Bioremediation of perchlorate-contaminated groundwater can occur via bacterial reduction of perchlorate to chloride. Although perchlorate reduction has been demonstrated in bacterial pure cultures, little is known about the efficacy of using perchlorate-reducing bacteria as inoculants for bioremediation in the field. A pilot-scale, fixed-bed bioreactor containing plastic support medium was used to treat perchlorate-contaminated groundwater at a site in Southern California. The bioreactor was inoculated with a field-grown suspension of the perchlorate-respiring bacterium Dechlorosoma sp. strain KJ and fed groundwater containing indigenous bacteria and a carbon source amendment. Because the reactor was flushed weekly to remove accumulated biomass, only bacteria capable of growing in biofilms in the reactor were expected to survive. After 26 days of operation, perchlorate was not detected in bioreactor effluent. Perchlorate remained undetected by ion chromatography (detection limit 4 mug L(-1)) during 6 months of operation, after which the reactor was drained. Plastic medium was subsampled from top, middle, and bottom locations of the reactor for shipment on blue ice and storage at -80 degrees C prior to analysis. Microbial community DNA was extracted from successive washes of thawed biofilm material for PCR-based community profiling by 16S-23S ribosomal intergenic spacer analysis (RISA). No DNA sequences characteristic of strain KJ were recovered from any RISA bands. The most intense bands yielded DNA sequences with high similarities to Dechloromonas spp., a closely related but different genus of perchlorate-respiring bacteria. Additional sequences from RISA profiles indicated presence of representatives of the low G+C gram-positive bacteria and the Cytophaga-Flavobacterium-Bacteroides group. Confocal scanning laser microscopy and fluorescence in situ hybridization (FISH) were also used to examine biofilms using genus-specific 16S ribosomal RNA probes. FISH was more sensitive than RISA profiling in detecting possible survivors from the initial inoculum. FISH revealed that bacteria hybridizing to Dechlorosoma probes constituted <1% of all cells in the biofilms examined, except in the deepest portions where they represented 3-5%. Numbers of bacteria hybridizing to Dechloromonas probes decreased as biofilm depth increased, and they were most abundant at the biofilm surface (23% of all cells). These spatial distribution differences suggested persistence of low numbers of the inoculated strain Dechlorosoma sp. KJ in parts of the biofilm nearest to the plastic medium, concomitant with active colonization or growth by indigenous Dechloromonas spp. in the biofilm exterior. This study demonstrated the feasibility of post hoc analysis of frozen biofilms following completion of field remediation studies.