This is an open access article published under an ACS AuthorChoice License, which permits copying and redistribution of the article or any adaptations for non-commercial purposes.

Article

pubs.acs.org/JAFC

Gastrointestinal Simulation Model TWIN-SHIME Shows Differences between Human -Metabotypes in Gut Microbiota Composition, Pomegranate Polyphenol Metabolism, and Transport along the Intestinal Tract † # § ∥ ⊥ # ‡ † † Rocío García-Villalba, , Hanne Vissenaekens, , , , Judit Pitart, María Romo-Vaquero, Juan C. Espín, § † ∥ ⊥ ‡ Charlotte Grootaert, María V. Selma, Katleen Raes, Guy Smagghe, Sam Possemiers, § † John Van Camp, and Francisco A. Tomas-Barberan*, † Research Group on Quality, Safety, and Bioactivity of Plant Foods, Laboratory of Food & Health, Department of Food Science and Technology, CEBAS-CSIC, 30100 Campus de Espinardo, Murcia Spain ‡ ProDigest BVBA, Ghent, Belgium § Department of Food Safety and Food Quality, Faculty of Bioscience Engineering, Ghent University, Ghent, Belgium ∥ Department of Industrial Biological Sciences, Faculty of Bioscience Engineering, Ghent University, Kortrijk, Belgium ⊥ Department of Crop Protection, Faculty of Bioscience Engineering, Ghent University, Ghent, Belgium

*S Supporting Information

ABSTRACT: A TWIN-SHIME system was used to compare the metabolism of pomegranate polyphenols by the gut microbiota from two individuals with different urolithin metabotypes. Gut microbiota, metabolism, short-chain fatty acids (SCFA), transport of metabolites, and phase II metabolism using Caco-2 cells were explored. The simulation reproduced the in vivo metabolic profiles for each metabotype. The study shows for the first time that microbial composition, metabolism of , and SCFA differ between metabotypes and along the large intestine. The assay also showed that pomegranate phenolics preserved intestinal cell integrity. Pomegranate polyphenols enhanced urolithin and propionate production, as well as Akkermansia and Gordonibacter prevalence with the highest effect in the descending colon. The system provides an insight into the mechanisms of pomegranate polyphenol gut microbiota metabolism and absorption through intestinal cells. The results obtained by the combined SHIME/Caco-2 cell system are consistent with previous human and animal studies and show that although urolithin metabolites are present along the gastrointestinal tract due to enterohepatic circulation, they are predominantly produced in the distal colon region. KEYWORDS: , urolithin, phenotypes, gut microbiota, intestinal cells

■ INTRODUCTION metabolites trigger different molecular and cell responses that may account, at least partially, for the antioxidant, anti- Dietary ellagitannins (ETs) and ellagic acid (EA) have been fl associated with important health effects and benefits in diseases in ammatory, anticancer, cardio-metabolic, and neuroprotec- 1,2 tive effects attributed to ETs and (or) to ET-containing including cardiovascular disease. They are present in dietary 2,5,12,13 3 foods. Therefore, the study of the mechanisms for sources in larger amounts than previously estimated. In fi fi humans, ETs are not absorbed as such, and the absorption of urolithin production, the identi cation of the speci c regions of EA is rather low.4 Both ETs and EA are catabolized by the gut the intestine where they are formed, and the gut microbiota microbiota leading to urolithin metabolites.5 The final involved are of special interest. After pomegranate polyphenols metabolites in this catabolic conversion are urolithin A (Uro- intake, several have been detected in human feces, urine, and also in biopsies taken from prostate and different A), (Uro-B), and isourolithin A (Isouro-A) (Figure 14,15 1). Not all individuals have the appropriate gut microbiota to regions of the colon in cancer patients. In vitro production fi ff of urolithins from EA by human fecal microbiota from both produce the nal urolithin metabolites, and three di erent 16 urolithin metabotypes (UMs), UM-A, UM-B, and UM-0, have metabotypes A and B has also been described. However, the been reported.6 Species of the genus Gordonibacter have been gastrointestinal tract site for urolithin production, the stability, identified as gut microbiota constituents that are involved in the and absorption of the metabolites in the gut are still unknown. 7,8 fi conversion of EA into intermediary urolithins. Fecal To date, Gordonibacter levels have only been quanti ed in Gordonibacter concentrations correlate positively with uroli- thin-A content in feces and urine,9 although other unknown Received: May 3, 2017 bacterial species are needed to produce the final urolithin Revised: June 13, 2017 metabolites. Urolithins reach concentrations within the micro- Accepted: June 15, 2017 molar range in human plasma,10,11 and these bioavailable Published: June 15, 2017

© 2017 American Chemical Society 5480 DOI: 10.1021/acs.jafc.7b02049 J. Agric. Food Chem. 2017, 65, 5480−5493 Journal of Agricultural and Food Chemistry Article

Figure 1. Gut microbiota catabolism of pomegranate ETs () to urolithin metabolites, and differences between metabotypes A and B. In blue, urolithin metabolites derived from ellagic acid. In red, transient intermediate metabolite not detected. Compound numbers as in the chromatograms of Figure 2. human fecal samples, but its distribution throughout the subjected to a stomach and small intestine (SI) digestion to digestive tract and its role in urolithin production are still estimate the bioavailability of native polyphenols and their unknown. Fecal Gordonibacter levels are higher in urolithin catabolism in the upper part of the gastrointestinal tract. Long- metabotype A (UM-A) individuals than in those with urolithin term microbial colon fermentation was also investigated in the 17 metabotype B (UM-B) and urolithin metabotype 0 (UM-0). TWIN-SHIME, thus determining the gut microbiota metabo- Modulation of some human fecal bacteria by consumption of lism of ETs in the colon, the urolithin production pathway, the ET- rich food such as pomegranate has recently been sites of transformation, and the metabolite profile of (poly)- described,18,19 and the increase of fecal Gordonibacter levels phenolics which have potential to be absorbed. The production was highlighted.19 However, modulation of Gordonibacter and of specific SCFA was also evaluated, as well as the modulation other bacterial groups by ET-rich foods along the digestive tract as well as their differences between metabotypes has not been of gut microbiota. The intestinal transport and cell metabolism explored and requires further research. of the (poly)phenolics were also evaluated through direct In the present study, a simulator of the human intestinal addition of diluted phenolics-containing SHIME matrix to microbial ecosystem (TWIN-SHIME) was used to shed light Caco-2-cells. Overall, our results are of interest to validate this on ET gut microbiota metabolism in the different regions of the system when compared with the results previously obtained in intestine. A pomegranate extract (PE) supplement was vivo.

5481 DOI: 10.1021/acs.jafc.7b02049 J. Agric. Food Chem. 2017, 65, 5480−5493 Journal of Agricultural and Food Chemistry Article ■ MATERIALS AND METHODS the limits of detection and quantification using this matrix (SHIME medium) (Suppl. Table 1). Pomegranate Extract (PE) and Chemicals. Characterized PE SCFA Analysis. TWIN-SHIME samples were analyzed as was provided by Laboratorios Admira S.L. (Alcantarilla, Murcia, 24 fl 19 previously described. Brie y, SCFA were extracted from the samples Spain). EA, punicalagin and 6,7-dihydroxycoumarin (DHC) were with diethyl ether, after the addition of 2-methyl hexanoic acid as an from Sigma-Aldrich (St. Louis, MO, USA). Urolithins were obtained 3 internal standard. Extracts were analyzed using a GC-2014 gas as previously described. Purity was higher than 95% for all tested chromatograph (Shimadzu, Hertogenbosch, The Netherlands), compounds. Organic solvents such as methanol, acetone, and equipped with a capillary fatty acid-free EC-1000 Econo-Cap column acetonitrile were from Merck (Darmstadt, Germany). All chemicals (dimensions, 25 mm × 0.53 mm; film thickness, 1.2 mM; Alltech, and reagents were of analytical grade. Laarne, Belgium), a flame ionization detector, and a split injector. The fi Volunteer Strati cation and Characterization. To select the injection volume was 1 mL, and the temperature profile was set from fecal donors for UM-A and UM-B, 13 individuals consumed 30 g 110 to 160 °C, with a temperature increase of 6 °C/min. The carrier walnuts/day for 3 days, and urine samples were collected on the third gas was nitrogen, and the temperature of the injector and detector was day. Urolithin production and metabolic profiles were evaluated using 100 and 220 °C, respectively. 20 HPLC-DAD-MS, and eight individuals were stratified as UM-A, four DNA Extraction and Microbial Analysis. Powerfecal DNA as UM-B, and one as UM-0. This distribution is consistent with isolation kit (Mo-Bio Laboratories, Carlsbad, CA, USA) was used to normal values as previously reported.20,6 Representatives of UM-A and isolate total DNA from different SHIME samples. An additional step UM-B were selected as fecal donors for the assay of ET metabolism in was done consisting of vigorous shaking using a FastPrep Instrument the gastrointestinal simulator. and 2 mL tubes containing special beads (MP Biomedicals, LLC, Metabolism of the ETs in the TWIN-SHIME and Sampling. A Ohio, USA). After DNA extraction, DGGE was used to monitor the SHIME setup (registered name from Ghent University and most prominent shifts within the overall microbial community ProDigest), simulating the entire human gastrointestinal tract, was together with group-specific shifts within the total bacteria25 and − used as previously reported.21 23 To investigate two different Clostridium cluster XIVa.26 After DNA extraction and PCR with metabotypes (UM-A and UM-B) at the same time, a TWIN-SHIME general or group-specific primers, DGGE was performed to separate setup23 was used by operating two systems in parallel. The first two PCR products.25,26 Gels had a denaturizing gradient from 45% to 60% reactors are of the fill-and-draw principle to simulate different steps in and were run using a DCodeTM Universal Mutation Detection food uptake and digestion, with peristaltic pumps adding a defined System (Bio-Rad). Data analysis was carried out using GelCompar amount of SHIME nutritional medium (140 mL 3×/day) and version 6.6 (Applied Maths, Sint-Martens-Latem, Belgium). Pearson pancreatic and bile liquid (60 mL 3×/day), respectively, to the correlation and UPGMA (Unweighted Pair Group Method using stomach and small intestine (SI) compartments and emptying the Arithmetic Mean) clustering were used to calculate dendrograms of respective reactors after specified intervals. The last three compart- DGGE profiles. ments simulate the ascending (AC), transverse (TC), and descending qPCR for total bacteria, Firmicutes, Bacteroidetes, Bacteroides spp., Bifidobacterium spp., Lactobacillus spp., Akkermansia spp., and (DC) colon. Inoculum preparation, retention time, pH, temperature 17,19 settings, and reactor feed composition were previously described.23 Gordonibacter spp., was performed as previously reported. Real- After inoculating the colon reactors of the TWIN-SHIME system with time qPCR was run on the ABI 7500 system (Applied Biosystems, ’ fresh fecal samples from UM-A and UM-B, a two-week stabilization ABI, Madrid, Spain) following the manufacturer s conditions. period allowed the microbial community to differentiate in the Maintenance of Intestinal Cell Culture. The human colon reactors. After the stabilization period, the standard SHIME nutrient adenocarcinoma cell line Caco-2 (HTB37) was obtained and cultured as reported previously.27 Cells were maintained in an incubator with a matrix was further dosed to the model for 2 weeks. Analysis of samples ° in this control period allowed us to determine the baseline microbial water saturated atmosphere of 10% CO2 at 37 C (Memmert CO2 community composition and activity in the different reactors, which incubator, Memmert GmbH & Co., Nurnberg, Germany). The cell were used as a control to compare with the results from the treatment culture medium was replaced 3 times a week, and when Caco-2 cells reached 80−90% confluence, the cells were subcultured using 0.25% period where the basic diet was supplemented with 1.8 g/day of PE (v/v) trypsin-EDTA solution (Sigma-Aldrich). during 3 weeks. TWIN-SHIME samples obtained from the different Cytotoxicity Measurements. Using MTT and SRB assays, the reactors (SI, AC, TC, and DC) were analyzed. Samples were taken for cytotoxicity of (i) PE-free intestinal matrix, (ii) digested PE, and (iii) chemical (ET derived metabolites and SCFA profile) and microbial 27,28 undigested PE on the intestinal Caco-2 cells was investigated. (denaturing gradient gel electrophoresis (DGGE) and quantitative Intestinal Caco-2 cells were maintained for 21 days to obtain confluent PCR (qPCR)) analyses. monolayers of differentiated intestinal cells as previously reported.27,28 EA and Urolithin LC-MS Analysis. Samples (2 mL) from the 21 Days postseeding, the Caco-2 monolayers were loaded for 4 h with different vessels of the TWIN-SHIME were extracted with 2 mL of fi 1/5 (v/v) dilutions of the intestinal matrix and the (un)digested PE in ethyl acetate acidi ed with 1.5% formic acid. The mixture was vortexed HBSS.29 Subsequently, the MTT and SRB assays were performed to for 2 min and centrifuged at 3500g for 10 min. The organic phase was monitor mitochondrial activity and protein content, respectively.27,28 separated and evaporated under reduced pressure to dryness. The dry Intestinal Transport. μ fi Caco-2 cells were seeded on the apical side samples were then redissolved in 400 L of methanol and ltered of the Transwell filters of 6-well Transwell plates (0.4 μm pore μ fi μ μ through a 0.22 m PVDF lter. Then, 5 Lof10 g/mL of internal diameter, 24 mm insert, Corning Costar Co., Elscolab, Kruibeke, μ standard (6,7-dihydroxycoumarin) was added to 50 L of the sample Belgium), and after 21 days, a 100% confluent monolayer was before the injection onto a column for HPLC-DAD-single Q analysis. obtained. Intestinal monolayer integrity measurements were per- Several samples were also analyzed after 1:10 dilution in methanol to formed as previously reported.27,28 To ensure that the intestinal Caco- quantify compounds present at very high concentrations (saturated 2 cell monolayer (i) is intact during the transport assays and (ii) is not compounds). Samples (0.5 mL) from the Caco-2 transport experiment permanently damaged by the treatment, the transepithelial electrical were extracted with 0.5 mL of acetonitrile/formic acid (2%). After resistance (TEER) of the monolayer was measured before, vortexing the samples for 2 min, they were centrifuged for 15 min at immediately after, and 24 h after the transport assays. The paracellular 14,000g. The supernatants were collected, dried using nitrogen, and transport was also assessed 24 h after the transport assays using the stored at −80 °C until HPLC-MS analyses. Samples were analyzed as fluorescent paracellular transport marker Lucifer Yellow. On the basis described previously,3,20 using HPLC-DAD-MS with a reversed-phase of in-house experience,27,28 Hank’s balanced salt solution (HBSS) column. All metabolites were quantified with their standards at 305 (Gibco Life Technologies) was selected as a transport medium. Cells nm, except for EA, punicalagin, and Uro-M7 at 360 nm. These were were washed and preincubated with the transport medium (HBSS) for quantified with an EA calibration curve at 360 nm. The method 1 h. Next, cells were treated with dilutions of (i) PE-free intestinal validation previously reported3,20 was adapted regarding recovery and matrix, (ii) digested PE, and (iii) undigested PE dissolved in HBSS (2

5482 DOI: 10.1021/acs.jafc.7b02049 J. Agric. Food Chem. 2017, 65, 5480−5493 Journal of Agricultural and Food Chemistry Article mL, pH 6.5), while fresh HBSS (2.5 mL, pH 7.5) was loaded in the LOQ (Suppl. Table 1) are similar to those previously ° 20 basal compartment and incubated for 4 h (37 C and 10% CO2). reported. The polyphenol constituents of the PE and the Samples (0.5 mL) from the apical and basal compartments were hydrolysis products were detected in very small quantities obtained after 2 and 4 h of treatment. (Figure 2). EA, however, was the main metabolite observed in Statistical Analysis. All samples were analyzed in triplicate. All ± all of the gut compartments (data not shown). The PE data are expressed as the mean value SD. Statistical analyses were ff performed using SPSS V.23 for Windows (SPSS, Chicago, IL, USA). metabolites produced in the di erent SHIMEcompartments Two-way ANOVA was performed to evaluate differences between (SI, AC, TC, and DC) were analyzed (Figure 3). The analysis groups. Bonferroni posthoc test was used to investigate differences of the SI samples showed a profile in which only small amounts between stages. Statistical significance was accepted at P < 0.05. To of PE ETs and hydrolysis products were detected, while no study differences between the different stages of the experiment, two- urolithins were produced apart from trace amounts of urolithin way ANOVA with Bonferroni test was applied to the data. For the M5 (Uro-M5). In contrast, the metabolic profile in large cytotoxicity measurements, the treatments were compared using one- ff fi intestine compartments di ered qualitatively and quantitatively way ANOVA followed by a Tukey posthoc test with signi cance levels between metabotypes. UM-B produced Isouro-A and Uro-B, of p < 0.05, p < 0.01, and p < 0.001. while these metabolites were not detected in the samples from ■ RESULTS UM-A (Figures 1 and 3). Uro-A and intermediate metabolites such as the urolithin M5, M6, D, E, C, and M7 were detected in Identification of ETs and Gut Microbiota Metabolites. the large intestine compartments from both metabotypes Characteristic HPLC chromatograms of the two metabotypes (Figures 1, 2, and 3), although in some cases at very low are shown in Figure 2. The main analytical parameters of concentrations. Differences in metabolic kinetics were observed urolithin metabolites in the SHIME medium and the LOD and along large intestine compartments for both metabotypes. Urolithin concentration before and after PE administration showed an increase in urolithin production. In fact, only in the DC, Uro-A was produced from the start of PE administration for both metabotypes (Figure 3). In the AC, urolithin-A was detected in UM-B during the third week of PE treatment but not in UM-A samples. In TC, Uro-A was produced during the second week of PE treatment in the case of UM-B, while almost no urolithins were observed in UM-A even after 3 weeks of PE treatment (Figure 3). In UM-B, Isouro-A production also started during the first, second, and third week of PE administration in the case of DC, TC, and AC, respectively. In DC, Uro-B appeared during the second week of PE treatment, in parallel with the Isouro-A decrease and the Uro-A increase. After 18 days, the main urolithin detected in DC was Uro-A for both metabotypes (100% in UM-A and 88% in UM- B). Production of SCFA. Concentrations of acetate, propio- nate, butyrate, and total SCFA were higher in the distal than proximal intestine (Figure 4). Differences between metabotypes were also observed. Initially, butyrate production was higher in UM-B than in UM-A in the case of AC and TC. These differences were maintained after 3 weeks of PE administration and also observed in DC. Changes in SCFA production were already observed after 1 week. In AC, there was a reduction of total SCFA and butyrate for both metabotypes. Total SCFA and butyrate levels were, however, similar to initial levels after 3 weeks of treatment (Figure 4). An increase in propionate was detected in TC from UM-A but also in DC from both metabotypes. A decrease in acetate was observed but only in UM-A. Gut Microbiota Composition. UM-A and UM-B clustered differently in the dendrogram obtained for total bacteria (Suppl. Figure 1). Moreover, for each metabotype, the AC profiles differed from their respective TC and DC profiles that were more similar between them. PE treatment induced a shift in microbial composition for both metabotypes, while the initial Figure 2. HPLC chromatograms at 305 nm of samples from the difference between proximal (AC) and distal colon regions (TC descending colon (DC) compartment of (A) SHIME 1 (UM-A) and and DC) became less pronounced. DGGE was also performed (B) SHIME 2 (UM-B) after 3 weeks of PE administration. Numbers: ff 1, -1 ; 2, punicalagin-1; 3, punicalagin-2; 4, valoneic acid- for Clostridium cluster XIVa. While no clear di erences between both metabotypes could be observed (Suppl. Figure 1), the 2; 5, ellagic acid hexose; 6, urolithin-M5; 7, gallagic acid; 8, urolithin- fi ff D; 9, ellagic acid pentose; 10, urolithin-E; 11, ellagic acid; 12, urolithin- community pro les of the AC were again di erent from those M6; 13, urolithin C; 14, urolithin-M7; 15, m/z 245; 16, m/z 402; 17, of the TC and DC. Further, in contrast with the total bacterial isourolithin-A; 18, urolithin-A; 19, urolithin-B. profiles, it followed that the two main clusters consisted of the

5483 DOI: 10.1021/acs.jafc.7b02049 J. Agric. Food Chem. 2017, 65, 5480−5493 Journal of Agricultural and Food Chemistry Article

Figure 3. Kinetics of production of different urolithin metabolites and evolution of Gordonibacter in the different SHIME compartments of UM-A (A, C, E, and G) and UM-B (B, D, F, and H) during 3 weeks of PE administration. (A,B) Small intestine; (C,D) ascending colon; (E,F) transverse colon; (G,H) descending colon. control versus the treatment samples, revealing a more metabotypes. In contrast, levels of Firmicutes, Bacteroidetes, as pronounced PE effect on the Clostridium cluster XIVa. This well as Bacteroides and Bifidobacteria were higher in AC than in distinct PE effect also resulted in reduced differences between DC. The comparison of microbiota changes before and after proximal (AC) and distal compartments (TC and DC). the 3 weeks of PE administration demonstrated a significant (P Different microbial groups (total bacteria, Firmicutes, Bacter- < 0.05) enhanced growth of total bacteria in both metabotypes oidetes, as well as Bacteroides,Bifidobacteria, Akkermansia, lactic in TC (Table 1). A total bacteria increase was also observed in acid bacteria, and Gordonibacter spp.) were monitored by qPCR DC but only in UM-A. PE induced changes mainly in the DC analysis (Table 1). Before PE administration, higher levels of and especially in UM-A. Thus, Bacteroidetes, Bacteroides, Akkermansia and Gordonibacter were observed in the distal Akkermansia, lactic acid bacteria and Gordonibacter increased colon (TC and DC) than in the proximal colon (AC) of both immediately after starting the PE treatment while Bifidobacte-

5484 DOI: 10.1021/acs.jafc.7b02049 J. Agric. Food Chem. 2017, 65, 5480−5493 Journal of Agricultural and Food Chemistry Article

Figure 4. Short chain fatty acid production in the ascending (A,B), transverse (C,D), and descending colon (E,F) compartments of SHIME 1 (UM- A) (A,C,E) and SHIME 2 (UM-B) (B,D,F) during 3 weeks of PE administration. Results are presented as the mean ± SD. *, **, and *** correspond to significant differences at P < 0.05, P < 0.01, and P < 0.001, respectively, from the beginning of the treatment. rium decreased. In UM-B, some significant changes were also measured by the MTT and SRB assays, were investigated. MTT observed in DC but only in lactic acid bacteria and values did not significantly change when Caco-2 cells were Gordonibacter which increased (Table 1). In AC, modulation treated with undigested PE. MTT values, however, significantly of different microbial groups was also more evident in UM-A increased when Caco-2 cells were exposed to 1/5 dilutions of than in UM-B, particularly in the case of Akkermansia, lactic the PE-containing SI digest of both metabotypes compared to acid bacteria, and Gordonibacter. However, Gordonibacter those of the untreated cells. The same trend was observed for increase was much higher in DC than in AC (Table 1). A 1/5 dilutions of the PE-containing AC digest of both positive correlation between Gordonibacter and Uro-A metabotypes. For the other compartments, decreased MTT production was observed in DC (P = 0.007; r = 0.599) while values were observed when the intestinal cells were exposed to such a correlation was not observed for Akkermansia (Table 2). 1/5 dilutions of the PE-free digests, especially for metabotype However, both Gordonibacter and Akkermansia were positively B, whereas the PE-containing digests caused an increased correlated between them (P = 0.018; r = 0.549) and also with mitochondrial activity compared to that of the untreated cells propionate production. In contrast, these bacteria were not (Figure 5A). correlated with total SCFA (Table 2) or with acetate or For SRB, the intestinal cells treated with the PE-free DC butyrate (data not shown). matrix of both metabotypes had a significant decrease in Cytotoxicity Measurements. Before transport assays, the protein content. Interestingly, this possible toxic effect was cytotoxicity of the intestinal matrix and the (un)digested PE, as prevented by the presence of PE for both metabotypes, in a

5485 DOI: 10.1021/acs.jafc.7b02049 J. Agric. Food Chem. 2017, 65, 5480−5493 Journal of Agricultural and Food Chemistry Article

Table 1. Average (± SD) Abundances (log CFU/mL) of Bacterial Groups in the AC, TC, and DC Compartments of SHIME 1 a (UM-A) and SHIME 2 (UM-B) during PE Administration

treatment ascending colon transverse colon descending colon days UM-A UM-B UM-A UM-B UM-A UM-B total bacteria 0 10.27 ± 0.09 10.08 ± 0.04 10.04 ± 0.19 9.82 ± 0.14 9.71 ± 0.19 9.55 ± 0.59 7 10.17 ± 0.54 9.62 ± 0.64 10.11 ± 0.22 9.89 ± 0.09 10.09 ± 0.18 9.67 ± 0.08 21 10.30 ± 0.09 10.29 ± 0.19 10.40 ± 0.06* 10.09 ± 0.05* 10.05 ± 0.08* 9.90 ± 0.07 Firmicutes 0 9.96 ± 0.04 9.70 ± 0.06 9.81 ± 0.25 9.31 ± 0.17 9.43 ± 0.20 8.81 ± 0.37 7 9.82 ± 0.58 9.63 ± 0.26 9.73 ± 0.15 9.44 ± 0.01 9.78 ± 0.15 9.40 ± 0.04 21 9.85 ± 0.05* 9.84 ± 0.17 9.93 ± 0.10 9.53 ± 0.05 9.55 ± 0.18 9.29 ± 0.19 Bacteroidetes 0 9.72 ± 0.26 9.47 ± 0.06 9.53 ± 0.13 9.17 ± 0.33 8.93 ± 0.39 8.38 ± 0.59 7 9.56 ± 0.55 9.34 ± 0.23 9.42 ± 0.31 9.35 ± 0.12 9.56 ± 0.16* 8.73 ± 0.29 21 9.65 ± 0.18 9.72 ± 0.27 9.88 ± 0.18 9.56 ± 0.05 9.55 ± 0.17* 9.22 ± 0.15* Bacteroides 0 9.79 ± 0.25 9.37 ± 0.04 9.56 ± 0.14 9.14 ± 0.25 9.07 ± 0.24 8.53 ± 0.42 7 9.66 ± 0.51 9.19 ± 0.21 9.62 ± 0.22 9.26 ± 0.14 9.58 ± 0.14* 8.79 ± 0.40 21 9.80 ± 0.18 9.58 ± 0.25 9.93 ± 0.08* 9.45 ± 0.02 9.60 ± 0.11* 9.22 ± 0.09* lactobacilli 0 2.06 ± 0.20 1.83 ± 0.11 1.75 ± 0.71 1.58 ± 0.30 1.31 ± 0.60 1.78 ± 0.16 7 2.54 ± 0.15** 2.25 ± 0.02 2.59 ± 0.18 2.38 ± 0.40 2.49 ± 0.18* 2.44 ± 0.11** 21 2.64 ± 0.05** 4.73 ± 0.29 2.68 ± 0.16 5.22 ± 0.11** 2.70 ± 0.15* 5.11 ± 0.36** Bifidobacterium 0 9.67 ± 0.16 9.49 ± 0.22 9.47 ± 0.30 9.31 ± 0.09 9.07 ± 0.18 8.99 ± 0.17 7 8.88 ± 0.21* 9.19 ± 0.44 9.02 ± 0.03 9.17 ± 0.02 8.99 ± 0.26 9.09 ± 0.06 21 9.09 ± 0.39 9.60 ± 0.03 9.14 ± 0.23 9.34 ± 0.16 8.65 ± 0.11* 9.08 ± 0.05 Gordonibacter 0 3.37 ± 0.19 3.51 ± 0.65 4.69 ± 0.11 4.81 ± 0.03 5.26 ± 0.01 5.03 ± 0.09 7 3.44 ± 0.29 3.13 ± 0.14 3.87 ± 0.57 3.57 ± 0.36** 6.51 ± 0.19** 6.82 ± 0.01*** 21 3.89 ± 0.09* 3.79 ± 0.79 4.08 ± 0.64 6.83 ± 0.02*** 6.83 ± 0.49** 7.10 ± 0.08** Akkermansia 0 5.36 ± 0.21 5.87 ± 0.26 8.11 ± 0.48 9.02 ± 0.31 7.99 ± 0.79 8.31 ± 0.66 7 8.72 ± 1.30** 5.78 ± 0.04 9.69 ± 0.59* 9.22 ± 0.38 9.86 ± 0.61* 9.27 ± 0.28 21 9.66 ± 0.63** 6.14 ± 0.39* 10.35 ± 0.33** 9.35 ± 0.24 10.04 ± 0.16** 9.19 ± 0.32 a*, **, and ***: significant increase or decrease at P < 0.05, P < 0.01, and P < 0.001, respectively, from the beginning of treatment.

Table 2. Significant Correlations between Microbial 42% in the monolayer integrity was observed (data not shown). Metabolites Produced in Descending Colon Vessel and One day after the assays, the TEER values were partially Concentration of Gordonibacter and Akkermansia during 3 restored to initial values, with an average recovery of 77%. a Weeks with PE Administration Interestingly, the TEER values recovered to a higher extent when the Caco-2 cells were treated with PE-containing digests Gordonibacter Akkermansia (maximal decrease of 36%) compared to PE-free digests rP-value rP-value (maximal decrease of 53%) (Figure 6A). Uro-Ab For all intestinal monolayers treated with PE digests of UM-A 0.721 0.019 NS metabotype A, the Papp values of apical-to-basolateral direction − UM-B NS NS direction ranged from 2.66 ± 0.12 × 10 6 cm/s to 5.78 ± 0.094 − all 0.599 0.007 NS × 10 6 cm/s, thereby still confirming a good quality of the Total Urolithinsb monolayer.28 The apparent permeability coefficient values of UM-A 0.721 0.019 NS monolayers treated with PE-containing digests of metabotype A − fi UM-B NS NS were signi cantly higher than the Papp values of monolayers All 0.726 0.000 NS treated with PE-free digests. Propionateb Intestinal monolayers treated with the PE digest of UM-A 0.918 0.000 0.860 0.003 metabotype B, especially of the TC and DC compartments, UM-B 0.833 0.005 0.631 0.068 had higher apparent permeability coefficient values compared All 0.863 0.000 0.720 0.001 to those of metabotype A (Figure 6B). This increase in Total SCFAb paracellular transport should be taken into account while UM-A NS NS interpreting the results of the transport assays. The apparent UM-B NS NS permeability coefficient values of intestinal monolayers loaded All NS NS with PE digests of the TC and DC compartment of metabotype a fi b fi NS, not signi cant. Pearson correlation. B were signi cantly higher than the Papp values of monolayers treated with PE-free digests of these compartments. similar way to the restoration of the mitochondrial activity Transport Assays. Caco-2 cells produced phase II under these conditions (Figure 5B). conjugates, such as Uro-A-sulfate, when the monolayers were Monolayer Integrity Measurement. After 1 h of treated with PE-containing UM-A digests from the TC and DC, preincubation with HBSS, the TEER values were overall well and Uro-A-sulfate, Isouro-A-sulfate, Uro-A-glucuronide, and preserved (90 ± 12% − 112 ± 19% of initial values). After 4 h Uro-B-sulfate in the case of UM-B (Suppl. Table 2A and B). of incubation with intestinal matrix and (un)digested PE Table 3 shows the basal recoveries of various constituents when (immediately after the transport assays), a maximal decrease of Caco-2 cells were exposed to (un)digested PE extract from

5486 DOI: 10.1021/acs.jafc.7b02049 J. Agric. Food Chem. 2017, 65, 5480−5493 Journal of Agricultural and Food Chemistry Article

Figure 5. Cytotoxicity of the intestinal matrix from the SI, AC, TC, and DC compartments of SHIME 1 (UM-A) and SHIME 2 (UM-B) without and with PE administration. (A) Mitochondrial respiration (MTT) and (B) protein content (SRB) expressed as percentage compared to the untreated condition. HBSS, undigested PE, UM-A, UM-B. Error bars indicate the SD of biological replicates. Statistical difference between the PE treated and PE untreated condition. *,(p < 0.05); **,(p < 0.01); and ***,(p < 0.001).

UM-A and UM-B. Both punicalagin and EA were transported carried out in DC. These results are consistent with those across the intestinal monolayer for all conditions, except when obtained in pigs where Uro-A and Uro-B were mainly produced the Caco-2 cell monolayers were exposed to the descending at the distal parts of the intestine30 and with the conclusions colon digest of UM-A for 2. EA hexoside and EA pentoside obtained from human intervention studies, in which urolithin transport were also observed when the intestinal monolayer started to appear in plasma and was excreted in urine at was treated with the SI digest of both UMs. The transport significant concentrations, around 24 h after EA intake.4,31 assays also demonstrated for the first time that Uro-A was Therefore, TWIN-SHIME was suitable to reproduce the UMs transported across the Caco-2 monolayer, as well as Isouro-A observed in humans consuming PE.19 and Uro-B. Moreover, the intestinal cells produced (Iso)-Uro-A Figure 3 shows that there is a clear difference between the phase II metabolites (Uro-A-sulfate, Isouro-A-sulfate and Uro- gut microbiota metabolisms of UM-A and UM-B. These are A-glucuronide) that were transported to the basal side (Figure differences in the type of metabolites produced, the place where 7). they are formed, and the kinetics of their production. The results of the present study show that the production of final ■ DISCUSSION urolithin metabolites started earlier in UM-B and that its The purpose of the study was to compare the intestinal response to PE administration was faster than that of UM-A. metabolism of ETs and EA from PE in different regions of the Final urolithins were also produced in the AC in the case of human gastrointestinal tract. Production and distribution of UM-B, while they were not observed in UM-A before TC. urolithins along the intestine were previously investigated in These differences between UMs indicate that the gut vivo in Iberian pigs fed with an ET-rich diet,30 but it is microbiota present in UM-B can start the synthesis of Uro-A unknown in humans. For this purpose, fecal samples from two and Isouro-A in the AC, while the microbiota in UM-A are not individuals, with distinct UMs, were introduced in the in vitro able to produce urolithins under the ecological and TWIN-SHIME system. The pathway of urolithin formation and physiological conditions of the AC and are very active under the place in the intestinal tract where they are produced were the conditions of the DC. In the same line, a study of the elucidated. For both UMs, the highest urolithin production was production of urolithins after the intake of ETs in cattle showed

5487 DOI: 10.1021/acs.jafc.7b02049 J. Agric. Food Chem. 2017, 65, 5480−5493 Journal of Agricultural and Food Chemistry Article

Figure 6. Caco-2 cell monolayer integrity measurement in the presence of the intestinal matrix from the SI, AC, TC, and DC compartments of SHIME 1 (UM-A) and SHIME 2 (UM-B) without and with PE administration. (A) Relative transepithelial electrical resistance (TEER 24 h after treatment/TEER before treatment) × 100%. (B) Apparent permeability coefficients of lucifer yellow after 90 min. HBSS, UM-A, and UM-B. Error bars indicate the standard deviation of technical replicates. Statistical difference between the HBSS treated and the digest treated conditions. *,(p < 0.05); **,(p < 0.01); ***,(p < 0.001). Statistical difference between the PE treated and the PE untreated condition. #,(p < 0.05); ##,(p < 0.01); ###,(p < 0.001). that Isouro-A was mainly produced in the ruminal compart- after 18 days of PE administration, the UM-B profile was quite ment, while Uro-A was mainly produced in the distal similar to that of UM-A as the main urolithin was Uro-A (100% intestine.32 Interestingly, urolithins and other pomegranate- in UM-A and 88% in UM-B). In UM-B, it seems that Isouro-A derived metabolites were previously detected and quantified in is produced earlier, then decreases at the same time that Uro-B malignant and normal colonic tissues from different colon and Uro-A increase (Figure 3), supporting previous observa- regions (AC, TC, and DC) in colorectal cancer patients that 15 tions in human intervention studies that suggested that Uro-B consumed PE for around 15 days. In this study, the location is much better produced from Isouro-A than from Uro-A.16 As of urolithins in the colonic tissues was independent of both the recentlyshowninhumans,14 this study confirms that colon region and metabotype of the patients. These results do fi intermediate metabolites such as urolithins M5, M6, D, E, C, not agree with the speci c detection of Uro-A in DC from the and M7, are produced by both UMs, showing that the bacterial UM-A individual. However, this could be explained by the strains responsible for the UM-B only affect the removal of the existence of an active enterohepatic circulation of urolithins as previously observed in the pig, whose bile was reported to be hydroxyl at C-8 position, as this removal is needed for the very rich in urolithin derivatives which can re-enter the small production of Isouro-A and Uro-B. Both Uro-E and Uro-M7 intestine.30 This throws a new hypothesis by which urolithins that were only recently reported to be produced by the human 16 fi should occur from the duodenum to the rectum of individuals gut microbiota were present in signi cant amounts in both fi upon chronic PE consumption. UMs, con rming that they are relevant intermediates in EA A three-week addition of PE enhanced urolithin production catabolism. In the present study, we showed that urolithins and in the TWIN-SHIME for both UMs. This was consistent with a ET precursors are transported across the Caco-2 cell recent trial in which overweight-obese subjects consumed the monolayer, which metabolized Uro-A into both Uro-A sulfate same PE and dose (1.8 g/d).19 In the present study, PE and glucuronide as well as Isouro-A into Isouro-A sulfate, enhanced, especially, Uro-A production in both UMs. Indeed, confirming previous results.33 Sulfated and glucuronidated

5488 DOI: 10.1021/acs.jafc.7b02049 J. Agric. Food Chem. 2017, 65, 5480−5493 a Chemistry Food and Agricultural of Journal Table 3. Basal Recoveries from Intestinal Caco-2 Cells Exposed for 2 and 4 h to Undigested PE and also SI, AC, TC, and DC Digests of PE of UM-A and UM-B

basal recovery (%) undigested pomegranate extract small intestinal digest ascending colon digest transverse colon digest descending colon digest 2h 4h 2h 4h 2h 4h 2h 4h 2h 4h UM-A valoneic acid 1 ---nq------valoneic acid 2 ------gallic acid nq ------ ---nq------punicalagin 2.03 ± 1.43 1.83 ± 1.84 0.36 ± 0.18 0.76 ± 0.20 3.79 ± 2.29 3.70 ± 2.55 1.45 ± 0.83 1.66 ± 1.04 - 4.70 ± 1.66 EA 3.74 ± 2.84 3.88 ± 4.02 0.51 ± 0.23 1.05 ± 0.22 1.52 ± 1.28 1.67 ± 1.11 1.03 ± 0.91 2.53 ± 0.79 8.47 ± 1.84 13.56 ± 3.58 EA hexose nq nq 0.57 ± 0.19 1.27 ± 0.34 nq nq nq nq - - EA pentose nq nq 0.54 ± 0.18 1.25 ± 0.27 nq nq nq nq - - Uro-A ------nqnq14.35 ± 0.45 23.62 ± 0.1 Uro-A sulfate ------nqnq Uro-A glucuronide ------Isouro-A ------Isouro-A sulfate ------Uro-C ------Uro- B ------

5489 Uro-E ------Uro-M7 ------UM-B valoneic acid 1 ---nq------valoneic acid 2 ---nq------gallic acid nq - nq nq ------punicalin - - 0.03 ± 0.03 ------punicalagin 2.03 ± 1.43 1.83 ± 1.84 2.57 ± 1.87 3.23 ± 1.93 2.14 ± 0.97 1.92 ± 0.75 2.69 ± 0.83 7.04 ± 3.11 3.30 ± 1.13 3.85 ± 1.68 EA 3.74 ± 2.84 3.88 ± 4.02 3.58 ± 2.41 5.34 ± 2.83 1.57 ± 1.01 3.02 ± 0.9 8.05 ± 6.84 14.09 ± 2.64 9.64 ± 6.21 12.05 ± 11.01 EA hexose nq nq 3.38 ± 2.35 4.97 ± 2.84 nq nq - nq - - EA pentose nq nq 3.31 ± 2.33 4.66 ± 2.58 nq nq nq - - nq Uro-A ----nqnq14.52 ± 1.19 20.97 ± 0.79 12.44 ± 0.37 19.34 ± 0.54 Uro-A sulfate ------nq2.79 ± 1.03* nq 2.10 ± 0.31* Uro-A glucuronide ------1.03 ± 0.02* 1.76 ± 0.15* 1.20 ± 0.15* 1.50 ± 0.04*

.Arc odChem. Food Agric. J. Isouro-A ----nqnq8.50 ± 0.82 10.12 ± 0.42 6.67 ± 0.42 8.89 ± 0.45 Isouro-A sulfate ------5.67 ± 2.03* - 5.19 ± 1.12* Uro-B ------23.08 ± 7.74 26.92 ± 1.04 15.63 ± 0.86 25.00 ± 0.98

DOI: Uro-B sulfate ------

10.1021/acs.jafc.7b02049 Uro-C ----nq-nqnq- - 07 5 5480 65, 2017, Uro-E ------Uro-M7 ----nq-nqnq- nq

a The data presented are average values ± SD (n = 3). Basal recovery (%) = (concentration after transport)/(initial concentration) × 100. *Basal recovery (%) = (concentration of metabolite after Article

− transport)/(initial concentration of the original compound) × 100. nq: not quantified due to too low intensity. -: not detected. 5493 Journal of Agricultural and Food Chemistry Article

Figure 7. Transport of metabolites obtained from the SI, AC, TC, and DC compartments of SHIME 2 (UM-B) in the apical (A, C, E, and G) and basolateral (B, D, F, and H) compartments in Caco-2 cells. (A,B) Small intestine; (C,D) ascending colon; (E,F) transverse colon; (G,H) descending colon. urolithins were also transported back across the Caco-2 cell proportions of SCFA were observed throughout the PE monolayer which is also consistent with in vivo results.19 treatment consisting of a propionate increase for both UMs Upon treatment of these communities with the PE, an initial in DC. Low levels of butyrate and propionate bacterial decrease in SCFA production was observed in the AC, both in producers have been associated with some diseases in which UM-A and UM-B. This reduction was absent in the TC and inflammation is involved.34 According to this, the results DC (Figure 5). This observation could be explained by an obtained in this experiment indicate that the PE had a positive initial toxic effect of the PE on the microbial communities effect on gut health. Other SCFA changes by PE treatment which would only be observed in the AC since this is the region were metabotype-dependent. In general, higher changes that received the highest dose of the product. By the time that it occurred in UM-A than in UM-B. For UM-A, acetate levels reached the TC and DC, the extract had already been diluted, decreased after dosing the PE extract. and therefore, the effect was likely attenuated. After the first According to differences in urolithin and SCFA production week of treatment, the bacterial groups in the AC probably along the intestinal tract, DGGEs (total and Clostridium cluster adapted to the presence of the PE, and consequently, the initial XIVa) and qPCR showed differences in bacterial composition activity and abundances were re-established. Changes in the along the large intestine, between UMs, and modulation of gut

5490 DOI: 10.1021/acs.jafc.7b02049 J. Agric. Food Chem. 2017, 65, 5480−5493 Journal of Agricultural and Food Chemistry Article microbiota by PE administration. This microbial difference PE has anti-inflammatory activities and can reduce TNF-α between proximal (AC) and distal (TC and DC) colon regions levels in rats with chronic colitis.42 In some studies, PE also is consistent with previous observations.23 Interestingly, the in stimulated the growth of Bifidobacterium spp. and Lactobacillus vitro colon compartments derived from the two UMs contained spp.,12,36 which are known to enhance the intestinal epithelial adifferent gut microbiota as previously shown in vivo.19 These barrier function.43 In this study, however, bifidobacteria were differences might be related to the interindividual variation in not positively affected by PE. gut microbiome metabolic activity between both donors (so- The in vitro intestinal transport of punicalagin and EA is in called urolithin metabotypes; UMs). Further, the treatment agreement with previous studies that reported free EA in with the PE caused a strong long-term effect (treatment weeks human plasma upon consumption of PE,4 although intact 2 and 3) on the composition and the activity of the microbial punicalagin was not detected in the human plasma samples. communities, especially in UM-A as previously observed in However, the observed in vitro basolateral transport of humans.19 For UM-A, there was an increase in Akkermansia and punicalagin is consistent with a study showing the presence Bacteroides (together with the corresponding phylum Bacter- of punicalagin in the blood of rats fed a diet complemented oidetes). As these microbial groups are potent propionate with a high amount of punicalagin-rich pomegranate husk for producers, this nicely correlates with the shift from acetate to one month.44 propionate for this donor in DC. The correlation of propionate In general, the results obtained are consistent with the with Gordonibacter is also interesting as PE treatment previous observations in animal models and humans,18,19,32,37 modulated this bacterial group. Regarding UM-B, there is a and, therefore, support the combined TWIN-SHIME/Caco-2 specific increase of Bacteroidetes in the DC which matches with cell system as a model to evaluate the gut microbiota the specific increase of propionate in that compartment. The metabolism of pomegranate ETs in the GI tract and described relative abundance of Bacteroidetes has also been linked to the preliminary differences between UMs. Differences in microbial total fecal propionate concentration.35 composition along the intestine and between UMs determine The capacity of Gordonibacter to transform EA into different urolithin profiles. The present study has some limitations as urolithins has been reported.8 To date, Gordonibacter has only only two donors from UM-A and UM-B were compared. been quantified in human fecal samples, and its distribution However, this study contributes to elucidate the distribution of throughout the digestive tract and its role in the formation of urolithins along the digestive tract, the mechanism involved in certain urolithins are still unknown. This study confirms that their production, the specific intestinal transport of urolithins Gordonibacter abundance is increased in both UMs along the and their phase-II conjugates, and the effects on the intestine after PE administration. Furthermore, a direct composition of gut microbiota under the ecological and relationship between Gordonibacter and urolithin-A production physiological conditions at different sections of the colon. along the digestive track was observed as shown previously in vivo.9 ■ ASSOCIATED CONTENT In the present study, we also observed an increase in *S Supporting Information Bacteroidetes and especially Akkermansia, in UM-A after 21 The Supporting Information is available free of charge on the days of PE administration. This increase confirms results ACS Publications website at DOI: 10.1021/acs.jafc.7b02049. obtained in other in vitro studies that evaluated the effect of PE ff 4,36 ff Validation HPLC analysis; transport of metabolites on human gut microbiota from di erent donors. The e ect obtained from SI, AC, TC, and DC compartments of of pomegranate polyphenols on Akkermansia in human SHIME; transport of constituents of undigested intervention trials with pomegranate juice of PE has not led 19,37,38 pomegranate extract; and clustering trees of total bacteria to concluding results and deserves more study. This and Clostridium cluster XIVa (PDF) positive bacterial modulation could contribute to the beneficial health effects of pomegranate as they are inversely correlated with body weight and type- 2 diabetes.39 ■ AUTHOR INFORMATION Studies on the Caco-2 monolayer integrity showed that the Corresponding Author presence of PE significantly empowered barrier function which *Phone: +34-968396200. Fax: +34-968396213. E-mail: is stressed with colon microbial environment. In all cases, the [email protected]. permeability coefficients of the monolayers treated with PE-free ORCID digest were higher compared to Papp values of Caco-2 cells Francisco A. Tomas-Barberan: 0000-0002-0790-1739 loaded with PE-containing digests, thereby showing that the Author Contributions presence of the PE counteracted the possible toxic effects of the # R.G.-V. and H.V. contributed equally to this work. intestinal matrix for both metabotypes. This protective effect was in line with the observed higher TEER recoveries for the Funding monolayers treated with PE digests compared to monolayers This work was funded by the Projects BACCHUS (FP7-KBBE- loaded with PE-free digests. These results agree with previous 2012-6-single stage, European Commission Grant Agreement observations showing that consumption of a PE-enriched diet 312090), CICYT AGL2011-22447, AGL201564124-R (MINE- ́ caused an improvement in the intestinal mucosa barrier CO, Spain), 201370E068 (CSIC, Spain), and Fundacion ́ function of rats with dextran sodium sulfate (DSS)-induced Seneca (19900/GERM/15). H.V. is a Ph.D. Fellowship inflammation12 or obstructive jaundice.40 This protective effect Strategic Basic Researcher of the Research Foundation-Flanders could be either due to a direct effect of the PE polyphenols on (FWO). We thank the Special Research Fund of Ghent the intestinal cells, or to a change in the microbial community University (BOF 01B04212) for the funding of the TEER and metabolism caused by the presence of the PE. Epithelial equipment. barrier dysfunction is associated with inflammation and an Notes increased TNF-α production,41 and it has been observed that The authors declare no competing financial interest.

5491 DOI: 10.1021/acs.jafc.7b02049 J. Agric. Food Chem. 2017, 65, 5480−5493 Journal of Agricultural and Food Chemistry Article ■ ABBREVIATIONS USED (11) Cerda,́ B.; Soto, C.; Albaladejo, M. D.; Martínez, P.; Sanchez-́ Gascon,́ F.; Tomas-Barberá n,́ F. A.; Espín, J. C. Pomegranate juice AC, ascending colon; ATCC, American-Type Culture supplementation in chronic obstructive pulmonar disease (COPD): a Collection; DC, descending colon; DGGE, denaturing gradient 5 week randomised, double blind, placebo-controlled trial. Eur. J. Clin. ’ fi gel electrophoresis; DMEM, Dulbecco s modi ed essential Nutr. 2006, 60, 245−253. medium; DMSO, dimethyl sulfoxide; EA, ellagic acid; ETs, (12) Larrosa, M.; Gonzalez-Sarrías,́ A.; Yań̃ez-Gascon,́ M. J.; Selma, ellagitannins; FBS, fetal bovine serum; HBSS, Hank’s balanced M. V.; Azorín-Ortuño, M.; Toti, S.; Tomas-Barberá n,́ F. A.; Dolara, P.; salt solution; Isouro-A, isourolithin-A; LY, Lucifer yellow; Espín, J. C. Anti-inflammatory properties of a pomegranate extract and MTT, 3-[4,5-dimethylthiazol-2-yl]-2,5 diphenyl tetrazolium its metabolite urolithin A in a colitis rat model and the effect of colon bromide; NEAA, nonessential amino acids; PBS, phosphate inflammation on the phenolic metabolism. J. Nutr. Biochem. 2010, 21, buffered saline; PE, pomegranate extract; Pen/Strep, penicillin/ 717−725. streptomycin; SCFA, short chain fatty acids; SI, small intestine; (13) Ishimoto, H.; Shibata, M.; Myojin, Y.; Ito, H.; Sugimoto, Y.; Tai, SRB, sulphorhodamine B; TC, transverse colon; TEER, A.; Hatano, T. In vivo anti-inflammatory and antioxidant properties of transepithelial electrical resistance; UM-A, urolithin metab- ellagitannin metabolite urolithin A. Bioorg. Med. Chem. Lett. 2011, 21, 5901−4. otype A; UM-B, urolithin metabotype B; UM-0, urolithin (14) Nuń̃ez-Sanchez,́ M. A.; García-Villalba, R.; Monedero-Saiz, T.; metabotype 0; UMs, urolithin metabotypes; Uro-A, urolithin-A; García-Talavera, N. V.; Gomez-Sá nchez,́ M. B.; Sanchez-Alvarez,́ C.; Uro-B, urolithin-B; Uro-C, urolithin-C; Uro-E, urolithin-E; García-Albert, A. M.; Rodríguez-Gil, F. J.; Ruiz-Marín, M.; Pastor- Uro-M7, urolithin-M7 Quirante, F. A.; Martínez-Díaz, F.; Yań̃ez-Gascon,́ M. J.; Gonzań̃ez- Sarrías, A.; Tomas-Barberá n,́ F. A.; Espín, J. C. Targeted metabolic ■ REFERENCES profiling of pomegranate polyphenols and urolithins in plasma, urine ́ ́ and colon tissues from colorectal cancer patients. Mol. Nutr. Food Res. (1) Larrosa, M.; García-Conesa, M. T.; Espín, J. C.; Tomas-Barberan, 2014, 58, 1199−1211. F. A. Ellagitannins, Ellagic acid and Vascular Health. Mol. Aspects Med. (15) Gonzalez-Sarrías,́ A.; Gimenez-Bastida,́ J.; García-Conesa, M. T.; 2010, 31, 513−539. ́ ́ ̃ Gomez-Sanchez, M. B.; García-Talavera, N.; Gil-Izquierdo, A.; (2) García-Munoz, C.; Vaillant, F. Metabolic fate of ellagitannins: Sanchez-Alvarez,́ C.; Fontana-Compiano, L. O.; Morga-Egea, J. P.; implications for health, and research perspectives for innovative Pastor-Quirante, F. A.; Martinez-Diaz, F.; Tomas-Barberan, F. A.; functional foods. Crit. Rev. Food Sci. Nutr. 2014, 54, 1584−1598. Espin, J. C. Occurrence of urolithins, gut microbiota ellagic acid (3) García-Villalba, R.; Espín, J. C.; Aaby, K.; Alasalvar, C.; Heinonen, metabolites, and proliferation markers expresion response in human M.; Jacobs, G.; Voorspoels, S.; Koivumaki,̈ T.; Kroon, P. A.; Pelvan, E.; prostate gland upon consumption of walnuts and pomegranate juice. Saha, S.; Tomas-Barberá n,́ F. A. Validated method for the character- Mol. Nutr. Food Res. 2010, 54, 311−322. ization and quantification of extractable and non-extractable (16) García-Villalba, R.; Beltran,́ D.; Espín, J. C.; Selma, M. V.; ellagitannins after acid hydrolysis in pomegranate fruits, juices and Tomas-Barberan, F. A. Time Course Production of Urolithins from extracts. J. Agric. Food Chem. 2015, 63, 6555−6566. (4) Gonzalez-Sarrías,́ A.; García-Villalba, R.; Nuń̃ez-Sanchez,́ M. A.; Ellagic Acid by Human Gut Microbiota. J. Agric. Food Chem. 2013, 61, ́ ́ ́ 8797−8806. Tome-Carneiro, J.; Zafrilla, P.; Mulero, J.; Tomas-Barberan, F. A.; ́ Espín, J. C. Identifying the limits for ellagic acid bioavailability: A (17) Selma, M. V.; Romo-Vaquero, M.; García-Villalba, R.; Gonzalez- Crossover pharmacokinetic study in healthy volunteers after Sarrías, A.; Tomas-Barberan, F. A.; Espin, J. C. The human gut microbial ecology associated with overweight and obesity determines consumption of pomegranate extracts. J. Funct. Foods 2015, 19, − 225−235. ellagic acid metabolism. Food Funct. 2016, 7, 1769 1774. (5) Tomas-Barberá n,́ F. A.; Gonzalez-Sarrías,́ A.; García-Villalba, R.; (18) Li, Z.; Henning, S. M.; Lee, R. P.; Lu, Q. Y.; Summanen, P. H.; Nuñez-Sanchez, M. A.; Selma, M. V.; García-Conesa, M. T.; Espin, J. Thames, S.; Corbett, K.; Downes, J.; Tseng, C.-H.; Finegold, S. M.; C. Urolithins, the rescue of ‘old’ metabolites to understand a ‘new’ Heber, D. Pomegranate extract induces ellagitannin metabolite concept; metabotypes as a nexus between phenolic metabolism, formation and changes stool microbiota in healthy volunteers. Food − 2017 Funct. 2015, 6, 2487 2495. microbiota dysbiosis and host health status. Mol. Nutr. Food Res. , ́ 61, 1500901. (19) Gonzalez-Sarrías, A.; García-Villalba, R.; Romo-Vaquero, M.; ̈ ́ ́ (6) Tomas-Barberá n,́ F. A.; García-Villalba, R.; Gonzalez-Sarrías,́ A.; Alasalvar, C.; Orem, A.; Zafrilla, P.; Tomas-Barberan, F. A.; Selma, M. Selma, M. V.; Espín, J. C. Ellagic acid metabolism by human gut V.; Espín, J. C. Clustering according to urolithin metabotype explains microbiota: Consistent observation of three urolithin phenotypes in the interindividual variability in the improvement of cardiovascular risk intervention trials, independent of food source, age, and health status. biomarkers in overweight-obese individuals consuming pomegranate: J. Agric. Food Chem. 2014, 62, 6535−6538. A randomised clinical trial. Mol. Nutr. Food Res. 2017, 61, 1600830. (7) Selma, M. V.; Tomas-Barberá n,́ F. A.; Beltran,́ D.; García-Villalba, (20) García-Villalba, R.; Espin, J. C.; Tomas-Barberan, F. A. R.; Espín, J. C. Gordonibacter urolithinfaciens sp. nov., a urolithin- Chromathographic and spectroscopic characterization of urolithin for producing bacterium isolated from the human gut. Int. J. Syst. Evol. their determination in biological samples after the intake of foods Microbiol. 2014, 64, 2346−2352. containing ellagitannins and ellagic acid. J. Chromatogr. A 2016, 1428, (8) Selma, M. V.; Beltran,́ D.; García-Villalba, R.; Espín, J. C.; Tomas-́ 162−175. Barberan,́ F. A. Description of urolithin production capacity from (21) Eeckhaut, E.; Struijs, K.; Possemiers, S.; Vincken, J. P.; ellagic acid of two human intestinal Gordonibacter species. Food Funct. Keukeleire,D.D.;Verstraete,W.Metabolismofthelignan 2014, 5, 1779−1784. macromolecule into enterolignans in the gastrointestinal lumen as (9) Romo-Vaquero, M.; García-Villalba; Gonzalez-Sarrías, A.; determined in the simulator of the human intestinal microbial Beltran,́ D. R.; Tomas-Barberá n,́ F. A.; Espín, J. C.; Selma, M. V. ecosystem. J. Agric. Food Chem. 2008, 56, 4806−4812. Interindividual variability in the human metabolism of ellagic acid: (22) Possemiers, S.; Verthe,́ K.; Uytthendaele, S.; Verstraete, W. Contribution of Gordonibacter to urolithin production. J. Funct. Foods PCR-DGGE-based quantification of stability of the microbial 2015, 17, 785−791. community in a simulator of the human intestinal microbial ecosystem. (10) Cerda,́ B.; Espín, J. C.; Parra, S.; Martínez, P.; Tomas-Barberá n,́ FEMS Microbiol. Ecol. 2004, 49, 495−507. F. A. The potent in vitro antioxidant ellagitannins from pomegranate (23) Van den Abbeele, P.; Grootaert, C.; Marzorati, M.; Possemiers, juice are metabolized into bioavailable but poor antioxidant hydroxy- S.; Verstraete, W.; Gerard,́ P.; Rabot, S.; Bruneau, A.; El Aidy, S.; 6H-dibenzopyran-6-one derivatives by the colonic microflora in Derrien, M.; Zoetendal, E.; Kleerebezen, M.; Smidt, H.; Van de Wiele, healthy humans. Eur. J. Nutr. 2004, 43, 205−220. T. Microbial community development in a dynamic gut model is

5492 DOI: 10.1021/acs.jafc.7b02049 J. Agric. Food Chem. 2017, 65, 5480−5493 Journal of Agricultural and Food Chemistry Article reproducible, colon region specific and selective for Bacteroidetes and Motilva, M. J. Effect of daily intake of pomegranate juice on fecal Clostridium cluster IX. Appl. Environ. Microbiol. 2010, 76, 5237−5246. microbiota and feces metabolites from healthy volunteers. Mol. Nutr. (24) De Weirdt, R.; Possemiers, S.; Vermeulen, G.; Moerdijk- Food Res. 2015, 59, 1942−1953. Pootvliet, T. C. W.; Boshker, H. T. S.; Verstraete, W.; Van de Wiele, T. (39) Plovier, H.; Everard, A.; Druart, C.; Depommier, C.; Van Hul, Human faecal microbiota display variable patterns of glycerol M.; Geurts, L.; Chilloux, J.; Ottman, N.; Duparc, T.; Lichtenstein, L.; metabolism. FEMS Microbiol. Ecol. 2010, 74, 601−611. Myridakis, A.; Delzenne, N. M.; Klievink, J.; Bhattacharjee, A.; van der (25) Muyzer, G.; Dewaal, E. C.; Uitterlinden, A. G. Profiling of Ark, K. C. H.; Aalvink, S.; Martinez, L. O.; Dumas, M. E.; Maiter, D.; complex microbial-populations by denaturing gradient gel-electro- Loumaye, A.; Hermans, M. P.; Thissen, J. P.; Belzer, C.; de Vos, W. phoresis of polymerase chain reaction-amplified genes-coding for 16S M.; Cani, P. D. A purified membrane protein from Akkermansia ribosomal-RNA. Appl. Environment. Microbiol. 1993, 59, 695−700. muciniphila or the pasteurized bacterium improves metabolism in (26) Maukonen, J.; Mattö ̈, J.; Satokari, R.; Söderlund, H.; Mattila- obese and diabetic mice. Nat. Med. 2017, 23, 107−113. Sandholm, T.; Saarela, M. PCR DGGE and RT-PCR DGGE show (40) Gumus, M.; Tekiv, P.; Firat, V.; Onder, A.; Kapan, M.; Boyuk, diversity and short-term temporal stability in the Clostridium A.; Aldemir, M.; Kilic,̧ C. The effect of pomegranate on bacterial coccoides−Eubacterium rectale group in the human intestinal micro- translocation in rats with obstructive jaundice. Eur. Rev. Med. Pharm. biota. FEMS Microbiol. Ecol. 2006, 58, 517−28. Sci. 2013, 17, 1488−1494. (27) Kamiloglu, S.; Grootaert, C.; Capanoglu, E.; Ozkan, C.; (41) Turner, J. R. Intestinal mucosal barrier function in health and Smagghe, G.; Raes, K.; Van Camp, J. Anti-inflammatory potential of disease. Nat. Rev. Immunol. 2009, 9, 799−809. black carrot (Daucus carota L.) polyphenols in a co-culture model of (42) Rosillo, M. A.; Sanchez-Hidalgo, M.; Cardeno, A.; Aparicio- intestinal Caco-2 and endothelial EA.hy926 cells. Mol. Nutr. Food Res. Soto, M.; Sanchez-Fidalgo, S.; Villegas, I.; De la Lastra, C. A. Dietary 2017, 61, 1600455. supplementation of an ellagic acid-enriched pomegranate extract (28) Toaldo, I. M.; Van Camp, J.; Gonzales, G. B.; Kamiloglu, S.; attenuates chronic colonic inflammation in rats. Pharmacol. Res. 2012, Ordignin-Luiz, M. T.; Smagghe, G.; Raes, K.; Capanoglu, E.; 66, 235−242. Grootaert, C. Resveratrol improves TNF-alpha-induced endothelial (43) Madsen, K.; Cornish, A.; Soper, P.; NcKaigney, C.; Jijon, H.; dysfunction in a coculture model of a Caco-2 with an endothelial cell Yachimec, C.; Doyle, J.; Jewell, L.; De Simone, C. Probiotic bacteria line. J. Nutr. Biochem. 2016, 36,21−30. enhance murine and human intestinal epithelial barrier function. − (29) Possemiers, S.; Pinheiro, I.; Verhelst, A.; Van den Abbeele, P.; Gastroenterology 2001, 121, 580 591. ́ Maignien, L.; Laukens, D.; Reeves, S. G.; Robinson, L. E.; Raas, T.; (44) Cerda, B.; Llorach, R.; Ceron, J. J.; Espin, J. C.; Tomas- Schneider, Y. J.; Van de Wiele, T.; Marzorati, M. A dried yeast Barberan, F. A. Evaluation of the bioavailability and metabolism in the fermentate selectively modulates both the luminal and mucosal gut rat of punicalagin, an antioxidant polyphenol from pomegranate juice. − microbiota and protects against inflammation, as studied in an Eur. J. Nutr. 2003, 42,18 28. integrated in vitro approach. J. Agric. Food Chem. 2013, 61, 9380− 9392. (30) Espín, J. C.; Gonzalez-Barrio,́ R.; Cerda,́ B.; Lopez-Bote,́ C.; Rey, A. I.; Tomas-Barberá n,́ F. A. Iberian pig as a model to clarify obscure points in the bioavailability and metabolism of ellagitannins in humans. J. Agric. Food Chem. 2007, 55, 10476−10485. (31) Cerda,́ B.; Tomas-Barberá n,́ F. A.; Espín, J. C. ‘Metabolism of antioxidant and chemopreventive ellagitannins from strawberries, raspberries, walnuts and oak-aged wine in humans: Identification of biomarkers and individual variability’. J. Agric. Food Chem. 2005, 53, 227−235. (32) Gonzalez-Barrio,́ R.; Truchado, P.; García-Villalba, R.; Hervas,́ G.; Frutos, P.; Espín, J. C.; Tomas-Barberá n,́ F. A. Metabolism of oak leaf ellagitannins and urolithin production in beef cattle. J. Agric. Food Chem. 2012, 60, 3068−3077. (33) Gonzalez-Sarrías,́ A.; Nuń̃ez-Sanchez,́ MÁ; García-Villalba, R.; Tomas-Barberá n,́ F. A.; Espín, J. C. Antiproliferative activity of the ellagic acid-derived gut microbiota isourolithin A and comparison with its urolithin A isomer: the role of cell metabolism. Eur. J. Nutr. 2017, 56, 831−841. (34) Ríos-Covian,́ D.; Ruas-Madiedo, P.; Margolles, A.; Gueimonde, M.; De los Reyes-Gavilan,́ C. G.; Salazar, N. Intestinal short chain fatty acids and their link with diet and human health. Front. Microbiol. 2016, 7,1−9. (35) Salonen, A.; Lahti, L.; Salojarvi, J.; Holtrop, G.; Korpela, K.; Duncan, S. H.; Date, P.; Farquharson, F.; Johnstone, A. M.; Lobley, G. F.; Louis, P.; Flint, H. J.; de Vos, W. M. Impact of diet and individual variation on intestinal microbiota composition and fermentation products in obese men. ISME J. 2014, 8, 2218−2230. (36) Bialonska, D.; Ramnani, P.; Kasimsetty, S. G.; Muntha, K. R.; Gibson, G. R.; Ferreira, D. The influence of pomegranate by-product and on selected groups of human intestinal microbiota. Int. J. Food Microbiol. 2010, 140, 175−182. (37) Henning, S. M.; Summanen, P. H.; Lee, R.-P.; Yang, J.; Finegold, S. M.; Heber, D.; Li, Z. Pomegranate ellagitannins stimulate the growth of Akkermansia muciniphila in vivo. Anaerobe 2017, 43,56− 60. (38) Mosele, J. I.; Gosalbes, M. J.; Macia,̀ A.; Rubio,́ L.; Vazquez-́ Castellanos, J. F.; Jimenez-Herná ndez,́ N.; Moya, A.; Latorre, A.;

5493 DOI: 10.1021/acs.jafc.7b02049 J. Agric. Food Chem. 2017, 65, 5480−5493