The consistent trend in community composition over time, across all the cohorts, indicates that an age-related process of ecological succession is the largest factor shaping the microbial community of post-weaning piglets, as found in this study where animals aged 20–63 days and were fed the same diet. A peak in unrooted phylogenetic diversity and drop in balance weighted phylogenetic diversity (BWPD) reflects the acquisition of new species with the loss of dominating species. This change, detected in the week following the piglets’ arrival at the trial site irrespective of the cohort, could be linked to the piglets being subjected to microbial interchange (e.g.: new pen mates 92,93) and/or to diet transition (peri-weaning transition to solid food 92,93) leading to the reshaping of the gut microbial community. The week following the drop of BWPD, a significant increase of BWPD was recorded, reflecting the acquisition of a larger proportion of the community by the newly introduced species. The strong changes in phylogenetic diversity detected in the first and the second week could as well be attributable to other post-weaning related physiological changes, as previous studies report 44,92−94.
The highest inter-individual differences among piglets are seen in the first week of life, irrespective of maternal or environmental effects. The microbiota of 3 week old piglets is still very dynamic, but environmental factors become evident 93. At 6 weeks of age, CD8 + T cells infiltrate the intestinal tissue and the mucosa and intestinal lining resemble that of an adult pig 115. In this study, piglets reached a comparable alpha diversity to the sows after the first week of the trial, at which time the piglets were aged between 3.8 and 4.6 weeks. Unrooted PD did not reach higher levels at later sampling time points. The highest BWPD accompanied by a high unrooted PD was reached after the second week of the trial when piglets were aged between 4.9 and 5.6 weeks. Age-dependent physiological changes could explain i) the major shifts we detected in alpha diversity during the first two weeks of the trial and, ii) the distinct differences in community composition with age, even with a narrow age difference between piglets (1–6 days). We were able to appreciate a significant trend of increasing unrooted PD and decreasing BWPD with age in piglets that separated a 6 days maximum by day of birth from each other. Since age groups were confounded with breeds in our study, we attempted to determine the correlation within single breeds. Unfortunately, although the correlations with age could still be detected, we could not determine the association at later time points due to the introduction of treatment effects.
Animal trials are often conducted in controlled environments so as to minimize environmental effects. However, individual variations such as breed and age are often unavoidable in large animal trials. Previously reported confounding factors include: individual variation 44, cohabitation 92,93, age 44,92,93, maternal effects 92,93,95, hormones 44, behavioural differences between breeds (e.g. coprophagy, mouth to mouth contact) and extent of long-term behavioural adaptation, which can differ between breeds for reasons not attributable to genetics 44,92,93. A litter effect was found in piglets at the start of the trial and was lost at later time points during the trial. This could be due to either of the aforementioned factors. Co-housing, aging and the splitting of the piglets in separate rooms to receive a different treatment, are possible causes for loss of the litter effect. In this study we confirm the importance of these factors in the contribution to inter-individual variability of gut microbial composition. Motta et al (2019) report a correlation of beta diversity with age and no correlation of genotype and litter effect with either alpha or beta diversity 9. On the contrary, we found the piglet samples to significantly cluster by litter, breed and by age up to the second and the fourth week post weaning, in alpha diversity and beta diversity, respectively. We conclude that even small age differences among post- weaning piglets, down to the day, must be accounted for in an experimental set up.
Three groups of piglets (cohorts neomycin, neomycin + D-Scour™ and neomycin + ColiGuard®) underwent 5 days of treatment with the broad-spectrum antibiotic neomycin, via intramuscular administration. Intramuscular neomycin poorly diffuses (< 10%) into a healthy gastrointestinal tract 96, therefore a direct effect of neomycin on the gut microbiome may not be expected. However, neomycin showed a different trend in unrooted PD between the second and the third week of the trial, corresponding to the week following the neomycin treatment period for the neomycin cohort. Taking this time frame into consideration, the neomycin cohort did not increase in BWPD to the extent of the Control cohort. Although statistically significant differences between neomycin and Control in alpha diversity were not reached, the BWPD of the neomycin cohort appears to follow a different trend to the Control from the first week (during neomycin treatment) where neomycin treated piglets show the lowest decrease of BWPD compared to the control cohort and all other cohorts. While all cohorts show an increase in absolute phylogenetic diversity accompanied by a decrease of diversity evenness during this time frame, the neomycin cohort piglets show a lower drop in BWPD, suggesting an increase of species richness, without a corresponding loss of species evenness. Furthermore the neomycin cohort significantly separated from the control cohort in beta diversity in the third week of the trial, showing a higher representation of Mollicutes. Numerous studies report the link of oral antibiotic use with dysbiosis 22,35− 38,44,45, as well as with host physiology changes 37. On the contrary, the effect of intramuscular antibiotic administration on the microbiome is less well investigated. Correlation between intramuscular antibiotic use and dysbiosis has been reported in fish 46, gorillas 47, humans 97, and pigs 48,98. In 1 day old piglets, a single IM injection of amoxicillin (penicillin class) is reported to have an effect on the intestinal microbiota, detectable 40 days post treatment 48. Zeineldin et al (2018) tested the effects of IM administration of several antibiotics of various classes (penicillin, macrolide, cephalosporin and tetracycline), in 8-week old piglets, reporting shifts of the Firmicutes/Bacteroidetes ratio following treatment (length of the treatment not reported)98. The effects of intramuscular administration of neomycin (aminoglycoside class) on the gut microbiota have to our knowledge not been investigated. Based on our results we conclude that a mild effect on phylogenetic diversity is appreciable post IM neomycin treatment, up to two weeks after termination of the treatment. Additional compositional and functional analysis is necessary to determine the source of this mild variation. Differences were not detected at later time points, based on our phylogenetic diversity analyses, suggesting a full recovery of the microbial communities after two week from the end of the treatment.
It is possible that the large shifts in phylogenetic diversity taking place in the first two weeks irrespective of the treatment (an increase, then decrease of unrooted PD, and an opposite trend of BWPD) have masked the milder effects of the treatment, despite our efforts to control for the effects of aging. This could be the reason why a significantly distinct alpha diversity trend was found in the neomycin + D-Scour™ cohort compared to the neomycin cohort, but not in the D-Scour™ cohort compared to the Control cohort. The neomycin + D-Scour™ cohort underwent 5 days of neomycin treatment followed by 2 weeks of D-Scour™ treatment. A significant increase of BWPD was detected in the two-week period of D-Scour™ treatment, indicating a possible enhancement of microbiome evenness following neomycin treatment. To our knowledge there are no studies reporting an increased evenness in piglet gut community composition following a specific probiotic treatment. There are instead multiple studies reporting beneficial effects of probiotic treatment in sucker and weaner piglets in terms of improved gut mucosal integrity 66,80, growth rate 80− 83, digestibility of proteins and water absorption 80,83, reduction of pathogen invasion efficiency 76,79,80, and decreased mortality 80,82. Although the assessment of physiologic changes from probiotic treatments was outside the scope of this study, we found significant separation of neomycin + D-Scour™ cohort samples to neomycin cohort samples in beta diversity 3 and 10 days after D-Scour™ treatment, where neomycin + D-Scour™ samples showed a higher representation of Lactobacillales compared to neomycin samples, suggesting a transient establishment of the probiotic strains in the piglet guts.
The second probiotic in this study, ColiGuard®, did not have an effect on alpha diversity, but clustering was detected in beta diversity, where ColiGuard® samples separated from Control cohort samples in the third principal component (explaining 7.88% of the variation) two weeks post probiotic treatment. Additionally, the ColiGuard® treatment correlated with a lower weight gain, whether or not it was preceded by the antibiotic treatment. However, when comparing the overall weight gain (from the start to the end of the trial) the weight gain in the cohorts receiving ColiGuard® did not differ from the other cohorts.
We extracted the 16S rRNA gene hypervariable regions from our dataset, obtained the counts, and ran a correlation analysis to discover taxa that correlated with the weight of the piglets. As a consequence of the library size normalization step, the use of correlation with compositional data can inflate the false discovery rate 99,100. For this reason it can be expected that some of the taxa we found to correlate with the weight of the piglets (eighty-three distinct species) could be spurious while other correlations may have been missed.
Technical controls in metagenomic studies and methodological limitations
Taxonomic assignment of the raw reads from the positive controls was performed with MetaPhlAn2101 which relies on a ca. 1M unique clade-specific markers derived from 17,000 reference genomes. Such a database to map against the positive controls suffices as these organisms are cultivable, and for this reason they are widely studied hence the sequences are known. This is not the case for real-world samples where mapping against a database (which completeness relies on studied and often cultivable organisms) would narrow the view on the true diversity within the sample.
Positive controls with well-studied members and known ratios within the samples, has proven to be a valuable tool to assess consistency among technical replicates across batches and to detect possible biases derived from the DNA extraction method.
Systematic taxonomic bias in microbiome studies, resulting from differences in cell wall structures between Gram positive and Gram negative bacteria, have previously been reported; sample treatment with enzymatic cocktails can modestly reduce this bias 102–104. Although we implemented this step in our workflow, it seems that, from the read abundance of our mock community, which contained three Gram negative and four Gram positive strains, a bias towards Gram negative taxa may still be present.
In terms of contamination we concluded that: a) contamination in our study was not batch specific; b) a problem of sample cross-contamination existed at the DNA extraction step between neighbouring wells. During the bead beating step of DNA extraction, the deep-well plate is sealed with a sealing mat, rotated and placed in a plate shaker for the bead beating to take place. We consider that sample cross-contamination is most likely to occur during this step.