3.1 DNA sequence coverage in colonic content
As shown in Table S1, 2,668,457 V3-V4 16S rRNA valid sequences reads were obtained from 32 samples, including 78,848 (GHP), 88,485 (GLP), 85,353 (BHP), and 80,871(BLP) raw reads. After removing chimeric sequences, 74,565, 82,306, 80,145, and 75,997 clean reads remained in the GHP, GLP, BHP, and BLP groups, respectively. The effective proportion of clean reads was 90.14–97.79%. The GC% among the clean reads was 70.16–82.57%. The results showed that there were no dramatic differences in the number clean reads among the groups.
The Venn diagram analysis of OTUs is shown in Fig.1. There were 342 common OTUs among all groups (Fig. 1A), and 20, 40, 73, and 129 unique OTUs were identified in the GHP, GLP, BHP, and BLP groups, respectively. From the perspective of dietary crude protein levels (Fig. 1B), there were 396 common OTUs in the gilt groups, 81 unique OTUs in the GHP group, and 772 in the GLP group. Similarly, there were 1015 common OTUs in barrow groups, and 164 unique OTUs in the BHP group, and 244 in the BLP group. These results suggested that there were fewer common OTUs in gilts than in barrows. With respect to pig sex (Fig. 1C), there were 425 common OTUs in the HP groups, with 804 unique OTUs in the BHP group, and 90 in the GHP group. Similarly, there were 1060 common OTUs in the LP groups, with 264 unique OTUs in the BLP group, and 158 in the GLP group. The results suggested that there were fewer common OTUs in the HP groups than in the LP groups.
3.2 Observed species, microbial α-diversity, and cluster analysis
As shown in Table 2, the low protein diet increased the number of observed species significantly compared with high protein diet (P < 0.05). In contrast, pig sex had no significant effects (P > 0.05). However, pig sex, rather than dietary protein levels, affected the Shannon index significantly (P < 0.05) and was higher in barrows than in gilts. Neither the dietary protein level nor sex affected the Simpson index. Sample richness indices (ACE and Chao1) were higher (P < 0.05) in the pigs fed the low protein diet. Both indices showed no remarkable differences (P > 0.05) between gilts and barrows. In terms of all the α-diversity indices, no significant interactions (P > 0.05) were found between dietary protein levels and pig sex.
As shown in Fig.2, the UPGMA cluster of community structures at the phylum level were analyzed among the four treatments. The results showed that the GLP and BLP groups were the closest, and then they clustered together with BHP, followed by GHP, indicating that dietary protein levels had greater effects on microbial community structures than pig sex.
3.3 Relative abundance of the predominant microbial community induced by dietary protein levels in the colonic content of barrows and gilts
The results for the relative abundance of top 10 members of the microbial community structure in the colonic contents at different levels (phylum and genus) are shown Fig.3 and Table 3 and Table 4. At the phylum level, reducing dietary protein levels significantly increased the abundance of Actinobacteria (P < 0.05) and decreased the abundance of unidentified bacteria (P < 0.01). No remarkable differences (P > 0.05) were found between gilts and barrows. For Proteobacteria, Gracilibacteria, and Synergistetes, there were significant interactions (P < 0.05) between dietary protein levels and pig sex. At the genus level, the top three microbial community, including unidentified Clostridiales (P < 0.05), Neisseria (P < 0.05), and unidentified Prevotellaceae (P = 0.00) were significantly affected by dietary protein levels. The abundance of unidentified Prevotellaceae was also different (P < 0.01) between gilts and barrows. In addition, dietary protein levels and pig sex showed significant interactions in Neisseria (P < 0.05) and unidentified Prevotellaceae (P = 0.00). In addition, the abundances of Gracilibacteria (P < 0.05) and unidentified bacteria (P = 0.00) were significantly affected by dietary protein levels, and showed significant interactions (P < 0.05) between dietary protein levels and pig sex.
3.4 Metabolome profiles and PCA of the main metabolites in the colonic content of barrows and gilts induced by dietary protein levels
To reveal the effects of dietary protein levels and pig sex on intestinal metabolic profiles, LC-MS was used to analyze the metabolome of the colonic content. As shown in Table S2, the score plot of LC‑MS (electrospray ionization negative (ESI-)) data with 2037 metabolite signals and LC-MS (ESI+) data with 3844 metabolite signals were detected.
From the perspective of dietary protein levels (Fig. 4A), the PCA results showed that dietary protein levels had a robust influence on main metabolites of pigs, especially between GHP and GLP groups, in which the metabolic communities were clustered. From the perspective of pig sex (Fig. 4B), the PCA results showed that the main metabolites between the two pairs of BHP and GHP, and BLP and GLP groups were mixed together. Especially for barrows, there was a marked variation among samples even in the same group. In contrast, samples of gilts were more gathered. The PLS-DA score plots (Fig. 5) also showed that the GHP and GLP groups were well-separated, suggesting that dietary protein levels caused more significant biochemical changes in gilts compared with that in barrows. These results suggested that main metabolites between gilts and barrows and within barrows fed different dietary protein levels had no significant differences. Therefore, in this study subsequent analysis on microbial different metabolites-related results mainly focused on the experimental gilts fed the HP and LP diets.
3.5 Identification and KEGG analysis of differently abundant metabolites in colonic content of gilts fed the high protein and low protein diets
Furthermore, the parameters of variable importance of projection (VIP) >1.0 and adjusted q < 0.05 were used to detect differentially abundant metabolites in response to different dietary protein levels in gilts. As shown in Fig.6 and Table S3, compared with those in the GHP group, a total of 156 differentially abundant metabolites in LC-MS (ESI-) were identified in the GLP group, including 32 increased and 124 decreased abundant metabolites. Similarly, 278 metabolites in LC-MS (ESI+) were identified, including 126 increased and 156 decreased abundant metabolites. These results suggested that the low protein diet induced more decreased and fewer increased abundant metabolites.
The KEGG was used to analyze the pathways of the differentially abundant metabolites between the two gilt groups. As shown in Fig.7 and Table S4, the metabolic pathways of the Phosphotransferase system (PTS), Ascorbate and aldarate metabolism, the HIF-1 signaling pathway, and Asthma and Glutathione metabolism were associated with four metabolites in LC-MS (ESI-) and were significantly affected by dietary protein levels. Inflammatory mediator regulation of TRP channels, the Fc epsilon RI signaling pathway, Linoleic acid metabolism, Degradation of aromatic compounds, and Biosynthesis of alkaloids derived from histidine and purine were associated with nine metabolites in LC-MS (ESI+) and were significantly affected by dietary protein levels. Interestingly, the metabolite vitamin C (also named ascorbic acid), was enriched and regulated the pathways of PTS, Ascorbate and aldarate metabolism, the HIF-1 signaling, and Glutathione metabolism.
3.6 Correlation between the predominant microbial community and differentially abundant metabolites in the colonic content of gilts induced by dietary protein levels
There were six genus-level microbial communities whose proportions showed significant differences in response to dietary protein levels (Table S5). Compared with the GHP group, the proportions of unidentified Clostridiales (P < 0.05) and Terrisporobacter (P < 0.01) were significantly increased in GLP and the proportions of the remaining communities, Neisseria (P < 0.05), unidentified Prevotellaceae (P = 0.00), Gracilibacteria (P < 0.05), and unidentified bacteria (P < 0.01), were decreased.
To further reveal the crosstalk between the microbiota and the host, the six communities were selected and used to analyze the correlation with 12 changed metabolites that were enriched in above KEGG analysis. As shown in Fig.8, the proportion of unidentified Clostridiales was associated positively with the levels of the Platelet-activating factor (P < 0.05), Cinnamaldehyde (P < 0.01), Carbazole (P < 0.01), and Arachidonic acid (P < 0.01). Neisseria was associated positively with Vitamin C (P < 0.01), Histamine (P < 0.01), Naphthalene (P < 0.01), and Acetophenone (P < 0.01), and negatively with Cinnamaldehyde (P < 0.05) and 3-Phenylpropanoic acid (P < 0.05). Unidentified Prevotellaceae was associated positively with Vitamin C (P < 0.05), D-Mannose 6-phosphate (P < 0.05), Dihomo-gamma-linolenic acid (P < 0.05), Naphthalene (P <0.05) and Dolichotheline (P < 0.05), but negatively with Platelet-activating factor (P < 0.01), Cinnamaldehyde (P < 0.01), 3-Phenylpropanoic acid (P < 0.01), Carbazole (P < 0.01), and Arachidonic acid (P < 0.01). The proportion of Terrisporobacter was associated positively with Platelet-activating factor (P < 0.01), Cinnamaldehyde (P < 0.01), Carbazole (P < 0.01), and Arachidonic acid (P < 0.01), but negatively with Vitamin C (P < 0.05) and Histamine (P < 0.01). Gracilibacteria were associated positively with Vitamin C (P < 0.01), Histamine (P < 0.01), Dihomo‑gamma‑linolenic acid (P < 0.01), Naphthalene (P < 0.01), and Acetophenone (P < 0.01), but negatively with Cinnamaldehyde (P < 0.05) and 3-Phenylpropanoic acid (P < 0.01). Unidentified bacteria were associated positively with Histamine (P < 0.01) and Naphthalene (p <0 .05) but negatively with Platelet-activating factor (P < 0.05), Cinnamaldehyde (P < 0.01), 3-Phenylpropanoic acid (P < 0.01), Carbazole (P < 0.01), Arachidonic acid (P < 0.01), and Dihomo-gamma-linolenic acid (P < 0.01).