We evaluated a strategy for achieving complete reduction of perchlorate (ClO4–) in the presence of much higher concentrations of sulfate (SO42–) and nitrate (NO3–) in a hydrogen-based membrane biofilm reactor (MBfR). Full ClO4– reduction was achieved by using a two-stage MBfR with controlled NO3– surface loadings to each stage. With an equivalent NO3– surface loading larger than 0.65 ± 0.04 g N/m2-day, the lead MBfR removed about 87 ± 4% of NO3– and 30 ± 8% of ClO4–. This decreased the equivalent surface loading of NO3– to 0.34 ± 0.04–0.53 ± 0.03 g N/m2-day for the lag MBfR, in which ClO4– was reduced to nondetectable. SO42– reduction was eliminated without compromising full ClO4– reduction using a higher flow rate that gave an equivalent NO3– surface loading of 0.94 ± 0.05 g N/m2-day in the lead MBfR and 0.53 ± 0.03 g N/m2-day in the lag MBfR. Results from qPCR and pyrosequencing showed that the lead and lag MBfRs had distinctly different microbial communities when SO42– reduction took place. Denitrifying bacteria (DB), quantified using the nirS and nirK genes, dominated the biofilm in the lead MBfR, but perchlorate-reducing bacteria (PRB), quantified using the pcrA gene, became more important in the lag MBfR. The facultative anaerobic bacteria Dechloromonas, Rubrivivax, and Enterobacter were dominant genera in the lead MBfR, where their main function was to reduce NO3–. With a small NO3– surface loading and full ClO4– reduction, the dominant genera shifted to ClO4–-reducing bacteria Sphaerotilus, Rhodocyclaceae, and Rhodobacter in the lag MBfR.