During in vitro susceptibility testing of clinical isolates of Proteus vulgaris, we noted that the MICs of several expanded-spectrum cephems were much higher in the broth microdilution method than in the agar dilution method (termed the MIC gap phenomenon). Here we investigated the mechanism of the MIC gap phenomenon. Cephems with the MIC gap phenomenon were of the oximino type, such as cefotaxime, cefteram, and cefpodoxime, which serve as good substrates for inducible class A -lactamase (CumA) enzymes produced by P. vulgaris; this finding suggests a relationship between the MIC gap phenomenon and CumA. Since peptidoglycan recycling shares a system common to that inducing CumA, we analyzed the mechanism of the MIC gap phenomenon using P. vulgaris B317 and isogenic mutants with mutations in the peptidoglycan recycling and -lactamase induction systems. The MIC gap phenomenon was observed in the parent strain B317 but not in B317G (cumG-defective mutant; defective peptidoglycan recycling) and B317R (cumR-defective mutant; defective CumA transcriptional regulator). No -lactamase activity was detected in B317G and B317R. -Lactamase activity and the MIC gap phenomenon were restored in B317G/pMD301 (strain transcomplemented by a cloned cumG gene) and B317R/pMD501 (strain transcomplemented by a cloned cumR gene). MICs determined by the agar dilution method increased when lower agar concentrations were used. Our results indicated that the mechanism of the MIC gap phenomenon is related to peptidoglycan recycling and CumA induction systems. However, it remains unclear how -lactamase induction of P. vulgaris is suppressed on agar plates.