Twelve Tibetan pigs (4 months of age) and twelve DLY [Duroc × (Landrace × Yorkshire)] pigs (4 months of age) in the growth stage were selected and maintained in the same house with one pig per pen which equips with fully slatted floors, a feeder, and a nipple drinker. All pigs were housed at Yunnan Agricultural University and fed by the same NRC (NRC 2012) diet for 28 days (Table 1), then euthanized via exsanguination. Jejunal chyme samples were collected by advancing the lumenal contents into a cryotube. The jejunum samples (2–3 cm in middle section) were collected and fixed in 4% neutral paraformaldehyde fix solution. The treatment of the pigs was approved by the Institutional Animal Care and Use Committee of Yunnan Agricultural University (No. YNAU20201304).

Each jejunal sample was cut into 4 μm thick slides and embedded in paraffin after fixation. And then three to four transverse slides from each sample were chosen to stain with hematoxylin and eosin. For immunohistochemical staining, slides were heated in microwaves at 750 W for 15 min in citrate buffer (pH 6.0) for antigen repairing after deparaffinization and rehydration. Following washing with phosphate-buffered saline (PBS), slides were treated for 25 min with 3% H₂O₂ at 20–22°C and in a photophobic environment to block endogenous peroxidase activity, washed with PBS, and trated for 30 min at 20–22°C with 3% bovine serum albumin to block. Then the slides were incubated with the primary antibody including rabbit anti-CD3 (Servicebio Technology, Wuhan, CN, #GB13014, 1:50), rabbit anti-MYD88, mouse anti-TLR4 and rabbit anti-TLR9 (#GB11269 and #GB11266, 1:100), rabbit anti-ZO-1 and rabbit anti-CD4 (#GB11195 and #GB13064-2, 1:200), rabbit anti-TLR2 (#GB11554, 1:500), rabbit anti-IL-1β and rabbit anti-IL-6 (#GB11113 and #GB11117, 1:800), and rabbit anti-MUC-2 (#GB11344, 1:1,000) diluted in PBS and at 4°C overnight. Next, the slides were washed with PBS and incubated for 50 min at 20–22°C with a secondary antibody HRP conjugated goat anti-rabbit IgG (#GB23303, 1:200). The immunohistochemical reaction was revealed for visual inspection using diaminobenzidine (DAB) as a chromogen (Vector Laboratories). Slides were further counterstained with hematoxylin and coverslipped. A digital microscopy scanner Pannoramic P250 (3DHISTECH Ltd., Budapest, Hungary) with a 20 × microscope objective was used to scan the immunostained slides. The corresponding whole-slide scanned images were evaluated using Pannoramic Viewer 1.15.2 (3DHISTECH Ltd., Budapest, Hungary). The final immunohistochemistry results were expressed as H-score as reported by Yang et al. (12).

Following the manufacturer's instructions, total microbial DNA extraction of jejunal contents sample was carried out using the HiPure Stool DNA Kits (Magen, Guangzhou, China). Amplicon pooled libraries for sequencing on an Illumina HiSeq 2500 (Illumina, San Diego, CA, USA) were prepared using primers 341F (5′-CCTACGGGNGGCWGCAG-3′) and 806R (5′-GGACTACHVGGGTATCTAAT-3′) targeted across the hypervariable V3-V4 regions of the 16S rRNA gene. The amplification was performed in the following steps: initial denaturation at 95°C for 2 min, followed by 27 cycles (98°C−10 s, 62°C−30 s, 68°C−30 s) and a final extension at 68°C for 10 min.

FASTP (https://github.com/OpenGene/fastp) was used to filter the raw sequences, and then with a minimal overlap of 10 bp and mismatch error ratio of 2%, paired-end reads were assembled as raw tags through FLSAH (version 1.2.11). QIIME (version 1.9.1) pipeline was used to filter noisy sequences of raw tags to obtain high-quality clean tags. Subsequently, with a similarity ≥97%, UPARSE (http://drive5.com/uparse/) was used to cluster the tags into OTUs (operational taxonomic units), while all chimeric tags were removed using Userach (version 7.0). Each representative sequence was assigned into organisms by RDP (Ribosomal Database Project) classifier (version 2.2) with a confidence threshold value of 0.8–1. Alpha diversity of observed species, Good's coverage, Chao1, Shannon, Simpson, PD whole tree, and beta diversity of principal component analysis (PCA) was determined using QIIME software (version 1.9.1). Based on high-quality reads, functional categories of KEGG ortholog were predicted by Tax4Fun (version 1.0).

The analysis of intestinal morphology and immune performance results was calculated by Welch's t-test using the SPSS 22.0 software. The Wilcoxon rank test was used to compute differential functional analysis among groups in R project (version 3.4.1). Correlations between variables were calculated using the Spearman rank correlation in GraphPad Prism 7.0.

In the jejunum, TP pigs had higher villus height and less deeper crypt depth than DLY pigs (p < 0.05). And TP pigs tended to have higher villus height to crypt depth ratio than DLY pigs (p = 0.162) (Table 2).

As shown in Table 3, at the protein level, the cluster of differentiation 4 (CD4+) and Toll-like receptor 9 (TLR9) were significantly higher in TP pigs than in DLY pigs (p < 0.05), as were the interleukin 10 (IL-10) and zonular occludens 1 (ZO-1) contents (p < 0.01).

We obtained 2,925,307 clean tags from the samples by using the IlluminaHiSeq2500 platform. Of these, 2,881,333 effective tags with an average length of 421 bp were selected after the quality control and filtering. This yielded an average of 120,055 effective reads per sample. The average effective tags to clean tags ratio was 94.68% (Table 4). We identified a total of 6,156 operational taxonomic units (OTUs) at a 97% similarity level. A Venn diagram (Figure 1) shows 249 OTUs are shared between TP and DLY groups, along with 188 and 276 unique OTUs in the TP and DLY groups, respectively.

Good's coverage for the two groups was >99.5%, indicating a satisfactory sequencing depth of the gut microbiota (Table 5). Alpha diversity indices were not significantly different between TP and DLY pigs (p > 0.05). For example, the Shannon index (p = 0.281) and simpson (p = 0.523) index of TP pigs were lower than those of DLY pigs.

The principal component analysis (PCA) score plot (Figure 2A) from sequences at the OTU level showed that the microbial composition of TP pigs was like that of DLY pigs, and as were the analysis of similarity (ANOSIM) at the OTU and phylum levels (p > 0.05) (Figures 2B,C). However, the TP pigs showed a significant different microbiome compared to DLY pigs at the genus level, as shown in Figure 2D (ANOSIM; p < 0.05).

The predominant phyla were Firmicutes, Proteobacteria, Actinobacteria, and Bacteroidetes (Figure 3). The relative abundance of Tenericutes in TP (0.76%) pigs was significantly higher than in DLY (0.24%) pigs (p < 0.05). The TP pigs displayed a 23.53% increase in the relative abundances of Firmicutes compared with DLY pigs (p = 0.07), while the relative abundance of Proteobacteria (p = 0.24), Actinobacteria (p = 0.25), Bacteroidetes (p = 0.06), and Cyanobacteria (p = 0.16) tended to be lower in the TP pigs than in the DLY pigs (Table 6).

At the genus level, the most predominant was Lactobacillus, followed by Escherichia-Shigella, Megasphaera, Pantoea, Acinetobacter, Pseudoscardovia, Pseudomonas, Bifidobacterium, Streptococcus, Olsenella, and Clostridium_sensu_stricto_1 (Figure 4). As shown in Table 7, TP pigs manifested an increase of 29.05 and 0.47% increase in the relative abundances of Lactobacillus and Solibacillus compared with DLY pigs, while the relative abundances of Pseudomonas, Stenotrophomonas, Phenylobacterium, and Sandaracinobacter in TP pigs were significantly lower than those in DLY pigs (p < 0.05).

Spearman's rank correlations (Figure 5) between the jejunal morphology phenotype, immune performance and relative abundance of differential microbiota were assessed, the relative abundance of Lactobacillus positively correlated with villus height (R = 0.65, p = 0.049), TLR9 expression at the protein level (R = 0.66, p = 0.044) negatively correlated with crypt depth (R = −0.89, p = 0.0011). The relative abundance of Solibacillus negatively correlated with crypt depth (R = −0.65, p = 0.049). The relative abundance of Pseudomonas positively correlated crypt depth (R = 0.78, p = 0.011). The relative abundance of Stenotrophomonas negatively correlated with villus height (R = −0.72, p = 0.024), TLR2 (R = −0.73, p = 0.021) and TLR9 expression at the protein level (R = −0.65, p = 0.047). The relative abundance of Sandaracinobacter negatively correlated with IL-10 (R = −0.84, p = 0.040). The relative abundance of Phenylobacterium negatively correlated with CD4+(R = −0.70, p = 0.022) and IL-6 expression at the protein level (R = −0.70, p = 0.022).

According to the Tax4Fun-based functional prediction, 16 differential KEGG pathways were found of the jejunal microbiota between the TP and DLY pigs (Figure 6). The pathways enriched in the TP pigs were mainly related to fructose and mannose metabolism, phosphotransferase system (PTS), glycerolipid metabolism, glycolysis/gluconeogenesis, RNA transport, glycosaminoglycan degradation, RIG-I-like receptor signaling pathway and flavonoid biosynthesis (p < 0.05). The pathways enriched in the DLY pigs were mainly related to Vitamin B6 metabolism, Toluene degradation, Meiosis – yeast, GABAergic synapse, Proximal tubule bicarbonate reclamation, Steroid biosynthesis, GnRH signaling pathway and Endocytosis (p < 0.05).