Nonstarch Polysaccharides Modulate Bacterial Microbiota, Pathways for Butyrate Production, and Abundance of Pathogenic Escherichia coli in the Pig Gastrointestinal Tract

A total of 8 crossbred Duroc-Landrace pigs (average weight, 22 ± 1.4 kg) from the herd of the Swine Research and Technology Centre, Edmonton, AB, Canada, were surgically fitted with a simple T-cannula at the distal ileum (23). The animal protocol was approved by the University of Alberta Animal Care and Use Committee for Livestock and followed the principles established by the Canadian Council on Animal Care (5). The pigs were assigned to one of four diets in a double 4-by-4 Latin square, resulting in 8 observations per diet. A semipurified diet based on cornstarch and casein was formulated to meet or to exceed the nutrient requirements for growing pigs (see Table S1 in the supplemental material) (34). The basal diet was supplemented with 5% active NSP ingredient of four purified NSP fractions: (i) low-fermentability, low-viscosity cellulose (CEL) (TIC pretested Ticacel MCC FG-100; TIC GUMS, White Marsh, MD), (ii) low-fermentability, high-viscosity carboxymethylcellulose (CMC) (TIC pretested Ticalose CMC 6000 F; TIC GUMS), (iii) high-fermentability, low-viscosity oat β-glucan (LG) (OatVantage; GTC Nutrition, Missoula, MT), and (iv) high-fermentability, high-viscosity oat β-glucan (HG) (Viscofiber; Cevena Bioproducts, Edmonton, AB, Canada). To reach 5% of the active NSP in the diet, the inclusion levels of the NSP fractions were 5.20, 6.25, 8.95, and 9.25% for CEL, CMC, LG, and HG, respectively. The NSP fractions were selected based on their in vitro viscosity levels, which were 0.3, 285, 20, and 210 mPa·s for CEL, CMC, LG, and HG, respectively, as determined with 0.5% NSP solution using a rheometer (UDS 200; Paar Physica, Glenn, VA) at a shear rate of 12.9/s and 20°C. The viscosity of β-glucans is linked to their molecular weight and has been reported to increase about 10-fold with a doubling of molecular weight (9). Titanium dioxide was added to the diet as a digestibility marker. The pigs were allowed to consume the experimental diets at a rate of approximately 3% of their maintenance requirement for energy (3 × 110 kcal digestible energy [DE]/kg metabolic body weight [BW^0.75]) (34). They were fed two equal meals in mash form twice daily (at 8 a.m. and 4 p.m.).

Each experimental period comprised 17 days; an adaptation period of 10 days was followed by collection of feces over 3 days and collection of ileal effluent for 8 hours a day, from 8 a.m. to 4 p.m., over 4 days. Feces were collected using plastic bags attached to the skin around the anus (45). The bags were changed each time feces were voided, and after subsamples of fresh feces were taken for analysis of short-chain fatty acids (SCFA), the feces were stored at −20°C until being freeze-dried. For bacterial DNA extraction, subsamples of freshly voided feces were taken and immediately stored at −20°C. The ileal collection procedure was adapted from the method of Li et al. (23), using plastic tubings attached to the barrel of the cannula by elastic bands. Twice during collection of ileal effluent (11 a.m. and 2 p.m.), subsamples of ileal effluent (approximately 50 ml) were collected for bacterial DNA analysis and immediately stored at −20°C. The ileal effluent was pooled within each animal and stored at −20°C. A subsample of fresh ileal effluent was stored separately for SCFA analysis. The remaining ileal and fecal samples were freeze-dried prior to analyses of dry matter and protein.

Total genomic DNA was extracted from ileal effluent and feces using a FastDNA kit (Qiagen, Mississauga, Ontario, Canada) according to the manufacturer's instructions. Partial fragments of bacterial 16S rRNA genes were amplified by PCR using universal forward primer S-D-Bact-0008-a-S-20 (AGAGTTTGATCMTGGCTCAG), labeled with 6-carboxyflourescein (FAM), and reverse primer S-D-Bact-0926-a-S-20 (CCGTCAATTCATTTGAGTTT) (37). The purified PCR product (200 ng) was digested at 37°C overnight using 15 U of MspI (Fermentas, Burlington, CA) in 2 μl reaction buffer and UV-sterilized Millipore water, made up to 20 μl. Two microliters of the digestion solution was subsequently mixed with 9 μl of formamide and 0.5 μl of an internal size standard (ABI GeneScan 600 LIZ size standard) and denatured at 95°C for 5 min, followed by cooling on ice for 2 min. Fragment sizes were analyzed using an ABI 3130xl genetic analyzer in gene scan mode and GeneMapper version 3.7 software (Applied Biosystems, Ontario, Canada). Fragments that were different at fewer than 3 bp were considered to be identical in accordance with binning criteria.

Nucleic acids were extracted from ileal effluents and feces of pigs by use of phenol-chloroform essentially as described by Knarreborg et al. (17). For DNA extraction, 200 mg of sample was weighed into a sterile tube containing 300 to 400 mg of sterile zirconium beads (diameter, 0.1 mm) and suspended in 1 ml of TN150 buffer (10 mM Tris-HCl, 150 mM NaCl [pH 8.0]). The suspension was vortexed and centrifuged at 14,600 × g for 5 min. The pellet was washed twice with 1 ml of TN150 buffer and was resuspended in 1 ml of TN150 buffer. The cells were lysed by physical disruption in a mini-bead beater (Biospec Products, Bartlesville, OK) at 5,000 rpm for 3 min and placed on ice to cool for 5 min. Subsequently, the sample was centrifuged at 14,600 × g for 5 min. A total of 900 μl of the supernatant was extracted twice with 1 ml TE (10 mM Tris, 1 mM EDTA [pH 8.5])-saturated phenol, followed by extraction with an equal volume of chloroform-isoamyl alcohol (24:1). The nucleic acids were precipitated with 2 volumes of cold ethanol (−20°C) and 0.1 volume of 5 M potassium acetate and stored overnight at −20°C. The DNA was collected by centrifugation at 14,600 × g for 20 min at 4°C, dried at room temperature for 1 h, and dissolved in 50 μl TE buffer (10 mM Tris-HCl, 1 mM EDTA [pH 7.5]). Prior to quantitative PCR (qPCR), DNA was diluted 5 times with sterilized Millipore water.

Quantitative PCR was performed on a model 7500 Fast real-time PCR system (Applied Biosystems, Foster City, CA) using the detection software (version 2.01; Applied Biosystems). Each reaction was run in duplicate in a volume of 25 μl in optical reaction plates sealed with optical adhesive film (Applied Biosystems). The reaction mixture consisted of 12.5 μl Fast SYBR green master mix (Applied Biosystems), 1 μl (10 μM) of primers (Table 1), and 1 μl of template DNA of ileal or fecal samples. To account for the degeneracy of the butyryl-CoA CoA transferase and butyrate kinase primers, higher primer concentrations were used in the reaction mixture (20 μM each primer). For amplification of the butyrate kinase gene and virulence factor genes, 12.5 μl QuantiFast SYBR green master mix (Qiagen, Mississauga, ON, Canada) was used. Amplification involved 1 cycle at 95°C for 5 min for initial denaturation, followed by 40 cycles of denaturation at 95°C for 15 s, primer annealing at the optimal temperatures (Table 1) for 30 s, extension at 72°C for 30 s, 1 cycle of 95°C for 1 min, 1 cycle at 55°C for 1 min, and a stepwise increase of the temperature from 55 to 95°C (at 10 s per 0.5°C) to obtain melt curve data. Data were collected at the extension step. Melting curves were checked after amplification in order to ensure correct amplification results. Standard curves were generated using serial dilutions of the purified and quantified PCR products generated by standard PCR using the primers shown in Table 1 and genomic DNA from pig intestinal contents (20). The detection limits were 10², 10⁴, and 10³ copies/g wet digesta for the group-specific primers, the butyrate enzyme gene primers, and the E. coli virulence factor primers, respectively.

Samples of diets, freeze-dried ileal effluent, and feces were finely ground to pass through a 0.5-mm mesh screen (Lab Retsch Mill, Haan, Germany). Dry matter, crude protein, and titanium dioxide were analyzed according to the method of the AOAC (1). Feces were analyzed for SCFA by gas chromatography as described by Htoo et al. (16).

Results for terminal restriction fragment length polymorphism (TRFLP) analysis and measurement of gene copy numbers for bacterial groups, butyrate enzymes, E. coli virulence factors, and fecal SCFA are expressed on the basis of wet weight of ileal effluent and feces rather than dry weight to illustrate the actual in situ situation in the gastrointestinal tract. Ileal protein flow was expressed on a dry-matter basis.

Ileal flow of dry matter and crude protein represents the amounts of dry matter and protein present in ileal effluent and was calculated according to the equation D_O = A_I × (I_D/I_I), where D_O is the total output of a nutrient in ileal effluent (g/kg of dry-matter intake), A_I is the concentration of a nutrient in ileal effluent (g/kg of dry-matter intake), I_D is the marker concentration in the assay diet (g/kg of dry matter), and I_I is the marker concentration in ileal effluent (g/kg of dry matter). Dry-matter disappearance in the large intestine was calculated as the difference between fecal and ileal dry-matter content levels.

Data were analyzed according to a double 4-by-4 Latin square design using the mixed procedure (PROC MIXED) of the Statistical Analysis System (SAS Institute, Inc., Cary, NC). The fixed effects included animal and treatment effects. Instances of period and animal within a square were considered random effects, assuming a compound symmetry variance-covariance structure. To detect any influential observation on the model, a test was performed using Cook's distance (Cook's D) as a criterion. Any observation with a Cook D greater than 0.5 was considered influential and hence excluded from further analysis. Degrees of freedom were approximated using the Kenward-Rogers method. P values of ≤0.05 were defined as significant, whereas P values that were both >0.05 and ≤0.10 were considered representative of a trend.

The TRFLP results were analyzed using Statistica software (version 6.0; Statsoft, Tulsa, OK). The profiles were normalized, and only fragments with a relative peak area ratio (P_i) of ≥1% were considered for further analyses. The total number of distinct fragments (richness [S]) was counted, and Shannon's index [H′ = −∑_{i = 1}^S(P_i)(log P_i)] and Simpson's index [1 − D = ∑_{i = 1}^S (P_i²)] were calculated as ecological measures of the relative distribution of bacterial groups in the community. The Shannon and Simpson indices take into account the number of species and the evenness (Shannon's index) or relative distribution (Simpson's index) of the species as represented by terminal restriction fragments (TRFs). The indices are increased either by having additional unique TRFs or by having a greater evenness of TRFs. The mean values for these parameters were compared by analysis of variance (ANOVA), followed by Tukey's honestly significant difference (HSD) test. Individual TRFs were assigned to microbial species by use of the MICA II online analysis tool (http://mica.ibest.uidaho.edu/trflp.php ), using the above-mentioned primer set and MspI for an in silico virtual digest against the Ribosomal Database Project (RDP) database.

The discrimination model was developed using linear discriminant analysis in the JMP software program (version 8.0.1; SAS Institute Inc., Cary, NC) to examine potential relationships between NSP fractions and TRFs and NSP fractions and gene copies for bacterial groups. Principal component analysis (PCA) was performed to examine any potential grouping of bacterial group gene copies, butyrate production pathway gene copies, E. coli virulence factors, ileal flow of dry matter, and postileal dry-matter disappearance according to the different diets used in this study by means of JMP software. The loading plot shows the variables responsible for the variation within the data set, and the correlations among individual variables of the first two eigenvalues (i.e., principal component 1 [PC 1] and PC 2). This gives a graphical representation of the extent to which each factor accounts for the variance in the data and shows the relationship between the different variables.

The dry-matter content levels of ileal effluent and feces were lower (P < 0.01) for CMC than for LG and HG (Table 2). The levels of ileal flow of dry matter and protein were higher (P < 0.01) for CEL, LG, and HG than for CMC. As a result, the levels of postileal dry-matter disappearance were greater (P = 0.01) for CEL, LG, and HG than for CMC. Fecal concentrations of total SCFA, including acetate, propionate, and butyrate, were lower (P < 0.05) in pigs fed the CMC diet than in those fed the CEL, LG, and HG diets. Inclusion of LG and HG additionally increased concentrations of isobutyrate and isovalerate in feces (P < 0.01).

The TRFLP analysis of ileal samples showed patterns dominated by relatively few major TRFs, whereas a total of 75 different TRFs were observed in feces (see Tables S2 and S3 in the supplemental material). Measures of species richness (4.1 ± 1.8 versus 14.5 ± 4.3), and Shannon's (0.34 ± 0.21 versus 0.86 ± 0.22) and Simpson's (0.40 ± 0.24 versus 0.76 ± 0.13) diversity were higher in feces than in ileum (Table 3). Among diets, diversity indices were not significantly different. However, a trend (P < 0.10) toward higher levels of diversity for CMC than for the other treatments was observed, whereas CEL showed relatively low diversity indices (Table 3), and similar trends for ecological measures were observed in ileal effluents and fecal samples.

Terminal restriction fragments identified in ileal effluent could be assigned to Streptococcus/Lactococcus spp., to Lactobacillus spp., to Clostridium clusters I, XIVa, and XVIII, and to Fibrobacter spp. Ileal communities were dominated by TRFs representing Streptococcus agalactiae-like species in almost all animals. In feces, the following bacterial groups and clusters could be identified: Streptococcus spp., Lactobacillus/Enterococcus/Oenococcus spp., Clostridium clusters I, IV, XI, XIVa, and XVIII. Furthermore, TRFs that contributed only about 1 to 2% to all TRFs were Corynebacterium, Collinsella, Fibrobacter, Selenomonas, and Desulfovibrio species-like phylotypes. Two TRFs in ileal effluent and 21 TRFs in feces could not be assigned to known species.

A set of 10 group-specific primers was employed to quantify bacterial populations in ileal effluent and feces. Bacterial populations in ileal effluent were only slightly affected by the NSP fractions (Table 4). The number of total bacterial gene copies in the distal ileum was lower (P < 0.05) for LG than for CEL, CMC, and HG. Clostridium cluster I was detectable only in the ileal effluents of pigs fed the CEL diet. Supplementation with CMC resulted in the highest number of Enterobacteriaceae family gene copies in ileal effluent (P < 0.05). In feces, the number of total bacterial gene copies was highest for CMC and lowest for LG (P < 0.05). The number of Clostridium cluster XIVa gene copies in feces was increased by CMC and reduced by LG in comparison to the number obtained with CEL, whereas the numbers of Clostridium cluster IV gene copies in feces were lower for both CMC and LG than for CEL (P < 0.05). Additionally, CMC and LG caused lower (P < 0.05) numbers of Clostridium cluster I gene copies in feces than CEL and HG. CMC increased the number of Bacteroides-Prevotella-Porphyromonas group gene copies in feces in comparison to the other NSP fractions, whereas HG reduced the number of gene copies in feces in comparison to CEL and CMC (P < 0.05). Finally, the number of Enterobacteriaceae family gene copies in feces was distinctly higher (P < 0.05) for CMC than for LG and HG.

Linear discriminant analysis of NSP fractions and TRFs (Fig. 1 a) divided the effects of NSP fractions into three clusters, comprising the CEL and LG diets (overlapping in their 95% confidence intervals), the CMC diet, and the HG diet. The CEL and LG diets were more related to TRF9 (Bacteroides species-like phylotype), whereas the CMC diet was linked to TRF22 (Clostridium polysaccharolyticum-like phylotype). The HG diet was more related to TRF14 (Clostridium innocuum-like phylotypes). In accordance with the linear discriminant analysis of NSP fractions and gene copy numbers for bacterial groups (Fig. 1b), samples were also divided into three clusters; however, CEL and HG formed a single cluster, as indicated by the intersecting 95% confidence intervals. The results for the LG diet were different from those for the CEL, HG, and CMC diets, whereas the results for CMC were drastically different from those for the other NSP fractions. Here, the gene copy numbers for Bifidobacterium spp. and Clostridium cluster I discriminated best for CEL and HG, whereas those for Clostridium cluster IV and Streptococcus spp. discriminated best for CMC and those for Clostridium cluster XIVa for LG.

Figure 1c depicts the loading plot showing the individual qPCR data, fecal SCFA, ileal flow, and disappearance of dry matter in the large intestine responsible for the variation of the first two eigenvalues (PC 1 and PC 2) and the correlations among individual variables. The first PC component accounted for 38.8% of the variation, and PC 2 explained 15.2%. The SCFA formed a cluster located on the right part of the graph and were highly influenced by PC 1. Lactobacillus spp., Bifidobacterium spp., Enterococcus spp., Clostridium clusters I and IV, and butyryl-CoA CoA transferase formed a second cluster at the bottom of the loading plot and were negatively correlated with PC 2. A third cluster was formed by total bacteria, Enterobacteriaceae, the Bacteroides-Prevotella-Porphyromonas group, Streptococcus spp., Clostridium cluster XIVa, E. coli virulence factors, and butyrate kinase, which were influenced by both PC 1 and PC 2 and were negatively correlated with the other two clusters (with angles of >90° between the arrows for these clusters). Variables within these three clusters were positively related among each other, as indicated by the small angles between the arrows for these variables (<90°).

Butyrate-producing bacteria in feces were determined by targeting genes for the enzymes butyryl-CoA CoA transferase and butyrate kinase (Fig. 2). The gene copy numbers for butyryl-CoA CoA transferase were higher (7.9 to 8.8 log₁₀ DNA gene copies/g [wet weight]) than those for the butyrate kinase (5.7 to 6.6 log₁₀ DNA gene copies/g [wet weight]). Dietary supplementation with CMC resulted in a lower (P < 0.05) gene copy number for the butyryl-CoA CoA transferase than supplementation with the CEL diet. In contrast, CMC increased (P < 0.05) the gene copy number for the butyrate kinase in comparison to LG and HG.

The heat-stable enterotoxin of enteroaggregative E. coli was the dominating virulence factor detected in both ileal and fecal samples, ranging from 6.0 to 8.9 log₁₀ DNA gene copies/g (wet weight). In ileal effluent, EAST1 levels were affected by NSP (P = 0.026) and were significantly higher for CMC and HG than for LG (6.8, 7.5, 6.0, and 7.3 ± 0.35 log₁₀ DNA gene copies/g [wet weight] for CEL, CMC, LG, and HG, respectively). The levels of virulence factors of enterotoxigenic E. coli heat-stable toxin A (STa), STb, and heat-labile toxin (LT) were below the detection limits in the ileal effluent. In feces, the EAST1 levels were higher by 1.4 to 2.5 log units for CMC than for the other NSP fractions (P < 0.05) (Fig. 3). The gene copy numbers for STa did not differ in feces, whereas the numbers of STb copies were higher (P < 0.05) for CMC than for LG. The LT was detectable only in the feces of pigs fed the CMC diet.