Article Cite This: J. Agric. Food Chem. XXXX, XXX, XXX−XXX
pubs.acs.org/JAFC
Arabinoxylo-Oligosaccharides and Inulin Impact Inter-Individual Variation on Microbial Metabolism and Composition, Which Immunomodulates Human Cells Pieter Van den Abbeele,† Bernard Taminiau,‡ Iris Pinheiro,† Cindy Duysburgh,† Heidi Jacobs,§ Loek Pijls,§ and Massimo Marzorati*,∥ †
ProDigest bvba, Technologiepark 3, 9052 Ghent, Belgium Department of Food Science, University of Liège (ULG), Quartier Vallée 2, Avenue de Cureghem 10, 4000 Liège, Belgium § Cosucra-Groupe Warcoing S.A., Rue de la Sucrerie 1, 7740 Pecq, Belgium ∥ Center of Microbial Ecology and Technology (CMET), Ghent University, Coupure Links 653, 9000 Ghent, Belgium ‡
ABSTRACT: Fecal batch fermentations coupled to cocultures of epithelial cells and macrophages were used to compare how arabinoxylo-oligosaccharides (AXOS) and inulin modulate gut microbial activity and composition of three different human donors and subsequently the epithelial permeability and immune response. Both inulin and AXOS decreased the pH during incubation (−1.5 pH units), leading to increased productions of acetate, propionate, and butyrate. Differences in terms of metabolites production could be linked to specific microbial alterations at genus level upon inulin/AXOS supplementation (i.e., Bif idobacterium, Bacteroides, Prevotella and unclassified Erysipelotrichaceae), as shown by 16S-targeted Illumina sequencing. Both products stimulated gut barrier and immune function with increases in TEER, NF-KB, IL-10, and IL-6. Ingredients with different structures selectively modulate the microbiota of a specific donor leading to differential changes at metabolic level. The extent of this effect is donor specific and is linked to a final specific modulation of the host’s immune system. KEYWORDS: arabinoxylan, fructan, gut microbiota, Caco-2 cells, In vitro
■
macrophages).15 This in vitro approach allows us to screen for potential immune-modulatory effects of prebiotics.14,16,17 While it is commonly accepted that diet coshapes the gut microbiome,18 there is little mechanistic insight in the interactions between diet and gut microbiome effects.19 Despite the physiological relevance, especially in vivo studies often fail in providing mechanistic insight in this matter due to several reasons. First, most in vivo data are restricted to fecal samples, which only provide end-point measurements, thereby limiting the dynamic monitoring of the gut microbiome at the site of fermentation. A second drawback of in vivo approaches is the inability to only focus on the gut microbiome as microbial composition is not just influenced by the diet but also by multiple host-derived factors such as the immune system, transit time, amount of food intake, or enzyme levels. All these factors result in a large variability during in vivo studies, thus obscuring potential links between diet and the gut microbiome. Many of these limitations could be addressed by using suitable animal models or biopsies, though with relevant ethical considerations. On the other hand, in vitro approaches allow mechanistic research under highly controlled conditions. However, as the inoculum for such studies is specific from human individuals, the interindividual variability in microbiome composition in in vitro studies is equally important. These
INTRODUCTION Intestinal micro-organisms may play several roles in human health. In the past decade, potential improvement of health by modulation of the human gut microbiota has gained a lot of attention. One of the strategies is the use of prebiotics, which are nondigestible substrates that are selectively used by the host micro-organisms conferring a health benefit.1 One of the most frequently studied prebiotics is inulin, a fructan-type carbohydrate. Fructan-type carbohydrates can be easily fermented by the intestinal microbiota, resulting in a predominantly proximal fermentation and attributed health effects.2−5 However, because many colonic diseases occur in the distal colon,6 other classes of prebiotics have been investigated including arabinoxylooligosaccharides (AXOS). This type of carbohydrate is potentially degraded more distally due to their complex chemical structure.7−9 Next to improvement of intestinal function itself, dietary nondigestible carbohydrates may also have systemic effects and modulate the immune system in both animals and humans.10−12 These nondigestible carbohydrates might stimulate the immune system via cytokine expression. Little research is conducted on the direct effects on cytokine expression, but many studies have assessed indirect effects through stimulation of beneficial micro-organisms and their metabolism in the gut.13 Recently, a technology platform was described which allows studying host-microbiota interactions in the gut.14 It combines microbial fermentation samples with a coculture human cell model, where intestinal epithelial cells (Caco-2 cells) are combined with immune cells (THP1 © XXXX American Chemical Society
Received: October 4, 2017 Revised: January 3, 2018 Accepted: January 8, 2018
A
DOI: 10.1021/acs.jafc.7b04611 J. Agric. Food Chem. XXXX, XXX, XXX−XXX
Article
Journal of Agricultural and Food Chemistry
SYBR FAST qPCR Kit (KapaBiosystems, Wilmington, U.S.A.) before normalization, pooling, and sequencing on a MiSeq sequencer using v3 reagents (ILLUMINA, U.S.A.). Positive control using DNA from 20 defined bacterial species and a negative control (from the PCR step) were included in the sequencing run. Sequence reads processing was used as previously described using respectively MOTHUR software package v1.35,28 and UCHIME algorithm29 for alignment and clustering and chimera detection. Clustering distance of 0.03 was used for OTU generation. 16S Reference alignment and taxonomical assignation were based upon the SILVA database (v1.28) of full-length 16S rDNA sequences.30 From 2 190 612 raw reads, we obtained 2 032 295 reads after cleaning (length and sequence quality). We retained 10 000 reads per sample as a subsampling process which were further screened for chimeric contaminants. Finally, 161 277 reads were used for OTU clustering and taxonomic assignment. Good’s coverage estimator was used as a measure of sampling effort for each sample, with a mean value of 92.73%. Effects on Host Cells in Coculture Model of Epithelial Cells and Macrophages. On the final time point (48 h), samples were collected for incubation studies with human epithelial cells and macrophages. The coculture experiment was performed as previously described.17 Briefly, Caco-2 cells (HTB-37; American Type Culture Collection, LGC Promochem, Molsheim, France) were seeded in 24well semipermeable inserts (0.4 μm pore size) at a density of 1 × 105 cells/insert. Caco-2 cell monolayers were cultured for 14 days until a functional cell monolayer with a Transepithelial electrical resistance (TEER) of more than 300 Ω.cm2 was obtained. Two days before coculture, THP1-Blue cells (InvivoGen, Toulouse, France) were seeded in 24-well plates at a density of 5 × 105 cells/well and treated with 100 ng/mL of phorbol 12-myristate 13-acetate (PMA) for 48 h to induce differentiation into macrophage-like cells. Then, the Caco-2bearing inserts were placed on top of the PMA-differentiated THP1Blue cells for further experiments. The apical compartment was filled with sterile-filtered (0.22 μm) samples from the fecal batch incubations (diluted 1:5 (v/v) in Caco-2 complete media (Dulbecco’s modified eagle medium containing 4.5 g/L glucose and supplemented with 20% heat inactivated fetal bovine serum (Gibco, Invitrogen), 10 mM HEPES, and 1X Antibiotic-Antimycotic)). Sodium butyrate (8 mM) was used as positive control. The basolateral compartment was filled with Caco-2 complete media. Cells were also exposed to Caco-2 complete media in both chambers as control. Cells were treated for 24 h, after which the TEER was measured. All 24-h values were normalized to its own 0-h value and are presented as percentage of initial value. Then, the basolateral supernatant was discarded, and cells were stimulated basolaterally with Caco-2 complete media containing 500 ng/mL of ultrapure lipopolysaccharide (LPS Escherichia coli K12, InvivoGen). Cells were also stimulated basolaterally with LPS and 1 μM hydrocortisone and media without LPS as controls. After 6 h of LPS stimulation, the basolateral supernatant was collected for cytokines measurement (human IL-6 and IL-10 by ELISA (eBioscience, Vienna, Austria)) and for assessing NF-κB activity by the QUANTI-Blue assay (InvivoGen), according to the manufacturers’ instructions. On the apical side, cellular activity was measured by using the WST-1 assay (Roche, Vilvoorde, Belgium). Briefly, the stable tetrazolium salt WST-1 was added on the apical side, where metabolically active cells reduced it to a soluble formazan, which was measured spectrophotometrically and correlated directly with cell viability. All treatments were done in triplicate. Statistics. Comparison of normally distributed data of the different test conditions (BNC, FIB, and AXOS) on microbial metabolic markers was performed with a Student’s t test. Differences were significant if p < 0.05. To evaluate the difference between BNC, FIB, and AXOS samples on host end points, a one-way Anova with Dunnett’s posthoc test was performed. Statistics were performed using GraphPad Prism version 7.00 for Windows (GraphPad Software, San Diego, U.S.A.). Principal component analysis (PCA) was performed using Analyzeit (v4.51) software. For each microbial metabolic marker (pH, gas production, SCFA), the increase or decrease from 0 h to 24 h and
interindividual differences in microbial composition can stimulate different microbial metabolic pathways when a given dietary component is provided.20,21 In order to elucidate the aforementioned microbial changes in sufficient detail, one heavily relies on molecular detection methods,22−24 and among them, high-throughput 16S rRNA gene sequencing can be used to look at changes at high phylogenetic resolution.25,26 In this study, the main objective was to evaluate changes in microbial activity and composition, gut barrier function and cytokine expression in response to inulin/AXOS fermentation, with the focus on interindividual variability. For this purpose, we made use of a standardized in vitro setup for fermentation of dietary compounds, coupled to cocultures of Caco-2 cells and THP1 macrophages and microbial analysis via 16S rRNA targeted Illumina.
■
MATERIALS AND METHODS
Chemicals and Carbohydrates. All chemicals were obtained from Sigma-Aldrich (Overijse, Belgium) unless stated otherwise. Cosucra Groupe Warcoing S.A. (Pecq, Belgium) provided: inulin (Fibruline instant - FIB) with a purity of 92% and a degree of polymerization between 3 and 60 with an average of 10; an AXOS-rich extract, containing 88% dietary fiber, of which 66% AXOS, a degree of substitution of 0.38 and an average degree of polymerization of 6. Description of In Vitro Fecal Batch Incubations. Fecal batch fermentations were performed for three different test conditions (blank control = BNC, FIB and AXOS) against three different human gut microbiota (donors 1, 2, and 3). Briefly, 63 mL of colonic background medium (K2HPO4 3.5 g/L; KH2PO4 10.9 g/L; NaHCO3 2 g/L; yeast extract 2 g/L; peptone 2 g/L; mucin 1 g/L; cysteine 0.5 g/L; Tween80 2 mL/L) was added to 120 mL penicillin bottles, already containing the correct amount of the test products (i.e. 0 mg or 350 mg) for obtaining a final concentration of 0 g/L (BNC) or 5 g/ L (for both FIB and AXOS), respectively. The bottles were sealed with rubber stoppers and anaerobiosis was obtained by flushing with N2. Subsequently, a human fecal inoculum was prepared for the different healthy volunteers (donor 1 = m, 31yr; donor 2 = m, 25 yr; donor 3 = m, 29 yr) by making a 1:5 (mass:volume) mixture of a freshly collected fecal sample with anaerobic phosphate buffer (K2HPO4 8.8 g/L; KH2PO4 6.8 g/L; sodium thioglycolate 0.1 g/L; sodium dithionite 0.015 g/L). After homogenization (10 min, BagMixer 400, Interscience, Louvain-LaNeuve, Belgium) and removal of big particles via centrifugation (2 min, 500g), 7 mL of inoculum was added to the different bottles. At that point, the actual incubation started for a period of 48 h during which temperature was controlled at 37 °C and continuous mixing was ensured by a shaker (90 rpm). All experiments were performed in triplicate. Microbial Metabolic Activity: pH, Gas Production, and Short-Chain Fatty Acids (SCFA). pH (Senseline F410; ProSense, Oosterhout, The Netherlands), gas pressure (Hand-held pressure indicator CPH6200; Wika, Echt, The Netherlands), and SCFA were measured at the start of the incubation, and after 24 and 48 h. Shortchain fatty acids (SCFA) were measured as described by De Weirdt et al.27 Microbial Community Analysis: 16S rRNA Gene Sequencing. On the final time point (48 h), samples were collected for microbial community analysis. PCR-amplification of the V1−V3 region of the 16S rDNA and library preparation were performed with the following primers (with Illumina overhand adapters), forward (5′-GAGAGTTTGATYMTGGCTCAG-3′) and reverse (5′-ACCGCGGCTGCTGGCAC-3′). Each PCR product was purified with the Agencourt AMPure XP beads kit (Beckman Coulter, Pasadena, U.S.A.) and submitted to a second PCR round for indexing, using the Nextera XT index primers 1 and 2. After purification, PCR products were quantified using the Quant-IT PicoGreen (ThermoFisher Scientific, Waltham, U.S.A.) and diluted to 10 ng/μL. A final quantification, by qPCR, of each sample in the library was performed using the KAPA B
DOI: 10.1021/acs.jafc.7b04611 J. Agric. Food Chem. XXXX, XXX, XXX−XXX
Article
Journal of Agricultural and Food Chemistry Table 1. Overall Metabolic Activity in Terms of Acidification and Gas Productiona
a Average absolute pH (± SD) on different time points (0, 24, and 48 h) and corresponding increase in gas pressure (kPa) between 0−24 h and 24− 48 h, upon fermentation of 5 g/L AXOS or FIB by a fecal microbiota of three different human donors (1, 2, and 3), versus their respective blank controls (BNC) (n = 3). For optimal observation of consistent effects over the different donors, also the average of the three donors (“Donor 1/2/ 3”) is presented (n = 9). Significant differences between the BNC, FIB and AXOS are indicated with a different letter (p < 0.05), with the highest values within each donor being shaded with a darker background.
Figure 1. Microbial metabolic activity in terms of SCFA production. (A) Average acetate, (B) propionate, (C) butyrat, and (D) branched SCFA production (mM) during the initial the 0−2 4h and 24−48 h time interval, upon fermentation of 5 g/L AXOS or FIB by a human fecal microbiota of three different donors (1, 2, and 3), versus their respective blank controls (BNC) (n = 3). For optimal observation of consistent effects over the different donors tested, also the average of the three donors (“Donor 1/2/3”) is presented. Data is presented as mean ± SD. Significant differences between the BNC, FIB, and AXOS are indicated with a different letter (p < 0.05); with the differences in the 0−24 h time interval indicated in or below the bar and in the 24−48 h time interval above the bar.
■
from 24 h to 48 h of incubation was used to create a joint PCA biplot. For host end points, the data obtained from the coculture experiments (TEER, NF-κB activity and cytokines) were used to create a joint PCA biplot. Statistical differences of bacterial population relative abundance were compared between multiple groups by 2-way ANOVA followed by Tukey-Kramer posthoc test. The analysis was performed again between each pair of groups by multiple t tests with Holm−Sidak False Discovery Rate using GraphPad Prism software (version 7.00, GraphPad Software, San Diego, U.S.A.). The analysis was first performed with the 3 donors grouped and then each donor separately. Differences were considered as statistically significant if p < 0.05.
RESULTS
Overall Microbial Metabolic Activity in Terms of Acidification and Gas Production. The overall degree of acidification and gas production are markers for the intensity of bacterial metabolism of FIB and AXOS (Table 1 and SI Figure 1). First, for all three donors, the decrease in pH and gas production with FIB and AXOS were more pronounced versus control incubations (BNC). Both test ingredients mainly affected pH and gas production during the first 24 h of incubation with (i) an average pH decrease from 6.7 to 5.5 for FIB and to 5.6 for AXOS (versus 6.7 to 6.6 for BNC), and (ii) C
DOI: 10.1021/acs.jafc.7b04611 J. Agric. Food Chem. XXXX, XXX, XXX−XXX
Article
Journal of Agricultural and Food Chemistry
Figure 2. Microbial community composition as assessed via 16S-targeted illumina sequencing. Results are based on microbiota analysis at phylum level, upon a 48 h fermentation of 5 g/L AXOS or FIB by a human fecal microbiota of three different donors (1, 2, and 3), versus their respective blank controls (BNC) (n = 3). For optimal observation of consistent effects over the different donors tested, the average of the three donors is presented (n = 9). Significant differences between the BNC, FIB, and AXOS are indicated with a different letter (p < 0.05).
Table 2. Microbial Community Composition As Assessed via 16S-targeted Illumina Sequencing, at Phylum Levela
a
Absolute average abundance (%) (± SD) of different microbial phyla upon a 48 h fermentation of 5 g/L AXOS or FIB by a human fecal microbiota of three different donors (1, 2, and 3), versus their respective blank controls (BNC) (n = 3). Significant differences between the BNC, FIB, and AXOS are indicated with a different letter (p < 0.05), with the highest values within each donor being shaded with a darker background.
donors 2 and 3 was noted between 24 and 48 h. Further, FIB increased butyrate levels after 48 h significantly more than AXOS did, over the three donors tested (p = 0.0003). Besides being lower (in contrast to FIB), butyrate production upon AXOS fermentation was donor-specific, with butyrate increasing less upon supplementation to the microbiota of donor 2. Finally, branched SCFA strongly decreased for both FIB and AXOS for the three donors tested (Figure 1D), with significantly lower levels for FIB as compared to AXOS after 48 h (p = 0.007). This was attributed to a slight increase in branched SCFA levels between 24 and 48h upon AXOS (versus FIB) supplementation, which was most apparent for donor 2. PCA of the metabolic data (pH, gas, and SCFA) revealed that both FIB and AXOS cluster very differently than BNC (Figure 4A). Nevertheless, FIB and AXOS still cluster together within donors. This donor-dependent effect is more pronounced for donor 1, while donors 2 and 3 are more similar regarding their FIB fermentation. Microbial Composition via 16S rRNA Gene Sequencing. Sequencing analysis compared the impact of FIB and AXOS on the abundances of different microbial groups upon fermentation of the test substrates. At phylum level (Figure 2 and Table 2), FIB and AXOS increased Actinobacteria while decreasing Lentisphaerae. FIB and especially AXOS induced a shift from Firmicutes to Bacteroidetes, mostly for donor 1 and 2,
an average increase in gas pressure of 86 kPa for FIB and 84 kPa for AXOS (versus 20 kPa for BNC). With respect to pH, the microbiota of both donor 2 and 3 reincreased the pH from 24 h to 48 h upon FIB or AXOS supplementation. Other donor-specific findings were that gas production with FIB and AXOS was overall much higher for donor 1, whereas for donor 3, a marked gas consumption was observed in the 24−48 h time interval. Microbial Metabolic Activity in Terms of SCFA Production. For the three donors, FIB and AXOS increased acetate, propionate, and butyrate levels while decreasing branched SCFA levels (Figure 1A/B/C/D). Acetate was similarly increased for FIB and AXOS for all three donors. It was mainly produced during the 0−24 h interval, except for donor 3 where a significant amount of acetate was still produced between 24 and 48 h (i.e., 8.1 mM and 9.5 mM for FIB and AXOS, respectively; Figure 1A). Propionate was similarly increased for FIB over the three donors tested, with the main increase taking place from 0 to 24 h (Figure 1B). In contrast, AXOS fermentation was characterized by donorspecific effects, with propionate being much more strongly increased upon supplementation to microbiota of donor 2. Butyrate levels were similarly increased for FIB for the three different donors (Figure 1C). While the main butyrate increase occurred between 0 and 24 h for donor 1, the main increase for D
DOI: 10.1021/acs.jafc.7b04611 J. Agric. Food Chem. XXXX, XXX, XXX−XXX
Article
Journal of Agricultural and Food Chemistry Table 3. Microbial Community Composition As Assessed via 16S-targeted Illumina Sequencing, at Genus Levela
a
Absolute average abundance (%) of different microbial genera upon a 48 h fermentation of 5 g/L AXOS or FIB by a human fecal microbiota of three different donors (1, 2, and 3), versus their respective blank controls (BNC) (n = 3). For optimal observation of consistent effects over the different donors, also the average of the three donors (“Donor 1/2/3”) is presented (n = 9). Significant differences between the BNC, FIB, and AXOS are indicated (p < 0.05), with the highest values within each donor being shaded with a darker background.
as was the decrease of Verrucomicrobia with FIB and AXOS. Finally, Proteobacteria responded in a donor-dependent manner, but mostly decreased upon FIB and AXOS supplementation (except for FIB in donor 2 and AXOS in donor 3). At family and genus level, changes in microbial groups are more meaningful as they relate to smaller groups often
possessing specific functionalities. We therefore focused on the phylogenetic level with highest resolution (Table 3). A first set of genera were those that increased under nutrientdepleted conditions (i.e., the BNC). The following genera consistently decreased upon FIB/AXOS supplementation for all three donors: Anaerotruncus, Flavonif ractor, Lachnospira, and Parasutterella. Other genera that correlated to the BNC of one E
DOI: 10.1021/acs.jafc.7b04611 J. Agric. Food Chem. XXXX, XXX, XXX−XXX
Article
Journal of Agricultural and Food Chemistry
Figure 3. Effect of fermentation-derived metabolites on barrier function and immunomodulation in vitro. (A) TEER (% of initial value) as a measure of barrier function of Caco-2 cells, (B) NF-κB activity of THP1-Blue cells and (C,D) basolateral cytokine levels (IL-10 and IL-6) on the cocultures that were treated with fermentation-derived metabolites collected 48 h after the start of incubation of 5 g/L FIB or AXOS in the presence of fecal microbiota of three different human donors (1, 2, and 3), versus their respective blank control (BNC) (n = 3). For optimal observation of consistent effects over the different donors tested, also the average of the three donors is presented. Data is presented as mean ± SEM ($), ($$), ($$$), and ($$ $$) represent statistically significant differences between the different test conditions with p < 0.05, p < 0.01, p < 0.001, and p < 0.0001, respectively.
remarkable were the increase of Bacteroides from 13% in the BNC up to 25% for FIB and 43% for AXOS. Both test ingredients also increased the abundance of Oscillospira, while decreasing levels of unclassified Ruminococcaceae and Subdoligranulum. Further, FIB specifically increased Escherichia-Shigella (from 19% to 30%), Blautia (from
