|Home | About | Journals | Submit | Contact Us | Français|
Pseudomonas aeruginosa is an opportunistic pathogen that causes nosocomial infections in intensive care units. Determining a system of typing that is discriminatory is essential for epidemiological surveillance of P. aeruginosa. We developed a method for the typing of Pseudomonas aeruginosa, namely, multiple-locus variable-number tandem-repeat (VNTR) typing with high-resolution melting analysis (HRMA). The technology was used to genotype a collection of 43 environmental and clinical strains isolated during an outbreak in a neonatal intensive care unit (NICU) that we report. Nineteen strains isolated in other departments or outside the hospital were also tested. The genetic diversity of this collection was determined using VNTR-HRMA, with amplified fragment length polymorphism (AFLP) analysis as a reference. Twenty-five and 28 genotypes were identified, respectively, and both techniques produced congruent data. VNTR-HRMA established clonal relationships between the strains of P. aeruginosa isolated during the outbreak in the NICU and proved, for the first time, the role of mineral water as the inoculum source. VNTR typing with one primer pair in association with HRMA is highly reproducible and discriminative, easily portable among laboratories, fast, and inexpensive, and it demonstrated excellent typeability in this study. VNTR-HRMA represents a promising tool for the molecular surveillance of P. aeruginosa and perhaps for molecular epidemiologic analysis of other hospital infections.
Pseudomonas aeruginosa is one of the most important opportunistic pathogens involved in nosocomial infections worldwide. P. aeruginosa is a ubiquitous bacterium which is able to survive under many conditions, and many hospitals create new niches for its development. Several reservoirs are frequently involved in infections, such as medical equipment, contaminated environments (sinks, taps, etc.), contaminated bottled water (10), contaminated products (eye drops, soaps, etc.), health care workers, other contaminated patients, and sometimes the patient him- or herself (4, 21, 39). In neonatal intensive care units, in addition to the previously described sources, P. aeruginosa strains were recovered in contaminated feeding bottles and were associated with breast milk feeding (5, 25, 34).
Typing of nosocomial pathogens is necessary to determine the source of an outbreak and to take effective control measures to prevent the spread of the pathogens. A good typing tool should be convenient, highly discriminatory, reproducible, fast, and inexpensive. For inoculum source investigations, it should target loci with a rather fast molecular clock so the epidemiological relationships are not overestimated. Phenotypic and biochemical methods based on serotyping, bacteriophage typing, or antimicrobial susceptibility have weak discriminatory power and were replaced with DNA-based techniques. Among these molecular methods, pulsed-field gel electrophoresis (PFGE) of restricted genomic DNA fragments is commonly considered the most appropriate technique available to type closely related isolates of P. aeruginosa. However, this method is time-consuming (2 or 3 days) and expensive and requires specialized equipment and qualified operators (16, 27). Consequently, other molecular methods have been used to discriminate strains with the same accuracy while avoiding the drawbacks inherent to PFGE.
Random amplified polymorphic DNA PCR typing is faster and less expensive than PFGE, but it is less discriminatory (35). Multilocus sequence typing, a method based on allelic differences among housekeeping genes, is less discriminatory than PFGE for detecting genetic differences among P. aeruginosa isolates (20). Genotyping using amplified fragment length polymorphism (AFLP) analysis was applied successfully to P. aeruginosa epidemiological investigations (22, 30). Moreover, AFLP typing is comparable to PFGE (6) and is more accurate under certain conditions (35). Other methods target short repeated DNA sequences that are interspersed in bacterial genomes. Repetitive DNA sequence-based PCR (rep-PCR) is fast and convenient but lacks interlaboratory reproducibility (36). Semiautomated rep-PCR seems to be highly reliable for typing of P. aeruginosa but requires specialized equipment (8). Minisatellites or variable-number tandem repeats (VNTR) are appropriate tools for developing a highly discriminatory and reproducible typing assay. Multiple-locus VNTR analysis (MLVA) displays a similar clustering ability to that of PFGE (28), and increasing the marker number was shown to enhance the typing efficiency (38).
High-resolution melting (HRM) technology is now available on many real-time PCR devices. HRM curve analysis (HRMA) is a fast, convenient, and cost-effective screening method for PCR products. Such a technology could be used advantageously to separate polymorphic PCR products generated by VNTR amplifications, using the same equipment. The aim of this work was to elucidate the source of a P. aeruginosa outbreak in a neonatal intensive care unit by using such an alternative technique. Data derived from HRMA of a VNTR locus were compared with data obtained by AFLP analysis. AFLP analysis was used as a reference technique that requires no prior knowledge of genomic DNA sequences and targets multiple interspersed sites over the whole genome. We believe that this is the first study to use VNTR typing with HRMA and is the first time that mineral water used to prepare baby bottles is described as the source of contamination in a neonatal intensive care unit.
P. aeruginosa strains were obtained from neonates admitted to the neonatal department (28 beds, including 10 for the neonatal intensive care unit and 2 for pediatric resuscitation; 350 admissions per year) of the Groupe Hospitalier Sud Réunion (the largest hospital on the island, with 1,300 beds). This department has an incidence of infections or colonizations with P. aeruginosa of <2 per 1,000 hospitalization days (14). Criteria for colonization in this outbreak report were the presence of P. aeruginosa in peripheral samples, no clinical signs of infection, and a normal value for C-reactive protein. During the first 3 months of 2006, the incidence was 1.4 per 1,000 hospitalization days. From April to June 2006, the incidence rose to 6.8 infections per 1,000 hospitalization days. During this prospective study, to investigate the outbreak, serial stool and conjunctival samples were taken once a week from 100 neonates admitted to the neonatal department during this period. Ocular swabs were chosen because these samples were positive in a previous epidemic due to P. aeruginosa (14). These samples are easy to obtain and are not invasive. Additional specimens were collected for clinical reasons. The presence of P. aeruginosa was investigated in 100 neonates admitted to the hospital. Forty-two had at least one sample positive for P. aeruginosa. Among these 42 patients, 40 were colonized and 2 were infected. In this study, 20 patients were included, among whom 19 patients were colonized and 1 was infected. Twenty-one strains isolated from stool, 3 strains from eyes, 1 strain from blood culture, and 1 strain from a tracheal aspirate were studied. Nine environmental samples positive for P. aeruginosa were also obtained during the study period to research the source of contamination, particularly in the aqueous environment of the department, where P. aeruginosa is usually present (3). Positive environmental samples were obtained from 2 sink drains, 4 tap water samples, 2 faucets, and 1 sink, with potential contamination throughout the neonatal department, including the room where milk and baby bottles were prepared. Three mineral water bottles (different brands and batches) purchased by the hospital, 4 mineral water samples with milk (in a baby bottle or not), and 1 whisk used to mix mineral water and milk were collected. Some strains were isolated from clinical samples (13 strains) and from the environment (1 strain) in other departments of the hospital and outside the hospital (5 strains isolated from sink drains in the houses of the authors) during the same period. Water from taps was collected in 250-ml sterile bottles with sodium thiosulfate. For sampling of surfaces, an agar contact technique (Count-Tact agar; bioMérieux, Marcy l'Etoile, France) was used. Other environmental samples were obtained with cotton swabs and were plated onto Mueller-Hinton plates (bioMérieux, Marcy l'Etoile, France). Water was concentrated by filtration through a 0.45-μm membrane and then seeded onto Cetrimid agar (bioMérieux, Marcy l'Etoile, France). Clinical stool samples were plated onto MacConkey agar (bioMérieux, Marcy l'Etoile, France), and conjunctival samples were plated onto Polyvitex agar (bioMérieux, Marcy l'Etoile, France). The strains were identified by the analytical profile index procedure (API 20NE test; bioMérieux, Marcy l'Etoile, France). Sixty-two strains were identified, and a colony of each strain was frozen at −80°C on cryo-beads (AES, Combourg, France). Figure Figure11 summarizes the epidemiological data for the P. aeruginosa strains studied. Strain ATCC 27853 was chosen as the root for tree construction.
Target DNA was prepared from cells grown overnight at 37°C on Polyvitex agar (bioMérieux, Marcy l'Etoile, France). Cells were resuspended in 200 μl of RNase- and DNase-free water (Sigma).
A DNeasy blood and tissue kit (Qiagen, Courtaboeuf, France) was used for DNA extraction, with an extended lysis step (overnight at 56°C). The concentration of DNA was measured with Quant-iT PicoGreen dsDNA reagents and kits (Invitrogen, Cergy Pontoise, France).
The AFLP experiments were performed in 96-well plates in a GeneAmp model 9700 thermocycler (Applied Biosystems, Foster City, CA), as previously described (1). Oligonucleotides were synthesized by Applied Biosystems. Digestions were carried out with SacI and MspI restriction enzymes. The first combination for the selective amplification step used selective primers MspI+A and 5′-6-carboxyfluorescein (FAM)-labeled SacI+C, the second combination used MspI+C and 5′-VIC-labeled SacI+C, the third combination used MspI+T and 5′-NED-labeled SacI+C, and the fourth combination used MspI+G and 5′-PET-labeled SacI+C. Samples were then prepared for capillary electrophoresis by adding 1 μl of each of the four final PCR products to 10.7 μl of formamide and 0.3 μl of LIZ500 DNA ladder (Applied Biosystems, Foster City, CA), used as an internal standard. The mixture was then denatured for 5 min at 95°C and placed on ice for at least 5 min. Electrophoresis was performed in an ABI Prism 3130 genetic analyzer (Applied Biosystems, Foster City, CA), using performance-optimized POP-7 polymer (Applied Biosystems, Foster City, CA) at 15,000 V for about 20 min at 60°C, with an initial injection time of 69 s. The AFLP fingerprints were analyzed visually using GeneMapper software 4.0 (Applied Biosystems, Foster City, CA). The amplified fragments were scored as present (1) or absent (0) to create binary matrices before analysis by R software, version 2.9.0 (32). Bands with intensities above a preset level (500 relative fluorescence units) were the only ones to be scored. To test the reproducibility of the AFLP technique, two independent DNA extractions were used for all strains, and nonreproducible markers were discarded from the analysis. Strain Pyo-7 was used as a control in each experiment. The genetic similarities were calculated with the ade4 package (9) in R software for each AFLP condition and for the concatenated data set, using the Dice similarity index (7).
Kendall's coefficient of concordance (W) (23) among the four distance matrices was computed and tested through a permutation test (9,999 permutations) with the ape package in R software (29). A posteriori permutation tests of the contribution of each of the four AFLP condition distance matrices to the overall concordance of the group were carried out using Mantel tests. Permutational probabilities were corrected using Holm's correction method. A weighted neighbor-joining tree was constructed with the ape package in R software for the concatenated data set, using Dice similarity coefficients as distances. Bootstrap values (12) were computed to assess the robustness of the tree (1,000 resamplings).
Six polymorphic tandem repeats (ms77, ms127, ms142, ms172, ms207, and ms209) with different motif lengths (39, 15, 115, 54, 6, and 6 bp, respectively) and PCR product lengths (442, 210, 890, 789, 146, and 148 bp, respectively, for reference strain PAO1), as described by Vu-Thien et al. (38), were tested with 15 clinical strains. Previously, 4 of these 6 markers were described by Onteniente et al. (28). First, we looked for the typeability of these VNTR. We chose tandem repeats of different motif lengths (micro- and minisatellites) which had PCR products of different sizes to compare the discriminatory powers of VNTR-HRMA and AFLP typing and to obtain the best discriminatory power with VNTR-HRMA. The influence of motif lengths of VNTR on HRMA is unknown. When the VNTR was chosen, the reproducibility of our technique was tested.
The conditions were optimized using a Light Cycler 480 high-resolution melting master kit (Roche Diagnostics, Meylan, France). The 20-μl reaction mix contained 1× master mix, 3 mM MgCl2, a 0.3 μM concentration of each VNTR primer, and LCGreen I dye.
PCRs were performed with a Light Cycler 480 instrument (Roche Diagnostics, Meylan, France) under the following conditions: an activation cycle at 95°C for 10 min; 10 cycles of denaturation at 95°C for 10 s, annealing at 60°C for 15 s, and elongation at 72°C for 25 s; and 40 cycles of 95°C for 10 s, 65°C for 15 s, and 72°C for 25 s.
HRMA was performed immediately after PCR cycling. The samples were heated to 98°C, rapidly cooled to 40°C, and heated again to 70°C, and then fluorescence was measured with continuous acquisitions (25 per °C; 1°C/s until 98°C).
LC480 genotyping software (Roche Diagnostics, Meylan, France) was used to calculate the melting curve data, and 3 groups were obtained (Fig. (Fig.2).2). The limits of each group (premelt/postmelt slider settings) were searched and fixed (Fig. (Fig.3)3) to obtain the most genotypes. The difference plot was analyzed by the software, and samples with similar melting profiles were grouped.
To assess the reproducibility of the analysis, five repeats were carried out with the chosen limits.
The discriminatory power of the different typing methods was evaluated using an index described by Hunter (19). Desirable techniques for inoculum-tracking purposes should have a high discriminatory power (index of >0.95) (4).
The Rand index (15, 31) estimates the proportions of agreements and disagreements found for two classifications over the total number of pairs. Hubert and Arabie (18) proposed a correction for chance for this index to ensure that its maximum value is 1 and expected value is 0 when the partitions are selected randomly (15). The corrected Rand index was computed to assess the agreement between classifications obtained with AFLP typing and HRMA (13), using fpc package 1.2-4 (17) in R software.
Six loci were tested on 15 clinical strains. Three loci were eliminated (ms77, ms127, and ms172) because many strains could not be interpreted by HRMA, probably due to low concentrations of PCR products (data not shown). Markers ms207 and ms209 were tested on the 62 strains and were eliminated because their discriminatory power in HRMA was low. In addition, the discriminatory power obtained when these markers were associated with ms142 was not superior to that for ms142 alone. In this study, better HRMA results were obtained with the VNTR with the greatest motif length and the greatest PCR product length. The influences of the PCR product length and the motif length have to be confirmed by typing other bacteria by VNTR-HRMA. The typeability with ms142 primers was 98.4%. Strain P2 was the only one that was not amplified with these primers. Therefore, ms142 primers were selected because their typeability and discriminatory power were high.
A total of 63 P. aeruginosa strains (62 isolated strains and strain ATCC 27853) were typed by AFLP typing, regarded as the “gold standard” in this study, and yielded 419 reproducible band positions (97.8% segregated) under the 4 AFLP conditions (99, 122, 97, and 101 positions for the SacI+C/MspI+A, SacI+C/MspI+C, SacI+C/MspI+T, and SacI+C/MspI+G conditions, respectively). The number of markers per strain ranged from 42 to 110. Twenty-nine genotypes (Fig. (Fig.1)1) were identified from the concatenated data set (genotypes A1 to A29). Dice's coefficient of similarity between genotypes ranged from 0.08 to 0.85 (median = 0.65). Kendall's coefficient of concordance between the four distance matrices revealed a high degree of correlation among them (W = 0.85; P < 0.0001). The mean of the Mantel correlations for each distance matrix with the others ranged from 0.74 (for the SacI+C/MspI+C condition) to 0.83 (for the SacI+C/MspI+A condition). The probability of these a posteriori permutation tests, corrected with Holm's method, was 0.0004, which indicated that the four distance matrices were also highly congruent with one another. Therefore, subsequent analyses were performed on a pooled data set.
The weighted neighbor-joining tree (Fig. (Fig.1)1) showed that 32 strains, which were all outbreak related, of the 63 strains clustered within a genetic cluster containing 31 strains identified as genotype A6 and one strain (P8) identified as genotype A7. This genotype differed from genotype A6 only by the absence of a marker (71 markers for A7 versus 72 markers for A6). Ten AFLP markers were found to be specific for these genotypes. This genetic cluster was supported by a bootstrap value of 100%.
The interreproducibility of VNTR-HRMA on the 61 strains tested five times was 100%. Based on a reference index of discriminatory power of 1.00 for AFLP typing, the index of discriminatory power calculated for 61 strains was 0.99 for VNTR-HRMA. The corrected Rand index between AFLP and HRMA classifications (0.95) showed a high level of agreement between these two classifications.
All of the suspected outbreak-related strains were identified as genotype H5 by VNTR-HRMA. The strain (P8) with genotype A7 (by AFLP typing) was classified into group H5. Among the 26 strains isolated from 21 neonatal patients during the study, three genotypes were identified by AFLP typing (A6, A7, and A27) and two others (H5 and H23) were identified by VNTR-HRMA, with H5 corresponding to A6 and A7 and H23 corresponding to A27. Genotypes A6 and H5 were isolated the most often, in 92.3% (24/26 strains) and 96% (25/26 strains) of strains, respectively.
AFLP typing showed that most patients (20/21) were colonized by strains with the A6 genotype; the same results were obtained by VNTR-HRMA for the H5 genotype. For 2 patients (patients 17 and 19), P. aeruginosa was isolated from several samples: it was isolated from 2 samples for patient 17 (conjunctival and stool samples) and from 4 samples for patient 19 (conjunctival, stool, blood culture, and tracheal aspiration samples), and these different isolates all exhibited genotype A6 by AFLP typing and genotype H5 by VNTR-HRMA. The death of patient 19 was attributed directly to a P. aeruginosa infection with a genotype A6 and H5 strain. P. aeruginosa was isolated from two stool samples from patient 20, and the 2 isolates exhibited two different genotypes (genotype A6 or the closely related genotype A7, which is probably a variant of the same strain, and genotype H5). Seventeen strains were isolated from the environment of the neonatal department during the study. AFLP typing of these strains revealed 8 genotypes, including the A6 genotype (47% of these strains), and 8 genotypes, including the H5 genotype (47% of these strains), were revealed by VNTR-HRMA. The classification of genotypes determined by AFLP typing and VNTR-HRMA had a concordance of 100%. Strains exhibiting genotypes A6 and H5 were isolated either from commercial bottled mineral water (bottled at 15:40 and 15:41 ratios), from milk prepared with this water, or from whisks used to mix water and milk. None of the strains isolated in other departments of the hospital (with 14 AFLP genotypes and 13 VNTR-HRMA genotypes; one strain [P2] was not amplified) or outside the hospital exhibited the AFLP genotype A6 or H5. Outside the hospital, five new genotypes were identified by AFLP typing and VNTR-HRMA.
The role of P. aeruginosa in outbreaks caused by contaminated water has been well described (2, 33, 37). Recently, a P. aeruginosa outbreak provoked by the contamination of feeding bottles was reported (34). To our knowledge, there has not been any reported outbreak of P. aeruginosa caused by mineral water used to prepare milk in a neonatal intensive care unit.
The integration of molecular genotyping in the epidemic investigations of P. aeruginosa in the neonatal intensive care unit established the role of two batches of a brand of mineral water. When another brand of mineral water was used, the outbreak stopped. The contamination of the mineral water was reported to health authorities and to the manufacturer. Analysis of the process of fabrication by the manufacturer has permitted an understanding of the reasons for the presence of P. aeruginosa in these two batches: the decontamination process after a stop of production due to a factory mechanical failure was not realized, and the length of the control culture was only 2 days (it was later increased to 5 days to avoid false-negative results).
The pathogen associated with the outbreak was rapidly identified, and the role of water was suspected. The source of the outbreak was more difficult to identify. During the outbreak, five patients died: four deaths were not attributed to a P. aeruginosa infection, but one death was attributed to a P. aeruginosa infection. The other patient infected by P. aeruginosa had a septicemia. Forty patients were colonized.
At first, two phenotypic methods were used, but the indices of discriminatory power were too weak (<95%) to determine the source of the outbreak (for antibiotyping, 60%; for serotyping, 82%; and for antibiotyping plus serotyping, 92% [data not shown]).
The high discriminatory power of AFLP typing permitted us to define genotype A6, which was found only in the environmental samples of mineral water, in preparations made with that water, and in neonates admitted to the neonatal department during the outbreak.
Although AFLP typing can be suitable for outbreak investigations of P. aeruginosa (11), easier and cheaper techniques could improve the real-time monitoring of an outbreak.
MLVA is now used widely for the genotyping of many different bacteria. Consequently, we tested the effectiveness of this technique for the genotyping of P. aeruginosa (28, 38). Onteniente et al. (28) showed with a set of 7 loci that MLVA has a high discriminatory power. Vu-Thien et al. (38) concluded by use of 15 VNTR that MLVA is an efficient, easy, and rapid typing method.
In this study, we targeted a single locus (ms142). The discriminatory power of the locus, analyzed by polyacrylamide gel electrophoresis, was 0.95 (17 genotypes [data not shown]). In the study of Vu-Thien et al. (38), agarose gel electrophoresis analysis showed a Hunter discriminatory index of 0.81 with ms142.
HRMA combined with VNTR typing permitted an increase of the discriminatory power of the typing system to 25 genotypes for the 62 strains tested, which is very close to the number of genotypes obtained by AFLP typing (28 genotypes). The index of discriminatory power obtained was 0.99, which is very close to that of AFLP typing. The same number of genotypes was identified by VNTR-HRMA and AFLP typing for distances superior to 0.5 (data not shown). In this study, VNTR-HRMA had a high discriminatory power and an excellent typeability and reproducibility.
In our study, VNTR-HRMA confirmed the role of two batches of a brand of mineral water. The technique has the advantages of requiring only PCR reagents, LCGreen I dye, 2 unlabeled primers hybridizing with the flanking region of a VNTR locus, and 2 min of post-PCR analysis. No other analysis, such as electrophoresis, is necessary, and the risk of contamination is reduced.
This technique is simple, rapid (amplification, 42 min; and HRMA, 2 min), and very cost-efficient and requires only melting instrumentation. One strain was not amplified, perhaps due to mutations in the VNTR locus or the deletion, in whole or in part, of the VNTR locus.
In this study, outbreak strains and other strains isolated in La Réunion, France, were tested. The ms142 locus was tested with 90 European strains, suggesting a good typeability for this technique (38). However, the high discriminatory power of HRMA with other strains must be confirmed.
Published ahead of print on 23 June 2010.
†The authors have paid a fee to allow immediate free access to this article.