pap Genotype and P Fimbrial Expression in Escherichia coli Causing Bacteremic and Nonbacteremic Febrile Urinary Tract Infection

Transcription

1 MAJOR ARTICLE pap Genotype and P Fimbrial Expression in Escherichia coli Causing Bacteremic and Nonbacteremic Febrile Urinary Tract Infection Gisela Otto, 1,2 Mattias Magnusson, 1 Majlis Svensson, 1 JeanHenrik Braconier, 2 and Catharina Svanborg 1 1 Division of Microbiology, Immunology, and Glycobiology, Department of Laboratory Medicine, and 2 Department of Infectious Diseases and Medical Microbiology, Lund University, Lund, Sweden Escherichia coli strains from patients with febrile urinary tract infections ( n p 73) were examined for pap genotype and P fimbrial expression in relation to bacteremia and patients background variables. Most isolates were pap + by DNA hybridization ( n p 51), and 36 were papg IA2+ and 18 prsg J96+ by polymerase chain reaction. The pap and papg genotypes of the infecting strain were shown to vary with host compromise, sex, and age. Bacteremia in noncompromised patients was caused by papg IA2+ strains, but compromised hosts carried a mixture of papg IA2+, prsg J96+, and pap strains. Women of all ages were infected with papg IA2+ strains. Infected men carried prsg J96+ or pap strains and were older, and most had compromising conditions. papg IA2+ strains predominated among patients with medical illness, whereas prsg J96+ strains predominated among patients with urinary tract abnormalities. These findings emphasize the strong influence of host factors on the selection of E. coli strains causing febrile urinary tract infections. Urinary tract infections (UTIs) vary in severity, depending on bacterial virulence and host resistance. Whereas some patients with bacteriuria have no or few symptoms, others develop acute pyelonephritis, bacteremia, or urosepsis. Blood culture results are positive for bacterial infection in 20% 30% of patients with Received 21 April 2000; revised 11 August 2000; electronically published 30 April The study was approved by the Research Ethics Committee of the Medical Faculty at the University of Lund, Sweden, and informed consent was obtained from all participants. Grant support: Swedish Medical Research Council; the Österlund, Wallenberg, and Crawford foundations; the Royal Physiographic Society; and the Medical Faculty, Lund University. C.S. was the recipient of a Bristol-Myers-Squibb unrestricted grant. Reprints or correspondence: Dr. Gisela Otto, Division of Microbiology, Immunology, and Glycobiology, Institute of Laboratory Medicine, Lund University, Sölvegatan 23, Lund, Sweden (gisela.otto@infek.lu.se). Clinical Infectious Diseases 2001; 32: by the Infectious Diseases Society of America. All rights reserved /2001/ $03.00 s [1 4], but few patients develop septic shock [3]. Epidemiological studies have shown that acute pyelonephritis is caused by uropathogenic Escherichia coli clones belonging to certain O:K:H (somatic [O], capsular [K], and flagellar [H] antigens) serotypes and with chromosomal pathogenicity islands that encode multiple virulence factors [5 10]. Bacteremia is caused by a subset of these clones that express P fimbriae more often than strains causing less severe forms of UTI [1, 2, 6, 7, 11 16]. Two P fimbrial subtypes predominate among uropathogenic E. coli. They differ in terms of the G adhesin at the tip of the fimbrial organelle, with slight variations in receptor specificity [17 20]. The class II G adhesins encoded by the papg IA2 sequences are associated with pyelonephritis and bacteremia [1, 13, 14, 21]. The class III G adhesins encoded by the prsg J96 sequences are associated with cystitis but have been found in pyelonephritis and bacteremia [1, 13 15, 21]. P Fimbrial Genotype and Febrile UTI CID 2001:32 (1 June) 1523

2 The virulence of E. coli strains causing and bacteremia is influenced by host compromise, a term denoting concomitant diseases in the patient. Host compromise allows less-virulent strains to cause infection [1 3, 7, 11 13], and P fimbrial strains are less common in compromised patients with bacteremia (52% 68%). Few studies have addressed the influence of host compromise on the presence of G adhesins, especially the prsg J96 adhesin [1, 13]. Earlier studies on P fimbriation, pap genotype, and bacteremia were retrospective or involved selected patient groups. This study examined the pap and papg genotypes and the P fimbrial expression in a prospectively enrolled, unselected patient population with s. The findings emphasize the influence of host compromise, sex, and age on the papg allele of the infecting strain. PATIENTS AND METHODS Patients. Seventy-three patients with febrile E. coli UTIs were enrolled in a prospective study [4]. The patients were years old and had clinically diagnosed s requiring hospitalization and parenteral antibiotic therapy. Their temperatures were 38C at the time of admission, and they had focal symptoms involving the urinary tract (flank pain, costovertebral angle tenderness, dysuria, or urinary frequency or urgency); laboratory findings of a positive result on a nitrite test or increased urine leukocyte counts; and/or urinary tract instrumentation or acute urinary retention preceding the onset of fever. All patients had significant bacteriuria at admission. Patients were assigned to the bacteremic group if the result of 1 blood culture was positive and the same bacterial species was recovered from the blood and urine. The patients were hospitalized, treated, and followed for 4 months after therapy. The patients represent a subset of the entire study sample of 97 patients. Eighty-three of those 97 patients had E. coli infections, and the E. coli strains from 73 patients were saved. Forty women and 33 men were included; the median age was 49 and 68 years, respectively. Bacteremia was diagnosed for 24 patients (33%): 15 women (38%) and 9 men (27%). The bacteremic patients were older but did not differ with regard to sex or host compromise from those without bacteremia (table 1). The urinary tracts of 66 of the 73 patients were evaluated by iv pyelography or ultrasonography. Host compromise. Forty-six patients (63%) with compromising conditions were assigned to 2 subgroups: those with medical illness ( n p 17; 12 women, 5 men) and those with urinary tract abnormalities ( n p 29; 5 women, 24 men) (table 1). Eight patients with medical illness and urinary tract abnormalities were assigned to the first subgroup. Twenty-seven patients were healthy. Patients assigned to the medical illness group had diseases or had received therapies that might influence the resistance to UTI (table 1). Patients assigned to the urinary tract abnormalities group had a history of noninfectious urinary tract disease and/or structural abnormalities of the urinary tract, confirmed by iv pyelography, ultrasonography, or urological examination (table 1). Noncompromised patients were younger (median age of 45 years vs. median age of 69 years among compromised patients), and 85% were women, whereas men predominated in the compromised group (63%). The 4 men designated as healthy were!50 years old and presented with fever (temperature, 38.4C 39.7C), lower urinary tract symptoms, C-reactive protein level 1130 mg/dl, and leukocytosis. They were all examined by pyelography and followed for 4 months, and no signs of compromising conditions were noted. The frequency of bacteremia did not differ between compromised and noncompromised patients (30% vs. 35%; not significant [NS]) or between patients with medical illness and those with urinary tract abnormalities (53% vs. 24%; NS). Urine and blood cultures. Cultures of urine and blood samples were performed at the time of enrollment [4]. Significant bacteriuria was defined by growth of a single bacterial Table 1. Host background variables among patients with bacteremic and nonbacteremic febrile urinary tract infections (UTIs). Variable Nonbacteremic (n p 49) Bacteremic (n p 24) Men/women 24/25 9/15 Age in years, median (range) 53 (19 83) 69 (29 85) No host compromise 19 (39) 8 (33) Host compromise 30 (61) 16 (67) Medical illness a 8 (16) 9 (38) Diabetes mellitus 6 (12) 3 (13) Cancer 2 (4) 2 (8) Immunosuppression 0 4 (17) Urinary tract abnormalities b 27 (55) 10 (42) With medical illness 5 (10) 3 (13) Alone c 22 (45) 7 (29) Urinary tract device d 5 (10) 2 (8) NOTE. Data are no. (%) of patients, unless otherwise indicated. No significant difference was seen between bacteremic and nonbacteremicpatients, except in age ( P!.05; the Mann-Whitney U test and Fisher s exact test were used for comparisons). a Includes diabetes mellitus, chronic lymphocytic leukemia, treatment with corticosteroids (for rheumatoid arthritis, bronchial asthma, cancer, and polymyalgia rheumatica), ethylismus, systemic lupus erythematus, cancer, and Crohn s disease. b Includes prostatic hyperplasia, operations of the urogenital tract, renal or ureteral stones, severe prolapse of the uterus, urinary tract instrumentation, artificial urinary sphincter, multiple sclerosis with neurogenic bladder dysfunction, and rectovesical fistula. c Patients with medical illness excluded ( n p 8). d Includes indwelling urinary catheter ( n p 4; 4 weeks for 2 patients) and intermittent catheterization ( n p 3) CID 2001:32 (1 June) Otto et al.

3 Figure 1. Identification of papg alleles by PCR. A, Sequence alignment of different papg alleles. Sequences amplified by primers used in this study are indicated by grey bars. Sequences amplified by primers used by Johnson et al. [15, 24] are indicated by white bars. Numbers signify the position of the first and last base in the amplified sequence. B, Primer sequences, positions, and length of the PCR products. C, Specificity of PCR reactions with the primers defined in B. The control strains Escherichia coli J96 and E. coli IA2 carried the prsg J96 and the papg IA2 sequences, respectively. The clinical isolate E. coli U33468 carried both the papg IA2 and the prsg J96 sequences, and the clinical isolate E. coli U22728 carried the papg IA2 sequence. E. coli HB101 was the negative control. species at 10 4 cfu/ml in women and 10 3 cfu/ml in men [22]. Serotyping of E. coli. The O, K, and H antigens of the E. coli isolates were determined using antisera to 165 O antigens, 72 K antigens, and 53 H antigens [23]. Nontypeable antigens were defined as ON, KN, and HN. Strains with no demonstrable capsule were designated as K, nonmotile cultures as H, and rough strains as Oru. P fimbrial phenotype. The P fimbrial phenotype was defined by P blood group dependent hemagglutination [14], after in vitro passage to enhance expression. Strains that agglutinated P 1 erythrocytes in an ABH blood group independent manner were assigned to class II. Strains agglutinating only A 1 P 1 and not OP 1 erythrocytes were assigned to class III. pap Genotype. The pap genotype was determined by DNA-DNA hybridization, with probes specific for the 5 (HindIII) and 3 (SmaI) regions of the pap operon and derived from the pap gene cluster of E. coli J96 [14]. papg genotype. The papg genotypes were defined by PCR, using primer pairs that matched unique regions of the papg IA2 or prsg J96 sequences (figure 1). P fimbrial E. coli IA2 and E. coli J96 and pap recombinants E. coli HB101 (papg IA2 ) and P Fimbrial Genotype and Febrile UTI CID 2001:32 (1 June) 1525

4 Table 2. pap Genotype and P fimbrial expression in Escherichia coli isolates from patients with bacteremic and nonbacteremic febrile urinary tract infections (UTIs). No. (%) of isolates Variable Total (n p 73) Bacteremic (n p 24) Nonbacteremic (n p 49) pap Genotype HindIII and SmaI 51 (70) 18 (75) 33 b (67) NS Negative 22 (30) 6 (25) 16 (33) NS papg allele(s) papg IA2 33 (45) 15 (63) 18 b (37).048 prsg J96 15 (21) 3 (12) 12 (24) NS papg IA2 and prsg J96 3 (4) 0 3 (6) NS Negative 22 (30) 6 (25) 16 (33) NS P fimbrial expression P fimbriae present c 46 (63) 17 (71) 29 (59) NS Other MR adhesins d 8 (11) 2 (8) 6 (12) NS Negative 19 (26) 5 (21) 14 (29) NS P fimbrial subtype Class II e 35 (48) 15 (63) 20 (41) NS Class III f 11 (15) 2 (8) 9 (18) NS NOTE. MR, mannose-resistant; NS, not significant. a Comparison between bacteremic and nonbacteremic isolates, with the Fisher s exact test. b One strain hybridized only with the HindIII probe and not with the SmaI probe. c Agglutinated human A 1 P 1 but not A 1 p erythrocytes. d Agglutinated A 1 p erythrocytes. These strains may or may not be coexpressing P fimbriae. e Agglutinated human OP 1,A 1 P 1, and sheep erythrocytes. f Agglutinated human A 1 P 1 and sheep erythrocytes. E. coli HB101 (prsg J96 ) were used as positive controls, and E. coli HB101 and E. coli AAEC191A(pPKL4) were used as negative controls. Primers amplifying the fima gene were used to control the yield of bacterial DNA. Whole bacterial cells provided template DNA. The reaction mixture (25 ml) contained 1 PCR buffer (Perkin-Elmer), 5 mm MgCl 2, 0.2 mm of each dntp, 0.1 mg/ml bovine serum albumin, 0.5 mm of each primer, and U/mL AmpliTaq Gold (Perkin-Elmer). The reaction mixture was denatured at 94C for 10 min and amplified by 5 touchdown cycles of denaturing, annealing, and extension at 94C for 1 min, 63C 59C for 1 min, and 72C for 2 min, followed by 45 cycles of 94C for1min,58c for 1min,72C for 2 min, and final extension at 72C for7min. The PCR products were separated by agarose gel electrophoresis (2.5%) with ethidium bromide and photographed under ultraviolet transillumination. Each strain was amplified with primers for papg IA2 or prsg J96 in separate tubes. Strains positive for papg IA2 or for prsg J96 were assigned to that genotype. The primers did not cross-amplify other papg sequences, as shown by the recombinant strains containing a single known copy of papg IA2 or prsg J96. P a Statistics. The Mann-Whitney U test and Fisher s exact test were used. A 2-tailed P value!.05 was considered statistically significant. RESULTS pap and papg genotypes of the clinical isolates. The majority ( n p 51, 70%) of the strains were pap, and all but one carried the entire pap gene cluster, as shown by hybridization with both probes (table 2). The remaining 22 isolates did not hybridize with the pap DNA probes. All the pap strains were papg by PCR. There were 33 papg IA2 strains, 15 prsg J96 strains, and 3 strains carrying both sequences (table 2). Twenty-two strains were negative for pap and papg. P fimbrial expression. P fimbriae were expressed by 46 strains (63%; table 2), with 35 strains belonging to class II and 11 belonging to class III. Eight strains (11%) expressed P blood group independent adhesins, as shown by agglutination of A 1 p erythrocytes. The remaining 26% of the strains did not express P fimbriae CID 2001:32 (1 June) Otto et al.

5 Table 3. pap Genotype and P fimbrial expression of Escherichia coli isolates from patients with febrile urinary tract infections. Genotype Phenotype HindIII SmaI papg IA2 prsg J96 A 1 P 1 OP 1 A 1 p Sheep No. (%) of isolates (n p 73) a b 32 (44) 1(1) 1(1) c 11 (15) 3(4) 2(3) 1(1) 19 (26) 2(3) 1(1) a One strain did not hybridize with the SmaI probe. b Eight papg IA2 strains agglutinated sheep erythrocytes only at 4C. c Seven prsg J96 strains agglutinated OP 1 erythrocytes at 4C but not at room temperature. Comparison of genotype and phenotype. The papg IA2 strains all expressed class II P fimbriae (table 3). All but 1 of the prsg J96 strains expressed class III P fimbriae, and the double papg IA2 and prsg J96 strains showed a class II phenotype. The pap and papg strains did not express P fimbriae. Agglutination of A 1 p erythrocytes was independent of pap/papg genotype. papg genotype in relation to O:K:H serotype. There were 56 different O:K:H serotypes, with 18 strains (25%) belonging to the classic pyelonephritis-associated O:K:H serotypes [5 8]. Eight O:K:H serotypes occurred more than once (table 4), with O1:K1:H7, O15,45:K1:H7, and O75:K5:H being most frequent. The O:K:H serotypes of the remaining 48 isolates occurred only once. Eighty-eight percent of strains with frequently occurring serotypes were papg IA2, compared with 23% of the strains with serotypes that appeared only once ( P!.0001). The majority of prsg J96 isolates (93%) and pap isolates (91%) had single-occurrence serotypes. pap/papg genotype and P fimbrial expression in relation to bacteremia. There was no difference in the pap frequency of isolates from patients with and from patients without bacteremia, but papg IA2 strains occurred more often in the bacteremic group (table 2). The prsg J96 strains accounted for 12% of the isolates in the bacteremic group and 24% in the nonbacteremic group. There was no difference in P fimbrial expression between bacteremic and nonbacteremic isolates (table 2). Twenty-one percent of the bacteremic and 4% of the nonbacteremic isolates had the O1:K1:H7 serotype ( P!.05), and all were papg IA2. pap/papg genotype and host compromise. There were marked differences in pap frequency and in papg allele distribution related to patients health (table 5). The pap and papg IA2 isolates were more common in noncompromised patients than in compromised patients. In contrast, prsg J96 isolates were more common in compromised patients. The subgroup with medical illness had fewer pap and papg IA2 isolates than did noncompromised patients, but there was no difference in the prsg J96 frequency. The patients with urinary tract abnormalities had fewer pap and papg IA2 isolates than did noncompromised patients but more prsg J96 isolates. Indeed, 87% of all prsg J96 isolates were from the group with urinary tract abnormalities. Isolates from patients with medical illness or urinary tract abnormalities showed no difference in pap frequency, but pap G IA2 isolates were more common in the group with medical illness and prsg J96 in the group with urinary tract abnormalities. The eight patients with both medical illness and urinary tract abnormalities did not differ, in terms of papg allele distribution, from other patients in the group with medical illness, but they differed significantly in this respect from the group with urinary tract abnormalities ( P!.05 for both papg IA2 and prsg J96 ). pap/papg genotype and host compromise in relation to bacteremia. All isolates causing bacteremia in noncompromised patients were pap and papg IA2. Isolates causing bacteremia in patients with compromising conditions carried pap sequences (62%) and papg IA2 sequences (44%; P!.01) less frequently. The prsg J96 isolates caused bacteremia only in compromised patients ( n p 3). pap/papg genotype and sex. There was a marked difference in pap and papg genotypes between isolates from women and men (figure 2). pap ( P!.05) and papg IA2 isolates ( P!.0001) were more common in women, whereas prsg J96 isolates were more common in men ( P!.01). This sex-related difference in papg allele distribution was marked in the nonbacteremic group, with 60% papg IA2 isolates and 8% prsg J96 isolates from women and 12% papg IA2 isolates and 42% prsg J96 isolates from men ( P!.001). papg IA2 isolates were also more common among bacteremic women (80% vs. 33%; P!.05), but the frequency of prsg J96 isolates did not differ between bacteremic women and men (7% vs. 22%; NS). Six men and 1 woman underwent instrumentation or catheterization. The sex difference in papg IA2 and prsg J96 distribution remained after exclusion of these 7 patients ( P!.001). pap/papg genotype and age. The pap frequency did not differ between patients who were 50 and those who were 150 years old (80% vs. 82%), but papg IA2 isolates were more common among younger patients (63% vs. 32%; P!.05). Patients P Fimbrial Genotype and Febrile UTI CID 2001:32 (1 June) 1527

6 Table 4. pap Genotype in relation to O:K:H serotype in Escherichia coli from patients with febrile urinary tract infections. O:K:H serotype Frequent occurrences a Total no. of isolates (n p 73) No. of isolates with indicated papg genotype papg IA2 prsg J96 papg IA2 and prsg J96 pap O1:K1:H O1:K1:H O2:K1:H O6:KN:H O7:K1:H O15,45:K1:H O75:K5:H Oru:K1:H Total 25 (34%) 22 (88%) b Single occurrence c 48 (66%) 11 (23%) b 14 (29%) 3 (6%) 20 (42%) NOTE. H, flagellar antigen; H, nonmotile antigen; K, capsular antigen; K, no demonstrable capsule; N, nontypeable; O, somatic antigen; Oru, rough strain. a O1:K1:H7, O1:K1:H, O7:K1:H, O15,45:K1:H7, and O75:K5:H were found in bacteremic and nonbacteremic patients. O2:K1:H7, O6:KN:H1, and Oru:K1:H7 were found only in nonbacteremic patients. b papg IA2 was more common in strains with frequently occurring serotypes than in the strains with O:K:H serotypes that occurred only once ( P!.0001; Fisher s exact test). c O2:K1:H4, O2:K1:H5, O2:K :H, O6:K5:H1, O6:K5:H, O6:K :H1, O6:K :H, O9:KN:H, O15:K52:H1, O20:K?: H, O25:K16:H4, O83:K14:H31, O120:K7:H31, and Oru:K1:H were found in bacteremic patients, and O2:K2:H1, O2:K5:H4, O2:KN:H6, O2:KN:H, O2:K :H4, O4:K :H1, O4:K :H5, O6:K2:H1, O6:K13:H, O6:K14:H31, O6:K14:H, O6:K :H?, O9:K30:H, O13,44,106:K :H5, O15:K4:H33, O15:K25:H5, O15:K :H1, O15:K :H, O16:K1:H6, O17:K : H18, O18:K1:H, O18:K5:H, O22:K13:H1, O22:K?:H5, O23:KN:H1, O25:K5:H1, O25,102:K5:H1, O75:KN:H7, O82: K?:H4, O88:K :H6, Oru:K1:H5, Oru:K :H33, Oru:K?:H, and Oru:K?:H1 were found in nonbacteremic patients. with papg IA2 strains had a median age of 49 years, compared with 68 years for patients with prsg J96 strains ( P!.05). papg IA2 strains predominated in women of all age groups, from 100% among patients years old to 44% among Table 5. Variable pap Genotype in relation to host compromise in patients with febrile urinary tract infections. Yes (n p 27) Host compromise patients years old (figure 3). In men, the frequency of prsg J96 isolates increased with age, up to 46% in patients years old. The majority of men (70%) were 150 years old, and none were!40 years old. No. (%) of isolates Urinary tract abnormalities Medical illness No Total Total (n p 46) P a (n p 29) P a (n p 17) P a P b pap Genotype pap 24 (89) 27 (59) (59) (59).030 NS pap 3 (11) 19 (41) (41) (41).030 NS papg allele(s) papg IA2 21 (78) 12 (26)! (14)! (47) prsg J96 1 (4) 14 (30) (45) (6) NS.008 papg IA2 and prsg J96 2(7) 1(2) 1(6) pap 3 (11) 19 (41) (41) (41).030 NS papg IA2 vs. prsg J96 21/1 12/ /13! /1 NS.003 NOTE. Definitions of host compromise are in Patients and Methods and footnotes b and c in table 1. The 3 isolates carrying both papg IA2 and prsg J96 were included in the analysis only when pap and pap strains were compared. The remaining comparisons were based on data for 70 patients. NS, not significant. a Compared with noncompromised patients (Fisher s exact test). b Compared with patients who had urinary tract abnormalities (Fisher s exact test) CID 2001:32 (1 June) Otto et al.

7 Figure 2. Distribution of papg alleles in Escherichia coli isolates from women and men with febrile urinary tract infections and in relation to host compromise. Black, papg IA2 isolates; grey, prsg J96 isolates; striped, both papg IA2 and prsg J96 isolates; white, pap isolates. Combined analysis of host compromise, sex, age, and pap/ papg genotype. The majority of papg IA2 strains (64%) were from noncompromised women (median age, 42 years). The majority of prsg J96 isolates (73%) were from men with urinary tract abnormalities (median age, 72 years). In fact, 82% of all papg IA2 isolates were from women, and 80% of all prsg J96 isolates were from men. The women included in the study were younger and healthier (58%) than the men (12%), who had a high frequency of urinary tract abnormalities (73%). The difference in pap genotype distribution between women and men was dramatic in the noncompromised patient group (figure 3). papg IA2 strains were recovered from 92% of the women but none of the men ( P!.001). The difference was less marked among compromised women and men who had a mixture of papg IA2, prsg J96, and pap strains (NS), and no differences were found among patients with urinary tract abnormalities or among patients with medical illness. DISCUSSION This study examined the papg genotype and P fimbrial expression in E. coli isolates from patients with s. The pap and papg genotypes of the infecting E. coli strain were shown to vary with host compromise, sex, and age. P fimbrial E. coli of the papg IA2 genotype infected healthy women and patients with medical illness, whereas older men and patients with compromising conditions of the urinary tract carried prsg J96 E. coli or pap strains. The findings confirmed the high frequency of pap and papg IA2 strains in bacteremic patients without compromising conditions. They also show that the host exerts a strong selective pressure on the pap genotype and the P fimbrial phenotype of strains causing. Analyses of pap genotype have developed to a great extent since the early use of large probes encompassing the pap J96 gene cluster [7, 25, 26]. papg-specific probes were used to investigate isolates from different types of UTIs and from the fecal flora [1, 14, 18, 27, 28]. Johnson et al. designed PCR primers specific for the different papg sequences and showed a high concordance with DNA hybridization [15, 24]. In the present study, papg genotyping was done by PCR with primers adjacent to and partially overlapping those used by Johnson et al. The papg primers gave a complete concordance with DNA hybridization, and all strains with amplified papg sequences could express P fimbriae after in vitro passage. Previous studies of, P fimbriae, pap genotype, and host factors often were based on patient populations selected in terms of age, sex, host compromise, or bacteremia [1, 2, 11 13, 28, 29]. The present study had a prospective design and enrolled adult patients of both sexes with s who were in need of hospitalization, which indicated severe illness. P Fimbrial Genotype and Febrile UTI CID 2001:32 (1 June) 1529

8 Figure 3. Influence of age and sex on the papg IA2 and prsg J96 alleles in febrile urinary tract infections. The total no. of isolates in each age group is indicated in parentheses. Black, papg IA2 isolates; grey, prsg J96 isolates. The identification of host compromise was supplemented by observations made during the 4-month follow-up period. The results confirmed that host compromise has a strong influence on the pap genotype of strains causing bacteremia [1, 2, 11 13] and added information on papg. All noncompromised bacteremic patients carried papg IA2 isolates, but prsg J96 isolates caused bacteremia only in compromised patients. This confirmed a suggested association between prsg J96 isolates and host compromise in urosepsis [13]. The pap/papg genotype differed according to the type of host compromise. Patients with medical illness resembled the noncompromised group in papg allele distribution but had more pap strains. Patients with urinary tract abnormalities also had pap strains, confirming the association between P fimbriae negative strains and host compromise [1, 3, 7, 11 13], but they also were shown to carry 87% of all prsg J96 strains. The predominance of prsg J96 isolates among patients with urinary tract abnormalities is in agreement with recent findings in a study of young children [29]. Women are susceptible to UTIs throughout adult life. In contrast, UTIs in men are mainly restricted to men of advanced age, coinciding with, for example, increasing urinary tract obstructions due to prostatic hyperplasia [3, 30]. This sex-related difference in UTI susceptibility has been attributed to a variety of factors, including anatomic differences, but few suggestions have been made as to possible molecular mechanisms. The present study demonstrated a marked sex-related difference in papg allele distribution. Healthy women were infected with papg IA2 strains (92%), whereas men were infected with prs G J96 or pap isolates. The mucosal receptor repertoire of men and women has not been compared, but the differential selection of papg genotypes may suggest that molecular differences in adhesion and receptor recognition determine mucosal interactions in women and men. Additional studies in larger populations may define these associations more clearly. The pap/papg allele distribution did not differ between compromised women and compromised men. Because the majority of compromised patients were men and because most had urinary tract abnormalities, the prsg J96 strains appeared to have an association with sex. However, analysis of women with urinary tract abnormalities showed no difference in the distribution of papg alleles between those women and the men, suggesting that host compromise overruled the selective effect of sex. Urinary tract abnormalities appeared to facilitate infection with prsg J96 strains in women and in men and explained the age-correlated increase of prsg J96 strains in men. To summarize, papg IA2 E. coli was the major uropathogen in women of all ages, the major cause of bacteremia in healthy women, and a frequent cause of bacteremia in women with compromising conditions. Men seldom developed s unless compromising conditions were present and had a less clear-cut pattern of papg allele selection. prsg J96 E. coli was a common uropathogen in men; it was associated with urinary tract abnormalities, and it caused bacteremia only in compromised patients. We conclude that host factors place a strong selective pressure on the establishment of uropathogenic E. coli and the selection of their virulence profiles. References 1. Otto G, Sandberg T, Marklund BI, et al. Virulence factors and pap genotype in Escherichia coli isolates from women with acute pyelonephritis, with and without bacteremia. Clin Infect Dis 1993; 17: Ikäheimo R, Siitonen A, Kärkkäinen U, et al. Community-acquired pyelonephritis in adults: characteristics of E. coli isolates in bacteremic and nonbacteremic patients. Scand J Infect Dis 1994; 26: Kunin CM. Urinary tract infections: detection, prevention, management. 5th ed. Baltimore: Williams & Wilkins, Otto G, Braconier JH, Andreasson A, et al. Interleukin-6 and disease severity in patients with bacteremic and nonbacteremic febrile urinary tract infection. J Infect Dis 1999; 179: Örskov I, Örskov F, Birch-Andersen A, Kanamori M, Svanborg-Eden C. O, K, H and fimbrial antigens in Escherichia coli serotypes associated 1530 CID 2001:32 (1 June) Otto et al.

9 with pyelonephritis and cystitis. Scand J Infect Dis Suppl 1982; 33: Väisänen-Rhen V, Elo J, Väisänen E, et al. P-fimbriated clones among uropathogenic Escherichia coli strains. Infect Immun 1984; 43: Johnson JR. Virulence factors in Escherichia coli urinary tract infection. Clin Microbiol Rev 1991; 4: Johnson JR, Örskov I, Örskov F, et al. O, K, and H antigens predict virulence factors, carboxylesterase B pattern, antimicrobial resistance, and host compromise among Escherichia coli causing urosepsis. J Infect Dis 1994; 169: Blum G, Ott M, Lischewski A, et al. Excision of large DNA regions termed pathogenicity islands from trna-specific loci in the chromosome of an Escherichia coli wild-type pathogen. Infect Immun 1994; 62: Kao J-S, Stucker D, Warren J, et al. Pathogenicity island sequences of pyelonephritogenic Escherichia coli CFT073 are associated with virulent uropathogenic strains. Infect Immun 1997; 65: Brauner A, Leissner M, Wretlind B, et al. Occurrence of P-fimbriated Escherichia coli in patients with bacteremia. Eur J Clin Microbiol 1985;4: Johnson JR, Roberts PL, Stamm WE. P fimbriae and other virulence factors in Escherichia coli urosepsis: association with patients characteristics. J Infect Dis 1987; 156: Johnson JR. papg alleles among Escherichia coli strains causing urosepsis: associations with other bacterial characteristics and host compromise. Infect Immun 1998; 66: Johanson IM, Plos K, Marklund BI, et al. Pap, papg and prsg DNA sequences in Escherichia coli from the fecal flora and the urinary tract. Microb Pathog 1993; 15: Johnson JR, Russo TA, Brown JJ, et al. papg alleles of Escherichia coli strains causing first-episode or recurrent acute cystitis in adult women. J Infect Dis 1998; 177: Domingue GJ, Roberts JA, Lancirica R, et al. Pathogenic significance of P-fimbriated Escherichia coli in urinary tract infections. J Urol 1985; 133: Lund B, Lindberg F, Marklund BI, et al. The papg protein is the a- D-galactopyranosyl-(1 4)-b-D-galactopyranose-binding adhesin of uropathogenic Escherichia coli. Proc Natl Acad Sci USA 1987; 84: Strömberg N, Marklund BI, Lund B, et al. Host-specificity of uropathogenic Escherichia coli depends on differences in binding specificity to Gala1-4Gal containing isoreceptors. EMBO J 1990; 9: Marklund BI, Tennent JM, Garcia E, et al. Horizontal gene transfer of the Escherichia coli pap and prs pili operons as a mechanism for the development of tissue-specific adhesive properties. Mol Microbiol 1992; 6: Johanson I, Lindstedt R, Svanborg C. Roles of the pap and prs encoded adhesins in Escherichia coli adherence to human uroepithelial cells. Infect Immun 1992; 60: Johnson JR, Brown JJ, Maslow JN. Clonal distribution of the three alleles of the Gal(a1 4)Gal-specific adhesin gene papg among Escherichia coli strains from patients with bacteremia. J Infect Dis 1998; 177: Hamilton-Miller JMT. Urinary tract infections. Curr Opin Infect Dis 1997; 10: Örskov F, Örskov I. Serotyping of Escherichia coli. In: Bergen T, ed. Methods in microbiology. Vol 14. London: Academic Press, 1984: Johnson JR, Brown JJ. A novel multiply primed polymerase chain reaction assay for identification of variant papg genes encoding the Gal(a1 4)Gal-binding papg adhesins of Escherichia coli. J Infect Dis 1996; 173: Plos K, Hull S, Hull R, et al. Distribution of the P-associated pilus (pap) region among Escherichia coli isolates from natural sources: evidence for horizontal gene transfer. Infect Immun 1989; 57: Plos K, Carter T, Hull S, et al. Frequency and organization of pap homologous DNA in relation to clinical origin of uropathogenic Escherichia coli. J Infect Dis 1990; 161: Ekbäck C, Mörner S, Lund B, et al. Correlation of genes in the pap gene cluster to expression of globoside-specific adhesin by uropathogenic Escherichia coli. FEMS Microbiol Lett 1986; 34: Plos K, Conell H, Jodal U, et al. Intestinal carriage of P fimbriated Escherichia coli and the susceptibility to urinary tract infection in young children. J Infect Dis 1995; 171: Jantunen ME, Siitonen A, Koskimies O, et al. Predominance of class II papg allele of Escherichia coli in pyelonephritis in infants with normal urinary tract anatomy. J Infect Dis 2000; 181: Lipsky B. Urinary tract infections in men: epidemiology, pathophysiology, diagnosis and treatment. Ann Intern Med 1989; 110: P Fimbrial Genotype and Febrile UTI CID 2001:32 (1 June) 1531