INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Lineages of Escherichia coli which are overrepresented among clinical isolates from patients with urinary-tract infections (UTI) compared with the fecal flora and which exhibit virulence-associated phenotypic traits such as adhesins, toxins, and certain somatic or capsular polysaccharides are commonly designated "uropathogenic clones" (15, 18, 19, 25, 27, 35, 36, 43, 45, 47). Such virulent clones traditionally have been identified based on their O:K:H serotypes (8, 35, 36, 45), although more recently they have been defined by using multilocus enzyme electrophoresis (MLEE) and ribotyping (2, 8, 10, 18, 19, 27, 39).

Serotype O15:K52:H1, virtually unrecognized until recently, has not traditionally been regarded as one of the classic uropathogenic O:K:H serotypes, which include O6:K2:H1, O4:K12:H5, O1:K1:H7, and others (36, 45). However, O15:K52:H1 achieved notoriety in 1986 to 1987 when strains of this serotype that expressed P fimbriae, produced aerobactin, and displayed an unusual multiple antimicrobial resistance phenotype caused a large-scale epidemic of UTI, septicemia, and diverse other serious extraintestinal infections in south London, England (34, 38). The subsequent recognition of O15:K52:H1 as the second most common serotype among E. coli bacteremia isolates at a Copenhagen hospital (in which setting the organism usually arose from a urinary-tract source), together with the observation that the Copenhagen isolates exhibited the same virulence factor profile as the south London outbreak strains, provided further evidence of the pathogenic potential of E. coli O15:K52:H1 and suggested that this serotype might constitute a widespread virulent clone (33).

Reinforcing the uropathogenic-clone hypothesis for E. coli O15:K52:H1 and extending the known geographic range of this serotype to include southern Europe was the finding that in the early 1990s E. coli O15:K52:H1 (always from a urinary-tract source) accounted for 7 of 160 E. coli bacteremia isolates (4.4%) at the Hospital Mutua de Terrassa in Terrassa, Spain (9), a remarkable prevalence for an organism not generally recognized as a uropathogen. In view of these results, we performed a prospective epidemiological study at the Hospital de la Santa Creu i Sant Pau in Barcelona to determine the prevalence of this serotype as a cause of UTI and to characterize the clinical correlates of UTI due to such strains. We also investigated the virulence traits and other bacteriological characteristics of O15:K52:H1 strains from Barcelona, in comparison with those of O15:K52:H1 strains from other geographic locales and those of reference strains of E. coli.

(This work was presented in part at the 37th Interscience Conference on Antimicrobial Agents and Chemotherapy, Toronto, Ontario, Canada, 1997 [abstr. K-195]).

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Urine study. A total of 24,462 urine specimens, comprising all urine specimens submitted to the clinical microbiology laboratory of the Hospital de la Santa Creu i Sant Pau, Barcelona, Spain, from June 1994 through May 1995, were examined microscopically for epithelial cells, leukocytes, erythrocytes, and microorganisms by using both unstained centrifuged urine and Gram-stained uncentrifuged urine. Pyuria was defined as the presence of >5 leukocytes per 400× field in the sediment of 10 ml of centrifuged urine. Culture was done by inoculating 10 µl of urine onto CLED agar (Oxoid, Unipath Ltd., Basingstoke, England) and other media when suggested by the Gram stain results. After incubation at 37°C for 24 h, colonies were enumerated. Cultures were interpreted as positive or negative for UTI according to standard criteria endorsed by the American Society for Microbiology based on the colony count, the urinalysis findings, and patient-specific clinical data as provided on the request slip (16). Organisms were identified by standard methods (28). Significant E. coli isolates were serotyped as described below.

Bacteremia isolates. Blood culture was performed by using the BacT/ALERT blood culture automatic management and reading system (Organon Teknika, Turnhout, Belgium). All E. coli isolates were serotyped as described below.

Serotyping. Serotyping was done according to the methods of Orskov and Orskov (35) by using O15, K52, and H1 antisera provided by the Statens Seruminstitut (Copenhagen, Denmark). All significant E. coli strains isolated from urine and blood during the study period were tested for the O15 somatic antigen by slide and tube agglutination with O15 antiserum. Strains that were O15 positive were tested for the K52 capsular antigen by countercurrent immunoelectrophoresis and for the H1 flagellar antigen by slide and tube agglutination. All E. coli O15 strains that were negative for the K52 antigen in the initial screen were sent to the Statens Seruminstitut or the National Institute of Public Health and Environmental Protection (Bilthoven, The Netherlands) for serotyping. Nonencapsulated and nonmotile strains were designated as K and H, respectively, whereas encapsulated strains for which the K or H antigen could not be typed were designated as KN or HN. E. coli O15 strains that had the K52 and/or the H1 antigen, and no other K or H antigen, were putatively considered to be O15:K52:H1.

Clinical data. Epidemiological data were collected from all patients with community-acquired E. coli UTI during a 3-month window within the study period and from all subjects with community-acquired UTI due to E. coli O15:K52:H1 at any point during the 1-year study period. (Community-acquired UTI was defined as a UTI episode in which, at the time the index urine sample was submitted, the patient either was not hospitalized or had been hospitalized for <48 h and had not been previously hospitalized during the preceding 28 days.) The data collected included age, gender, clinical manifestations of UTI (dysuria, frequency, urgency, fever, and fever plus flank pain or tenderness, i.e., pyelonephritis), and putative predisposing factors (sexual intercourse, pregnancy, and underlying urological or medical conditions). Data collection was done by medical record review or by interviews with the subject's primary physician, in accordance with the guidelines of the hospital's institutional review board.

Other O15:K52:H1 strains and non-O15:K52:H1 controls. For comparison with the O15:K52:H1 strains isolated in the present study, 22 other O15:K52:H1 strains were collected from various geographic sources, including seven bacteremia isolates from Terrassa, Spain (9), nine UTI isolates from Lugo, Spain (generously provided by J. Blanco), one ascitic fluid isolate from Barcelona, and five isolates of unknown clinical source from London (generously provided by I. Phillips) (38). Also studied were eight E. coli strains from the 72-member ECOR reference collection of natural E. coli isolates (32), which have been assigned to four major phylogenetic groups (A, B1, B2, and D) according to their relatedness by MLEE (14). The eight strains included ECOR strains 7 (group A; O85:HN), 10 (group A; O6:H10), 24 (group A; O15:H), 30 (group B1; O113:H21), 41 (group D; O7:H), 59 (group B2; O4:H40), 62 (group B2; O2:H), and 72 (group B1; O144:H8). (O:H serotypes for the ECOR strains were as determined by R. Wilson using standard methods.) ECOR strains were obtained from the American Type Culture Collection (Manassas, Va.) as ATCC 35326, ATCC 35329, ATCC 35343, ATCC 35349, ATCC 35360, ATCC 35378, ATCC 35381, and ATCC 35391, respectively. As putative nonpathogenic controls, 15 E. coli strains with O:K serotypes not corresponding to any recognized E. coli pathogenic group (43) were obtained from the National Institute of Public Health and Environmental Protection of The Netherlands. These included strains H93-425 (O122:KN), H93-428 (O145:K), H93-435 (O23:K18), H93-443 (O39:K1), H93-465 (O64:KN), H93-466 (O14:K1), H94-27 (O132:KN), H94-106 (O73:K), H94-122 (O5:K), H94-130 (O17:K1), H94-170 (O169:KN), H95-41 (O5:KN), H95-75 (O12:K), H95-84 (O108:K), and H95-133 (O69:KN).

Antimicrobial resistance profiles. Susceptibility to 27 antimicrobial agents was determined by a disk diffusion method. The agents assayed were ampicillin, amoxicillin-clavulanic acid, ticarcillin, piperacillin, cefazolin, cefuroxime, cefoxitin, cefixime, cefotaxime, ceftazidime, aztreonam, imipenem, chloramphenicol, tetracycline, streptomycin, kanamycin, neomycin, gentamicin, tobramycin, amikacin, sulfonamide, trimethoprim, trimethoprim-sulfamethoxazole, nalidixic acid, norfloxacin, ciprofloxacin, and nitrofurantoin. Procedures and interpretive criteria were as proposed for Enterobacteriaceae by the National Committee for Clinical Laboratory Standards (30). The antibiotic resistance score was the number of agents to which a strain was found to be resistant.

Biotyping. The biotype for 12 metabolic reactions was determined according to the system proposed by Richard (40). Similarities between pairs of strains with respect to 12-digit binary biotype results were scored by the simple matching coefficient, and strains were clustered on the basis of their degree of similarity by the method of weighted average linkage by using the TAXAN program (Information Resources Group, University of Maryland, College Park).

Production of hemolysin and aerobactin. Isolates were considered alpha-hemolysin positive if they showed clearing around or beneath bacterial colonies on 5% sheep blood agar after 18 h of incubation at 37°C (3). Aerobactin production was determined by using a standard cross-feeding bioassay in minimal medium agar containing 5 g of Casamino Acids/liter, 2 g of glucose/liter, 50 mg of thiamine/liter, 20 mg of tryptophan/liter, and 160 µM 2,2'-dipyridyl (17). The aerobactin-requiring strain E. coli CCUG 29422 was used as an indicator strain, and the pColV+ aerobactin-producing strain E. coli CCUG 29423 was used as a positive control. (All CCUG control strains in the present study were obtained from the Culture Collection, University of Göteborg, Göteborg, Sweden.)

Expression of type 1 fimbriae and MRHA. Expression of type 1 fimbriae was determined, as previously described, by 5% alpha-methyl-D-mannoside-sensitive agglutination of Saccharomyces (11). Expression of P and non-P mannose-resistant (NPMR) adhesins was determined by using microscope slide assays as previously described (20, 23) with plate-grown bacteria suspended in phosphate-buffered saline (pH 7.4) (PBS) plus 5% alpha-methyl-D-mannoside as the agglutinator, 5% human A1P1 erythrocytes in PBS as the agglutination substrate, pigeon egg white as a P adhesin-specific inhibitor, and PBS and 3% (wt/vol) bovine serum albumin in PBS (BSA) as negative control inhibitors. Slides were rocked at 4°C for at least 1 min and examined both grossly and microscopically for evidence of mannose-resistant hemagglutination (MRHA), which was graded by intensity as 0 to 5+. P-pattern MRHA was defined as a 3 level reduction in MRHA intensity in the presence of pigeon egg white compared with PBS or BSA. NPMR hemagglutination (HA) was defined as a 2 level reduction in MRHA intensity in the presence of pigeon egg white.

Study of serum sensitivity. Resistance to the bactericidal effect of normal human serum (RSB) was determined by a rapid turbidimetric assay. The growth of bacteria in the presence of 35% serum was tested by spectrophotometer absorbance values (Epy System; Sorin bioMedica, Saluggia, Italy) at 0, 30, 60, 90, 120, and 180 min, as described elsewhere (37). All test strains were compared with resistant and sensitive control strains CCUG 31246 and CCUG 31251, respectively.

Multiplex PCR assays for virulence genes and papA alleles. The genes encoding uropathogenic virulence factors such as pilus associated with pyelonephritis (pap), hemolysin (hly), aerobactin (aer), cytotoxic necrotizing factor 1 (cnf1), S and F1C fimbriae (sfa), and afimbrial and other Dr-specific adhesins (afaI) were amplified as previously described by Yamamoto et al. (48). Strains C7 (pap sfa cnf1 hly) and C149 (afaI aer hly), generously provided by S. Yamamoto, were used as positive controls (48). The presence of the three papG alleles was assessed by using an allele-specific PCR assay as previously described (21, 22). Control strains included wild-type strains IA2 (allele II), U5 (allele III), and J96 (alleles I and III), and recombinant strains JJ48 (allele I), HB101/pDC1 (allele II), and P678-54/pJFK102 (allele III) (21). Selected O15:K52:H1 strains were tested for the 11 known alleles of papA (which encodes PapA, the major structural subunit and major antigenic determinant of P fimbriae) by using an allele-specific multiplex PCR assay, as previously described (J. R. Johnson, F. Scheutz, C. C. Fasching, L. van Dijk, and W. Gaastra, Abstr. Am. Soc. Antimicrob. Annu. Meet., abstr. 12200, p. 147, 1998).

Ribotyping. DNA was extracted as described by Nastasi et al. (29). Purified DNA (4 mg) was digested with BglII under the conditions recommended by the manufacturer (Pharmacia Biotech, Uppsala, Sweden). DNA fragments were separated on a 0.6% agarose gel at 30 V overnight. Restriction fragments were transferred under vacuum to nylon membranes (Hybond-N; Amersham, Little Chalfont, England) by Southern blotting. Prehybridization and hybridization were performed as described elsewhere (13). The acetylaminofluorene-labelled 16S plus 23S rRNA from E. coli (Eurogentec, Liège, Belgium) was added to the hybridization solution. The hybridization reactions were visualized colorimetrically by immunoenzymatic detection as recommended by the manufacturer. Xenorhabdus sp. strain 278 chromosomal DNA (kindly provided by P. D. Grimont, Pasteur Institute, Paris, France) digested with EcoRI was used on each gel as a molecular size standard. Ribotyping patterns were compared with the Bio Image system (Genomic Solutions Inc., Ann Arbor, Mich.), and isolates were considered to be identical if there was complete concordance of profiles. Similarities in ribotyping patterns between two strains were scored by the Dice coefficient, and strains were clustered on the basis of their degree of similarity by the unweighted pair-group mean arithmetic method (UPGMA).

PCR fingerprinting. Selected strains were subjected to randomly amplified polymorphic DNA (RAPD) analysis using as a primer the decamer oligonucleotide 1290, as previously described (10, 46). Banding patterns in ethidium bromide-stained agarose gels were analyzed digitally by using the applications Molecular Analyst and Molecular Fingerprinting (Bio-Rad, Hercules, Calif.). Dendrograms were constructed by using UPGMA based on Pearson similarity coefficients as derived from pairwise comparisons between densitometric tracks representing each gel lane.

Statistical analysis. Differences involving categorical variables were tested for significance with Yates' corrected chi square or, when appropriate, with Fisher's exact test. Differences involving resistance scores were tested by the Mann-Whitney U test. The diversity (H) of genotypes or phenotypes was calculated as H = (1   x_i²) (n/n1), where x_i is the frequency of the ith type and n is the number of types (31).

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Prevalence of E. coli O15:K52:H1 in UTI. The O15 antigen was detected in 25 (1.3%) of the 1,871 E. coli isolates obtained from the 3,664 urine specimens that were positive for UTI according to standardized criteria (16) among the 24,462 total urine specimens tested during the 1-year study period. All 25 O15-positive E. coli strains met the study definition for O15:K52:H1. They either expressed all three antigens, i.e., were O15:K52:H1 (n = 14) or were missing only the K or the H antigen, i.e., were O15:K:H1 (n = 7) or O15:K52:H (n = 4). O15 strains were similarly prevalent among E. coli isolates from community-acquired UTI (18 of 1,259 [1.4%]) and nosocomial UTI (7 of 612 [1.1%]).

Epidemiology. To assess the clinical manifestations and underlying host characteristics associated with community-acquired UTI due to E. coli O15:K52:H1, a case-control study was done in which subjects who developed community-acquired UTI due to E. coli O15:K52:H1 at any time during the study (n = 18) were compared with all evaluable control subjects who developed community-acquired UTI due to E. coli of other serotypes during a 3-month observation period (n = 411). Compared with UTI episodes due to non-O15 E. coli, episodes due to E. coli O15:K52:H1 were significantly more likely to meet criteria for pyelonephritis (Table 1) and to occur in nonelderly hosts (Table 2) (P < 0.005 for both comparisons). The mean ages of case and control subjects were 38 years (range, 28 days to 83 years) and 55 years (range, 3 months to 97 years), respectively. Other underlying host characteristics, including predisposing urological and medical conditions, were similarly distributed among case and control subjects (Table 2).

Virulence factor profiles. The 25 O15:K52:H1 strains from the present study exhibited little diversity with respect to virulence factor profile (H = 0.28). Nineteen (76%) displayed the group's modal profile, which included the presence of papEF, papG allele II, P pattern MRHA, type 1 fimbrial expression, aer, aerobactin production, and serum resistance, plus the absence of afa, sfa, hly, and cnf. Four of the remaining six O15 strains from the present study deviated from this pattern by lacking only a single trait, i.e., type 1 fimbriae, Aer, or RSB. All five papEF-positive O15:K52:H1 strains from the present study that were tested by multiplex PCR for the 11 known alleles of papA were positive for the F16 allele only (data not shown).

Antimicrobial resistance. Among the 25 E. coli O15:K52:H1 isolates from the present study, eight different antimicrobial resistance patterns were detected (Table 5). The predominant pattern included resistance to penicillins (compatible with plasmid-mediated beta-lactamase production, as supported in five strains by the detection in isoelectric focusing of an enzyme compatible with TEM-1 [data not shown]), accompanied by resistance to streptomycin, kanamycin, neomycin, sulfonamides, trimethoprim, tetracycline, and nalidixic acid. The 22 comparison O15:K52:H1 strains also exhibited diverse susceptibility patterns (n = 10), ranging from fully susceptible (the predominant pattern) to multiply resistant, although none exhibited the predominant resistance pattern of the isolates from the present study. Compared with isolates from the present study, the other O15:K52:H1 strains had a lower prevalence of resistance to any antimicrobial agent (13 of 22 versus 23 of 25; P = 0.01), a lower median resistance score (3 versus 10; P = 0.003), and a lower prevalence of quinolone resistance (0 of 22 versus 15 of 25; P < 0.001). The Terrassa isolates, which were isolated in 1988 to 1991, i.e., chronologically between the English strains (1986 to 1987) and those from the present study (1994 to 1995), were generally quite sensitive (four of seven were fully susceptible). The hallmark Amp^r Str^r Sul^r Tmp^r Tec^r Chl^r pattern of the 1986-to-1987 London outbreak was confined to the English O15 strains (4 of 5 versus 0 of 42; P < 0.001).

Biotyping. In a similarity dendrogram based on the 12-digit binary biotype code, all but 3 (i.e., 94%) of the 47 O15:K52:H1 strains were clustered together at the 83% similarity level, with no intermixed control strains (Fig. 1). The two major biotypes, which accounted for 41 (87%) of the O15:K52:H1 strains, differed only with respect to lactose fermentation. In contrast, biotypes were more numerous and more diverse among the control strains (H = 0.84 versus H = 0.51 for the O15:K52:H1 strains).

View larger version (18K): [in this window] [in a new window]   FIG. 1.   Dendrogram depicting the degrees (percentages) of biotype homology among the 70 studied strains. The O15-P, O15-L, O15-T, O15-B, and O15-E, O15:K52:H1 strains were from the present study (P), Lugo (L), Terrassa (T), Barcelona (B), and England (E), respectively. NC, nonpathogenic control; C, category, SN, study number.

Ribotyping. Ribotyping was used as a molecular typing method for a subset of the O15:K52:H1 strains (nine from the present study and nine from other locales) and for the eight ECOR strains. All but 3 of the 18 O15:K52:H1 strains exhibited a single ribotype (A), which was unique to these strains (Fig. 2). The three other O15:K52:H1 strains (36-P, 8-P, and 42-P) exhibited unique ribotypes which clustered with ribotype A at homology levels of 89, 60, and 49%, respectively (Fig. 2). The farthest removed of these strains (42-P) was from the present study, had serotype O15:K:H1, lacked pap, and was aer⁺; the intermediate strain (8-P) was from Lugo, had serotype O15:K52:H, and was pap⁺ aer⁺; and the closest strain (36-P) was from the present study, had serotype O15:K52:H1, and was pap⁺ aer⁺. All three strains had biotypes that fell within the major O15 biotype cluster (Fig. 1).

View larger version (19K): [in this window] [in a new window]   FIG. 2.   Dendrogram depicting the degrees (percentages) of ribotype homology among the 27 studied strains. The O15-P, O15-L, O15-T, and O15-E, O15:K52:H1 strains were from the present study (P), Lugo (L), Terrassa (T), and England (E), respectively. C, category, SN, study number.

RAPD fingerprinting. RAPD fingerprints were generated for five O15:K52:H1 strains, including two representatives of ribotype A (one from the present study and one from England) plus the sole representatives of each of the three outlier O15:K52:H1 ribotypes, and for five ECOR strains, i.e., strains 24 (group A), 30 (group B1), 41 (group D), 62 (group B2), and 72 (group B1) (14). Cluster analysis yielded groupings that were largely similar to those seen with ribotyping (data not shown). In replicate runs all the O15:K52:H1 strains (excepting occasionally outlier 42-P) clustered with one another and apart from the ECOR strains. The nearest neighbor to the O15:K52:H1 strains was either ECOR strain 41 (group D) alone, as in ribotyping, or a cluster containing ECOR strains 41 (group D) and 62 (group B2) (data not shown).

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

In the present study we determined the prevalence of E. coli O15:K52:H1 as a cause of UTI and bacteremia during a 1-year comprehensive surveillance of all urine and blood isolates at a hospital clinical microbiology laboratory in Barcelona, Spain, and we assessed the clinical manifestations and underlying host characteristics associated with community-acquired UTI due to this organism. We also compared the O15:K52:H1 UTI isolates from the present study with O15:K52:H1 isolates from other locales in Spain and England and with two groups of non-O15:K52:H1 control strains, analyzing multiple phenotypic and genotypic traits. We found that although O15:K52:H1 strains caused only a small fraction of all UTI, when they did cause community-acquired UTI they behaved as aggressively as or more aggressively than other E. coli strains. We also found that O15:K52:H1 strains of diverse geographic origins were similar to one another but distinct from control E. coli strains with respect to virulence factor profiles, biotypes, ribotypes, and amplotypes, evidence that E. coli O15:K52:H1 constitutes a widely disseminated uropathogenic clone.

Several observations testify to the virulence of E. coli O15:K52:H1. Our finding that this organism accounted for only a small fraction (1.3%) of the total 1,871 E. coli isolates from patients with UTI during the 1-year prospective surveillance period might be interpreted as evidence that it is not a significant uropathogen. However, although precise data are lacking, 1.3% is probably higher than the frequency of most serotypes in the E. coli population as a whole, which has been estimated to contain as many as 1,000 distinct clones (41). Thus, O15:K52:H1 strains indeed may be overrepresented among UTI isolates, which is one criterion of uropathogenicity. Additionally, in our case-control study of community-acquired UTI, E. coli O15:K52:H1 strains exhibited a level of clinical virulence equal to or greater than that of other urinary E. coli strains. Compared with UTI episodes due to other E. coli strains, the O15-associated episodes were as likely to be accompanied by pyuria, voiding symptoms, and fever and were more likely to present as pyelonephritis and to occur in young hosts. Finally, in the present study E. coli O15:K52:H1 appeared to be as prevalent a cause of E. coli bacteremia (2.9%) as of E. coli UTI (1.3%), a finding consistent with previous reports of the substantial contribution of this serotype to E. coli bacteremia in Copenhagen (4.6%) and Terrassa (4.5%) (9, 33). Although based on small numbers (two bacteremia isolates), this observation is consistent with the hypothesis that this organism may have an invasive potential as great as or greater, on average, than that of other urinary E. coli strains.

The comparative homogeneity with respect to virulence factor profile, biotype, ribotype, and RAPD fingerprints observed among the 47 O15:K52:H1 strains examined in the present study is consistent with the hypothesis that these strains, whether from Barcelona, elsewhere in Spain, or England, are clonally derived. The finding of a common virulence factor profile and a fairly uniform biotype largely confirms the results of previous studies of O15:K52:H1 strains elsewhere and extends these observations to include Spanish O15:K52:H1 isolates. Novel aspects of the present study are the demonstration of papG allele II in pap-positive O15:K52:H1 strains; the finding of the F16 papA allele in Spanish clone members (which corresponds with the reported F16 fimbrial antigen positivity of Danish O15:K52:H1 strains [33]); the documentation of the absence of sfa, afa, and cnf1; and the ribotype and RAPD results, which provide the first direct genetic evidence of clonality among O15:K52:H1 strains.

Previous investigators have proposed (correctly, as it turns out) that O15:K52:H1 strains are clonally derived (33). This hypothesis has been based in part on the three-antigen serotype itself and in part on these strains' typical homogeneity with respect to virulence traits, biotype, plasmid profile, and (in some instances) antimicrobial resistance profile (33, 38). However, all of these traits are unreliable as indicators of clonality because of their instability, horizontal mobility, and/or nonspecificity (1, 4, 8, 24, 26, 41).

For example, the distinctive multiple antimicrobial resistance phenotype which initially was regarded as a hallmark of the outbreak strain in the London O15:K52:H1 epidemic of 1986 to 1987 (38) was supplanted by a less-resistant phenotype toward the end of that outbreak (34) and was not noted at all among the subsequently reported O15:K52:H1 bacteremia isolates from Copenhagen (33) or Terrassa.

Similarly, the existence of O15:K52:H1 strains that lack pap illustrates the variability of virulence factor profiles within a clone (41) and is consistent with the known localization of pap on pathogenicity-associated islands (PAIs) (5-7, 44). Although PAIs presumably provide a vehicle through which pap and other linked virulence genes can move en bloc into new E. coli lineages (7), they also can facilitate spontaneous deletion of virulence genes (6), as may have occurred in some O15:K52:H1 clone members.

Finally, the same O:K:H serotype sometimes occurs in strains from distantly related evolutionary lineages, presumably because of recombination or convergent evolution (8, 41). Thus, the previously available evidence for clonality among O15:K52:H1 strains, although strongly suggestive, has been inconclusive. Since both ribotyping and RAPD fingerprinting correspond to MLEE in defining clonal relationships (2, 10, 27, 39, 46), the present study provides genetic confirmation that O15:K52:H1 strains, irrespective of geographic locale, are clonally derived. Further support for this conclusion is provided by our observation that all O15 strains were refractory to pulsed-field gel electrophoresis (PFGE), an uncommon characteristic observed with certain serovars of Salmonella and certain phage types of E. coli O157:H7 (B. Swaminathan, personal communication). Analysis of polymorphism of chromosomal macrorestriction fragments by PFGE using standard methods (42) was unsuccessful, because all E. coli O15:K52:H1 DNAs were completely sheared during the PFGE process despite the introduction of technical modifications intended to circumvent this problem (not shown). In contrast, control E. coli strains of other serotypes were readily analyzed by PFGE (not shown).

The antimicrobial susceptibility results contrast with the results of the other bacteriological typing that was done on the O15 and control strains. Whereas the virulence factor profiles, biotypes, ribotypes, and RAPD amplotypes of the O15:K52:H1 strains were markedly less diverse than those of the control strains and were similar among O15 strains from different geographic locales, the resistance profiles of the O15 strains were actually more diverse than those of the control strains and differed significantly between locales. This is consistent with the expectation that among members of a clone the background genomic structure (as reflected in the ribotype or RAPD amplotype) should be the most highly conserved characteristic, followed by such potentially mobile or variably expressed traits as virulence factor profiles and biotypes, followed finally by the most mobile and selection-prone trait of all, antimicrobial resistance.

The seeming locale-specific nature of the resistance patterns of the O15 strains suggests that in different geographic regions clone members may have encountered distinctive selective environments to which they have adapted extensively, possibly over as short a time span as the past 50 years of antibiotic use by humans. Alternatively, locale-specific differences in the reservoirs of mobile resistance elements available for acquisition by clone members could have driven divergent evolution of resistance patterns. Time of isolation also may be relevant, since the O15 strains from the present study, which were the most highly resistant of all, also were the most recently isolated, and hence may have had more opportunity to acquire resistance than O15 strains isolated earlier, which in contrast may have had more opportunity to lose resistance markers during subculture. Finally, anatomic site of infection may be relevant, since in several bacterial species the bacteremic strains are characteristically more antibiotic sensitive than the nonbacteremic strains (12). This would be fully consistent with the reported susceptibility of the Danish (33) and Terrassa bacteremia isolates.

Our inclusion of diverse members of the ECOR reference collection as control strains permits us to tentatively place the O15:K52:H1 clone on the E. coli evolutionary tree within ECOR group D (14). This hypothesis is consistent with the observed pap-positive but hly-, sfa-, and cnf-negative virulence profile of this clone, which is characteristic of ECOR group D strains and contrasts with the more extensive virulence factor profiles typical of ECOR group B2 strains (7, 39). It should be noted that the ECOR collection contains a single O15-positive strain, strain 24 (group A), which in the present study was only distantly related to the O15:K52:H1 strains according to ribotype homology (Fig. 2).

In summary, we have shown that strains of E. coli O15:K52:H1 constitute a genetically restricted clone that exhibits diverse antimicrobial susceptibility patterns but a limited range of virulence factor profiles and biotypes irrespective of geographic locale. Epidemiological data suggest that although the clone is an infrequent cause of UTI in a non-outbreak setting, when it does cause UTI it exhibits a level of virulence comparable to or greater than that of other E. coli strains, which qualifies it to be regarded as a uropathogenic clone.