Long-read epicPCR enhances species-level host identification of clinically relevant antibiotic resistance genes in environmental microbial communities.

Identifying clinically relevant antibiotic resistance gene (ARG) hosts in complex microbial communities is crucial for environmental health. EpicPCR (emulsion, paired isolation, and concatenation PCR), a single-cell technology, has advanced this field. However, its traditional format, which links target genes to the V4 region of 16S rRNA genes (∼300 bp), limits species-level identification. To overcome this, we developed "long-read" epicPCR, which links target genes to 16S segments spanning the V4-V9 regions (∼1000 bp) by refining primer pairing strategies to balance amplification length and specificity. We validated this approach by targeting seven clinically relevant ARGs (optrA, tet(X4), mcr-3, NDM-5, KPC-2, IMP-4, and VIM-1), an efflux pump gene (tmexD), and an insertion sequence gene (IS1216E), all confirming correct sequence fusion. Using the optrA gene as a model target, long-read epicPCR demonstrated greater precision and fewer false positives than the short-read method in mock communities. It also significantly improved the identification rate of optrA host species from 29.0 % to 54.4 % in anaerobic digestion reactors, while maintaining consistency with short-read epicPCR in profiling host bacterial communities. Moreover, long-read epicPCR identified two novel optrA host species, Lactobacillus amylotrophicus and Streptococcus alactolyticus, in anaerobic effluents, highlighting potential dissemination risks. Notably, this versatile method is envisioned to enhance targeted antimicrobial surveillance and microbial functional dynamics monitoring in the environment.