ABSTRACT
Two microsatellite tandem repeated tetramers, (GACA)4 and (CAAT)4, were used for Proteus mirabilis strain differentiation. The microsatellite-based PCR tests were applied for the examination of interstrain diversity for 87 P. mirabilis strains. Forty-six of the investigated strains were clinical isolates (5 were hospital isolates and 39 were outpatient clinic isolates); 42 strains were derived from the Kauffmann-Perch collection of laboratory strains. Fingerprinting done with the tetramers had a high discrimination ability [0.992 and 0.940 for (GACA)4 and (CAAT)4, respectively]. The distributions of clinical isolates among well-defined laboratory strains, determined by numerical analysis (unweighted pair-group method with arithmetic averages; Dice similarity coefficient), proved their genetic similarity to reference strains in the Kauffmann-Perch collection. This analysis also indicated that it is possible to estimate some phenotypic properties of P. mirabilis clinical isolates solely on the basis of microsatellite fingerprinting.
Proteus spp. are mobile gram-negative bacteria common in both the natural environment and animal or human intestinal tracts. Proteus spp. are also known etiologic agents for meningitis and numerous bacteremias (8, 20-23, 43). Urinary tract infections are among the most frequent bacterial infections (19), and Proteus mirabilis strains are one of the most common causes of urinary tract infections (7%), third after Escherichia coli (52%) and Enterococcus spp.(12%) (11). Such infections occur commonly among patients with structural defects of the urinary tract (6, 38, 39). The presence of P. mirabilis rods within a urease-induced bladder stone matrix was visualized recently (24). Moreover, some results suggest a possible etiopathogenic role of P. mirabilis in rheumatoid arthritis (9, 31), and some nosocomial transmission events have been reported (31). Because of the increasing spread and clinical significance of P. mirabilis rods (13, 15, 30, 31, 32), studies of effective methods for epidemiological investigations are of great importance.
Out of the numerous types of simple sequence repeats proposed as tools for very sensitive bacterial fingerprinting (25, 27, 48, 50, 51, 54), many microsatellites have been described as being useful for microbial differentiation, especially below the level of species (1, 10, 26, 28, 33, 34, 47, 48, 49, 53).
Most of the molecular fingerprinting methods applied for the differentiation of Proteus (35, 36, 44), however, are not sensitive enough for more detailed interstrain differentiation. In particular, no specific method allowing for P. mirabilis differentiation, especially below the serotype level, has been described so far.
In this study, we have focused on microsatellite-based methods supplying patterns specific for particular P. mirabilis strains. The aim of the study was to verify how effective microsatellites are for P. mirabilis fingerprinting; in particular, we examined whether tandem tetramer-based PCR is applicable to Proteus strain differentiation or typing as well as the sensitivities of PCR methods based on tandem repeated tetramers. In addition, we compared the efficiencies of these methods and other important Proteus typing methods. Finally, we examined how informative these patterns are in relation to other properties of P. mirabilis strains.
Two microsatellite sequences were used for P. mirabilis laboratory strain differentiation: (GACA)4 and (CAAT)4. The studies were performed with 40 P. mirabilis strains from the serologically defined Kauffmann-Perch (23) collection and with 42 P. mirabilis clinical isolates.
MATERIALS AND METHODS
Bacterial strains. P. mirabilis laboratory strains were from the Czech Collection of Type Cultures, Institute of Microbiology and Epidemiology, Prague, Czech Republic. P. mirabilis strain S1959 was obtained from the Institute of Microbiology and Immunology, University of Łódź, Łódź, Poland. Thirty-six clinical isolates of P. mirabilis from urine were obtained from outpatient clinics in Łódź and given the prefix “ZOZ”; they were kindly supplied by Halina Skulimowska (Table 1). An additional six clinical isolates were derived from the Military Medical Academy Hospital, Łódź, Poland, and given the prefix “WAM”; they were kindly supplied by Maria Kowalska (Table 1).
Proteus strains examined in these studies
Bacterial culture and DNA isolation.Bacteria were cultivated in 3 ml of Luria-Bertani (LB) medium for 12 h at 37°C. Then, 1 ml of the culture was centrifuged for 3 min. The pellet was resuspended in 100 μl of Tris-EDTA buffer. After 30 min of incubation with 10 μl of proteinase K solution (20 mg/ml) at 37°C, chromosomal DNA was isolated with a genomic DNA isolation kit (A&A Biotechnology, Gdansk, Poland) and then dissolved in 200 μl of Tris (pH 8.2). DNA samples were kept at −20°C until PCR was performed.
The amount of isolated DNA was verified with a UV spectrophotometer (Ultraspec 2000; Pharmacia LKB) at 260 nm and by electrophoresis in 2% agarose (Serva; analytical grade) in 0.04 M Tris-acetate-1 mM EDTA buffer (pH 7.8).
Primers.Sequences of oligonucleotides (synthesized by Ransom Hill Bioscience Inc., Ramona Calif.) for genomic DNA analysis were as follows: (GACA)4, 5′-GACAGACAGACAGACA-3′ (16 nucleotides) (26), and (CAAT)4, 5′-CAATCAATCAATCAAT-3′ (16 nucleotides) (33).
Primer target site computer analysis.The presence of the doubled tetramer repeats (GACA)4 and (CAAT)4 in the known part of the Proteus genome was confirmed. The National Center for Biotechnology Information GenBank database was explored for the presence of P. mirabilis DNA sequences. This search was followed by a search for tandem repeated tetramers among the 84 P. mirabilis sequences found. For this analysis, two programs, Quicksearch and Nesearch, from the PC/Gene packet, were used.
PCR conditions.A master mixture, the same for each reaction type, contained all reagents except for genomic DNA. Each thin-walled vial (MJ Research) contained 25 μl of reaction solution, which consisted of 2.5 mM deoxynucleotide triphosphate mixture (dATP, dTTP, dCTP, and dGTP; TaKaRa), 2.5 μl of 10× reaction buffer, 100 pM primer, 1 U of thermostable polymerase (Dynazyme; Finnzyme), 18MΩ ultrapure water (Millipore), and 20 ng of template DNA. Amplification was carried out with a UNO II thermocycler (Biometra). An initial 7 min of denaturation at 95°C was followed by 32 cycles of annealing (40°C for 1 min), extension (65°C for 1 min), and denaturation (92°C for 30 s). The reaction was completed by 16 min of extension at 65°C. During gel electrophoresis in an MGU 602T unit (CBS Scientific), aliquots of amplification products (5 μl) were resolved against molecular weight markers (Ideal, Poland, Gdansk) in 2% agarose (Serva; analytical grade). The gels were stained with ethidium bromide solution and photographed with the aid of a BioDoc system (Biometra) and the computer program SM Camera (FAST Multimedia AG 1993, version1.1).
Electrophoretic pattern analysis.For investigation of PCR product diversity, computer-assisted pattern analysis was carried out (GelCompar, version 4.0; Applied Maths, Kortijk, Belgium). The bands chosen for the analysis were selected manually from the hard-copy photograph and the densitometric curves of the appropriate electrophoretic paths. For all electropherograms, the same background subtraction procedure (the rolling-disk procedure, as suggested by the program authors) was used. Normalization procedures included the internal and external reference band sets as shown in Fig. 1. Electrophoretic patterns were normalized to common internal sets of amplicons next to the external molecular weight markers placed on the peripheral paths of both sides of the gels. The correlations among the investigated species were based on the electrophoretic band distribution and the Pearson product-moment correlation coefficient. For comparison of the electrophoretic patterns and determination of their similarities (construction of dendrograms), the UPGMA (unweighted pair-group method with arithmetic averages) clustering algorithm was used (45). For the band pattern analysis, the Dice similarity coefficient was used.
Electrophoretic resolution by (GACA)4 PCR of six representative P. mirabilis strains derived from the Kauffmann-Perch collection. The fingerprinting procedure accuracy was tested three times for each strain. Lanes: 1 to 3, PrK 75/57; 4 to 6, PrK 66/57; 7 to 9, PrK 62/57; 10 to 12, PrK 34/57; 13 to 15, PrK 18/57; and 16 to 18, PrK 15/57. The repeated procedures included culturing, DNA extraction, and amplification. Experiments were done with 3 ml of inoculum (lanes 3, 6, 9, 12, 15, and 18), 1 liter of secondary inoculum (lanes 2, 5, 8, 11, 14, and 17), and 8 liters of culture (lanes 1, 4, 7, 10, 13, and 16). Lanes M, molecular weight markers. The white arrows on the left indicate bands applied as internal references for normalization procedure in the UPGMA analysis.
The levels of effectiveness of the applied fingerprinting methods were compared by using the formula described by Hunter (14):$$mathtex$$\[D{=}1{-}\left[\frac{1}{N\ (N{-}1)}\right]{{\sum}_{j\ {=}\ 1}^{N}}\ a_{j}\]$$mathtex$$where N is the total number of investigated strains, aj is the number of indistinguishable strains in the experiment, and D is the discrimination power.
RESULTS
Several strains were arbitrarily chosen to examine the reproducibility of the typing method and to test the stability of the bacterial strains. The reproducibility of PCR patterns was confirmed in two series of reactions performed on a DNA matrix isolated from six P. mirabilis laboratory strains, PrK 15/57 (O7), PrK 18/57 (O9), PrK 34/57 (O38), PrK 62/57 (O36), PrK 66/57 (O40), PrK 75/57 (O49), with the (GACA)4 primer and on a DNA matrix isolated from nine P. mirabilis strains, PrK 15/57 (O7), PrK 18/57 (O9), PrK 34/57 (O38), PrK 38/57 (O20), PrK 62/57 (O36), PrK 66/57 (O40), PrK 75/57 (O49), PrO 10/52 (O3), and S1959 (O3), with the (CAAT)4 primer. Each of the strains was isolated three times to check genetic stability. Electrophoresis performed on one gel only with 18 probes (six triplets) resulted in 100% reproducibility (i.e., identical strains produced identical outputs) (Fig. 1 and 2). For the (GACA)4 PCR, a lack of full homology was observed within the repeats of the same-strain analysis (Fig. 3, series designed A, B, and C). The inclusion of several paths from another gel resulted in a total 6.7% difference in the same-strain analysis (Fig. 3, paths not described). Hence, the cutoff was set at an arbitrary value of 90% interpattern homology. The weaker reproducibility in the last test resulted from the larger number of similar amplicons in particular paths as well as from the higher background smear intensity.
Accuracy of (CAAT)4 PCR fingerprinting for six representative P. mirabilis strains derived from the Kauffmann-Perch collection. Lanes: 1 to 3, PrK 66/57; 4 to 6, PrK 62/57; 7 to 9, PrK 38/57; 10 to 12, PrK 34/57; 13 to 15, PrK 18/57; and 16 to 18, PrK 15/57. The testing procedures included culturing, DNA extraction, and amplification. See the legend to Fig. 1 for a further description of lanes. Lanes M, molecular weight markers. The white arrows on the left indicate bands used as internal reference standards for normalization in the UPGMA analysis.
Reproducibility of (GACA)4 PCR (left) and (CAAT)4 PCR (right) in UPGMA band pattern analysis (the Dice similarity coefficient was used). An 0.8% position tolerance value was used. The calculation program GelCompar, version 4.0, was used. Three probes of each strain (A, B, and C) resolved on the same gel and one probe of some of the same strains from another gel (not labeled) are compared. The scales represent the level of homology between the investigated probes.
For the set of six P. mirabilis laboratory strains, the differentiation indices were very high and equal (0.966) with both primers. When more electropherograms were investigated simultaneously (38 and 89), the appropriate differentiation indices were 0.992 with the (GACA)4 primer and 0.940 with the (CAAT)4 primer when a cutoff value of 90% was used and 0.954 with the (CAAT)4 primer when a cutoff value of 93% was used, respectively (Table 2; see also Fig. 5). For further analysis of P. mirabilis clinical isolates, only the (GACA)4 primer was used.
Differentiation indices (DI) for two investigated fingerprinting methods based on tandem repeated tetramersa
The patterns obtained for the clinical isolates with the (GACA)4 primer were very similar to those obtained for the laboratory strains. Five bands common to most of the patterns were located at about 2,555, 1,241 (two bands), and 700 bp. Three shorter bands (two pairs separated by a single band) were located at about 500 to 100 bp (Fig. 6). These patterns were similar to the patterns obtained with the (CAAT)4 primer, where five bands positioned between 1,827 and 900 bp (Fig. 4) were common to most of the investigated strains. Most of the strain-specific bands were shorter than 900 bp. Numerical analysis of laboratory strains resulted in no evident clusters, and the dendrogram had the “stair-shape” structure (Fig. 5). The observed interstrain homologies were mostly in the range of about 60 to 95%, similar to those obtained with the dendrogram produced by (GACA)4-based fingerprinting. The patterns resulting from the (CAAT)4 test were more distinct, although some of the strains were still indistinguishable from others, making the interpretation of results difficult (Fig. 5).
Electrophoretic resolution by (CAAT)4 PCR of representative P. mirabilis strains derived from the Kauffmann-Perch collection. Lanes: 1, PrK 74/57 (O48); 2, PrK 69/57 (O43); 3, PrK 64/57 (O38); 4, PrK 61/57 (O35); 5, PrK 58/57 (O32); 6, PrK 56/57 (O31) (P. vulgaris); 7, PrK 53/57 (O30); 8, PrK 52/57 (O29); 9, PrK 51/57 (O28); 10, PrK 50/57 (O27); 11, PrK 49/57 (O26); 12, PrK 47/57 (O24); 13, PrK 46/57 (O24); 14, PrK 45/57 (O24); 15, PrK 43/57 (O23); 16, PrK 41/57 (O23); 17, PrK 38/57 (O20); and 18, PrK 34/57 (O18). Lanes M, molecular weight markers.
UPGMA (Dice) cluster analysis of 35 P. mirabilis laboratory strains by (CAAT)4 PCR. The calculated values for clustering errors are boxed. The scale at the left represents the homology level. The broken line indicates the accuracy of the method. The arrows indicate the main groups of indistinguishable strains for the cutoff value established at 90% similarity.
Representative electrophoretic patterns of (GACA)4 PCR for out-of-clinic isolates of P. mirabilis. Lanes: 1, ZOZ 63b; 2, ZOZ 42; 3, ZOZ 203; 4, ZOZ 216; 5, ZOZ 256; 6, ZOZ 58; 7, ZOZ 352; 8, ZOZ 304; 9, ZOZ 303; 10, ZOZ 302; 11, ZOZ 87; 12, ZOZ 253; 13, ZOZ 191; 14, ZOZ 72; 15, ZOZ 14; 16, ZOZ 13; 17, ZOZ 19; 18, ZOZ 630; 19, ZOZ 670; 20, ZOZ 168; 21, ZOZ 173; 22, ZOZ 367; and 23, ZOZ 105. Lanes M, molecular weight markers.
DISCUSSION
The studies reported here showed that tandem repeated tetramers might be used as effective tools for PCR-based P. mirabilis typing. The results of the cluster analysis partially corresponded to the distributions of important bacterial surface properties.
The levels of effectiveness of the most important Proteus typing methods were compared recently by Senior (42, 43) using phage typing, bacteriocin typing, and Dienes phenomenon-based incompatibility grouping. It was demonstrated that, when a combination of different typing methods is applied, highly sensitive differentiation may be obtained. Pignato et al. (36) found that with two ribotyping methods, 10 examined P. mirabilis strains were clustered together in one ribogroup. The effectiveness of ribotyping for P. mirabilis was compared with those of other genetic differentiation methods by Pfaller et al. (35). The discriminatory indices established were relatively high: 0.92 for ribotyping, 0.979 for PFGE, and 0.980 for the Dienes test. The sensitivity of molecular typing methods used to determine electrophoretic profiles of outer membrane or total cell proteins of Proteus strains (17) is much more limited.
PCR-based differentiation methods that have been evaluated include randomly amplified polymorphic DNA (RAPD) analysis and repetitive sequence-based PCR. For some Proteus species, repetitive sequences can be use as an effective tool for fingerprinting (18, 44). For P. mirabilis, however, the repetitive sequence-based PCR fingerprinting methods described so far (ERIC-PCR, BOXA1R-PCR, BOXA2R-PCR) have a lower sensitivity and only REP-PCR supports greater efficiency (44). Microsatellite-based PCR methods usually have a higher sensitivity. Nevertheless, for P. mirabilis, the discrimination ability of (GTG)5 microsatellite-based PCR analysis is much lower (data not shown).
Both tandem repeated tetramers are short enough for the presented fingerprints to resemble RAPD types (2, 29). The resolution of the RAPD method was found to be identical to that of ribotyping, since both generated the same numbers of different electrophoretic profiles for the investigated P. mirabilis strains (36). Therefore, the microsatellite-based fingerprinting method described here, with high discriminatory abilities [0.992 with the (GACA)4 primer and 0.940 with the (CAAT)4 primer], seems to be promising. Numerical analysis of the total number of investigated strains showed that most clinical isolates were interspersed among laboratory reference strains.
For some species, the correspondence between strain phenotypes and fingerprinting patterns based on sequences not directly related to surface antigens has already been reported. The application of ribotyping techniques to Listeria monocytogenes has demonstrated a significant relationship between serotypes and genetic lineages (C. Nadon, D. Woodward, C. Young, F. Rodgers, and M. Wiedmann, Abstr. 100th Gen. Meet. Am. Soc. Microbiol., abstr. P-96, 2000). The efficacy of phenotypic feature differentiation for some Abiotrophia strains on the basis of several genotyping methods has also been established (16). Also, for uropathogenic Escherichia coli strains, ribotype or RAPD-generated profiles consistent with their common clonal origins have been detected. However, interstrain serological similarities may originate from convergent evolution or recombination events (37) and therefore do not reflect common clonal origins.
The small series of P. mirabilis strains were genetically stable in the genome loci including both types of applied sequences: (GACA)4 and (CAAT)4. Moreover, our preliminary results indicate that there is some correspondence between microsatellite and repetitive sequence genomic distributions and phenotypic properties of Proteus laboratory strains (4, 5). These results correspond to some other findings indicating that a large number of P. mirabilis properties are correlated with, e.g., serological identity and proticine production and sensitivity type (7). Therefore, as one may expect, a considerable portion of genome sequences should be ordered steadily (i.e., the general order of genes and/or operons should be to be conserved). If this notion is true, then estimation of one set of bacterial properties based on a statistical distribution of others is well grounded. In particular, microsatellite-based PCR fingerprinting may then serve for the estimation of probable serological properties of clinical isolates or the selection of clinical isolates which are more strongly associated with uropathogenicity. It has been shown that some serological types of P. mirabilis are more frequent (8, 20-23, 41). The question of whether this association really results from general P. mirabilis genome stability or from local linkage disequilibria might be answered only by investigations based on a larger number of subtracted sequence types, i.e., additional types.
In conclusion, tetramer-based PCR fingerprinting results are independent of strain storage and culturing, and the method used here shows efficient discrimination ability. The presented findings suggest that at least some microsatellite-based fingerprints are specific enough for the rapid differentiation of P. mirabilis strains at the level of serogroup and for the effective differentiating of particular clones. Moreover, dendrogram structures determined independently of the applied calculation algorithm (40, 42, 52) include additional information which may be interpreted in relation to important surface properties. Our results are in agreement with reports of correlations between known surface antigens (of different bacteria) and fingerprinting results based on numerous, distinct sequence types (ribotyping, restriction fragment length polymorphism analysis, RAPD analysis, and pulsed-field gel electrophoresis) and, along with previous findings, suggest similar correlations with several PCR-based fingerprinting methods.
ACKNOWLEDGMENTS
The work was partially supported by grant 4 P05A 092-19 from the State Committee for Scientific Research.
Special appreciation is extended to Maria Olszewska (Institute of Cytology and Plant Physiology, University of Łódź), Leon Sedlaczek (Centre of Microbiology and Virology, Polish Academy of Sciences), and Antoni Różalski (Institute of Microbiology and Immunology, University of Łódź) for encouragement of our research as well as for comments, advice, and critical discussions. We also thank Tomasz Sakowicz (Institute of Cytology and Plant Physiology, University of Łódź) for kindly reading the manuscript.
FOOTNOTES
- Received 25 July 2002.
- Returned for modification 15 October 2002.
- Accepted 7 January 2003.
- Copyright © 2003 American Society for Microbiology