Antimicrobial Resistance and Transferability in Oral-Care Probiotic Strains and Products: A Narrative Review

Objective: To synthesize evidence on phenotypic non-susceptibility, antimicrobial-resistance genes (ARGs), genetic mobility, transferability, and oral-resistome effects associated with oral-care probiotic strains and products. Design: A focused narrative search of PubMed, Google Scholar, publisher platforms, and reference lists was conducted through 3 August 2026, covering product-level, strain-safety, oral-biofilm/resistome, regulatory, and mechanistic studies; non-oral probiotic data were treated as indirect evidence. Results: A survey of eight commercial oral probiotic lozenges reported non-susceptibility phenotypes and PCR-detected ARG-like sequences without demonstrated transfer to selected bacterial recipients. Strain-level assessments of Streptococcus salivarius K12, DB-B5, G7, and M18, and of Weissella cibaria CMU/CMS1, generally did not identify acquired transferable determinants. Limosilactobacillus reuteri DSM 17938 illustrates risk mitigation: its parent strain's tet(W)/lnu(A) plasmids were removed by classical curing, and subsequent antibiotic-selected reduced erythromycin susceptibility was not linked to known transferable genes or transferred to Enterococcus faecalis. A three-dimensional oral biofilm study found a transient reduction in active ARG abundance during peak K12 colonization, with no increased putative mobilization. Transfer between non-oral probiotic donors and bacterial recipients has been demonstrated only after laboratory adaptation under high streptomycin pressure. Conclusions: No study identified in this review demonstrated donor-to-recipient ARG transfer from an oral-care probiotic during human use, though this absence of evidence should not be read as proof that transfer cannot occur. Several products, including the widely used ATCC PTA 5289, remain incompletely characterized, and long-term effects on the human oral resistome are unknown. Standardized whole-genome, phenotypic, mobility, finished-product, and longitudinal human assessments are needed.