Characterization of Multidrug Resistant ESBL-Producing Escherichia coli Isolates from Hospitals in Malaysia

The emergence of Escherichia coli that produce extended spectrum β-lactamases (ESBLs) and are multidrug resistant (MDR) poses antibiotic management problems. Forty-seven E. coli isolates from various public hospitals in Malaysia were studied. All isolates were sensitive to imipenem whereas 36 were MDR (resistant to 2 or more classes of antibiotics). PCR detection using gene-specific primers showed that 87.5% of the ESBL-producing E. coli harbored the blaTEM gene. Other ESBL-encoding genes detected were blaOXA, blaSHV, and blaCTX-M. Integron-encoded integrases were detected in 55.3% of isolates, with class 1 integron-encoded intI1 integrase being the majority. Amplification and sequence analysis of the 5′CS region of the integrons showed known antibiotic resistance-encoding gene cassettes of various sizes that were inserted within the respective integrons. Conjugation and transformation experiments indicated that some of the antibiotic resistance genes were likely plasmid-encoded and transmissible. All 47 isolates were subtyped by PFGE and PCR-based fingerprinting using random amplified polymorphic DNA (RAPD), repetitive extragenic palindromes (REPs), and enterobacterial repetitive intergenic consensus (ERIC). These isolates were very diverse and heterogeneous. PFGE, ERIC, and REP-PCR methods were more discriminative than RAPD in subtyping the E. coli isolates.


2
Journal of Biomedicine and Biotechnology However, the genotypic characterization of other resistant isolates has not been reported so far. The objectives of this study were to determine the antimicrobial resistance and ESBL profiles of E. coli isolated from 5 public Malaysian hospitals and to determine their genetic diversity using PCRbased fingerprinting techniques and PFGE. The presence of resistance genes and integrons was also determined via PCR, and their transferability was determined by conjugation and transformation.

Genotyping by RAPD, REP, and ERIC.
Crude DNA from the E. coli obtained by direct cell lysate was used for random amplified polymorphic DNA (RAPD) analysis using primers OPAB04 [7] and OPB17 [8] with cycling conditions as previously described (Table 1).
Each PCR reaction was carried out in a 25 μL volume using 1.5 U of Taq DNA polymerase (Promega, Madison, Wis, USA) in the reaction buffer provided by the manufacturer containing 2.5 mM MgCl 2 , 50 μM of each deoxynucleoside triphosphate, 0.3 μM of the selected primer and 5 μL of DNA template. Aliquots (10 μL) of each PCR product were subjected to electrophoresis on 1.5% agarose gel.
2.3. Genotyping by PFGE. PFGE was performed according to previously described protocols [20] with minor variations. Equal volumes of 1% Seakem Gold agarose (Cambrex Bio Science, Rockland, USA) and standardized cell suspension (OD 610 = 1.4; approximately 1 × 10 8 cfu/mL) were mixed to form agarose plugs, and the bacteria lysed within the plugs with cell lysis buffer (50 mM Tris; 50 mM EDTA (pH 8.0), 1% Sacrosine, 1 mg/mL proteinase K) and incubated at 54 • C for 3 hours. Plugs were then washed with sterile deionised water (twice) and TE buffer (4 times), then digested with 12 U of XbaI (Promega, Madison, Wis, USA), and incubated overnight at 37 • C. The XbaI-digested DNA was separated on a CHEF-DRIII (BioRad, Hercules, CA, USA) with pulse times of 2.2-54.2 seconds at 200 V for 24 hours, and gels were photographed under UV light after staining with 0.5 μg/mL ethidium bromide.

Fingerprint Pattern
Analysis. The banding patterns generated by RAPD, ERIC-PCR, REP-PCR, and PFGE were analyzed using GelCompar II, version 2.5 (Applied Maths, Kortrijk, Belgium). PCR fingerprints and PFGE profiles were assigned arbitrary designation and analyzed by defining a similarity (Dice) coefficient F [21]. Cluster analysis based on the unweighted pair group method with arithmetic averages (UPGMA) with a position tolerance of 0.15 was done using the GelCompar II software.  [22]. The double-disk synergy test was performed according to established protocols and results interpreted as described previously [23].
Two ESBL-producing isolates (EC19 and EC31) were further tested by using primers specific for bla ACT , bla GES , bla VIM , bla PER , and bla FOX genes using conditions as listed in Table 1.
Clinical isolates of E. coli were used as the positive controls for bla TEM , bla SHV , bla CTX-M , and bla OXA genes. No positive controls were available for detection of bla VEB ,  Journal of Biomedicine and Biotechnology

Detection of Class 1, 2, and 3
Integrons. Class 1, 2, and 3 integrons were detected by PCR using established primers and conditions as listed in Table 1. Selected amplified products were sequenced to corroborate their identities.
Size determination of the plasmids from transconjugants and transformants was carried out by digestion with EcoRI or SphI (Promega, Madison, Wis, USA), and the products separated in 0.8% agarose gels at 70 V for 4 hours.

PCR-Based
Fingerprinting. Three PCR-based DNA fingerprinting methods were used to subtype the 47 E. coli isolates. ERIC-PCR analysis differentiated the 47 isolates into 45 unique profiles (F = 0.54-1.0) whereas RAPD using the OPAB04 and OPB17 primers generated 44 and 43 profiles, respectively, (F = 0.41-1.0 for the OPAB04 primer and F = 0.36-1.0 for the OPB17 primer, see Figures 1(a) and 1(b)). REP-PCR differentiated the 47 isolates into 45 distinct profiles (F = 0.53-1.0, see Figure 1(c)). All three PCR-based methods were reproducible as identical profiles were obtained in separate experiments using the same set of isolates.
Two isolates, EC14 and EC34, from the same hospital but from different wards, yielded identical profiles by all the 3 methods. Two other blood isolates, EC12 and EC24, from 2 different patients in the same ward, were also indistinguishable by their ERIC, REP, and RAPD profiles. ESBL-producing isolates EC4, EC9, and EC20 were clonally related by both RAPD and REP-PCR. Isolates EC4 and EC9 were indistinguishable by RAPD using the OPAB04 primer whereas isolates EC4 and EC20 were indistinguishable by RAPD using the OPB17 primer. However, isolates EC4 and EC9 were in the same cluster (92% similarity) based on their ERIC-PCR profiles.

PFGE with XbaI-Digested Genomic DNA. XbaI-digested
genomic DNA of the 47 E. coli isolates resulted in 44 distinct pulsed-field profiles (PFPs) comprising 12-26 restriction fragments. The 2 E. coli strains that had identical ERIC-PCR, REP-PCR and RAPD profiles (i.e., EC12 and EC24) were similarly indistinguishable by their PFPs with both sharing all 14 restriction fragments.
Two other isolates, EC1 and EC4 which were indistinguishable by PFGE but were distinguishable in their ERIC, RAPD and REP-PCR profiles, had 39%-68% similarities. Similarly, ERIC, RAPD, and REP-PCR differentiated isolates EC37 and EC39 that displayed identical PFPs. Both EC37 and EC39 were isolated from the same hospital. On the other hand, isolates EC14 and EC34 that displayed identical ERIC, RAPD, and REP-PCR profiles could be differentiated by PFGE (F = 0.81).

Combined Analysis.
A dendrogram based on the combined fingerprints generated by ERIC-PCR, RAPD, REP-PCR, and PFGE was constructed ( Figure 2). All the 47 isolates were differentiated into 46 combined subtypes ( Table 2). Two isolates, EC12 and EC24, with the combined profile E12R12A11B12X11 were identical in their ERIC, RAPD, REP, and PFGE profiles. Three other isolates (EC4, EC9, and EC20) were grouped within the same cluster and were clonally related (more than 85% similarity). Isolates EC14 and EC34 were also grouped together within the same cluster and were clonally related (more than 94% similarity).  Thirty-six (76.5%) isolates were presumptive ESBL producers based on initial screening. Using the doubledisk synergy test, only 3 ESBL producing isolates were detected. However, based on the phenotypic confirmatory test, 10 isolates were found to be ESBL producers. All isolates that tested positive for ESBL were also multidrugresistant.

Detection of Genes Encoding ESBLs.
Established primers were used on the genomic and plasmid DNA of the 47 E. coli isolates for the following ESBL-encoding genes: bla TEM , bla SHV , bla OXA , bla CTX-M , bla DHA , and bla VEB . The bla TEM-1 gene was detected in 35 (74.5%) whereas bla SHV -, bla CTX-M -, and bla OXA -specific amplicons were detected in only 3, 8, and 2 isolates, respectively. Of the 35 bla TEM isolates, 7 also harbored bla CTX-M and 1 had bla SHV . Three ESBL genes were detected in E. coli isolate EC7: bla SHV , bla OXA , and bla CTX-M . bla DHA , and bla VEB were not detected in any of the isolates. All bla TEM , bla SHV , and bla OXA genes were carried on plasmids whereas 6 of the 8 bla CTX-M genes were plasmid-borne. No    1, 2, and 3 Integrons. Forty-seven E. coli isolates were screened for the presence of integrases encoded on class 1, 2, and 3 integrons. The class 1 integron-encoded intI1 integrase gene was detected in 25 isolates while 4 isolates tested positive for class 2-encoded intI2 integrase. One isolate, EC24, was found to harbor both intI1 and intI2.
No class 3 integron was detected. Majority of the integrons were found to be plasmid-encoded (16 of the 25 intI1 and 3 of the 4 intI2 detected).
Isolates that were positive for class 1 and 2 integrons were further analyzed for the presence of inserted gene cassettes within the variable region by using the primer pair 5 CS/3 CS for class 1 integrons and primer pair orfx/attI2 for class 2 integrons. Amplified products of different sizes were obtained from 17 of the 25 intI1-positive isolates and sequencing indicated the presence of 5 different types of known gene cassettes: aadA5-dfrA17, dfrA7, aadA1-aadB-cmlA6, dfrA12-aadA2-orfF and aadA1. Using the attI2/orfX primer pair for intI2-positive isolates resulted in a 2 kb amplified product which, when sequenced, contained the dfrA1-aadA1-sat2 gene cassette. The aadA2, aadA5, and aadB genes encode resistance to aminoglycosides whereas sat2 encode resistance to strepthoricin. Both dfrA12 and dfrA17 encode resistance to trimethoprim and cmlA6 to chloramphenicol. Although trimethoprim and strepthoricin were not used in our study, the presence of gene cassettes encoding resistance to aminoglycosides and chloramphenicol coincided with the resistance profiles of the respective isolates. The majority of the integron-positive isolates (24 of 27) were multidrug-resistant.

Transfer of Resistance Determinants.
Conjugation experiments were carried out for 7 randomly-selected ESBLproducers and transfer of this phenotype to the recipient nalidixic acid-resistant E. coli JM109 was successful in only 3 of the 7 isolates (38%). However, it should be noted that only broth matings were carried out in this study and not filter matings and thus the transmissibility potential for the other 4 isolates could not be fully ascertained. All transconjugants were resistant to ampicillin and piperacillin except for the EC18 which remained susceptible to piperacillin. Streptomycin resistance was cotransferred in the EC46 transconjugant whereas for the EC31 transconjugant, trimethoprimsulfomethaxazole and cefoperazone resistances were also transferred ( Table 2).
Identical EcoRI and SphI restriction profiles were obtained from plasmids extracted from the donor E. coli and their respective transconjugants except for EC31. In this case, based on the restriction profiles obtained, the plasmid extracted from the transconjugant was smaller than the plasmid obtained from the donor EC31 strain (approximately 40 kb from the transconjugant as compared to ∼90 kb from the donor) although they shared a number of common restriction bands (Table 2). Plasmids extracted from the transconjugants were used to transform electrocompetent E. coli DH10B cells. Plasmids extracted from the resulting ampicillin resistant DH10B transformants showed identical EcoRI restriction profiles with those from their respective transconjugants. The DH10B transformants also displayed identical antibiotic resistance profiles as their respective transconjugants, strongly implying that these antibiotic resistance determinants were plasmid-borne.
Both bla TEM and intI1 were detected on the plasmids extracted from the EC18 and EC46 donor, transconjugants as well as their subsequent transformants indicating that these 2 genes were likely present on the plasmid that was transferred. None of the common ESBL-encoding genes were detected in EC31 although this isolate harbored class 1 integronencoded intI that was detected from the plasmid in the EC31 transconjugant and transformant suggesting that the transferred plasmid harbored a class 1 integron.
Transformation was carried out for the 4 isolates in which conjugation was not successful and another ESBLpositive randomly chose isolate. Plasmids were extracted from these 5 isolates and electroporated into recipient E. coli DH10B cells. All transformants obtained were resistant to ampicillin, piperacillin, tetracycline, and trimethoprimsulfamethoxazole. Plasmids extracted from the transformants showed identical EcoRI restriction profiles when compared to their respective donor plasmids except for isolate EC7 in which the plasmid isolated from the transformant was smaller (∼55 kb) when compared to the hospital isolate (∼135 kb) ( Table 2). Both bla TEM and intI1 genes were also detected from the plasmids isolated from the donor isolates as well as the transformants except for isolate EC7 and its transformant.

Discussion
Genotyping by PCR-based methods and PFGE showed the 47 E. coli clinical isolates to be genetically diverse and heterogeneous. This is expected as the isolates were randomly selected from different hospitals and sources. Similar observations were reported by Mugnaioli et al. [24] and Woodford et al. [16]. We found that ERIC and REP-PCR yielded practically identical results and able to differentiate strains which were indistinguishable by RAPD. Although PCR-based fingerprinting is rapid, it is more susceptible to technical variations than PFGE especially reproducibility. Thus PFGE is considered the better method for subtyping of E. coli, even though it is relatively laborious and timeconsuming compared to PCR-based methods.
Isolates EC12 and EC24 were indistinguishable by both PFGE and PCR fingerprints. These two blood isolates were from different patients in the same ward, strongly suggesting a possible nosocomial spread. Interestingly, both isolates harbored plasmids of different sizes with EC12 harboring a plasmid of ∼50 kb whereas EC24 contained a plasmid of ∼60 kb. On the other hand, there were isolates that were identical in their PCR fingerprints but were differentiated by their PFGE profiles (e.g., isolates EC14 and EC34) and conversely, isolates that were indistinguishable by PFGE were differentiated by PCR (isolates EC1 and EC4, and EC37 and EC39). This shows that these methods are complementary and a combined analysis would give a finer perspective bearing in mind the drawbacks of each of these methods.
All 47 E. coli isolates analyzed in this study were susceptible to imipenem. This finding is similar to previous reports [16,25]. Cefepime was equally sensitive at 98% which is a rate higher than the 80% sensitivity reported in Colombia [25] and 70% in China [26].
The resistance rates of E. coli isolates to ceftazidime (11%) and amoxillin-clavulanic acid (17%) were lower when compared to that from China-28% for ceftazidime and 84% for amoxillin-clavulanic acid [26]. E. coli resistance to amikacin in Malaysia (2% in this study) was still relatively low compared to 27% in Colombia [25] or 22.4% in Israel [28].
Forty E. coli isolates were classified as ESBL producers based on phenotypic or genotypic detection of ESBL: 38 isolates were classified as ESBL producers based on the genotypic detection of ESBL-encoding genes, and 2 others were categorized as ESBL producers based on the doubledisk synergy and phenotypic confirmatory tests. However, these two isolates, EC7 and EC31, did not harbor any of the tested ESBL-encoding genes and may instead harbor other genes such as bla TLA , bla IMP , and bla CMY [1,29] that were not included in our test. Although the double-disk synergy test had been reported to be reliable and easy-to-use, its major disadvantage is that the distance of disk placement for optimal sensitivity has not been standardized [20,29].
Analysis of the ESBL-encoding genes indicated that the majority of the ESBL-positive isolates harbored TEM-1 (88%) followed by CTX-M (20%), SHV (8%), and finally, OXA (5%). TEM-1 has been reported to be responsible for 90% of ampicillin resistance in E. coli [30]. The bla CTX-M gene is considered the most prevalent ESBL-encoding gene worldwide and is replacing TEM and SHV types as the predominant ESBL in many European countries [31]. The presence of the bla SHV-5 gene in 11 ceftazidime resistant E. coli isolates from one Malaysian teaching hospital has been reported [6]. The specific SHV subtypes could not be confirmed in this study as the primers used only amplified a portion of the bla SHV reading frame.
Analysis of integron-encoded integrases indicated that class 1 integron was the principal integron class in the Malaysian strains. Class 2 integron was in the minority, and no class 3 integron-encoded integrases were detected, a trend that has previously been reported [14,32]. Four different gene cassettes, namely, the aadA1, dfrA17-aadA5, dfrA12-orfF-aadA2, and aadA1-aadB-cmlA6 were found in the class 1 integron-positive isolates, and these have been previously described in E. coli as well as in other Enterobacteriaceae [15]. All 4 isolates positive for class 2 integron-encoded intI2 harbored the dfrA1-sat2-aadA1 gene cassette which has been reported [1,33]. Conjugation and transformation experiments indicated that the majority of the integrons and some of the ESBL-encoding genes (in particular bla TEM ) were plasmid-encoded and transmissible. Plasmids that were isolated from the E. coli hospital isolates were estimated to be larger than 50 kb in agreement with the sizes reported previously [34,35]. In 2 of the 3 successful conjugation experiments, plasmids with identical restriction profiles were isolated from the donor and the transconjugants. In the remaining case, based on the restriction profiles obtained, the plasmid that was isolated from the transconjugant was about 50 kb smaller than the donor EC31 strain (∼90 kb), indicating either that the donor strain harbored more than a single type of plasmid and that only the plasmid of about 40 kb was transferred, or that only a ∼40 kb portion of the original plasmid was successfully transferred by conjugation. The latter appeared to be a stronger possibility as separation of undigested plasmid DNA extracted from the parental EC31 strain seemed to indicate the presence of a single plasmid. Similar observations were also noted for the transformation experiments. Further characterization of these plasmids is clearly needed and is the subject of our on-going investigations. Nevertheless, our results indicate that most of the ESBL-encoding genes especially bla TEM are carried on plasmids which are transmissible suggesting that the spread of ESBL and other antibiotic resistance determinants is likely to be plasmid-mediated in agreement with the conclusions made by other reports [36,37] that plasmids are one of the main vehicles for spread of antibiotic resistance genes. This may have led to the high prevalence of ESBL-producers and multidrug resistance among E. coli hospital isolates in Malaysia.