Comparison of PCR-based methods for typing Escherichia coli

Comparison of PCR-based methods for typing Escherichia coli

ORIGINAL ARTICLE Comparison of PCR-based methods for typing Escherichia coli D. Jonas1,2, B. SpitzmuÈller1, K. Weist2,3, H. RuÈden2,3 and F. D. Daschn...

139KB Sizes 22 Downloads 201 Views

ORIGINAL ARTICLE Comparison of PCR-based methods for typing Escherichia coli D. Jonas1,2, B. SpitzmuÈller1, K. Weist2,3, H. RuÈden2,3 and F. D. Daschner1,2 1

Institute of Environmental Medicine and Hospital Epidemiology, Freiburg University Hospital, Freiburg, 2National Reference Center for Surveillance of Nosocomial Infections, Berlin, and 3 Institute Institute of Hygiene, Free University of Berlin, Berlin, Germany

Objective To establish a library typing system appropriate for studying cross-transmis-

sion of Escherichia coli.

Methods Eighteen epidemiologically unrelated isolates were genotyped by means of

pulsed-®eld gel electrophoresis (PFGE), random ampli®ed polymorphic DNA (RAPD), repetitive (rep) PCR, and ¯uorescent ampli®ed fragment length polymorphism (AFLP). Fingerprints were analyzed either by Pearson correlation or, in the case of AFLP, by Dice coef®cients employingthenovel `uncertain band' softwaretool from GelCompar II. Duringa nine-month period, 112 isolates taken from 93 patients hospitalized in ®ve intensive care units were analyzed by use of the two most discriminative PCR typing methods.

Results Genotyping by RAPD and rep-PCR revealed insuf®cient discrimination. Among 18 epidemiologically unrelated strains with 17 different PFGE patterns, IS3 rep-PCR and AFLP distinguished 10 and 18 types, respectively. Comparison of the different methods for analysis of AFLP ®ngerprints showed that the Dice coef®cients, which ignore `uncertain bands', offered the best concordance with visual interpretation. Consecutive isolates originating from the same patient differed in less than three fragments. Conclusions AFLP analysis showed the highest discriminative capacity for PCR typing of E. coli isolates. Analysis of ®ngerprints employing the Dice coef®cients proved the most ef®cient method for an automated software-based retrieval of visually indistinguishable genotypes in an AFLP ®ngerprint database. Keywords Escherichia coli, molecular typing, AFLP, uncertain bands tool Accepted 12 September 2002

Clin Microbiol Infect 2003; 9: 823±831

INTRODUCTION PCR-based typing has proved to be a useful alternative to macrorestriction analysis (MRA) by pulsed-®eld gel electrophoresis (PFGE) in many Gram-negative bacterial species [1]. There are three ways to generate PCR ®ngerprints: (1) random ampli®ed polymorphic DNA (RAPD), which ampli®es DNA sequences of unknown origin under low-stringency conditions; (2) repetitive Corresponding author and reprint requests: D. Jonas, Institute of Environmental Medicine and Hospital Epidemiology, University Hospital Freiburg, Hugstetter Str. 55, D79106 Freiburg, Germany Tel: ‡49 761270 5445 Fax: ‡49 761270 5485 E-mail: [email protected]

PCR (rep-PCR), which represents DNA spacers between two repetitive elements, e.g. insertion sequences (IS); and (3) PCR-ampli®ed restriction fragment length polymorphism, a highly reproducible method that has been demonstrated to be useful for many different bacterial species [2]. PCR amplicons can be separated ef®ciently by employing automated DNA sequencers followed by an analysis of the primarily digital data by use of pattern recognition software. Without the assignation of bands, the patterns are compared in a curve-based manner. This Pearson correlation generally distinguishes different genotypes at a level below 70±80% [3]. When genotyping Escherichia coli, interest is focused mainly on enteropathogenic isolates. However, genotyping methods have been employed for

ß 2003 Copyright by the European Society of Clinical Microbiology and Infectious Diseases

824 Clinical Microbiology and Infection, Volume 9 Number 8, August 2003 non-enteropathogenic strains, to investigate different issues such as the spread of ¯uoroquinoloneresistant strains [4,5], association of virulence traits with certain clonal groups [6,7], and cross-transmission in hospitals [8±10]. The evaluation study of different PCR genotyping methods presented here was initiated by a project concerning the association between patient exposure variables, nosocomial infection, and the transmission of ten different bacterial pathogens [11]. These included E. coli, which is the most frequently encountered Gram-negative species in intensive care units (ICUs). The ®ndings of this epidemiologic network project will be published in detail elsewhere. The aim of this study was three-fold. First, different published protocols for typing by means of RAPD, rep-PCR and ¯uorescent ampli®ed fragment length polymorphism (AFLP) were compared with PFGE, using a small panel of epidemiologically unrelated strains [7,9,10,12±15]. Second, an ef®cient method for analysis of AFLP ®ngerprints was sought, as well as a sensible, though preliminary, estimate of a cut-off value to discriminate genotypically related and non-related isolates. Third, the two most discriminative PCR methods were routinely used to type isolates originating from ®ve ICUs during a nine-month period. MATERIALS AND METHODS Bacterial strains The collection of epidemiologically unrelated strains included 16 strains from 13 different laboratories and the type strains ATCC 11229 and 11775 [16]. During the study presented here, 112 clinical isolates with unique patient identi®cation numbers were obtained from 93 patients with suspected nosocomial infections. Following biochemical identi®cation, they were genotyped. Stock cultures were stored at 70 8C in de®brinated horse blood (Acila GMN mbH, MoÈrfelden, Germany). Prior to being used in experiments, strains were grown on Columbia sheep blood agar for 16 h (Heipha Diagnostika, Heidelberg, Germany). Pulsed-field gel electrophoresis Extraction of intact chromosomal DNA from isolates and the preparation of agarose plugs for subsequent PFGE on a CHEF DRIII system

(Bio-Rad, Munich, Germany) were done according to standard protocols [17]. Following incubation in proteinase K and digestion with SpeI (New England Biolabs, Frankfurt, Germany), the PFGE parameters used were 200 V for 23 h at 14 8C with switch times of 5 s (initial) and 60 s (®nal). Bacteriophage lambda concatemers (Bio-Rad) were used as size standards. RAPD and rep-PCR Colonies (3±5) from a freshly grown culture were resuspended in 100 mL of dH2O and incubated for 15 min at 95 8C. After chilling on ice, bacterial debris was pelleted by centrifugation at 15 000g for 20 s. The supernatant was transferred into a fresh microfuge tube. The typing methods investigated were RAPD (primers: M13, DAF4, 1247), rep-PCR (primers: ERIC-2, IS3A in combination with IS3B), and AFLP (EcoRI-0 ‡ MseI-TA). The different PCR reactions were essentially performed as described previously [7,9,10,12±15]. The oligonucleotides, their DNA sequences and the PCR annealing temperatures (Tann) are listed in Table 1. The primers used for typing were ¯uorescently labeled with Cy-5 during manufacture (MWG-Biotech AG, Ebersberg, Germany), except for the adapter oligonucleotides and the MseI-TA primer employed in AFLP reactions. PCR products were detected by separation of a 1.2-mL portion on an ALF Express DNA Sequencer (Pharmacia Biotech, Freiburg, Germany), as described previously [18]. Fingerprints in the range of the molecular sizes given in Table 1 were analyzed by means of GelCompar software (Applied Maths BV, Kortrijk, Belgium). After conversion, normalization by use of external (100-bp ladder) and internal (103 and 1064 bp) standards, as well as background subtraction, the degree of similarity between ®ngerprints was calculated with the Pearson product moment correlation coef®cient. Different genotypes were distinguished at a given level of 70% Pearson correlation. Cluster analysis was performed with the unweighted pair-group method using arithmetic averages (UPGMA). Fluorescent amplified fragment length polymorphism Typing of E. coli by AFLP was performed according to a previously described protocol, except that

ß 2003 Copyright by the European Society of Clinical Microbiology and Infectious Diseases, CMI, 9, 823±831

Jonas et al Genotyping of E. coli by means of PCR methods 825 Table 1 Oligonucleotides used for PCR-based typing Oligonucleotide

Sequence in 50 to 30 direction

ERIC2 M13 DAF4 1247 IS3A IS3B EcoRI-0 MseI-TA EcoRI-Adp1b EcoRI-Adp2b MseIAdp1b MseIAdp2b


PCR annealing temperature

Zones used for fingerprint analysis

25 50 45 36 35

120±900 120±900 120±900 120±900 120±900

8C 8C 8C 8C 8C

66±56 8Ca

bp bp bp bp bp

120±600 bp


Stepwise decrease of annealing temperature during the first 11 cycles. Oligonucleotides used as adapters for the ligation reaction in the AFLP method.


Cy-5 ¯uorescently labeled EcoRI-0 primers were used for DNA ampli®cation [15]. Puri®ed DNA was prepared by means of the Qiagen Blood Kit (Qiagen, Hilden, Germany). A 1-mL portion of DNA was restricted with 5 U of EcoRI (Roche Molecular Biochemicals, Mannheim, Germany) and MseI (New England Biolabs) in a volume of 12.5 mL for 2 h at 37 8C. After heat inactivation at 70 8C for 15 min, the sample was chilled on ice. Then the same volume containing 1 U of T4 DNA ligase (Roche), 2.5 mL of 10  ligation buffer, 2.5 pmol of EcoRI adapter and 25 pmol of MseI adapter was added for ligation. After incubation for 2 h at 20 8C, one 10-fold diluted microliter of the ligation reaction was ampli®ed in a 20-mL PCR reaction mixture containing 13.3 pmol of the EcoRI-0 primer and 80 pmol of the MseI-TA primer. The PCR comprised 11 cycles of 20 s at 94 8C, with a stepwise decreasing Tann from 66 to 56 8C for 30 s and an extension step for 2 min at 72 8C, followed by 24 cycles at the lower Tann, and a ®nal extension step at 60 8C for 30 min. Detection of the AFLP products was performed by automated laser ¯uorescence analysis, as described above. AFLP ®ngerprint data were converted to TIFF ®les with the Alf2tiff software tool (Pharmacia Biotech) and processed with the GelCompar software. Following conversion, normalization employing external (100-bp ladder) and internal (103 plus 651 bp) standards, the background was subtracted with mathematical algorithms. Designation of bands to the primarily densitometric curves was done with the novel GelCompar II `uncertain band' software tool, whereby missing

and uncertain bands are not penalized in the pairwise comparisons of the ®ngerprints. Levels of similarity between ®ngerprints were calculated with the Pearson product±moment correlation coef®cient, and the Dice coef®cient, including or ignoring regions of uncertain bands. Cluster analysis was performed with the UPGMA. The use of clustering algorithms for grouping of similar ®ngerprints poses the risk of a relationship between two ®ngerprints being overlooked in some cases, due to lower cophenetic values. In particular, this may occur when many ®ngerprints have to be compared. Therefore, in the study presented here, every ®ngerprint was compared with all the other stored ®ngerprints obtained from isolates from the same ICU. Finally, ®ngerprints with Dice coef®cients of >90% were analyzed visually for the number of fragment differences. RESULTS RAPD and rep-PCR The discriminative abilities of the different PCR methods were analyzed in order to distinguish 18 non-associated strains displaying 17 distinct PFGE patterns (data not shown). None of the RAPD and rep-PCR protocols was able to distinguish all of the epidemiologically unrelated strains (Table 2). The most discriminative approach was IS3 rep-PCR. Even after combination of the IS3 PCR results with the other four RAPD or rep-PCR results, just 14 genotypes were identi®ed (data not shown). In three of ®ve investigated protocols, reproducibility experiments with three strains analyzed in three different experiments demonstrated an

ß 2003 Copyright by the European Society of Clinical Microbiology and Infectious Diseases, CMI, 9, 823±831

826 Clinical Microbiology and Infection, Volume 9 Number 8, August 2003 Table 2 Discrimination and reproducibility of RAPD- and rep-PCR-based methods: typing results of 18 non-associated strains with 17 different PFGE patterns were analyzed PCR primers



DAF4 M13 1247 ERIC-2 IS3A/IS3B

9 6 6 9 10

80% ND 65% 76% 85%

ND, not determined.

interassay mean of similarity values of identical strain above 70% of Pearson correlation, which was chosen as the minimum value of similarity between identical genotypes. IS3 PCR achieved the highest reproducibility. AFLP serial reproducibility Since none of the less complex PCR methods discriminated the non-associated strains satisfactorily, in a further attempt an AFLP protocol was investigated. In order to test the serial reproducibility, DNA ®ngerprints of one strain were generated in six independent experiments. The visual comparison demonstrated no differences in the banding pattern, with the exception of one experiment, in which two de®nite fragments of about 600 bp and 450 bp were missing (Figure 1). Software-based comparison of AFLP patterns Reliable software-based recognition of highly similar or identical DNA ®ngerprints is a prerequisite

Figure 1 Fluorigram view of the AFLP banding patterns of type strain ATCC 11775 from six independent experiments. Arrows denote visually missing `certain' bands.

Figure 2 Computer-based comparison of the ®rst AFLP ®ngerprint with the subsequent ®ve reproducibility experiments (see Figure 1). Percentages of similarity were calculated as Pearson correlation (black bars), Dice coef®cients including all bands (hatched bars), and Dice coef®cients ignoring uncertain bands (empty bars).

for con®ning at least the number of similar ®ngerprints, which need to be con®rmed as belonging to the same genotype by visual assessment. However, in the study presented here, automated software-based comparison of these visually nearly identical AFLP ®ngerprints corresponded to lower similarity values, depending upon the kind of calculation method used. Comparison by means of Pearson correlation revealed the lowest concordance with visual comparison (Figure 2). After designation of bands, the calculated Dice coef®cient resulted in higher similarity values. Common uncertainties in the analysis of PCR ®ngerprints are small peaks just exceeding the background noise in the densitogram view, or arguable bands in the ¯uorigram view. This problem can be avoided by employing the `uncertain band' tool available on the GelCompar II software, because weak bands can be designated as uncertain bands. This allows the comparison of two banding patterns, whereby ambiguous uncertain bands are ignored and ®ngerprints are considered to be identical if they differ only in uncertain bands. Most reliable computer-aided comparisons of AFLP patterns from consecutive typing experiments were achieved by means of a band-based analysis with the `uncertain band' tool of the GelCompar II software (Figure 2). AFLP fragment differences in clinically associated isolates There were 16 of 93 patients from whom more than one isolate (n ˆ 35) was obtained during their

ß 2003 Copyright by the European Society of Clinical Microbiology and Infectious Diseases, CMI, 9, 823±831

Jonas et al Genotyping of E. coli by means of PCR methods 827

Table 3 Genotyping of 23 consecutive E. coli isolates from 11 patients by AFLP



Number of days after the first isolate

Number of fragment differences


U, U U, B TS, U U, B U, TS, B TS, B U, TS TS, US U, B iop, U env, TS

0 6 27 1 9, 14 68 26 5 2 0 5

0 0 2 0 0, 0 0 0 0 0 1 0

U, catheter urine; B, blood culture; TS, tracheal secretion; iop, taken under surgery; env, environmental sample from the patient's ventilator device.

hospital stay. These 35 isolates were mostly obtained from various body sites or at different times. Comparison of the primary isolates with the consecutive isolates from the same patients by employment of the two different typing methods, AFLP and IS3 PCR, revealed a highly similar pattern in 12 of the consecutive strains from 11 patients (Table 3). In the majority of cases, no fragment differences were seen, even when the second isolate was obtained 68 days after the ®rst (patient VI). Two fragment differences were noted in just one case of an isolate collected 27 days after the ®rst specimen (patient III). Therefore, in the following analysis, isolates with more than two fragment differences were assigned to different genotypes.

most discriminative PCR genotyping procedure in this study. Use of AFLP in a transmission study Finding two different isolates belonging to the same type is a prerequisite for de®ning a transmission event, but not suf®cient. Therefore, in this study, no attempt was made to determine the number of transmission events, as this would have gone beyond the scope of this methodology evaluation study. During a nine-month period, 112 isolates were obtained from 93 patients, hospitalized in ®ve ICUs (Table 4). Isolates were assigned ICU-based to genotypes. The number of types determined by means of AFLP exceeded the number of distinct IS3 PCR types. In this study, the latter did not distinguish any more types with identical AFLP patterns, but only con®rmed them. In order to give an indication of the discriminatory power of this typing method, the Simpson's index of diversity (D) was calculated on the basis of the AFLP genotypes. Isolates with indistinguishable types collected from different patients were obtained in varying numbers from the ®ve ICUs.

AFLP discrimination According to this rule, all of the 18 epidemiologically non-associated strains that had been investigated by RAPD and rep-PCR could be distinguished, with the results shown in Table 2. In most cases, comparison of AFLP patterns revealed more than ®ve, and at least three, fragment differences. This proved the AFLP approach to be the Table 4 Genotyping of 112 E. coli isolates from patients in different ICUs during a 9-month period



Identical types (isolated from different patients)


11 34 42 18 7

1 3 5 0 2

(3) (6) (12) (1)

Genotypes IS3


Simpson's index of diversity (D)

7 19 21 15 5

8 29 26 18 5

0.89 0.99 0.96 1.00 0.90

ß 2003 Copyright by the European Society of Clinical Microbiology and Infectious Diseases, CMI, 9, 823±831

828 Clinical Microbiology and Infection, Volume 9 Number 8, August 2003

DISCUSSION The objective of this study was to evaluate different PCR methods that have already been used in typing studies of E. coli. The investigation of a larger number of isolates in an ongoing transmission study requires a method that is both ef®cient and discriminative, and also meets the demands of a `library typing system', such as serial reproducibility [19]. Traditionally, E. coli isolates have been characterized by serotyping with the O-, H- and K-antigens [20]. However, not all isolates are typable, e.g. rough or non-motile isolates. Meanwhile, restriction fragment length polymorphism (RFLP) of the ¯iC gene and the O-antigen gene clusters has been proposed as an alternative molecular approach for typing the O- and H-antigens [21,22]. Moreover, there are strains within a serotype, which cannot be distinguished [23±25]. In recent years, traditional methods such as phage typing and serotyping have been succeeded by molecular techniques. To date, PFGE has been found to be the most effective molecular typing method available for the subdivision of serotypes of E. coli [26±28]. PCR-based typing is an effective approach in the epidemiologic investigation of various Gramnegative bacilli and an alternative to PFGE. Automated laser ¯uorescence analysis has been proposed as a time-saving method of examining PCR ®ngerprints [29,30]. Essentially, this means a separation of the PCR products with high resolution in automated DNA sequencing equipment. Irregularities during electrophoresis of the primary digital data are corrected electronically by use of internal and external standards. These normalized curve patterns can be compared pairwise by employing the Pearson correlation, i.e. curvebased without decisions being taken on designation of bands. Eventually, isolates with similarities above 70±80% are assigned to the same genotypes [18,30,31]. However, clinical isolates of the genetically clonal species E. coli cannot be suf®ciently discriminated by simple PCR approaches. The E. coli species has a comparatively clonal population structure, which can make the distinction of different isolates more dif®cult [32]. This view was supported in this study by the typing results obtained when 18 non-related strains were used to examine ®ve different RAPD

and rep-PCR protocols. None of them distinguished all strains at a given level of similarity value of 70% Pearson correlation. Owing to the limited interassay reproducibility, higher threshold values could not be employed to increase the discrimination. Even the combination of the most discriminative IS3 PCR results with the results of other methods could not differentiate all of the 18 strains. These poor results are in contrast to other reports [12,14]. The reason for this could be the different range of fragment sizes present when analyzing the ®ngerprints by use of DNA sequencing machines, or the fact that the method was originally actually devised for enteropathogenic strains. However, our results are supported by other reports. Even after optimization, single rep-PCR methods have been described as having insuf®cient discrimination [7]. By combining up to ten different RAPD experimental results, individual isolates can be assigned to phylogenetic groups [6,32], but this would seem to be expensive and error prone. Hence, a more elaborate AFLP protocol was investigated that uses the restriction enzymes EcoRI and MseI, where evaluation of different PCR primer extensions had revealed a combination of EcoRI-0 and MseI-TA as appropriate for a reproducible differentiation of E. coli [33]. Furthermore, this protocol was successfully employed in the demonstration of the phylogeny of the strains of the ECOR collection [15]. In the study presented here, these ®ngerprinting and data analysis methods were adapted for the investigation of transmission events in hospital epidemiology. Based on AFLP genotypes, the Simpson's index of diversity of those isolates revealed values of D > 0.95, which had been supplied by different ICUs in substantial numbers (n  18). This indicates the appropriate discriminatory power of this technique. Essentially, PCR-generated ®ngerprints can be analyzed with either a band-based method, by employing the Dice coef®cient, or a curve-based method, by means of Pearson product±moment correlation. The latter approach is simpler, because it is possible to avoid the tedious assignment of bands to weak, questionable ampli®cation products, which is a common problem with PCRgenerated ®ngerprints. If the densitograms have the same pattern of dips and rises, unequal

ß 2003 Copyright by the European Society of Clinical Microbiology and Infectious Diseases, CMI, 9, 823±831

Jonas et al Genotyping of E. coli by means of PCR methods 829

amounts of the ampli®cation products will not have an effect on this similarity value [34]. However, it is affected by differences in the background signals [3]. This work has demonstrated that the Pearson correlation method is not suitable for the reliable retrieval of AFLP ®ngerprint data, even when the same type strain has been repeatedly analyzed in consecutive experiments. Low reproducibility of the ®ngerprints or insuf®cient compensation for irregularities during electrophoresis by the process of data normalization could be excluded as the primary reason, because this should also have worsened the results of the Dice-based analysis. Beside minute differences in the background, a possible explanation for the lower correspondence of the Pearson-based to the visual results might be the fact that the product±moment correlation depends upon the ranking order of the peak heights of all the fragments. For instance, two patterns with a fragment of the same size will not reveal 100% correlation if the corresponding peaks do not have the same position in the ranking order of the peak heights among all the remaining peaks in both patterns. Furthermore, AFLP ®ngerprinting generates complex fragment patterns with few characteristic, yet highly reproducible, differences. Since the number of fragment differences between distinct patterns can be low compared to the total number of fragments, results of commonly used curvebased analyses were inferior to simple visual comparisons. However, a software-based ®ngerprint comparison becomes indispensable with a larger number of isolates. GelCompar II `uncertain band' tool software offers a possible alternative, in avoiding the limitations of curve-based or simple band-based comparisons in AFLP analysis. In the search for a preliminary working de®nition, 23 clinically associated ®ngerprints of patients were compared. These AFLP ®ngerprints of consecutive isolates from the same patients differed by a maximum of two fragments. The assumption that these consecutive isolates belong to the same genotype is based on highly similar AFLP and IS3 ®ngerprints. Both methods have two entirely different genetic bases; that is, ®rst, an RFLP is probed throughout the whole chromosome, and second, spacer fragments are demonstrated between neighboring insertion elements. This avoids the circular argument when the cut-off

value for associated isolates is de®ned, because it seems highly improbable that consecutive isolates with similar ®ngerprints from the same patient were simply indistinguishable because the two methods employed simultaneously displayed a low discriminatory capacity. In conclusion, AFLP and IS3 rep-PCR offer a feasible approach for typing E. coli within two days. A band-based comparison employing GelCompar II `uncertain band' tool software offers an alternative approach to an analysis by means of Pearson correlation. This type of band-based analysis of AFLP ®ngerprints offered high serial reproducibility and superior retrieval of identical genotypes, which could be con®rmed by visual examination. AFLP seems to be the best discriminative and reproducible PCR typing method, and was employed here for the ®rst time in hospital epidemiology. ACKNOWLEDGMENTS Data were presented in part at the 53rd conference of the Deutsche Gesellschaft fuÈr Hygiene und Mikrobiologie on 1±4 October 2001, Aachen, Germany. We wish to thank Ms Beck and Ms Nestler for their help in collecting the isolates and the basic identi®cation data, as well as Deborah LawrieBlum for assistance with the manuscript. This study was supported by a grant from the German Bundesministerium fuÈr Bildung und Forschung (01 KI 9907). REFERENCES 1. Olive DM, Bean P. Principles and applications of methods for DNA-based typing of microbial organisms. J Clin Microbiol 1999; 37: 1661±9. 2. Janssen PJD. Selective restriction fragment amplification by AFLP. In: Dijkshoorn L, Towner KJ, Struelens M, eds. New approaches for the generation and analysis of microbial typing data. Amsterdam: Elsevier Science, 2001: 177±210. 3. Towner K, Grundmann H. Generation and analysis of RAPD fingerprinting profiles. In: Dijkshoorn L, Towner KJ, Struelens M, eds. New approaches for the generation and analysis of microbial typing data. Amsterdam: Elsevier, 2001: 135±57. 4. Oethinger M, Conrad S, Kaifel K et al. Molecular epidemiology of fluoroquinolone-resistant Escherichia coli bloodstream isolates from patients admitted to European cancer centers. Antimicrob Agents Chemother 1996; 40: 387±92.

ß 2003 Copyright by the European Society of Clinical Microbiology and Infectious Diseases, CMI, 9, 823±831

830 Clinical Microbiology and Infection, Volume 9 Number 8, August 2003 5. Perrin M, Donnio PY, Heurtin-Lecorre C, Travert MF, Avril JL. Comparative antimicrobial resistance and genomic diversity of Escherichia coli isolated from urinary tract infections in the community and in hospitals. J Hosp Infect 1999; 41: 273±9. 6. Hilali F, Ruimy R, Saulnier P et al. Prevalence of virulence genes and clonality in Escherichia coli strains that cause bacteremia in cancer patients. Infect Immun 2000; 68: 3983±9. 7. Johnson JR, O'Bryan TT. Improved repetitive-element PCR fingerprinting for resolving pathogenic and nonpathogenic phylogenetic groups within Escherichia coli. Clin Diagn Lab Immunol 2000; 7: 265±73. 8. Chetchotisakd P, Phelps CL, Hartstein AI. Assessment of bacterial cross-transmission as a cause of infections in patients in intensive care units. Clin Infect Dis 1994; 18: 929±37. 9. Grundmann H, Hahn A, Ehrenstein B, Geiger K, Just H, Daschner FD. Detection of cross-transmission of multiresistant Gram-negative bacilli and Staphylococcus aureus in adult intensive care units by routine typing of clinical isolates. Clin Microbiol Infect 1999; 5: 355±63. 10. Webster CA, Towner KJ. Use of RAPD-ALF analysis for investigating the frequency of bacterial cross-transmission in an adult intensive care unit. J Hosp Infect 2000; 44: 254±60. 11. Baerwolff S, Gastmeier P, Geffers C. Correlation between transmissions and the nosocomial infection (NI) rate in five different intensive care units (ICUs) in a German university hospital [abstract 52471]. In: Program and Abstracts of the Annual Meeting of the Society for Healthcare Epidemiology of America, Salt Lake City. Mt. Royal, NJ: Society for Healthcare Epidemiology of America, 2002: 93. 12. Vogel L, Van Oorschot E, Maas HM, Minderhoud B, Dijkshoorn L. Epidemiologic typing of Escherichia coli using RAPD analysis, ribotyping and serotyping. Clin Microbiol Infect 2000; 6: 82±7. 13. Wang G, Whittam TS, Berg CM, Berg DE. RAPD (arbitrary primer) PCR is more sensitive than multilocus enzyme electrophoresis for distinguishing related bacterial strains. Nucleic Acids Res 1993; 21: 5930±3. 14. Thompson CJ, Daly C, Barrett TJ, Getchell JP, Gilchrist MJ, Loeffelholz MJ. Insertion element IS3-based PCR method for subtyping Escherichia coli O157: H7. J Clin Microbiol 1998 ; 36: 1180±4. 15. Arnold C, Metherell L, Willshaw G, Maggs A, Stanley J. Predictive fluorescent amplified-fragment length polymorphism analysis of Escherichia coli: high-resolution typing method with phylogenetic significance. J Clin Microbiol 1999; 37: 1274±9. 16. Frank U, Jonas D, Lupke T et al. Antimicrobial susceptibility among nosocomial pathogens isolated in intensive care units in Germany. National Reference Center Study Group on Antimicrobial




20. 21.

22. 23.







Resistance. Eur J Clin Microbiol Infect Dis 2000; 19: 888±91. Pfaller MA, Hollis RJ, Sader HS. PFGE of chromosomal DNA. In: Isenberg HD, ed. Clinical microbiological procedures handbook. Washington, DC: American Society for Microbiology, 1992: 1±12. Grundmann HJ, Towner KJ, Dijkshoorn L et al. Multicenter study using standardized protocols and reagents for evaluation of reproducibility of PCR-based fingerprinting of Acinetobacter spp. J Clin Microbiol 1997; 35: 3071±7. Struelens MJ, De Gheldre Y, Deplano A. Comparative and library epidemiological typing systems: outbreak investigations versus surveillance systems. Infect Control Hosp Epidemiol 1998; 19: 565±9. Orskov I, Orskov F, Jann B, Jann K. Serology, chemistry, and genetics of O and K antigens of Escherichia coli. Bacteriol Rev 1977; 41: 667±710. Coimbra RS, Grimont F, Lenormand P, Burguiere P, Beutin L, Grimont PA. Identification of Escherichia coli O-serogroups by restriction of the amplified O-antigen gene cluster (rfb-RFLP). Res Microbiol 2000; 151: 639±54. Machado J, Grimont F, Grimont PA. Identification of Escherichia coli flagellar types by restriction of the amplified fliC gene. Res Microbiol 2000; 151: 535±46. Pacheco AB, Ferreira LC, Pichel MG, Almeida DF, Binsztein N, Viboud GI. Beyond serotypes and virulence-associated factors: detection of genetic diversity among O153:H45 CFA/I heat-stable enterotoxigenic Escherichia coli strains. J Clin Microbiol 2001; 39: 4500±5. Achtman M, Heuzenroeder M, Kusecek B et al. Clonal analysis of Escherichia coli O2:K1 isolated from diseased humans and animals. Infect Immun 1986 ; 51: 268±76. Campos LC, Whittam TS, Gomes TA, Andrade JR, Trabulsi LR. Escherichia coli serogroup O111 includes several clones of diarrheagenic strains with different virulence properties. Infect Immun 1994; 62: 3282±8. Arbeit RD, Arthur M, Dunn R, Kim C, Selander RK, Goldstein R. Resolution of recent evolutionary divergence among Escherichia coli from related lineages: the application of pulsed field electrophoresis to molecular epidemiology. J Infect Dis 1990; 161: 230±5. Meng J, Zhao S, Zhao T, Doyle MP. Molecular characterisation of Escherichia coli O157:H7 isolates by pulsed-field gel electrophoresis and plasmid DNA analysis. J Med Microbiol 1995; 42: 258±63. Arbeit RD. Laboratory procedures for the epidemiologic analysis of microorganisms. In: Murray PR, Baron EJ, Pfaller MA, Tenover FC, Yolken RH, eds. Manual of clinical microbiology. Washington, DC: American Society for Microbiology, 1999: 116±37. Grundmann H, Schneider C, Daschner FD. Fluorescence-based DNA fingerprinting elucidates

ß 2003 Copyright by the European Society of Clinical Microbiology and Infectious Diseases, CMI, 9, 823±831

Jonas et al Genotyping of E. coli by means of PCR methods 831

nosocomial transmission of phenotypically variable Pseudomonas aeruginosa in intensive care units. Eur J Clin Microbiol Infect Dis 1995; 14: 1057±62. 30. Grundmann H, Schneider C, Tichy HV et al. Automated laser fluorescence analysis of randomly amplified polymorphic DNA: a rapid method for investigating nosocomial transmission of Acinetobacter baumannii. J Med Microbiol 1995; 43: 446±51. 31. Webster CA, Towner KJ, Humphreys H, Ehrenstein B, Hartung D, Grundmann H. Comparison of rapid automated laser fluorescence analysis of DNA fingerprints with four other computer-assisted approaches for studying relationships between Acinetobacter baumannii isolates. J Med Microbiol 1996; 44: 185±94.

32. Desjardins P, Picard B, Kaltenbock B, Elion J, Denamur E. Sex in Escherichia coli does not disrupt the clonal structure of the population: evidence from random amplified polymorphic DNA and restriction-fragment-length polymorphism. J Mol Evol 1995; 41: 440±8. 33. Arnold C, Metherell L, Clewley JP, Stanley J. Predictive modelling of fluorescent AFLP: a new approach to the molecular epidemiology of E. coli. Res Microbiol 1999; 150: 33±44. 34. van Ooyen A. Theoretical aspects of pattern analysis. In: Dijkshoorn L, Towner KJ, Struelens M, eds. New approaches for the generation and analysis of microbial typing data. Amsterdam: Elsevier, 2001: 31±45.

ß 2003 Copyright by the European Society of Clinical Microbiology and Infectious Diseases, CMI, 9, 823±831