Deciphering pervasive domestication and active “dark matter” of proviruses in prokaryotes via precision border mapping

Proviruses are widespread components of prokaryotic genomes, yet their characterization is impeded by border identification errors. We developed ProBord (Provirus Border Delimiter), a tool achieving single-base-precision border mapping via cross-genome attB site alignment. Using ProBord, we built a database of 430,134 border-delimited proviruses (BDPs). Our analysis revealed that BDPs carry diverse functional genes, including virulence and antibiotic resistance determinants, as well as defense and antidefense systems, with position-specific distributions. Furthermore, comparing BDPs with border-undelimited proviruses (BUPs), whose boundaries could not be precisely delimited by ProBord, revealed a universal pattern of proviral domestication: BUPs selectively lose viral core genes but retain host-beneficial functions through rapid deletion or progressive mutation. Critically, integrating the tool ProAct (Provirus Activity Detector) with ProBord enabled reverse mining of 101 “dark” proviruses (DPs), which are overlooked by mainstream prediction tools. Most of these DPs were associated with pathogenic bacterial hosts. This integrated framework combines tools, theory, and resources—including more than 430,000 BDPs, 4910 active proviruses, and 101 DPs—to decipher the composition, function, and evolution of proviruses in prokaryotes.