Figures
Abstract
In vitro gut modeling provides a useful platform for a fast and reproducible assessment of treatment-related changes. Currently, pig intestinal fermentation models are mainly batch models with important inherent limitations. In this study we developed a novel in vitro continuous fermentation model, mimicking the porcine proximal colon, which we validated during 54 days of fermentation. This model, based on our recent PolyFermS design, allows comparing different treatment effects on the same microbiota. It is composed of a first-stage inoculum reactor seeded with immobilized fecal swine microbiota and used to constantly inoculate (10% v/v) five second-stage reactors, with all reactors fed with fresh nutritive chyme medium and set to mimic the swine proximal colon. Reactor effluents were analyzed for metabolite concentrations and bacterial composition by HPLC and quantitative PCR, and microbial diversity was assessed by 454 pyrosequencing. The novel PolyFermS featured stable microbial composition, diversity and metabolite production, consistent with bacterial activity reported for swine proximal colon in vivo. The constant inoculation provided by the inoculum reactor generated reproducible microbial ecosystems in all second-stage reactors, allowing the simultaneous investigation and direct comparison of different treatments on the same porcine gut microbiota. Our data demonstrate the unique features of this novel PolyFermS design for the swine proximal colon. The model provides a tool for efficient, reproducible and cost-effective screening of environmental factors, such as dietary additives, on pig colonic fermentation.
Citation: Tanner SA, Zihler Berner A, Rigozzi E, Grattepanche F, Chassard C, Lacroix C (2014) In Vitro Continuous Fermentation Model (PolyFermS) of the Swine Proximal Colon for Simultaneous Testing on the Same Gut Microbiota. PLoS ONE 9(4): e94123. https://doi.org/10.1371/journal.pone.0094123
Editor: Markus M. Heimesaat, Charité, Campus Benjamin Franklin, Germany
Received: January 13, 2014; Accepted: March 13, 2014; Published: April 7, 2014
Copyright: © 2014 Tanner et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: This study was carried out thanks to the financial support of the Commission of Technology and Innovation of Switzerland (CTI-Project Nr. 11962.1 PFLS-LS) (URL: http://www.kti.admin.ch/?lang=en). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Introduction
The pig gut microbiome is a complex ecosystem, dominated by members of the phyla Firmicutes and Bacteroidetes [1]. The vast quantity and diversity of the gut bacterial community provides the host with a large set of metabolic functions and is considered to play a key role in host health and disease [2]. Diet is a principal factor shaping the gut bacterial composition and functionality [3], thereby also impacting overall animal health. This in turn largely determines productivity and efficiency of swine livestock production. Given the complexity of this interplay, it is of paramount importance to precisely evaluate the effects of specific feed ingredients or additives on composition and functionality of the gut microbiome and further elucidate the role and function of the gut bacterial ecosystem in animal health to improve productivity and efficiency of swine livestock production.
It is obvious that animal testing is one of the most prominent strategies to predict effectiveness and impact of dietary additives on the gut microbiota, but ethical concerns and costs can restrict these applications [2]. Intestinal in vitro models are able to partly evade these restrictions by enabling reproducible experimentation under standardized conditions, and more importantly, giving the yet host-uncoupled opportunity to investigate the complexity of gut microbiomes and the functional relatedness of specific bacterial species [4].
To date, most porcine in vitro models are using simple batch cultures with the aim to examine the fermentation capacity of intestinal ecosystems on a given substrate, using the cumulative gas production technique [5]–[8]. However, batch fermentations are limited in terms of experimental duration and the amount of substrate supply to avoid negative feedback mechanisms [9]. Batch cultures are also highly dependent on the inoculation density as it directly impacts microbial growth in these closed systems [4]. In contrast, continuous culture systems are superior in modeling the dynamic nature of the gastrointestinal tract, allowing the adaptation of various parameters, including dilution rate, retention time, pH and temperature, to meet and maintain optimal growth conditions [9]. Substrate replenishment and toxic waste removal, further, is continuous and facilitates studies on the modulation of microbial composition and activity [4]. So far, only few porcine semi-continuous or continuous intestinal fermentation models inoculated with feces or cecal content have been described, focusing on changes in gut bacterial communities tested over limited fermentation periods [10], the inhibition of Salmonella by medium-chain fatty acids [11] or the effect of live yeast on fermentation parameters [12]. Long-term continuous in vitro models of the swine gut are still lacking, likely due to the difficulty of generating highly stable free-cell suspension fermentations, inoculated with fecal extract, over a long experimental period.
To overcome possible drawbacks associated with free-cell suspension cultures, such as limited stability and washout of less dominant or slow growing bacterial species, fecal microbiota was immobilized in polysaccharide gel beads to set up stable continuous intestinal fermentation models [13], [14]. The major benefits shown for immobilized fecal microbiota models include high cell density, maintenance of bacterial diversity and high stability over extended fermentation periods, tested for up to 71 days [4], [15]. Besides stability requirements, reproducibility and parallel testing of treatments with the same gut microbiota are of major importance for gut in vitro research, but difficult to apply with classical continuous models inoculated with fecal microbiota. We have recently set up and validated a novel PolyFermS model of the child proximal colon in a two-stage system for parallel testing of treatments on the same microbiota [16]. In this model, a first-stage reactor containing immobilized fecal microbiota and operated with conditions mimicking the first section of the proximal colon was used to constantly inoculate up to four second-stage reactors, operated with conditions mimicking the remaining proximal colon. Our data demonstrated that this PolyFermS model produced reproducible and stable intestinal microbiota over a 38 days test period [16]. The microbial diversity of reactor effluents tested with the HITChip phylogenetic array was comparable to the feces of the healthy donor, whereas a high response to pH was demonstrated. However, this study used a two-stage model for upper and lower sections of the proximal colon, shown to be highly dependent on pH in the first reactor, which had to be arbitrary set at 5.5 to reach a metabolic balance for the child microbiota.
In the present study we aimed to enhance the original design of the PolyFermS model validated with child microbiota [16], and to adapt to a new model of the swine proximal colon. The novel PolyFermS model consisted of two stages; a first stage for the inoculum reactor (IR) seeded with immobilized swine fecal microbiota and used to inoculate at 10% (v/v) five parallel second-stage reactors, which were also fed with 90% fresh chyme medium. Each reactor was operated under identical conditions, selected to mimic swine proximal colon, and allowing the parallel testing of multiple treatments on the same gut microbiota. The stability of the complex intestinal microbiota in the multiple reactors was monitored over 54 days of fermentation by analyzing its composition (qPCR and 454 pyrosequencing) and metabolic activity (HPLC).
Material and Methods
Ethical statement
No specific permits were obtained for the collection of the fecal sample. The animal was not harmed during fecal sample collection and oral consent for sample collection was obtained from the owner of the farm.
Feces collection and immobilization
Feces from a healthy 5 month old sow (80 kg), raised under farming conditions and not subjected to any antibiotic treatment for the last 3 months, were collected in a sterile 50 mL Falcon tube. Anaerobiosis was maintained using anaerobic gas pack systems (Oxoid AnaeroGen TM, Oxoid AG, Basel Switzerland) during transport to the laboratory and until immobilization was performed. The entire immobilization procedure was carried out under anaerobic conditions (anaerobic chamber; Coy Laboratories, Ann Arbor, MI, USA). Briefly, a 20% (w/v) suspension of feces in pre-reduced peptone water (0.1%, pH 7) was prepared, homogenized and further immobilized in a polymer solution consisting of gellan gum (2.5%, w/v), xanthan (0.25%, w/v) and sodium citrate (0.2%, w/v) for production of 1–2 mm diameter gel beads using a two phase dispersion process as described previously [13], [16].
Nutritive medium
The nutritive medium described by Macfarlane et al. [17] was modified for its carbohydrate and protein concentration to more closely mimic the ileal chyme of a swine (Table S1), using a similar approach as described previously [15]. For calculation of ingredient concentrations, a standard cornstarch based diet with corn (641 g kg−1) as main carbohydrate and soybean meal (331 g kg−1) as main N-source was used [18]. Digestibility indices of 97% for cornstarch [19] and 82.5% for soybean meal [20] were applied while considering a 2 kg/day feed intake per pig, resulting in a cornstarch:N-compound (soybean meal) ratio of 25:75. The amount of cornstarch and soy peptone supplied daily to the model was calculated by applying a scale factor of 0.09 to account for the actual volume of the proximal colon in vivo (approx. 2.9 L [21] compared to the proximal reactor volume of the model (260 mL). The final concentration of the two compounds in the chyme medium was estimated for a mean retention time of 9 hours in the reactors, giving a daily chyme medium supply per reactor of 693 mL. To avoid excess of nitrogen compounds, the soy peptone concentration was fixed to 13 g L−1 resulting in a carbohydrate (cornstarch) concentration of 4.3 g L−1. Yeast extract and mucin, mimicking the contribution of endogenous secretion [13], and the non-starch polysaccharides (pectin, xylan, arabinogalactan and guar gum) were excluded from this calculation.
A volume of 0.5 mL L−1 of a filter-sterilized (Minisart pore size 0.2 μm, Sartorius, VWR International AG, Dietikon, Switzerland) vitamin solution described by Michel et al. [22] was added to the sterilized medium (20 min, 120°C). All components of the nutritive medium were purchased from Sigma-Aldrich Chemie (Buchs, Switzerland), except for soy peptone (Labo-Life Sàrl, Pully, Switzerland), yeast extract (I2CNS GmbH, Urdorf, Switzerland) and KH2PO4 (VWR International AG).
Experimental set-up of the PolyFermS model
The continuous fermentation was carried out for 54 days using a two-stage design with a total of six reactors (Sixfors, Ismatec, Glattbrugg, Switzerland) (Figure 1). Each reactor was aimed to simulate conditions of the swine proximal colon fermentation.
IR: inoculum reactor, containing immobilized swine feces (30% v/v); CR: control reactor; TR1-TR4: test reactors 1-4; M: fresh nutritive medium supply; S: effluent sampling; F: flow rate; Stab: stabilization period; T: treatment period; W: wash period
The inoculum reactor (234 mL) was seeded with 30% (v/v) swine fecal beads and connected via a peristaltic pump (Masterflex L/S, Fisher Scientific SA, Wohlen, Switzerland) to one control reactor (CR) and four test reactors (TR1-4) operated in parallel. IR was supplied with 100% fresh nutritive medium (26 mL h−1) whereas the second-stage reactors CR and TR1-4 were continuously supplied with 90% (26 mL h−1) fresh nutritive medium and 10% (2.9 mL h−1) effluent from IR for continuous inoculation. The remaining effluent (50%) from IR was discarded. The inoculation rate of CR and TR1-4 was accurately controlled using an in-house designed distributor device equipped with valves regulated chronometrically.
Fermentation procedure
The PolyFermS model was run under conditions of the swine proximal colon with a controlled constant pH of 6.0 through addition of 2.5 M NaOH. The mean retention time was fixed at 9 hours, while temperature was maintained at 38°C (being in the middle of the range previously applied in porcine in vitro models) [7], [10], [11], [23]–[25]. Anaerobic conditions were ensured by constantly flushing the headspace of the reactors with CO2 and constant stirring was performed at 120 rpm. During the first 72 hours, IR was operated in batch mode to colonize the fecal beads and the nutritive medium was replaced by fresh medium every 12 hours. After colonization, continuous operation (26 mL h−1) in IR was started, followed by a stabilization period of 5 days before connection to CR and TR1-4. The entire two-stage system was then stabilized for another 5 days.
The 54 days continuous fermentation was split into stabilization, treatment and washing periods (Figure 1). IR and CR were operated with constant conditions to assess the temporal stability of the system and were not subjected to any manipulation during the entire fermentation. In addition, CR served as a control reactor for TR1-4 that were subjected to different parallel treatment periods (data not shown).
Between two treatment periods, TR1-4 were subjected to a washing procedure with 10% chlorine to kill microbes and remove any historical effect of the previous periods. Briefly, TR1-4 were disconnected from IR, the entire medium was removed and reactors were filled with 10% freshly prepared chlorine solution. After stirring for one hour, the reactors were rinsed twice by adding sterile bidistilled water and stirring for another hour. After complete removal of water and chlorine residues, reactors were filled with sterile fresh nutritive medium and reconnected to IR. Thereafter, the system was allowed to stabilize for 3 days until reaching steady state before starting the next treatment period.
Effluent samples of all reactors were collected daily. HPLC samples were processed immediately whereas samples for DNA extraction were stored at −80°C.
HPLC analysis for metabolite determination
Short-chain fatty acids (SCFA; acetate, propionate, butyrate, valerate, formate, iso-butyrate and iso-valerate) as well as lactate concentrations in fermentation effluent samples from all reactors were determined by HPLC analysis (Thermo Fisher Scientific Inc. Accela, Wohlen, Switzerland). Briefly, effluent samples were centrifuged (14 000 g) for 10 min at 4°C. The pellet was used for DNA extraction while the supernatant was diluted 1∶10 with ultrapure water and filtered directly into vials through a 4 mm HPLC filter with a 0.45 μm nylon membrane (Infochroma AG, Zug, Switzerland). The analysis was run at a flow rate of 0.4 mL min−1 using an Aminex HPX-87H column (Bio-Rad Laboratories AG, Reinach, Switzerland) and 10 mM H2SO4 as eluent. Mean metabolite concentrations were calculated from duplicate analyses and expressed in mM.
DNA extraction and qPCR analyses
Genomic DNA was extracted from effluent samples using the FastDNA SPIN Kit for soil (MP Biomedicals, Illkirch, France). DNA extracts were subjected to quantitative real-time PCR (qPCR) for enumeration of specific bacterial target groups comprising total bacteria, Bifidobacterium spp., Bacteroides-Prevotella group, Enterobacteriaceae, Lactobacillus/Pediococcus/Leuconostoc spp. and Clostridium Cluster IV (Table 1). In addition, the Succinivibrio dextrinosolvens group was quantified to verify results obtained by 454 pyrosequencing. Standard curves for each target group were prepared as described previously [26]. All assays, were performed using the 2 x SYBR Green PCR Master Mix (Applied Biosystems, Zug, Switzerland) in a 25 μl volume and an ABI PRISM 7500-PCR sequence detection system (Applied Biosystems).
454 pyrosequencing
Selected samples were analyzed by 454 pyrosequencing for their microbial 16S rRNA based composition profile. IR and CR (days 19/20, 30/31, 42/43 and 52/53) samples were selected to assess bacterial diversity and temporal stability of the model as well as to compare the microbial composition to the fecal inoculum. To demonstrate the re-establishment of the microbiota after washing, day 25 (last day of 3rd stabilization period) was chosen as a representative day for all reactors. For IR and CR stability samples, effluent from 2 consecutive days were pooled at a ratio of 1∶1. DNA was extracted with the FastDNA Spin Kit for soil (MP Biomedicals) and sent to DNAVision SA (Charleroi, Belgium) for 454 pyrosequencing analysis and subsequent taxonomic assignment of 16S rRNA gene reads. 454 pyrosequencing was performed using a 454 Life Science system combined with Titanium Chemistry (Roche) as described previously [27]. The complete 454 pyrosequencing dataset has been deposited to the National Center for Biotechnology Information (NCBI) Sequence Read Archive (SRA) under accession number SRP034540.
Statistical analysis
All statistical analyses were performed using PASW Statistics for Windows version 18.0 (SPSS Inc., Chicago). qPCR data were log10-transformed and expressed as means ± SD of the last three days of each stabilization period. To assess reproducibility of the microbial composition in CR and TR1-4 prior to a treatment period qPCR data were subjected to the non-parametric Mann-Whitney U test with exact significance and a p-value <0.05 was considered significant. The same test was used to verify results obtained from 454 pyrosequencing for Succinivibrionaceae. Comparisons were made between qPCR data from before (d11-37) and after (d38-54) appearance in CR. To assess temporal metabolite stability in IR and CR, linear regression of total SCFA, acetate, butyrate, propionate, iso-valerate, valerate and iso-butyrate concentrations versus time were calculated over the time period d11-54 and difference from 0 of slope coefficients was tested using the t-test (P<0.05).
Results
Microbial activity by HPLC
To assess the microbial activity and temporal stability of the novel PolyFermS model for the pig proximal colon during 54 days continuous fermentation, daily effluents for each reactor were analyzed using HPLC. IR and CR were operated with constant conditions and not subjected to any treatment or washing period during the entire fermentation. Therefore, the microbial activity in these two reactors was used to assess the metabolic stability of the two-stage model.
After an initial stabilization period of 10 days for reaching pseudo-steady state conditions, high and stable metabolic activity was measured in IR and CR throughout the fermentation (Figure 2). In IR the mean total SCFA concentration (d11-54) was 180.1±5.8 mM with mean values for acetate, propionate and butyrate of 102.3±7.3 mM, 45.1±4.4 mM and 20.4±1.2 mM, respectively. Iso-valerate was produced at 5.9±0.5 mM. In contrast, valerate and iso-butyrate were not detected until day 13 but were measured at concentrations of 4.5±1.3 mM and 1.9±0.6 mM over d13-54, respectively. Lactate and formate, were not detected throughout the fermentation. For CR similar mean total SCFA (174.4±7.6 mM) as well as main SCFA acetate (92.9±7.6 mM), propionate (47.1±3.8mM) and butyrate (23.1±2.4 mM) concentrations were detected. Detection of minor metabolites, iso-valerate, valerate and iso-butyrate, was delayed for 9, 14 and 13 days in CR, respectively, resulting in mean values of 5.7±1.2 mM (d11-54), 4.2±1.6 mM (d14-54) and 1.8±0.9 mM (d13-54).
Initial stabilization: stabilization period in continuous mode to reach pseudo steady-state; closed symbol: IR; open symbol: CR; (▴, Δ) total SCFA, (•, ○) acetate, (♦, ◊) propionate, and (▪, □) butyrate.
To test the effect of culture time on metabolite concentrations, linear regressions of daily concentration data (d11-54) versus time were calculated. Highly significant time effects (P<0.001) were determined for most measured metabolites in IR and CR, except for the total SCFA concentration (Table 2). Acetate concentrations in IR and CR (P<0.001) and butyrate in CR (P<0.05) decreased significantly over time whereas propionate, butyrate in IR, iso-valerate, valerate and iso-butyrate showed significant (P<0.001) concentration increases. Corresponding slope coefficients remained small (-0.493 mM/day – 0.265 mM/day) indicating a moderate time effect on tested metabolites.
Microbial composition by qPCR
Microbial composition of daily reactor effluents was assessed by analyzing the 16S rRNA gene copy numbers of bacterial target groups using qPCR. IR and CR data were used to assess the time stability of the model.
Total 16S rRNA gene copy numbers mL−1 effluent were approximately 1.6 log10 units higher in the fecal inoculum compared to the reactor effluents from IR and CR (Table 3). The fecal inoculum was dominated by the Bacteroides-Prevotella group and the Lactobacillus/Pediococcus/Leuconostoc spp. group, followed by Clostridium Cluster IV, Enterobacteriaceae and Bifidobacterium spp. In reactor effluents from IR and CR, stable copy numbers of the targeted bacterial groups were recorded during continuous fermentation (d11-54) after an initial stabilization as indicated by low standard deviations of the mean. Similar to the fecal inoculum, Bacteroides-Prevotella group was predominant in IR and CR effluents while Bifidobacterium spp. was the least abundant group. Enterobacteriaceae were approximately 1.8 log10 higher in reactor effluents compared to the fecal inoculum and displayed equally high copy numbers as for the Clostridium Cluster IV. In contrast, the Lactobacillus/Pediococcus/Leuconostoc spp. population was detected at ca. 3 log10 lower copy numbers in reactor effluents during the stabilized period (d11-54) compared to the fecal inoculum.
Microbial diversity by 454 Pyrosequencing
Microbial diversity and composition analyses of the fecal inoculum and selected samples from IR and CR (days 19/20, 30/31, 42/43, 52/53) were performed by 454 pyrosequencing. Sequences were aligned with the RDP classifier v 2.1 using a confidence cutoff level of 80%. After quality check the number of reads per sample was decreased from 11661±2758 to 7762±4585 and mean read length per sample was 256±2 base pairs (bp).
Relative abundance detected by 454 pyrosequencing revealed the predominance of three major phyla in all samples tested (Figure S1A). The fecal inoculum was predominated by the phylum Firmicutes whereas the Bacteroidetes phylum was most abundant in reactor effluents, followed by the Firmicutes and Proteobacteria phyla, except for samples 19/20 in IR and 30/31 in CR where the two phyla Bacteroidetes and Firmicutes were almost equally abundant. The phylum Proteobacteria was increasing from less than 1% relative abundance in the fecal inoculum to up to 29% in sample 52/53 from CR. At family level (Figure 3), the highest abundance in IR was recorded for Prevotellaceae (33–66%), Lachnospiraceae (7–17%), Ruminococcaceae (5–14%) and Enterobacteriaceae (4–8%), with unclassified reads accounting for 10–21% of total reads. The same pattern of relative abundances on family level was observed in CR, with Prevotellaceae (38–51%), Lachnospiraceae (13–18%), Ruminococcaceae (5–13%) and Enterobacteriaceae (5–9%) being the most abundant families in all samples and unclassified bacteria accounting for 10–20% of the reads. The detected families were composed of the predominant genera Prevotella, Escherichia/Shigella, Ruminococcus, Roseburia, Blautia, Bacteroides and Oscillibacter (Figure S1B). Remarkably, the family Succinivibrionaceae increased from 0.1% on day 19/20 in IR to up to 8% at the end of the fermentation (52/53). In CR the family Succinivibrionaceae even accounted for as much as 21% of total reads on day 52/53. In contrast, reads assigned to the family Ruminococcaceae decreased in CR from 10% on day 42/43 to 4% on day 52/53. This observation was confirmed by qPCR with specific primers targeting the 16S rRNA gene of S. dextrinosolvens. Similarly, the 16S rRNA gene copy numbers for the S. dextrinosolvens group in CR significantly increased (P<0.001) from 7.47±0.2 log10 copies mL−1 (mean ± SD; d11-37) to 9.2±0.3 log10 copies mL−1 (mean ± SD; d38-54).
The relative abundance on family level is shown. Values <1% are summarized in the group "others".
In the fecal inoculum, the family Clostridiaceae (26%) was most abundant and predominantly represented by the genus Clostridium, followed by Ruminococcaceae and Lactobacillaceae (both 13%) and Lachnospiraceae (8%). The unclassified bacteria accounted for 25% of total reads. Enterobacteriaceae were not detected in the fecal inoculum.
Reproducibility of microbial composition and activity
The novel PolyFermS model was designed to allow reproducible testing of different treatments in parallel test reactors compared to a control reactor inoculated with the same microbiota. Consecutive multiple treatment periods could be tested by subjecting the test reactors to a washing procedure using 10% chlorine and followed by a stabilization period to regenerate comparable microbiota in CR and all test reactors (Figure 1). Therefore, to test reproducibility of control and test reactors after a washing procedure, bacterial composition (qPCR data) and activity (HPLC data) in TR1-4 after three day re-stabilization were compared with data measured in CR. No significant differences of bacterial composition of CR and TR1-4 (P<0.05) were detected using the Mann-Whitney U test (Table S2). Moreover the analysis of effluent samples of CR and TR1-4 by 454 pyrosequencing on day 25 (last day of 3rd stabilization period) showed similar microbiota composition (Figure S2). Metabolite concentrations in TR1-4 progressively re-established at similar levels to that in CR, reaching comparable values three days after restarting the system (Figures 4 and S3). High and stable bacterial concentrations were measured already after one day re-stabilization as indicated by total 16S rRNA gene copy numbers in TR1-4 over three days stabilization periods.
(◊) total SCFA; (○) acetate; (□) propionate; (Δ) butyrate; (black) CR; (blue) TR1; (red) TR2; (green) TR3; (pink) TR4.
Discussion
The large field of research to evaluate the impact of feed ingredients and additives on the gastrointestinal health and development of pigs [28] emphasizes the need for potent porcine in vitro fermentation models to study feed-related impacts on gut microbial community and functionality in a fast and reproducible setting.
In this study we tested a novel porcine in vitro fermentation model for the proximal colon, advanced from the PolyFermS model presented by Zihler Berner et al. [16]. The porcine in vitro model was designed to generate self-contained parallel fermentations in each of the reactors, simulating the proximal colon, which is the primary site of fermentation in the gastrointestinal tract [29]. The first PolyFermS model validated by Zihler Berner et al. [16] included an inoculum reactor operated under conditions of the first section of the proximal colon which was used to continuously feed control and test reactors with fermented effluent, operated with conditions selected to mimic the distal section of the proximal colon. Therefore in this model a combination of two stages, IR and CR or TR, was used to model the proximal colon. In contrast, test reactors in the porcine PolyFermS tested in this study were operated in conditions of the proximal colon and constantly inoculated with only 10% fecal microbiota produced in IR, while 90% of the feed was fresh medium to simulate the dynamic process of chyme inflow. This design is likely more suitable for studying the fate of dietary treatments on the fermentative capacity of the gut microbial community, since the complete response of the proximal colon microbiota, constantly supplied with fresh substrates from the small intestine, is of interest. In our study, the combination of the reactor set-up and the nutritive medium, adapted to simulate swine chyme, allowed the establishment of six self-contained parallel fermentations for the porcine proximal colon with a high metabolic and compositional stability, diversity and reproducibility throughout 54 days of fermentation.
The stable metabolite concentrations obtained during the continuous fermentation, indicate balanced microbial growth and the maintenance of gut microbiota functional capacity. In addition, metabolite ratios for IR and CR (61∶27∶12 and 57∶29∶14, acetate:propionate:butyrate) after initial stabilization were very similar and also in agreement with values reported for the pig proximal colon (60∶25∶15 [30]) and for fecal SCFA ratios (63∶25∶12 [31]) in vivo. Small but significant time effects were recorded for all detected metabolites, except total SCFA concentrations that remained constant during the entire stabilized fermentation. The ability to detect even small time changes is directly related to the sensitivity of the analysis, permitted by continuous operation of a stabilized fermentation system and achieved with a high number of time points analyzed [32].
The slight but significant increase of propionate concentrations during continuous fermentation may directly be related to the observed increase of the family Succinivibrionaceae, which was already detected in the fecal inoculum but at a low relative abundance of 0.1% by 454 pyrosequencing. Succinivibrionaceae belong to the γ-subclass of the phylum Proteobacteria and play an important role in starch digestion in sheep and cattle [33]. While Kim et al. [34] assigned the genus Succinivibrio to the group of the less abundant genera in the swine fecal microbiome, other studies in contrast have grouped the genus as a member of the core microbiota of the porcine proximal colon [35] or porcine cecum [36]. Different carbohydrate sources can be metabolized by Succinivibrionaceae resulting in the main fermentation products acetate and succinate [33]. Further decarboxylation of succinate can lead to propionate [37], likely due to cross-feeding reactions in the complex intestinal environment.
Microbial composition and diversity determined in the model effluents by qPCR and 454 pyrosequencing showed no major changes in the bacterial groups between days 11 and 54 in both, IR and CR, after the initial stabilization period of 10 days. The establishment of a microbial pseudo steady-state is an important factor to gain reliable data on the modulating potential of a specific treatment in order to avoid false positive conclusions related to the microbiota adaptation to in vitro conditions [38]. Compared to the fecal inoculum all bacterial groups targeted by qPCR were reduced in reactor effluents from IR and CR, except for Enterobacteriaceae that displayed higher copy numbers. Changes in microbiota composition and diversity may reflect the transfer from in vivo (feces) to in vitro (proximal colon) conditions, the adaptation to a new environment, which depends on the conditions in the host during collection of the fecal sample, and the lack of host effects in vitro [16], [39]. Additionally, the shift in microbial composition may favor more robust species due to a competitive advantage in adaptation and may open niches, possibly occupied by Enterobacteriaceae, which can explain their increase. Using 454 pyrosequencing, Clostridiaceae exhibited the most remarkable decrease in relative abundance between the fecal inoculum and reactor effluents, whereas Prevotellaceae increased markedly in reactor effluents. The high prevalence of Prevotellaceae was directly linked to an increased Bacteroidetes and decreased Firmicutes ratio in reactor effluents compared to the fecal inoculum. The high occurrence of Bacteroidetes in intestinal in vitro models, has already been reported previously in the Twin-SHIME [39] and TNO intestinal model [40] and may be the result of the higher micromolar levels of oxygen and the less adhesive capacity of Bacteroidetes.
Studies on the swine fecal and cecal microbiome reported Firmicutes, Bacteroidetes and Proteobacteria as the most prevalent phyla in swine [1], [34], [36], [41]–[43], which is in accordance with our study. Firmicutes and Bacteroidetes represented between 80–90% of assigned reads in all effluent samples, except for CR on day 52/53, when Proteobacteria increased to 29% and Firmicutes/Bacteroidetes decreased to 70%, respectively. On family level, Prevotellaceae was the most abundant family detected in all reactor samples, which is in accordance with other studies [1], [35], [36], [41], [44] and supported by our qPCR data. On genus level the predominant genera in the porcine PolyFermS (mean relative abundance in IR and CR >1%) were Prevotella, Escherichia/Shigella, Roseburia, Ruminococcus, Oscillibacter, Succinivibrio, Blautia and Bacteroides, genera that have previously been described as members of the cecal [36], [43], [45] and fecal [1], [44] gut microbiota. We thus conclude that the microbial composition is representative for the porcine proximal colon, as demonstrated by the high accordance of bacterial genera between our in vitro study and previous in vivo studies on the cecal porcine microbiota. Furthermore, a dominant fraction of glycan degraders was identified, consisting of the predominant genera Prevotella, Ruminococcus and Roseburia. The high prevalence of fibrolytic bacteria and their associated carbohydrate utilization systems in swine has already been reported previously [1], [46] and is possibly related to the high amount of complex polysaccharides found in the pig diet. Starch degraders play a key role in gut microbial ecosystems as they represent the top of the trophic chain by providing simple sugars from the breakdown of complex carbohydrates, thus directly rendering them accessible to other members of the microbial community [47].
Assessing the reproducibility of in vitro gut fermentation models is a permanent challenge [4] and is difficult to achieve with classical models. In this study, we used a constant 10% inoculation rate from IR to the subsequent reactors, which allowed reproducing similar and parallel evolving self-contained ecosystems in IR and the second-stage reactors. Due to the consecutive treatment periods with in-between washing of the test reactors, a fast re-establishment of reproducible environment in the test reactors was required. This is of particular importance for studying and more importantly comparing treatment-related responses of the microbial community in vitro, which presumes comparable experimental conditions in the different reactors. While the bacterial groups in TR1-4 targeted by qPCR reached similar numbers to that of CR after only one day of re-stabilization, complete metabolic activity was recovered after a longer time of approximately three days. Such delay of functionality response has also been reported with the Twine-SHIME model [39]. The short re-stabilization period of PolyFermS compared to 5–8 days for Twin-SHIME points on the benefits of immobilized fecal microbiota to provide bacterial stability and diversity [4].
To conclude, in the present study we validated a novel PolyFermS continuous intestinal fermentation model of the swine proximal colon, inoculated with immobilized fecal microbiota. This model operated with a nutritive medium designed to mimic pig chyme allowed to stably reproduce the microbiota and metabolic activity of swine proximal colon for at least 54 days. The particular model set-up allows comparing different treatments and a control, run with the same inoculated microbiota, simultaneously. Furthermore, our data demonstrate a considerable interplay between functionality and taxonomic composition and highlight the stringent potential of the model for compositional as well as functionality related studies. This in vitro gut model can further be expanded to simulate multiple stages of the large intestine (proximal, transverse, distal) and a number of consecutive treatment periods. It should be particularly suitable to accurately investigate the effects of dietary factors such as pro-and prebiotics, as well as environmental parameters or drugs on the porcine gut microbiota in highly controlled settings.
Supporting Information
Figure S1.
Microbial composition in the fecal inoculum (FI), IR and CR measured by 454 pyrosequencing on (A) phylum level and (B) genus level. Values <1% are summarized in the group “others”.
https://doi.org/10.1371/journal.pone.0094123.s001
(TIF)
Figure S2.
Microbial composition on family level in CR and TR1-4 on day 25 (last day of 3rd stabilization period) measured by 454 pyrosequencing. Values <1% are summarized in the group “others”.
https://doi.org/10.1371/journal.pone.0094123.s002
(TIF)
Figure S3.
Mean main metabolite concentrations and total 16S rRNA gene copy numbers from TR1-4 during the last three days of each stabilization period. Data are depicted as mean values ± SD from TR1-4. (♦) total 16S rRNA gene copies mL−1 effluent; (○) acetate; (□) propionate; (Δ) butyrate.
https://doi.org/10.1371/journal.pone.0094123.s003
(TIF)
Table S1.
Composition of the nutritive medium simulating the swine ileal chyme.
https://doi.org/10.1371/journal.pone.0094123.s004
(PDF)
Table S2.
16S rRNA gene copy numbers (log10 mL−1 of effluent) of specific bacterial groups measured by qPCR in effluent samples from CR and TR1-4.
https://doi.org/10.1371/journal.pone.0094123.s005
(PDF)
Acknowledgments
We kindly thank the Genetic Diversity Center (GDC), ETH Zurich, for providing equipment for quantitative real-time PCR analysis and Eun-Hee Doo, Lukas Meile and Simon Galenda for technical assistance.
Author Contributions
Conceived and designed the experiments: SAT AZB FG CC CL. Performed the experiments: SAT AZB ER. Analyzed the data: SAT AZB ER CC CL. Wrote the paper: SAT AZB CC CL.
References
- 1. Lamendella R, Domingo JW, Ghosh S, Martinson J, Oerther DB (2011) Comparative fecal metagenomics unveils unique functional capacity of the swine gut. BMC Microbiol 11: 103.
- 2. Macfarlane GT, Macfarlane S (2007) Models for intestinal fermentation: association between food components, delivery systems, bioavailability and functional interactions in the gut. Curr Opin Biotechnol 18: 156–162.
- 3. Russell WR, Hoyles L, Flint HJ, Dumas ME (2013) Colonic bacterial metabolites and human health. Curr Opin Microbiol 16: 246–254.
- 4. Payne AN, Zihler A, Chassard C, Lacroix C (2012) Advances and perspectives in in vitro human gut fermentation modeling. Trends Biotechnol 30: 17–25.
- 5. Jha R, Bindelle J, Van Kessel A, Leterme P (2011) In vitro fibre fermentation of feed ingredients with varying fermentable carbohydrate and protein levels and protein synthesis by colonic bacteria isolated from pigs. Anim Feed Sci Technol 165: 191–200.
- 6. Jonathan MC, van den Borne JJGC, van Wiechen P, da Silva CS, Schols HA, et al. (2012) In vitro fermentation of 12 dietary fibres by faecal inoculum from pigs and humans. Food Chem 133: 889–897.
- 7. Lin B, Gong J, Wang Q, Cui S, Yu H, et al. (2011) In-vitro assessment of the effects of dietary fibers on microbial fermentation and communities from large intestinal digesta of pigs. Food Hydrocolloid 25: 180–188.
- 8. Zhu WY, Williams BA, Konstantinov SR, Tamminga S, De Vos WM, et al. (2003) Analysis of 16S rDNA reveals bacterial shift during in vitro fermentation of fermentable carbohydrate using piglet faeces as inoculum. Anaerobe 9: 175–180.
- 9. Williams BA, Bosch MW, Awati A, Konstantinov SR, Smidt H, et al. (2005) In vitro assessment of gastrointestinal tract (GIT) fermentation in pigs: Fermentable substrates and microbial activity. Anim Res 54: 191–201.
- 10. Ricca DM, Ziemer CJ, Kerr BJ (2010) Changes in bacterial communities from swine feces during continuous culture with starch. Anaerobe 16: 516–521.
- 11. Messens W, Goris J, Dierick N, Herman L, Heyndrickx M (2010) Inhibition of Salmonella typhimurium by medium-chain fatty acids in an in vitro simulation of the porcine cecum. Vet Microbiol 141: 73–80.
- 12. Pinloche E, Williams M, D'Inca R, Auclair E, Newbold CJ (2012) Use of a colon simulation technique to assess the effect of live yeast on fermentation parameters and microbiota of the colon of pig. J Anim Sci 90: 353–355.
- 13. Cinquin C, Le Blay G, Fliss I, Lacroix C (2004) Immobilization of infant fecal microbiota and utilization in an in vitro colonic fermentation model. Microb Ecol 48: 128–138.
- 14. Cinquin C, Le Blay G, Fliss I, Lacroix C (2006) New three-stage in vitro model for infant colonic fermentation with immobilized fecal microbiota. FEMS Microbiol Ecol 57: 324–336.
- 15. Le Blay G, Chassard C, Baltzer S, Lacroix C (2010) Set up of a new in vitro model to study dietary fructans fermentation in formula-fed babies. Br J Nutr 103: 403–411.
- 16. Zihler Berner A, Fuentes S, Dostal A, Payne AN, Vazquez Gutierrez P, et al. (2013) Novel Polyfermentor Intestinal Model (PolyFermS) for controlled ecological studies: validation and effect of pH. PLoS One 8: e77772.
- 17. Macfarlane GT, Macfarlane S, Gibson GR (1998) Validation of a three-stage compound continuous culture system for investigating the effect of retention time on the ecology and metabolism of bacteria in the human colon. Microb Ecol 35: 180–187.
- 18. Nyannor EK, Williams P, Bedford MR, Adeola O (2007) Corn expressing an Escherichia coli-derived phytase gene: a proof-of-concept nutritional study in pigs. J Anim Sci 85: 1946–1952.
- 19. Lee TT, Huang YF, Chiang CC, Chung TK, Chiou PW, et al. (2011) Starch characteristics and their influences on in vitro and pig prececal starch digestion. J Agric Food Chem 59: 7353–7359.
- 20. Wang JP, Hong SM, Yan L, Cho JH, Lee HS, et al. (2011) The evaluation of soybean meals from 3 major soybean-producing countries on productive performance and feeding value of pig diets. J Anim Sci 89: 2768–2773.
- 21. Kararli TT (1995) Comparison of the gastrointestinal anatomy, physiology, and biochemistry of humans and commonly used laboratory animals. Biopharm Drug Dispos 16: 351–380.
- 22. Michel C, Kravtchenko TP, David A, Gueneau S, Kozlowski F, et al. (1998) In vitro prebiotic effects of Acacia gums onto the human intestinal microbiota depends on both botanical origin and environmental pH. Anaerobe 4: 257–266.
- 23. Sakata T, Kojima T, Fujieda M, Takahashi M, Michibata T (2003) Influences of probiotic bacteria on organic acid production by pig caecal bacteria in vitro. Proc Nutr Soc 62: 73–80.
- 24. Bindelle J, Buldgen A, Boudry C, Leterme P (2007) Effect of inoculum and pepsin-pancreatin hydrolysis on fibre fermentation mecasured by the gas production technique in pigs. Anim Feed Sci Technol 132: 111–122.
- 25. von Heimendahl E, Breves G, Abel HJ (2010) Fiber-related digestive processes in three different breeds of pigs. J Anim Sci 88: 972–981.
- 26. Dostal A, Fehlbaum S, Chassard C, Zimmermann MB, Lacroix C (2013) Low iron availability in continuous in vitro colonic fermentations induces strong dysbiosis of the child gut microbial consortium and a decrease in main metabolites. FEMS Microbiol Ecol 83: 161–175.
- 27. Jost T, Lacroix C, Braegger CP, Chassard C (2012) New insights in gut microbiota establishment in healthy breast fed neonates. PLoS One 7: e44595.
- 28. Pluske JR (2013) Feed- and feed additives-related aspects of gut health and development in weanling pigs. J Anim Sci Biotechnol 4: 1.
- 29. Guarner F, Malagelada JR (2003) Gut flora in health and disease. Lancet 361: 512–519.
- 30.
Grieshop CM, Reese DE, Fahey Jr GC (2000) Nonstarch polysaccharides and oligosaccharides in swine nutrition. In: Lewis AJ and Southern LL, editors. Swine Nutrition. Boca Raton, FL: CRC Press. pp. 107–130.
- 31. Bird AR, Vuaran M, Crittenden R, Hayakawa T, Playne MJ, et al. (2009) Comparative effects of a high-amylose starch and a fructooligosaccharide on fecal bifidobacteria numbers and short-chain fatty acids in pigs fed Bifidobacterium animalis. Dig Dis Sci 54: 947–954.
- 32. Mozzetti V, Grattepanche F, Moine D, Berger B, Rezzonico E, et al. (2012) Transcriptome analysis and physiology of Bifidobacterium longum NCC2705 cells under continuous culture conditions. Benef Microbes 3: 261–272.
- 33. Hippe H, Hagelstein A, Kramer I, Swiderski J, Stackebrandt E (1999) Phylogenetic analysis of Formivibrio citricus, Propionivibrio dicarboxylicus, Anaerobiospirillum thomasii, Succinimonas amylolytica and Succinivibrio dextrinosolvens and proposal of Succinivibrionaceae fam. nov. Int J Syst Bacteriol 49 Pt 2: 779–782.
- 34. Kim HB, Borewicz K, White BA, Singer RS, Sreevatsan S, et al. (2012) Microbial shifts in the swine distal gut in response to the treatment with antimicrobial growth promoter, tylosin. Proc Natl Acad Sci U S A 109: 15485–15490.
- 35. Li RW, Wu S, Li W, Navarro K, Couch RD, et al. (2012) Alterations in the porcine colon microbiota induced by the gastrointestinal nematode Trichuris suis. Infect Immun 80: 2150–2157.
- 36. Buzoianu SG, Walsh MC, Rea MC, O'Sullivan O, Cotter PD, et al. (2012) High-throughput sequence-based analysis of the intestinal microbiota of weanling pigs fed genetically modified MON810 maize expressing Bacillus thuringiensis Cry1Ab (Bt maize) for 31 days. Appl Environ Microbiol 78: 4217–4224.
- 37. Hosseini E, Grootaert C, Verstraete W, Van de Wiele T (2011) Propionate as a health-promoting microbial metabolite in the human gut. Nutr Rev 69: 245–258.
- 38. Possemiers S, Verthe K, Uyttendaele S, Verstraete W (2004) PCR-DGGE-based quantification of stability of the microbial community in a simulator of the human intestinal microbial ecosystem. FEMS Microbiol Ecol 49: 495–507.
- 39. Van den Abbeele P, Grootaert C, Marzorati M, Possemiers S, Verstraete W, et al. (2010) Microbial community development in a dynamic gut model is reproducible, colon region specific, and selective for Bacteroidetes and Clostridium cluster IX. Appl Environ Microbiol 76: 5237–5246.
- 40. Rajilic-Stojanovic M, Maathuis A, Heilig HG, Venema K, de Vos WM, et al. (2010) Evaluating the microbial diversity of an in vitro model of the human large intestine by phylogenetic microarray analysis. Microbiology 156: 3270–3281.
- 41. Looft T, Johnson TA, Allen HK, Bayles DO, Alt DP, et al. (2012) In-feed antibiotic effects on the swine intestinal microbiome. Proc Natl Acad Sci U S A 109: 1691–1696.
- 42. Riboulet-Bisson E, Sturme MH, Jeffery IB, O'Donnell MM, Neville BA, et al. (2012) Effect of Lactobacillus salivarius bacteriocin Abp118 on the mouse and pig intestinal microbiota. PLoS One 7: e31113.
- 43. Poroyko V, White JR, Wang M, Donovan S, Alverdy J, et al. (2010) Gut microbial gene expression in mother-fed and formula-fed piglets. PLoS One 5: e12459.
- 44. Kim HB, Borewicz K, White BA, Singer RS, Sreevatsan S, et al. (2011) Longitudinal investigation of the age-related bacterial diversity in the feces of commercial pigs. Vet Microbiol 153: 124–133.
- 45. Haenen D, Zhang J, da Silva CS, Bosch G, van der Meer IM, et al. (2013) A diet high in resistant starch modulates microbiota composition, SCFA concentrations, and gene expression in pig intestine. J Nutr 143: 274–283.
- 46. Leser TD, Amenuvor JZ, Jensen TK, Lindecrona RH, Boye M, et al. (2002) Culture-independent analysis of gut bacteria: the pig gastrointestinal tract microbiota revisited. Appl Environ Microbiol 68: 673–690.
- 47. Chassard C, Lacroix C (2013) Carbohydrates and the human gut microbiota. Curr Opin Clin Nutr Metab Care 16: 453–460.
- 48. Guo X, Xia X, Tang R, Zhou J, Zhao H, et al. (2008) Development of a real-time PCR method for Firmicutes and Bacteroidetes in faeces and its application to quantify intestinal population of obese and lean pigs. Lett Appl Microbiol 47: 367–373.
- 49. Ramirez-Farias C, Slezak K, Fuller Z, Duncan A, Holtrop G, et al. (2009) Effect of inulin on the human gut microbiota: stimulation of Bifidobacterium adolescentis and Faecalibacterium prausnitzii. Br J Nutr 101: 541–550.
- 50. Furet JP, Firmesse O, Gourmelon M, Bridonneau C, Tap J, et al. (2009) Comparative assessment of human and farm animal faecal microbiota using real-time quantitative PCR. FEMS Microbiol Ecol 68: 351–362.
- 51. Bartosch S, Fite A, Macfarlane GT, McMurdo ME (2004) Characterization of bacterial communities in feces from healthy elderly volunteers and hospitalized elderly patients by using real-time PCR and effects of antibiotic treatment on the fecal microbiota. Appl Environ Microbiol 70: 3575–3581.
- 52. Cleusix V, Lacroix C, Dasen G, Leo M, Le Blay G (2010) Comparative study of a new quantitative real-time PCR targeting the xylulose-5-phosphate/fructose-6-phosphate phosphoketolase bifidobacterial gene (xfp) in faecal samples with two fluorescence in situ hybridization methods. J Appl Microbiol 108: 181–193.
- 53. Stevenson DM, Weimer PJ (2007) Dominance of Prevotella and low abundance of classical ruminal bacterial species in the bovine rumen revealed by relative quantification real-time PCR. Appl Microbiol Biotechnol 75: 165–174.