Study design

The use of animals in this study was approved by the Animal Care Committee at the University of Guelph. Ten conventionally-raised Yorkshire-Landrace pigs (4 males, 6 females) at a 300-sow batch-farrowing facility were enrolled on the day of birth from ten different litters (1 pig per litter). Each pig in the herd had a unique identifier assigned at birth and the pigs enrolled in the present study were randomly selected (random number generator) from a stratified sampling frame of pigs that weighed ≥1 kg at birth and did not demonstrate clinical signs of disease. Each pig was raised with their mother and litter-mates for 21 days without creep feed and, at weaning, randomized into 4 pens (located in the same nursery room) containing 11–13 pen-mates. Pigs were given a first phase ration for 14 days followed by a second phase ration for the remainder of the nursery period (Additional file 3: Table S1). The rations did not contain antimicrobial compounds and pigs were not administered any antimicrobials during the trial.

Samples were collected on days 1, 3, 7, 14, 21 (prior to weaning), 28, 35, 42, and 49. Samples were collected at the same time of day by the same investigator. At each sampling, a swab of the rectum was collected, as well as whole feces when possible, and a swab of the interior of both nares (approximately 10 mm penetration). Samples were transported at 4 °C and stored at −80 °C before being processed.

DNA extraction

Extraction of DNA from fecal and nasal samples was completed using a commercially prepared kit and following the manufacturer’s protocol for stool DNA extraction for pathogen detection (E.Z.N.A. Stool DNA Kit, Omega Bio-Tek Inc., Doraville, Georgia, USA). For nasal samples, the whole tip of the swab was processed through the lysis stage of extraction.

Amplification and sequencing of bacterial 16S rRNA gene

The V4 region of the 16S rRNA gene was amplified using forward (5′-AYTGGGYDTAAAGNG-3′) and reverse (5′-TACNVGGGTATCTAATCC-3′) primers previously designed [18]. The 16S primers contained adapter regions (Forward: TCGTCGGCAGCGTCAGATGTGTATAA-GAGACAG, Reverse: GTCTCGTGGGCTCGGAGATGTGTATAAGAGACAG) for annealing to Illumina universal index sequencing adaptors that were added in a later PCR.

Amplification of the 16S rRNA V4 region was completed in a 25 μl reaction consisting of 12.5 μl of KAPA 2G Fast HotStart ReadyMix 2X (KapaBiosystems), 9.0 μl of molecular-grade water, 2.5 μl template DNA, and 0.5 μl each of both the forward and reverse 16S rRNA V4 primers (10.0 μM). The reaction conditions for PCR were 94 °C for 10 min, and 27 cycles of 94 °C for 45 s, 53 °C for 60 s, and 72 °C for 90 s, followed by a final period of 72 °C for 10 min. The product was then purified using AMPure X (Beckman Coulter Inc, Mississauga, Ontario, Canada). The AMPure (20 μl) was mixed with the amplicon (25 μl) and incubated at room temperature for 2 min. After applying a magnetic field, the supernatant was discarded and the beads were washed twice with 80 % ethanol. The beads were then incubated at room temperature for 10 min before eluting with 50 μl of 10 mM Tris buffer.

Illumina universal adapters (Forward: AATGATACGGCGACCACCGAGATCTACAC-index-TCGTCGGCAGCGTC, Reverse: CAAGCAGAAGACGGCATACGAGAT-index-GTCTCGTGGGCTCGG) were then added to the purified 16S rRNA gene product by PCR using a 25 μl reaction consisting of 12.5 μl KAPA 2G Fast HotStart ReadyMix 2X (KapaBiosystems), 8.0 μl of molecular-grade water, 2.5 μl template DNA, and 1.0 μl each of the forward and reverse sample-specific Illumina universal adapters. The PCR conditions were as follows; a single cycle of 94 °C for 3 min, 8 cycles of 94 °C for 45 s, 50 °C for 60 s, and 72 °C for 90 s, followed by a final cycle of 72 °C for 10 min. The product was purified with AMPure X (Beckman Coulter Inc, Mississauga, Ontario, Canada) as previously described and DNA was eluted into 30 μl of 10 mM Tris buffer. The samples were then quantified by spectrophotometry (Nanodrop, Roche, Mississauga, Canada) and normalized to a final concentration of 2 nM. Sequencing of the library pool was performed using an Illumina MiSeq (San Diego, USA) and 2 × 250 chemistry at the University of Guelph’s Advanced Analysis Centre.

Analysis of sequencing data

Sequencing data were analyzed using the mothur software package v.1.33.0 [19]. The paired-end reads were aligned and screened to remove sequences with the following irregularities: contiguous sequence lengths >245 bp or <239 bp, ambiguous base calls, stretches of homopolymers >8 bp, and misalignment with the target region. Sequences were then screened for chimeras using the UCHIME tool [20] and sequences belonging to non-bacterial domains, including chloroplasts, mitochondria, Archaea and Eukaryotes, were removed. Operational taxonomic units (OTUs) were created using a 3.0 % dissimilarity threshold and average neighbour algorithm, then assigned taxonomy using the RDP reference database (http://rdp.cme.msu.edu). The observed relative abundance of taxa was analyzed prior to subsampling and the mean relative abundance of the 10 pigs was plotted graphically and in a chart to demonstrate changes in phyla, classes, and genera over time. Random subsampling of sequences from each individual sample was completed to normalize the sequence count. Community diversity (inverse Simpson index) and richness (CatchAll; [21]) were computed and sampling coverage was assessed by Good’s coverage value. The core microbiota (OTUs with >1 % relative abundance and shared amongst ≥80 % of pigs) was also explored for each sample type at different ages. The Jaccard index of dissimilarity was used as an assessment of community membership and the Yue and Clayton measure of dissimilarity was used to as an assessment of community structure. These indices were visualized by principle coordinate analysis, and compared between groups using analysis of molecular variance (AMOVA). Probabilistic modelling using Dirichlet multinomial mixtures (DMM) was utilized in examining community types associated with phase of production [22]. Additionally, the linear discriminant analysis (LDA) effect size (LEfSe) algorithm was used to identify OTUs associated with pre-weaning and post-weaning microbiotas [23] and cladograms were produced using the online LEfSe tool (http://huttenhower.sph.harvard.edu/galaxy/). Non-parametric statistical analysis (Spearman’s rank correlation) was performed in STATA 10.0 I/C (Stata Corporation, College Station, TX) to determine age-related differences in taxonomical abundance, diversity, and richness. The null hypothesis for all statistical tests was rejected at P < 0.05.