Full-assembly screening reveals mobile antibiotic-resistance cargo missed by chromosome-only genomes

BackgroundProbiotic bacteria occupy the same gut niches as enteric pathogens, prompting concern that probiotic strains might carry or contribute mobile antibiotic-resistance genes (ARGs). Genome-based screening is routinely used to assess this risk, but many screens use chromosome-level assemblies that may omit plasmid-borne, high-mobility cargo. We quantified this effect and compared the mobile context resistomes of probiotic-associated and pathogen reference genomes. MethodsWe screened 50 bacterial reference genomes (25 probiotic-associated, 25 pathogen/comparator) using a reproducible workflow with the CARD nucleotide catalog, PlasmidFinder replicons, and ISfinder insertion sequences. Each ARG was assigned a fourtier in silico mobile-context risk category from plasmid co-localization and insertion-sequence (IS) flanking. The identical strain panel was screened in matched full-assembly and chromosome-only modes. Acquired calls were curated against intrinsic/efflux/biocide determinants and cross-checked with AMRFinderPlus, ResFinder, and targeted BLAST, with MOB-suite as a plasmid/mobility overlay. ResultsFull-assembly screening detected 373 ARG loci versus 338 in chromosome-only mode on the same strains, increasing High-risk calls from 4 to 15 and recovering 32 plasmid replicons (chromosome-only: 0). All 15 High-risk mobile-context loci occurred in pathogen/comparator genomes and none in probiotic-associated genomes; no ARG was shared across groups at [&ge;]95% nucleotide identity (0/175 edges). Per-strain ARG burden was higher in pathogen genomes (mean 12.32 versus 0.52 loci; Mann-Whitney U = 606.5, P< 0.001). Most priority High-risk loci were corroborated by one or more external tools, with discordant calls retained explicitly as flagged records. ConclusionsChromosome-only screening materially undercounts mobile ARG cargo. In this reference-genome panel, high-risk mobile-context loci were concentrated in pathogen/comparator genomes -- an in silico reference-genome-level safety signal rather than evidence about commercial products or genetic transfer. Data SummaryNo new sequencing data were generated; all genomes are publicly available reference assemblies from NCBI RefSeq. O_LIGenome accessions for all 50 genomes are in Supplementary Table S1. C_LIO_LISource code (screening, validation, and figure scripts, including make_figures.py) is available at https://github.com/abdullahak07/1dprob. C_LIO_LIProcessed result tables are summarised in Supplementary Tables S2-S10. C_LIO_LIExternal validation outputs (AMRFinderPlus, ResFinder, BLAST, MOB-suite) are provided as supplementary data. C_LIO_LITool/database versions and detection thresholds are in Supplementary Table S9. The authors confirm that all supporting data, code, and protocols are available within the article, the cited repositories, or the supplementary material. C_LI O_TBL View this table: org.highwire.dtl.DTLVardef@b0f2beorg.highwire.dtl.DTLVardef@110c676org.highwire.dtl.DTLVardef@5578ceorg.highwire.dtl.DTLVardef@16e448forg.highwire.dtl.DTLVardef@573000_HPS_FORMAT_FIGEXP M_TBL O_FLOATNOSupplementary Table S9:C_FLOATNO O_TABLECAPTIONtool/database versions and detection thresholds used in the SCALE50 analysis. Database snapshots and external validation outputs should be archived with the supplementary data to ensure reproducibility. C_TABLECAPTION C_TBL Impact StatementGenome-based screening is widely used to judge whether a bacterial strain carries transferable antibiotic-resistance genes, including in the safety assessment of probiotic-associated species. We show that the assembly level chosen for such screening materially affects the result: on an identical panel of 50 reference genomes, chromosome-only analysis recovered fewer than a third of the high-risk, mobile-context resistance loci detected when plasmid replicons were included. Applying full-assembly screening with transparent, multi-tool external validation, high-risk mobile-context resistance genes were concentrated in pathogen/comparator genomes and absent from the probiotic-associated genomes in this panel, with no cross-group sharing at high identity. These observations argue for full-assembly inputs and explicit mobile-context interpretation in genome-based resistance screening and provide a cautious, reference-genome-level safety signal; they are not claims about commercial products and do not demonstrate genetic transfer.