All experimental procedures on piglets were in accordance with European and Belgian regulations concerning the care and use of animals for research purposes and were performed at the facilities of the Nuscience Group (Melle, Belgium) under the license number A04/357 granted by the ethical commission of the University of Ghent to perform animal experiments.

A total of 128 male Large White × Danish Landrace × Piétrain piglets weaned at 21 days and with an average weight of 5.2 ± 0.5 kg (mean ± SD) were obtained from a commercial farm (Flanders, Belgium). Only male piglets were considered in this study to limit sex disparity in immunological function and susceptibility (31). The animals were allocated to 32 pens of 4 littermates based on body weight and were fed ad libitum during a 5-day adaptation period with a commercially formulated creep feed diet (Babistar Flex©, Nuscience Group, Melle, Belgium). Piglets were housed in pens with fully slatted plastic floors and wire-mesh sides in a temperature-controlled animal house, including eight identical rooms (with four pens each). Each pen was equipped with a feeder and a drinking bowl. Following the 5-day adaptation period, piglets were assigned to one of the four treatments (n = 8 pens): a control diet, an IN diet (control feed additionally supplemented with 0.2% of IN; Cosucra Group Warcoing SA, Warcoing, Belgium), and two CP diets (control feed additionally supplemented with either 0.2% or 2% of the dried CP (CP0.2/CP2%); Nuscience Group, Melle, Belgium). The dose of IN was chosen as suggested by the producer and based on the previous studies found in the literature that used 0.4–0.6% IN supplementation (32, 33). Moreover, Metzler-Zebeli et al. (34) showed that 0.2% IN was the median dose used in overall 43 studies, which was used in this study as a minimal effective dose. Therefore, the same dose was chosen for CP, and 10 times higher dose to reach similar amounts as described in few earlier studies (25–27). In each of the eight rooms, there was one pen per experimental diet. The basal diet was formulated to reach the nutritional requirements of animals following the NRC recommendations (Table 1). For the chemical composition of IN and CP, we refer to Uerlings et al. (35). Throughout the experiment, the pigs had ad libitum access to feed and water.

Six piglets facing too severe weight loss or with leg injuries after weaning had to be treated with antibiotics while one piglet suddenly died. Therefore, these were excluded from the experiment without affecting the number of experimental units, reaching eight pens per treatment for zootechnical performance parameters and eight piglets per treatment for the other parameters. A power analysis was performed using G^*Power software to calculate the sample size required for the experiment.

Moisture, crude ash, crude fiber, crude protein, crude fat, sugar, and starch contents in the feed were measured using near-IR spectroscopy (Tango-R, Bruker Belgium N.V., Kontich, Belgium). Minerals and trace elements were analyzed by inductively coupled plasma atomic emission spectroscopy (ICP-AES; Avio 500, PerkinElmer, Waltham, MA, USA) according to the EN15510 method.

Feed intake and body weight gain were weekly recorded throughout the entire experiment (n = 8 pens per treatment). The diarrhea status of the piglets was assessed visually, by a single observer, with fecal scoring every 2 days using a scale going from 1 to 3 (1 = hard or soft dry pellets; 2 = soft wet-shaped pellets; and 3 = unshaped soft pellets and watery feces). Each score was given per pen (n = 8 pens per treatment) and was converted into a weekly occurrence.

One pig per pen, with an average pen body weight, was euthanized on post-weaning days 10–11, i.e., 5–6 days after receiving the experimental diets, and d31–32 post-weaning, when piglets were 31–32 and 52–53 days of age, respectively (n = 8 animals per treatment). The sampling points were organized in 2 consecutive days with a randomized scheme and an equal distribution of animals from each experimental group to be euthanized per day. Anesthesia was applied by an intramuscular injection of a mix of Xylazine (Dopharma, Raamsdonkveer, the Netherlands) and Zoletil 100 (Virbac, Barneveld, the Netherlands) resulting in doses of 4 mg kg⁻¹ BW of xylazine, 2 mg kg⁻¹ BW of zolazepam, and 2 mg kg⁻¹ BW of tilamine. After anesthesia, piglets were euthanized with an intracardiac injection of sodium pentobarbital (0.2 ml kg⁻¹ BW; Release®, ECUPHAR NV/SA, Oostkamp, Belgium) and were immediately exsanguinated by the severance of the carotid arteries and jugular veins.

Tissue samples were collected from the duodenum (about 15 cm after the pyloric junction), jejunum (middle section of the small intestine), and ileum (about 20 cm from the ileocecal valve), rinsed with a saline solution, and fixed in phosphate-buffered formalin (10%, pH 7.6) for 48 h prior to storage in 70% ethanol for histomorphometric measurements. Colonic tissues were rinsed with saline, snap-frozen, and stored at −80°C until the gene expression assay. Ileal, cecal, and colonic contents were collected. Ileal content was immediately stored on wet ice until viscosity measurements. Furthermore, ileal, cecal, and colonic contents were snap-frozen and stored at −80°C until further analyses of metabolites.

Soluble fibers are more extensively and rapidly fermented than insoluble polysaccharides due to their higher water-holding capacity, which increases the viscosity of the digested and allows bacteria to easily penetrate the matrix (36). Therefore, the viscosity of the ileal digesta was determined. Ileal digesta (n = 8 animals per treatment) collected on d31–32 were centrifuged for 15 min at 21,000 g, and the viscosity of a 500-μl supernatant was measured with a CP40 cone and a constant shear rate of 450 s⁻¹ (Brookfield DV II+ viscometer; Brookfield, Middleboro, MA, USA). Viscosity measurements could not be achieved on d10–11 because of the lack of sufficient ileal content.

Tissues from the duodenum and jejunum (n = 8 animals per treatment/sampling time) collected on d10–11 and d31–32, were embedded in paraffin wax, cut at a 5-μm thickness with a microtome using Thermo MX35 Ultra blades (Thermo Fisher Scientific, Waltham, MA, USA), and stained with Alcian Blue-Periodic Acid Schiff. Twenty well-oriented villus-crypt units were selected on each slide to determine villus height and width as well as crypt depth (μm) by 10-fold microscopy (Olympus Corporation, Tokyo, Japan). The number of acidic and neutral goblet cells (count per crypt), the density in total goblet cells (count per 100 μm of the crypt), the thickness of the muscularis mucosae and tela submucosa, and the tunica muscularis thickness (μm) were determined.

Total RNA from colonic tissues (n = 8 animals per treatment/sampling time) collected on d10–11 and d31–32 was extracted using the ReliaPrep RNA Tissue Miniprep System (Promega Corporation, Madison, WI, USA) according to the protocol of the manufacturer. RNA concentration and quality were determined by Nanodrop dosage (Thermo Fisher Scientific, Waltham, MA, USA) and agarose gel (1%), respectively. The extracted RNA (75 ng) was converted into cDNA using the Reverse Transcription Master Mix (Fluidigm Corporation, South San Francisco, CA, USA). Samples were thereafter pre-amplified according to the PreAmp MasterMix instructions provided by the manufacturer (Fluidigm Corporation, South San Francisco, CA, USA), followed by an exonuclease I treatment (New England Biolabs, Ipswich, MA, USA).

High-throughput quantitative PCR (qPCR) was performed as previously described (35) with intron-spanning primer pairs (Supplementary Table 1) designed using Primer-BLAST (NCBI) and validated through agarose gel electrophoresis and through melting curves. High-throughput qPCR was performed in 48 × 48 dynamic array-integrated fluidic circuits (Fluidigm Corporation, South San Francisco, CA, USA) following the protocol: 60 s at 95°C, followed by 30 cycles (5 s at 96°C and 20 s at 60°C). Quantification cycles (Cq) were acquired using the Fluidigm real-time PCR analysis software 3.0.2 (Fluidigm Corporation, South San Francisco, CA, USA).

First, all housekeeping genes were evaluated, and the four most stable genes between treatments were determined by NormFinder (37). The selected four reference genes consisted of hypoxanthine phosphoribosyltransferase 1 (HPRT1), ribosomal protein L4 (RPL4), TATA-box binding protein (TBP), and tyrosine 3-monooxygenase/tryptophan 5-monooxygenase activation protein zeta (YWHAZ). For each target and housekeeping gene analyzed, the relative gene expression level was calculated using the Pfaffl method (38), and the geometrical mean of the relative expression of the four housekeeping genes was used to normalize all samples.

Ileal, cecal, and colonic contents collected on d10–11 and d31–32 of the experiment (n = 8 animals per treatment/sampling time/intestinal segment) were six-fold diluted in ultrapure water prior to metabolite determination by high-performance liquid chromatography (HPLC). Intermediate metabolites (lactate, pyruvate, succinate, and formate), SCFAs (acetate, propionate, and butyrate), and branched-chain fatty acids (BCFAs; i-butyrate, i-valerate, and valerate) were analyzed by isocratic HPLC using an Alliance System e2695 (Waters Corporation, Milford, MA, USA) with an Aminex HPx-87H column (BioRad, Hercules, CA, USA) combined with a UV detector (210 nm), with H₂SO₄ (5 mM) as a mobile phase at an eluent flow of 0.6 ml min⁻¹ and with a temperature of 60°C. Each peak was integrated using the Empower 3 software (Waters Corporation, Milford, MA, USA) and quantified using an external standard calibration. Although valerate is not a BCFA per se, the organic acid is usually classified as a metabolite from the proteolytic fermentation and is therefore included within the BCFA group. Intermediate metabolites and the sum of SCFAs were expressed in mg g⁻¹ of fresh content. Acetate, propionate, butyrate, and BCFA amounts were expressed as a ratio (%) of the sum of SCFAs (39). As there were only differences in SCFA and BCFA in the colonic content, these samples were chosen for further metabolome analyses by gas chromatography-mass spectrometry (GC-MS) and liquid chromatography-mass spectrometry (LC-MS).

The GC-MS analysis was performed at the VIB metabolomics core Ghent (Ghent, Belgium). Samples were dissolved in 1 ml of methanol and divided into two aliquots of 400 μl each. For each sample, one of the aliquots was evaporated to dryness under vacuum, and the obtained residue was derivatized by adding 10 μl of pyridine and 50 μl of N-Methyl-N-(trimethylsilyl)-trifluoroacetamide (Sigma-Aldrich, Saint Louis, MO, USA). The GC-MS analysis was carried out using a 7,890B GC system equipped with a 7,693A Automatic Liquid Sampler and a 7,250 Accurate-Mass Quadrupole Time-of-Flight (QTOF) MS system (Agilent Technologies, Santa Clara, CA, USA). About 1 μl of the sample was injected in a split-less mode with an injector port set to 280°C. Separation was achieved with a VF-5ms column (30 m × 0.25 mm × 0.25 μm; Varian CP9013; Agilent Technologies, Santa Clara, CA, USA) with helium as a carrier gas at a constant flow of 1.2 ml/min. The oven was held at 80°C for 1 min, ramped to 280°C at 5°C/min, held at 280°C for 5 min, ramped to 320°C at 20°C/min, held at 320°C for 5 min, and finally cooled to 80°C at 50°C/min at the end of the run. The MSD transfer line was set to 280°C, and the electron ionization energy was 70 eV. Full EI-MS spectra were recorded between m/z 50–800 at a resolution of >25,000.

Data processing was done in MassHunter Profinder (Agilent Technologies, Santa Clara, CA, USA) and includes feature extraction, combined with a chromatographic alignment across multiple data files. The appearance of both false positive and false negative features is minimized by “binning” the features in a chromatographic time domain. The NIST 17 Mass Spectral Library was accessed to screen for the spectra that matched the compounds present in the GC-MS chromatogram.

The LC-MS analysis was performed at the VIB metabolomics core Ghent (Ghent, Belgium). Samples were dissolved in 1 ml of methanol and divided into two aliquots of 400 μl each. For each sample, one of the aliquots was evaporated to dryness under vacuum, and the obtained residue was reconstituted in cyclohexane/water (1:1 v/v) (Sigma-Aldrich, Saint Louis, MO, USA). The extract in the water phase was used for the following ultra-high-performance liquid chromatography (UHPLC) analysis. The latter was performed on an ACQUITY UPLC I-Class system (Waters Corporation, Milford, MA, USA) consisting of a binary pump, a vacuum degasser, an autosampler, and a column oven. Chromatographic separation was carried out on an ACQUITY UPLC BEH C18 (150 × 2.1 mm, 1.7 μm) column (Waters Corporation, Milford, MA, USA) at 40°C. A gradient of two buffers was used: buffer A (99:1:0.1 water:acetonitrile:formic acid, pH 3) and buffer B (99:1:0.1 acetonitrile:water:formic acid, pH 3), as follows: 99% A for 0.1 min decreased to 50% A in 30 min, decreased to 30% in 5 min, and decreased to 0% in 2 min. The flow rate was set to 0.35 ml min⁻¹, and the injection volume was 10 μl. The UHPLC system was coupled to a Vion IMS QTOF hybrid mass spectrometer (Waters Corporation, Milford, MA, USA). The LockSpray ion source was operated in negative electrospray ionization (ESI) mode under the following specific conditions: capillary voltage, 2.5 kV; reference capillary voltage, 3 kV; cone voltage, 40 V; source offset, 50 V; source temperature, 120°C; desolvation gas temperature, 600°C; desolvation gas flow, 800 L h⁻¹; and cone gas flow, 50 L h⁻¹. Mass range was set from 50 to 1,000 Da. The collision energy for full HDMSe was set at 6 eV (low energy) and ramped from 20 to 70 eV (high energy), the intelligent data capture intensity threshold was set at 5. Nitrogen (greater than 99.5%) was employed as desolvation and cone gas. Leucin-enkephalin (250 pg μl⁻¹ solubilized in water:acetonitrile 1:1 [v/v], with 0.1% formic acid) was used for the lock mass calibration, with scanning every 2 min at a scan time of 0.1 s.

Profile data were recorded through a UNIFI Scientific Information System (Waters Corporation, Milford, MA, USA). ESI interface was employed in both positive and negative modes. Data processing was performed with Progenesis QI software version 2.4 (Waters Corporation, Milford, MA, USA). The in-house Mass Spectral Library (PhytoComp) and external spectral libraries (MONA; https://mona.fiehnlab.ucdavis.edu/) were used to screen for the spectra that match the compounds present in the LC-MS chromatogram.

DNA from colonic contents collected on d10–11 and d31–32 of the experiment (n = 8 animals per treatment/sampling time) was extracted using the QIAamp PowerFaecal Pro DNA kit (Qiagen, Hilden, Germany) following the instructions of the manufacturer. The DNA concentration and quality were determined by Nanodrop dosage (Thermo Fisher Scientific, Waltham, MA, USA) and agarose gel (1%), respectively. Following DNA extraction, 16S rRNA gene sequencing was performed by DNAVision (Gosselies, Belgium), using the Illumina MiSeq technologies (2 × 250 nt). The F-primer (5′-TCGTCGGCAGCGTCAGATGTGTATAAGAGACAGCCTACGGGNGGCWGC AG-3′) and the R-primer (5′-GTCTCGTGGGCTCGGAGATGTGTATAAGAGACAGGACTA CHVGGGTATCT AATCC-3′) were used to amplify the hypervariable regions V3-V4 according to the 16S Metagenomic Sequencing Library Preparation protocol from Illumina.

Quantitative Insights into Microbial Ecology (QIIME) software package (version 1.9.0) was used for operational taxonomic unit (OTU) clustering, with an identity cutoff of 97% by comparison to the Greengenes reference database 13.8. Beta diversity was measured using weighted UniFrac distance metrics with principal coordinate analysis (PCoA) using QIIME. The ADONIS method was used to determine whether communities differed significantly between groups of samples. Microbiota results were analyzed per time point using a Kruskall–Wallis test with the treatment as a fixed factor. The false discovery rate (FDR) correction was used to calculate the adjusted p-values. Microbial alpha diversity metrics, including Chao1, phylogenetic diversity whole tree, observed OTU, and Shannon indexes, were calculated using QIIME. Raw sequences have been uploaded to the European Nucleotide Archive database (project number PRJEB38284). Pearson's correlation coefficients between metabolites and microbiota communities were also calculated.

A second experiment was run in Gembloux Agro-Bio Tech, University of Liège, in Gembloux (Belgium) to determine intestinal permeability. All experimental procedures on piglets were in accordance with European and Belgian regulations concerning the care and use of animals for research purposes and were approved by the Animal Ethical Committee of Liège University, Belgium (protocol number: 1,860).

A total of 32 male Landrace × Piétrain piglets weaned at 21 days and with an average weight of 5.9 ± 0.6 kg (mean ± SD) were obtained from the Walloon Agricultural Research Center in Gembloux (Belgium). The animals were allocated to 16 pens of 2 littermates based on body weight. Piglets were housed in a temperature-controlled animal house with slatted floors and with a feeder and a drinking nipple per pen. Following the 5-day adaptation period with the commercial creep feed diet, piglets were assigned to one of the four treatments (n = 4 pens): a control diet, an IN diet at 0.2% (IN), and two CP diets (CP0.2/CP2%). Throughout the experiment, the pigs had ad libitum access to feed and water.

Intestinal permeability was assessed in vivo after 7 days of treatment with a sugar absorption test. Pigs (n = 8 per treatment) were initially fasted overnight and were subsequently administered with an oral dose of 5 ml kg⁻¹ of sugar solution, containing D-xylose (100 mg kg⁻¹; VWR International, Oud-Heverlee, Belgium) dissolved in water. One hour after oral administration, a 5-ml blood sample was collected from the jugular vein into gel and clot activator vacuum tubes (VWR International, Oud-Heverlee, Belgium). After centrifugation (10 min, 2,000 g at 4°C), serum samples were stored at −20°C until the analysis of serum concentration of D-xylose as a marker of intestinal absorptive capacities and mucosal integrity.

Serum D-xylose was determined as described by Eberts et al. (40). Briefly, D-xylose standard solutions were prepared by dissolving D-xylose in saturated benzoic acid to reach 0, 50, 100, 200, 400, and 800 mg L⁻¹ concentrations. The phloroglucinol color reagent solution was prepared with 0.5 g of phloroglucinol, 100 ml of glacial acetic acid, and 10 ml of concentrated hydrochloric acid (all from Sigma-Aldrich Co., St Louis, MO, USA).

The standard and serum samples (50 μl) were added to 5 ml of phloroglucinol color reagent solution, were heated at 100°C for 4 min, and were allowed to cool down in a water bath at room temperature. The absorbance was determined at 570 nm in a spectrophotometer (VICTOR plate reader, PerkinElmer, Waltham, MA, USA). The standard solution of 0 mg L⁻¹ D-xylose was considered as blank, and xylose-free serum was used to obtain the net absorptive concentrations.

Weekly body weight gain and feed intake remained unaffected by diets (p > 0.05). Piglets fed with CP2% displayed a significantly higher feed conversion ratio (1.66 ± 0.05) in comparison to their control counterparts (1.45 ± 0.06) after 1 week of treatment (p < 0.05), which was no longer observed during the 3 following weeks and for the entire experimental period (Supplementary Table 2). Piglet receiving CP2% demonstrated significantly lower proportions of score 1 feces, dry pellets, during the 2nd week of treatment (Figure 1A). No other significant differences were observed for weeks 1, 3, and 4 (data not shown) or for the entire experimental period (Figure 1B).

The CP2% diet (2.69 ± 0.36 cP) induced a significant rise in ileal chyme viscosity in comparison with control (1.48 ± 0.06 cP) and CP0.2% (1.48 ± 0.09 cP) treatments on d31–32 (p <0.0001), while IN-treated pigs (2.03 ± 0.03 cP) demonstrated intermediate values.

At d10–11, histological measurements showed a significantly reduced crypt depth with increased VH:CD ratio due to the treatment with CP and IN in all the studied segments of the intestine, including the duodenum, jejunum, and ileum (Table 2). The measurements at d31–32 showed a decrease in crypt depth in the duodenum (p < 0.05) and ileum (0.05 < p <0.1) and a significant increase of VH:CD ratio due to the treatment in the duodenum, jejunum, and ileum (p < 0.05).

Inulin-treated pigs displayed the thickest and the thinnest duodenal muscularis mucosae and tela submucosa layers on d10–11 and d31–32, respectively, reaching statistical trends (p < 0.1; Supplementary Table 3). No further differences were observed in other histomorphometrical measurements.

Acidic goblet cells showed a significant decrease in counts due to the treatment at days 10–11 in the duodenum and ileum (p < 0.05) (Table 2). Total goblet cell counts were also significantly affected by the treatment in the ileum section of the intestine (p < 0.05), while the duodenum showed a tendency. At the sampling times of 31–32 days, the acidic goblet cell count was significantly reduced due to the treatment in the jejunum section (p < 0.05), while it remained unaffected in the other sections.

Apoptosis-related genes in a colonic tissue (Figure 2) showed downregulation of caspase 1 (CASP1) in the IN treatment compared to control pigs on d31–32 (Figure 2B). None of the 11 genes involved in signaling pathways of inflammation were altered following IN or CP supplementation, neither on d10–11 nor on d31–32, as seen in Supplementary Figure 1 (p > 0.05). The inflammation target gene (Figure 3) β-defensin 2 expression was downregulated in the colonic tissue of CP2%- and IN-fed piglets in comparison to control piglets on d31–32 (Figure 3B). Barrier integrity genes (Figure 4) showed an upregulation of tricellulin (MARVELD2) on d10–11 in CP2%-treated pigs, which was significantly different from that in the CP0.2% treatment (Figure 4A). On d31–32, claudin-3 and mucin 2 (MUC2) mRNA levels were lower with CP 0.2% in comparison to the control diet. At the same time, IN-fed pigs showed decreased expressions of MUC2 and occludin compared to their control counterparts (Figure 4B). The FDR correction did not reveal any difference between diets for any of the genes.

The ileal and cecal SCFA and BCFA profiles remained unaffected by treatments, regardless of the sampling time (Supplementary Table 4). On d10–11, piglets receiving the CP2% diet had significantly lower colonic BCFA proportions in comparison to their control counterparts (p < 0.01; Table 3). At that time, significantly higher colonic acetate ratios were found with the CP2% treatment in comparison to the other treatments (p < 0.01; Table 3). At the same time, colonic lactate levels significantly increased following IN supplementation in comparison to CP0.2% and CP2% treatments (p < 0.05; Table 3). On d31–32, the CP2% diet tended to induce lower colonic propionate proportions, reaching a statistical trend (p < 0.1; Table 3).

The top significant features obtained with the LC-MS results are shown in Table 4 and can be visualized in Figure 5. The ESI+ list showed 12 features at d10–11 and 11 features at d31–32. The ESI– list showed 14 features at d10–11 and 14 at d31–32, which mostly include flavonoids such as hesperidin, neohesperidin, naritin, and naringenin.

Moreover, LC-MS resulted in an ESI– list containing 89 significant feature ions between the treatments at d10–11, while at d31–32 there were 100 features significantly different at an FDR < 0.05 and p-value < 0.05 (data not shown). Eighty-three features were in common between both time points. For an ESI+ list, at d10–11, there were 114 features significantly different, while at d31–32, 127 features were significantly different between treatments, at an FDR < 0.05 and p-value < 0.05. For this list, 104 features were in common between both time points (data not shown).

For all the colonic content samples analyzed, GC-MS resulted in the detection of 329 compounds. The list of the top 10 identified compounds and heatmaps obtained at d10–11 and d31–32 that had a p-value < 0.05 for the treatment effect are presented in Table 5 and visualized as (Supplementary Figure 2). None of them were significantly different between treatments after the FDR correction. Nonetheless, some of these identified compounds, such as octodecenoic acid, indole-acetic acid (IAA), and pipecolic acid, were interesting.

On d10–11, the observed OTU and phylogenetic diversity whole tree indexes were significantly impacted by treatments, whereas no significant differences were found for the Chao1 and Shannon indexes (Table 6), neither on d10–11 nor on d31–32. On d10–11, according to both the observed OTU and phylogenetic diversity whole tree indexes, higher alpha diversity was found with the IN treatment in comparison to the CP0.2% treatment (p < 0.05). Concerning the observed OTU index, the IN treatment was also significantly higher than the control treatment (p < 0.05).

Sequencing of the 16S rRNA genes in colonic digesta produced a total of 3,900,165 quality-filtered sequences, with a mean number of 60,940 ± 29,645 reads (mean ± SD) per sample. Communities did not significantly differ between treatments, and no visual clustering could be detected on the weighted UniFrac PCoA plots, neither on d10–11 nor on d31–32 of the experiment (data not shown).

Colonic microbiota composition with a relative abundance of at least 1% was included in Table 7. At the first sampling time (d10–11), the main phyla observed were Firmicutes, Bacteroidetes, and Proteobacteria; only two genera significantly differed between treatments, which disappeared after the FDR correction. The relative abundance of Faecalibacterium spp. was significantly increased in the colon of piglets fed with CP2% in comparison to the other diets. The colonic abundance in Anaerovibrio spp. was greater with IN and CP0.2% pigs compared to control pigs. When extending the analysis to microbiota communities with a relative abundance over 0.1% of the total colonic microbiota (Supplementary Table 5), six genera significantly differed at this sampling time although the FDR correction did not reveal any difference. The population of Ruminococcus spp. was lower in the colon of CP0.2%- and IN-fed piglets in comparison to the CP2%-supplemented piglets. The inclusion of CP2% induced the highest abundances of Lachnospira spp. (followed by CP0.2%) and unclassified Peptostreptococcaceae. Unclassified p-2534-18B5 (followed by CP0.2%) and unclassified Enterobacteriaceae levels were significantly increased under the IN treatment.

On the second sampling time (d31–32), colonic bacterial sequences present over 1% of the total microbiota predominantly included the Firmicutes, Bacteroidetes, Cyanobacteria, and Proteobacteria phyla (Table 7). The abundance in colonic Cyanobacteria was significantly affected by treatments with control piglets, showing greater proportions of the phylum than CP2%- and IN-fed piglets. Five genera significantly differed in the colon of pigs administered with the treatments on d31–32, considering a relative abundance of over 1% of the total microbiota (Table 7). The FDR correction did not reveal any difference between diets except for the Cyanobacteria phylum and unclassified 2YS2 (p < 0.05). Unclassified 2YS2 belonging to the Cyanobacteria phylum was significantly lower with CP2% and IN treatments compared to the control feed. Decreased levels of unclassified Lachnospiraceae were found in the colon of piglets offered CP2% in comparison to the control and IN-fed piglets. CP2% induced a higher abundance of Faecalibacterium spp., significantly differing from control and CP0.2% treatments. The population of Megasphaera spp. was increased by CP2% compared to the other treatments. The Bulleidia spp. the genus was present in higher proportions in the colon of IN and CP0.2%-fed piglets compared to the other two treatments. Considering genera present with a relative abundance of over 0.1% of the total colonic microbiota, five genera significantly differed between treatments although the FDR correction did not reveal any difference on d31–32 (Supplementary Table 5). Enterococcus spp. and unclassified Christensenellaceae were more abundant in the colon of piglets under IN supplementation compared to the other treatments. The colonic population of Lachnospira spp. was significantly higher with CP2% compared with IN and control diets. The genera Clostridium spp. and Shuttleworthia spp. were present in the lowest proportions with the CP2% treatment.

At 7 days after dietary treatment, piglets showed lower permeability by decreased serum xylose concentrations when fed with IN (0.45 ± 0.03 mmol/L) in comparison to control (0.69 ± 0.11 mmol/L), CP0.2% (0.77 ± 0.10 mmol/L), and CP2% (0.83 ± 0.14 mmol/L) treatments (p = 0.0868).