Characteristics of the study samples
In our study, 1,119 high-quality microbial samples were collected and used for downstream analysis including piglet feces (n = 482), sow feces (n = 86), milk (n = 139), vaginal (n = 56), breast skin (n = 136), water (n = 25), air (n = 27) and floors (n = 168) (Fig. 1a, Additional file 1: Table S1). The microbiome of all samples was analyzed by 16S rRNA gene sequencing, yielding 74,032,942 high-quality sequences after quality control, with an average of 66,160 ± 391 sequences per sample (ranging from 7,746 to 92,011). The overall number of OTUs detected by the analysis reached 40,533 based on ≥ 97% nt identity. Rarefaction curves based on the Chao1 and Shannon indexes of all samples nearly reached a plateau, indicating that the sampling depth was sufficient to characterize the bacterial communities (Additional file 2: Figure S1).
Similarity Of The Microbial Community Structure Between Sample Types
Alpha diversity analysis revealed that the species richness and diversity of the microbial communities were distinct at different sources. The Chao1 index of the piglet microbiome was significantly lower than all environmental and maternal microbiomes except for the vaginal microbiome, and the Shannon diversity index of the sow fecal microbiome was significantly higher than that of other samples (Fig. 1b). In addition, we observed higher alpha diversity indexes in the piglet fecal microbiome at the first time point (day 0), and then they decreased shortly (to the bottom in day 5 for Chao 1 and in day 1 for Shannon diversity index) and rebounded over time.
The NMDS ordination based on Bray-Curtis dissimilarity showed distinct clusters between the sample types, and the early piglet microbiome did not consistently resemble one specific sow or environmental sample (Fig. 2a). For example, the early piglet fecal samples (at days 0 and 1) clustered with the sow vaginal samples, while they gradually shifted towards sow fecal samples as the piglets aged. Permutation Multivariate Analysis of Variance (PERMANOVA) showed that sample types significant effect on the bacterial community structure (stress = 0.16, P = 0.001). The fecal microbiomes in the piglets were relatively divergent from each other and had high intersubject variability, particularly on days 0, 1 and 3, compared with those of the sows and the environment (Fig. 2a, Additional file 3: Figure S2). Similarity percentage (SIMPER) analysis on the microbial community dissimilarity further confirm the result in the NMDS (Additional files 4: Figure S3). SIMPER analysis of the sow and environmental microbiota compared with the piglet fecal microbiota indicated that the piglet fecal microbiota on day 0 was more similar to the vaginal (index = 13.3), milk (index = 11.0) and breast skin microbiota (index = 12.6) than to the other microbiota groups. This high similarity between the piglet feces (day 0) and vagina and the sow milk and breast skin were attributed to the dominance of Proteobacteria (30.3%-41.1%, Additional files 5–7: Table S2-4). The similarity between the piglet fecal and sow vaginal microbiota increased within the first three days after birth and then gradually decreased, but the similarity between the piglet fecal and the sow fecal microbiota gradually increased as the piglets aged. At day 28, the piglet fecal microbiota was more similar to the sow fecal microbiota (index = 9.3) than to other samples, which was attributed to the dominance of Firmicutes (36.3%, Additional files 8: Table S5). The bacterial profiles at the genus level were analyzed using the UPGMA method to measure the similarity of bacterial community compositions between different sample types. Consistent with the NMDS analysis, the UPGMA clustering analysis revealed that the microbiota of piglet feces on days 0 and 1 clustered with sow vaginal samples and appeared to be more similar to feces on day 28 (Fig. 2c).
Firmicutes, Proteobacteria, Bacteroidetes and Actinobacteria were the four most abundant phyla in all samples except sow feces and accounted for 93.6% − 95.9% of the different sample types (Fig. 2b, Additional files 9: Table S6). Firmicutes, Bacteroidetes, Tenericutes and Spirochaetes were the most abundant phyla and accounted for 94.0% of the sow fecal microbiota. At the genus level, 23, 20, 25, 21, 22, 26, 22 and 26 predominant bacterial taxa (average relative abundance of > 1%) were identified in piglet feces (65.6% of the total sequences), sow feces (78.9%), milk (52.5%), breast skin (53.7%), vagina (55.4%), air (50.9%), slatted floor (55.9%) and water (55.6%) samples, respectively (Fig. 2c).
SourceTracker analysis highlights the contribution of sow and environmental sources to the piglets
SourceTracker, a Bayesian probability tool [29], was used to predict the relative contributions of the sow and delivery environment microbiome to the piglet fecal microbiota. The results revealed that the vaginal microbiota contributed the most to the meconium (day 0) microorganisms compared with other sources, followed by the slatted floor (9.6%), milk (9.4%) and air (8.5%) (Fig. 3a). The relative contribution of the vaginal microbiota to the piglet fecal microbiome increased in the first three days from 69.0–89.3% and then gradually decreased to 0.28% on day 28. Interestingly, the relative contribution from sow feces gradually increased after day 5 and finally reached the highest on day 28 (62.1%). However, the relative contribution of bacteria from sow milk was increased only on days 0 (9.4%) and 21 (15.0%). Apart from the vertical transmission of the sow microbiota, the neonatal piglets were also exposed to a wide variety of environmental microbiota. The environment (water, air and slatted floor) contributed 18.1% of the bacterial communities on day 0, rapidly decreased in contribution to 4.0% on day 3, and gradually increased in contribution to 34.1% on day 28, indicating that the slatted floor was the primary environmental source of bacterial communities in piglets, especially five days after birth, while air and water contributed less to the colonization of piglet bacterial communities than the slatted floor.
Ternary plot was used to more intuitively reflect the contribution of various bacterial sources to each fecal microbiome of piglets on different days. As shown in the plot, the piglet fecal samples were more closely related to the sow vaginal sample on the first 7 days (Fig. 3b). The piglet fecal samples diverged in their distributions among the vertices in the ternary plot at days 14 and 21, indicating that the bacterial sources during these times were more complex. At day 24, almost all fecal samples were uniformly distributed between the slatted floor and sow fecal samples, and most of the piglet fecal samples were close to the sow fecal samples in the plot at day 28. Analysis of the OTU cooccurrence patterns showed a hierarchy among the sample sources that were shared with piglet feces (Fig. 3c). This result indicates that more OTUs of piglet feces were shared in the milk, breast skin and slatted floor samples than in the water and air samples. The disparity between the similarity and the proportion of the OTUs shared between the piglet and sow fecal microbiota might be related to reduced diversity and therefore competition in the piglet gut microbiota [31]. Overall, the relative contribution of various sources of bacteria to the microbial composition of the piglet gut gradually changed as the piglets aged, and the main source of microbes in the fecal microbiota of the piglets was the vagina of the sow within 3 days after birth, which was gradually replaced by the sow feces and the slatted floor.
Maturation Of The Piglet Fecal Microbiota
Tracking individual OTUs within different phyla revealed distinct temporal dynamics within Firmicutes, Bacteroidetes and Proteobacteria. Many of the Firmicutes OTUs displayed dynamic volatility, with 16.4% disappearing between days 0 and 1; 77.8% of those that disappeared eventually reappeared at later times (Fig. 4a, left panel). A smaller proportion of the Bacteroidetes and Proteobacteria OTUs also showed dynamic changes. The greatest number of Bacteroidetes OTUs disappeared at days 0 and 1 (49.3%) but reappeared at later time points (Fig. 4a, middle panel). The greatest number of Proteobacteria OTUs disappeared from days 3 to 5 (35.1%) but reappeared at later time points (Fig. 4a, right panel). We used BugBase to further predict phenotypes in the piglet fecal microbiomes [32] BugBase predicted the fecal microbiome of the piglets to have a higher proportion of facultative aerobic bacteria than aerobic and anaerobic bacteria on days 0 and 1 (Additional files 10: Figure S4). The proportion of anaerobic and facultative anaerobic bacteria showed a contrasting trend, in which the proportion of anaerobic bacteria gradually increased from 22.3–68.0% during the first four weeks postpartum, while the proportion of facultative anaerobic bacteria gradually decreased from 55.75 to 11.8%. The proportion of aerobic bacteria in the piglet fecal microbiomes was only 15.2% at day 0, and this proportion decreased over the first five days postpartum before recovering over time (Additional files 10: Figure S4).
The relative abundances of OTUs were regressed against the chronologic age of each piglet using the Random Forest machine-learning algorithm to probe the age-dependent development of the piglet fecal microbiota. The regression explained 98.4% of the variance related to chronologic age. The top-ranking age-discriminatory taxa were selected according to their variable importance measures using 10-fold cross-validation. Thus, the top 30 age-discriminatory taxa were identified and used for the subsequent construction of the microbiota-based model for discriminating the degree of microbiota maturity, as inclusion of any taxa beyond these top taxa produced only minimal improvement in model performance (Fig. 4b). This model consisted of 21 genera that distinguished the maturity of the gut microbiota during the 28 days of the experiment. Although the natural development of the gut microbiota exhibited a smooth curve that gradually increased, the curve did not reach a plateau until day 28 (Additional files 11: Figure S5), indicating that the gut microbiota had not reached maturity by the end of this study. These age-discriminatory taxa were primarily affiliated with Lachnospiraceae and Erysipelotrichaceae, and the relative abundance of these age-discriminatory taxa significantly changed across the sampling times (Fig. 4b).
To explore bacterial interactions within piglet feces and environment samples, we used network analysis based on strong (Spearman’s rs < − 0.7 or rs > 0.7) and significant (P < 0.01) correlations of genera. In this network, it was assumed that cooccurring genera interacted with each other in either a positive or negative manner. The piglet feces network consisted of 53 nodes (genera) and 211 edges (relations) with an average degree (the mean number of connections per node) of 3.98 (Fig. 4c). According to the modularity algorithm, the piglet feces genera were partitioned into five modularity structures, where major age-discriminatory taxa such as Actinomyces and Bacteroides were part of the same subcommunity and had positive correlations. In addition, Actinobacillus, Epulopiscium and Pasteurella were also major age-discriminatory taxa, which were part of the same subcommunity and positively correlated. Most of the piglet fecal age-discriminatory taxa were also identified in the network of other sow and environmental samples (Additional files 12: Figure S6).
Diversification Of The Microbial Community Function
We next sought to examine how the microbial metabolic and functional pathways of the early piglet fecal metagenome changed over time. The majority of functional genes of the piglet fecal microbiota were associated with transporters (6.90%), ABC transporters (3.58%) and DNA repair and recombination proteins (2.66%) (Fig. 5a). The relative abundance of transporters was also the highest in the other samples. Principal coordinates analysis (PCoA) showed that the functional profiles of piglet fecal microbiota clustered more closely to the vaginal microbiota of sows at days 0 and 1, while they were more similar to the sow fecal microbiota at days 24 and 28 (Fig. 5b). The LEfSe analysis revealed that 63 differentially abundant bacterial functions were present across the piglet sampling times (Additional files 13: Figure S7).
The piglet fecal microbiota at day 0 was enriched for several microbial pathways, including secretion systems, pore ion channels, bacterial secretion systems, fatty acid metabolism, tryptophan metabolism and butanoate metabolism. In comparison, the piglet fecal microbiota at day 28 was significantly enriched for pathways related to sporulation, metabolism and biosynthesis, including starch and sucrose metabolism, methane metabolism, lysine biosynthesis and terpenoid backbone biosynthesis (Fig. 5c). There were no significant differences in the metabolic pathways of functional genes in the microbiota among piglet feces at day 0, sow vaginal samples, and piglet and sow feces at day 28 according to the LEfSe analysis. However, 62 differentially abundant bacterial functions were observed between the piglet and sow feces at day 0 (Fig. 5d). Metabolic functions, including fatty acids, tryptophan, glutathione and butanoate and valine, leucine, isoleucine, lysine, geraniol and caprolactam degradation, were overrepresented in the piglet fecal microbiota. In contrast, ribosome, methane metabolism, transcription machinery, DNA replication proteins and amino acid-related enzymes were underrepresented in the sow feces. The relative abundance of hypertrophic cardiomyopathy (HCM), the renin angiotensin system and the ubiquitin system were overrepresented in the vaginal microbiota of sows compared to the piglet fecal microbiota at day 28 (Fig. 5e).