Prophage-AMR Co-mobilization Atlas

Five completed AMR projects characterized resistance gene distribution, conservation, fitness costs, and environmental patterns. Separately, the prophage_ecology project mapped prophage distribution across phylogeny. This project connects the two: testing whether AMR genes near prophage markers are more likely to be accessory (mobile) and whether prophage-rich species have broader AMR repertoires. Uses bakta_amr for curated AMR hits, bakta_annotations product-field keywords for prophage marker detection, and gene coordinate proximity for co-localization scoring.

Pangenome-Scale Census

Gene Neighborhood Co-localization

Across 100 species (20 genomes sampled per species, 1,953 genomes total), 36,041 AMR gene instances were analyzed:

H1: Conservation × Proximity

Fisher's exact test (one-sided): OR=1.10, p=0.005. Bootstrap 95% CI: [1.024, 1.185].

H2: Species-Level Breadth

The strongest finding is that prophage density is a powerful species-level predictor of AMR repertoire breadth (H2). This is consistent with two non-exclusive mechanisms: (1) prophages directly mobilize resistance genes via specialized or generalized transduction, and (2) species with high recombination potential (indicated by many prophages) are also better at acquiring genes from other mobile elements. The effect persists after controlling for genome count and is consistent across all five major phyla, arguing against a purely phylogenetic explanation.

The gene-level co-localization result (H1) is more nuanced. While statistically significant in aggregate, the effect is modest (2.1 percentage point difference in accessory fraction) and heterogeneous across species. The reversal at very close range (3-5 genes) may reflect that genes immediately adjacent to phage structural genes are core phage components rather than recently acquired cargo. The strengthening at broader thresholds (15-50 genes) could capture "genomic islands" that include both phage remnants and laterally transferred genes.

Literature Context

Based on articles retrieved from PubMed:

• Our H2 finding aligns with Rendueles et al. (2018) (DOI), who showed that bacteria encoding capsules have more prophages AND more antibiotic resistance genes across >100 pangenomes. Our work extends this by demonstrating the prophage-AMR correlation at much larger scale (4,770 species vs ~100) and showing it persists after controlling for genome count rather than capsule presence.

• Chen et al. (2018) (DOI) found ARG-MGE co-occurrence on assembled contigs from river metagenomes, including associations with prophages. Our pangenome-scale analysis provides a genomic-reference counterpart to their metagenomic observations, confirming that the ARG-prophage association is encoded in reference genomes, not just environmental assemblies.

• Bearson & Brunelle (2015) (DOI) demonstrated that fluoroquinolone exposure induces prophage in multidrug-resistant Salmonella, enabling phage-mediated transduction of resistance plasmids. This provides a direct mechanistic link supporting our correlative finding.

• Fisarova et al. (2021) (DOI) showed that S. epidermidis phages transduce antimicrobial resistance plasmids at high frequency, providing additional mechanistic evidence for phage-mediated AMR spread in clinically relevant organisms.

Novel Contribution

1. Scale: This is the first pangenome-scale analysis of prophage-AMR gene-neighborhood co-localization across 100 species (1,953 genomes, 36,041 AMR gene instances).
2. Species-level predictor: Prophage density explains 30% of variance in AMR breadth across 4,770 species — a stronger predictor than previously documented.
3. Threshold sensitivity: The reversal of the proximity-accessory association at very close range (≤5 genes) suggests that immediate phage neighbors are core phage genes, not acquired cargo. This nuance was not captured in prior contig-level studies.

Limitations

• Gene position proxy: Distances are in ordinal gene positions parsed from gene_id format, not base-pair resolution. True genomic distances may differ.
• Prophage identification: Uses keyword/Pfam matching on bakta_annotations rather than dedicated prophage prediction tools (e.g., PHASTER, geNomad). May include false positives (e.g., phage-defense systems) and miss divergent prophages.
• Sampling: 20 genomes sampled per species for co-localization (NB02). While representative, exhaustive analysis of all 293K genomes would strengthen the findings.
• Core/accessory labels: Depend on species-level pangenome calling (motupan). The same gene in different species may have different conservation status.
• Fitness data gap: H3 could not be tested due to limited overlap between fitness browser organisms and GTDB species.
• Correlation vs causation: The H2 association does not prove phage-mediated AMR transfer; species with open pangenomes may independently accumulate both prophages and AMR genes.

1. Dedicated prophage prediction: Apply geNomad or PHASTER to BERDL genomes and ingest results as a new table. This would replace keyword-based prophage detection with a validated tool and reduce false positives.
2. Base-pair resolution: Use contig sequences (scaffoldseq) to calculate true genomic distances rather than ordinal gene positions.
3. Expanded fitness analysis: As fitness browser coverage expands, revisit H3 to test whether prophage-proximal AMR genes have distinct fitness costs.
4. Plasmid vs phage partitioning: Distinguish between AMR genes mobilized by prophages vs plasmids vs ICEs using the pangenome-level co-localization data.
5. Clinical strain focus: Analyze the WHO priority pathogen subset (K. pneumoniae, A. baumannii, P. aeruginosa, E. coli) in greater depth with all available genomes.

Sources

Generated Data

• Rendueles O, de Sousa JAM, Bernheim A, Touchon M, Rocha EPC. (2018). "Genetic exchanges are more frequent in bacteria encoding capsules." PLoS Genet. 14(12):e1007862. DOI. PMID: 30576310.
• Chen H, Chen R, Jing L, Bai X, Teng Y. (2018). "A metagenomic analysis framework for characterization of antibiotic resistomes in river environment." Environ Pollut. 245:398-407. DOI. PMID: 30453138.
• Bearson BL, Brunelle BW. (2015). "Fluoroquinolone induction of phage-mediated gene transfer in multidrug-resistant Salmonella." Int J Antimicrob Agents. 46(2):201-4. DOI. PMID: 26078016.
• Fisarova L, et al. (2021). "Staphylococcus epidermidis Phages Transduce Antimicrobial Resistance Plasmids and Mobilize Chromosomal Islands." mSphere. 6(3):e00223-21. DOI. PMID: 33980677.
• Price MN, et al. (2018). "Mutant phenotypes for thousands of bacterial genes of unknown function." Nature. 557:503-509. DOI. PMID: 29769716.
• Arkin AP, et al. (2018). "KBase: The United States Department of Energy Systems Biology Knowledgebase." Nat Biotechnol. 36:566-569. DOI. PMID: 29979655.