Proteus mirabilis - analysis of a concealed source of carbapenemases and development of a diagnostic algorithm for detection.

Objectives: To analyse carbapenemases in Proteus mirabilis and assess the performance of carbapenemase detection assays.

Methods: Eighty-one clinical P. mirabilis isolates with high-level resistance at least to ampicillin (>32 mg/L) or previous detection of carbapenemases were selected and investigated by three susceptibility testing methods (microdilution, automated susceptibility testing, and disk diffusion), six phenotypic carbapenemase assays (CARBA NP, modified carbapenemase inactivation method [CIM], modified zinc-supplemented CIM, simplified CIM, faropenem, and carbapenem-containing agar), two immunochromatographic assays, and whole-genome sequencing.

Results: Carbapenemases were detected in 43 of 81 isolates (OXA-48-like [n = 13]; OXA-23 [n = 12]; OXA-58 [n = 12]; New Delhi metallo-β-lactamase (NDM) [n = 2]; Verona integron-encoded metallo-β-lactamase (VIM) [n = 2]; Imipenemase (IMP) [n = 1]; Klebsiella pneumoniae carbapenemase (KPC) [n = 1]). Carbapenemase-producing Proteus were frequently susceptible to ertapenem (26/43; 60%), meropenem (28/43; 65%), ceftazidime (33/43; 77%), and some even to piperacillin-tazobactam (9/43; 21%). Sensitivity/specificity of phenotypic tests were 30% (CI: 17-46%)/89% (CI: 75-97%) for CARBA NP, 74% (CI: 60-85%)/82% (CI: 67-91%) for faropenem, 91% (CI: 78-97%)/82% (CI: 66-92%) for simplified CIM, and 93% (CI: 81-99%)/100% (CI: 91-100%) for modified zinc-supplemented CIM. An algorithm for improved detection was developed, which demonstrated sensitivity/specificity of 100% (CI: 92-100%)/100% (CI: 91-100%) on the 81 isolates, and 100% (CI: 29-100%)/100% (CI: 96-100%) in a prospective analysis of additional 91 isolates. Interestingly, several OXA-23-producing isolates belonged to the same clonal lineage reported previously from France.

Discussion: Current susceptibility testing methods and phenotypic tests frequently fail to detect carbapenemases in P. mirabilis, which could result in inadequate antibiotic treatment. In addition, the non-inclusion of bla in many molecular carbapenemase assays further impedes their detection. Therefore, the prevalence of carbapenemases in P. mirabilis is likely underestimated. With the herein proposed algorithm, carbapenemase-producing Proteus can be easily identified.