Influence of the Intestinal Microbiota on Colonization Resistance to Salmonella and the Shedding Pattern of Naturally Exposed Pigs

Salmonella is a global threat for public health, and pork is one of the main sources of human salmonellosis. However, the complex epidemiology of the infection limits current control strategies aimed at reducing the prevalence of this infection in pigs. The present study analyzes for the first time the impact of the gut microbiota in Salmonella infection in pigs and its shedding pattern in naturally infected growing pigs. Microbiome (16S rRNA amplicon) analysis reveals that maturation of the gut microbiome could be a key consideration with respect to limiting the infection and shedding of Salmonella in pigs. Indeed, seronegative animals had higher richness of the gut microbiota early after weaning, and uninfected pigs had higher abundance of strict anaerobes from the class Clostridia, results which demonstrate that a fast transition from the suckling microbiota to a postweaning microbiota could be crucial with respect to protecting the animals.

IMPORTANCE Salmonella is a global threat for public health, and pork is one of the main sources of human salmonellosis. However, the complex epidemiology of the infection limits current control strategies aimed at reducing the prevalence of this infection in pigs. The present study analyzes for the first time the impact of the gut microbiota in Salmonella infection in pigs and its shedding pattern in naturally infected growing pigs. Microbiome (16S rRNA amplicon) analysis reveals that maturation of the gut microbiome could be a key consideration with respect to limiting the infection and shedding of Salmonella in pigs. Indeed, seronegative animals had higher richness of the gut microbiota early after weaning, and uninfected pigs had higher abundance of strict anaerobes from the class Clostridia, results which demonstrate that a fast transition from the suckling microbiota to a postweaning microbiota could be crucial with respect to protecting the animals. KEYWORDS microbiome, colonization resistance, metagenome, pathogen, pig S almonella species is a ubiquitous enterobacterium which colonizes the intestine of animals (1). Nontyphoidal serovars such as Salmonella enterica serovar Typhimurium and Salmonella enterica serovar Enteritidis, which are frequently present in the gastrointestinal tract of production animals, are a major source of human salmonellosis (2), and recent studies identify pork products as one of the main sources of infection (3). Pigs are a natural host for Salmonella (4), and infection can occur at any production stage (5). With the exception of infections caused by Salmonella enterica serovar Choleraesuis, the serovar which is host adapted to pigs (6), the course of clinical infection is restricted to intestinal disease and is usually subclinical (7). Salmonella infection in swine is characterized by an early acute phase in which the pathogen is shed in relatively high concentrations in the feces (8). This then progresses to intermittent shedding or carriage with reactivation of shedding under adverse circumstances (9).
Salmonella epidemiological studies demonstrate that infected and noninfected pigs cohabit within the same herd (10,11). Differences in disease outcome are also observed in animals monitored during field trials (12,13). This phenomenon may reflect individual variability in susceptibility in pigs exposed to Salmonella under the same environmental conditions. Furthermore, infected animals exhibit variations in the concentration and duration of pathogen excretion in the feces (5,8,14). This could, at least in part, be due to colonization resistance, a concept first proposed in the 1950s (15,16) but which may be relevant to the interpretation of recent studies investigating the influence of the microbiome on disease outcomes (17). The principle underlying colonization resistance is that normal gut symbionts can form a barrier in the gastrointestinal tract, which limits the invasion of nonnative bacteria such as pathogens (18)(19)(20).
The observed differences in gastrointestinal colonization of pigs by Salmonella under natural conditions, including variations in shedding pattern and resistance to colonization, could therefore be ascribed, at least partially, to the resident microbiota of the host. Recently, high-throughput sequencing has enabled metagenomic cataloguing of pig intestinal samples, thereby providing insights into the microbial species present within the porcine intestinal tract (21)(22)(23). This has revolutionized our ability to study the gut microbiome, under different conditions, including deliberate Salmonella infection (24,25).
In this paper, we present a novel study of the fecal microbiome of naturally infected weaned pigs from a Salmonella-positive herd, in which apparent differences in susceptibility to Salmonella infection and divergence in shedding pattern among penmates were observed. The overall aim was to identify, for the first time, particular groups of bacteria associated with the outcome of Salmonella infection in pigs naturally infected with the pathogen.

RESULTS
Diversity of the microbiota in pigs categorized according to Salmonella infection-associated variables. After filtering, 16S rRNA amplicon sequences were assigned to 1,493 taxa across seven taxonomic ranks. Alpha-diversity of the fecal microbiota was measured using three different estimators ( Fig. 1; see also Table S1 in the supplemental material). The Shannon and Simpson indices (P Ͻ 0.01) but not Chao1 (P ϭ 0.103) revealed progressive increases in diversity at each sampling time point (Fig. 1A). Analysis of variables associated with Salmonella (infection status, serology, shedding pattern and shedding group) and alpha-diversity measures suggested a link between serological status and Shannon index value (P ϭ 0.0653, Fig. 1B). This trend was influenced by the significant differences observed at sampling 1 (18 days postweaning [pw]) when the Shannon index was clearly higher in seronegative pigs (P Ͻ 0.05, Fig. 1C), a result not observed at sampling 2 or sampling 3 ( Fig. 1B; Table S1). No differences in alpha-diversity of the fecal microbiota were observed for infection status, Salmonella shedding pattern (Fig. 1D), or shedding group during the course of the study (Fig. S1).
Influence of study variables on sample ordination. Bray-Curtis ( Fig. 2A and B) and weighted UNIFRAC distance analysis ( Fig. 2C and D) provided consistent insights in relation to the ordination of pig fecal samples according to the different factors under study. Fitting these environmental factors (infection status, shedding group, serology, pig or sampling time point, feed type) revealed an influence of sampling time point (P Ͻ 0.001) (linked in part to the change in diet [P Ͻ 0.01]) and serology (P ϭ 0.035) variables on the Bray-Curtis ordination of samples (Table S2). The strong influence of sampling time point on the ordination of samples was corroborated by performing a multivariate ANOVA based on dissimilarities (Table S3; P Ͻ 0.01). Neither Salmonella infection status nor shedding group influenced ordination of the pig fecal samples (P Ͼ 0.05; Table S3). Core microbiome analysis. The core microbiome was established for each category within the variables "infection status" and "shedding group" (Fig. 3). Data were not split by sampling time point, in order to obtain a complete picture of the bacteria defining the core microbiome of each group, irrespective of any modifications occurring over time. Salmonella-infected and noninfected pigs shared more than half of the operational taxonomic units (OTUs) included in the core microbiome, while 7 and 36 OTUs were present only within the core microbiome of infected and noninfected pigs, respectively (Fig. 3A). The seven unique OTUs found in the infected pigs belonged to the genera Lachnospira and Prevotella. Operational taxonomic units from these two genera were also present in the core microbiome of noninfected pigs. In contrast, the core microbiome of noninfected pigs included OTUs from Phascolarctobacterium, Roseburia, and Blautia, genera which were not present within the core microbiome of infected pigs. The core microbiome was also established for pigs categorized according to Salmonella shedding. Most of the OTUs were shared among the three shedding categories (A, B, and C). Twenty-three OTUs were part of the core microbiome in nonshedder pigs only (group A), 5 OTUs were unique to single-point shedders (group B), and 10 OTUs were unique to intermittent/persistent shedder pigs (group C). Operational taxonomic units from Roseburia, Lachnospira, or Phascolarctobacterium were present only in the core microbiome of group A, while the genus Oscillospira was present only within the core microbiome of shedding groups B and C.

Differences in abundance of OTUs between pigs in different categories. (i) Differences according to infection status.
When the fecal microbiota of Salmonellainfected and noninfected pigs was compared, we observed differences in abundance of OTUs belonging mainly to the phylum Firmicutes but also to Bacteroidetes and Euryarchaeota (Table 1; Table S4). Most of these OTUs belonged to the class Clostridia. An OTU classified within the family Ruminococcaceae and another two belonging to the genera Coprococcus and Lachnospira, both from the family Lachnospiraceae, were relatively more abundant (with a Ͼ1.5-log fold change) in noninfected pigs.
Analysis of infection status-associated abundance differences by sampling time point revealed changes in the genera linked to Salmonella infection throughout the study (Table S5). At sampling 1 (weaning period), OTUs in the family Lachnospiraceae were associated with noninfected pigs. Similarly, at this time point, the genus Sutterella was also more abundant in noninfected pigs, a result which was not observed at subsequent samplings. At the end of the weaning (sampling 2) and growing (sampling 3) periods, OTUs belonging to Ruminococcus and Prevotella were more abundant in noninfected pigs. Two genera were exclusively more abundant in infected pigs, Lactobacillus and Oscillospira, both at sampling 2.
(ii) Differences according to serological status. Table 1 lists the OTUs associated with the factor "serological status" (seropositive or seronegative pigs). Most of the OTUs found to be more abundant in seronegative pigs belonged to the phylum Firmicutes, in particular to the families Lachnospiraceae and Ruminococcaceae. Methanobrevibacter, Prevotella, Lachnospira, Anaerovibrio, and Ruminococcus were the genera most frequently linked to the seronegative category, although the latter also had an OTU significantly more abundant in seropositive pigs. In contrast, an OTU from the genus Lactobacillus was significantly more abundant (1.5 log) in seropositive than seronegative pigs (Table 1; Table S4).
Analysis of the microbiome by sampling time point revealed shifts in some of the main genera found to be differentially abundant between categories within the variable "serological status." While Lachnospira, Ruminococcus, and Prevotella OTUs were more abundant in seropositive pigs at the beginning of the study (sampling 1), their abundance shifted at subsequent samplings and we observed significantly higher abundance in seronegative pigs at the end of the weaning (sampling 2) and growing (sampling 3) periods for both genera.
(iii) Differences associated with Salmonella shedding pattern. Analysis of differences in OTUs among the three shedding groups (group A [nonshedders], group B [single-point shedders], and group C [intermittent and persistent shedders]) revealed four differentially abundant OTUs. However, the limited number of pigs in which these OTUs were present, as well as the low abundances observed, calls into question the relevance of this finding (Table S4). We therefore further analyzed differences using the shedding group variable by limiting the analysis to differences in abundance between groups A and C. No differences in clustering by differentially abundant OTUs were observed between groups A and C (Fig. 4A). In fact, samples from both groups were equally distributed within the two major clades observed in the heat map. However, we did observe particular differences for several taxa. The genera Lactobacillus and Succinivibrio were more abundant in group C pigs (intermittent/persistent shedders) while OTUs belonging to the families Lachnospira and Ruminococcaceae were more abundant in group A pigs (nonshedders). Inconclusive results were observed for Prevotella OTUs, which were significantly abundant in both groups ( Table 1). Analysis of data by sampling time point ( Fig. 4B; Table S5) showed that Coprococcus abundance was associated with nonshedder pigs (group A) across all three samplings performed (Fig. 4C). At sampling 1, the genera Treponema and Fibrobacter were also linked to nonshedders while in subsequent samplings at the end of the weaning and growing stages Prevotella and Dialister were more abundant in nonshedder pigs. Lactobacillus was more abundant at sampling 1 in single-point shedder pigs (group B) as were some OTUs of Prevotella at samplings 2 and 3 (Fig. 4C). The order YS2 (Cyanobacteria) and the genus Anaerobiospirillum were noticeably increased in the intermittent/continuous shedder group (group C). In particular, the genus Anaerobiospirillum was increased in abundance in group C at sampling 1 (Fig. 4B).
Differentially abundant OTUs shared across the Salmonella infectionassociated variables under study. Finally, we compared the data sets for the OTUs that were differentially abundant according to the three Salmonella infectionassociated variables studied. Five OTUs were shared among those differentially abundant according to the variables "infection" and "serological status" ( Table 2). It is noteworthy that noninfected and seronegative pigs shared OTUs from the genera Lachnospira, Ruminococcus, and Methanobrevibacter. Similarly, the same OTUs belonging to the families Lachnospiraceae and Ruminococcaceae and the genus Prevotella were present in seronegative and nonshedder (group A) pigs (Table 2). On the other hand, seropositive and intermittent/persistent Salmonella shedder (group C) pigs shared a differentially abundant OTU from Lactobacillus.

DISCUSSION
The pig gastrointestinal tract is colonized by many different types of microorganisms which contribute to a range of host physiological processes, such as metabolism, integrity of the epithelial barrier, immune homeostasis, and protection against pathogens (18,20,26). Competition for niche and nutrients and production of bacteriocins and metabolites are considered the main mechanisms of pathogen exclusion by the resident microbiota (27). Recent studies using 16S rRNA gene sequencing have demonstrated the role of the commensal microbiota in conferring resistance to gut colonization by pathogens (19,26). In addition, there are already findings from deliberate infection studies in pigs that suggest the influence of microbiota composition on the concentration of Salmonella excreted in pig feces (24). Here, for the first time, by parallel monitoring of the Salmonella infection and shedding status and microbiome composition of pigs naturally exposed to Salmonella under field conditions, we aimed to identify microbial communities associated with infection susceptibility. Disease susceptibility was defined by the combination of data, shedding of the pathogen in feces, and development of specific anti-Salmonella antibodies, together with three potential shedding scenarios. It was anticipated that the combination of this information and microbiome data would provide insight into the differences in infection outcome observed among individuals, thereby informing new control strategies for Salmonella in pigs.
Microbiome diversity may prevent early infection of pigs postweaning. The outcome of Salmonella infection in pigs is usually evaluated through direct methods which include microbiological detection of the bacterium and indirect methods which look for markers such as antibodies within the host (14). The two methods complement each other and were used in the present study to categorize pigs according to their infection outcome. Interestingly, microbiome diversity and ordination were associated with differences in the serological status of the animals. Two weeks after weaning, diversity of the fecal microbiota was higher in seronegative pigs, i.e., those without antibodies (anti-LPS IgG) to the pathogen. Two factors increase disease susceptibility postweaning: (i) the transition from milk to a solid diet, which shifts the microbial composition of the gut toward what is often considered a transitory dysbiosis (28); and (ii) the loss of maternal protective immunoglobulins provided by sows' milk (29). Therefore, taking our results and the first factor above into consideration, early establishment of a diverse and healthy microbiota may hamper the colonization success of pathogens such as Salmonella. The suggestion that gut health correlates with microbiome richness is in line with previous studies (30,31). This is further supported by our study, which also showed an association between serological status and microbial ordination. From our results, we can infer that a more complex early-life microbiome may provide a more challenging environment for pathogens such as Salmonella, limiting their infective capacity and thereby preventing intestinal invasion and activation of the humoral immune response (32).

Members of the class Clostridia may prevent colonization of the gastrointestinal tract by Salmonella. Members of the class Clostridia, including the genera
Roseburia and Blautia from the family Lachnospiraceae and the genera Ruminococcus and Anaerovibrio, were more abundant in Salmonella-negative pigs. Furthermore, Roseburia and Blautia (Lachnospiraceae) together with the genus Phascolarctobacterium were part of the core microbiome of noninfected pigs. Previous studies link these taxa to a healthy gut configuration in mammals (25,(33)(34)(35). In addition, metagenomic studies have found a negative correlation between anaerobe counts and epithelial damage in the ileal mucosa (25) and a higher abundance of Ruminococcaceae prior to challenge in pigs shedding low Salmonella concentrations (24). Another common feature of these genera is that they are preferentially or strict anaerobes and producers of short-chain fatty acids (SCFA) (33,(35)(36)(37). These two factors, together with limited oxygen and high concentrations of SCFA, such as butyrate, prevent the expansion of facultatively anaerobic Enterobacteriaceae such as Salmonella (38). The dominance of anaerobes in Salmonella-negative pigs is supported by the higher counts of the genus Methanobrevibacter in seronegative pigs. Methanogens such as Methanobrevibacter obtain energy from hydrogen molecules produced by strict anaerobes (39), and their abundance is linked to the presence of these anaerobes (20). In contrast, Salmonellainfected pigs had higher counts of Lactobacillus and Oscillospira OTUs. Lactobacillus and Oscillospira are both characteristic of the gut microbiome of nursing pigs (40,41). We propose that their presence in high counts in feces 18 days pw (ϳ6 weeks of age) could be indicative of immaturity of the microbiota and a lack of commensal organisms that restrict Salmonella colonization. Another potential explanation could be the presence of more favorable conditions for Lactobacillus growth in infected pigs, a result already reported by Drumo and colleagues (42). However, the mechanisms by which Salmonella infection might boost the growth of these taxa remain unknown.
Changes in the competitive microbiota throughout different life stages of the pig. Under intensive production conditions, the microbiota of the pig gastrointestinal tract evolves, shifting from Bacteroidetes to Firmicutes as the animal grows (40,43). Putative components of the microbiota which participate in competitive exclusion of pathogens may also vary from one life stage to another. In our study, we also observed a strong effect of sampling time point on microbiome composition. Thus, sampling time point was first used as a cofactor in the statistical analyses, and then longitudinal differences in microbiota abundance were analyzed over time. Early after weaning (18 days pw; ϳ6 weeks of age), we observed a higher abundance of the genus Sutterella in noninfected pigs. Sutterella is a member of the Proteobacteria which predominantly inhabits the small intestine, at least in humans (44). Although its role in intestinal health is unclear, it is considered a commensal, is capable of adhering to the epithelium, and has a mild proinflammatory capacity. Members of this genus could therefore potentially occupy the niche of pathogenic Proteobacteria or positively stimulate the immune response, although more research is needed to explore this. In samplings at the end of the weaning (ϳ8 weeks of age) and growing (ϳ11 weeks of age) periods, OTUs of Ruminococcus and Prevotella were enriched in the noninfected pigs. Both genera are cellulolytic microbiota that increase in abundance during maturation of the gut microbiota in pigs (40,45). These two genera were also part of the healthy gut configuration in patients who had recovered from Vibrio cholerae infection (46,47), adding further support to the theory proposed in the present study that anaerobic cellulolytic SCFA-producing bacteria limit the success of Salmonella in colonizing the pig gastrointestinal tract.
Persistence of Salmonella shedding may be influenced by the presence of synergistic bacteria. The final goal of the present study aimed to investigate the role of the gut microbiome in determining the distinct shedding patterns observed in the animals under study. After an acute phase of infection, characterized by continuous shedding of high concentrations of Salmonella (48), nontyphoidal Salmonella infection in pigs progresses to a chronic phase, with no evident clinical signs and intermittent shedding of the pathogen in feces. This is a result of the combination of lower concentrations of Salmonella in feces and the limitation of microbiological methods to detect this low burden of Salmonella (49). By monitoring naturally infected pigs throughout the weaning and growing periods, we observed differences in Salmonella shedding patterns from single-point shedder pigs to a pig which was positive at all five samplings performed. Although clustering analysis by abundance of OTUs did not split samples according to the shedding group, there were differences in particular taxa. In addition to higher abundance of Lactobacillus, both in single-point and in intermittent/ continuous shedder groups, we observed another two taxa that were overrepresented in these groups: the genus Anaerobiospirillum in single-point shedder pigs and the order YS2 in intermittent/continuous shedder pigs. Anaerobiospirillum can be considered a pathobiont, an indigenous microbe that is able to promote disease under certain circumstances (18,50), and its presence has been linked to diarrhea in humans (51,52). However, little is known about YS2, an order included in the phylum Cyanobacteria, and the reason why it is more abundant in the gastrointestinal tract of pigs shedding Salmonella is unclear.
Conclusion. Field studies allow the investigation of diseases under natural conditions, revealing interesting information which may not come to light in challenge studies. The present field study is the first in which Salmonella colonization resistance was studied in pigs. Although the study includes a limited number of animals, the information that they provide is of potentially great value. Our results suggest that early establishment of a diverse core microbiome enriched in anaerobes capable of producing SCFA metabolites and subsequent enrichment of cellulolytic bacteria may impede Salmonella gut colonization and invasion and limit fecal shedding. On the other hand, a lack of maturation of the microbiome, with a predominance of microorganisms normally associated with suckling, may increase susceptibility to infection and persistence of pathogen shedding in the feces. Overall, these results suggest that certain taxa within the porcine intestinal microbial community could potentially be targeted in the future to manipulate the intestinal microbiome so as to increase resistance to infection with Salmonella in pigs.

MATERIALS AND METHODS
Study design. The study was conducted under license from the Department of Health and Children (number B100/2982) and received ethical approval from the University College Dublin Animal Research Ethics Committee (AREC 13-37).
The study was performed on an Irish 750-sow commercial farrow-to-finish farm with endemic Salmonella (S. Typhimurium) infection. One batch of 15 pigs was monitored from approximately 2 weeks after weaning (18 days pw orϳ45 days of age) to the end of the growing phase (48 days pw or ϳ75 days of age) (Fig. 5). Seven days after pigs were transferred to the weaning facility, eight pens were swabbed using sponges, as previously described (53), and tested for the presence of Salmonella, as outlined below. Pigs from two Salmonella-positive pens (7 pigs from one pen and 8 from the other) were ear tagged for identification purposes and fecally sampled by digital rectal stimulation 18 days pw (sampling 1) and on four further occasions during the weaning and growing periods (Fig. 5). All fecal samples were tested for the presence of Salmonella as outlined below, while for microbiota analysis a subsample was flash-frozen in dry ice at sampling 1, sampling 2, and sampling 3 and stored at Ϫ80°C until analysis. Blood samples were collected at the beginning (sampling 1) and end (sampling 3) of the study by jugular venipuncture using whole-blood plastic Vacutainers (BD Vacutainer; Becton Dickinson, Oxford, United Kingdom) for the detection of Salmonella antibodies, as outlined below. Throughout the course of the study, none of the study animals received antibiotics or displayed clinical signs compatible with Salmonella infection or any other disease of pigs. Animals received two different diets during the weaning and growing stages, respectively.
Salmonella detection in pen swabs and feces. All pen swabs and fecal samples were tested for the presence of Salmonella according to Annex D of the ISO 6579 method, as previously described (53). Based on the results from the fecal samples, pigs were classified into four "shedding patterns" (Fig. 5): (i) nonshedders (7 pigs), defined as animals which were Salmonella negative at all samplings; (ii) single-point shedders (2 pigs), pigs which shed Salmonella at only one of the sampling time points; (iii) intermittent shedders (5 pigs), defined as pigs which shed Salmonella at between two and four of the sampling time points; and (iv) persistent shedders (1 pig), which were pigs that shed Salmonella at all samplings. These shedding patterns were grouped into three "shedding groups," group A (nonshedders), group B (single-point shedders), and group C (intermittent and persistent shedders), for statistical analysis of microbiome differences.
Salmonella serological analysis and definition of infected/noninfected pigs. Serum was obtained from the blood samples and analyzed in duplicate using an in-house indirect enzyme-linked immunosorbent assay (ELISA) as previously described (53). Briefly, the method detects immunoglobulin G against the O side chain of the lipopolysaccharide of Salmonella (54). Optical density percentages (OD%) were determined by relating each serum absorbance value at 650 nm to that of the positive control.
According to the ELISA results, pigs were defined as seropositive (8 pigs) when the ELISA OD value was Ն10% and seronegative (7 pigs) when the OD value was Ͻ10% (Fig. 5).
By combining the Salmonella shedding and serological data, pigs were defined by a new variable referred to as "infection status": "noninfected" pigs (6 pigs) were defined as those which were bacteriologically and serologically negative on all occasions, while "infected" (9 pigs) comprised all other pigs which were Salmonella positive in the feces, seropositive, or both.
16S rRNA amplicon sequencing of fecal microbiota. Total DNA was extracted from all fecal samples (ϳ200 mg) using the QIAamp DNA stool minikit (Qiagen, Crawley, United Kingdom) according to the manufacturer's instructions apart from adding a bead-beating step after sample addition to the InhibitEX buffer and increasing the lysis temperature to 95°C to increase the DNA yield (23). All samples were prepared for MiSeq compositional sequencing using the specifications outlined by Illumina (Illumina Inc., Cambridge, United Kingdom). The V3-V4 region of the 16S rRNA gene was amplified, and Illumina index primers was . These shedding patterns were grouped into three "shedding groups": group A (nonshedders), group B (single-point shedders), and group C (intermittent and persistent shedders). Serology cutoff was set at 40% of ELISA optical density. Finally, pigs were categorized by "infection status," with noninfected pigs referring to those that were nonshedders as well as seronegative and infected referring to the other combinations. attached in two separate PCRs (55). All PCR conditions and cleanup procedures using AMPure XP (Labplan, Kildare, Ireland) were as outlined by Illumina. Quantified samples were then sequenced using an Illumina MiSeq system at the Teagasc Sequencing Centre (Fermoy, Ireland).
Bioinformatic processing and analysis. Raw sequence reads generated by MiSeq were processed using version 1.9.1 of the Quantitative Insights Into Microbial Ecology (QIIME) pipeline (56) by using the subsampled open-reference OTU calling approach (57). Demultiplexing and trimming of MiSeq reads were performed using the default QIIME parameters (58). After trimming, the reads were merged into a single FASTA file and clustered into OTUs against the Greengenes database (59) (release 2013-08; gg_13_8_otus) by using the parallel uclust_ref method. Reads that failed this step were clustered into de novo OTUs using the uclust method (60). The filtering of chimeric OTUs was performed using Chime-raSlayer (61) against the Greengenes reference alignment. After removing singleton and doubleton OTUs, only those OTUs representing Ͼ0.005% of the total filtered were retained as suggested by Bokulich et al. (58). For analysis at the genus level, OTUs were collapsed into genus taxonomic level using the tax_glom function in Phyloseq (62).
Statistical analysis. Statistical analyses were performed in R v3.4.2. Microbiota and study variables (shedding pattern, shedding group, serology, infection status, pen, and sampling time point) were included in the estimation of alpha-diversity richness (Shannon, Simpson, and Chao1 indices) by the Nmle, Vegan, and Phyloseq R packages (62,63). For richness values, assumption of normality was checked using the Shapiro-Wilk test, and potential differences in richness of factors included in the study were estimated by repeated-measures analysis of variance (ANOVA), using either sampling time point or pig as a cofactor and a Tukey multiple-comparison test. Dissimilarities in beta-diversity between pairs of samples were estimated with the Bray-Curtis dissimilarity index (64) and weighted UniFrac index (65) and analyzed with nonlinear multidimensional scaling (NMDS) in Vegan. The Vegan envfit function, which fits environmental vectors or factors onto an ordination, was used to evaluate if the factors sampling day and infection status were associated with the NMDS ordinations; the significance of the fitted factors was estimated using 999 permutations. Core microbiome was established as those OTUs present in Ն50% of the samples and presenting a mean relative abundance of Ͼ1% within the corresponding groups, using the core function in the R package Microbiome (66). Differences in taxon abundance were analyzed after OTU count normalization by calculating the scaling factors equal to the sum of counts in the metag-enomeSeq R package (67). The zero-inflated, log-normal distribution (fitFeatureModel function) and the zero-inflated Gaussian distribution mixture-model (fitZig function) were used to estimate differences in variables under study, using sampling time point and pig factors as covariates and with a false-discovery rate (FDR) threshold of 0.05.
Accession number(s). The full data sets have been submitted under BioProject accession no. PRJNA521510.
FIG S1, TIF file, 1.1 MB.  We gratefully acknowledge the Central Veterinary Research Laboratory (CVRL) Backweston for their expert help with the serological analyses.
H.A. participated in the study design, sample collection and processing, data analysis, and manuscript writing. F.C.L. participated in study design and manuscript correction. J.E. participated in data analysis and manuscript revision. P.D.C. and F.C. participated in microbiome sequencing and manuscript revision. O.O. performed part of the data analysis and manuscript revision. H.L. and K.W. participated in sample collection and processing. G.D. and P.G.L. participated in the study design and manuscript correction. G.E.G. participated in the study design, data analysis, and manuscript writing.