Abstract
AmpC β-lactamases (BlaAmpC) are an emerging group of antimicrobial resistance determinants. The lack of an agreed BlaAmpC detection method hinders investigation of their epidemiology and understanding of their clinical significance. This study compared the sensitivity and specificity of phenotypic methods of BlaAmpC detection in a collection of 246 Enterobacteriaceae with a diverse range of β-lactam resistance profiles. The BlaAmpC screening methods evaluated were based on cephamycin, ceftazidime and cefepime susceptibility. These were compared with BlaAmpC screening using conventional ESBL detection methods. The confirmatory methods evaluated were biologically based assays, inhibitor-based assays, an AmpC Etest and a rapid chromogenic assay. A multiplex nucleic acid amplification test and the three-dimensional enzyme extraction assay were used as reference methods. BlaAmpC activity was present in 74 isolates. The majority of the enzymes were plasmid-encoded and belonged to the CMY, DHA and EBC families. The screening methods had sensitivities between 47 and 99 % and specificities of 45–95 %. The performance of confirmatory tests varied widely, ranging in sensitivity from 19 % to 97 % and in specificity from 88 % to 100 %. Only the Tris-EDTA and MAST ID D68C disc tests had a sensitivity and a specificity above 90 %. Further investigation is needed to establish the most suitable enzyme substrates, inhibitor types, inhibitor concentrations and interpretative cut-offs in order to refine the inhibitor-based methods. A simple disc-based protocol using cefoxitin non-susceptibility as a screening tool, followed by the Tris-EDTA method for confirmation, detects BlaAmpC activity with 95 % sensitivity and 98 % specificity.
Introduction
In spite of the recent decision by both the Clinical and Laboratory Standards Institute (CLSI) and the European Committee on Antimicrobial Susceptibility Testing (EUCAST) to lower expanded-spectrum cephalosporin breakpoints for Enterobacteriaceae, there remains a need to detect and characterize specific resistance determinants such as extended-spectrum β-lactamases (ESBLs) and AmpC β-lactamases (BlaAmpC) for epidemiological surveillance and infection control purposes (Doi & Paterson, 2007). Although less prevalent than ESBLs (Jacoby, 2009), BlaAmpC have a comparatively broader substrate range, are not inhibited by traditional β-lactamase inhibitors and bacteria producing them are more likely to become resistant to carbapenems (Philippon et al., 2002). Whereas standardized screening and confirmatory methods for ESBL identification are agreed upon (CLSI, 2010), no such methods for BlaAmpC detection exist.
Members of the family Enterobacteriaceae such as Enterobacter spp., Citrobacter freundii and Serratia marcescens possess chromosomally encoded BlaAmpC, obviating the need for specific BlaAmpC testing for these species (Thomson, 2010). Chromosomally encoded BlaAmpC in Escherichia coli is accompanied by negligible expression because of the presence of a transcriptional attenuator coupled with a weak promoter (Tan et al., 2009). Infrequently, hyperproduction of BlaAmpC in E. coli arises following mutations in these regulatory regions or acquisition of imported BlaAmpC genes (Mulvey et al., 2005). Most of the remaining clinically important Enterobacteriaceae, including Klebsiella spp., Proteus mirabilis and Salmonella spp., lack chromosomally encoded BlaAmpC but the emergence of plasmid-encoded BlaAmpC in the last few decades has altered the situation. The mobility afforded by transferable genetic elements has enabled rapid global dissemination reminiscent of the emergence of ESBLs (Kohner et al., 2009). For example, in China the prevalence of E. coli and Klebsiella pneumoniae expressing plasmid-encoded BlaAmpC increased from 2 % to 9 % between 2005 and 2006 (Ding et al., 2008).
A number of detection methods for BlaAmpC have been proposed. Potential screening tools include reduced susceptibility to cephamycins and/or ceftazidime (Moland et al., 2002), cefepime susceptibility (Doi & Paterson, 2007) and a positive ESBL screening test (Munier et al., 2010) with or without a negative confirmatory test (Doi & Paterson, 2007), or combinations of these. More specific, but more time- and labour-intensive tests may be used to confirm the presence of BlaAmpC, including biological assays [e.g. the modified cefoxitin Hodge test (Yong et al., 2002) and Tris-EDTA disc test (Black et al., 2005a)], inhibitor-based assays [e.g. using boronic acid compounds (Tan et al., 2009) or cloxacillin (Brenwald et al., 2005)], and rapid chromogenic assays (Livermore et al., 2007). Some methods have entered commercial production [e.g. the AmpC Etest (Jacoby, 2009) and MAST ID D68C disc test (Ellem et al., 2009a)], whereas others are less readily available [e.g. Syn 2190 (Doi & Paterson, 2007), Ro-48-1220 (Doi & Paterson, 2007) or LN 2-128 (Black et al., 2005b) inhibitor-based assays]. The comparative sensitivity and specificity of these methods has not been comprehensively studied.
This study aimed to compare the test performance of phenotypic methods for BlaAmpC detection in Enterobacteriaceae isolates from a large clinical microbiology laboratory.
Methods
Two hundred and forty-six non-duplicate isolates, comprising 165 K. pneumoniae, 63 E. coli, 11 P. mirabilis, 6 Salmonella spp. and one Raoultella sp. were evaluated. The identification of all isolates was confirmed by API 20E and/or Vitek 2 (bioMérieux, France). Ninety-six per cent were clinical isolates, collected between 2006 and 2009 at PathWest Laboratories, QEII Medical Centre, Western Australia. Isolates were cultured from urine (76 %), blood (8 %), wound swabs (6 %), sputum (4 %), other sterile sites (4 %) or faeces (3 %). The isolates were selected to represent a diverse range of β-lactam resistance profiles, including cefoxitin resistant (77), cefoxitin susceptible (51), ESBL screening test positive/confirmatory test negative (87), confirmed ESBL (23), metallo-β-lactamase (MBL) carbapenemase (7) and K. pneumoniae carbapenemase (KPC) production (1: K. pneumoniae BAA 1705). These isolates had been characterized using either standard CLSI (CLSI, 2010) or published molecular detection methods (Senda et al., 1996). Isolates were stored at −70 °C in brain-heart infusion broth with 15 % gycerol then subcultured twice onto Columbia agar base supplemented with 6 % horse blood prior to testing. All tests were performed on Mueller–Hinton (MH) agar plates of 4 mm depth. All discs and reagents were stored at 2–8 °C.
The phenotypic tests that were performed once each on all 246 isolates were as follows.
Cephalosporin/aztreonam susceptibility.
MICs (cefoxitin, ceftriaxone, cefotaxime, ceftazidime and aztreonam) and/or inhibitory zone sizes (cefoxitin, cefotetan, cefpodoxime) were determined by agar dilution and disc diffusion respectively, as per CLSI recommendations (CLSI, 2009). MICs could not be obtained for P. mirabilis isolates due to the effects of swarming. Susceptibility results were interpreted according to current CLSI criteria (CLSI, 2010).
ESBL screening and confirmatory testing.
As per CLSI criteria (CLSI, 2010), isolates with either MIC of ≥2 µg ml−1 (ceftazidime, cefotaxime, ceftriaxone, or aztreonam) or cefpodoxime inhibitory zone diameter ≤17 mm (or ≤22 mm for P. mirabilis) were considered to have a positive ESBL screen. Confirmatory testing was performed using the CLSI disc potentiation method (CLSI, 2010).
Carbapenemase testing.
Isolates were screened for carbapenemase production by disc diffusion using a 10 µg meropenem disc. In accordance with CLSI recommendations, isolates with an inhibitory zone ≤21 mm had carbapenemase production confirmed using the modified (meropenem) Hodge test (CLSI, 2010).
Modified (cefoxitin) Hodge test.
This was performed as described by Yong et al. (2002).
Tris-EDTA test.
This test, also known as the ‘AmpC disc test’, was performed as described by Black et al. (2005a) (Fig. 1a).
(a) A positive result obtained from BlaAmpC confirmatory testing using the Tris-EDTA test. This shows a zone of inhibition of E. coli ATCC 25922 surrounding a cefoxitin disc, with indentation adjacent to a Tris-EDTA-impregnated disc smeared with several colonies of the test isolate. (b) A positive result obtained from BlaAmpC confirmatory testing using the MAST ID D68C disc test. This shows a difference in inhibitory zone size of ≥5 mm between discs containing cefpodoxime with (C, D) and without (A, B) an unidentified BlaAmpCL inhibitor.
Saline disc test.
This was performed as described by Singhal et al. (2005).
Inhibitor-based tests.
Commercially available 6 mm antibiotic discs were supplemented with either boronic acid compounds or cloxacillin, both known to inhibit BlaAmpC activity. A 0.5 McFarland suspension of the test isolate was inoculated evenly on MH agar, and both unsupplemented and supplemented discs were applied. After incubation for 16–18 h in room air at 35 °C the increase in zone size around the supplemented disc compared to the unsupplemented disc was recorded. The antibiotic substrate, inhibitor type, inhibitor concentration and proposed interpretative cut-offs vary among published methods. For the purposes of this study, all zone sizes were reviewed to establish alternative cut-offs that maximized both sensitivity and specificity for BlaAmpC detection.
The combinations of substrate and inhibitor used were phenylboronic acid (400 µg) with either cefoxitin (30 µg) (Tan et al., 2009) or cefotetan (30 µg) (Coudron, 2005), phenylboronic acid (300 µg) with either cefotaxime (30 µg) or ceftazidime (30 µg) (Yagi et al., 2005), aminophenylboronic acid hemisulfate (APBA) (300 µg) with either cefotetan (30 µg) or ceftazidime (30 µg) (Tenover et al., 2009), 2-benzothienylboronic acid (64 µg) with cefpodoxime (10 µg) both with and without 1 µg of clavulanic acid (Brenwald et al., 2005), cloxacillin (100 µg) with cefoxitin (30 µg) (Brenwald et al., 2005), and cloxacillin (200 µg) with cefoxitin (30 µg) (Tan et al., 2009). Blank discs with inhibitor alone were used to detect intrinsic antimicrobial activity of the inhibitor.
MAST ID D68C disc test.
A 0.5 McFarland suspension of the test isolate was inoculated evenly on MH agar, then one each of four discs (A–D) were placed onto the agar in accordance with the manufacturer’s instructions (Mast Group). Disc A contained cefpodoxime (10 µg), disc B contained cefpodoxime (10 µg) and an unidentified ESBL inhibitor, disc C contained cefpodoxime (10 µg) and an unidentified AmpC inhibitor, and disc D contained cefpodoxime (10 µg) and both the AmpC and ESBL inhibitors. Incubation was for 18–24 h in room air at 35 °C. A zone difference of ≥5 mm between disc C and disc A, or between disc D and disc B, was taken to indicate the isolated presence of a BlaAmpC (Fig. 1b), whereas a zone difference of ≥5 mm between disc D and disc C, but <5 mm difference between disc B and disc A, was taken to indicate the presence of both a BlaAmpC and an ESBL.
AmpC Etest.
Double-sided Etest strips (BioMérieux) containing cefotetan (0.5–32 µg ml−1) with and without cloxacillin were applied to MH agar that had been inoculated evenly with a 0.5 McFarland suspension of the test isolate in accordance with the manufacturer’s instructions. After overnight incubation, an eightfold reduction in cefotetan MIC in the presence of cloxacillin, the presence of a ‘phantom’ zone of inhibition or deformation of the inhibition ellipse was interpreted as presence of a BlaAmpC. The result for isolates with a cefotetan MIC >32 µg ml−1 both with and without cloxacillin were considered indeterminate.
Cica-Beta chromogenic test.
This test, which utilizes a chromogenic oxyiminocephalosporin substrate (HMRZ-86), was performed as described by Livermore et al. (2007).
Three-dimensional enzyme extraction assay.
This assay was performed as described by Coudron et al. (2000). In view of the fact that carbapenemases may hydrolyse cefoxitin, the presence of enhanced growth of organisms at the point on the agar surface where the slit intersected the zone of inhibition was only taken as evidence of BlaAmpC when in combination with negative carbapenemase testing.
PCR.
Detection of plasmid-borne genes encoding BlaAmpC was performed based upon the method described by Pérez-Pérez & Hanson (2002). This multiplex nucleic acid amplification assay utilizes PCR primers specific for blaDHA, blaACC, blaMOX, blaCIT, blaFOX and blaEBC and is capable of detecting six phylogenetic families of plasmid-mediated BlaAmpC on the basis of amplicon size. All positive PCR results were confirmed by monoplex PCR using individual primer sets with a primer concentration of 0.2 µM. The identity of PCR products was established by DNA sequencing performed using the Big Dye terminator system and a 3130XL instrument (Applied Biosystems), followed by a GenBank blastn search.
Performance and interpretation of the phenotypic testing was performed without reference to the results of PCR assays. E. coli ATCC 25922 (wild-type) was used as a negative control for all tests. Genotypically well-characterized BlaAmpC-expressing K. pneumoniae served as positive controls for phenotypic testing. For PCR testing, genotypically confirmed representatives of each of six phylogenetic BlaAmpC families acted as positive controls. For this purpose, an E. coli expressing FOX β-lactamase (J53 pMG252) was kindly provided by George Jacoby (Lahey Clinic, Burlington, MA, USA). John Iredell (ICPMR, Westmead Hospital, Sydney, Australia) kindly provided K. pneumoniae expressing CMY-2 (Jie091) and DHA β-lactamases (Jie142), an Enterobacter sp. expressing the EBC β-lactamase (W27Agr06), an Aeromonas sp. expressing the MOX β-lactamase (Ahc09) and a Hafnia alvei expressing the ACC β-lactamase (Hav09).
For the purpose of analysis, for all isolates other than E. coli, the reference method was detection of plasmid-encoded BlaAmpC using the PCR assay. For E. coli, two reference methods were used for BlaAmpC detection: a positive PCR assay was interpreted as evidence of a plasmid-encoded BlaAmpC, whereas the combination of a negative PCR assay, a positive three-dimensional enzyme extraction assay and a negative carbapenemase test was taken to represent hyperproduction of chromosomally encoded BlaAmpC. The three-dimensional enzyme extraction assay selected for this purpose had previously been shown to have 100 % sensitivity and specificity for BlaAmpC detection (Yong et al., 2002).
Fisher’s exact test was used to compare proportions using GraphPad Instat 3.06.
Results
Seventy-four (30 %) isolates demonstrated evidence of a BlaAmpC using the reference method, including 47 E. coli, 23 K. pneumoniae, 3 P. mirabilis and 1 Salmonella sp. (Table 1). Of the 47 E. coli, 22 (47 %) had hyperproduction of chromosomal BlaAmpC and 25 (53 %) had plasmid-encoded BlaAmpC. Of the study isolates with plasmid-encoded BlaAmpC, the proportion with CMY (77 %), DHA (20 %) and EBC (4 %) enzymes was in keeping with the known epidemiology of plasmid-encoded BlaAmpC in Australia (Ellem et al., 2009b). A previously recognized (Tan et al., 2009) association between certain bacterial species and BlaAmpC type was observed (i.e. K. pneumoniae and DHA-1, E. coli and CMY-2).
The test performance of various proposed screening and confirmatory tests for BlaAmpC detection is shown in Tables 2 and 3, respectively.
Comparison of the methods for detecting different BlaAmpC families revealed that the inhibitor-based methods utilizing cloxacillin and cefoxitin were significantly less sensitive for detection of DHA enzymes compared to CMY enzymes (0 % vs 58 % for 100 µg cloxacillin, P = 0.0036; 0 % vs 78 % for 200 µg of cloxacillin, P<0.0001).
Fig. 2(a, b) are histograms demonstrating the inhibitory zone sizes for cefoxitin and cefotetan as determined by disc diffusion, separated according to the presence or absence of BlaAmpC. Using CLSI breakpoints, non-susceptibility to cefoxitin rather than cefotetan was a more sensitive (97 % vs 47 %, P<0.0001) but less specific (64 % vs 95 %, P<0.0001) means of screening for the presence of BlaAmpC.
(a) Cefoxitin inhibitory zone sizes for BlaAmpC-containing (dark grey bars) and non-BlaAmpC-containing (light grey bars) isolates. The CLSI breakpoint for non-susceptibility is <18 mm (dotted line) (CLSI, 2010). (b) Cefotetan inhibitory zone sizes for BlaAmpC-containing (dark grey bars) and non-BlaAmpC-containing (white bars) isolates. The CLSI breakpoint for non-susceptibility is <16 mm (dotted line) (CLSI, 2010).
Five isolates co-expressed both BlaAmpC and ESBL activity. Compared to isolates that expressed only BlaAmpC, the performance of detection methods for these five isolates did not differ significantly, with the exception of the inhibitor-based method utilizing cefpodoxime and 2-benzothienylboronic acid, which was less sensitive for detection of co-producers of BlaAmpC and ESBL (0 % vs 69 %, P = 0.0159).
AmpC Etest results for 10 (4 %) of the isolates were indeterminate. Of these, six expressed carbapenemases, one an ESBL and one a BlaAmpC, whilst the remainder had no identifiable β-lactamases.
When tested in isolation, none of the BlaAmpC inhibitors displayed intrinsic antimicrobial activity against the study isolates.
Discussion
A lack of consensus regarding optimal methods for BlaAmpC detection is a major barrier to defining their epidemiology and understanding their clinical significance (Sidjabat et al., 2009). A mechanism-based approach to BlaAmpC detection may reduce erroneous cephalosporin susceptibility reporting (Thomson, 2010), and enable laboratories to explain a significant proportion of false-positive ESBL screening and false-negative ESBL confirmatory tests (Robberts et al., 2009). In the face of reducing development of new antibiotics, the early identification of, and application of basic infection control principles to, patients harbouring bacteria with broad-spectrum resistance mechanisms becomes paramount (Thomson, 2010). For these reasons, the development of standardized guidelines for detection of broad-spectrum β-lactamases was recently identified as a priority for both national and international laboratory method committees (Giske et al., 2009).
The large number of isolates tested, the breadth of BlaAmpC detection methods assessed and the inclusion of isolates possessing ESBL and carbapenemases enabled a comprehensive comparison of test performance in our study. Akin to ESBL detection, the low prevalence of BlaAmpC favours a two-step process in which a highly sensitive screening test is followed by a more specific confirmatory test (Doi & Paterson, 2007). Interpretative cut-off points used in this study were selected to support a two-step evaluative test process.
Cephamycin hydrolysis distinguishes BlaAmpC from many other β-lactamases. We found that non-susceptibility to cefoxitin, rather than cefotetan, was a more useful screening tool for BlaAmpC detection, with a sensitivity of ≥97 % using either an MIC or disc-based method. The use of a cefoxitin inhibitory zone size less than the CLSI breakpoint (or a higher MIC) improved specificity with marginal loss of sensitivity (Fig. 2a, Table 2), as has been shown elsewhere (Coudron et al., 2000). Other mechanisms of cephamycin resistance (e.g. porin deficiency, carbapenemases) and the fact that the cefoxitin CLSI breakpoint is only twice the modal wild-type cefoxitin MIC (Livermore, 1995) weakens the specificity of a screening approach using cefoxitin non-susceptibility. A positive ESBL screen is an adequate alternative (sensitivity 96 %) that is familiar to most laboratories. The use of combined screening tests improved specificity, but with variable impact on sensitivity. The optimal screening test combination was a positive ESBL screen plus cefepime susceptibility, which had a sensitivity of 96 % and specificity of 91 %.
The performance of confirmatory tests for BlaAmpC detection varied widely, ranging in sensitivity from 19 % to 97 % and in specificity from 88 % to 100 %. Only two methods, the Tris-EDTA and MAST ID D68C disc test, had both sensitivity and specificity >90 %. The Tris-EDTA test performed better than the other biological assays, presumably because Tris-EDTA improves the release of β-lactamases by permeabilizing Gram-negative cells (Thomson, 2010). Another property of Tris-EDTA, the inhibition of MBL carbapenemase activity (Thomson, 2010), improved the specificity of this assay by preventing cefoxitin hydrolysis by this enzyme, as was demonstrated in six of the seven MBL-expressing isolates. The high sensitivity and specificity of this method has been reported in other studies (Black et al., 2005a). Whilst the commercial MAST ID D68C disc test performed as well as the in-house Tris-EDTA method, lower acquisition costs may make the latter more cost effective. The AmpC Etest was easy to use, but was unsuitable for testing for BlaAmpC in 4 % of isolates, although the majority of those did not possess an BlaAmpC. Test performance could be improved by developing an Etest strip with a broader MIC range. The poor performance of the chromogenic Cica-Beta test is consistent with findings of other studies (Montgomery et al., 2008).
Inhibitor-based assays are the most widely published BlaAmpC confirmatory method. We found the sensitivity and specificity of all the inhibitor-based methods to be lower than previously described (Tan et al., 2008). In contrast to other studies suggesting that cefotaxime was the optimal substrate (Hope et al., 2008), we found ceftazidime to be superior, perhaps explained by preferential hydrolysis of this cephalosporin by BlaAmpC (Philippon et al., 2002). When used at lower concentrations, cloxacillin appeared to have inferior sensitivity compared to the boronic acid compounds. However, comparisons must be made with caution as the concentration of each inhibitor that optimizes test performance remains to be defined. When used with the optimal cephalosporin substrate, phenylboronic acid and ABPA were superior to 2-benzothienylboronic acid, but still lacked sensitivity. Notably, all the boronic-acid-based assays produced false-positive results when tested on the KPC carbapenemase-producing isolate. Boronic acid is a known inhibitor of this enzyme (Thomson, 2010). Establishing the optimal inhibitor type, inhibitor concentration, cephalosporin substrate(s) and interpretative cut-offs for this group of tests requires further research.
Variation in test performance between BlaAmpC families is known to occur. For example, due to differing substrate specificity, cephamycin non-susceptibility is a less sensitive screening tool for the ACC family of enzymes (Tan et al., 2009). We observed differing sensitivity of cloxacillin inhibitor-based assays for detection of CMY and DHA enzymes, which may be explained by variation in the inhibitory capacity of cloxacillin between BlaAmpC families (Philippon et al., 2002). These unique findings warrant further investigation on a broader range of bacterial species and BlaAmpC families.
Our study has several limitations. We included relatively few isolates with co-production of both BlaAmpC and ESBL. This infrequently described combination is known to weaken the performance of inhibitor-based methods of BlaAmpC detection (Tenover et al., 2009), as was observed for the cefpodoxime and 2-benzothienylboronic acid combination in our study. To address this, further investigation of methods simultaneously utilizing clavulanic acid and BlaAmpC inhibitors is warranted. Secondly, the geographical distribution of plasmid-encoded BlaAmpC families varies widely (Alvarez et al., 2004) and as we demonstrated, the performance of phenotypic tests may differ between these families, potentially restricting the applicability of our findings to other settings. However, in our study the majority of plasmid-encoded BlaAmpC were of the CMY type, the most prevalent type in all other regions of the world (Jacoby, 2009). Finally, it is possible that other BlaAmpC families, not detected by the multiplex PCR assay used, were present in our study. A novel plasmid-encoded BlaAmpC, CFE-1, has recently been described, although its prevalence remains to be defined (Nakano et al., 2004).
In conclusion, a number of BlaAmpC screening options exist. These have excellent sensitivity but the performance of confirmatory tests differs widely. Further investigation of optimal substrates, inhibitory type, inhibitor concentration and interpretative cut-offs is needed in order to improve the performance of the inhibitor-based assays. A simple disc-based protocol utilizing cefoxitin non-susceptibility as a screening method, then using the Tris-EDTA method for confirmation, could conveniently be incorporated into laboratory workflow and would detect BlaAmpC with a sensitivity of 95 % and a specificity of 98 %.
Acknowledgements
This work was supported in part by a PathWest QEII Medical and Scientific Research Grant. We would like to thank Sue-Ellen Blackaby (Blackaby Diagnostics, Australia), who provided the Cica-Beta chromogenic test kits, and bioMérieux (France) for supplying the AmpC Etests. We appreciate the assistance of our colleagues in the Department of Microbiology, PathWest Laboratory Medicine, WA, with collection of clinical isolates. We thank George Jacoby (Lahey Clinic, Burlington, MA) for kindly reviewing the manuscript prior to submission.