High-Resolution Melting-Curve Analysis of Ligation-Mediated Real-Time PCR for Rapid Evaluation of an Epidemiological Outbreak of Extended-Spectrum-Beta-Lactamase-Producing Escherichia coli

Abstract

Methods for the confirmation of nosocomial outbreaks of bacterial pathogens are complex, expensive, and time-consuming. Recently, a method based on ligation-mediated PCR (LM/PCR) using a low denaturation temperature which produces specific melting-profile patterns of DNA products has been described. Our objective was to further develop this method for real-time PCR and high-resolution melting analysis (HRM) in a single-tube system optimized in order to achieve results within 1 day. Following the optimization of LM/PCR for real-time PCR and HRM (LM/HRM), the method was applied for a nosocomial outbreak of extended-spectrum-beta-lactamase (ESBL)-producing and ST131-associated Escherichia coli isolates (n = 15) and control isolates (n = 29), including four previous clusters. The results from LM/HRM were compared to results from pulsed-field gel electrophoresis (PFGE), which served as the gold standard. All isolates from the nosocomial outbreak clustered by LM/HRM, which was confirmed by gel electrophoresis of the LM/PCR products and PFGE. Control isolates that clustered by LM/PCR (n = 4) but not by PFGE were resolved by confirmatory gel electrophoresis. We conclude that LM/HRM is a rapid method for the detection of nosocomial outbreaks of bacterial infections caused by ESBL-producing E. coli strains. It allows the analysis of isolates in a single-tube system within a day, and the discriminatory power is comparable to that of PFGE.

INTRODUCTION

Antibiotic resistance among bacteria is an increasing problem in hospitals and other health care facilities. In order to prevent nosocomial infections, basic hygiene procedures must be strictly followed by all staff members. When outbreaks occur, it is crucial to identify and isolate patients as soon as possible (5). In order to identify potential nosocomial outbreaks, several molecular methods have been described (2, 14, 20). These epidemiological typing methods are based on both genomic and phenotypic principles. Many of the methods are time-consuming and expensive and require special equipment. When designing new typing methods, several factors need to be considered, depending on the purpose. These factors include stability, discriminatory power, reproducibility, speed, accessibility, cost-efficiency, and user efficiency (2, 19, 20). In addition, its appropriateness in a given situation (e.g., an outbreak situation) must be evaluated. As of today, pulsed-field gel electrophoresis (PFGE) is considered the gold standard for a large number of bacterial species because of its discriminatory power and high typeability (19, 20). The drawbacks of PFGE are that it is laborious and time-consuming and that the interpretation can be complex and requires rigorous standardization and experienced personnel in order to achieve reproducible results that are comparable over time and place. Clear criteria to determine whether two or more isolates are identical during a restricted time and place have been developed (15, 18). New genotyping methods are continuously being developed. One approach is repetitive-sequence-based PCR (rep-PCR), where repetitive sequences in the genome are amplified and subjected to electrophoretic separation (21). Other approaches leading to a much higher reproducibility among laboratories, as they are based on DNA sequencing, include multilocus sequence typing (MLST) and multilocus variable-number tandem-repeat analysis (MLVA) (10). Previously, Masny and Płucienniczak (12) described a novel method based on ligation-mediated PCR (LM/PCR) using low denaturation temperatures (T_Ds), resulting in specific melting-profile DNA product patterns for bacterial and fungal isolates. The method is based on genomic cleavage by restriction enzymes and the amplification of several fragments by ligation-mediated PCR followed by an analysis of the DNA fragment patterns by gel electrophoresis (8). The method has performed successfully for Enterobacteriaceae, Candida spp., Staphylococcus aureus, and Enterococcus spp., with a discriminatory power that is comparable to that of PFGE (6–8, 16).

Our objective was to further develop the LM/PCR method for real-time PCR and high-resolution melting analysis (HRM) in a closed, single-tube system optimized to achieve results within 1 day in small-sized hospitals. Furthermore, we have validated this typing method using isolates from a nosocomial outbreak together with control isolates for which there were no epidemiological links to the outbreak.

MATERIALS AND METHODS

Bacterial isolates.

Isolates of extended-spectrum-beta-lactamase (ESBL)-producing Escherichia coli from a well-characterized nosocomial outbreak (n = 15) and control isolates of ESBL-positive (n = 16) and ESBL-negative (n = 13) E. coli, including strain ATCC 17620, were collected at the Department of Clinical Microbiology, Kalmar County Hospital, Sweden, and the Swedish Institute for Communicable Disease Control, Solna, Sweden. Susceptibility testing and confirmation of ESBL production were performed according to national guidelines at the time of the study (www.srga.org).

The ESBL-producing E. coli isolates from the nosocomial outbreak were collected in 2008 from a hospital in Kalmar County, Sweden. An outbreak of ESBL-producing E. coli was suspected by the department of hospital hygiene because the patients carrying these isolates were epidemiologically related in time and place. Moreover, the isolates showed an unusual susceptibility pattern not commonly encountered in the local area (ESBL-producing E. coli with resistance to gentamicin and ciprofloxacin but susceptibility to trimethoprim-sulfamethoxazole). The majority of the outbreak isolates were urine samples from elderly patients at the medical department and two nursing homes surrounding a hospital within Kalmar County. Two patients (isolates 1 and 2) (Table 1) suffered from septicemia caused by the outbreak isolate.

Table 1.

Summary of analyses performed^a

Cluster	Isolate	LM/HRM pattern	Electrophoresis pattern (% similarity)	PFGE pattern	CTX-M group	ST131	TEM
1	1	1	1	1	1	Pos	Pos
1	2	1	1	1	1	Pos	Pos
1	3	1	1	1	1	Pos	Pos
1	4	1	1	1	1	Pos	Pos
1	5	1	1	1	1	Pos	Pos
1	6	1	1	1	1	Pos	Pos
1	7	1	1	1	1	Pos	Pos
1	8	1	1 (81)	1 (81)	1	Pos	Neg
1	9	1	1	1	1	Pos	Pos
1	10	1	1	1	1	Pos	Pos
1	11	1	1	1	1	Pos	Pos
1	12	1	1	1	1	Pos	Pos
1	13	1	1 (81)	1 (68)	1	Pos	Neg
1	14	1	1	1	1	Pos	Pos
1	15	1	1	1	1	Pos	Pos
Ctrl	16	Unique	Unique	Unique	9	Neg	Pos
Ctrl	17	Unique	Unique	Unique	9	Neg	Pos
Ctrl	18	Unique	Unique	Unique	9	Neg	Pos
Ctrl	19	[3]	Unique	Unique	9	Neg	Pos
Ctrl	20	Unique	Unique	Unique	1	Neg	Pos
Ctrl	21	Unique	Uniquec	Unique	9	Neg	Neg
Ctrl	22	Unique	Uniquec	Unique	9	Neg	Neg
Ctrl	23	Unique	Uniquec	Unique	9	Neg	Neg
A, urine	24	2	2	2	Neg	Neg	Neg
A, blood	25	2	2	2	Neg	Neg	Neg
Ctrl	26	[4]	Unique	Unique	Neg	Neg	Neg
Ctrl	27	Unique	[3]	Unique	Neg	Neg	Pos
Ctrl	28	Unique	Unique	Unique	Neg	Neg	Neg
Ctrl	29	Unique	Unique	Unique	Neg	Neg	Neg
Ctrl	30	[4]	[4]	NAb	Neg	Neg	Neg
Ctrl	31	[3]	[3]	Unique	Neg	Neg	Pos
Ctrl	32	Unique	Unique	Unique	Neg	Neg	Neg
Ctrl	33	Unique	Unique	Unique	Neg	Neg	Neg
Ctrl	34	Unique	Unique	Unique	Neg	Neg	Neg
Ctrl	35	Unique	[4]	Unique	Neg	Neg	Neg
Ctrl (ATCC)	36	Unique	Uniquec	Unique	Neg	Neg	Neg
2	37	5	5c	3	1	Pos	Pos
2	38	5	5c	3	1	Pos	Pos
3	39	6	6c	4	1	Pos	Neg
3	40	6	6c	4	1	Pos	Neg
4	41	7	7c	5	1	Neg	Pos
4	42	7	7c	5	1	Neg	Pos
5	43	8	8c	6	9	Neg	Pos
5	44	8	8c	6	9	Neg	Pos

^aThe outbreak isolates (n = 15) were from a hospital in Kalmar County, Sweden, in 2008. The remaining isolates were included as controls (n = 29). Numbers without brackets indicate clusters, whereas numbers with brackets indicate control isolates which showed clustering above the cutoff limit (80%) but without an epidemiological connection and where LM/HRM/electrophoresis and PFGE results were discordant. LM/HRM, ligation-mediated real-time PCR with high-resolution melting-curve analysis; Neg, negative; Pos, positive; ESBL, extended-spectrum beta-lactamase; A, patient A.

^bNA, not analyzed. Isolate 30 repeatedly caused a smeared pattern in the PFGE gel and could not be analyzed.

^cDetermined by conventional gel electrophoresis only. The remaining isolates were also run on capillary gel electrophoresis gels.

For control purposes, clinical E. coli isolates, both with (n = 8) and without (n = 13) ESBL, were collected from sequential urine samples from another hospital in Kalmar County in 2008. Two isolates of ESBL-producing E. coli from each of four previous clusters (isolates 37 to 44) (Table 1) were collected from the Swedish Institute for Communicable Disease Control, Solna, Sweden. Isolates 37 and 38 were from a previously described hospital outbreak (1). E. coli isolates were collected from both the blood and urine of a patient with urosepticemia (patient A). All isolates presented in Table 1 were coded prior to analyses to ensure blinding.

Ligation-mediated real-time PCR with high-resolution melting analysis.

In short, 5 to 10 isolated colonies were transferred into purified water, subsequently incubated for 10 min at 95°C, and centrifuged briefly. DNA extraction was carried out with a Magna Pure Compact system (Roche Diagnostics Scandinavia AB, Bromma, Sweden) according to the manufacturer's instructions, and the DNA concentration was measured by use of a UV/Vis spectrophotometer (Techtum Lab AB, Sweden). The DNA concentration range was kept at 100 to 500 ng/μl. DNA was digested by using HindIII (10 U/μl; Fermentas, St. Leon-Rot, Germany), by mixing 1 μg DNA with 2.5 μl buffer R (10×) (10 mM Tris-HCl [pH 8.5], 10 mM MgCl₂, 100 mM KCl, 0.1 mg/ml bovine serum albumin [Fermentas]) and 5 U HindIII in a total volume of 15 μl. The digestion mix was incubated for 30 min at 37°C. After digestion, ligation was carried out by the addition of 2.5 μl ligation buffer (10×) (330 mM Tris-acetate [pH 7.8], 660 mM potassium acetate, 100 mM magnesium acetate, 5 mM dithiothreitol [Epicentre]), 1.6 μM each oligonucleotide POW (5′-CTC ACT CTC ACC AAC GTC GAC-3′) and HINDLIG (5′-AGC TGT CGA CGT TGG-3′), 5 mM ATP (25 mM; Epicentre, WI), and 5 U T4 DNA ligase (10 U/μl; Epicentre) to a final volume of 25 μl. The samples were incubated for 60 min at room temperature, followed by enzymatic deactivation at 70°C for 10 min and cooling to room temperature for 10 min. Subsequent PCR was carried out with a 20-μl reaction mixture containing 2 μl ligation mix, 50 μM POW-AGCTT primer (5′-CTC ACT CTC ACC AAC GTC GAC AGC TT-3′), and 2× master mix (2× Type-It HRM PCR with Eva Green; Qiagen, Hilden, Germany). Real-time PCR with HRM (Rotorgene 6000; Corbett Research, Techtum Lab AB, Sweden) was performed by using a preincubation time of 7 min at 72°C to release unligated HINDLIG oligonucleotides and to fill in the single-stranded ends and create amplicons, followed by initial denaturation and enzyme activation at 95°C for 5 min and 22 cycles of denaturation either at 84°C for 60 s (optimized for the ligation-mediated real-time PCR and high-resolution melting analysis [LM/HRM]) or at 87.2°C for 60 s (optimized for electrophoresis detection as described in reference 9), with a subsequent elongation step at 72°C for 120 s. After the last cycle prior to HRM, samples were incubated at 72°C for 5 min and at 50°C for 30 s. Melting-curve analysis was performed by using HRM at 75°C to 90°C, with stepwise increases of 0.05°C.

Four reference isolates (isolates 1 and 2 from the suspected outbreak [cluster 1] and isolates 16 and 17, which were not epidemiologically linked to the outbreak) (Table 1) were used in all experiments for quality control purposes. All experiments were repeated at least six times in duplicates by two investigators.

Detection of DNA products by use of LM/HRM.

In order to objectively evaluate if HRM curves of two isolates (referred to as isolate A and isolate B below) were different within the same run, the following HRM evaluation algorithm was developed and applied: (i) no data below a baseline of 0.3 df/dt fluorescence units is evaluated; (ii) a peak is defined as an increase of more than 0.10 df/dt in fluorescence units from the previous baseline within 1°C; (iii) if two individual peaks of isolate A are located more than 0.5°C from the peaks of isolate B, the isolates are considered different; (iv) if isolate A shows only one peak, and this peak is located more than 0.5°C from more than two peaks in isolate B, the isolates are considered different; (v) if isolates A and B show only one or two peaks, the isolates should be analyzed by gel electrophoresis; and (vi) two isolates are highly likely to be identical if three peaks are colocalized at a distance of less than 0.2°C, and such isolates should be confirmed by gel electrophoresis.

Visualization of PCR products from LM/HRM.

Following LM/HRM, the DNA products were analyzed by agarose gel electrophoresis (E-gel EX, 2%; Invitrogen) and capillary electrophoresis (QIAxcel DNA high-resolution kit; Qiagen) according to the instructions provided by the manufacturers. In order to determine the denaturation temperature that produced the largest numbers of evaluable peaks for HRM, a range from 80°C to 90°C was tested. Optimized temperatures were 87.2°C for both agarose and capillary gel electrophoreses and 84°C for LM/HRM. Analyses of band patterns from agarose E-gels were done manually. Analysis of results from capillary gel electrophoresis was done by use of BioNumerics software, v. 6.01 (Applied Maths NV, Sint-Martens-Latem, Belgium), and the Pearson correlation with 3% optimization for curve-based analyses. A dendrogram was constructed by using the unweighted-pair group method with arithmetic means (UPGMA). Isolates with >80% similarity were defined as being related.

PFGE analysis.

Briefly, bacterial isolates and control strains were analyzed by pulsed-field gel electrophoresis (PFGE) (Table 1) using the restriction enzyme XbaI as described previously (11). XbaI-digested DNA from Salmonella enterica serovar Braenderup H9812 (www.cdc.gov/pulsenet; last accessed September 2010) was included as a normalization standard on every gel. DNA banding patterns were analyzed by use of BioNumerics software (Applied Maths NV). The Dice similarity coefficient and UPGMA were used for cluster analysis; isolates with >90% similarity were defined as being related, and isolates with >80% similarity were defined as being probably related.

Molecular detection of CTX-M, TEM, and O25b-ST131.

The association of the isolates with O25b-ST131 was analyzed according to methods described previously by Clermont et al. (4). The molecular detection of CTX-M to the group level and TEM was performed by PCR amplification and amplicon sequencing as described previously (13, 17).

RESULTS

All 15 outbreak isolates (cluster 1) clustered with LM/HRM results based on a blinded analysis (Table 1). Furthermore, these isolates belonged to CTX-M group 1 and were associated with O25b-ST131. Representative HRM curves for outbreak and control isolates are shown in Fig. 1 A. LM/HRM showed good reproducibility for outbreak and control isolates in six separate experiments performed by two investigators. As defined by the HRM evaluation algorithm, isolates with similar patterns were verified by capillary gel electrophoresis (see Fig. 3) and conventional gel electrophoresis (see Fig. 4b). Using PFGE as a gold standard, all outbreak isolates (cluster 1) except for isolate 13 showed a high level of similarity (>80%) to the outbreak isolates based on blinded analysis (Table 1 and Fig. 2). Isolate 13 originated from a patient in the same medical ward and during the same week as the other isolates of the outbreak were identified. For this isolate, the similarity was 81% for LM/HRM with capillary gel electrophoresis but 68% with PFGE (Table 1). Similar to PFGE, LM/HRM and subsequent capillary gel electrophoresis showed a similarity with the outbreak strains of 81% for isolate 8. Isolate 8 was obtained from the only outpatient identified in the outbreak (obstetrics and gynecology department from the same hospital and during the same period as the outbreak). Both isolate 8 and isolate 13 were TEM negative, in contrast to the other outbreak isolates, which were TEM positive (Table 1).

Fig. 1.

(A) Results from ligation-mediated real-time PCR with a T_D of 84°C and consecutive high-resolution melting analysis (LM/HRM) in a single-tube system. The HRM spectrum from a nosocomial outbreak of ESBL-producing E. coli is shown in panels A (isolate 2) and B (isolate 4), with a merging of the two in panel C. The HRM curves in panels D to F (isolates 16, 20, and 31) are representative control isolates of ESBL-producing E. coli, and in panel G, all spectra shown are merged. (B) LM/HRM results for four control isolates, which were not unique by application of the HRM evaluation algorithm. Upon visual inspection, panels A (isolate 19) and B (isolate 31), which are merged in panel C, could be evaluated as being different, whereas panels D (isolate 26) and E (isolate 30) were evaluated as being similar (panel F). Theses isolates were confirmed to be different by electrophoresis according to the algorithm.

Fig. 3.

Ligation-mediated real-time quantitative PCR (LMQ-PCR) and high-resolution melting analysis (LM/HRM) followed by subsequent capillary gel electrophoresis using a T_D of 87.2°C. All isolates were run in duplicates. The numbers refer to isolates in Table 1. Isolates 37 to 44 were confirmed by conventional electrophoresis only.

Fig. 4.

(a) Results from ligation-mediated real-time PCR and high-resolution melting analysis (LM/HRM) at a T_D of 84°C with agarose E-gel EX electrophoresis analysis. The LM real-time PCR HRM experiments were analyzed by agarose gel electrophoresis (E-Gel EX, 2%; Invitrogen). The optimal denaturation temperature used for gel electrophoresis was 87.2°C, as previously described (7), and the optimal temperature for LM/HRM was 84°C. C, cluster isolates (C1, C2, and C5); Ctrl, control isolate; ATCC, E. coli strain ATCC 17620. The numbers correspond to isolates in Table 1. (b) Results from LM real-time PCR at a T_D of 87.2°C with agarose E-Gel EX electrophoresis analysis.

Fig. 2.

Pulsed-field gel electrophoresis (PFGE) of outbreak and control isolates. The numbers refer to isolates in Table 1.

No control isolates clustered with the outbreak isolates by any of the methods tested (Table 1). Of the ESBL-producing control isolates, nine belonged to CTX-M group 9, and seven belonged to CTX-M group 1. The non-ESBL-producing control isolates were all negative for CTX-M and TEM. As expected, the urine and blood isolates from a patient with clinical urosepticemia (isolates 24 and 25) clustered by PFGE and LM/HRM. In addition, two representatives each from four previously identified clusters of ESBL-producing E. coli included for control purposes showed perfect agreement with PFGE and LM/HRM including confirmatory gel electrophoresis. These isolates included CTX-M group 1 isolates that were both positive and negative for O25b-ST131 and TEM.

In the group of control isolates without epidemiological relationships (n = 21), two pairs (isolates 19 and 31 in addition to isolates 26 and 30) (Table 1) could not be separated by using the LM/HRM evaluation algorithm. Evaluation of the complete HRM pattern for one pair of these isolates (isolates 19 and 31) suggested that they were different (Fig. 1B). Isolates 26 and 30 were confirmed to be different by conventional gel and capillary gel electrophoreses. Isolate 30 caused a smeared pattern by PFGE and could not be analyzed. The remaining 17 control isolates were all unique by LM/HRM. Among the control isolates investigated by LM/HRM and subsequent electrophoresis, isolate 34 showed an 86% similarity with the blood and urine isolates from the same patient, and isolates 27 and 35 showed an 81% similarity. However, all control isolates that could not be separated by use of the LM/HRM algorithm were confirmed to be different by use of LM/HRM and capillary gel electrophoresis (Table 1 and Fig. 3). Conventional gel electrophoresis of the DNA products also confirmed these results (Fig. 4).

DISCUSSION

In this study, we present a newly developed method based on restriction enzyme cleavage and ligation-mediated quantitative PCR followed by subsequent high-resolution melting-pattern analysis of DNA fragments (LM/HRM). We have shown that the resolution of the LM/HRM method for the identification of a nosocomial outbreak of O25b-ST131-associated ESBL-positive E. coli isolates was comparable to that of PFGE.

The LM/HRM method was developed from a previously described strategy applying conventional PCR and gel electrophoresis analysis of DNA products (6–8, 16). This method could essentially be described as a PCR-based PFGE method where the resolution may be altered based on the denaturation temperature. Our development of the method to a closed, single-tube, real-time PCR method with consecutive high-resolution melting-pattern analysis minimizes the hands-on time and risk of carryover contamination and allows the reporting of results within 1 day, also when confirmatory electrophoresis was required. The method does not require special equipment other than an HRM device attached to a real-time PCR machine, thereby making it accessible and affordable for the majority of small- and middle-sized hospital-based laboratories. An additional advantage of the LM/PCR method is that it is a restriction enzyme-based method, which allows application to various pathogens with small adjustments (6–9, 16), as no prior knowledge of DNA sequence data is necessary.

To allow analysis by HRM, we modified the original protocol (8) by lowering the T_D from 87.2°C to 84°C. During the optimization of the T_D, it was clear that the resolution of HRM patterns is highly dependent on the denaturation temperature of the PCR (data not shown). We are currently evaluating the LM/HRM method for other species, and for each group of bacteria or fungi to be analyzed by the method, an individual T_D needs to be defined, as previously shown (6–9, 16). The reason why the original temperature for E. coli did not provide enough resolution by HRM in the present report is probably due to the synthesis of too many amplicons, which were visualized by the gel electrophoresis pattern of the DNA products (Fig. 4). In an effort to allow an analysis of the HRM results objectively, we developed an HRM evaluation algorithm, which was validated against PFGE data. However, as shown in Fig. 1, a comparison of the whole HRM pattern is usually enough to separate isolates that are not related. As previously described (22), in relation to HRM of several DNA products, it is essential to choose a DNA-binding fluorophore that can be used at saturated concentrations to minimize the risk of relocalization during the melt phase. For the optimal resolution of HRM, Eva Green (Qiagen) was used, as this is a third-generation intercalating dye with low toxicity, which enables binding to higher concentrations of DNA during PCR, resulting in increased saturation, thus increasing the sensitivity (22).

In total, 44 different E. coli isolates have been analyzed with the newly developed LM/HRM method. Isolates belonging to the nosocomial outbreak were shown to be identical by the LM/HRM method, and the results were fully reproducible within six independent experiments by two separate investigators. All outbreak isolates belonged to O25b-ST131, confirming the successful spread of this clone. Two isolates from the outbreak (isolates 8 and 13) were confirmed to be similar by the LM/HRM method but showed clustering at a lower level by LM/HRM with capillary gel electrophoresis confirmation and PFGE. These two isolates had an epidemiological relationship, including clustering in time and place, as well as a similar unusual susceptibility pattern for Kalmar County (ESBL-producing E. coli with resistance to gentamicin and ciprofloxacin but susceptibility to trimethoprim-sulfamethoxazole). However, these two outbreak isolates (isolates 8 and 13) were TEM negative, in contrast to the other isolates in the cluster, and showed a distinctly different replicon type (performed according to methods described previously by Carattoli et al. [3]) compared to those of the other outbreak isolates (data not shown), indicating that these isolates are genetically different from the other outbreak isolates. One possible explanation for this finding might be that the overall backbone of the genome for samples 8 and 13 is the same as that of the outbreak strain but that they have lost a plasmid or a fragment of a plasmid carrying the TEM gene. A final answer in regard to how samples 8 and 13 are related to the outbreak strain would require whole-genome sequencing.

The primary aim of the LM/HRM method is to analyze isolates for which there is an epidemiological relationship. Thus, we chose a predefined cutoff of an 80% similarity for the dendrogram, since we had epidemiological data to support the nosocomial outbreak. However, for isolates for which there was not any information on epidemiological relationships, such as in randomly collected isolates, a cutoff of a 90% similarity could be more accurate to avoid false associations by using PFGE in particular. The most suitable cutoff may differ between species, by the availability of epidemiological background data, and by the analytical method used (20).

To further explore the resolution of the methods, we included several control isolates, both ESBL- and non-ESBL-producing E. coli isolates, without epidemiological relationships with the outbreak isolates. Moreover, we also included isolates from previously defined clusters of ESBL-producing E. coli isolates. For these isolates, there was an excellent agreement between PFGE and LM/HRM data, including a previously described ST131-associated outbreak from southern Sweden (1). Based on the PFGE analysis, there was more than a 90% similarity for the previously defined clusters (clusters 2 to 5) as well as for all isolates except two in the nosocomial outbreak (cluster 1). It was previously shown that ST131-associated isolates present a challenge for typing methods, as they show a ≥60% similarity by PFGE analysis (2), which could also be confirmed in the present analysis. No clustering between outbreak isolates and control isolates was observed. Within the group of control isolates, four of them (isolates 19 and 31 in addition to isolates 26 and 30) (Table 1) could not be clearly separated by use of the LM/HRM evaluation algorithm. However, by performing an evaluation of the complete HRM patterns for two of these isolates, in addition to the use of the predefined algorithm, they were interpreted as being different (Fig. 1B). For both of these pairs, differentiation could be done when results from LM/HRM and capillary gel electrophoresis analyses were evaluated (Table 1). This finding indicates that until more data have been generated by use of the LM/HRM method, isolates evaluated as being similar based on the HRM evaluation algorithm should be confirmed by electrophoresis, which can be conveniently performed directly on the DNA fragments from the LM/HRM analysis. Either capillary gel electrophoresis or premade E-gel EX (Qiagen) could be used to generate final results from the DNA fragments within an hour.

In conclusion, the non-sequence-based LM/HRM method can rapidly and accurately determine whether two or more isolates are similar or not similar for nosocomial outbreak investigations. The system is well protected against cross-contamination in the laboratory environment, and the results can be achieved within a day.