This study was carried out in accordance with the recommendations of the experimental animal administration and ethics committee of South China University of Technology of guidelines. The protocol was approved by the Institutional Animal Care and Use Committee of South China University of Technology.

This study was conducted at two commercial swine farms, each with more than 10,000 pigs. The two farms located at least 100 km apart in Guangdong province, Southern China. Both farms practiced intensive, farrow-to-finish, open-house systems. The piglets from the same farm were raised under the same standard indoor conditions according to their growth stages. In detail, the piglets were raised in farrowing house from birth to 28 days old. Then, the piglets were moved to nursery pens until their weight reached approximately 30 kg. Subsequently, the animals were moved to growing–finishing pens to be sold out as breeders. Dominant breeds in the two farms were crosses between Landrace and Yorkshire (sows) with Duroc or Hampshire (boars). Sample collection was carried out in March 2017. Oropharyngeal swabs were taken from 33 piglets at 2–3 weeks of age before weaning from the two swine farms. Twenty four oropharyngeal samples (18 PRD and 6 healthy) were collected from farm A and 9 oropharyngeal samples (5 PRD and 4 healthy) were collected from farm B. Veterinarians with over 5 years of experience diagnosed PRD by physical examination. The piglets with PRD presented with clinical symptoms, including cough, fever, joint swelling, and wheezes. Only those piglets that had never experienced PRD and other diseases were included in the healthy groups. Deep oropharyngeal swabs were collected from the piglets with the help of the veterinarians using 20 cm DNA-free sterile swabs covered with a sterile plastic sheath. All samples were kept on ice until they were transferred to the laboratory where they were resuspended in 500 μL PBS and stored at -80°C until processing.

DNA was extracted from the oropharyngeal swabs from the piglets using the QIAamp Mini Kit (Qiagen, Hilden, Germany) following the manufacturer’s instructions. The final DNA concentration was determined on the basis of the absorbance at 260 and 280 nm using a Nanodrop 2000 UV-vis spectrophotometer (Thermo Fisher Scientific, Wilmington, United States), and the DNA quality was evaluated by 1% agarose gel electrophoresis. Negative controls were performed for the extraction process and evaluated by gel analysis after PCR amplification. For metagenomic analysis, the V3-V4 region of the 16S rRNA gene was amplified using forward (338F, 5′ACTCCTACGGGAGGCAGCAG-3′) and reverse (806R, 5′-GGACTACHVGGGTWTCTAAT-3′) primers on a GeneAmp^® 9700 (ABI, United States) PCR system. PCR reactions were performed in triplicate 20 μL mixture containing 10 ng template DNA, 4 μL 5 × FastPfu buffer, 2 μL dNTPs (2.5 mmol/L), 0.8 μL each primer (5.0 μmol/L), 0.4 μL FastPfu polymerase and ddH₂O. The reaction conditions for PCR were 95°C for 3 min for an initial denaturation, followed by 27 cycles of denaturation at 95°C for 30 s, primer annealing at 55°C for 30 s, extension at 72°C for 45 s, and a final elongation for 10 min at 72°C. The PCR products were extracted from 2% agarose gels and further purified using the AxyPrep DNA Gel Extraction Kit (Axygen Biosciences, Union City, CA, United States) and quantified using QuantiFluor^TM -ST (Promega BioSciences LLC, Sunnyvale, CA, United States) according to the manufacturer’s protocol. Samples that contained no template and those that contained known 16S rRNA gene sequences were used as negative and positive controls in the PCR reactions. DNA was stored at -80°C prior to downstream processing.

The DNA samples were quantified by spectrophotometry (Thermo Fisher Scientific, Wilmington, United States) and diluted to a final concentration of 2 nM. Purified amplicons were sequenced by paired-end sequencing (2 × 300) on an Illumina MiSeq platform (San Diego, CA, United States) for metagenomic analysis.

All raw reads were demultiplexed and quality-filtered by Trimmomatic and merged by FLASH with the following criteria: (1) The reads were truncated at any site receiving an average quality score less than 20 over a 50 bp sliding window. (2) Primers were exactly matched allowing two nucleotides mismatching and reads containing ambiguous bases were discarded from the analysis. (3) Sequences whose overlap was longer than 10 bp were merged according to their overlap sequence.

Operational taxonomic units (OTUs) were clustered with 97% similarity cutoff using UPARSE (version 7.1¹) and chimeric sequences were identified and removed using the UCHIME tool. RDP Classifier² was used to analyze the phylogenetic affiliation of each 16S rRNA gene sequence, against the silva (SSU115) 16S rRNA database with a confidence threshold of 70%. Sequences belonging to non-bacterial domains, including chloroplasts, mitochondria, archaea, and eukaryotes, were removed.

The analysis was performed at different taxonomic levels: kingdom, phylum, class, order, family, genus, and species. OTUs that reached 97% similarity were used for alpha diversity estimations, which included observed OTUs (Sobs), diversity (Shannon and Simpson indices), richness (Chao I), coverage (Good’s coverage), and rarefaction curve analysis using Mothur (Version 1.30.2 ³). Alpha diversity metrics between different groups were compared using Student’s t-test and p < 0.05 was considered a statistically significant difference. The core microbiota (OTUs with >1% relative abundance and shared in all healthy piglets) was explored for the healthy piglets in the two farms. Bray–Curtis distance metrics analysis was performed using OTUs for each sample, and principal coordinates analysis (PCoA) analysis was conducted according to this distance. Venn diagram analysis was performed to show unique and shared OTUs in the microbial population structure between healthy and diseased groups. Heatmaps were generated with the R-package gplots at the phylum and genus level. Non-parametric Wilcoxon rank-sum test was used to identify OTUs that shown significant differences in abundance between groups (confidence interval method).

The 16S rRNA gene sequence information in this paper was deposited in the GenBank Sequence Read Archive database under accession number SRP169602.

All oropharyngeal samples were analyzed by 16S rRNA gene sequencing, and comparisons were made between healthy piglets and piglets with PRD in the two farms. In this study, 1,491,208 quality-filtered and chimera-checked sequences were obtained with an average length of 446.78 bp across all samples. The mean number of reads per sample was 45,188, ranging from 30,717 to 63,827 reads. A total of 1658 unique OTUs (97% sequence similarity) were detected among all samples. Based on the relative abundance, the taxonomic analysis revealed 22 bacteria phyla, 48 classes, 111 orders, 202 families, and 444 genera across all samples.

All samples’ coverage (Good’s coverage) was more than 99.50%, which indicated that the accuracy of sequencing was reliable. Multiple rarefaction curves were measured using several metrics namely Sobs, Chao1, Simpson, and Shannon, which confirmed adequate sequence coverage for all samples (Supplementary Figure S1). Table 1 indicates the specific bacterial indices (Shannon, Simpson, OTUs, and Chao 1) made from each group as well as the p-value for PRD and healthy groups. At the OTU level, there were no statistically significant differences in bacterial diversity and richness between the healthy and PRD groups in the two farms (p > 0.05), except for bacterial diversity measured by Shannon index in farm A (p < 0.05). The comparison of bacterial indices between all groups is shown in Supplementary Figure S2.

Beta diversity of each group was calculated through cluster tree analysis and PCoA. The microbial community structure of all samples was analyzed using the phylogeny-based Bray–Curtis method and visualized using PCoA (Figure 1). The first two factors (PC1 and PC2) accounted for 41.28% of the sample variation, which demonstrated that the microbial communities were distinguishable from one another, especially the samples from different farms (Figure 1).

The Venn diagrams displayed a high number of unique and shared OTUs in suckling piglets. In addition, a clear overlap pattern for oropharyngeal samples from the two farms was also observed (Figure 2). In farm A, 1094 OTUs were identified, of which 456 OTUs (41.68%) were shared between the two groups. The total number of OTUs in the PRD group was much higher than that in the healthy group (1000 OTUs versus 550 OTUs in controls) (Figure 2A). In farm B, 1375 OTUs were identified totally and 567 OTUs (41.24%) were shared between the two groups (Figure 2B). There were 1658 OTUs among all groups, of which 811 OTUs (48.91%) were shared between the two farms regardless of health status (Figure 2C). This illustrated that some common OTUs may be the permanent inhabitants of the oropharynx of piglets.

There were 22 different phyla identified across all samples, of which only five phyla had more than 1% overall relative abundance- Firmicutes (53.11%), Proteobacteria (27.89%), Bacteroidetes (12.17%), Fusobacteria (3.15%), and Actinobacteria (2.29%), which comprised on average 98.61% of the total reads and formed the main microbiota of porcine oropharynx (Figure 3A). In the Figure 3A, ‘others’ represents bacterial phyla that formed less than 1% of the total abundance. The relative abundance was different at the phylum level in piglets from both farm A and farm B (Supplementary Figure S3). The heatmap profile indicated that oropharyngeal microbial community structures of the four groups were similar at the phylum level (Supplementary Figure S4).

In farm A, the most abundant bacterial genera in the PRD group were Streptococcus (26.25%), Actinobacillus (17.63%), Moraxella (12.02%), Veillonella (7.16%), Bergeyella (4.60%), and Escherichia-Shigella (4.19%), while Lactobacillus (37.28%), Streptococcus (32.26%), Actinobacillus (15.09%), Bergeyella (3.10%), and Escherichia-Shigella (2.24%) were the predominant genera in healthy group (Figure 3B and Supplementary Table S1). In farm B, the prevalent genera found in the PRD group were Streptococcus (22.22%), Moraxella (15.82%), Actinobacillus (9.66%), Bergeyella (7.46%), Fusobacterium (4.64%), Porphyromonas (4.31%), Leptotrichia (4.04%), whereas the genera Lactobacillus (30.44%), Streptococcus (23.60%), Actinobacillus (11.48%), Moraxella (4.28%), and Bacteroides (2.00%) were predominant in the healthy group (Figure 3B and Supplementary Table S1). In the Figure 3B, ‘others’ represents bacterial genera that formed less than 2% of the total abundance. Heatmap analyses of abundant genera in each group are displayed in Figure 4. There was remarkable variability in the oropharyngeal microbial composition at the genus level (Supplementary Figure S5).

There was a relative increase in Proteobacteria and Bacteroidetes in conjunction to a decrease in Firmicutes in the PRD groups from both farms. The increase in Proteobacteria corresponded to a higher abundance of Moraxella and Pasteurella as well as other bacterial components present in low abundance. Meanwhile, the increase in Bacteroidetes was mainly due to the relatively higher abundance of Bergeyella and Porphyromonas in piglets with PRD. In addition, within Firmicutes, the principal change was in Lactobacillus, which decreased from 37.28 and 30.45% in healthy groups to 1.26 and 0.60% in PRD groups from the two farms, respectively (Supplementary Table S1).

The core microbiota was defined as those OTUs with more than 1% relative abundance and present in all healthy samples. The core microbiota in the oropharynx of healthy piglets from farm A and farm B was described in Table 2. Some remarkable differences were observed in the core genera. Bergeyella and Escherichia-Shigella were relatively abundant in farm A while Bacteroides and Prevotella_1 were part of the core genera in farm B. The five core oropharyngeal genera for farm A were Streptococcus, Lactobacillus, Actinobacillus, Bergeyella, and Escherichia-Shigella. For farm B, Streptococcus, Lactobacillus, Actinobacillus, Bacteroides, and Prevotella_1 formed the core oropharyngeal microbiota.

To explore the oropharyngeal microbiota differences between the healthy and PRD piglets, non-parametric Wilcoxon rank-sum test was used to compare the mean relative abundance of the predominant bacteria at the phylum and genus levels. Some differences were marked in the microbiota communities between the two groups depending on the health status. In farm A, at the phylum level, there was a greater relative abundance of Firmicutes in healthy piglets (38.34% versus 2.45% in PRD group, p < 0.05), while in PRD group, we observed a predominance of Proteobacteria (40.80% versus 20.20% in healthy group, p < 0.05) and Actinobacteria (3.15% versus 0.84% in healthy group, p < 0.05) (Supplementary Figure S6). In farm B, only Fusobacteria was significantly different between the two groups (8.67% in PRD_B versus 1.94% in Healthy_B, p < 0.05) (Supplementary Figure S6).

In farm A, the percentages of Moraxella (12.02%) and Veillonella (7.16%) in the PRD group were higher as compared to the healthy group (p < 0.05). However, we observed a statistically significant predominance of Lactobacillus in the healthy group (37.28% versus 1.26% in PRD group, p < 0.05) (Figure 5A and Supplementary Table S1). In farm B, the genus Porphyromonas (4.31% versus 0.62% in Healthy_B, p < 0.05) was predominant in PRD group, whereas Lactobacillus (30.45% in Healthy_B versus 0.60% in PRD_B) was statistically increased in healthy group as well as Blautia and Prevotella_1 (p < 0.05) (Figure 5B and Supplementary Table S1). Therefore, the data demonstrated that the oropharyngeal microbial community composition differed significantly between the PRD and healthy piglets and between different farms.