Eijkman-Winkler Institute for Medical Microbiology, University Hospital Utrecht HP G04-614, 3584 .... scores were compared by the Mann-Whitney U test and.
JOURNAL
OF
CLINICAL MICROBIOLOGY, Dec. 1993, p. 3174-3178
Vol. 31, No. 12
0095-1137/93/123174-05$02.00/0 Copyright X 1993, American Society for Microbiology
Escherichia coli in Bacteremia: O-Acetylated Ki Strains Appear to Be More Virulent than Non-O-Acetylated Ki Strains HELMA FRASA,1* JOHN PROCEE,' RUURD TORENSMA,12 ANJO VERBRUGGEN,3 ALE ALGRA,4 MAJA ROZENBERG-ARSKA,l KEES KRAAIJEVELD, AND JAN VERHOEF' Institute for Medical Microbiology, University Hospital Utrecht HP G04-614, 3584 Utrecht, Winkler EijkmanU-gene Research B. V.,2 and Clinical Epidemiology Unit, University Department of Public Health and Epidemiology, 4 Utrecht, and National Institute for Public Health and Environmental Protection, Bilthoven, 3 The Netherlands Received 13 April 1993/Returned for modification 29 June 1993/Accepted 31 August 1993
A total of 174 blood isolates of Escherichia coli, collected during a 5-year period at the University Hospital Utrecht, were serotyped with rabbit sera against 171 0 antigens and 73 capsule (K) antigens. The four most prevalent 0-antigen serotypes were 06 (n = 22), 018 (n = 19), 01 (n = 19), and 02 (n = 15). Thirty-one strains were not typeable with any of the 0-antigen-typing sera. Of the 148 strains that were subjected to K-antigen serotyping, 34 strains lacked a K antigen and 41 were not typeable with the K-antigen-specific antisera used in the study. Ki was by far the most frequently found K-antigen serotype; this was followed by K2, K53, K5, K13, K7, K(A)28, and K15. Strains possessing a Kl antigen were further classified as either 0-acetyl-positive (n = 12) or 0-acetyl-negative (n = 21) strains. Retrospective analysis of patients infected with different E. coli isolates-nonencapsulated (n = 23), 0-acetylated Ki (n = 12), and non-0-acetylated Ki (n = 21)-revealed clinical differences. More patients suffered from sepsis (94% versus 74%), and a higher rate of mortality was found in the group infected with Kl isolates (18 versus 9%v) than in the group infected with nonencapsulated isolates. More patients with severe sepsis (25 versus 10%) and a higher mortality (33 versus 10%o) were found in the group infected with 0-acetylated Kl isolates than in the group infected with non-0-acetylated isolates. Also, the hospitalization of these patients was prolonged. Thus, 0-acetylated E. coli Kl strains seem to be more virulent than non-0-acetylated Kl strains.
Escherichia coli is one of the most commonly isolated microorganisms from patients with gram-negative bacteremia (7, 16). There is a great diversity of 0 and K antigens among E. coli strains, but there appears to be some correlation between the serotype of the strain and/or the presence of capsular (K) antigens (12). E. coli Kl has been reported to be a major pathogen (2). Approximately 80% of the cases of neonatal E. coli meningitis are caused by Kl-encapsulated E. coli (13). It is well-known that Kl-encapsulated strains have a higher level of resistance to serum than nonencapsulated strains (11). Prior surveys of blood isolates have shown that the Kl capsule is an essential virulence factor, associated with bacteremic infections in adults, particularly in strains which possess the rough lipopolysaccharide phenotype (3). In a rat model, a >105-fold difference in the 50% lethal dose for the Kl-encapsulated parent compared with that for the nonencapsulated one for certain 0 serotypes was noticed (3). The Kl capsule can be subdivided into 0-acetylated and non-Oacetylated capsules. In rabbits, non-0-acetylated Kl capsules are less immunogenic than the 0-acetylated ones (10). We studied the distribution of E. coli serotypes among human blood isolates and correlated them with the clinical signs of infection in various patient groups. A high preva-
acetylated Kl capsule, and with an 0-acetylated Kl capsule were compared to determine whether the non-O-acetylated Kl antigen is associated with a higher level of virulence.
MATERIALS AND METHODS Bacteria. A total of 174 E. coli isolates were consecutively isolated from blood specimens of 174 patients sent to the microbiology laboratory of the University Hospital Utrecht between 1986 and 1991. Routine biochemical identification was performed with Enterotube II Roche (Hoffmann-La Roche, Basel, Switzerland). When no reaction was observed with any of the 0-antigen-typing sera, the strain was confirmed as an E. coli strain isolate by the following biochemical reactions: triple sugar iron, urea, indole production, malonate production, and citrate utilization. Serotyping. Serotyping was performed by the method described by Guinee et al. (4). Bacteria were grown in bovine broth, incubated overnight at 37°C, and then treated with heat at 100°C for 1 h. The 0-antigen group of these boiled E. coli isolates was determined by using agglutination reactions with antisera to each 0 antigen. Polyclonal-absorbed rabbit antisera to 171 E. coli serotypes were used throughout the study. Strains that did not show agglutination with any of the typing sera were again grown in broth for 24 h and were then autoclaved for 2.5 h at 120°C to remove certain heat-resistant K antigens; agglutination assays were performed with antisera only against 08, 09, 020, and 0101 (12). All strains were titrated in twofold dilutions into the 0 serum or sera in question. If the titer was 640 or higher, the strain was assigned to that group. The strains were regis-
lence of a limited number of 0 and K antigens was found. The Kl capsule was identified most frequently. Subsequently, the clinical parameters of patients with bacteremia caused by E. coli strains without a capsule, with a non-0*
Corresponding author. 3174
VOL. 31, 1993
E. COLI SEROTYPES IN HUMAN
tered as nontypeable when they were autoagglutinable or did not show agglutination. K antigens were determined by agglutination of nonheated bacteria with the 0-antigen-typing serum that was found to agglutinate the boiled strain. Bacterial strains that failed to agglutinate because of the presence of a K antigen were serotyped with polyclonal-absorbed rabbit antisera directed against 73 different K antigens. Kl agglutination. Bacteria were grown overnight at 37°C on a blood agar plate to determine whether the E. coli bacteria contained 0-acetylated or non-O-acetylated Kl capsules. Then, a suspension of bacteria and ascitic fluid diluted 1:10, anti-O-acetylated Kl monoclonal antibody suspension, subclass immunoglobulin G2b (15), or 40 ,ul of antimeningococcal group B monoclonal antibody suspension (Wellcome Diagnostics, Dartford, England) was mixed on a glass slide at room temperature. When clumping occurred within 2 min, the result was considered positive. Suspensions of bacteria in phosphate-buffered saline served as negative controls. Definitions and criteria. The criteria used in the present study for determining bacteremia, sepsis, and severe sepsis were published by Bone et al. (1). Briefly, bacteremia was defined as the presence of viable bacteria in the blood. Sepsis was the systemic response to infection, manifested by two or more of the following conditions as a result of infection: temperature, >38°C or 90 beats per minute; respiratory rate of >20 breaths per minute (PACO2), 12,000/mm3, 10% immature (band) forms. Severe sepsis was defined as sepsis associated with organ dysfunction and hypotension. APACHE-II scores were used to study the acute physiology and to evaluate the chronic health of the patients. These scores were based on three different criteria, namely, acute physiology score, age points, and health points (5), and were calculated within 24 h of determining the severity of the bacteremia. The length of stay in the hospital was determined from the day that blood for culture was drawn from the patient until the day that the patient was discharged from the hospital. Data analysis. Frequencies of characteristics were compared between E. coli subgroups by using relative risks and 95% confidence intervals (CIs). For continuous variables, t-test-based confidence intervals were applied. APACHE-II scores were compared by the Mann-Whitney U test and linear regression to adjust for age differences. Mortality rates were compared with hazard ratios calculated from the Cox proportional hazards model. RESULTS
Serotyping of E. coli strains from human blood. (i) 0-antidistribution. Of the 174 strains isolated from blood, 143 (82%) were typeable with 0-specific antisera. The distributions of the 0 antigens and capsular antigens of the strains are presented in Table 1. Thirteen percent (22 of 174) of the strains belonged to the 06 antigen type, while 11% (19 of 174) possessed either the 018 or the 01 antigen. Nine percent (15 of 174) of the isolates had the 02 antigen. Fifty percent of strains were agglutinated by at least one of the six O groups (06, 018, 01, 02, 08, and 015), which were arbitrarily designated major 0 groups. (ii) Capsule distribution. The distributions of the different capsule antigens are shown in Tablg 2. A total of 80% (118 of 148) of the E. coli strains were encapsulated; 65% (77 of 118) of these strains were typeable with K-specific antisera.
gen
BLOOD
3175
TABLE 1. Distribution of 0 antigens and K antigens of 174 E. coli strains isolated from human blood arranged by frequency 0 antigen
06
No. (%) of isolates
22 (12.6)
Capsule antigen
(no. of isolates) K- (1)
K2 (4) K13 (5) K15 (2) K53 (6) K? (4) 19 (10.9)
018
K- (7)
Kl (4) K5 (2) K? (6) 19 (10.9)
01
K- (3)
Kl (13) K? (3) 02
15 (8.6)
08
7 (4.0)
K- (1) K(A)27 (1) K(A)28 (2) K(A)? (3)
015
6 (3.4)
K- (1) K? (5)
083
6 (3.4)
K- (2) Kl (3) K? (1)
07
6 (3.4)
K- (1)
Kl (6) K2 (1) KS (4) K7 (4)
Kl (4) K? (1) 025
5
(2.9)
K- (1)
K13 (1) K23 (1) K? (2) 4 (2.3)
04
K2 (1)
K53 (1) K? (2) 023
3
(1.7)
K- (2) K18 (1)
09
3
(1.7)
016
3 (1.7)
092
2 (1.1)
013
2 (1.1)
K(A)28 (1) K(A)32 (1) K(A)? (1) K- (1) Kl (2) K- (1) K? (1) K- (2)
Other
0
groups
Not typeable with 0-specific antisera
Total I
ND, not determined.
21 (12.1)
31 (17.8)
(7) (1) (1) (12) Kl (5)
KK2 K7 K?
ND- (26) 174
3176
J. CLIN. MICROBIOL.
FRASA ET AL.
TABLE 2. Distribution of K antigens of 148 E. coli strains isolated from human blood arranged by frequency K antigen
Kl
No. (%) of ioae isolates
K .37 (25.0) K2 ....................................................................... 7 (4.7) K53 ...................................................................... 7 (4.7) KS ....................................................................... 6 (4.0) K13 ...................................................................... 6 (4.0) K7 ....................................................................... 5 (3.4) K(A)28 ................................................................. 3 (2.0) K15 ...................................................................... 2 (1.4) K18 ...................................................................... 1 (0.7) K23 ...................................................................... 1 (0.7) K(A)27 ................................................................. 1 (0.7) K,32 ...................................................................... 1 (0.7) K(A)? ................................................................... 4 (2.7) K? ....................................................................... 37 (25.0) K-........................................................................ 30 (20.3) Total
TABLE 3. Demographic characteristics of patients with nonencapsulated E. coli or E. coli Kl bacteremia
148
Characteristic
No. of patients
12
21 58 ± 16a 33
Age (yr) % Male
Underlying disease (no. [%] of patients) Cardiovascular Renal Metabolic/endocrine Neurological Bronchopulmonary Oncologic Hepatic and biliary Traumatic Pancreatic None
Most encapsulated isolates (31%; 37 of 118) were of the Kl capsular polysaccharide type, that is, 21% (37 of 174) of all strains. Following in decreasing order of frequency were K2 (seven strains), K53 (seven strains), KS (six strains), K13 (six strains), K7 (five strains), K(A)28 (three strains), and K15 (two strains). The other types were relatively rare. Seventy-three of the 148 strains (49%) analyzed for capsular serotyping were agglutinated by only eight K types [Kl, K2, K53, KS, K13, K7, K(A)28, and K15], that is, 62% of the encapsulated strains. (iii) KI capsules. A differentiation could be made between 0-acetylated and non-0-acetylated K1 antigens. Of the Kl antigens that were determined, 57% (21 of 37) were non-0acetylated. The serotype 01K1 was most frequently isolated, namely, 9% (13 of 148). Of the 13 OlKl strains, 10 possessed the non-0-acetylated Kl capsule, while 3 possessed the 0-acetylated capsule. Patient evaluation: bacteremia with nonencapsulated or Kl strains. Patients suffering from bacteremia caused by nonencapsulated or K1 strains were studied and evaluated with regard to demographic characteristics (Table 3). Since there were no clinical data available for seven of the initial 30 patients infected with nonencapsulated strains, 23 patients were used for evaluation. Four neonates in the group infected with the Kl isolates were excluded from analysis, so that 33 patients were available for study. No significant differences existed between the groups with regard to age, sex, or APACHE-II scores (P = 0.46, Mann-Whitney U test). The urinary tract was the main source of the E. coli infection in both groups (Table 3). In Table 4 the severity of the bacteremia, the mortality of the patients, and the length of hospital stay are shown. Patients infected with E. coli K1 had a 1.27 (95% CI, 0.98, 1.64) times greater risk of developing sepsis than those infected with nonencapsulated E. coli. The relative risk of dying was 2.1 times greater (95% CI, 0.5, 9.5) in the group infected with K1 isolates than in the group infected with nonencapsulated isolates. Six patients in the group infected with Kl isolates died; three were infected with an isolate of the 02 serotype, two with an isolate of the 07 serotype, and one with an isolate of the 083 serotype. Subsequently, the complete group infected with isolates of the 02 serotype was studied. Only one
Non-O-acetylated 0-acetylated
APACHE-II score (no. [%] of patients)' , 0.8 0.6 D0 0
Xu 0.4
._
0.2
0
0
5
10
15
20
25
30
35
40
45
Days FIG. 1L. Mortality in the groups of patients infected with Kl isolates. The survival probabilities for patients infected with either 0-acetylaated E. coli Kl (*; n = 12) or non-0-acetylated E. coli Kl (@;n = 211) isolates are presented by Kaplan-Meier survival curves. The diffe rence between both groups was not significant (see text).
Patients ,in the group infectedwith 0-acetylated Ki isolates average 9.8 years older (95% CI, -1.2, 20.8) than those in the group infected with non-0-acetylated isolates. Patients with 0-acetylated E. coli bacteremia had cardiovascular di,sease as the underlying disease 2.2 times (95% CI, 0.7, 6.6) more often than patients with non-0-acetylated E. coli ba cteremia. APACHE-II scores were significantly higher itn the group infected with 0-acetylated isolates (P = 0.02, M: ann-Whitney U test). The acute physiology component of the APACHE-II score also differed significantly betweeni the two groups when age was taken into account (P = 0.046: ; linear regression model). Bacteremia originated 1.9 times (915% CI, 1.1, 3.4) more often in the urinary tracts of patients ,infected with 0-acetylated Kl E. coli isolates than it did in tihe urinary tracts of patients infected with a non-0acetylated Kl strain. The relative risk of developing severe sepsis vwas 2.6 times higher (95% CI, 0.5, 13.6) and the durationiof hospitalization of the survivors was 7 days longer (95% C] I, 2.0, 15.4) in the group infected with 0-acetylated isolates than in the group infected with non-0-acetylated isolates. Also, the mortality rate in the former group was higher: ]hazard ratio, 3.8 (95% CI, 0.7, 21.2) (Fig. 1).
were on
DISCUSSION In thi: s report, we described the serotyping of 174 E. coli strains iisolated from blood and provided clinical data for three paitient groups: those infected with nonencapsulated E. coli as vvell as those infected with encapsulated 0-acetylated and noni-O-acetylated E. coli Kl isolates. Six major 0 types (06, 0118, 01, 02, 08, and 015) representing 50% of the isolates,, were found. McGabe et al. (6) carried out a similar study o f 149 patients and found the following order of frequen cy of 0 antigens in adults: 06 (13%), 04 (8%), 02 (7%), 0 18 (6%), 016 (5%), 01 (5%), 07 (5%), 015 (4%), 025 (4%), 0 8 (3%), and 075 (3%). The 06 antigen was also the most pr evalent 0 antigen in our study, while the 04 antigen was not found as often. The other 0 antigens were identified at perce ntages comparative to those given above. Cross et al. (2) di etermined the serotypes of clinical isolates present in
3178
J. CLIN. MICROBIOL.
FRASA ET AL.
risk of dying compared with patients infected with nonencapsulated isolates. Thus, we could confirm that Kl strains are more virulent than nonencapsulated strains. This observation is not in agreement with a study by McGabe et al. (6), in which E. coli strains were related to sensitivity to serum and the severity of underlying disease. They found that K-antigen-containing strains did not appear to be associated with an enhanced severity of bacteremia. However, those researchers used only 18 K-antigen-typing serum samples compared with our 71, and thus, they may have missed some of the capsulated strains. 0rskov et al. (10) found that E. coli strains containing non-O-acetylated Kl antigens were less immunogenic in rabbits and therefore may be more virulent than 0-acetylated Kl strains. To investigate whether 0-acetylation of E. coli Kl is associated with severity of disease, we compared the clinical data for patients infected with 0-acetylated strains with data for patients infected with non-O-acetylated strains. The group infected with 0-acetylated isolates was older than the group infected with non-O-acetylated isolates and their APACHE-II scores were significantly higher (age points are included). Cardiovascular disease as the underlying disease was found more often in the group infected with 0-acetylated strains than in the group infected with non-Oacetylated strains. Most of the infections caused by 0-acetylated bacteria originated in the urinary tract. More patients with severe sepsis were found in the group infected with 0-acetylated isolates than in the group infected with non-Oacetylated isolates. Also, the duration of hospitalization was longer and the relative risk of dying was higher in the group infected with 0-acetylated isolates. In summary, the results of the present study indicate that the 0-acetylated Kl E. coli strains are associated with more severe disease than are the non-O-acetylated strains. This conclusion does not agree with that found in studies in animals, namely, that the 0-acetylated strains are more immunogenic and presumably less virulent than the non-Oacetylated strain. Thus, it would be worthwhile to confirm our observations with a larger number of clinical isolates. REFERENCES 1. Bone, R. C., R. A. Balk, F. B. Cerra, R. P. Dellinger, A. F. Fein, W. A. Knaus, R. M. H. Schein, and W. J. Sibbald. 1992. Definitions for sepsis and organ failure and guidelines for the use of innovative therapies in sepsis. Chest 101:1644-1655. 2. Cross, A. S., P. Gemski, J. C. Sadoff, F. 0rskov, and I. 0rskov. 1984. The importance of the Kl capsule in invasive infections caused by Escherichia coli. J. Infect. Dis. 149:184-193.
3. Cross, A. S., K. S. Kim, D. C. Wright, J. C. Sadoff, and P. Gemski. 1986. Role of lipopolysaccharide and capsule in the serum resistance of bacteremic strains of Escherichia coli. J. Infect. Dis. 154:497-503. 4. Guinee, P. A. M., C. M. Agterberg, and W. H. Jansen. 1972. Escherichia coli 0 antigen typing by means of a mechanized microtechnique. Appl. Microbiol. 24:127-131. 5. Knaus, W. A., E. A. Draper, P. Douglas, D. P. Wagner, and J. E. Zimmerman. 1985. APACHE II: a severity of disease classification system. Crit. Care Med. 13:818-829. 6. McGabe, W. R., B. Karser, S. Oiling, M. Uwaydah, and L. A. Hanson. 1978. Escherichia coli in bacteremia: K and 0 antigens and serum sensitivity from adults and neonates. J. Infect. Dis. 138:33-41. 7. McGowan, J. E., Jr., M. W. Barnes, and M. Finland. 1975. Bacteremia at Boston City Hospital: occurrence and mortality during 12 selected years (1935-1972), with special reference to hospital acquired cases. J. Infect. Dis. 132:316-335. 8. Opal, S. M., A. Cross, P. Gemski, and L. W. Lythe. 1988. Survey of purported virulence factors isolated of Eschenchia coli from blood, urine and stool. Eur. J. Clin. Microbiol. Infect. Dis. 7:425-427. 9. 0rskov, F., and I. 0rskov. 1975. Escherichia coli O:H serotypes isolation from human blood. Acta Pathol. Microbiol. Scand. 83:595-600. 10. 0rskov, F., I. 0rskov, A. Sutton, R. Schneerson, W. Lin, W. Egan, G. E. Hoff, and J. B. Robbins. 1979. Form variation in Escherichia coli Kl: determined by 0-acetylation of the capsular polysaccharide. J. Exp. Med. 149:669-685. 11. 0rskov, I., and F. 0rskov. 1985. Escherichia coli in extraintestinal infections. J. Hyg. Camb. 95:551-575. 12. 0rskov, I., F. 0rskov, B. Jann, and K. Jann. 1977. Serology, chemistry and genetics of 0 and K antigens of Eschenchia coli. Bacteriol. Rev. 41:667-710. 13. Schiffer, M. S., E. Oliveira, M. P. Glode, G. H. McCracken, L. M. Sarff, and J. B. Robbins. 1976. Relation between invasiveness and Kl capsular polysaccharide of Eschenichia coli. Pediatr. Res. 10:82-87. 14. Siitonen, A. 1992. Escherichia coli in fecal flora of healthy adults: serotypes, P and type 1C fimbriae, non-P mannoseresistant adhesins, and hemolytic activity. J. Infect. Dis. 166:
1058-1065. 15. Torensma, R., A. Van Wjk, M. J. C. Visser, A. Bouter, M. Rozenberg-Arska, and J. Verhoef. 1991. Monoclonal antibodies specific for the phase-variant 0-acetylated Kl capsule of Escherichia coli. J. Clin. Microbiol. 29:1356-1358. 16. Weinstein, M. P., J. R. Murphy, L. B. Reller, and K. A. Lichtenstein. 1983. The clinical significance of positive blood cultures: a comprehensive analysis of 500 episodes of bacteremia and fungemia in adults. II. Clinical observations, with special reference to factors influencing prognoses. Rev. Infect. Dis. 5:54-70.
