Implications
Animal studies seek to replace in vivo methods with in vitro methods. In this study, we show that in vitro models, used to investigate the influence of the diet on the digestive physiology and intestinal health of pigs, can omit some specific chemicals called reducing agents to induce the oxygen-free environment required for intestinal microbes. Therefore, building on this experiment, future research using such in vitro models will yield results that are more relevant, helping in reducing the number of animals usually used for such studies on intestinal health.
Introduction
In vitro fermentation models are increasingly used to characterize the fermentation of fiber-rich ingredients by intestinal microbes and its influence on digestive physiology (e.g. Bindelle et al., Reference Bindelle, Pieper, Montoya, Van Kessel and Leterme2011; Jha et al., Reference Jha, Woyengo, Li, Bedford, Vasanthan and Zijlstra2015) and intestinal health of pigs (e.g. Pieper et al., Reference Pieper, Bindelle, Rossnagel, Van Kessel and Leterme2009a; Cardarelli et al., Reference Cardarelli, Martinez, Albrecht, Schols, BDGM, SMI and Smidt2016). Although in vitro and in vivo methods are not directly comparable (Weiss et al., Reference Weiss, Rist, Rink and Mosenthin2015), Koecher et al. (Reference Koecher, Noack, Timm, Klosterbuer, Thomas and Slavin2014) showed that both types of models provide complementary information on fermentation in the gut. In in vitro models, after pre-digestion by pepsin and pancreatic enzymes to simulate digestion in the stomach and in the small intestine (Bindelle et al., Reference Bindelle, Buldgen, Wavreille, Agneessens, Destain, Wathelet and Leterme2007; Sappok et al., Reference Sappok, Pellikaan, Verstegen and Sundrum2009), ingredients are fermented in a closed system (syringe, flask or bottle) by a bacterial inoculum made from fresh or snap-frozen ileal, cecal or fecal material diluted in a mineral buffer solution. The required anaerobic environment is generated: (i) by carrying out feces collection and preparation of microbial inoculum under a constant stream of CO2 or N2 and (ii) by, except in a small number of early studies (e.g. Tilley and Terry, Reference Tilley and Terry1963), adding a reducing solution. Usually this solution contains Na2S, alone (Menke and Steingass, Reference Menke and Steingass1988) or in combination with cysteine-HCl (Theodorou et al., Reference Theodorou, Williams, Dhanoa, McAllan and France1994).
Such in vitro models lack the response of the host when exposed to the bacterial metabolites produced during the fermentation of the different ingredients, which limited their use in health related studies. To solve this limitation, in vitro adenocarcinoma intestinal cell culture models (Caco-2 cells or mucus-secreting HT29-MTX cell layers) have been combined to in vitro fermentation models (e.g. Payne et al., Reference Payne, Zihler, Chassard and Lacroix2012).
However the addition of Na2S and/or cysteine-HCl in the fermentation model releases metabolites such as H2S, ammonia, thiols or amines that could be detrimental to some intestinal microbes and to the intestinal epithelium while not being present in such amounts in natural digestive physiological processes. This could induce a bias in the results from cell culture models, limiting their relevancy. H2S, for example, is toxic to the intestinal mucosal barrier via DNA damage (Attene-Ramos et al., Reference Attene-Ramos, Wagner, Plewa and Gaskins2006), alteration of the cellular respiration (Medani et al., Reference Medani, Collins, Docherty, Baird, O’Connell and Winter2011) and inhibition of the butyrate oxidation in colonocytes (Roediger et al., Reference Roediger, Moore and Babidge1997). Ammonia is a metabolic disruptor due to its ability to inhibit mitochondrial oxygen consumption (Andriamihaja et al., Reference Andriamihaja, Davila, Eklou-Lawson, Petit, Delpal, Allek, Blais, Delteil, Tomé and Blachier2010) and short-chain fatty acids (SCFA) oxidation (Cremin et al., Reference Cremin, Fitch and Fleming2003) in colonic epithelial cells. Moreover, the in vitro microbial activity could be influenced by the reducing agent as they can precipitate essential metal ions and produce potentially toxic intermediates that may induce an imbalance between bacterial species (Fukushima et al., Reference Fukushima, Weimer and Kunz2003). Finally, reducing agents produce molecules that are also end-products of intestinal protein fermentation, blunting the ability of in vitro methods to quantify the production of metabolites originating from protein fermentation. H2S is produced by fermentation of sulfur-containing amino acids (AA) (Christl et al., Reference Christl, Murgatroyd, Gibson and Cummings1992) by bacterial species commonly present in the large intestine (e.g. Escherichia coli, Clostridium spp., Enterobacter aerogenes) (Kumagai et al., Reference Kumagai, Sejima, Choi, Tanaka and Yamada1975; Awano et al., Reference Awano, Wada, Mori, Nakamori and Takagi2005). Ammonia is generated in the large intestine by deamination of AA whereas the decarboxylation of AA by Bifidobacterium, Clostridium, Bacteroides, Streptococcus, Lactobacillus members can lead to the production of amines (Hughes et al., Reference Hughes, Magee and Bingham2000).
In this context, the present study aimed to investigate the possibility of using an in vitro gas production method in combination with a model of the host’s epithelium without the addition of any reducing agent in the incubation medium. For this purpose, the influence of the reducing agent (Na2S, cysteine-HCl, or none) on kinetics of gas and SCFA production, on intestinal microbial populations and on the toxicity of fermentation residues for epithelial cells was evaluated when protein (casein or soybean proteins) and carbohydrate (cellulose or potato starch) ingredients were fermented in vitro.
Material and methods
Ingredients
Four ingredients were used: soybean proteins (Soycomil; ADM, Rotterdam, the Netherlands), casein (C7078; Sigma-Aldrich, St. Louis, MO, USA), potato starch (S4251; Sigma-Aldrich) and cellulose (Mikro-technik GmbH & Co. KG, Bürgstadt, Germany).
Kinetics of gas and short-chain fatty acids production during in vitro large intestine fermentation
In vitro fermentation
In vitro fermentation was conducted as described in Bindelle et al. (Reference Bindelle, Buldgen, Wavreille, Agneessens, Destain, Wathelet and Leterme2007) with changes in the use of reducing agents as described hereafter. In brief, fresh feces were collected from three sows directly from the rectum and placed in plastic syringes. The air was chased from syringes and they were placed in a water-bath (39°C) for transportation to the lab. After <1 h, feces samples of the three sows were pooled in equal proportions on fresh-weight basis and added to three incubation media, prepared according to Menke and Steingass (Reference Menke and Steingass1988) (Na2HPO4, 1.423 g/l; KH2PO4, 1.548 g/l; MgSO4×7 H2O, 0.150 g/l; NaHCO3, 8.738 g/l; (NH4) HCO3, 0.999 g/l; CaCl2×2 H2O, 1.669 mg/l; MnCl2×4H2O, 1.264 mg/l; CoCl2×6 H2O, 0.126 mg/l; FeCl3×6 H2O, 0.101 mg/l; resazurin, 0.129 mg/l) but differing in reducing agent: (i) Na2S (14.3 mg/l), (ii) cysteine-HCl (25 mg/l) or (iii) control without reducing agent. The preparation was carried out under a constant stream of CO2, and subsequently, the three media were bubbled for 30 min with CO2.
The fermentation was initiated by mixing 200 mg of an ingredient with 30 ml of one of the three incubation media in a 140-ml glass bottle equipped with a pressure sensor module (Gas production system; Ankom Technology, Macedon, NY, USA). The experimental scheme was as follows: four ingredients (+ one blank without ingredient)×three incubation media×six replicates.
During 72 h, pressure data were regularly recorded. After 24 h, the fermentation of two replicates for each combination of ingredient×incubation medium was stopped. Fermentation broth was centrifuged (13 000×g, 5 min, 4°C), the supernatant and the pellet were stored at −20°C until further use, for SCFA analysis and for extraction of bacterial genomic DNA, respectively. Short-chain fatty acids were also analyzed after 72 h.
Gas production
Gas production curves were modeled according to Groot et al. (Reference Groot, Cone, Williams, Debersaques and Lantinga1996)
$${\rm G\,{\equals}}\,{{\rm A} \over {1{\plus}_}{{{\rm B}^{{\rm c}}\over{\rm t^c}}}}\!\!: {\rm if} \,\,t>0,$$
where G (ml/g of dry matter (DM)) denotes cumulative gas production v. time; A (ml/g of DM), maximal gas volume for t=∞; B (h) time at which 50% of A is reached; and C, a constant determining shape of the curve. From this equation, two other parameters were calculated: R max, maximum rate of gas production (ml/g of DM×h) and T max, time to reach R max (h).
Measurement of short-chain fatty acids production
Fermentation supernatants were analyzed for SCFA (acetate, propionate, butyrate, isobutyrate, valerate, isovalerate) production with a Waters 2690 HPLC system (Waters, Milford, MA, USA) fitted with an Aminex HPX 87 H column (Bio-Rad Laboratories, Hercules, CA, USA) combined with an UV absorbance detector (Waters 486 tunable absorbance detector; Milford) set at 210 nm. The analysis was performed at a flow rate of 0.6 ml/min and at 40°C using 3 mM H2SO4 at 5% CH3CN as eluant.
Cytotoxicity of fermentation residues
Cell culture
A Caco-2 cell line (ATCC n°HTB-37), derived from a human colon adenocarcinoma, was obtained from the American Type Culture Collection (Rockville, MD, USA). Cells were routinely cultured at 37°C in a humidified atmosphere with 5% to 10% (v/v) CO2 in air in Dulbecco’s Modified Eagle Medium (DMEM) (Lonza, Verviers, Belgium) supplemented with 10% (v/v) heat-inactivated fetal bovine serum (FBS) (Hyclone Perbio-Sciences, Erembodegem, Belgium), 1% (v/v) L-glutamine 200 mM and 1% (v/v) nonessential amino acids (NEAA) (Lonza) with weekly passage.
Preparation of fermentation supernatants
Cytotoxicity assay was performed on fermentation supernatants of the four tested ingredients and blank stopped after 24 h of fermentation. Each supernatant was filtered through a 0.22 μm filter (Millipore, Bedford, MA, USA) and diluted with 10% (v/v) of 10× concentrated DMEM (4.5 g/l glucose, 10% (v/v) FBS, 1% (v/v) L-glutamine 200 mM, 1% (v/v) NEAA (Lonza)). Five dilutions (10×, 30×, 100×, 300× and 1000×) were prepared with 1× complete DMEM from each initial supernatant mixed with 10× concentrated DMEM.
Cytotoxicity assay
Caco-2 cells were inoculated at a density of 20 000 cells/well in 96-well culture plates (Corning Costar #3596; Corning Costar Corp., Cambridge, MA, USA), and cultivated until 8 days post-confluence. Then, cells were incubated for 6 h (37°C, 5% (v/v) CO2/air) with 100 μl of supernatants (1×, 10×, 30×, 100×, 300× and 1000× dilutions), with the culture medium alone (negative reference), or with 1% (v/v) Triton X-100 (Sigma-Aldrich) (positive reference). At the end of the incubation, the possible cytotoxic effect of fermentation supernatants was determined using a Cytotoxicity Detection Kit (Roche Diagnostics GmbH, Mannheim, Germany) according to the manufacturer’s instructions. This assay is based on the measurement of the activity of cytosolic lactate dehydrogenase (LDH) released in the extracellular medium of Caco-2 cells upon cell damage or necrosis. During the experiment, cell morphology was observed by phase contrast microscopy.
Determination of bacterial composition
Total bacterial DNA extraction
Centrifugation pellets collected after 24 h of fermentation were extracted for total bacterial genomic DNA using a commercial kit (QIAamp® DNA Stool Mini Kit; QIAGEN, Crawley, UK), following the manufacturer’s instructions. DNA concentration and purity were measured by optical density using a NanoDrop ND-1000 (Isogen, Sint-Pieters-Leeuw, Belgium).
Bacterial 16S ribosomal RNA gene amplification and pyrosequencing
Pyrosequencing analyses were conducted as detailed in Tran et al. (Reference Tran, Boudry, Everaert, Théwis, Portetelle, Daube, Nezer, Taminiau and Bindelle2015), on bacterial V1 to V3 regions of the 16S rRNA gene amplified in a Ep Master system gradient apparatus (Eppendorf, Hamburg, Germany) and sequenced in the same titanium pyrosequencing reaction using the Roche 454 GS Junior Genome Sequencer (Roche Diagnostics, Vilvoorde, Belgium).
Bioinformatics analysis
Data processing from amplicon sequencing was carried out with the Genome Sequencer FLX System Software Package 2.3 (Roche Diagnostics). The raw reads were processed until operational taxonomic unit (OTU) binning and taxonomical assignation with MOTHUR v1.32 (Schloss et al., Reference Schloss, Westcott, Ryabin, Hall, Hartmann, Hollister, Lesniewski, Oakley, Parks, Robinson, Sahl, Stres, Thallinger, Van Horn and Weber2009). Raw reads denoising was performed with the Pyronoise algorithm implemented in MOTHUR. The presence of chimeric sequences was checked using ChimeraSlayer developed by the Broad Institute (http://microbiomeutil.sourceforge.net/#A_CS) (Su et al., Reference Su, Bian, Zhu, Smidt and Zhu2014). A taxonomical identification based on the SILVA database V1.15 (Pruesse et al., Reference Pruesse, Quast, Knittel, Fuchs, Ludwig, Peplies and Glöckner2007) to the genus level was assigned to each OTU (80% homogeneity cutoff).
In the second step, all unique sequences for each OTU were compared with the SILVA data set 1.15 gene microbial database using the Basic Local Alignment Search Tool algorithm (Altschul et al., Reference Altschul, Gish, Miller, Myers and Lipman1990). For each OTU, a consensus detailed taxonomic identification was given based upon the identity (<1% of mismatch with the aligned sequence) and the metadata associated with the best hit. Rarefaction curves and biodiversity parameters were obtained from subsampled datasets using MOTHUR. Chao index and inverted Simpson index were used for richness and alpha-diversity measures, respectively. Depth of coverage was estimated with Good’s coverage index (Rodriguez et al., Reference Rodriguez, Taminiau, Korsak, Avesani, Van Broeck, Brach, Delmée and Daube2016).
Statistical analyses
The influence of the ingredient and the reducing agent on gas fermentation parameters and on total SCFA production and molar ratio after 24 and 72 h of fermentation was analyzed using the MIXED procedure of SAS 9.4 software (SAS Institute Inc., Cary, NC, USA). Means were separated using the LSMEANS statement with a GLM based on two criteria of classification (ingredient, reducing agent) and their interaction. Cytotoxicity data as well as microbial composition (phylum and genus levels) of fermentation broth after 24 h of fermentation were analyzed similarly.
In all cases, the fermentation flask was used as experimental unit. Post hoc pair-wise comparisons were done using Student t test. Significance was judged at P<0.05.
In addition, a principal component analysis (PCA) was performed using the PRINCOMP procedure of SAS 9.4 software (SAS Institute Inc.) to determine the relationships between the ingredient and the reducing agent upon gas fermentation parameters, SCFA production, cytotoxicity and microbial composition of the fermentation broth. The correlation among the variables was determined with the loading plots of the first four eigenvalues.
Results
Kinetics of gas and short-chain fatty acids production during in vitro large intestine fermentation
Carbohydrate and protein ingredients differed in gas production profiles. Fermentation of carbohydrates yielded higher final gas production (A) than proteins (P<0.001; Table 1) but A was similar (P>0.10) among the three reducing agents for each ingredient. With higher B and T max and lower R max values, cellulose and soybean proteins fermented slower than potato starch and casein. For each ingredient, fermentation patterns with Na2S and in the control without a reducing agent were similar and differed (P<0.05) from cysteine-HCl, specifically for R max and T max.
Gas fermentation parameters modeled according to Groot et al. (Reference Groot, Cone, Williams, Debersaques and Lantinga1996) of four ingredients fermented by pig fecal bacteria in three incubation media: with Na2S or cysteine-HCl (Cys) or control (Ctrl) without reducing agent (n=2)
A=maximal gas volume (ml/g dry matter (DM)); B=time to reach 50% of A (h); C=constant; R max=maximum rate of gas production (ml/g DM×h); T max=time to reach R max (h); I=ingredient; R=reducing agent.
a,b,c,d,e,f,g Values within a row with different superscripts differ significantly at P<0.05.
For all ingredients, SCFA production was similar (P>0.10) after 24 and 72 h with Na2S and in the control without reducing agent (Table 2). Cysteine-HCl yielded higher total SCFA production for some ingredients. Branched-chain fatty acids (BCFA) ratio was similar (P>0.05) between incubation media for protein ingredients and higher than with carbohydrates (P<0.05).
Total short-chain fatty acids (SCFAtot) production (mg/g dry matter (DM)) and molar ratios (%) of acetate (C2), propionate (C3), butyrate (C4) and branched-chain fatty acids (BCFA) when four ingredients are fermented by pig fecal bacteria in three incubation media: with Na2S or cysteine-HCl (Cys) or control (Ctrl) without reducing agent (n=2)
I=ingredient; R=reducing agent.
a,b,c,d,e,f,g,h Values within a row with different superscripts differ significantly at P<0.05.
Cytotoxicity of fermentation residues
For the initial dilution, the LDH release was higher (P<0.001) after exposure to fermentation supernatants of casein than to the other ingredients and were over two-fold higher that the LDH release from the negative reference, which indicated a possible cytotoxic effect (Figure 1). Increasing dilutions of supernatants did not reach the threshold of cytotoxicity. None of the reducing agents induced a different release of LDH in the culture media (P>0.10) (data not shown).
Cytotoxic effects on intestinal mucosa upon exposure of Caco-2 cells to fermentation supernatants (initial dilution) of the four tested ingredients and blank of fermentation in three incubation media: with Na2S or cysteine-HCl (Cys) or control without reducing agent. Comparison with incubation medium alone (negative reference) or supplemented with 1% Triton X-100 (positive reference). The cytotoxicity was determined by assaying the lactate dehydrogenase (LDH) activity released in the culture medium and is given as means ± SD (n=8). Values are representative of two independent experiments. The horizontal discontinuous line represents the threshold of toxicity.
Determination of bacterial composition
The sequencing of 16S amplicons led to the analysis of 88 088 raw reads. After cleaning and chimera removal, 67 666 reads, that is, a mean of 2255 reads/sample, were binned into 13 728 OTUs with 0.03 clustering distance. Rarefaction curves are provided in the Supplementary Figure S1. The mean value of 0.99 for the Good’s coverage index at the genus level (Supplementary Table S1) shows that if the sequencing effort was deep enough to capture the dominant populations of the microbiota at the genus level which were fermenting the different substrates in the presence of the different reducing agents. Regardless of the ingredient and the reducing agent, the microbiota expressed as relative proportions of each phylum in fermentation broth was dominated by the two most abundant bacterial phyla in the porcine gut microbiota: Firmicutes (77.6±11.19%) and Bacteroidetes (10.6±5.94%). Other phyla included Proteobacteria (4.2±8.20%), Spirochaetes (0.3±0.23%) and Verrumicrobia (0.3±0.15%). Actinobacteria, Fusobacteria, Lentisphaerae, Fibrobacteres, Synergistetes each contributed to less than 0.1% of the population. Unclassified sequences accounted for 6.6±3.63%.
Relevant differences were found in the relative proportions of these phyla according to the ingredient. By comparison with the fermentation blank, the proportion of Firmicutes was reduced for potato starch, soybean proteins and cellulose (86.34% v. 62.99%, 71.50%, 78.22%, respectively; P<0.001) whereas, Bacteroidetes were more abundant for potato starch and soybean proteins (6.06% v. 13.16%, 20.82%; P<0.05). Fermentation broths of potato starch presented a higher proportion of Proteobacteria as compared with the blanks (17.45% v. 0.75%; P<0.0001).
In addition to the ingredient, the microbial communities changed with the reducing agent. At the phylum level, the comparison of the three reducing solutions for a given ingredient (Table 3) did not reveal any difference between Na2S and control without reducing agent (P>0.10). But the use of cysteine-HCl with potato starch modified the share of Firmicutes and Proteobacteria. The reduction in Firmicutes (47.56% v. 71.70%, 69.71%; P<0.0001) was compensated by a three-fold increase in Proteobacteria with cysteine-HCl compared with Na2S or control (31.77% v. 10.74%, 9.83%, respectively; P<0.0001). When casein was fermented with cysteine-HCl, the fermentation broth contained higher proportions of Proteobacteria than with Na2S or without any reducing agent (2.77% v. 0.33%, 0.35%; P<0.05).
Distribution of different phyla in the bacterial population (%) after 24 h of fermentation of the four tested ingredients and blank of fermentation by pig fecal bacteria in three incubation media: with Na2S or cysteine-HCl (Cys) or control (Ctrl) without reducing agent (n=2)
I=ingredient; R=reducing agent.
a,b,c,d Values within a row with different superscripts differ significantly at P<0.05.
1 Others bacteria include Actinobacteria, Fusobacteria, Lentisphaerae, Synergistetes, Candidate_division_TM7, Fibrobacteres.
At the genus level, only proportions of genera for which significant differences (P<0.05) were observed between the three reducing solutions for at least one ingredient are given in Tables 4 and 5. Similarly to the phylum level, the most significant differences in genera were found with potato starch and casein. For potato starch, several genera belonging to the Firmicutes (Streptococcus, Blautia, Lachnospiraceae unclassified, Veillonellaceae unclassified, Erysipelotrichaceae Incertae sedis) were less abundant with cysteine-HCl than with Na2S and/or the control (P<0.05). Higher proportions of Proteobacteria with potato starch with cysteine-HCl than with Na2S and control could be attributed to an increase in Succinivibrio (28.87% v. 9.08%, 8.24%; P<0.0001). Within Proteobacteria, Escherichia increased with cysteine-HCl compared with Na2S and control for the fermentation of both protein ingredients (P<0.01). For casein, Lachnospiraceae Incertis sedis and Dorea were more abundant when casein was fermented with cysteine-HCl (P<0.01), whereas the opposite trend was observed for Fusobacterium (P<0.0001).
Distribution of different genera in the bacterial population (%) after 24 h of fermentation of the four tested ingredients and blank of fermentation by pig fecal bacteria in three incubation media: with Na2S or cysteine-HCl (Cys) or control (Ctrl) without reducing agent (n=2)
i.s.=incertae sedis; uncl.=unclassified; I=ingredient; R=reducing agent.
a,b,c,d,e,f Values within a column with different superscripts differ significantly at P<0.05.
Distribution of different genera in the bacterial population (%) after 24 h of fermentation of the four tested ingredients and blank of fermentation by pig fecal bacteria in three incubation media: with Na2S or cysteine-HCl (Cys) or control (Ctrl) without reducing agent (n=2)
uncl.=unclassified; i.s.=incertae sedis; I=ingredient; R=reducing agent.
a,b,c,d,e Values within a column with different superscripts differ significantly at P<0.05.
Principal component analysis of the influence of the ingredient and the reducing agent
A PCA was performed based on the correlation matrix from the complete data set (Figures 2 and 3). The first two principal components (PC1 to PC2), which explained 61.3% of the variability in the data set, clearly discriminated samples according to the ingredient. High values along PC1 reflected high degrees of cytotoxicity of casein, combined to high BCFA production that was counterbalanced by reduced acetate and propionate ratios compared with the other ingredients. PC2 mainly discriminated ingredients according to the fermentation kinetics measured through B, T max and R max. Similarly, PC3 was mainly influenced by the ingredient and explained 16.3% of the observed variability. High proportion of Bacteroidetes after 24 h for soybean proteins was translated into high negative scores along PC3. PC4, accounting for 5.8% of the variance, showed a separation according to the reducing agent for casein and potato starch. High proportions of Fusobacterium and Escherichia, as observed when casein was fermented with cysteine-HCl, were associated with low values along PC4. Similarly, potato starch fermented with cysteine-HCl showed a high proportion of Succinivibrio, leading to a negative value along PC4 whereas a positive value was obtained with Na2S or the control.
Score plot from the first four principal components (PC1 to PC4). Different symbols indicate the scores of the four ingredients according to the reducing agent incorporated in the fermentation broth: Na2S, cysteine-HCl (Cys) or control without reducing agent (control). Symbols are: ●=Casein; ▲=Cellulose; *=Potato starch; ■=Soybean proteins.
Loading plot from the first four principal components (PC1 to PC4) describing the relationships between fermentation parameters, short-chain fatty acids (SCFA) production after 24 and 72 h of fermentation, cytotoxicity (a and b) and microbial composition (c and d). A=maximal gas volume; B=time to reach 50% of A; C=constant; R max=maximum rate of gas production; T max=time to reach R max; SCFAtot=total SCFA production; BCFA=branched-chain fatty acids; 1=Xylanibacter spp.; 2=RF16 unclassified; 3=Streptococcus spp.; 4=Mogibacterium spp.; 5=Blautia spp.; 6=Dorea spp.; 7=Lachnospiraceae Incertae Sedis; 8=Roseburia spp.; 9=Lachnospiraceae unclassified; 10=Ruminococcaceae unclassified; 11=Veillonellaceae unclassified; 12=Erysipelotrichaceae Incertae Sedis; 13=Fusobacterium spp.; 14=Succinivibrio spp.; 15=Escherichia spp.
Discussion
Comparing fermentation kinetics, SCFA production, cytotoxicity of the fermentation products and finally the microbiota fermenting several contrasted ingredients under three different reducing conditions (with cysteine-HCl, with Na2S, without any reducing agent) showed that the different reducing agents accounted for a limited part of the variability. Differences in the investigated response variables that can be obtained from such in vitro models were by far more influenced by the ingredient than the reducing agent since the three first PCs accounted for more than 75% of the variability. This was highlighted by the low share of variability (5.8%) explained by PC4 in the PCA and the fact that clusters grouped the different ingredients whatever the reducing agent along the first three PCs. Looking more specifically at individual response variables, final fermentability of the ingredients as measured through total gas production (A) was not influenced by the reducing agent added to the fermentation broth. Nonetheless, Morgan et al. (Reference Morgan, Kliem and Mould2004) observed with some feed ingredients a negative impact of the use of cysteine-HCl on rumen microbes activity in vitro, which were reflected in this study by changes in fermentation kinetics. The addition of cysteine-HCl to the buffer solution reduced the maximum rate of gas production (R max) for the carbohydrate ingredients. The second reducing agent (Na2S) did not influence the fermentation parameters for any of the tested ingredients. Casein and potato starch fermented faster than soybean proteins and in turn, soybean proteins fermented faster than cellulose. This is explained, in the absence of any enzymatic pre-digestion of the ingredients, by differences in solubility of the nutrients and water-holding capacity of the polymers.
Although the total SCFA production was not influenced by the reducing agent for casein and potato starch, the molar ratio of propionate was higher when cysteine-HCl was used by comparison with the control devoid of reducing agent. Higher propionate was counterbalanced by a reduced ratio of acetate and is explained by the higher proportion of Proteobacteria for casein and potato starch fermentation with cysteine-HCl since some Proteobacteria are specifically involved in propionate production (e.g. Succinivibrio dextrinosolvens) (Watanabe et al., Reference Watanabe, Suzuki, Koike, Nagashima, Mochizuki, Forster and Kobayashi2010).
From a phylum perspective, the microbiota, dominated by Firmicutes (77.6±11.19%) followed by Bacteroidetes (10.6±5.94%), was in agreement with Guo et al. (Reference Guo, Xia, Tang, Zhou, Zhao and Wang2008) and Kim et al. (Reference Kim, Borewicz, White, Singer, Sreevatsan, Tu and Isaacson2011), regardless of the ingredient and the reducing agent. The domination of the bacterial communities after 24 h of fermentation by these two phyla, composed principally of anaerobes, confirms the strict anaerobic conditions in all fermentation bottles, including in the controls devoid of reducing agent. Differences in microbial communities and SCFA profiles between the three reducing conditions after 24 h were more important with ingredients for which the fermentation started earlier, namely casein and potato starch. This means that although the influence of the reducing agent is less important than that of the ingredient, it significantly affects the fermentation pathways and bacteria that are growing in the broth, especially when substrates are being metabolized intensively by the microbial communities. So for casein and potato starch, the proportion of Proteobacteria was considerably higher when cysteine-HCl was used as reducing agent. This was ascribed to a proliferation of the Escherichia and Succinivibrio genera for the fermentation of casein and potato starch, respectively. Moreover among the Fusobacteria phylum, the proportion of the Fusobacterium genus was increased for the fermentation of casein in the presence of cysteine-HCl in the fermentation broth. These three genera are presumed to have the capacity to degrade cysteine-HCl. Indeed, Escherichia are able to degrade cysteine-HCl by means of specific desulfhydrases (Awano et al., Reference Awano, Wada, Mori, Nakamori and Takagi2005). Gomez-Alarcon et al. (Reference Gomez-Alarcon, O’Dowd, Leedle and Bryant1982) studied the nutrient requirements of S. dextrinosolvens and reported that the addition of cysteine to the culture medium promotes its growth, leading to increased propionate production as mentioned before. Regarding the Fusobacterium genus, F. nucleatum, found in the oral cavity and in infected sites of healthy and sick people, was also reported as able to degrade cysteine with specific desulfhydrases (Fukamachi et al., Reference Fukamachi, Nakano, Yoshimura and Koga2002).
Using culture-independent molecular techniques, some studies demonstrated that dietary changes affect the complex microbiota of the gastro-intestinal tract and lead to shifts in bacterial communities (Leser et al., Reference Leser, Lindecrona, Jensen, Jensen and Møller2000; Pieper et al., Reference Pieper, Janczyk, Urubschurov, Korn, Pieper and Souffrant2009b). So the higher differences in phyla according to the ingredient than the reducing agent are not really surprising. For example, potato starch led to higher proportions of Proteobacteria as compared with the other ingredients that could be attributed mainly to Succinivibrio, since its proliferation in the rumen was associated to a high-starch diet (O’Herrin and Kenealy, Reference O’Herrin and Kenealy1993). Also, Bacteroidetes were more abundant with potato starch and soybean proteins than in the blanks devoid of ingredients. This observation can be explained by the requirements in polysaccharides of the Bacteroidetes. In the human large intestine, the stability and coexistence of closely related members in this phylum was shown to be based on an synergetic interaction network related to the breakdown and utilization of dietary polysaccharides (Rakoff-Nahoum et al., Reference Rakoff-Nahoum, Coyne and Comstock2014).
Results from the Caco-2 assay indicate that the reducing agent incorporated in the fermentation broth has a negligible impact on intestinal cytotoxicity of the fermentation supernatants. Despite their colorectal origin, Caco-2 cells were proven a valuable in vitro model of the human and pig intestinal epithelium to assess toxicity of fermentation products (Artursson et al., Reference Artursson, Palm and Luthman2012). Indeed, differences in ingredients were observed comforting the relevancy of the test. The higher cytotoxicity of casein can be ascribed to its fast fermentation and by the production of toxic metabolites by the bacterial population such as ammonia, H2S, thiols and amines, which are usually produced by protein fermenting bacteria (Hughes et al., Reference Hughes, Magee and Bingham2000).
Conclusion
It can be concluded that in vitro intestinal fermentation models used in pig studies to investigate shifts in microbial fermentation and communities should avoid the use of cysteine-HCl as reducing agent. The addition of Na2S seems useless if appropriate CO2 saturation is realized and, since it probably blunts the ability to detect some protein fermentation metabolites, it can easily be omitted. Finally, when fermentation broths are applied to Caco-2 cells culture to simulate the effect of bacterial metabolites on the intestinal epithelium, the impact of the ingredient is detectable regardless of the reducing agent used, as for most of the response variables measured in the present study.
Acknowledgments
The project was supported by the Fund for Scientific Research (FNRS, Brussels, Belgium – Research Credit 1.5180.12). The research was performed in the framework of the collaborative Food4Gut excellence research program of the Walloon Government. The results were partially presented at the 12th International Symposium on Digestive Physiology of Pigs in Keystone, CO, USA, 29 May 29–1 June 2012 and published in the Supplement 4 (Volume 90, 2012) of the Journal of Animal Science.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/S1751731117002749
