Abstract / Summary
Abstract Background Species-level microbiome profiling may overlook strain-level genomic variation relevant to rheumatoid arthritis (RA). We investigated whether microbial structural variants (SVs) provide additional information beyond microbial taxonomy. Methods We analyzed shotgun metagenomic data from four independent RA cohorts comprising 491 samples. Microbial SVs were identified using SGV-Finder and quantified across cohorts. Species- and SV-level associations were evaluated by cross-cohort meta-analysis. Meta-analysis-prioritized features were subjected to LASSO selection and random forest classification using combined training cohorts, followed by validation in an independent cohort. Selected SV-associated loci were functionally characterized by genomic annotation, motif analysis, structural modeling, and biolayer interferometry. Results We identified 4,105 SVs, including 2,860 variable and 1,245 deletion SVs. Meta-analysis identified 49 species and 59 SVs associated with RA. Most RA-associated SVs were not accompanied by differential species abundance, with only 7 variable and 4 deletion SVs overlapping species-level associations. Incorporating SVs improved RA discrimination, with AUCs of 0.82 versus 0.71 in training cohorts and 0.62 versus 0.48 in the validation cohort. Functional characterization identified strain-level loss of a retron-associated reverse transcriptase locus in Faecalibacterium prausnitzii and an XRE-family transcriptional regulator locus in Agathobacter rectalis . Biolayer interferometry confirmed sequence-specific DNA binding by the XRE regulator, which was attenuated by motif mutation. Conclusions Microbial SVs capture RA-associated strain-level genomic variation beyond species abundance and improve disease discrimination, providing an additional genomic layer for microbiome-based characterization of RA.