The Groote Schuur NHLS microbiology laboratory serves the southern part of the greater Cape Town area with a catchment population of approximately 2 million people. The laboratory receives approximately 3600 blood culture samples a month and from these maintains a stored collection of selected organisms. From the 2015–2016 collection of bloodstream isolates, we randomly selected 50 isolates of K. pneumoniae and E. coli, including 40 ESBL-producing and 10 non-ESBL-producing strains of each organism. This categorisation was based on the original identification and susceptibility testing (with Vitek 2 AST-N255 Gram-negative susceptibility card) that was reported in routine diagnostic testing. An isolate was considered an ESBL producer if the phenotypic interpretation by the AES of the Vitek 2 included ESBL with or without decreased outer membrane permeability and not an ESBL if only wild type or β-lactamases other than ESBLs were suggested by AES. For rapid identification of blood culture isolates, the laboratory uses a previously validated method of direct inoculation from a concentrated suspension of the bottle fluid.¹¹ Isolates had been stored as glycerol stocks at -80 °C and were subcultured and re-tested with the Vitek 2 to confirm identification and susceptibility test results. The same inoculum was used for concurrent phenotypic and genotypic testing as described below. This second Vitek 2 result was considered the definitive result for comparison purposes. Control strains for ESBL detection, as described below, were tested concurrently.

Extended-spectrum β-lactamase production was detected by disc diffusion and broth microdilution methods, performed and interpreted according to CLSI criteria.⁶ Appropriate quality control organisms, namely E. coli ATCC ^(R) 25922 and K. pneumoniae ATCC ^(R) 700603, were included for each method in each run.

Disc diffusion testing was performed using the Kirby–Bauer method. Cefotaxime and ceftazidime discs with or without clavulanic acid were utilised, with an increase of ≥ 5 mm in zone inhibition diameter for either cephalosporin in the presence of the inhibitor, indicating the presence of an ESBL.

Broth microdilution was performed using the Sensititre™ ESBL plate format (Trek Diagnostic systems, ThermoScientific Waltham, MA, USA) in accordance with the manufacturer’s instructions. The Sensititre™ ESBL plate includes wells containing cefotaxime and ceftazidime with or without clavulanic acid. A ≥ 3 twofold concentration decrease in MIC for either cephalosporin in the presence of the inhibitor indicates the presence of an ESBL.

The ranges of MICs (in µg/mL) that can be determined using the Sensititre ESBL plate are ≤ 0.25–> 64, ≤ 0.12/4-–>64/4, ≤ 0.25–> 128 and ≤ 0.12/4–> 128/4 for cefotaxime, cefotaxime + clavulanic acid, ceftazidime and ceftazidime + clavulanic acid, respectively. An indeterminate result was reported when it was not possible to calculate the ratio accurately at the limits of MIC range, for example if cefotaxime MIC ≤ 0.25, and cefotaxime + clavulanic acid MIC ≤ 0.12/4. If an indeterminate result was obtained for either cefotaxime or ceftazidime, the isolate was categorised according to the result of the other antibiotic. If indeterminate results were obtained for both antibiotics, the ESBL status was determined by the MICs of the antibiotics, that is, if both cefotaxime and ceftazidime MICs were at the lower limit of the MIC range, the isolate was reported as ESBL negative, whereas if both MICs were at the upper limit of the range, the isolate was reported as ESBL-positive.

Bacterial DNA was extracted from colonies grown on 2% blood agar using the QiaSymphony SP automated extraction platform with the QiaSymphony DSP Virus/Bacteria mini kit, according to the manufacturer’s instructions (Qiagen, Hilden, Germany). To test for the presence of the most commonly occurring ESBL genes bla_CTX-M, bla_SHV and bla_TEM, conventional PCR assays using primers designed to target internal fragments of these genes were performed. Details of the primer sequences, expected amplicon sizes and amplification conditions are given in the Appendix. The previously described positive and negative controls were included in each run. Selected amplicons including any discrepant genotypic-phenotypic results were submitted for DNA sequencing (Inqaba Biotech, Muckleneuk, Pretoria, South Africa) and the data analysed to confirm the gene identity and genotype, where possible.

The sensitivity and specificity of the Vitek 2 AST-N255 Gram-negative susceptibility card for ESBL detection was compared to a composite reference standard, in which an isolate was defined as an ESBL if either an ESBL was detected by both phenotypic methods, that is, by disc diffusion and by broth microdilution, or if an ESBL was detected by either phenotypic method as well as genotypically. The 95% confidence intervals for proportions were calculated according to Newcombe method.¹²

Among the 46 E. coli isolates tested, bla_CTX-M and bla_TEM were detected in 23 and 13 isolates, respectively. Five isolates contained both bla_CTX-M and bla_TEM, whilst bla_SHV was not detected in any isolate. Among the 50 K. pneumoniae isolates tested, bla_SHV was detected in all, with 11 isolates containing no other bla gene. The remaining 39 isolates all contained bla_CTX-M and 27 also contained bla_TEM.

Sequence analysis of a limited number of gene products confirmed the identification of the ESBL-encoding gene bla_CTX-M-15 in four bla_CTX-M amplicons in E. coli and in two bla_CTX-M amplicons in K. pneumoniae. All bla_SHV and bla_TEM amplicons sequenced were identified as narrow-spectrum beta-lactamases in both E. coli (7 bla_TEM-1 isolates) and in K. pneumoniae (5 bla_SHv-1, 1 bla_LEN-17, 1 bla_LEN-19 and 1 bla_TEM-1). No other amplicons were sequenced.

All isolates were tested for ESBL production using both phenotypic methods. Using the Sensititre method, an indeterminate result was obtained with both cefotaxime and ceftazidime for 10 E. coli and 10 K. pneumoniae isolates. These isolates were reclassified according to the MICs of both antibiotics.

The composite reference standard comprised either ESBL detection by both phenotypic methods or by a combination of one phenotypic method together with genotypic detection. Given the detection of non-ESBL genes in bla_SHV and bla_TEM amplicons, bla_CTX-M was the sole target included for genotypic ESBL detection.

The sensitivity of Vitek 2 system for detection of ESBLs as compared to the composite reference standard was 33/36 or 92% (78% – 97%) for E. coli and 40/40 or 100% (91% – 100%) for K. pneumoniae, whilst specificity was 10/10 or 100% (72% – 100%) and 9/10 or 90% (60% – 98%), respectively (see Tables 1 and 2).

Detailed analysis of three E. coli and one K. pneumoniae isolates that were misclassified by the Vitek 2 is shown in Appendix Table 1-A1: The only isolate mis-categorised as an ESBL by Vitek 2 was a K. pneumoniae isolate-resistant to ceftazidime only and lacking inhibition by clavulanic acid. The Vitek phenotype listed SHV hyper-production as an alternative resistance phenotype alongside the ESBL resistance phenotype and bla_SHV was the sole bla gene detected.

Three E. coli isolates classified by the Vitek 2 as having acquired penicilllinases were classified as ESBLs according to the composite reference standard, based on the detection of bla_CTX-M plus phenotypic detection of an ESBL by one method. The antibiotic susceptibility profile of these isolates according to Vitek 2 showed non-susceptibility to ampicillin, co-amoxiclav and cefuroxime and susceptibility to cefotaxime and ceftazidime. The three isolates were also generally susceptible to cefotaxime and ceftazidime by both phenotypic methods with some inconsistent exceptions.