A cross-sectional study of the nasal and fecal microbiota of sows from different health status within six commercial swine farms

Study population and sampling protocol

This study has been approved by The Ohio State University’s Institutional Animal Care and Use Committee (IACUC Protocol #2017A00000060), as well as by the Institutional Biosafety Committee (IBC Protocol #2017R00000041).

A cross-sectional observational study was conducted during July of 2017. The project included enrollment of sows distributed within six privately owned commercial farrow-to-wean swine farms from the same production company located in the US Mideast region. A commercial farm was defined as a swine farm that raised pigs with the intent to commercialize them (in this case of farrow-to-wean farms, the farms were producing and selling weaned piglets to downstream farms that would raise them to market). All sows were fed commercial diets from the same source that met the Nutrient Requirements of Swine (National Research Council, 2012). Basic farm descriptors are shown in Table 1. Healthy sows were defined as adult females with no evidence of clinical signs of disease at the time of data collection. Healthy sows were included regardless of stage of production (i.e., pregnant, lactating or weaned). Cull sows were defined as adult females that were identified to be removed from the herd based on the standard operating procedure implemented by the farm staff. Investigators were using sow cards (paper records) in order to identify culled sows. The primary reason for animals to be culled was if they did not achieve acceptable reproductive performance. Compromised or clinical sows (the two terms are used interchangeably throughout the manuscript) were defined as sows presenting acute clinical signs of disease that could be visually assessed by the attending veterinarian or farm staff. This included the following categories: lameness, severe shoulder ulcers, active abortion, and poor body condition score. Each farm contributed an equal number of sampled sows for each health status examined herein (healthy, cull and compromised / clinical); with 10 sows sampled for each health status in each farm.

Farms were visited once during the month of July of 2017 for collection of fecal samples and nasal swabs from all sows within a farm on the same day. Fecal samples were used in this study as a non-invasive way to examine the gut microbiota. A total of 30 fecal samples and 30 nasal swabs were obtained per farm (ten for each type of animal) for a total of 360 samples (180 fecal samples and 180 nasal swabs). A schematic of sampling design is presented in Fig. 1, along with relative geographical location for the sampled farms.

For nasal swab sampling, pigs were restrained using a snare and the swab (Puritan™ PurFlock™ Ultra Sterile Flocked Swabs) was inserted into the ventral passage of each nostril. The nasal swab was kept in the cavity for three seconds (per nasal passage) and placed in a dry tube. Following this, a fecal sample was collected using digital palpation of the rectum to remove feces with a sterile glove. Fecal samples were then placed into sterile bags.

Sample pooling, genomic DNA extraction and sequencing

Fecal and nasal swabs were collected and placed in tubes on dry ice immediately after collection. Samples were then transported to the Food Animal Health Research Program (The Ohio State University) laboratory in Wooster, Ohio, and transferred into a −80 °C freezer within 24 h, where they remained until extraction. For nasal swabs, two milliliters of phosphate-buffered saline (PBS) were added to each nasal swab tube, tubes were vortexed for 30 s. The samples were divided into 2 subsamples: one 1 ml sample was stored at −80 °C for future use and the other 1 ml sample was centrifuged at 13,000 g for 15 min. The resulting pellet was resuspended in 250 µl of PBS and used for DNA extraction. Fecal and nasal swab samples were pooled across 5 animals of the same status by farm; so each farm contributed two pools of each health status containing samples of five animals each, as shown in Fig. 1. The samples selected to be part of the same pool were randomly selected, and the nasal and fecal pool sets were composed of the same animals. The pools consisted of 0.5 ± 0.015 g of feces per sample and 45 µl of nasal swab suspension per sample. There were 72 pooled samples in total (36 fecal and 36 nasal swab pooled samples) and the pooled sample was considered our unit of analysis. Genomic DNA extraction of the pooled fecal (0.5 ± 0.015 g) or nasal (225 µl) swab samples was conducted using the PureLink Microbiome DNA Purification Kit (Life Technologies, Invitrogen Corp.) followed by RNAse treatment (10 units for 1 h), as previously described (Deblais et al., 2018). Four blank (empty) extractions were also performed as negative controls and sequenced along with the rest of the samples. It is important to note that this commercial kit lyses the cells using chemical (lysis buffer + enhancer [guanidine isothiocyanate]), thermic (65 °C for 10 min), and mechanical (bead beating at full speed for 10 min) approaches. Next generation sequencing library preparations and Illumina MiSeq sequencing were conducted at the Bioscience Division, Los Alamos National Laboratory, NM.

Briefly, DNA samples were quantified using a Qubit 2.0 Fluorometer (Invitrogen, Carlsbad, CA) and DNA quality was checked on a 0.6% agarose gel using E-gel electrophoresis system (ThermoFischer Scientific, Grand Island, NY). A total of 50 ng DNA was used to generate amplicons of the V3–V4 hypervariable regions of the bacterial and Archaeal 16S rRNA gene using primers 341F and 806R. The first round of PCR amplified the V3-V4 region using KAPA HIFI HotStart Ready Mix (Kapa Biosystems, Wilmington, MA) with following PCR conditions; 95 °C for 3 min, 20 cycles of 95 °C for 30 s, 55 °C for 30 s and 72 °C for 30 s and an extension of 72 °C for 5 min. The second round of PCR added the Illumina specific sequencing adapter sequences and unique indexes using the Nextera XT Index Kit v2 (Illumina, San Diego, CA) and KAPA HIFI HotStart Ready Mix (Kapa Biosystems, Wilmington, MA) with following PCR conditions; 95 °C for 3 min, 8 cycles of 95 °C for 30 s, 55 °C for 30 s and 72 °C for 30 s and an extension of 72 °C for 5 min. The amplicons were cleaned up using AMPure XP beads (Beckman Coulter, Indianapolis, IN). The concentration of the amplicons pool was obtained using the Qubit dsDNA HS Assay (ThermoFisher Scientific, Grand Island, NY). The average size of the library was determined by the Agilent High Sensitivity DNA Kit (Agilent, Santa Clara, CA). An accurate library quantification was determined using the Library Quantification Kit –Illumina/Universal Kit (KAPA Biosystems, Wilmington, MA). The amplicon pool was sequenced on the Illumina MiSeq generating paired end 300 bp reads. The amplicon pool was demultiplexed using Illumina’s bcl2fastq. Sequencing data are publicly available at NCBI SRA: PRJNA674564.

Taxonomic microbiota analyses and statistical analyses

Raw, single-end sequence reads were processed using QIIME2 v. 2021.4 (Bolyen et al., 2019). The DADA2 plugin was used to truncate reads at 280 bp and to denoise the single-end reads (Callahan et al., 2016). We opted for single end reads over paired end due to the poor quality of the R2 reads. Taxonomy was assigned in QIIME2 using SILVA version 138.1, with a 99% similarity threshold using the full length 16S rRNA gene classifier (Quast et al., 2013; Yilmaz et al., 2014). All fecal samples with fewer than 5,600 reads and all nasal samples with fewer than 500 reads were removed from analyses. To compare alpha diversity between nasal and fecal samples, all samples were rarefied at 500 reads. We also rarefied at 500 reads to compare nasal samples by farm and health status. To compare fecal samples by farm and health status, we rarified at 5,600 reads. In total, we retained all fecal samples (n = 36) and 19 out of 36 nasal samples for further analysis. To assess the effects of rarefaction, we also analyzed fecal and nasal samples without rarefaction (excluding fecal samples with fewer than 5,600 reads and nasal samples with fewer than 500 reads). We also analyzed nasal samples (rarefied and non-rarefied) excluding samples with fewer than 1,000 reads.

Alpha diversity was analyzed using Shannon and Faith’s Phylogenetic Diversity (PD) indices. A Kruskal–Wallis test was used to compare alpha diversity and read numbers by health status and farm. An adjusted P-value lower than 0.05 was considered significant.

Beta diversity indices were compared using permutational multivariate analysis of variances (PERMANOVA) and permutational analysis of multivariate dispersions (PERMDISP) between Bray Curtis distance matrices. P-values were corrected for multiple comparisons using the Benjamini–Hochberg FDR correction, and values less than 0.05 were considered significant. PERMANOVAs were performed on the fecal and nasal samples by farm and by health status. An analysis of composition of microbes (ANCOM) was used to identify differentially abundant nasal and fecal taxa by farm and health status. An ANCOM performs pairwise comparisons of microbial species ratios and generates a W-value, which indicates how many times a null hypothesis is rejected across comparisons. The maximum W value in any data set is limited by the total number of microbial taxa within that data set. For all samples, we filtered out taxa that had fewer than 10 reads and occurred in fewer than two samples prior to ANCOM analysis. No additional filtering was performed on nasal samples due to the low microbial diversity in the majority of samples. ANCOMs were performed at the L7 level, which is roughly equivalent to a species level; although deeper genome sequencing is necessary for true species differentiation.

16S rRNA gene amplicon sequencing

We obtained a total of 1,759,761 reads with a mean read length of 294 base pairs (Table S1). Of the four negative controls, only one yielded any reads (3 reads total), and all 3 reads were identified as chloroplasts. Chloroplasts, mitochondria, and eukaryotic reads were then removed bioinformatically prior to further analyses. A total of 493 taxa were identified across all samples. The mean number of reads per nasal swab pool (n = 36) was 2,572; the median was 914.5, with a range of 1–27,870 reads (Table S2). Nasal samples with fewer than 500 reads (n = 17) were excluded from analysis. For the 36 pooled fecal samples, we obtained a total of 901,440 reads. The mean number of reads per fecal pooled sample was 25,040, and the median was 22,256, with a range of 5,694–62,677 reads (Table S3). All fecal samples were included in our analyses, which were performed rarefied at 5,600 reads and unrarified. To facilitate a direct comparison between nasal and fecal samples, we also performed a separate analysis with fecal and nasal samples together rarefied at 500 reads and unrarefied.

Alpha diversity results

Overall, nasal samples harbored a significantly lower microbial diversity compared to the fecal samples (Faith’s PD, p < 0.001; Shannon, p < 0.001; Fig. S1A). No significant differences in fecal or nasal diversity (Shannon index) were detected by health status (healthy, cull and compromised/clinical) or by farm (Kruskal–Wallis, all p > 0.05, Fig. 2, Fig. S1B, Table S4).

Beta diversity results

There were significant differences in overall nasal microbial composition by farm (Bray–Curtils PERMANOVA, pseudo-F = 1.83, p = 0.002, Fig. 3A, Fig. S2A, Table S4); although, no pairwise comparisons were significant. Overall nasal microbial composition by health status also differed significantly (PERMANOVA, pseudo-F = 1.64, p = 0.017, Fig. 3B, Fig. S2B) with healthy versus cull as the only significant pairwise comparison (p = 0.033). These differences were not explained by dispersion (PERMDISP by farm: pseudo-F = 0.577, p = 0.795; by health status: pseudo-F = 2.11, p = 0.148) or number of reads by farm or health status (Kruskal–Wallis by farm: p = 0.36; by health status: p = 0.28). When we examined the pooled fecal microbiota of sows, we found significant differences in overall microbial composition by farm (PERMANOVA, pseudo-F = 3.16, p = 0.001, Fig. 4A, Fig. S3A, Table S4) with significant pairwise differences between all farms (p < 0.05) with the exception of farms 4 and 5 (p = 0.08) and farms 5 and 6 (p = 0.1). No significant differences in pooled fecal microbial composition were found based on health status (p = 0.663, Fig. 4B, Fig. S3B).

In terms of relative abundances, the most abundant phyla in the pooled nasal samples was Proteobacteria, followed by Firmicutes, Actinobacteria and Bacteroidetes (Fig. 5A). The most abundant nasal genera across all pooled samples was Moraxella. At the L7 level, we identified one differentially abundant nasal taxa by farm: a taxa in the Micrococcaeae family, genera Nesterenkonia (ANCOM, W = 32), which was only present on farm 6 (Table S5). No nasal taxa were differentially abundant by health status. In terms of relative abundance in the pooled fecal samples, Firmicutes was the most abundant phyla, followed by Bacteroidetes, Euryarchaeotes, Spirochetes, Actinobacteria and Proteobacteria (Fig. 5B). Within the Firmicutes phyla, the most abundant family was Lachnospiraceae, and the most abundant genera across all fecal samples were Lactobacillus and Methanobrevibacter (an Archaea). At the L7 (roughly species) level in fecal samples, we identified one differentially abundant taxa by farm: it was a taxa in the order Bacteroidales, genera F082 (ANCOM, W = 140), which was broadly present at low abundances across all samples from farms 1, 4, 5 and 6 but almost entirely absent on farms 2 and 3 (Table S6). No fecal taxa were differentially abundant by health status.