The potential of intestinal bacteria to hydrolyze ginsenoside Rb-1 to 20-O-beta-D-glucopyranosyl-20(S)-protopanaxadiol (I) was found in 79 % of the fecal specimens from 58 human subjects whose age ranged from 1 to 64 years. Following a ginsenoside-Rb-1-hydrolyzing activity assay, Prevotella oris strains were then isolated as a major bacterial species possessing the potential. All the intestinal isolates converted ginsenosides Rb-1 and Rd to I, ginsenoside Rb-2 to 20-O-[alpha-L-arabinopyranosyl-(1-->6)-beta-D-glucopyranosyl]-20(S)-protopanaxadiol (II), and ginsenoside Re to 20-O-[alpha-L-arabinofuranosyl(1-->6)-beta-D-glucopyranosyl]-20(S)-protopanaxadiol (III) like fecal microflora, but did not attack ginsenosides Re or Rg(1) (protopanaxatriol-type). The isolates were susceptible to colimycin (MIC, 3.13 mu g/(m)l) and then the treatment of specific pathogen free mice with colimycin (20 mg/kg/day) decreased intestinal bacterial Rb-1-hydrolyzing potential from 22.1 +/- 1.2 % to 4.7 +/- 2.7 %, while the decreased potential was restored to 30.7 +/- 3.7% by the inoculation with P. oris isolates, These results suggest that the metabolism of protopanaxadiol saponins to metabolites I-III in the intestines seems most partly due to intestinal P. oris. In addition, the fact that neither intact ginsenoside Rb, nor its middle metabolic derivatives but only the final metabolite I was detected at 1.0 - 7.3 mu g/ml in blood after oral administration of mice with ginsenoside Rb-1 (125 mg/kg) allows us to speculate that metabolites I-III are the most likely forms of protopanaxadiol saponins absorbed from the intestines.