All experimental procedures were carried out with the approval (IACUC-150701) of the Animal Care and Use Committee in Guangdong Academy of Agricultural Sciences, China. A total of 60 male weaned piglets (Duroc × Landrace × Large White) with an average weight of 6 kg and 21-day old were used in this study. The control (CTR) group was the piglets fed the basal diet (Supplementary Table S1); the antibiotics and ZnO (AZ) treatment group was fed the basal diet supplemented with aureomycin (30 mg/kg), polymyxin E (12 mg/kg) and ZnO (3,000 mg/kg); the LC group was those given the basal diet added with low-molecular-weight (20,000~30,000 Da) chitosan (50 mg/kg), which was provided by Jiaxing Korui Biotech Co. Ltd, Zhejiang, China (http://www.korui-china.com/). The product of LC (KR901, Korui) is fabricated by physical methods and in the form of fine powder, water insoluble but soluble in dilute acidic solution, and recommended dosage 50 g/T for piglet's feed. Five replicates were used for each treatment (4 piglets per replicate, totaling 20 animals). The feeding trials lasted for 28 days. The food and water were daily processed ad lib. At day 28, 1 randomly selected piglet in individual replicate for each treatment (total 5 animals/treatment) were slaughtered under anesthesia. The contents of cecum were collected and proceeded. The cecal pH was immediately determined.

Microbial genomic DNA was extracted from 200 mg of the sample using the QIAamp DNA stool minikit (Qiagen, Germany) according to the manufacturer's recommendation. DNA quality was evaluated by the agarose gel electrophoresis. The V4 hypervariable regions (the forward primer was 520F 5-AYTGGGYDTAAAGNG-3 and the reverse primer was 802R 5-TACNVGGGTATCTAATCC-3) of 16S rDNA were PCR amplified from microbial genomic DNA (Caporaso et al., 2011). Briefly, 2 μL of diluted DNA sample (~20 ng/μL) was used for PCR amplification (25 μL mixtures). The PCR conditions were as follows: one pre-denaturation cycle at 98°C for 2 min, 25 cycles of denaturation at 98°C for 15 s, annealing at 50°C for 30 s, and elongation at 72°C for 30 s, and one post-elongation cycle at 72°C for 5 min. The PCR amplicon products were purified using 2% agarose gels and used to construct the sequencing library. The libraries of amplicons were attached to Illumina sequencing adapters using the NEBNext Ultra™ II DNA Library Prep Kit for Illumina (E7645L), purified using AMPure XP beads (Biomek, USA) and quality controlled on an Agilent 2100 Bioanalyzer (Agilent, USA). The pooled libraries were pair-end sequenced on the Illumina MiSeq platform with using 2 × 250 bp MiSeq reagent kit v2 (Illumina, USA).

Raw sequences were filtered with the average base quality lower than Q₂₀, the quality of the head or tail bases lower than Q₂₀, sequence lengths shorter than 150 bp, or reads with N. Then FLASH (v1.2.7) was used to assemble the paired-end sequences (Magoc and Salzberg, 2011). Reads with homopolymer runs exceeding 8 bp, primer mismatches, ambiguous bases and sequence lengths shorter than 150 bp were further removed in QIIME (v1.9.0) (Caporaso et al., 2010). The UCHIME (Chiu et al., 2017) of mothur (v1.31.2) (Edgar et al., 2011) was used to remove the chimera sequences.

Operational taxonomic units (OTUs) were counted for all samples with a cutoff of 97% identity using the UCLUST function (Schloss et al., 2009) in QIIME. Any reads <7 times (0.001%) were removed to minimize the impact of rare OTUs on our data analysis (Edgar and Baterman, 2010). The representative sequences of each clustered OTU were selected according to the maximum length, aligned to Greengenes 16S rRNA gene database (v13.8) (Bokulich et al., 2013) and classified by RDP classifier (v.2.2) (Mc Donald et al., 2012). The alpha-diversity indices (PD, Shannon, observed species and Chao1) were calculated for each sample, and beta-diversity (non-metric multi-dimensional scaling, NMDS) analysis was performed to show the group differences in microbial community structure. Beta-diversity statistical analyses were also tested using permutational multivariate analysis of variance (PERMANOVA) based on Bray–Curtis dissimilarities and 999 permutations in the vegan package (v. 2.3.2) (Wang et al., 2007).

The microbial function was predicted using Phylogenetic Investigation of Communities by Reconstruction of Unobserved States (PICRUSt) (v1.1.0) (Oksanen et al., 2015). Based on the pre-calculated Greengenes (v13.5) database (Langille et al., 2013), PICRUSt was performed on the abundance predictions of Kyoto Encyclopedia of Genes and Genomes (KEGG) orthologs and KEGG pathways of bacterial communities. The functional differences among groups were compared using Statistical Analysis of Metagenomic Profiles (STAMP) (Parks et al., 2014). Statistical analysis of two groups was conducted for Two-sided Welch's t-test and Benjamini-Hochberg FDR correction. Multiple groups were conducted for ANOVA with Tukey-Kramer test and Benjamini-Hochberg correction. Differences were considered significant at P < 0.05. Heatmap diagrams were plotted in R environment (v3.1.2) (http://www.r-project.org).

Raw paired-end reads were submitted to the Sequence Read Archive of the NCBI (accession number: SRP104359).

The cecal pH value in the piglets fed with LC was determined to be 6.19, which was significantly lower than that in the control (6.55) (P < 0.05). Similarly, compared to that in the control, the cecal pH (6.33) was significantly lower in the piglets fed with AZ (P < 0.05). The pH between LC and AZ treatments was not significantly different (P > 0.05) (Supplementary Table S2).

At least 47,083 sequence reads were obtained per sample and a total of 706,251 high-quality sequences were used for later analysis (Supplementary Table S3). Based on 97% sequence similarity, these sequences were assigned to 17,890 OTUs. After removing the rare OTUs (lower than 0.001% of total sequences), 2,242, 2,134, and 2,129 OTUs were retained from CTR group, LC group, and AZ group, respectively (Supplementary Table S4). From the Veen analysis of OTU, out of total 2,624 OTUs, 1,585 (~60% of the total OTUs) existed in three groups (Figure 1A). Instead, 148, 72, and 108 unique OTUs were identified in CTR, LC, and AZ groups, respectively.

We compared bacterial diversity (PD and Shannon index) and richness (observed species and Chao index) indices for alpha-diversity. PD index in the LC group or AZ group was significantly lower than that in the CTR group (P < 0.05), showing that the bacterial diversity in cecum was decreased by AZ or LC (Figure 1B). Shannon index between the two treatments did not show statistical difference while the value in the treatments was lower than the control (Supplementary Table S5). However, there was no significant difference in diversity index (PD and Shannon index) between LC and AZ (P > 0.05), indicating both of the treatment strategies decreased bacterial diversity under our testing conditions. About community richness indices (Observed species and Chao), there was no significant difference between LC and CTR, indicating that LC did not affect bacterial richness. However, the number of observed species or Chao index in the AZ group was significantly lower than that in CTR (P < 0.05) (Figures 1C,D), indicating that AZ decreased bacterial richness in the cecum. Collectively, our data suggested that piglets fed AZ had a decreased bacterial richness and diversity. Additionally, supplementary LC only decreased the gut bacterial diversity rather than bacterial richness.

For beta diversity analysis, we examined the relationships in cecal microbiome between the control, AZ supplement and LC addition using NMDS. The distribution of microbiota at each group was distinctly clustered separately along principal coordinate (Figure 1E). Moreover, PERMANOVA was used to test the Beta-diversity statistical analysis. The results showed that three groups were distinctly different in the distribution of microbiota (P < 0.05) (Table 1), indicating that LC and AZ significantly affected the cecal bacterial structure when compared to CTR.

Twenty-two bacterial phyla were assigned in the LC, AZ, and CTR groups and 14 of them were shared among them (Supplementary Table S6). Bacteroidetes, Firmicutes, and Proteobacteria accounted for more than 98% of the total sequence reads (Figure 1F). Approximately 1% of sequences in the cecum samples could not be assigned to a certain rank (unclassified) (Supplementary Table S6), suggesting that there were uncharacterized bacterial taxa existing in the cecal microbiome. The proportion of Bacteroidetes in LC and AZ was 1.5- and 1.9-fold higher than that in the CTR group (P < 0.05), respectively (Figure 1G). Instead, the proportion of Firmicutes in the LC and AZ treatments was around 0.8- and 0.6-fold lower than that in the CTR group (P < 0.05), respectively (Figure 1G). The proportion of Proteobacteria in the AZ treatment was 0.2-fold lower than that in the CTR group (P < 0.05), but in the LC treatment, it was in accordance with CTR group (Figure 1G).

Moreover, 41 bacterial classes (Supplementary Table S7), 68 orders (Supplementary Table S8), 96 families (Supplementary Table S9), and 124 genera (Supplementary Table S10) were assigned to all the three treatments. On the genus level, the relative abundance of Prevotella in the LC group (30% of the total bacterial community) and AZ group (29%) was significantly higher than that in the CTR group (10%) (P < 0.05) (Figure 1H). Compared to CTR and AZ groups, the relative abundance of Succinivibrio was significantly increased by LC treatment (P < 0.05) (Figure 1H). Similarly, the relative abundance of Anaerovibrio in the LC group was significantly higher than that in AZ group (P < 0.05) (Figure 1H). However, the relative abundance of CF231 in AZ group was significantly higher than that in CTR as well as LC (P < 0.05) (Figure 1H). The relative abundance of Desulfovibrio in treatments LC and AZ was significantly lower than that in the control group (P < 0.05) (Figure 1H).

Given the distinct microbiome changes within the LC and AZ pigs compared to CTR pigs, we tested whether the different treatments would lead to functional changes in each microbiome. Prediction of metagenome functional contents was conducted by applying PICRUSt to insight our 16S rDNA sequences. Compared to CTR, both LC and AZ had significantly higher function enrichments of energy metabolism, metabolism of terpenoids and polyketides, digestive systems and cell growth and death (P < 0.05) (Figures 2A,B). However, both treatments dramatically reduced the function enrichment of environmental information processing such as membrane transport (P < 0.05) (Figures 2A,B). In addition, LC group significantly increased the abundance of glycan biosynthesis and metabolism when compared to that in CTR group (P < 0.05) (Figure 2A). The relative abundance in the metabolism of cofactors and vitamin in LC treatment was higher than that in CTR or AZ group (P < 0.05) (Figures 2A,C). Furthermore, we also analyzed the microbial genes connected to KEGG Orthology (KO) terms. 1,373 of 6,014 KO terms (23%) were found to be significantly different among AZ, LC, and CTR groups based on Tukey-Kramer's ANOVA Test (P < 0.05). Remarkably, out of 1,373, 462 KO terms were assigned to bacteria Prevotella. It is worthy of highlighting that most of them (462 different KO terms from any two-group comparisons) were enriched on carbohydrate metabolism (15% different KO terms), metabolism of cofactors and vitamins (9%), energy metabolism (8%), glycan biosynthesis and metabolism (6%) (Figure 3), suggesting that the drastically increased Prevotella by LC and AZ played an important function among the intestinal microbiota. Further, 79 different KO terms whose relative abundance were >0.1% in any group were screened out, and most of them including iron complex were increased in LC than in CTR group (Figure 4). In the metabolism pathways, there were 62 and 29 different KO terms, whose relative abundance was >0.01% in any group, involved in the metabolism of cofactors and vitamins (Supplementary Figure S1) and glycan biosynthesis and metabolism (Supplementary Figure S2), respectively.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.