 77% in frozen sweet cherries subjected to gastrointestinal digestion (Fazzari et al. 2008;Bouayed et al. 2012;Ju aniz et al. 2017). This is also in accordance with the recorded recovery of caffeoylquinic acids (59-77%) in ileal fluid after coffee or apple ingestion by ileostomists (Stalmach et al. 2010;Erk et al. 2014). The highest release after incubation with digestive fluids was found for coumaroylquinic acids, followed by feruloylquinic and caffeoylquinic acids (Table 1; Figure 2). These results are in accordance with Monente et al. (2015) and Tagliazucchi et al. (2012) who found that feruloylquinic and caffeoylquinic acids were quite stable under gastrointestinal conditions. No data about the stability or bioaccessibility of coumaroylquinic acids have been published until now. Otherwise, the minor hydroxycinnamic acids identified in cherries (such as di-hydroxycinnamic acids and lactone derivatives) were not bioaccessible probably because of their instability in the gastrointestinal environment (Tagliazucchi et al. 2012;Monente et al. 2015). The differences between the recoveries of hydroxycinnamic acids in the two different cherry cultivars could be indicative of a food matrix impact ( Figure 2; Table 1). As reported above, overall coumaroylquinic, caffeoylquinic and feruloylquinic acids were easily released from food matrices, although changes mainly due to variations in isomers level were recorded after in vitro digestion. While the 5-and 3-caffeoylquinic acid isomers were found at lower concentrations after in vitro digestion compared to the chemical extracts, the 4-caffeoylquinic acid isomers concentration increased significantly (Figure 2(A,B); Table 1). Acyl migration of caffeoylquinic acids has been already demonstrated both in vitro with pure standard compounds or after digestion of coffee and in vivo in ileostomy volunteers (Tagliazucchi et al. 2012;Erk et al. 2014;Monente et al. 2015). Similar results were observed for feruloylquinic acids (Figure 2(C,D); Table  1). Otherwise, a different behaviour was observed for coumaroylquinic acids. In this case, we found remarkable higher concentrations of 4-and 5-coumaroylquinic acid isomers after simulated digestion ( Figure  2(E,F); Table 1). On the contrary, the concentration of 3-coumaroylquinic acid isomers was lower after digestion than that recorded after chemical extraction. Previously, Kahle et al. (2011) reported isomerisation of 4-coumaroylquinic acid to 3-and 5-coumaroylquinic acids after incubation of standard compounds with intestinal fluids. These results, instead, suggested that coumaroylquinic acids were subjected to isomerisation from the 3-acyl to 4-and 5-acyl structures.
Simple hydroxycinnamic acids (coumaric, caffeic and ferulic acids) were not found after intestinal digestion, suggesting that relevant hydrolysis of ester bond did not occur during in vitro digestion. Conversely, incubation of caffeoylquinic and coumaroylquinic acids standards resulted in the appearance of simple acids after 6 h of incubation in intestinal fluids (Kahle et al. 2011).
Flavan-3-ols, which were the second most abundant class of phenolic compounds in cherry chemical extracts, were characterised by a very low bioaccessibility (Figure 1(C); Table 1). In the cultivar Celeste, only (epi)catechin-3-gallate was detected after gastrointestinal digestion. In the cultivar Durone Nero I, we also found monomeric and dimeric (epi)catechins, although (epi)catechin-3-gallate was still the flavan-3-ol with the highest bioaccessibility. The bioaccessibility of monomeric catechins seems to be highly dependent on the food matrix. During in vitro digestion of green tea, Green et al. (2007) found that only the 32% of epicatechin was degraded. High stability of monomeric catechins was also described Values represent means ± standard deviation of triplicate determination. Ã Hydroxycinnamic acids were quantified as 5-caffeoylquinic acid equivalent; flavan-3-ols were quantified as epicatechin equivalent; flavonols and other flavonoids were quantified as quercetin-3-glucoside equivalent; hydroxybenzoic acids were quantified as gallic acid equivalent; anthocyanins and derivatives were quantified as cyanidin-3-glucoside equivalent. Different superscript letters within the same row indicate that the values are significantly different (p < .05). <LOQ means the compound was detected but it was below the limit of quantification. n.d.: means not detected.
when the standard compounds were incubated with gastrointestinal fluids (Tagliazucchi et al. 2010). However, in vitro digestion of apple varieties, grape or wine resulted in an important degradation of monomeric catechins (Bouayed et al. 2012;Lingua et al. 2018). Whether catechin and epicatechin were degraded or not released from cherry matrices remained to be established. Cultivar Durone Nero I was also characterised by a high content of anthocyanins. In this cherry cultivar, anthocyanins represented the phenolic class with highest bioaccessibility (55.0%), constituting the 28.2% of total phenolic compounds after in vitro digestion (Figure 1(G,H); Table 1). Anthocyanins are generally considered very unstable compounds under slightly alkaline conditions as found in the intestinal milieu. For example, in vitro digestion of red wine, chokeberry and pomegranate juice resulted in a substantial degradation of monomeric anthocyanins (P erez-Vicente et al. 2002;Berm udez-Soto et al. 2007;Lingua et al. 2018). Anthocyanins stability in the digestive system appeared to be related to their molecular structure. The presence of a methoxy group in the B-ring stabilised the anthocyanin structure (Yang et al. 2018). The acetylated derivatives of anthocyanins showed relatively higher stability during the gastrointestinal transit than simple anthocyanins (Yang et al. 2018). As shown in Table 1, the rutinoside derivatives of cherry anthocyanins were more stable than the corresponding glycosidic forms. These results are in agreement with Fazzari et al. (2008). It can be speculated that fruits containing high levels of rutinoside derivatives are likely to provide larger amounts of anthocyanins in the colon. Indeed, peonidin-3-O-rutinoside, which is characterised by the presence of a rutinose moiety and a methoxy group in the B-ring showed 100% of bioaccessibility.
Some newly identified compounds were indicative of anthocyanins degradation (Table 1). Protocatechuic acid appeared after in vitro digestion of cherry cultivars and plausibly came from the degradation of cyanidin-3-O-glucoside. Moreover, cyanidin-3-O-glucoside chalcone was also found after in vitro digestion of Durone Nero I due to the presence of its [M þ H] þ precursor ion at 467 and product ions at m/z 305 (-162 amu; loss of a glucosyl moiety), m/z 449 (-18 amu; loss of water) and m/z 287 (-18-162 amu; loss of water and glucosyl moiety) (Lopes et al. 2007). After in vitro digestion of Durone Nero I, a new compound with a [M þ H] þ mass signal at 737 was detected. In ESI positive experiments, it gave a base peak product ion at m/z 575 (-162 amu) corresponding to the loss of a glucosyl moiety. Additional product ions were detected at m/z 557 (loss of water from m/z 575) and m/z 287 (loss of a (epi)catechin unit from m/z 575). Finally, a product ion at m/z 423 was detected corresponding to the loss of 152 amu from m/z 575 resulting from the retro Diels-Alder decomposition of the flavan-3-ol unit. The described fragmentation pattern corresponded to that previously observed for a (epi)catechin-cyanidin-3-O-glucoside dimer (Sentandreu et al. 2010).
The bioaccessibility of the other phenolic classes was very low with the exception of flavonols (52.2% and 39.3% of bioaccessibility in the cultivars Celeste and Durone Nero I, respectively) (Figure 1(D); Table  1). However, due to their low concentration in the cherries, they represented only the 1.5% and 4.4% of total phenolic compounds determined after in vitro digestion in the cultivars Celeste and Durone Nero I, respectively (Figure 1(H); Table 1).

Effect of in vitro digestion on the antioxidant properties
To fully characterise the antioxidant properties of the phenolic-rich fractions of cherries, the ability to scavenge some physiologically relevant radicals (superoxide anion and hydroxyl radical) and the organic nitro-radical ABTS as well as the reducing power were evaluated before and after digestion (Figure 3). The hydroxyl radical scavenging capacity of phenolic-rich extracts decreased after in vitro digestion of both the cherry cultivars (Figure 3(A)). This means that cherry phenolic compounds in the chemical extracts were more efficient scavengers of hydroxyl radical than the phenolic compounds present in the digested cherries. This decrease may reflect the different composition of the chemical and in vitro digested phenolic-rich extracts. In both the cherry cultivars, the in vitro digested extracts were enriched in hydroxycinnamic acids (Table 1; Figure 1(H)) that showed the lowest hydroxyl scavenging activity among phenolic compounds (Ozy€ urek et al. 2008). On the contrary, ferric reducing power of the phenolic-rich extracts increased after in vitro digestion ( Figure  3(B)). This clearly reflected the different composition of the chemical and digested extracts, since the ferric reducing ability of hydroxycinnamic acids is considered higher than that of flavan-3-ols (Pulido et al. 2000). The superoxide anion scavenging activity instead was similar between the phenolic-rich extracts before and after in vitro digestion (Figure 3(C)). Finally, the ABTS radical scavenging activity assay displayed a different behaviour in the two cherry cultivars (Figure 3(D)). In Celeste cultivar, the phenolicrich extract obtained after in vitro digestion showed a lower scavenging ability than the extract before digestion. This is in keeping with the observed percent increase in hydroxycinnamic acids that showed lower ABTS radical scavenging activity respect to flavan-3ols (Rice-Evans et al. 1996). Instead, in the cultivar Durone Nero I we observed an increase in the ABTS radical scavenging activity in the phenolic-rich extract after in vitro digestion. This can be related to the percentage increase of anthocyanins and flavonols in the digested extracts (Table 1; Figure 1(H)), which displayed greater ABTS radical scavenging activity than flavan-3-ols (Rice-Evans et al. 1996).
Indeed, it is important to note that the different incidence of the phenolic classes in the digested phenolic-rich extracts gave rise to significant different behaviours in the scavenging capacity or reducing power in relation to the type of assay and mechanism involved, leading to an increase or decrease of the considered activities.
Antiproliferative activity of cherry phenolic-rich fractions on human Colon adenocarcinoma cell lines The effect of cherry phenolic-rich fractions extracted before and after in vitro gastrointestinal digestion on the proliferation of human colon adenocarcinoma Caco-2 and SW480 cells was investigated. Caco-2 and SW480 cells were incubated with different concentrations of phenolic-rich extracts for 24 h. Cherry phenolic-rich extracts did not affect the proliferation of Caco-2 (data not shown). Instead, the phenolic compounds from the chemical extracts of both the cultivars were able to inhibit in a dose-dependent manner the proliferation of SW480 cell line (Figure 4). The chemical extract of cultivar Durone Nero I was about 2.5 times more efficient than Celeste with IC 50 values of 48.01 ± 3.02 and 121.90 ± 5.03 lmol/litre, respectively. In vitro digestion positively affected the antiproliferative activity of Celeste phenolic-rich fraction. As reported in Figure 4, the IC 50 value of phenolic compounds extracted after digestion of cultivar Celeste halved to 61.22 ± 4.02 lmol/litre. No significant differences were found between the IC 50 values calculated for cultivar Durone Nero I before and after digestion.
According to the literature, this is the first report investigating the antiproliferative activity of sweet cherry phenolic-rich extracts against human colon adenocarcinoma Caco-2 and SW480 cell lines and the influence of in vitro digestion processes.
Previous in vitro studies have shown antiproliferative properties of sweet cherry extracts against colon cancer cell lines such as HCT-15 and HT29 (Serra et al. 2011;Bastos et al. 2015). This effect has been attributed to the anthocyanin content of the cherry extracts (Wang and Stoner 2008). However, Olsson et al. (2004) found no correlation between proliferation of HT29 cells and the concentration of anthocyanins in ten fruits and berries (including cherries). In accordance, also in this study we did not find any correlation between the anthocyanin content and the antiproliferative activity. In fact, Durone Nero I chemical extract contained five times more anthocyanins than Celeste chemical extract but exhibited only 2.5 times more antiproliferative activity. Indeed, in vitro digested phenolic fraction from Celeste was as effective as Durone Nero I chemical extract although it did not contain anthocyanins.
Our results suggested that other compounds rather than anthocyanins can be responsible for the observed antiproliferative effect of phenolic-rich fractions extracted before and at the end of the in vitro gastrointestinal digestion.

Metabolism of cherry phenolic compounds in cell cultures
In an attempt to unravel the metabolism of cherry phenolic compounds by Caco-2 and SW480, cell media were collected after 24 h of incubation with in vitro digested Durone Nero I phenolic-rich fraction and subjected to LC-MS ion trap analysis. Some parent compounds and newly formed metabolites were detected in both cell types and reported in Table 2.
Total coumaroylquinic acids final concentration accounted for the 13.9 and 14.7% of the initial concentration in the cultured Caco-2 and SW480, respectively. In addition to the coumaroylquinic acid isomers, one newly formed metabolite was tentatively identified as coumaroylquinic acid sulphate in the cell media of both the cell lines. The rate of sulphation of coumaroylquinic acid isomers in SW480 was 0.5%, while in Caco-2 coumaroylquinic acid sulphate was found only at trace levels. To the best of our knowledge, this is the first report showing sulphation of coumaroylquinic acids by Caco-2 and SW480. Coumaric and (iso)ferulic acids were found in the media after incubation with both cell lines, whereas caffeic acid only after incubation with SW480 ( Table  2). The appearance of these simple hydroxycinnamic acids, which were not present in the digested sample, was accompanied by the disappearance of caffeoylquinic and feruloylquinic acids (with the exception of trace amounts of 3-feruloylquinic acid) and the decrease in concentration of coumaroylquinic acids. These results can be indicative of hydroxycinnamoylquinic acids degradation in the media or to the hydrolysis catalysed by esterases. Considering the initial concentration of feruloylquinic acids, their complete hydrolysis during incubation with cells should result in a (iso)ferulic acid concentration of 10.47 lmol/litre. However, the amount of (iso)ferulic acid found in cell medium after 24 h of incubation with Caco-2 was about 5.5 times higher (Table 2). Previous studies have shown that Caco-2 possess Table 2. Phenolic compounds extracted from in vitro digested Durone Nero I identified in the cell media after 24 h of incubation with Caco-2 and SW480. Results are expressed as lmol/litre. Ã catechol-O-methyltransferase able to metabolise caffeic acid to (iso)ferulic acid (Kern et al. 2003;Farrell et al. 2011;Monente et al. 2015). Therefore, the exceeded amount of (iso)ferulic acid can be a consequence of the methylation of caffeic acid released from hydrolysis of caffeoylquinic acids. The mass balance recovery of (iso)ferulic acid also considering the initial amount of caffeoylquinic acids present in the digested sample was 35.6%, suggesting the partial methylation of caffeic acid released after hydrolysis of caffeoylquinic acids. This conclusion is also supported by the lack of identification of caffeic acid in the medium after 24 h of incubation with Caco-2.
The same conclusions cannot be drawn for SW480. In the medium of this cell line, we found some residual caffeic acid and the amount of (iso)ferulic acid was 1.2 times higher respect than the amount found at the end of the digestion (Table 2). It can be speculated that SW480 expressed lower amount of catechol-O-methyltransferase or that the isoform in these cells had a lower affinity for caffeic acid respect to the isoform in Caco-2 cells.
Another difference between the two cell lines was related to the ability to metabolise glycosylated flavonols. Caco-2 cells were found to be able to hydrolyse quercetin-3-O-glucoside to the corresponding aglycone. Furthermore, a newly formed metabolite, quercetin-3-O-glucoside sulphate appeared in the cell medium after 24 h of incubation with Caco-2. The ability of Caco-2 cells to hydrolyse glycosylated flavonols was previously reported (Boyer et al. 2004). Finally, glycosylated flavonols tended to be accumulated after 24 h of incubation with SW480 cell line.

Conclusions
Bioactivity of phenolic compounds is primarily conditioned by their bioaccessibility in the gastrointestinal tract, and second on their cellular uptake and internal transformation. The present study determined the amounts of bioaccessible phenolic compounds from two cherry cultivars after in vitro gastrointestinal digestion. Results showed that the digestion process modified from a qualitative and quantitative point of view the profile of phenolic compounds in the samples. Hydroxycinnamic acids and flavonols showed the highest bioaccessibility. Isomerisation of caffeoylquinic and coumaroylquinic acids was observed after in vitro gastrointestinal digestion. We have also demonstrated that Caco-2 and SW480 cell lines showed metabolic activity resulting in a partial modification of cherry phenolic compounds. Methylation, sulphation, de-glycosylation and hydrolysis of esters were the main pathways of phenolic compounds metabolism in the tested cell lines. Moreover, we demonstrated that bioaccessible cherry phenolic compounds showed antiproliferative activity against two models of human colon adenocarcinoma cell lines. The presence of (iso)ferulic acid in Caco-2 cell medium at higher concentration than in SW480 cell medium or the presence of caffeic acid only in SW480 cell medium and quercetin and quercetin-3-O-glucoside sulphate only in Caco-2 cell medium may suggest the intrinsic differences among the two cell lines and the metabolic mechanisms involved. The different effects on the two tested cell lines can be related to the accumulation of glycosylated flavonols in the cell medium of SW480 or to a different sensitivity of the cell lines to the phenolic compounds. Nevertheless, it could be interesting to underline how the sensitivity to the presence of phenolic-rich extracts increased after digestion, where the hydroxycinnamic acids were the major phenolic class present.
Further studies are necessary to confirm the proposed pathways of metabolism of cherry phenolic compounds during incubation with cell lines and the potential role of hydroxycinnamic acids.

Disclosure statement
The authors report no conflict of interest.