Alterations of fecal microbiome characteristics by dietary soy isoflavone ingestion in growing pigs infected with porcine reproductive and respiratory syndrome virus

Abstract

Porcine reproductive and respiratory syndrome virus (PRRSV) is an economically important disease, and the ingestion of soy isoflavones (ISF) may benefit PRRSV-infected pigs due to demonstrated anti-inflammatory and antiviral properties. The objective of this study was to quantify the effects of ISF consumption on fecal microbiome characteristics at different timepoints across a disease challenge and determine whether any changes, if present, elude to potential biological mechanisms for previously observed performance benefits. In total, 96 weaned barrows were group-housed in a Biosafety Level-2 containment facility and allotted to one of three experimental treatments that were maintained throughout the study: noninfected pigs receiving an ISF-devoid control diet (NEG, n = 24) and infected pigs receiving either the control diet (POS, n = 36) or that supplemented with total ISF in excess of 1,600 mg/kg (ISF, n = 36). Following a 7-d adaptation, pigs were inoculated intranasally with either a sham-control (phosphate-buffered saline) or live PRRSV (1 × 10⁵ median tissue culture infectious dose[TCID]₅₀/mL, strain NADC20). Fecal samples were collected from 48 individual pigs at pre-infection (−2 d post-inoculation [DPI]), peak-infection (10 DPI), and post-infection (144 DPI) timepoints. Extracted DNA was used to quantify fecal microbiota profiles via 16S bacterial rRNA sequencing. Differences in bacterial communities among diet groups were evaluated with principal coordinate analysis and permutational multivariate analysis of variance using UniFrac distance matrices based on both unweighted and weighted UniFrac distances using QIIME 2. All other data were analyzed by one-way ANOVA performed on square root transformations using R. Across all timepoints, only a few differences were observed due to ISF alone mainly in lowly abundant genera. The most notable differences observed were decreased relative abundance of Actinobacteria at 144 DPI between noninfected and infected treatments (P < 0.05), which is consistent with various dysbioses observed in other disease models. Our findings indicate that the differences present were mainly due to PRRSV-infection alone and not strongly influenced by diet, which implies that previously observed performance benefits conferred by dietary ISF are not likely due to the changes in microbiome composition.

Introduction

Porcine reproductive and respiratory syndrome virus (PRRSV) continues to be one of the leading pathogens with a high economic impact in modern swine production (Holtkamp et al., 2013). Due to its prevalence and a lack of highly effective vaccines (Loving et al., 2015), nutritional intervention or management strategies are of high interest for PRRSV mitigation due to relatively low cost and ease of implementation. Previously, it has been demonstrated that increased levels of soybean meal in diets fed to pigs under immune stress reduce the adverse effects of illness and may improve growth performance (Boyd et al., 2010; Rochell et al., 2015). However, it still remains unclear if improved growth performance in those studies were due to differences in dietary bioavailable amino acid content or the presence of other dietary components. Among these components, soy isoflavones (ISF) are naturally occurring flavonoid compounds that possess anti-inflammatory and anti-oxidative properties (Smith and Dilger, 2018). Under PRRSV-challenge, weaned pigs that received dietary ISF have previously exhibited improved clinical and immune responses to PRRSV, including improved growth performance, reduced viral load, and improved adaptive immune cell responses by measurements of effector T-cell proportions during peak infection periods (Greiner et al., 2001; Smith et al., 2019). More recently, when ISF were fed to growing pigs infected with PRRSV over the entire growth period from weaning to market, dietary ISF improved PRRSV clearance during recovery and reduced pathogen-related mortality by approximately 50% (Smith et al., 2020).

In light of these effects on the response to PRRSV-infection in an in vivo model, it is unclear if dietary ISF and their metabolites are producing these effects directly following absorption into the body or if these effects are due to alterations of the gastrointestinal microbiome. While it is fairly well accepted that the gastrointestinal microbiome can influence systemic immune system function through the “gut-lung axis” (Collado et al., 2018), there is also evidence that soy ISF can influence the composition of the gastrointestinal microbiome (Nakatsu et al., 2014; Huang et al., 2016; Paul et al., 2017; Vázquez et al., 2017; Kolátorová et al., 2018). These alterations may impact the modulation of immune function and thus the systemic response to infections. For those reasons, the objective of this experiment was to evaluate how consumption of soy ISF by PRRSV-infected pigs affected the composition of the gastrointestinal microbiota at three timepoints across the entire growth period. Our hypothesis was that both PRRSV-infection status and consumption of ISF would result in detectable differences in global microbiome composition in growing pigs.

Materials and Methods

The protocol for this experiment was approved by the Institutional Animal Care and Use Committee and the Institutional Biosafety Committee of the University of Illinois at Urbana-Champaign.

Animal husbandry and experimental design

This experiment is a continuation of the one described by Smith et al. (2020) and utilized the same animals for sample collection. In total, 96 weanling pigs (96 barrows; 6.77 ± 0.94 kg initial body weight) were obtained from a PPRSV-negative, non-PRRSV-vaccinated commercial herd (1050 Cambro genetics; Carthage Veterinary Service, Ltd.) and group-housed at the Innovative Swine Solutions Veterinary Research Facility (ISS-VRF), a Biosafety Level 2 production containment facility in Champaign, IL. Pigs were managed at this facility through market weight (approximately 27 wk of age, 118 to 129 kg average body weight) and then transported to the Meat Science Laboratory at the University of Illinois for slaughter. The ISS-VRF consisted of four individual production rooms (9.1 × 4.3 m) with four pens (up to six pigs per pen) available to house pigs in each room. Each room was equipped with a high-efficiency particulate air filtration system, plastic slotted flooring appropriate for market-sized pigs, and each pen provided appropriate floor space for pigs taken to market weight. Lighting and ambient temperature were maintained as appropriate for life stage.

Upon entry to the facility, pigs were weighed and uniformly allotted to one of three experimental treatments (n = 24 for non-PRRSV-infected controls [NEG], n = 36 each for PRRSV-infected controls [POS], and PRRSV-infected receiving soy isoflavones [ISF]) by completely randomized design based on initial body weight. All replicates of the NEG treatment (six pigs per replicate, four replicates total) were housed in a single room to account for biosecurity while replicates for POS and ISF treatments (six pigs per replicate, six replicates total) were equally represented in the remaining three rooms at ISS-VRF. Two levels of supplemental ISF (none vs. ≥1,600 mg/kg Novasoy 400; ADM, Decatur, IL) constituted the total of two dietary treatments (Table 1). The ISF concentrations used in the ISF treatment were reflective of those typical for a commercially relevant corn–soybean meal diet fed to pigs with approximately 20% soybean meal inclusion (USDA-ARS, Nutrient Data Laboratory, 2016). Both diets contained soy protein concentrate as a protein source, as this ingredient is practically devoid of soy ISF. Experimental diets were formulated to be isocaloric and, with the exceptions of corn and the isoflavone-enriched product, were identical in ingredient and nutrient composition. Experimental diets were fed over seven feeding phases, including nursery phases 1 and 2 (N1-N2) and finisher phases 1 to 5 (F1-F5). For additional information regarding experimental diet formulations and nutrient composition, please refer to the paper by Smith et al. (2020).

Table 1.

Experimental treatments

Treatment	Dietary treatment	Infection status
NEG	Control diet	Uninfected
POS	Control diet	PRRSV-Infected
ISF	Control diet + ISF1	PRRSV-Infected

¹ISF provided by Novasoy 400 (ADM, Decatur, IL) at >1,600 mg/kg complete diet across all feeding phases.

Following a 1-wk adaptation period to experimental diets (days post-inoculation [DPI] −7 to 0), blood was collected from two pigs per pen to establish 0 DPI baseline measurements and to ensure that a representative number of pigs were confirmed PRRSV-negative at experiment initiation. Immediately following blood collection, pigs were administered via intranasal inoculation either 2 mL of a 2% fetal bovine serum + phosphate-buffered saline solution (sham-control) or 1.0 × 10⁵ 50% tissue culture infective dose of PRRS virus (strain NADC20, courtesy of Dr. Federico Zuckermann, University of Illinois, Urbana, IL).

Fecal sample collection

Individual pigs served as the experimental unit for all fecal microbiota outcomes described herein. Fecal samples were collected manually by trained personnel with a clean fecal loop from up to three pigs per pen (12 to 18 pigs/treatment, same pigs collected at each timepoint when possible) on −2, 10, and 144 DPI to represent baseline, peak infection, and post-PRRSV clearance timepoints, respectively. Samples were stored at −80 °C pending DNA extraction.

Fecal bacterial DNA extraction and analyses

For DNA extraction, 180 to 220 mg of feces from individual samples was homogenized using MP Biomedicals Lysing Matrix E 2 mL tubes (MP Biomedicals, Santa Ana, CA). DNA was extracted from homogenized samples using the QIAamp FastDNA Stool Mini Kit (Qiagen, Valencia, CA) as per the manufacturer’s instructions and concentrations of isolated DNA were quantified using a Nanodrop 1000 spectrophotometer (NanoDrop Technologies, Wilmington, DE). Quality of the DNA was checked on a 1% agarose gel following staining with ethidium bromide. When at sufficiently high quality, individual samples were then diluted to 20 ng/µL using RNAse/DNAse-free water and 15 µL of diluted DNA was transferred to a 96-well plate for submission to the University of Illinois High-Throughput Sequencing and Genotyping Unit at the Roy J. Carver Biotechnology Center for polymerase chain reaction amplification and sequencing of 16S bacterial rRNA genes for microbial profiling. The V3-V4 regions of bacterial 16S rRNA genes were amplified using primers V3f (5′-CCTACGGGAGGCAGCAG-3′) and V4r (5′-GGACTACHVGGGTWTCTAAT-3′) as previously described (Monaco et al., 2018). Sequencing was performed on an Illumina MiSeq flow cell for 251 cycles from each end of the fragments with a MiSeq Reagent Nano Kit version 2 (Illumina, San Diego, CA) (2 × 250 nt paired end reads).

Sequence processing

Sequences were processed and visualized using the QIIME 2 pipeline (Bolyen et al., 2018). Paired-end reads obtained from the sequencing facility were imported into QIIME 2 and demultiplexed using the plugin demux with emp-paired method. Demultiplexed reads were primer removed, chimera checked, and dereplicated using command DADA2 denoise-paired (Callahan et al., 2016). The exact sequence variant (ESV) table was generated and representative sequences were picked from each ESV. The representative sequences were aligned with MAFFT (Katoh and Standley, 2013) and the alignment was filtered to remove unconserved and gapped columns with the use of mask (Lane, 1991). The phylogenetic tree was constructed from filtered alignment via fasttree (Price et al., 2010). Alpha diversity and UniFrac distance metrics were computed through the plugin diversity with the core metrics-phylogenetic method on the ESV table and phylogenetic tree. Samples were rarefied to an equal number of reads when alpha-diversity and UniFrac distance metrics were calculated. Taxonomic assignments were performed using plugin feature-classifier with classify sklearn method (Pedregosa et al., 2011) in which a pre-trained Naïve-Bayes classifier with SILVA 132 99% operational taxonomic units (OTUs) from V3-V4 regions of sequences was used (https://www.arb-silva.de/;Quast et al., 2013).

Statistical analyses

Differences in bacterial communities among diet groups were evaluated with principal co-ordinate analysis (PCoA) and permutational multivariate analysis of variance (PERMANOVA) using UniFrac distance matrices based on both unweighted and weighted UniFrac distances using QIIME 2 (2019.4) (Caporaso et al., 2010). All other data were analyzed by one-way ANOVA performed on square root transformations using R (version 3.5.4, Vienna, Austria) and sliced by DPI. Please note that due to greater loss in mortality than expected (Smith et al., 2020), several animals were not able to be sampled at each timepoint. Thus, we chose to analyze the microbial data individually at each timepoint, rather than across timepoints. Post hoc mean separation with a Tukey adjustment was performed if the ANOVA resulted in a P-value < 0.05. Mean separation was performed by main effect of treatment and significance was accepted at P ≤ 0.05. Only bacterial phyla and genera with at least one mean with relative abundance >0.3% and >1.0%, respectively, were interpreted and summarized in data tables. Data are presented as least-squares means with pooled standard errors of the mean (SEM). For relative abundance microbiota outcomes, data are presented as raw least-square means plus SEM, along with P-values reflective of the square root transformed values.

Results

For the sake of clarity, the measures of alpha diversity (i.e., species richness or number of distinguishable taxa within an individual sample) and beta diversity (i.e., difference between global microbial communities or differences in taxonomic abundance profiles between individual samples) are discussed separately, with beta diversity being discussed first. Please note that some of the bacterial genera detected through our analyses are newly identified and have not yet been fully classified, so that they may be identified by higher taxonomic level names, particularly at the family level. In those cases, the bacterial genera names are preceded by the letters “F” or “O” in parentheses, indicating the related taxonomic order or family.

Beta and alpha diversity measures

Beta diversity measures are represented by experimental timepoint in Figure 1. During the pre-infection period (−2 DPI), PERMANOVA analysis revealed that the overall bacterial composition in fecal samples did not differ between treatments (P > 0.05). However, at peak infection (10 DPI), PERMANOVA analysis revealed that the overall bacterial composition in fecal samples differed (unweighted, P = 0.012; weighted, P = 0.02) between noninfected (NEG) and infected treatments (POS and ISF); the two infected treatments did not differ from each other. By the post-infection period, PERMANOVA analysis revealed that the overall bacterial composition in fecal samples differed but only for unweighted UniFrac measures (unweighted, P = 0.003; weighted, P = 0.172).

Figure 1.

PCoA of unweighted and weighted UniFrac distances generated from fecal samples collected from individual pigs (n = 12 to 18 per treatment) at −2, 10, and 144 DPI.

Alpha diversity was evaluated using (OTUs at the phyla and genera level. The differences in relative abundance of bacteria at these taxonomic levels are discussed in the following section. Visualization of variation in phyla relative abundance among individual samples across sampling timepoints can be found in Figure 2. Additionally, heat maps depicting the differences in relative abundance of both common (>1.0%) and rare (<1.0%) genera among treatments across sampling timepoints can be found in Figures 3 and 4, respectively.

Figure 2.

Percent relative abundance of bacterial phyla detected in fecal samples collected from individual pigs (n = 12 to 18 per treatment) at −2, 10, and 144 DPI.

Figure 3.

Heat map depicting the differences in relative abundance of common genera (>1.0% relative abundance) between experimental treatments at −2, 10, and 144 DPI. .

Figure 4.

Heat map depicting the differences in relative abundance of rare genera (<1.0% relative abundance) between experimental treatments at −2, 10, and 144 days post-inoculation.

Relative abundance of bacteria at the phyla and genera level

At the pre-infection period (−2 DPI), 8 phyla and 30 genera were detected at or above our interpretation thresholds (0.3% and 1.0%, respectively). There were no differences in relative abundance at the phyla level between treatments at the pre-infection timepoint (Supplementary Table S1). At the genera level, the NEG treatment had a greater (P < 0.05) relative abundance of Roseburia bacteria than POS and ISF treatments (1.35 ± 0.11% vs. 0.59 ± 0.11% and 0.58 ± 0.15%, respectively; P = 0.039) (Supplementary Table S2), but no other differences were present.

At peak infection (10 DPI), 6 phyla and 26 genera were detected at or above our interpretation thresholds. At the phyla level, the ISF treatment had greater (P < 0.05) relative abundance of Proteobacteria bacteria compared with the NEG treatment, but did not differ from the POS treatment (Table 2). At the genera level, there were five differences detected among experimental treatments (Table 3). The NEG treatment had greater (P < 0.05) relative abundance of Prevotella 9 compared with POS and ISF treatments and greater (P < 0.05) relative abundance of (F) Prevotellaceae Unassigned and Anaerovibrio compared with the ISF treatment only, with the POS treatment exhibiting intermediated relative abundances for those genera. The ISF treatment had greater (P < 0.05) relative abundance of Prevotellaceae NK3B31 group compared with the NEG treatment, with the POS treatment exhibiting an intermediate relative abundance. Alternatively, the ISF treatment had lower (P < 0.05) relative abundance of Dialister compared with the NEG treatment, again with the POS treatment exhibiting an intermediate relative.

Table 2.

Percent relative abundance of bacterial phyla detected in fecal samples at 10 DPI¹

	Treatment
Bacterial phyla2	NEG	POS	ISF	P-value
Bacteroidetes	50.1 ± 2.32	46.4 ± 2.7	42.4 ± 2.64	0.123
Elusimicrobia	0.41 ± 0.40	0.16 ± 0.10	0.09 ± 0.05	0.934
Epsilonbacteraeota	1.65 ± 0.42	3.62 ± 1.38	2.48 ± 0.72	0.536
Firmicutes	41.1 ± 2.76	36.8 ± 2.25	38.7 ± 3.00	0.533
Proteobacteria	5.15 ± 1.23b	10.4 ± 2.22ab	10.7 ± 1.71a	0.045
Spirochaetes	1.07 ± 0.48	1.72 ± 0.47	4.66 ± 1.57	0.070

¹Data are expressed as least-square mean ± SEM for each treatment group.

²Only phyla with >0.30 relative abundance were considered for interpretation.

^a,bMeans without a common superscript letter differ (P < 0.05).

Table 3.

Relative abundance of bacterial genera detected in fecal samples at 10 DPI¹

	Treatment
Bacterial genus by phyla2	NEG	POS	ISF	P-value
Bacteroidetes
Alloprevotella	4.46 ± 0.77	4.18 ± 0.81	3.32 ± 0.87	0.399
dgA-11 gut group	0.58 ± 0.30	0.79 ± 0.22	1.07 ± 0.46	0.563
(F) Muribaculaceae Porphyromonadaceae Unassigned	2.46 ± 0.39	5.98 ± 3.15	3.31 ± 0.73	0.593
(F) Muribaculaceae Unassigned	0.53 ± 0.16	1.08 ± 0.35	0.65 ± 0.15	0.243
Prevotella 1	2.52 ± 1.20	2.26 ± 0.57	2.52 ± 0.43	0.914
Prevotella 2	2.80 ± 0.32	2.30 ± 0.39	1.78 ± 0.21	0.100
Prevotella 9	18.4 ± 3.19a	9.56 ± 1.75b	7.63 ± 1.15b	0.002
Prevotellaceae NK3B31 group	1.71 ± 0.35b	3.64 ± 0.63ab	3.97 ± 0.61a	0.040
Prevotellaceae UCG-003	3.31 ± 0.91	4.28 ± 0.72	4.97 ± 0.86	0.408
(F) Prevotellaceae Unassigned	2.68 ± 0.37a	1.71 ± 0.28a	0.79 ± 0.15b	<0.001
Rikenellaceae RC9 gut group	7.25 ± 1.10	7.09 ± 2.72	7.79 ± 1.86	0.804
Firmicutes
Acidaminococcus	1.14 ± 0.39	0.69 ± 0.27	1.49 ± 0.83	0.723
Agathobacter	1.48 ± 0.32	3.55 ± 1.75	1.78 ± 0.51	0.574
Anaerovibrio	1.25 ± 0.27a	0.65 ± 0.21ab	0.42 ± 0.14b	0.011
Dialister	2.83 ± 0.90a	1.40 ± 0.40ab	0.52 ± 0.13b	0.011
Faecalibacterium	1.53 ± 0.50	0.76 ± 0.18	0.62 ± 0.10	0.099
Lachnospiraceae NK4A136 group	1.23 ± 0.22	3.15 ± 0.88	2.82 ± 0.58	0.111
(F) Lachnospiraceae Unassigned	1.91 ± 0.48	2.37 ± 0.69	2.67 ± 1.07	0.869
Lactobacillus	5.72 ± 1.48	3.42 ± 0.65	3.78 ± 0.74	0.604
Megasphaera	4.14 ± 1.19	5.85 ± 1.85	5.21 ± 1.94	0.865
Roseburia	0.41 ± 0.16	0.78 ± 0.27	1.38 ± 0.59	0.246
Ruminococcaceae UCG-002	2.71 ± 0.37	1.51 ± 0.21	2.31 ± 0.51	0.173
(F) Veillonellaceae Unassigned	2.05 ± 0.45	1.43 ± 0.47	1.94 ± 0.44	0.319
Proteobacteria
Campylobacter	1.58 ± 0.39	3.52 ± 1.37	2.31 ± 0.72	0.546
Succinivibrio	3.28 ± 1.27	7.81 ± 2.25	7.82 ± 1.88	0.155
Spirochaetes
Treponema 2	1.04 ± 0.48	1.67 ± 0.46	4.64 ± 1.57	0.060

¹Data are expressed as least-square mean ± SEM for each treatment group.

²Only genera with at least one mean with >1.0% relative abundance were considered for interpretation. Bacterial genera names preceded by (F) indicates that genera are associated with a specific taxonomic family and has not yet been further classified.

^a,bMeans without a common superscript letter differ (P < 0.05).

At the post-infection timepoint (144 DPI), 7 phyla and 27 genera were detected at or above our interpretation thresholds. At the phyla level, there was a difference observed for Actinobacteria, with the NEG treatment expressing greater (P < 0.05) relative abundance compared with POS and ISF treatments, though the total relative abundance of this phylum across treatments was low (Table 4). At the genera level, there were seven differences detected among experimental treatments (Table 5). The NEG treatment had greater (P < 0.05) relative abundance of CAG-873 and Prevotellaceae UCG-004 compared with both the POS and ISF treatments. Both the NEG and POS treatments had greater (P < 0.05) relative abundance of Clostridium sensu stricto 1 compared with the ISF treatment. The NEG treatment had greater (P < 0.05) relative abundance of Ruminococcaceae UCG-002 and Lachnospiraceae XPB1014 group compared with the POS treatment, with the ISF treatment exhibiting an intermediate relative. The ISF treatment had greater (P < 0.05) relative abundance of (F) Lachnospiraceae Unassigned and Ruminococcaceae UCG-014 compared with the NEG treatment, with the POS treatment exhibiting an intermediate relative abundances.

Table 4.

Percent relative abundance of bacterial phyla detected in fecal samples at 144 DPI¹

	Treatment
Bacterial phyla2	NEG	POS	ISF	P-value
Actinobacteria	0.43 ± 0.11a	0.20 ± 0.03b	0.24 ± 0.05ab	0.049
Bacteroidetes	30.6 ± 1.88	29.2 ± 1.09	29.3 ± 1.30	0.787
Elusimicrobia	0.05 ± 0.02	0.20 ± 0.05	0.49 ± 0.23	0.109
Firmicutes	30.1 ± 2.37	27.9 ± 1.27	30.1 ± 1.67	0.593
Kiritimatiellaeota	0.86 ± 0.18	1.33 ± 0.19	1.26 ± 0.18	0.129
Proteobacteria	25.3 ± 2.73	23.7 ± 2.44	18.7 ± 1.96	0.185
Spirochaetes	12.2 ± 2.69	16.9 ± 2.79	19.0 ± 3.14	0.346

¹Data are expressed as least-square mean ± SEM for each treatment group.

²Only phyla with at least one mean with >0.30% relative abundance were considered for interpretation.

^a,bMeans without a common superscript letter differ (P < 0.05).

Table 5.

Relative abundance of bacterial genera detected in fecal samples at 144 DPI¹

	Treatment
Bacterial genus by phyla2	NEG	POS	ISF	P-value
Bacteroidetes
Bacteroides	1.09 ± 0.21	0.79 ± 0.17	0.99 ± 0.21	0.445
CAG-873	2.98 ± 1.49a	0.33 ± 0.12b	0.16 ± 0.05b	<0.001
dgA-11 gut group	0.88 ± 0.26	1.29 ± 0.45	1.53 ± 0.44	0.576
(F) Muribaculaceae Porphyromonadaceae Unassigned	1.30 ± 0.40	1.38 ± 0.38	1.12 ± 0.30	0.871
(F) Muribaculaceae Unassigned	3.20 ± 0.40	2.93 ± 0.66	2.41 ± 0.37	0.372
Parabacteroides	1.10 ± 0.15	0.75 ± 0.09	1.16 ± 0.16	0.149
Prevotella 1	0.68 ± 0.15	1.09 ± 0.19	1.59 ± 0.53	0.141
Prevotellaceae NK3B31 group	5.63 ± 1.19	5.66 ± 0.54	6.66 ± 0.82	0.599
Prevotellaceae UCG-001	3.35 ± 0.63	5.44 ± 0.94	4.24 ± 0.64	0.306
Prevotellaceae UCG-004	1.13 ± 0.22a	0.59 ± 0.05b	0.55 ± 0.06b	0.002
Rikenellaceae RC9 gut group	4.70 ± 0.47	4.33 ± 0.73	3.57 ± 0.42	0.337
Firmicutes
Christensenellaceae R-7 group	3.43 ± 0.83	2.62 ± 0.50	2.10 ± 0.27	0.363
(F) Christensenellaceae Unassigned	0.79 ± 0.27	1.87 ± 0.31	2.15 ± 0.61	0.085
Clostridium sensu stricto 1	1.26 ± 0.24a	1.15 ± 0.21a	0.64 ± 0.09b	0.038
Family XIII AD3011 group	1.03 ± 0.18	1.63 ± 0.44	1.55 ± 0.28	0.471
Lachnospiraceae NK4A136 group	2.59 ± 0.64	1.75 ± 0.46	1.35 ± 0.30	0.155
(F) Lachnospiraceae Unassigned	1.60 ± 0.32b	2.87 ± 0.51ab	3.17 ± 0.50a	0.044
Lachnospiraceae XPB1014 group	2.07 ± 0.42a	1.05 ± 0.23b	1.12 ± 0.22ab	0.039
Ruminococcaceae NK4A214 group	2.19 ± 0.23	2.08 ± 0.31	1.78 ± 0.15	0.470
Ruminococcaceae UCG-002	1.54 ± 0.31a	0.83 ± 0.15b	0.92 ± 0.12ab	0.039
Ruminococcaceae UCG-005	2.09 ± 0.40	1.49 ± 0.20	1.58 ± 0.17	0.266
Ruminococcaceae UCG-014	0.45 ± 0.10b	0.79 ± 0.11ab	1.12 ± 0.25a	0.043
Ruminococcus 1	1.21 ± 0.36	1.01 ± 0.12	1.21 ± 0.17	0.782
(F) Veillonellaceae Unassigned	1.93 ± 0.64	1.07 ± 0.28	1.93 ± 0.45	0.170
Proteobacteria
Succinivibrio	21.8 ± 3.06	18.3 ± 2.26	12.7 ± 1.99	0.060
Succinivibrionaceae UCG-001	2.77 ± 1.68	4.40 ± 1.35	4.20 ± 1.05	0.481
Spirochaetes
Treponema 2	12.2 ± 2.68	16.9 ± 2.78	19.0 ± 3.14	0.348

¹Data are expressed as least-square mean ± SEM for each treatment group.

²Only genera with at least one mean with >1.0% relative abundance were considered for interpretation. Bacterial genera names preceded by (F) indicates that genera are associated with a specific taxonomic family and has not yet been further classified.

^a,bMeans without a common superscript letter differ (P < 0.05.

Discussion

In the live-animal phase of this study (Smith et al., 2020), ISF consumption over the entire growth period resulted in earlier complete viral clearance and an approximately 50% reduction in pathogen-related mortalities in pigs infected with PRRSV, though there were no growth performance benefits observed. Those results suggest that when included at levels that are typical for standard commercial swine diets, soy ISF may support the immune response to PRRSV, particularly in the presence of severe secondary bacterial infections. While it is understood that ISF can interact directly with cell signaling cascades utilized by the immune system (Smith and Dilger, 2018), within the context of this live-animal model it was unclear whether such beneficial health effects were due to direct action of ISF in circulation or if changes to the microbiome of the gastrointestinal tract (GIT) may be contributing to these systemic effects.

Regarding the role of the gastrointestinal microbiome in immune function, the “gut-lung axis” is a well-documented phenomenon that represents the bidirectional communication between microbial populations of the gastrointestinal and respiratory tracts (Collado et al., 2018). In general, the gastrointestinal microbiome has been shown to play a role in the development of innate immunity (e.g., maintenance of homeostasis, epithelial function at mucosal surfaces), T-cell differentiation, modulation of local and systemic inflammation, and regulate homeostasis of the adaptive immune response (Kau et al., 2011; Hooper et al., 2012; Hauptmann and Schaible, 2016; Niederwerder, 2017). These effects are not only driven by the composition of microbial populations in the GIT but also through the production of different metabolites through fermentation dependent on predominant species present. These metabolites, most notably short-chain fatty acids, have been shown to not only have local effects on mucosal immune and barrier function in the intestinal tract but also influence the development of peripheral immune cells such as dendritic cells and B-cells, thus impacting their overall function (Collado et al., 2018). When considering immune function at the level of the respiratory tract, it has been suggested that the GIT and respiratory tract share a common mucosal response, which involves the effects of the gastrointestinal microbiome on mucosal immunity driving immune responses at distal mucosal sites (Mcghee and Fujihashi, 2012; Hauptmann and Schaible, 2016; Date et al., 2017). In practice, several dysbioses (i.e., microbial imbalances or maladaptation on or inside the body) in the gastrointestinal microbiota have been associated with lung disorders and respiratory infections in humans, including viral pathogens such as influenza virus (Collado et al., 2018). Within respiratory viral pathogen disease models, microbiome composition and diversity have been shown to be directly associated with clinical outcomes, including disease progression, airway inflammation, immune response, and morbidity (Niederwerder, 2017; Ober et al., 2017).

This relationship between respiratory infections and gastrointestinal dysbioses is also observed in swine respiratory disease models (Niederwerder et al., 2016, 2018; Niederwerder, 2017; Ober et al., 2017). Considering disease models utilizing PRRSV, microbial composition has been linked to differences in pig performance following pathogenic challenge. In a PRRSV and porcine circovirus type 2 (PCV2) coinfection model in weanling pigs, researchers found that pigs with the best clinical outcomes had greater fecal microbial diversity than pigs with the worst clinical outcomes. In the worst clinical outcomes group, pigs experienced delayed peak PRRSV-infection with prolonged and increased viremia, higher peak PCV2 viremia levels (nonsignificant), and increased serum concentrations of Bacillus cereus, suggesting that these animals experienced a more severe systemic infection and compromised mucosal integrity. Regarding microbiome differences observed, in addition to differences in the overall measures of microbiome diversity, pigs in the worst clinical outcomes group notably had no Escherichia coli detected in their fecal samples, which has been found to influence feed efficiency and growth performance in growing pigs (Niederwerder et al., 2016). Continuation of the same research found that even early microbiome properties (i.e., composition and diversity prior to disease challenge) influenced the performance outcomes of PRRSV/PCV2 co-infected pigs, suggesting that the effects of the gastrointestinal microbiome on immune function extend beyond current GIT conditions. Following subclinical infection, there was a clear divergence between high and low growth rate groups during the mid-to-late infection period, with pigs in the high growth rate group exhibiting reduced viral replication rates and decreased lung lesion severity. Like previous findings from this group, pigs in the high growth rate group had increased fecal microbial diversity (Ober et al., 2017). In general, a common finding across these studies and other respiratory disease models is increased global microbial diversity at the level of the GIT being associated with improved clinical outcomes in infected individuals.

Within this experiment, we observed differences in beta diversity or global microbial diversity at peak-infection and post-infection timepoints (10 and 144 DPI) between experimental treatments. These differences were most notable at the peak-infection timepoint, with significance achieved for both weighted and unweighted UniFrac distance matrices, and appear to be mainly due to PRRSV-infection as infected treatments did not differ from one another. At the post-infection timepoint, there was an effect of treatment on beta diversity with the NEG treatment differing from our infected treatments, but only for unweighted UniFrac distance matrices. These differences are in line with the PRRSV coinfection models described above but suggest that at the inclusion level utilized for this experiment ISF do not result in dramatic shifts of global microbial diversity in the GIT.

While there were differences that appeared to be due to PRRSV alone, alpha diversity or relative abundance at both the phyla and genera levels were not strongly influenced by ISF supplementation. Of the differences observed due to ISF supplementation, a vast majority of differences in the relative abundance occurred at post-inoculation timepoints. At the peak-infection timepoint, we detected a single difference at the phyla level between the NEG and ISF treatment for Proteobacteria with the ISF treatment showing greater relative abundance. Interestingly, the increased abundance of Proteobacteria has been associated with improved clinical outcomes under swine pathogen challenge (Niederwerder et al., 2016; Smith et al., 2020). However, because the ISF treatment did not differ from the POS treatment, it is unclear whether this increased relative abundance contributed to any of the performance differences observed in the live animal phase of this experiment. At the genera level, reduced relative abundance of Prevotella 9, (F) Prevotellaceae Unassigned, and Anaerovibrio genera appear to be mainly driven by PRRSV-infection. Among these findings, the difference in Prevotella 9 abundance is most notable due to its reduction from nearly 20% in noninfected control pigs to less than 10% in infected pigs. However, knowing the specific function of this Prevotella genera is difficult due to genetic and overall microbiota differences in the host. Prevotella, while having been identified as potentially beneficial microbes, are also linked to inflammatory conditions such as colitis and arthritis under immune suppressive states (Dillon et al., 2016). Based on those findings, the expected relationship for Prevotella abundance with a PRRSV-infection model would be opposite of what was observed. Additionally, previous nutrition studies in humans suggest that the effects of single diet ingredients on Prevotella are not easily predicted (Kovatcheva-Datchary et al., 2015). Regarding dietary supplementation, ISF increased the relative abundance of the Prevotellaceae NK3B31 group and decreased the relative abundance of Dialister compared with the NEG treatment. Of these two genera quantified in intestinal dysbioses models, Prevotellaceae species has been found to be a driver of chronic inflammation at the level of the GIT. In the context of this experimental model, one could argue that ISF in the presence of a PRRSV-infection support the growth of this particular bacteria species (Palm et al., 2015).

Moving to the post-infection timepoint, there was a single difference at the phyla level for the relative abundance of Actinobacteria between the NEG treatment and our infected treatments, with PRRSV-infection appearing to decrease the relative abundance of this bacteria. Reduction in the relative abundance of Actinobacteria has been linked to various dysbioses for which changes typically represent a loss of microbial richness that may directly impact amino acid synthesis, cell junction integrity, and inflammatory responses (Rinninella et al., 2019). At the genera level, reduced relative abundance of the CAG-873 and Prevotellaceae UCG-004 genera appeared to be driven by PRRSV-infection alone. The reduction in Prevotellaceae UCG-004 specifically here is unexpected as other Prevotellaceae genera measured in this experiment demonstrated the opposite relationship. Regarding ISF supplementation, ISF appeared to increase the relative abundance of (F) Lachnospiraceae Unassigned and Ruminococcaceae UCG-014 genera relative to the NEG treatment and decreased the relative abundance of Clostridium sensu stricto 1 genera relative to both the NEG and POS treatments. Alternatively, the ISF treatment exhibited intermediate relative abundance of Ruminococcaceae UCG-002 and Lachnospiraceae XPB1014 group genera compared with the NEG and POS treatments. In previous PRRSV-challenge studies, pigs demonstrating improved performance exhibited increased species diversity in the Ruminococcaceae family, which was previously associated with improved growth performance (Ober et al., 2017). In the context of this experimental model, it may be that ISF in the presence of a PRRSV-infection supports the growth of certain Ruminococcaceae bacteria within the GIT, though it is difficult to speculate what the specific role these individual genera play in the global response to PRRSV-infection. The increased relative abundance of Lachnospiraceae bacteria have been associated with gastrointestinal dysbioses, specifically irritable bowel syndrome in humans (Rinninella et al., 2019), though within this model there appear to be inconsistent effects of ISF supplementation on its presence under PRRSV-challenge.

Though a number of differences in relative abundance were detected within this model based on our interpretation criteria, the interpretation of these findings must be done conservatively. It is important to note that the majority of differences detected at the genera level and involved bacterial genera that on average had a relative abundance of less than 3% (range: 0.16% to 18.5% across all treatments). Additionally, many of these genera are either newly identified or not well described regarding their role in swine specifically. Due to that limitation and the relatively low abundance of these bacteria, it is difficult to make conclusions on their specific role in the clinical response to PRRSV-infection within our model.

While this experiment did result in changes to global microbial diversity, the magnitude of those differences appear to be less than previously observed in PRRSV-challenge models (Niederwerder et al., 2016; Ober et al., 2017). This is particularly noticeable when referencing visualizations of phyla and genera relative abundance across our experimental timepoints (Figures 2–4); differences between experimental treatments are not particularly robust, especially when considering pre- and post-infection timepoints. This lack of detectable differences compared with previous research could be due to the differences in how global diversity was measured between our experiment and previous models or simply unaccounted for the differences in viral strain properties and influence on the GIT. Additionally, while we can appreciate that there are noticeable visual differences in relative abundance distributions across experimental timepoints (e.g., phyla relative abundance at −2 DPI vs. 144 DPI), we analyzed each experimental timepoint individually, which prevented us from taking the effect of time into account on individual microbiome composition.

Implications

In conclusion, there were few differences detected at either the phyla or genera level between treatments regardless of time point. Of the differences observed, some have been previously associated with gastrointestinal dysbioses and differences in pig performance under respiratory disease challenge. Our observations suggest that most of the differences present were likely driven by PRRSV-infection and not strongly influenced or inconsistently influenced by diet, which implies that the performance benefits observed within infected treatments conferred by dietary isoflavones in the companion study, such as reduced mortality, were not likely due to the changes in microbiome composition or function.

Glossary

Abbreviations

DPI

days post-inoculation

ESV

exact sequence variant

GIT

gastrointestinal tract

ISF

isoflavones/PRRSV-positive experimental treatment

ISS-VRF

Innovative Swine Solutions Veterinary Research Facility

NEG

PRRSV-negative control treatment

OTU

operational taxonomic unit

PCoA

principal coordinate analysis

PCV2

porcine circovirus type 2

PERMANOVA

permutational multivariate analysis of variance

POS

PRRSV-positive control treatment

PRRSV

porcine reproductive and respiratory syndrome virus

SID

standardized ileal digestible

TCID₅₀

median tissue culture infectious dose

Conflict of interest statement

The authors declare that there is no conflict of interest.