Halogen-transforming potential in prokaryotes partitions into mobile respiratory loci and a broad resident reservoir

Reductive dehalogenase homologs (RdhA) are widespread and structurally diverse across prokaryotes, but it remains unclear whether their major architectural classes differ systematically in how mobile they are, how often they transfer, where they sit taxonomically, and which habitats they occupy. We screened GTDB r220 and detected candidate rdhA in 1,576 genomes across 39 phyla; 2,392 homologs in 1,409 genomes across 36 phyla passed motif validation and constitute the analysed set. Classification by motif architecture and neighbourhood revealed four genomic classes: canonical respiratory rdhA-rdhB loci (rdhA + rdhB + TAT signal; 163 proteins), solitary catabolic-like homologs (1,845), tandem-array loci (153), and other architectures (231). Canonical respiratory rdhA-rdhB loci were the most mobile subset: 48.5% carried flanking mobility elements compared with 11.3% of catabolic-like homologs, a contrast that holds within individual phyla (Pseudomonadota 42.9% vs. 11.2%; Chloroflexota 65.5% vs. 11.8%) and is therefore not a taxonomic artefact. Their transfers also reached further: 56% crossed phylum boundaries against 13% for catabolic-like homologs, a 2.8-fold enrichment once the phylum composition of the dataset is accounted for. Five phylogenetic tests confirmed pervasive horizontal transfer across the family, with the strongest discordance in the canonical respiratory subset. DTL reconciliation assigned 74% of events as transfers, the AU test rejected the species tree, and normalized Robinson-Foulds distance was 0.82 on the full tree, rising to 0.90 in the respiratory motif class. These results support a model in which the RdhA superfamily contains a disproportionately mobile canonical respiratory subset embedded within a broader and less mobile set of catabolic-like homologs, while tandem arrays are associated mainly with within-lineage expansion. Metagenomic prediction of dechlorination potential should therefore distinguish these architectural classes rather than treating all rdhA detections equivalently.