For validation of the CIM, a selection of 30 Gram-negative isolates was used. This selection included isolates obtained from different institutes across the world carrying known carbapenemase encoding genes and carbapenem susceptible isolates, according to the submitter (Table 1). In addition, 694 isolates submitted to the National Institute for Public Health and the Environment for the national surveillance of carbapenemase-producing Enterobacteriaceae (CPE) by Dutch medical microbiology laboratories (MMLs) during the first six months of 2012 and the first six months of 2013 were used. For the national surveillance of CPE in the Netherlands, Dutch MMLs are requested to submit Enterobacteriaceae isolates with an MIC for meropenem > 0.25 μg/ml. However, more than half of the isolates (411/694, 59%) sent in for CPE surveillance were non-fermenting Gram-negatives belonging to the genera Pseudomonas and Acinetobacter. Furthermore, 35% of the isolates had MICs below 0.25 μg/ml. Nevertheless, all isolates were included in this study.

The species identification, as performed by the MMLs, was confirmed using MALDI-TOF (Bruker Daltonics GmbH, Bremen, Germany) and the MIC for all isolates was confirmed by E-test (BioMerieux Inc., Marcy L’Etoile, France). Culturing of isolates was done on Columbia Sheep Blood (bioTRADING Benelux BV, Mijdrecht, The Netherlands) and Mueller-Hinton agarplates (Oxoid Ltd, Hampshire, United Kingdom). An overview of all CPE surveillance isolates and their characteristics is displayed in Tables 2 and 3.

To perform the CIM, a suspension was made by suspending a full 10 μl inoculation loop of culture, taken from a Mueller-Hinton or blood agar plate in 400 μl water. Subsequently, a susceptibility-testing disk containing 10 μg meropenem (Oxoid Ltd, Hampshire, United Kingdom) was immersed in the suspension and incubated for a minimum of two hours at 35°C. After incubation, the disk was removed from the suspension using an inoculation loop, placed on a Mueller-Hinton agar plate inoculated with a susceptible E. coli indicator strain (ATCC 29522) and subsequently incubated at 35°C. Inoculation of the Mueller-Hinton agar plate with the indicator strain was done with a suspension of OD₅₉₅ 1.25 (correlates with a McFarland value of 0.5) streaked in three directions using a sterile cotton swab. If the bacterial isolate produced carbapenemase, the meropenem in the susceptibility disk was inactivated allowing uninhibited growth of the susceptible indicator strain. Disks incubated in suspensions that do not contain carbapenemases yielded a clear inhibition zone. If results are required within the same day, they can be read after six hours, but within the setting of our laboratory, we prefer reading results after overnight incubation (Fig. 1).

A multiplex PCR that detects the genes encoding the predominant carbapenemases KPC, NDM, OXA-48like, VIM and IMP was used. To serve as an internal positive control, primers detecting the bacterial 23S rRNA gene were incorporated. This multiplex PCR was developed in-house and combines previously published (OXA-48L [10], IMP [11] and 23S [16]) and newly designed primers (KPC, NDM and VIM). For PCR, 12.5 μl of QiaGen Multiplex mix (QiaGen GmbH, Hilden, Germany) was mixed with 2 pmol of forward and reverse primers for KPC, OXA-48, VIM and 23S and with 6 pmol of the forward and reverse primers for NDM and IMP and supplemented with water to reach a final volume of 23 μl. Two μl of boilate from an isolate was added to the mix and 25 PCR cycles of 60 sec denaturation at 95°C, 60 sec annealing at 57.5°C and 60 sec elongation at 72°C were run. In addition, Acinetobacter isolates were tested for the presence of the gene encoding for OXA-23 in a separate PCR [12]. Finally, for isolates that demonstrated carbapenemase activity using the CIM, but did not yield a PCR product using the multiplex PCR described above, three additional PCRs were performed. These separate PCRs, targeting BIC [11] and additional variants of OXA-48L and IMP [17], were performed using 10 pmol of each primer and the same PCR conditions as the multiplex PCR.

The primer sequences and the expected PCR product sizes are listed in Table 4.

Four isolates with discrepant results for CIM in comparison to PCR, an E. coli, K. pneumoniae, A. baumannii and a Proteus mirabilis isolate, have been analyzed using next-generation sequencing.

DNA was extracted using the QuickExtract Bacterial DNA Extraction Kit (Epicentre, Madison, USA) according to the manufacturer’s instructions, followed by proteinase K (QiaGen GmbH, Hilden, Germany) digestion and precipitation. Library preparation and sequencing of bacterial genomes was performed by BaseClear (Leiden, the Netherlands) using the Illumina Nextera XT kit and the HiSeq 2500 with a paired-end 100 cycles protocol. For detection of beta-lactamase genes, a reference list of relevant genes was composed based on those listed on the Lahey Clinic website (http://www.lahey.org/Studies/). Trimmed sequence reads were mapped to this reference list, after which mapped reads were extracted, assembled into contigs and antibiotic resistance determinants were identified by BLAST analysis of the resulting contigs against the same reference list. Sequence reads were deposited to the European Nucleotide Archive (ENA) under project accession PRJEB8575 (http://www.ebi.ac.uk/ena/data/view/PRJEB8575).

A selection of isolates has also been tested for carbapenemase activity with Carba NP and CarbAcineto NP using the most recently published protocols [18,19]. In short, a suspension of a fourth to a third of a 10 μl loop of culture in 100 μl of commercially available lysis buffer (B-PERII, Thermo Scientific Pierce, Rockford, USA) was mixed with 100 μl of a phenol red solution of pH 7.8 containing 0.6 mg imipenem monohydrate. If a carbapenemase is present in the solution, the hydrolysis of the imipenem will lower the pH, causing the phenol red to turn from red to orange or yellow. For Acinetobacter, this protocol has two modifications: the use of a 5 M NaCl solution instead of the B-PERII lysis buffer and a full 10 μl loop of culture as an inoculum.

To validate the robustness of the CIM, the influence of a number of variables on the outcome of the assay was determined. First, the required density of the bacterial suspension used to immerse the meropenem disk was assessed. For this purpose, we used three isolates: two isolates yielding an OXA-48 PCR product: a K. pneumoniae with an MIC of 32 μg/ml and an E. cloacae with an MIC of 1.5 μg/ml for meropenem and E. coli ATCC 25922 as a negative control. Bacterial suspensions with optical density OD₅₉₅ 2.6, OD₅₉₅ 13 (density obtained with a loop full of bacteria), and OD₅₉₅ 65 were used (OD₅₉₅ 2.6 correlates with a McFarland value of 11.3). The suspension with OD₅₉₅ 2.6 yielded negative results for both OXA-48 positive isolates, whereas both OD₅₉₅ 13 and OD₅₉₅ 65 yielded positive CIM results. The CIM remained negative for all densities of E. coli ATCC 25922.

A subset of ten isolates from the selection in Table 1 was used to determine the influence of a number of other variables. First, the incubation time of the meropenem disk in the bacterial suspension was varied from 30 minutes to overnight. Results indicated that for this subset, 30 minutes were sufficient to obtain correct results. After overnight incubation, the results could still be assessed, although decreased inhibition zones for the negative control indicated partial auto-hydrolysis of the meropenem in the disk. However, this was only tested with a small subset to illustrate the robustness of the method. To improve detection in cases of low-level carbapenemase activity and to obtain an assay that can still be interpreted within one working day, a minimum incubation time of 2 hours was chosen. Incubation temperatures of 30°C, 35°C and 37°C all yielded identical results. Suspensions made from isolates grown on Mueller-Hinton agar and Columbia sheep blood agar plates both resulted in the correct detection of carbapenemase activity. However, in cultures grown on Oxoid CRE plates (Oxoid Ltd, Hampshire, United Kingdom), which contain a chromogenic substrate and a carbapenem, carbapenemase activity could not be detected in four out of nine carbapenemase positive isolates. Suspensions made from cultures that had been maintained refrigerated on agar plates for up to two weeks gave identical results as suspensions made from fresh cultures. Similarly, bacterial suspensions made from fresh cultures that were subsequently stored for up to two weeks at 4°C, retained their capability to inactivate the carbapenem in the disk. The CIM was performed using susceptibility-testing disks containing 10 μg meropenem from two different suppliers (Oxoid Ltd, Hampshire, United Kingdom and Mast Group Ltd, Merseyside, United Kingdom), which yielded identical results. The addition of zinc sulfate to the suspension, used to stimulate metallo-betalactamases such as NDM, was tested but no beneficial effect could be demonstrated. Finally, to assess the influence of human skill on the performance of the CIM, different laboratory staff tested the previously mentioned selection of 30 Gram-negative isolates (Table 1) and no discrepancies were found.

Of the 283 Enterobacteriaceae included in the study, 70 (24.7%) were shown to carry a carbapenemase encoding gene by PCR and all PCR-positives (100%) were shown to produce carbapenemase by the CIM (Table 2). A Proteus mirabilis isolate with an MIC of 0.094 μg/ml for meropenem yielded a VIM PCR product, but was initially negative for CIM. Like all isolates for which PCR and CIM yielded discrepant results, the analysis was repeated and CIM was positive in the second analysis. Two other isolates with MICs ≤ 0.25 μg/ml (K. pneumoniae and E. coli) yielded an OXA-48 PCR product and both isolates were shown to produce carbapenemase. Three isolates (two K. pneumoniae and one E. coli) produced carbapenemase, but were PCR-negative using our multiplex PCR. However, they did yield an OXA-48-like PCR product using a modified reverse primer (OXA-48L-R2, Table 4). Sanger sequencing of the PCR product obtained with this modified primer revealed that all three had allele OXA-181, which cannot be detected with the original primer set. Nearly half of the K. pneumoniae (47.7%) and E. coli (48.8%) isolates included in this study yielded a PCR product representing a carbapenemase encoding gene and all these isolates were CIM-positive. However, we detected a carbapenemase encoding gene in only a minor fraction of the E. cloacae (5.6%) and the other Enterobacteriaceae (3.8%). Again, all PCR-positives were shown to produce carbapenemase by CIM. Of the 26 OXA-48 positive K. pneumoniae isolates, five had a mucoid colony morphology, but this did not affect the ability of CIM to detect carbapenemase activity. Forty Enterobacteriaceae isolates (14.1%) had MICs for meropenem that were above the EUCAST clinical breakpoint of resistance (>8 μg/ml). In 14 of these (35.0%) no carbapenemase encoding gene or carbapenemase activity was detected.

The collection contained a large number of non-fermenting Gram-negatives all belonging to the Pseudomonas and Acinetobacter genera (Table 3). Sixty-seven of the 411 isolates (16.3%) yielded a PCR product and 65 of these samples (97.0%) yielded a positive CIM. Only in two A. baumannii isolates carrying bla_OXA-23, carbapenemase activity could not be detected with CIM. All 67 isolates had MICs >0.25 μg/ml. Among the Pseudomonas isolates, VIM was the most frequently detected carbapenemase encoding gene, present in 49 of the 387 isolates (12.7%). In A. baumannii, OXA-23 was the predominant carbapenemase encoding gene, detected in 12 of the 18 isolates (66.7%). None of the other A. spp. were PCR positive. Of the 344 PCR-negative non-fermenter isolates, only seven were CIM-positive (2.0%) and all of these isolates had MICs >0.25 μg/ml. Three of these isolates were found to contain a carbapenemase-encoding gene using additional PCRs; two Pseudomonas isolates were BIC positive [11] and one Pseudomonas isolate was found to carry an IMP variant not detected by our multiplex PCR) [17]. In addition, one CIM-positive A. baumannii isolate was found to contain an OXA-72 gene by next-generation sequencing. Compared to the Enterobacteriaceae, a larger portion of the non-fermenter isolates had MICs for meropenem that were above the EUCAST clinical breakpoint for resistance (222/411, 54.0%). Yet, 164 (73.8%) of these were negative in both PCR and CIM.

A selection (n = 116) of the isolates subjected to CIM has also been tested with Carba NP or CarbAcineto NP using the most recently published protocols [18,19]. These isolates include all isolates used for validation (Table 1), all OXA-48 positive K. pneumoniae isolates as well as all CIM/PCR discrepant isolates, supplemented with a random selection of other isolates used (Table 5). Four isolates (3.4%) gave different results for CIM than for Carba NP. Three of these were isolates found to contain a carbapenemase encoding gene (two bla_BIC and one bla_OXA-48like) where CIM was positive and Carba NP was (false-) negative. The fourth isolate was a CIM positive Pseudomonas spp. for which no carbapenemase encoding gene was identified, but carbapenemase activity could not be confirmed with Carba NP.