Characterization of the pig lower respiratory tract antibiotic resistome

Antibiotic resistance genes (ARGs) pose a growing threat to public health, but most research has focused on the gut microbiota. With increased antibiotic use for respiratory diseases, it is unknown whether ARGs accumulate in the lower respiratory tract microbiome, how they relate to mobile genetic elements (MGEs), and whether they associate with disease severity. Pigs are an important agricultural species with respiratory diseases that affect production and have microbiome similarities to humans, making them a relevant model. This study aims to systematically characterize the composition, diversity, and distribution of the ARG resistome in the swine lower respiratory tract, determine relationships between ARGs and MGEs, identify bacterial hosts of ARGs, compare resistomes between bronchoalveolar lavage (BAL) and tracheal lavage samples, assess associations with lung lesions, and evaluate potential inter-site (lung–gut) and inter-host (pig–human) horizontal ARG transfer mediated by MGEs.

Prior work has shown that antibiotic exposure reduces microbial diversity and promotes multidrug-resistant bacteria. Studies have linked human gut resistome composition to disease progression, including diabetes and autism spectrum disorder. The respiratory tract has been less studied due to sampling challenges; available studies suggest airways in chronic respiratory diseases are important ARG reservoirs. In pigs, lung microbial community composition differs by lesion severity and relates to growth and physiological traits. MGEs including insertion sequences, transposons (e.g., Tn916), integrons, plasmids, and integrative conjugative elements mediate horizontal gene transfer and influence resistomes, but their diversity and role in swine lungs were unclear. Mycoplasma hyopneumoniae is a primary pathogen in swine respiratory disease; although often antibiotic-sensitive, its pathogenic mechanisms and potential to harbor or acquire ARGs had not been fully delineated.

• Samples: 745 lower respiratory tract metagenomes (670 BAL, 74 tracheal lavage, 1 esophageal lavage) from 675 pigs across five populations (F7 mosaic, Erhualian, Berkshire × Licha line, wild boars, Tibetan). Seventy-four tracheal samples included 69 pigs with paired BAL. No antibiotics for two months before sampling. Twelve control samples (six PBS blanks and six reagent/library controls) were sequenced.
• Lung lesion phenotyping: Lungs photographed; six panelists scored 13 sections with weighted contributions (apical 20%, cardiac 20%, diaphragmatic 50%, intermediate 10%), scores 0–5 by lesion extent; high-agreement panelists’ scores averaged. Pigs grouped: healthy (HL), slight (SLL), moderate (MLL), severe (SVLL).
• DNA extraction and sequencing: QIAamp Fast DNA Stool Mini Kit. DNBSEQ-T7, 150 bp PE. Average 70.3 Gb raw/sample; after host decontamination and QC, mean 1.52 Gb clean/sample due to host DNA contamination.
• Assembly and gene catalog: MEGAHIT single-sample and co-assembly; 19,685,103 contigs ≥500 bp (avg length 1136 bp, N50 1626 bp). ORFs predicted with Prodigal; clustered with CD-HIT at 90% identity; filtered to yield a non-redundant gene catalog of 10,337,194 genes.
• Abundance and taxonomy: Reads mapped with BWA; counts by featureCounts; abundance normalized considering gene length and depth. Taxonomy via DIAMOND against UniProt TrEMBL, parsed with BASTA; stringent thresholds (e-value ≤ 1e-10, identity >80%, match length >25 aa, ≥60% consensus).
• MAGs: Binning with MetaBAT2, MaxBin2, CONCOCT (single-sample) and MetaBAT2, MaxBin2, VAMB (co-binning), followed by refinement, reassembly, dereplication; 397 non-redundant MAGs (≥50% completeness, ≤10% contamination) with GTDB-Tk taxonomy; annotation by Prokka; MAG abundances by CoverM.
• ARG identification: RGI (CARD v3.1.4) on gene catalog and MAGs. Multidrug class defined for ARGs conferring resistance to ≥2 classes (MLS grouped separately). Ten clinically important ARGs (including mcr, sul1, tet variants, ErmB/T, aminoglycoside phosphotransferases/adenylyltransferases) profiled via BLASTP to SARG seed sequences.
• MGE identification: DIAMOND against MobileGeneticElementDatabase (e-value ≤ 1e-5, typically identity/coverage >80%; relaxed e-value-only threshold used for M. hyopneumoniae MAGs due to sensitivity). Prophages in M. hyopneumoniae via PHASTER.
• VFG identification: BLASTP against VFDB (e-value ≤ 1e-5, identity ≥80%).
• ARG–MGE relationships: Co-localization on contigs; co-abundance networks via Spearman correlations (r ≥ 0.5, FDR < 0.05) across samples.
• Cross-microbiome comparison: Pig gut (6339 MAGs from 500 metagenomes), human lung BAL metagenomes (46 children with pneumonia; 118 adult COVID-19 patients) analyzed for ARG–MGE linkages.
• qPCR validation: Primers for ARGs (tetM, APH(6)-Id, ANT(3")-IIa), MGEs (Tn916-orf13, Int-Tn916), and MAGs (Moraxellaceae MAG_21, Jeotgalicoccus A schoeneichii MAG_26, Escherichia coli MAG_340) in 23 samples; 16S rRNA control; ΔCt compared to metagenomic abundances (Spearman).
• Statistics: Alpha diversity (richness, Shannon) with vegan; Procrustes for β-diversity alignment; Wilcoxon tests for group comparisons (FDR correction); network visualization with Gephi; gene maps with ggplot2/gggenes.

• Catalog and prevalence: From 745 metagenomes, 10,337,194 non-redundant genes were cataloged; 1228 ORFs mapped to 372 distinct ARGs spanning 24 resistance classes. 55% of ARGs conferred resistance to one class; 45% to ≥2 classes (including 4% MLS). Tetracycline (33%) and aminoglycoside (22%) resistance dominated by abundance, followed by phenicol (14%) and multidrug (13%). Twelve ARGs were present in >75% of samples; >57% of ARGs occurred in <10% of samples. Top-abundant ARGs included floR, tet(39), tet(L), tetQ, tet(D). Major mechanisms were antibiotic efflux (39.8%) and inactivation (33.3%). Control samples yielded only adeF, TEM-181, TEM-237; adeF sequence in controls differed from those in experimental samples (≤72.3% identity), minimizing contamination concerns.
• Clinically important ARGs: Multiple mcr subtypes (mcr-1, mcr-1.2, mcr-3.4, mcr-4) detected with variable prevalence. sul1 and ErmB exhibited multiple ORFs with variable prevalence across samples.
• MGEs: Identified 3016 MGE ORFs representing 83 MGEs across 23 types. Transposases dominated (79% of MGE abundance). Plasmid genes were identified (105 ORFs; 19 types) but low abundance (~1%). MGE richness correlated with ARG alpha-diversity.
• ARG–MGE relationships: Direct contig co-localization was rare (16 ARG ORFs co-occurring with MGEs) due to short contigs. Co-abundance analysis revealed significant correlations between 59 ARGs and 25 MGEs (r ≥ 0.5, FDR < 0.05). Twelve tetracycline ARGs (e.g., tet(W/N/W), tetQ, tetM) were linked to Tn916-family elements (multiple orfs). Close physical linkage (<5 kb) between Tn916-family genes and tet was shown in MAGs classified as Moraxellaceae and Jeotgalicoccus schoeneichii. floR associated with Tn916-family MGEs. aminoglycoside ARGs (aadA27, APH(6)-Id, APH(3")-Ib, ANT(3")-IIc/IIa) correlated with multiple MGEs. adeF correlated with 13 MGEs, including transposases (tnpA variants). The transposase tnpA associated with 20 ARGs across aminoglycoside, tetracycline, and sulfonamide classes; close contig proximity exemplified by tnpA–sul2 in Acinetobacter towneri (MAG_75). repUS12 (plasmid) associated with ANT(4)-Ib, ANT(6)-Ia, APH(2")-If, ErmT, tet(45), tet(L); tet(L) coexisted with repUS12 on a contig.
• qPCR validation: Metagenomic abundances were corroborated by qPCR for nearly all tested ARGs and MGEs (except tetM showed a trend: R=0.35, P=0.099). Relationships among ARGs, MGEs, and their MAG hosts were supported.
• Cross-site and cross-host transfer potential: In pig gut MAGs, Tn916 (orf8/9) linked to tet on contigs, and tnpA was proximal to various ARG types, especially aminoglycoside. In human BAL metagenomes (pneumonia and COVID-19), adeF and tnpA were prevalent; tnpA closely linked to aminoglycoside (APH(3)-IIIa, ANT(4')-Ib, APH(6)-Id), MLS (ErmB), and phenicol (QnrSI) ARGs. Tn916–tet linkages were observed in Streptococcus contigs and COVID-19 BAL contigs (multiple Tn916 orfs adjacent to tetM). Survey of 3878 isolated genomes showed widespread tetM co-occurrence with Tn916 in Enterococcus faecalis (105/168), E. faecium (179/297), Streptococcus suis (14/94), Staphylococcus aureus (124/976), and Escherichia coli (38/1172); transposase tnpA comprised 72.5% of 78,514 MGE ORFs and was commonly near diverse ARGs.
• ARG host ranges: Of 1209 contigs with ARG ORFs, 1183 were bacterial; Proteobacteria (notably Gammaproteobacteria) carried 53% of ARGs. No ARGs were detected in Tenericutes despite high abundance of Mycoplasma. Pseudomonas aeruginosa carried 61 ARG ORFs, E. coli 52. Fourteen Acinetobacter species carried 97 ARGs. The five most abundant ARGs (floR, tet(39), tet(L), tetQ, tet(D)) had broad host ranges across phyla, suggesting HGT. From 397 MAGs, 115 carried 416 ARG ORFs (152 ARGs); 62 MAGs had >2 ARGs, 11 had >5. Three E. coli MAGs (MAG_340/368/389) each harbored >50 ARGs and multiple homologous multidrug efflux clusters (emr, mdt, Acr families); they also contained numerous MGEs (e.g., MAG_340 with 35 MGEs including multiple tnpA), with dense ARG–MGE co-localization within a 22-kb region.
• Lung vs trachea: In 69 paired pigs, tracheal microbiomes had significantly higher ARG richness, total abundance, and Shannon diversity than lungs; 78% of ARGs had higher prevalence in trachea. Forty ARGs differed significantly, largely higher in trachea and often multidrug/aminoglycoside resistance; nine plasmid-borne OXA beta-lactamases were common in trachea but nearly absent in BAL. MGEs showed analogous patterns (higher richness, abundance, diversity in trachea). Distinct ARG and MGE compositions were observed by β-diversity/PCoA.
• Population differences: Among five pig populations, no significant differences in ARG/MGE richness, total abundance, or alpha diversity, but distinct β-diversity and PCoA separation. Dominant ARGs varied by population (e.g., floR in F7; tet(L) in Berkshire × Licha and wild boars; tet(39) in Erhualian; tetQ in Tibetan). tnpA was the most abundant MGE across populations.
• ARGs and lung lesions: In 613 BAL samples (F7 population), ARG alpha-diversity decreased with increasing lesion severity, paralleling decreased microbial diversity and the dominance of Mycoplasma hyopneumoniae (which carried no ARGs). ARG alpha-diversity correlated strongly with species richness (r=0.69–0.86, P<2.2×10^-16) and moderately with Shannon (r≈0.27–0.30) and microbial gene richness. Procrustes analysis showed weak alignment of ARG and microbial β-diversity (r=0.15), improving after excluding M. hyopneumoniae (r=0.57). Phenicol resistance abundance increased with severity; aminoglycoside resistance decreased. Seventy-three ARGs differed among lesion groups (FDR<0.05); floR and Haemophilus influenzae PBP3 beta-lactam resistance (Hinf_PBP3_BLA) were enriched in severe lesions.
• Mycoplasma hyopneumoniae virulence and MGEs: No ARGs were found in seven M. hyopneumoniae MAGs or 23 RefSeq genomes. Nine virulence factor genes (VFGs) were detected; adhesion-related VFGs (P97/P102 paralog family, P146, P159, EF-Tu, PDH-B, LppT) were common, with multiple P97/P102 paralogs in most MAGs. Five MGEs (pEC4115, IS91, ISBf10, tnpA, prophage) were detected in these MAGs (under relaxed thresholds), with several VFGs located within 10 kb of MGEs on contigs. Prophage-encoded P146 and P97/P102 paralogs were observed, suggesting MGE-mediated VFG transfer that could enhance pathogenicity. tnpA and IS91, which co-occurred with VFGs, were also broadly associated with ARGs across the microbiomes, indicating potential for ARG acquisition in the future.

This work establishes a comprehensive view of the swine lower respiratory tract resistome and its mobilome context. The predominance of tetracycline and aminoglycoside resistance reflects historical and current swine antibiotic usage. The pervasive associations between tetracycline ARGs and Tn916-family elements, and the broad linkage of diverse ARGs to the transposase tnpA across pig lungs, pig gut, and human lungs, highlight MGEs as key vectors driving horizontal ARG dissemination both across body sites and between species. The identification of E. coli MAGs harboring extensive ARG and MGE repertoires underscores their role as potent reservoirs for multidrug resistance. Differences between tracheal and lung resistomes indicate that the upper airway may serve as a richer reservoir and conduit for ARGs and MGEs, including plasmid-borne beta-lactamases. The decrease in ARG diversity with increasing lung lesion severity appears to be driven by overgrowth of Mycoplasma hyopneumoniae, which lacks ARGs yet harbors adhesion-related VFGs that co-occur with MGEs, suggesting potential MGE-mediated virulence transfer. Population-specific ARG composition patterns further emphasize the influence of host genetics, management, and environment. Collectively, these findings support pigs as a model for studying MGE-mediated ARG transfer relevant to human health and point to MGEs, particularly Tn916 and tnpA, as critical targets for surveillance.

This study assembled the first large-scale catalog of ARGs in the swine lower respiratory tract microbiome and mapped their distribution across taxa and MGEs. Key contributions include: (1) identification of 372 ARGs (1228 ORFs) dominated by tetracycline and aminoglycoside resistance; (2) revelation of extensive ARG–MGE linkages, especially Tn916–tet and tnpA–multiple ARGs, conserved across pig lung, pig gut, and human lung microbiomes; (3) delineation of Proteobacteria, particularly Gammaproteobacteria such as E. coli, as major ARG carriers with high MGE content; (4) demonstration that tracheal microbiomes harbor higher ARG/MGE richness and abundance than lung microbiomes; (5) association of specific ARGs, including floR, with lung lesion severity; and (6) evidence that M. hyopneumoniae, though lacking ARGs, contains adhesion-related VFGs that co-occur with MGEs, suggesting potential for virulence gene transfer. Future work should leverage long-read sequencing to improve contig continuity and resolve ARG–MGE synteny, experimentally validate HGT events and MGE functionality, assess the dynamics of ARG/MGE transfer under antibiotic pressures, and develop interventions focusing on key MGEs and high-risk bacterial hosts to mitigate ARG spread.

The principal limitation is the short contig length (average 1136 bp) due to substantial host DNA contamination and limited microbial sequencing depth, which likely underestimates direct co-localization of ARGs and MGEs and constrains MAG completeness. Some MGE annotations in Mycoplasma hyopneumoniae relied on relaxed thresholds (e-value only), potentially affecting specificity. Sampling was cross-sectional; functional transfer events were inferred from co-abundance and proximity rather than experimentally demonstrated. Antibiotic usage history was controlled only for the two months prior to sampling, and environmental exposures were not exhaustively profiled.