βlasEN: Microdilution Panel for Identifying β-Lactamases Present in Isolates of Enterobacteriaceae

Abstract

A dried investigational use-only microdilution panel named βlasEN (a short named derived from the panel’s purpose, to identify β-lactamases in Enterobacteriaceae) containing 10 β-lactam drugs with and without β-lactamase inhibitors was developed to identify β-lactamases among clinical isolates of Escherichia coli, Klebsiella pneumoniae, Klebsiella oxytoca, Citrobacter koseri, Citrobacter freundii group, Enterobacter spp., and Serratia marcescens. The MICs obtained with a collection of 383 organisms containing well-characterized β-lactamases were used to develop numeric codes and logic pathways for computerized analysis of results. The resultant logic pathways and βlasEN panel were then used to test and identify β-lactamases among 885 isolates of Enterobacteriaceae recovered in cultures obtained at six different hospital laboratories across the United States. β-Lactamases present in 801 (90.5%) of the 885 isolates were identified by βlasEN by using the existing logic pathways and codes or after minor modifications were made to the existing codes. The 84 strains that gave codes that βlasEN could not identify were collected, reidentified, and retested by using βlasEN. Three strains had been misidentified, 54 strains gave different codes upon repeat testing that could be identified by βlasEN, and 27 strains repeated new codes. The β-lactamases in these strains were identified, and the new codes were added to the βlasEN logic pathways. These results indicate that βlasEN can identify clinically important β-lactamases among most isolates of Enterobacteriaceae. The results also show that good quality control and attention to proper performance of the tests are essential to the correct performance of βlasEN.

Over the past 10 years, there has been a significant increase in the numbers and types of β-lactamases encountered among clinical isolates of Enterobacteriaceae (4–7, 9, 14–16, 18, 23, 31, 37; H. Kurokawa et al., Letter, Lancet 354:955, 1999). Not only have the older plasmid-mediated β-lactamases such as TEM-1 and SHV-1 become more prevalent but also new derivatives of these enzymes capable of producing resistance to expanded-spectrum β-lactam antibiotics have appeared (9, 19). Plasmid derivatives of chromosomal β-lactamases have also appeared (2, 9, 14), as have enzymes capable of producing resistance to the carbapenems (9, 12, 22, 28; G. Cornaglia et al., Letter, Lancet 353:899–900, 1999). Unfortunately, many Enterobacteriaceae producing these new β-lactamases do not show frank resistance in routine susceptibility tests with certain β-lactam antibiotics despite clinical evidence that the drugs do not provide effective therapy (8, 10, 21, 29, 44). Thus, it has become imperative to design tests that will help microbiologists identify which β-lactamase(s) may be present in a clinical isolate of Enterobacteriaceae (8, 11, 12, 17, 20, 25, 33, 34, 38–40).

A series of studies have been performed to determine whether or not results of microdilution panels with or without β-lactamase inhibitors could be used to determine the presence of certain β-lactamases among species within the family Enterobacteriaceae (13, 25, 41). Results have shown that a broad variety of β-lactam drugs would be required for a β-lactamase identification panel and that accurate identification would require complex logic pathways involving multiple drugs. Therefore, a two-phase study was designed to develop a dried investigational use-only microdilution panel and associated software for the purpose of identification of β-lactamases among certain species of Enterobacteriaceae.

MATERIALS AND METHODS

Study design.

A two-phase study was designed to develop a dried microdilution panel for the identification of β-lactamases in clinical isolates of Enterobacteriaceae. In the first phase, our objectives were (i) to produce a microdilution panel containing a variety of β-lactam drugs with and without β-lactamase inhibitors, (ii) to test the panel against a collection of Enterobacteriaceae containing well-characterized β-lactamases, (iii) to determine which drug(s) were most useful in identifying β-lactamases in each species and develop a method for converting the MICs of these drugs to a numerical code, and (iv) to develop logic pathways for identifying numerical codes associated with specific β-lactamases in individual species and adapt the pathways for computerized use. The product of the first phase was the βlasEN system, which includes a microdilution panel with computerized analysis of results.

In the second phase, our objectives were (i) to have six clinical laboratories test the microdilution panel against 20 consecutive nonduplicate isolates of eight species of Enterobacteriaceae and send the MICs to a data collecting site, (ii) to run the MIC results through βlasEN software for analysis of results, (iii) to determine how well the existing logic pathways identified β-lactamases among the Enterobacteriaceae, and (iv) to modify the logic pathways to enable the βlasEN system to identify most β-lactamases encountered. The clinical test sites included Hunter Holmes Maguire VA Medical Center, Richmond, VA (site VA); the University of California at Los Angeles, Los Angeles (site CA); the Duke University Medical Center, Durham, N.C. (site NC); the Robert Wood Johnson Medical School, New Brunswick, N.J. (site NJ); St. Vincent’s Hospital and Medical Center of New York, New York, N.Y. (site NY); and the Cleveland Clinic Foundation, Cleveland, Ohio (site OH).

βlasEN system.

Antibiotic susceptibilities were determined according to the manufacturer’s recommendations by an overnight microdilution method with a dehydrated investigational panel provided by Dade MicroScan, Inc. (Sacramento, Calif.). The panels contained doubling dilutions of ceftazidime, ceftriaxone, ticarcillin, and cefpodoxime, with each drug tested alone and in combination with a fixed concentration of 2 μg of clavulanate/ml. Ceftriaxone was also tested in combination with a fixed concentration of 8 μg of sulbactam/ml. Piperacillin, meropenem, ampicillin, cefepime, cefoxitin, and cephalothin were also tested alone. MIC results were analyzed by computerized logic pathways (βlasEN) to identify types of β-lactamase production.

Strains.

Tests were performed with 383 isolates of Enterobacteriaceae with well-characterized β-lactamases (Table 1) and with 885 clinical isolates from six U.S. Medical Centers. Many of the characterized strains, as well as the methods used to identify the β-lactamases, have been described in detail previously (3, 13, 24, 27, 32, 35). The β-lactamase groups in the study were as follows: AmpC, including hyperproducers of both inducible and constitutively expressed chromosomal genes as well as plasmid-mediated derivatives; ESBL, extended-spectrum β-lactamases; OSBL, “older-spectrum” β-lactamases (e.g., TEM-1, SHV-1, OXA-1, etc.); LowA (for Escherichia coli only), low but elevated expression of chromosomal AmpC; wild type (WT), no detectable β-lactamase activity (E. coli or Klebsiella pneumoniae), low levels of β-lactamase (Klebsiella oxytoca), low levels of penicillinase (Citrobacter koseri), or low basal levels of AmpC which is inducible (Citrobacter freundii, Enterobacter cloacae, Enterobacter aerogenes, and Serratia marcescens); Hi K1, high-level production of K1 β-lactamase (K. oxytoca); and Carbapenemase, a β-lactamase capable of hydrolyzing carbapenems such as meropenem and imipenem. Quality control strains included E. coli ATCC 25922 (WT), CU-EC1 (ESBL-producing strain), and CU-EC2 (high-level chromosomal AmpC-producing strain). β-Lactamases were identified with the groups listed in Tables 1 and 2.

TABLE 1.

β-Lactamase groups identified by the βlasEN system for 383 characterized strains^a

β-Lactamase groupb	No. of strains correctly (incorrectly) identified for:
	EC	KP	KO	CK	CF	EbA	EbC	SM
AmpC	29	9			22	3	6	8
ESBL	27	40 (1)d	11	9	4	7	4	8
WT	14	1	10 (1)e	10	9	4	6	8
OSBL	25	27	3	11
Low AmpC	1
Low AmpC + OSBL	4 (2)e
Low AmpC ± OSBLf	9
Hi K1			17
WT + OSBLg					3		1	6
ESBL/mutant + porinf						4
ESBL/OSBLf							5
WT + Carbapenenase							1	2
ST (mutant)f								5
WT (mutant) + OSBLf								6
Total	111	78	42	30	38	18	23	43

^aEC, E. coli; KP, K. pneumoniae; KO, K. oxytoca; CK, C. koseri; CF, C. freundii; EbC, E. cloacae; EbA, E. aerogenes; SM, S. marcescens.

^bIncludes hyperproducers of chromosomal AmpC, plasmid-mediated derivatives of AmpC, and derepressed mutants of species producing inducible AmpC β-lactamases. WT strains of E. coli and K. pneumoniae possess no detectable β-lactamase activity; WT strains of K. oxytoca produce low levels of K1 β-lactamase; WT strains of C. koseri produce low-level penicillinase; WT strains of E. aerogenes, E. cloacae, C. freundii, and S. marcescens produce low basal levels of AmpC which are inducible. Low AmpC groups are for E. coli only; chromosomal AmpC produced at low but elevated levels from the wild type.

^cTwo porin mutants producing OSBLs.

^dOne high-level SHV-1 (OSBL) producer.

^eOne OSBL producer.

^fGroup includes two possibilities that βlasEN could not discriminate between due to similarities of MICs. If one possibility is much less likely, it is shown in parentheses. If both possibilities are equally likely they are shown with a slash.

^gUsed for those species for which the WT produces an inducible AmpC β-lactamase.

TABLE 2.

Final β-lactamase groups identified by βlasEN^a

βlasEN group	No. of strains (n = 883) positive for:
	EC	KP	KO	CK	CF	EbA	EbC	SM	Total
AmpC	8	8			32	12	35	10	105
ESBL	7	20	11	2	1		8	12	61
Low AmpCd
Low AmpC + OSBLd	4								4
Low AmpC ± OSBLabd	4								4
OSBL	39	89	4	5	10	5	5	6	163
Wild type	55	1	81	98	63	86	61	79	524
Hi K1			10						10
WT(M)bce								6	6
WT(M) + OSBLbe								3	3
OSBL/ESBLbf
OSBL/ESBL/Mbf							1		1
ESBL/M + porinbf						1			1
Total	117	118	106	105	106	104	110	116	882

^aIncludes results for 883 strains. Strains are abbreviated as defined in Table 1, footnote a.

^bGroup includes various possibilities that βlasEN cannot discriminate between. If one possibility is much less likely than the others, it is shown in parentheses. If all possibilities are equally likely, they are shown with a slash.

^cM, derepressed mutant (AmpC) of a strain normally producing an inducible β-lactamase.

^dGroup specific for E. coli.

^eGroup specific for S. marcescens.

^fGroup specific for E. aerogenes or E. cloacae.

RESULTS

Phase 1.

A total of 383 isolates of Enterobacteriaceae with well-characterized β-lactamases were used to design and produce the βlasEN system (Table 1). Selection of drugs for inclusion in the panels was based upon results of previous studies (25, 41) and those obtained in the previous study in tests with characterized strains. Among the extended-spectrum cephalosporins, ceftazidime, ceftriaxone, and cefpodoxime with or without β-lactamase inhibitors were chosen because they provided maximum discrimination between the various β-lactamases encountered. Since cefotaxime did not improve discrimination of β-lactamase groups beyond that provided by ceftazidime, ceftriaxone, and cefpodoxime, it was not included in the final part. The β-lactamase groups among the 383 characterized strains identified by βlasEN are shown in Table 1. The logic pathways utilized to determine these groups were species specific except for the two Enterobacter species where the same logic pathways could be used for both. Some of the β-lactamase groups were found among each of the species tested, while some were species specific. In a few instances, it was not possible to develop a logic pathway that could discriminate between two β-lactamase groups. Most of these were species specific. Due to similarities in MICs observed among certain groups, it was not possible to discriminate between certain strains of Enterobacteriaceae producing an ESBL that also had a porin mutation affecting carbapenem susceptibility and mutants producing high levels of AmpC that also had a similar porin mutation (denoted as ESBL/mutant + porin in Table 1). There were only four E. aerogenes in this category. It was also not possible to discriminate among the five strains of E. cloacae that possessed an ESBL (two strains) or an OSBL (three strains). Among the S. marcescens strains it was not possible to discriminate between one mutant and four WTs—the former being highly susceptible to many agents. It was also not possible to discriminate between one mutant producing an OSBL and five WTs producing an OSBL—the former again being highly susceptible to many agents. This problem with S. marcescens probably reflects the highly important role porin mutations play upon the background susceptibility of the species to a number of agents and the numerous mutations that can occur in the porins which alter susceptibility (36). It is likely that these rare S. marcescens derepressed mutants have porins that are responsible for their unusual high susceptibility to numerous agents.

To assess the robustness of the codes, tests were repeated with a number of the characterized strains to see whether the same code would be generated in each replicate. Results (not shown) indicated that codes generated by most strains were quite reproducible, whereas others varied. These variations were therefore built into the codes wherever possible. Codes for certain β-lactamase groups were so similar that replicate runs for some strains gave different β-lactamase groups. This variation was allowed to exist in the codes if it was a rare occurrence and/or if the implications of the groups were similar. For example, with a K. pneumoniae strain producing both an ESBL and a plasmid-mediated AmpC β-lactamase, a code indicating the ESBL was generated in one of two runs, while a code indicating the presence of the ESBL+AmpC was generated in one of two runs. With a strain of E. aerogenes producing an ESBL, 7 of 20 replicates gave a code indicating high levels of AmpC, while 13 of 20 replicates indicated the ESBL. For S. marcescens, one strain producing high-level AmpC+ESBL+OSBL coded as high-level AmpC in one of two replicates and as ESBL in one of two replicates, and one strain producing an ESBL coded as OSBL in one of two replicates and as WT (mutant)+OSBL in one of two replicates.

Among the characterized strains, there were several errors in identification by βlasEN (Table 1). For E. coli, two strains coded as Low AmpC+OSBL that were porin mutants plus OSBL. The porin mutation in this situation yields MICs similar to those produced by low levels of AmpC. One strain of K. pneumoniae producing high levels of SHV-1 (OSBL) generated codes indicating ESBL in three of three replicates. This error can occur in tests such as these since the high levels of the OSBL in this organism give elevated ceftazidime MICs that are reduced by clavulanate (30). One strain of K. oxytoca producing an OSBL generated codes indicating a WT in two of two replicates. This was due to an unusually high level of ticarcillin-clavulanate susceptibility in the strain. In fact, a probe was performed on the strain to confirm the continued presence of the OSBL.

At the end of phase 1, it appeared that the codes and logic pathways of βlasEN were sufficiently robust to take to a trial on clinical isolates. Alterations had been made in the codes to allow for inherent variability in codes, and errors had been minimized.

Phase 2.

A total of 885 clinical isolates were tested utilizing the βlasEN system. These included 119 isolates of E. coli, 117 isolates of K. pneumoniae, 107 isolates of K. oxytoca, 105 isolates of C. freundii, 104 isolates of E. aerogenes, 111 isolates of E. cloacae, and 115 isolates of S. marcescens. The existing logic pathways in the βlasEN system were able to identify a β-lactamase group for 711 (80.3%) isolates. The MIC codes generated by an additional 90 (10.2%) were so similar to existing codes (within the twofold error expected for MICs) that a β-lactamase group was identifiable for these strains after making only minor modifications to the logic pathways of βlasEN. Thus, for 801 (90.5%) isolates, a β-lactamase group could be identified by βlasEN.

We found that 84 (9.5%) isolates produced MIC codes that could not be identified by βlasEN. These strains (capture strains) were sent to a reference laboratory for further study and included 5 E. coli, 2 K. pneumoniae, 4 K. oxytoca, 21 C. koseri, 13 C. freundii, 8 E. aerogenes, 18 E. cloacae, and 13 S. marcescens strains. One C. koseri strain was identified as Citrobacter amalonaticus, one E. cloacae strain was identified as a Providencia stuartii, and one E. cloacae strain was identified as a C. koseri. Since the first two misidentified organisms are not in the database, these were dropped from further study. An additional 54 isolates generated different codes upon repeat testing with βlasEN, and these were readily identifiable by βlasEN. A total of 27 of the 885 isolates (3.1%) still gave new codes upon repeat testing with βlasEN. These were characterized for their β-lactamases, and the codes were added to the logic pathways for those β-lactamase groups. These included one E. coli (OSBL), two K. oxytoca (high K1), one C. koseri (OSBL), nine C. koseri (WT), one C. freundii (mutant), one C. freundii (WT+OSBL), two C. freundii (WT), four E. aerogenes (mutant), one E. cloacae (mutant), and two S. marcescens (WT) strains.

Within the 801 isolates initially identified by βlasEN, 21 strains were also collected for repeat testing (interest strains) either because they generated rare codes for β-lactamase groups commonly encountered for the species or because they generated codes for β-lactamases rarely encountered in the species. Of the 21 strains, only 4 generated the same code on repeat testing. Five were misidentified, and repeat testing generated different codes with 12, 7 of which indicated a different β-lactamase group from the original code.

When all results obtained with the clinical isolates (including repeats) were compiled, there were 882 evaluable study strains (Table 2). This reflected dropping the three strains that when reidentified, belonged to species outside of the database. The β-lactamases in the final 882 evaluable study strains included the following: high-level AmpC in 105 (11.9%) strains; ESBL in 61 (6.9%) strains; low chromosomal AmpC +OSBL in 4 isolates of E. coli; OSBL in 163 (18.5%) isolates; WT β-lactamase in 524 (59.4%) strains; and high K1 in 10 strains of K. oxytoca. In 15 (1.7%) strains, βlasEN could not definitively identify the β-lactamase present (Table 2). These included four strains of E. coli with Low AmpC β-lactamase that may or may not have also had an OSBL, six strains of S. marcescens that could have been WT or a hypersensitive mutant, three strains of S. marcescens with OSBLs that could have been WT or a hypersensitive mutant, and one strain of E. aerogenes that could have been a porin mutant and produced either high levels of AmpC or an ESBL. One strain of E. cloacae fell into a new group of OSBL/ESBL/mutant—created when a characterized capture strain was found to be a partially derepressed mutant and gave the same code as three OSBL-producing and three ESBL-producing E. cloacae strains in the characterized Creighton collection.

For C. freundii there had been a repeat code created, i.e., if the code was generated by a strain, a message told the operator to repeat the βlasEN test since the code suggested an error. This repeat code was not encountered among the 105 clinical isolates of this species tested. Because of the difficulty in discriminating among certain β-lactamase groups with some Enterobacter codes, a repeat code was created for this group as well. Four clinical isolates of E. aerogenes generated these repeat codes. All four were captured and found to be mutants. The codes generated upon repeat testing identified three to be mutants and one to be an OSBL—the only discrepancy in the capture strains.

Quality control was performed with three strains of E. coli: ATCC 25922, a WT; CU EC-1, an ESBL-producing strain; and CU EC-2, a producer of high levels of chromosomal AmpC β-lactamase. A total of 64 replicate runs with ATCC 25922 generated a WT code. Another 64 replicate runs with CU EC-1 generated an ESBL code. However, only 56 of 64 replicate runs with CU EC-2 generated AmpC codes. Among the eight out-of-control runs, seven generated codes not recognized by βlasEN and one coded for ESBL. All of these problems were due to lower-than-anticipated MICs in the ceftazidime-clavulanate and ceftriaxone-clavulanate combinations. Two of the participating laboratories accounted for six of the eight out-of-control results.

DISCUSSION

The results of this study clearly showed that it was possible to design a reliable microdilution panel with associated software for the identification of specific β-lactamases among certain species of Enterobacteriaceae. This system, the βlasEN system, was able to provide a definitive identification for β-lactamases in 867 of 882 (i.e., 98.3%) clinical isolates tested and a preliminary identification (more than one possibility) in the remaining 15 isolates. Given that members of the family Enterobacteriaceae have becoming increasingly complex in their mechanisms of resistance to β-lactam antibiotics (42), there is an increasing need for clinical laboratory methodologies that provide more sophisticated, and clinically relevant, β-lactamase identifications.

Of the 105 isolates recovered for repeat testing, 74 (70.5%) gave results that differed from the original test site. Of these isolates, 8 (7.6%) had been misidentified and 66 (62.9%) generated different codes on repeat testing that could be identified by βlasEN. It is possible that some of these differences in codes could have been due to the loss one or more resistance mechanisms during storage and transport between the initial and repeat test sites (1, 26, 43).

βlasEN is the first system involving microdilution with computer-assisted interpretation of results designed for the purpose of identification of β-lactamases among Enterobacteriaceae. It should be noted that the codes and logic pathways of βlasEN are specific for MICs obtained in this particular dried microdilution panel. They could not be used with other MIC procedures to identify β-lactamase groups.

The clinical utility of a system such as βlasEN warrants investigation. It would be of great value in hospitals with Enterobacteriaceae harboring β-lactamases that do not produce frank resistance in routine susceptibility tests. It could also generate very important information to surveillance programs both within a single hospital and across specific geographic regions. Knowledge of the presence of specific β-lactamases in addition to the β-lactam susceptibility of Enterobacteriaceae encountered in a given environment would be very valuable in directing therapy for specific patients, as well as in designing drug utilization programs for larger areas. To be maximally effective, βlasEN would need to be expanded to include additional species of Enterobacteriaceae currently not in the database and also updated on a regular basis to include new β-lactamases as they are encountered in clinical isolates. However, based on the results presented here, this approach appears to be very promising and should be pursued further.