Fast and Accurate Large-Scale Detection of β-Lactamase Genes Conferring Antibiotic Resistance

Abstract

Fast detection of β-lactamase (bla) genes allows improved surveillance studies and infection control measures, which can minimize the spread of antibiotic resistance. Although several molecular diagnostic methods have been developed to detect limited bla gene types, these methods have significant limitations, such as their failure to detect almost all clinically available bla genes. We developed a fast and accurate molecular method to overcome these limitations using 62 primer pairs, which were designed through elaborate optimization processes. To verify the ability of this large-scale bla detection method (_large-scaleblaFinder), assays were performed on previously reported bacterial control isolates/strains. To confirm the applicability of the _large-scaleblaFinder, the assays were performed on unreported clinical isolates. With perfect specificity and sensitivity in 189 control isolates/strains and 403 clinical isolates, the _large-scaleblaFinder detected almost all clinically available bla genes. Notably, the _large-scaleblaFinder detected 24 additional unreported bla genes in the isolates/strains that were previously studied, suggesting that previous methods detecting only limited types of bla genes can miss unexpected bla genes existing in pathogenic bacteria, and our method has the ability to detect almost all bla genes existing in a clinical isolate. The ability of _large-scaleblaFinder to detect bla genes on a large scale enables prompt application to the detection of almost all bla genes present in bacterial pathogens. The widespread use of the _large-scaleblaFinder in the future will provide an important aid for monitoring the emergence and dissemination of bla genes and minimizing the spread of resistant bacteria.

INTRODUCTION

The emergence and spread of multiple antibiotic resistance among pathogenic bacteria is a global health crisis (1). β-Lactam antibiotics are some of the most successful drugs used for the treatment of bacterial infections and represent approximately 65% of the total world market for antibiotics (2). Therefore, resistance to β-lactam antibiotics through the acquisition of genes that encode β-lactamases is one of the most serious problems in Gram-negative pathogenic bacteria, such as in members of the family Enterobacteriaceae and Pseudomonas spp. and Acinetobacter baumannii (3, 4). Since the first report observing a β-lactamase was published in 1940 (5), >1,300 distinct β-lactamase (bla) genes have been identified in clinical isolates, showing the remarkable diversity of bla genes due to their continuous mutations (4). They can be separated into the four major Ambler classes, A to D, based on their amino acid sequences. Of these enzymes, the β-lactamases that attract the largest amount of clinical concern are extended-spectrum β-lactamases (ESBLs) and carbapenemases (6–8).

For earlier detection of outbreaks and minimization of the spread of resistant bacteria, the availability of fast diagnostic methods to detect resistance genes is important. Determining susceptibility or resistance using classical culture-based phenotypic tests is the general method used in clinical microbiological laboratories, but this procedure is time-consuming and cannot easily detect ESBLs and carbapenemases produced by Enterobacteriaceae, owing to varied levels of enzyme expression and the poor specificity of some antibiotic combinations (9, 10). The implementation of molecular-based diagnostic methods easily overcomes these limitations and can increase the speed and accuracy of detecting resistance genes, which is important for infection control in hospital and community settings (9). However, previously developed methods for bla gene detection are restricted to the detection of only limited types of bla genes, and the molecular diagnostic method for bla gene types, including almost all clinically available bla genes, is not available (11–17).

In this study, we designed 62 ready-to-use PCR primer pairs, which have optimal features that are readily usable for fast and accurate detection of bla gene types, including almost all clinically available bla genes, with perfect specificity and sensitivity.

MATERIALS AND METHODS

Design of 62 primer pairs for the large-scale β-lactamase (bla) detection method.

Reference gene sequences (see Table S1 in the supplemental material) for each of 62 bla gene types were obtained from GenBank (see http://www.ncbi.nlm.nih.gov/GenBank). Based on multiple alignments of the sequences of all genes belonging to each gene type using Clustal W (http://www.genome.jp/tools/clustalw/), primer pairs were specifically designed within conserved sites to amplify all alleles of each bla gene type (_large-scaleblaFinder, with one colony multiplex PCR run per clinical isolate). The melting temperatures of the designed primer pairs were calculated using Primer3 (18) and OligoCalc (19). To avoid amplification of false-positive or -negative PCR products and to check the specificity of the designed primer pairs, a software tool called Primer-BLAST was used (http://www.ncbi.nlm.nih.gov/tools/primer-blast/) (20). Table S2 in the supplemental material details the sequences of 62 primer pairs and the sizes of the expected products. To confirm the specificity of the PCR assay, 62 primer pairs were evaluated in 62 simplex PCR assays to ensure that they correctly amplified the expected bla genes.

To check for cross-contamination, we included negative-control tubes containing 0.1% Triton X-100 with each PCR run. PCR mixtures, preparation of the template DNAs, PCR assays, analysis of the PCR products, and sequencing of the PCR products were separated by the use of dedicated rooms, laboratory coats, and instruments to minimize the risk of contamination, as previously described (21). For the chemical decontamination of surfaces and equipment, 0.5% sodium hypochlorite (bleach) was used every other day. The hypochlorite solution was removed with warm water after incubation for 5 to 10 min (21).

Template for the _large-scaleblaFinder using a single colony.

After much trial and error, the _large-scaleblaFinder for multiplex PCRs using a single colony (size, >1 mm) was established, as follows: (i) a single colony grown overnight on an agar plate was emulsified in 20 μl of 0.1% Triton X-100 using a 10-μl pipette tip and heated at 100°C for 10 min; (ii) after cell debris was removed by a centrifugation step of the cell suspension at 18,000 × g for 1 min, the supernatant (1 μl) was used as the template DNA for the multiplex PCR.

Optimization processes of multiplex PCR conditions.

To select the optimal Taq DNA polymerase, a variety of Taq DNA polymerases were used: Expand high-fidelity PCR system (Roche Diagnostics; DNA-free and high-purity enzyme without premixture), 2× PrimeSTAR premix (TaKaRa; DNA-free and high-purity enzyme), 2× EmeraldAmp GT PCR master mix (TaKaRa), 2× DiaStar multiplex PCR smart mix (SolGent; DNA-free), and 2× Solg multiplex PCR smart mix (SolGent; not DNA-free). Multiplex PCRs using the heated colony supernatant as the template DNA were performed under the various conditions recommended by each manufacturer. As a result, the 2× DiaStar multiplex PCR smart mix and 2× Solg multiplex PCR smart mix were selected as a Taq DNA polymerase for multiplex PCRs. Uracil-DNA glycosylase (UDG) was used to minimize the risk of false-positive results due to carryover contamination, as previously described (22, 23). According to manufacturer's recommendations, only the Solg multiplex PCR smart mix was used for prevention of amplicon contamination. According to our previous report (24), commercial Taq DNA polymerase can be contaminated with the TEM-type bla gene. Therefore, Solg multiplex PCR smart mix was treated with DNase I (RQ1 RNase-free DNase [1 unit]; Promega, USA), as previously described (24). Otherwise, a false-positive PCR product could be detected due to contamination of the TEM-type bla gene. After much trial and error, a single optimal annealing temperature (64°C) for multiplex PCRs was selected. Because Dallenne et al. (12) used various concentrations of primers (0.2 μM to ∼0.5 μM), a variety of primer concentrations (0.1 μM to ∼0.5 μM) were tested. Consequently, 0.2143 μM was selected as the optimal primer concentration for multiplex PCRs. In summary, the heated colony supernatant (1 μl) was added to each multiplex PCR tube (34 μl per PCR mixture of each tube [A1, A2, B1, B2, B3, C1, C2, D1 or D2]) containing DNase I-treated 1× Solg multiplex PCR smart mix and 0.2143 μM bla type-specific primers (A1, 16 primers; A2, 10 primers; B1, 16 primers; B2, 16 primers; B3, 16 primers; C1, 10 primers; C2, 10 primers; D1, 16 primers; D2, 14 primers; Table 1; see also Table S2 in the supplemental material). Nine multiplex PCR tubes were needed to test a single isolate/strain, and the total volume of each multiplex PCR mixture was 35 μl (Table 1). Amplification was performed under the following thermal cycling conditions: 3 min at 50°C to allow the UDG to break down the amplicons; initial denaturation at 95°C for 5 min; 30 cycles of 95°C for 30 s, 64°C for 40 s, and 72°C for 50 s; and a final elongation step at 72°C for 7 min. Amplicons were analyzed by electrophoresis on a 2% agarose gel at 100 V for 1 h and ethidium bromide staining. A 100-bp DNA ladder (Biosesang, South Korea) and 100-bp Plus DNA ladder (Bioneer, South Korea) were used as size markers.

TABLE 1.

Components of multiplex PCR (_large-scaleblaFinder)^a

Component	Result (μl) for multiplex tubeb:
	A1	A2	B1	B2	B3	C1	C2	D1	D2
Multiplex PCR master mixturec	18	18	18	18	18	18	18	18	18
Primer mixtured	12	7.5	12	12	12	7.5	7.5	12	10.5
RNase-free water	3	7.5	3	3	3	7.5	7.5	3	4.5
UDGe	1	1	1	1	1	1	1	1	1
Template DNAf	1	1	1	1	1	1	1	1	1

^aThe targeted genes were 1,335 bla genes, shown in Table S1 in the supplemental material.

^bThe total volume per tube was 35 μl.

^cConsisting of DNase I-treated 1× Solg multiplex PCR smart mix (0.3 mM each dNTP [dATP, dCTP, dGTP, and dUTP/dTTP {4/1}], 2.5 mM MgCl₂, and 100 U/ml Solg h-Taq; SolGent, South Korea).

^dConsisting of 0.2143 μM bla type-specific primer mixture (16 primers in A1, 10 primers in A2, 16 primers in B1, 16 primers in B2, 16 primers in B3, 10 primers in C1, 10 primers in C2, 16 primers in D1, and 14 primers in D2). Each concentrated primer (0.75 μl at a concentration of 10 μM) was added to this primer mixture; thus, the final concentration of each primer was 0.2143 μM. In the case of multiplex tube A1, 0.75 μl of each of the 16 primers resulted in a total of 12 μl.

^eUracil-DNA glycosylase (1,000 units; SolGent, South Korea).

^fTemplate DNA (supernatant shown in Fig. 1) was finally added to each multiplex PCR tube (34 μl) after mixing the multiplex PCR master mixture, primer mixture, and RNase-free water.

Elaborate optimization processes of 62 primer pairs.

Even though each primer pair was designed to be bound to target bla genes only, several false-positive bands among 108 control isolates were detected in some multiplex PCR tubes, including A1, A2, C1, C2, and D1 multiplex tubes (see Fig. S2 in the supplemental material). DNA sequence analysis showed that false-positive amplicons were PCR products of chromosomal genes from control isolates but not specific bla genes. To identify which primer was responsible for the false-positive band(s), multiplex PCR assays were performed using primer mixtures lacking one primer. In case of the false-positive band (see Fig. S2C in the supplemental material) detected in multiplex tube A2 containing 10 different primers, 10 multiplex PCR assays with nine primers were needed, and each assay mixture did not have one different primer.

Positive- or negative-control isolates/strains.

For the validation of 62 designed primer pairs and optimization of the multiplex PCR assays, 108 well-characterized bacterial isolates known to harbor bla genes were used as positive controls, and 81 bacterial isolates/strains without any targeted bla gene were used as negative controls for the assays (see Table S3 in the supplemental material).

Clinical isolates used for assay applicability.

The _large-scaleblaFinder was further assessed using 403 clinical isolates, which were isolated from a university hospital in Busan, Republic of Korea. All clinical isolates were characterized by phenotypic analysis, such as MIC determination, but they were not characterized by any molecular method to identify bla genes (see Table S4 in the supplemental material). Therefore, there was no information about any bla genes in all isolates. MICs and their interpretation (resistance) were determined by the following method. Susceptibility was determined on Mueller-Hinton agar plates (Difco Laboratories) containing serially 2-fold-diluted β-lactams. The plates were inoculated with a Steers replicator (Craft Machine), and ca. 10⁴ CFU per spot were incubated at 37°C for 18 h. The results were interpreted using the Clinical and Laboratory Standards Institute (CLSI) criteria (38).

Sequencing analysis of multiplex PCR products.

To confirm the exact bla gene type, all PCR products amplified in multiplex PCR assays were identified by the sequencing of PCR products. Amplified PCR products were purified using a PCR purification kit (Qiagen) in case of a single multiplex PCR amplicon and a gel extraction kit (Qiagen) when multiple PCR amplicons appeared. Bidirectional sequencing of purified DNAs was performed using the ABI Prism BigDye Terminator cycle sequencing kit (Applied Biosystems), according to standard procedures. Sixty-two designed primer pairs were used for the bidirectional sequencings.

Determination of specific bla genes (allelic variants).

To determine the exact bla genes after detecting bla gene types using the _large-scaleblaFinder, 87 gene-specific primer pairs (see Table S5 in the supplemental material) were designed, as previously described, in the case of 62 type-specific primer pairs (see Table S2 in the supplemental material). The heated colony supernatant (1 μl; Fig. 1) was added to each simplex PCR tube (50 μl per PCR mixture of each tube [e.g., one tube in case of GES type, nine tubes in case of IMP type]) containing 1× PrimeSTAR MAX DNA polymerase (2.5 units; TaKaRa), 200 μM (each) deoxynucleoside triphosphate (dNTP) (TaKaRa), and 0.3 μM bla gene-specific primers (GES type, one primer pair in one simplex PCR tube; IMP type, nine primer pairs in nine simplex PCR tubes; see Table S5). Amplification was performed under the following thermal cycling conditions: initial denaturation at 98°C for 5 min; 30 cycles of 98°C for 10 s, 55°C for 30 s, and 72°C for 15 s; and a final elongation step at 72°C for 7 min. Amplicons were analyzed by electrophoresis on a 2% agarose gel at 100 V for 1 h and ethidium bromide staining. A 100-bp Plus DNA ladder (Bioneer) was used as a size marker. Amplified PCR products were purified using a PCR purification kit (Qiagen). Bidirectional sequencing of purified DNAs was performed using the ABI Prism BigDye Terminator cycle sequencing kit (Applied Biosystems), according to standard procedures. Eighty-six designed primer pairs were used for the bidirectional sequencings.

FIG 1.

Large-scale bla detection method (_large-scaleblaFinder) using a single colony. A single colony from an overnight culture was suspended in a solution containing 0.1% Triton X-100, immediately followed by heating of the cell suspension at 100°C for 10 min. After cellular debris was removed through a centrifugation step at 18,000 × g for 1 min, the supernatant (1 μl, template) was subjected to a single multiplex PCR run with nine multiplex PCR tubes (A1, A2, B1, B2, B3, C1, C2, D1, and D2). The 34-μl reaction mixture of each multiplex PCR tube contained DNase I-treated 1× Solg multiplex PCR smart mix and 0.2143 μM bla type-specific primers (e.g., 16 primers in the case of multiplex PCR tube A1; see Table S2 in the supplemental material). All amplifications in nine multiplex PCR tubes were performed under the identical thermal cycling conditions: 3 min at 50°C to allow the uracil-DNA glycosylase (UDG) to break down the amplicons; initial denaturation at 95°C for 5 min; 30 cycles of 95°C for 30 s, 64°C for 40 s, and 72°C for 50 s; and a final elongation step at 72°C for 7 min. All multiplex PCR products (5 μl) were separated on a 2% agarose gel. Each bla gene type was determined by comparing each band size on agarose gels with the corresponding size shown in Fig. 2 (or the corresponding amplicon size in Table S2). In case of false-positive and/or -negative bands, optimization processes of 62 primer pairs were performed using our elaborate method (see Materials and Methods)..

RESULTS

Primer design for detecting almost all clinically available bla genes.

Our strategy for bla gene typing is the primer design using gene type-specific regions. We divided previously reported bla genes into 62 types, composed of 13 in class A, 24 in class B, 10 in class C, and 15 in class D (see Table S2 in the supplemental material). The sequences of all genes belonging to each gene type were aligned using Clustal W, and gene type-specific conserved regions were identified. Using these regions, primer pairs specific for each gene type were designed in silico, and they were compared with all members of the different primer pairs to avoid cross-hybridization. To easily define gene types of amplicons, primer pairs within each multiplex PCR tube (A1, A2, B1, B2, B3, C1, C2, D1, and D2) were designed to make different PCR product sizes (expected amplicon sizes shown in Table S2). Furthermore, a single multiplex PCR run was designed with nine reaction tubes including each primer mixture set (e.g., eight primer pairs in case of the multiplex PCR tube A1; see Table S2) to detect bla gene types, including 1,352 previously reported bla genes. Sequencing of the PCR products amplified using 87 gene-specific primer pairs can distinguish 1,352 bla genes. The number of total detectable bla gene accession numbers was 7,247, but the total number of different and detectable bla genes was 1,352, due to the assignment of multiple accession numbers to a bla gene (see Table S1 in the supplemental material). This method, to our knowledge, is the first large-scale method showing that a single multiplex PCR run with nine multiplex PCR tubes can detect and distinguish almost all clinically available bla genes.

bla detection method using a single colony.

To avoid time-consuming steps, such as genomic DNA extraction, we introduced the simple _large-scaleblaFinder using a single colony (Fig. 1). A single colony (size, >1 mm) always yielded enough genomic DNA for the _large-scaleblaFinder when the heated colony supernatants from control strains (or isolates) and clinical (test) isolates (Escherichia coli, Enterobacter cloacae, Citrobacter freundii, Providencia rettgeri, Klebsiella pneumoniae, Pseudomonas aeruginosa, Acinetobacter baumannii, Enterobacter aerogenes, and Serratia marcescens) were used as the template (see Tables S3 and S4 in the supplemental material). Through precise optimization processes of colony multiplex PCRs, a DNA Taq polymerase, an annealing temperature, and concentrations of components were selected for the most effective use of the _large-scaleblaFinder, as described in Materials and Methods. In case of a false-positive and/or -negative band(s), optimization processes of 62 primer pairs were performed using our elaborate method (see Materials and Methods). The Solg multiplex PCR smart mix was compatible with both dUTP incorporation and subsequent UDG treatment (see Fig. S3A and B in the supplemental material). Negative-control strains/isolates and a 0.1% Triton X-100 control had no PCR products (see Fig. S3A), showing that our PCR experimental conditions prevented the risk of false-positive results occurring due to carryover contamination.

Evaluation and optimization of the _large-scaleblaFinder using control isolates/strains previously reported to a harbor bla gene(s).

To verify the ability of this method, PCR assays were performed on previously reported bacterial strains/isolates that have been reported to have a bla gene(s) on a chromosome or plasmid (Fig. 2). All primer pairs were validated as simplex PCRs before being employed in a multiplex PCR. Expectedly, only one amplification product of each bla gene with the expected size was detected in 62 simplex PCRs for detecting 62 bla gene types (Fig. 2). Besides the size of the PCR product, the sequencing of 62 PCR products confirmed the exact detection of 62 bla genes (Fig. 2). Notably, although there was no optimization process, the multiplex PCR experiments using several templates (5 to 8; e.g., 5 in C1 and 8 in A1; see Table S2 in the supplemental material) obtained from 62 positive-control isolates showed that 62 bla type-specific PCR products were exactly detected in nine multiplex PCR assays (each lane 1, Fig. 2A to H), suggesting the good quality of the 62 designed primer pairs. Similarly, the exact detection of bla genes was confirmed through the sequencing of PCR products (Fig. 2). However, each multiplex PCR assay cannot discriminate whether one clinical isolate harbors only one bla gene or more than two. If one of eight isolates harbors two bla gene types (e.g., A1-1 and A1-5, Fig. 2) and another isolate harbors only one bla gene type (e.g., A1-5) in the case of a multiplex PCR tube A1 containing eight primer pairs and eight templates obtained from eight positive-control isolates, one band (A1-5) overlaps in an agarose gel of this multiplex PCR assay. To solve this problem, a single multiplex PCR run per control isolate had to be performed with nine multiplex PCR tubes (A1 to D2) and only one template obtained from a control isolate.

FIG 2.

Validation of primer pairs by simplex and multiplex PCR assays (with UDG). To confirm whether primer pairs can correctly amplify their respective loci, primer pairs were evaluated in simplex and multiplex PCR assays. The supernatant obtained from a single colony (Fig. 1) with each bla gene type was used as the template. The supernatants were obtained from 62 representative and positive-control isolates, each harboring a type-specific bla gene. Each (expected) size of the 62 type-specific bla PCR products (A1-1 to D2-7, indicated next to arrows on the right side of the panels) is shown in Table S2 in the supplemental material. The results from optimized multiplex PCR assays in 108 (including the above-mentioned 62) positive-control isolates are shown in Fig. S1 in the supplemental material. Multiplex (lane 1) and simplex (lanes other than 1) PCR products were separated on a 2% agarose gel. The sequencing of all PCR products in 62 simplex and nine multiplex PCRs confirmed the exact detection of 62 bla gene types. Lane M, molecular size marker. (A) Templates of multiplex and simplex PCRs for the detection of A1-specific bla genes, as follows: lane 1, eight templates used in lanes 2 to 9; lane 2, GES-5 from K. pneumoniae isolate CHAK36; lane 3, TEM-30 from E. coli isolate CF0022; lane 4, IMI-1 from E. coli pCCLLimiA; lane 5, KPC-3 from K. pneumoniae CL5761; lane 6, SHV-2a from E. coli isolate ECLA-4; lane 7, VEB-2 from E. coli strain TOPVEB02; lane 8, PER-2 from E. coli strain TOPPER02; lane 9, SME-1 from E. coli strain TOPSME01. (B) Templates of multiplex and simplex PCRs for the detection of A2-specific bla genes, as follows: lane 1, five templates used in lanes 2 to 6; lane 2, CTX-M-27 from E. coli strain TOPC027; lane 3, CTX-M-114 from P. rettgeri strain PS022; lane 4, CTX-M-25 from E. coli strain TOPC025; lane 5, CTX-M-43 from E. coli strain TOPC043; lane 6, CTX-M-8 from E. coli strain TOPC008. (C) Templates of multiplex and simplex PCRs for the detection of B1-specific bla genes, as follows: lane 1, eight templates used in lanes 2 to 9; lane 2, THIN-B from E. coli strain TOPTHINB; lane 3, CAU-1 from E. coli strain TOPCAU01; lane 4, SIM-1 from E. coli strain TOPSIM01; lane 5, CphA from E. coli strain TOPCPHA; lane 6, IND-6 from E. coli strain TOPIND06; lane 7, IMP-1 from E. coli strain JAEE1; lane 8, NDM-1 from E. coli strain TOPNDM01; lane 9, VIM-2 from C. freundii strain 11-7F4560. (D) Templates of multiplex and simplex PCRs for the detection of B2-specific bla genes, as follows: lane 1, eight templates used in lanes 2 to 9; lane 2, AIM-1 from E. coli strain TOPAIM01; lane 3, KHM-1 from E. coli strain TOPKHM01; lane 4, FEZ-1 from E. coli strain TOPFEZ01; lane 5, GOB-1 from E. coli strain TOPGOB01; lane 6, BlaB-1 from E. coli strain TOPBLAB01; lane 7, SPM-1 from E. coli strain TOPSPM01; lane 8, EBR-1 from E. coli strain TOPEBR01; lane 9, JOHN-1 from E. coli strain TOPJOHN01. (E) Templates of multiplex and simplex PCRs for the detection of B3-specific bla genes, as follows: lane 1, eight templates used in lanes 2 to 9; lane 2, SMB-1 from E. coli strain TOPSMB01; lane 3, DIM-1 from E. coli strain TOPDIM01; lane 4, TMB-1 from E. coli strain TOPTMB01; lane 5, GIM-1 from E. coli strain TOPGIM01; lane 6, BJP-1 from E. coli strain TOPBJP01; lane 7, CGB-1 from E. coli strain TOPCGB01; lane 8, EFM-1 from E. coli strain TOPEFM01; lane 9, POM-1 from E. coli strain TOPPOM01. (F) Templates of multiplex and simplex PCRs for the detection of C1-specific bla genes, as follows: lane 1, five templates used in lanes 2 to 6; lane 2, ACC-4 from E. coli strain TOPACC04; lane 3, DHA-1 from E. coli isolate E07-10537; lane 4, MIR-1 from E. coli isolate C600; lane 5, PDC-3 from E. coli strain TOPPDC03; lane 6, ADC-2 from E. coli strain TOPADC002. (G) Templates of multiplex and simplex PCRs for the detection of C2-specific bla genes, as follows: lane 1, five templates used in lanes 2 to 6; lane 2, CMY-3 from E. coli strain JM109; lane 3, Ear1 from E. coli strain TOPEAR01; lane 4, 520R from E. coli isolate 520R; lane 5, CMY-10 from E. aerogenes isolate K9911729; lane 6, BER from E. coli isolate BER. (H) Templates of multiplex and simplex PCRs for the detection of D1-specific bla genes, as follows: lane 1, eight templates used in lanes 2 to 9; lane 2, OXA-23 from A. baumannii isolate K0420859; lane 3, OXA-2 from P. aeruginosa isolate 08PAE4; lane 4, OXA-10 from P. aeruginosa isolate WK20; lane 5, OXA-69 from E. coli strain TOPOXA069; lane 6, OXA-31 from E. coli strain TOPOXA031; lane 7, OXA-40 from E. coli strain TOPOXA040; lane 8, OXA-63 from E. coli strain TOPOXA063; lane 9, OXA-42 from E. coli strain TOPOXA042. (I) Templates of multiplex and simplex PCRs for the detection of D2-specific bla genes, as follows: lane 1, seven templates used in lanes 2 to 8; lane 2, OXA-235 from E. coli strain TOPOXA235; lane 3, OXA-228 from E. coli strain TOPOXA228; lane 4, OXA-58 from E. coli strain TOPOXA058; lane 5, OXA-48 from E. coli strain TOPOXA048; lane 6, OXA-214 from E. coli strain TOPOXA214; lane 7, OXA-211 from E. coli strain TOPOXA211; lane 8, OXA-20 from E. coli strain TOPOXA020.

Unlike the simplex PCR assays, the multiplex PCR assay with only one template triggered two problems, the weak intensity of an amplification product of GES-type genes and some false-positive bands by chromosomal cross-hybridization. Elaborate optimization processes were needed to solve these problems. The exact in silico primer design to detect almost all clinically available bla genes did not generate false-negative bands. However, false-positive bands cannot be prevented without our elaborate optimization processes (see Fig. S2 in the supplemental material). Although the weak intensity of an amplification product was easily solved by increasing the length of GES-type-specific primer pairs (see Fig. S2A in the supplemental material), much effort was required to eliminate the amplifications of all false-positive bands. As a result, we found that false-positive PCR products were generated by chromosomal cross-hybridization between the forward (or reverse) primer of one bla gene type and the reverse (or forward) primer of another type (see Fig. S2B to E in the supplemental material). These elaborate optimization processes removed all false-positive bands identified in multiplex PCR assays using 108 control isolates (Table 2; see also Fig. S1 in the supplemental material).

TABLE 2.

Evaluation of results from the large-scare bla detection method (_large-scaleblaFinder) in 189 control isolates/strains

Control isolate/strain type	Detection result of bla genea
	Positive	Negative
Positive isolates (n = 108)	108	0
Negative isolates (or strains) (n = 81)	0	81
Total (n = 189)	108	81

^aSensitivity, 100%; specificity, 100%; positive predictive value, 100%; negative predicted value, 100%.

Unlike the positive controls, in the 81 negative-control isolates/strains without any targeted bla genes (see Table S3 in the supplemental material), no PCR product was detected in all isolates/strains tested, except (i) ampC genes already existing in the genome of an E. coli strain K-12 and its derivative strains (E. coli TOP 10, E. coli MG1655, E. coli ATCC 25922, E. coli DH5α, E. coli BL21[DE3], E. coli HB4, E. coli JF701, E. coli JF703, etc.) (25), and (ii) additional bla genes newly detected by the _large-scaleblaFinder, because previously reported primers detected only limited types of bla genes (see Table S3 and Fig. S1 in the supplemental material). Therefore, these results suggest that the _large-scaleblaFinder can detect almost all clinically available bla genes with complete specificity and sensitivity (Table 2).

Notably, multiplex PCR experiments in 12 isolates/strains detected 24 additional unreported bla genes in the isolates/strains that were previously studied (18 genes, Table 3; 6 genes, see Table S3 in the supplemental material). For example, although the report in 2004 showed that E. coli isolate K986110 had a CMY-1 type (CMY-11) bla gene, the _large-scaleblaFinder detected three additional bla gene types, such as OXA-1 type, TEM type, and BER (EC, KL, and AmpC-10) type. The sequence of each additional bla gene was identified by sequencing PCR products. Similarly, a previous report showed that K. pneumoniae isolate CL5761 isolated at Tisch Hospital, NY, was resistant to carbapenems and harbored a K. pneumoniae carbapenemase (KPC) type bla gene of class A. However, through the _large-scaleblaFinder, two additional bla genes (TEM and SHV types) were found in this KPC-3-producing isolate (Table 3). After detecting bla gene types using the _large-scaleblaFinder, the sequencing of all simplex PCR products amplified using 87 gene-specific primer pairs determined the exact bla genes (allelic variants) of the detected bla gene types (see Table S3 and Fig. S4 in the supplemental material). These results demonstrate that previous methods detecting only limited types of bla genes can miss unexpected bla genes existing in pathogenic bacteria. In summary, because molecular methods that are able to detect only a limited number of bla gene types cannot detect bla genes beyond the bounds of previous methods, it is possible for these methods to give researchers an insufficient result, as in the case of E. coli K986110. In contrast, our method can detect almost all clinically available bla genes, including unexpected bla genes undetected by previous methods. Therefore, these results suggest that our method can give researchers the exact information about bla genes existing in pathogenic bacteria.

TABLE 3.

Additional bla genes that were not found in previous studies

Isolate/strain (reference)	bla gene(s) reported by previous investigators	Additional bla gene type(s) detected by large-scaleblaFinder	Position in Fig. S1 in the supplemental material	Additional bla gene(s) determined by sequencing of simplex PCR products to distinguish allelic variants (accession no.)a
E. coli ECLA-4 (30)	SHV-2a	TEM type, BER (EC, KL, AmpC-10) type	A-18	TEM-116 (U36911), AmpC (NC_000913)
E. coli ECZP-1 (30)	SHV-12	TEM type, BER (EC, KL, AmpC-10) type	A-20	TEM-116 (U36911), AmpC (NC_000913)
K. pneumoniae CL5761 (31)	KPC-3	TEM type, SHV (LEN, OKP) type	A-21	TEM-216 (AHJ78622), SHV-5 (GU732833)
K. pneumoniae CHAK36 (32)	GES-5, SHV-12, OXA-17	TEM type	A-23	TEM-1 (KP453775)
E. coli A15R(+) (33)	CTX-M-3	TEM type, BER (EC, KL, AmpC-10) type	B-10	TEM-116 (U36911), AmpC (NC_000913)
E. coli K0519020 (34)	CTX-M-15	TEM type, OXA-1 type, BER (EC, KL, AmpC-10) type	B-12	TEM-1 (KP453775), OXA-1 (GU119958), AmpC (NC_000913)
E. coli K986110 (35)	CMY-11	TEM type, OXA-1 type, BER (EC, KL, AmpC-10) type	G-13	TEM-1 (KP453775), OXA-4 (EU380316), AmpC (NC_000913)
E. coli J53-2R(+) (36)	FOX-3	TEM type, BER (EC, KL, AmpC-10) type	G-18	TEM-116 (U36911), AmpC (NC_000913)
A. baumannii K0420859 (37)	OXA-23	OXA-51 type, ADC (AmpC-5) type	H-17	OXA-109 (CP001921), ADC-5 (AJ575184)

^aSimplex PCR products are shown in Fig. S4 in the supplemental material.

Applicability of the _large-scaleblaFinder to 403 clinical isolates.

To confirm the applicability of the _large-scaleblaFinder, multiplex PCR assays were performed on 403 clinical isolates, as determined by phenotypic analysis and not characterized by molecular methods to identify bla genes (see Table S4 in the supplemental material). As a result, all isolates exhibited resistance to one or more β-lactams and had at least a single bla gene, suggesting the surprising accuracy of this method. A single bla gene type was detected in 52 isolates, and 351 isolates had more than two bla gene types. The sequencing of all PCR products confirmed that this _large-scaleblaFinder can determine the exact bla gene typing of clinical isolates without any false-positive or false-negative results (Table 4). Therefore, our method is sufficient for application as a molecular diagnostic technique for the detection and gene typing of β-lactamases in bacterial pathogens.

TABLE 4.

Applicability of the _large-scaleblaFinder to 403 clinical isolates

Species	No. of isolates tested	No. of positive isolates detected by large-scaleblaFinder	No. of positive isolates confirmed through sequencing of PCR productsa	Detection (%)
E. coli	199	199	199	100
K. pneumoniae	144	144	144	100
A. baumannii	38	38	38	100
S. marcescens	22	22	22	100

^aAll bla genes detected in 403 clinical isolates are shown in Table S4 in the supplemental material.

Solution of the major problem in studying β-lactam resistance.

To date, several molecular methods for bla gene typing have been developed to detect the existence of bla genes in clinical isolates (Table 5). These methods can detect only limited bla genes (<539 bla genes, Table 5), such as ESBL genes (11–15). Because these methods cannot detect bla gene types, including almost all clinically available bla genes, they cannot perfectly explain the results of the culture-based phenotypic tests (11–15, 26–28). This is a big problem in studying β-lactam resistance, as β-lactam resistance can increase due to inappropriate β-lactam use. However, our _large-scaleblaFinder, which was designed to solve this problem, can detect bla gene types, including almost all clinically available 1,352 bla genes, with 100% specificity and 100% sensitivity, and the availability of this method in 403 clinical isolates was also proven (Table 5). Although perfect specificity and sensitivity were shown in four of eight previous methods, these methods had a low number of control and test (clinical) isolates (Table 5, except for clinical isolates of Ellem et al. [13]). Furthermore, unlike previous methods (12, 26, 28), our method needs one optimized multiplex PCR run per clinical isolate and thus can be a fast molecular method.

TABLE 5.

Comparison between detection methods of bla genes

Source or study	Method type(s)	No. of detectable bla genes	No. of control isolates tested	No. of clinical isolates tested	Specificity in control isolates (%)	Specificity in clinical isolates (%)
Pérez-Pérez and Hanson (11)	Multiplex PCR	30	29	22	NDa	ND
Dallenne et al. (12)	Multiplex and simplex PCRs	539	42	31	ND	ND
Poirel et al. (14)	Multiplex PCR	105	13	27	100	100
Monteiro et al. (15)	Multiplex real-time PCR	80	30	28	100	100
Ellem et al. (13)	Multiplex real-time PCR	123	41	617	100	100
Grimm et al. (28)	Microarray with simplex PCR	102	1	12	ND	ND
Leinberger et al. (26)	Microarray with dualplex and multiplex PCRs	156	14	60	100	100
Barišić et al. (27)	Padlock probes	33	33	70	ND	98.6
This study	Multiplex PCR (large-scaleblaFinder)	1,352	189	403	100	100

^aND, not determined.

DISCUSSION

The global health crisis arising from the upsurge of multiple antibiotic resistance strongly necessitates the development of a fast and accurate molecular method for detecting antibiotic resistance genes. Although several methods to detect bla genes have been reported, these methods can detect only a small number of bla genes. In this study, we present a new _large-scaleblaFinder that can recognize the existence of almost all clinically available bla genes and identify the exact type of detected bla genes. Furthermore, DNA sequence analysis of PCR products distinguished bla genes within bla gene types. This optimized multiplex PCR method is a fast and accurate technique for the screening of bla gene types in clinical isolates. This method would be suitable for microbiological laboratories of antimicrobial surveillance without sophisticated instruments, allowing improved surveillance studies and infection control measures.

The big problem in investigating β-lactam resistance is that most researchers detect bla genes in clinical isolates using primer pairs only for several interesting bla gene types. An important limitation of this tendency is that these methods cannot detect the unexpected bla genes that exist within the target isolate but do not belong to the primer-specific bla gene types. Because the exact detection of existing bla genes is important for improved surveillance studies and infection control measures, this problem can increase antibiotic resistance. Susceptibility tests using classical culture-based phenotypic tests are the routine method for determining the susceptibilities to most β-lactam antibiotics. However, this procedure can provide inaccurate information about which bla gene types exist in a clinical isolate, especially in isolates with ESBL and carbapenemase genes. A recent study demonstrated that ESBL and carbapenemase detection based on susceptibility tests causes a failure of therapy to reach levels similar to cases of success (29). Additionally, although ESBL and carbapenemase detection for epidemiological purposes is continuously advocated, some laboratories do not seek these enzymes for treatment purposes (29). Therefore, the inaccuracy and experimental difficulty of the phenotypic test can be complemented by the molecular gene typing method. Our method can accurately detect bla gene types, including almost all clinically available bla genes, using only one optimized multiplex PCR run (with nine PCR tubes) per clinical isolate. Thus, this technique can solve this major problem in detecting β-lactam resistance.

To remove false-positive and/or false-negative results in the multiplex PCR assays, we carried out the elaborate optimization processes of 62 primer pairs, such as the removal of chromosomal cross-hybridization between primers belonging to different types (see Fig. S2B and C in the supplemental material). Based on these optimization processes, new bla genes that will be reported in the future can be included in this detection method. Therefore, our method probably will become a molecular diagnostic tool to be easily upgradeable to overcome some limitations. (i) The _large-scaleblaFinder cannot differentiate SHV- and TEM-type ESBLs and non-ESBLs and GES-type carbapenemase and noncarbapenemase variants. However, after detecting bla gene types using the _large-scaleblaFinder, the sequencing of all simplex PCR products amplified using 87 gene-specific primer pairs determined the exact bla genes of the detected bla gene types. These sequencing results can differentiate ESBLs and non-ESBLs and carbapenemase and noncarbapenemase variants. Furthermore, ESBL-/carbapenemase-specific primer pairs for detecting ESBLs and carbapenemases can be designed, as previously described, in the case of 62 type-specific primer pairs. (ii) Certain bla genes (e.g., all OXY-type variants, 12 of 40 OKP-type variants, and 10 of 37 LEN-type variants) are not included in the _large-scaleblaFinder. Type-specific primer pairs for detecting these types can be also designed, as previously described, in the case of 62 type-specific primer pairs.

Our method can be successfully applied to all clinical isolates exhibiting resistance to any β-lactam antibiotic. In this study of 403 clinical isolates, a single (or more than two) bla gene type was detected in every isolate showing phenotypic resistance to β-lactams, suggesting the highly sensitive and specific features of our molecular method (see Table S4 in the supplemental material). This ability to quickly identify antibiotic-resistant pathogens will be one of major strategies for this success in reducing antibiotic resistance.

In conclusion, we developed a molecular diagnostic method that can detect almost all clinically available bla genes in various pathogenic isolates using only one unique multiplex PCR condition and sequencing of PCR products. This method enables fast and accurate detection of almost all clinically available bla genes, such as ESBL and carbapenemase genes. Thus, this method can promptly be used as an effective molecular diagnostic technique for the detection of bla genes in bacterial pathogens and will provide an important aid for monitoring the emergence and dissemination of bla genes and minimizing the spread of resistant bacteria.