Host factors versus virulence-associated bacterial characteristics in neonatal and infantile bacteraemia and meningitis caused by Escherichia cdi

Transcription

1 J. Med. Microbiol. - Vol. 36 (1992), The Pathological Society of Great Britain and Ireland Host factors versus virulence-associated bacterial characteristics in neonatal and infantile bacteraemia and meningitis caused by Escherichia cdi K. TULLUS, ANNELIE BRAUNER", BlRGlTA FRYKLUNDt, T. MUNKHAMMARS, W. RABSCH, R. REISSBRODT and L. G. BURMANt Departments of Pediatrics, St Goran's Children3 Hospital, PO Box 12500, S Stockholm and SDanderyd Hospital, t The National Bacteriological Laboratory and the "Central Microbiological Laboratory, Stockholm, Sweden and lnstitut fur Experimentelle Epidemiologie, Wernigerode, Germany Summary. Possession of P fimbriae, virulence-associated 0 and K antigens, haemolysin and aerobactin production, and susceptibility to 10 antimicrobial agents were studied in 63 Escherichia coli strains isolated from blood or CSF of infants who were grouped according to their clinical characteristics. These isolates were compared with 35 faecal E. coli strains from healthy infants. Individual virulence factors showed a relatively weak association with invasive infection except for P fimbriae in urosepsis and aerobactin production in meningitis. Combinations of factors were generally more predictive for defining virulent clones, particularly in infants defined as being at normal risk of developing septicaemia. Thus, 62% of isolates from such infants had characteristics typical of previously described uropathogenic or meningitis-associated clones of E. cofi, compared with 32% of the isolates from high-risk infants (i.e., those defined as being at high risk of developing septicaemia) and only 9% of the faecal isolates (p<o.ool and <0-05, respectively). Overall, 45% of the episodes of invasive infection were caused by such clones, whereas risk factors (conditions considered to be associated with increased risk of invasive infection) were present in 59% of the infected infants (39% in meningitis and urosepsis, 78% in cryptogenic septicaemia and untreated bacteraemia). The results indicated that bacterial factors played a significant causative role in neonatal meningitis and urosepsis, particularly in normal-risk infants, whereas predisposing host factors contributed greatly to cryptogenic septicaemia and untreated bacteraemia. Introduction Invasive bacterial infection is an important problem in infants, particularly in neonates, and Escherichia coli continues to be one of the dominating pathogens in septicaemia and meningitis during the first year of life.'.' Bacterial infection is a consequence of a disturbed balance between the host defences and the colonising micro-organism. Infants most at risk from invasive bacterial infection, i.e., high-risk infants such as prematurely born babies, are not fully immunocompetent and are often exposed to invasive procedures in the hospital. Thus, they may become infected with bacteria of lower virulence than those which are able to attack healthy children. During the last decade our knowledge of the bacterial characteristics associated with virulence in E. coli has greatly increased. P firnbriae, certain 0 and K antigens, haemolysin and aerobactin production are typical of E. coli strains causing pyeionephriti~~-~ Received 19 Feb ; revised version accepted 12 June Downloaded from by and ba~teraemia,~-'* and such strains appear to belong to a limited number of widespread E. coli clones. 3-1 Clones causing meningitis seem to differ somewhat from those causing pyelonephritis. In previous studies of E. coli virulence in neonatal septicaemia and meningitis, the isolates have been examined without relating bacterial characteristics to the clinical status of the infected infants. In the present study, the virulence characteristics of invasive isolates of E. coli from infants were compared with those of isolates from the faecal flora of healthy infants. The presence of the virulence characteristics in the infecting strains was also assessed in relation to clinical data of the babies. Materials and methods Patients and isolates Of the 33 bacteriological laboratories in Sweden, 26 contributed recent E. coli isolates from the blood or CSF of infants. Clinical data were obtained from

2 204 K. TULLUS ET AL. hospital records. The 63 infants studied were assigned to one of four diagnostic categories : meningitis (organism in CSF), urosepsis (same organism in urine and blood), cryptogenic septicaemia (without a known focus of the invasive infection), and untreated bacteraemia (a child who had been regarded, by the pediatrician in charge of the child, as not having septicaemia and therefore left untreated) (table I). According to clinical criteria, the infants were grouped by one of us (KT) as infants either with high risk for developing septicaemia or with normal risk. The conditions considered to be associated with increased risk of invasive infection were prematurity (1 8), hyaline membrane disease (lo), severe asphyxia (4), premature rupture of the amniotic membranes with intra-uterine infection (4), small for gestational age (3), myelomenigocele (3), vesico-ureteric reflux grade I11 or more (2), Down s syndrome (2), leukaemia, exchange transfusion, incarcerated hernia, posterior urethral valve, Hirschsprung s disease, necrotising enterocolitis and a meningeal fistula (one case each). Many infants had several of these risk factors. Faecal isolates of E. coli were obtained from healthy infants aged 0-18 months and without a history of septicaemia or urinary tract infection. Bacteriological methods 0 serogroups were determined by bacterial agglutination with antisera to 690 antigens.20 As previously postulated, the 0 groups 1, 2, 4, 6, 7, 16, 18 and 75 were considered as virulent 0 groups.2 K antigens were typed by a modification of the serum agar technique in which 22 different antisera to capsular acidic polysaccharide K antigens were used.22 The K5 antigen was identified with a K5-specific phage.23 Isolates with K antigens 1,2,3,5,12,13 were regarded as virulent. Expression of P fimbriae was tested, after cultivation of isolates on colonisation factor antigen (CFA) agar, with the P-fimbriae-specific latex particle agglutination test (PPA-test). 24 Haemolysin production was determined after incubation overnight on blood agar plates containing Blood Agar Base No. 2 (Oxoid) and thrice-washed sheep red blood cells at a final concentration of 5% v/v. The production of aerobactin was determined with the test E. coli strain LG The strains were defined as uropathogenic if they expressed P fimbriae and haemolysin and shared 0 serogroup and K antigen with eight previously described uropathogenic clones. 8*2 Strains expressing the Kl antigen and of 0 groups 7, 16, 18, 83 or rough were defined as meningitis strains The infecting strains were tested for their susceptibility to ampicillin, cefotaxime, cefpirome, cefuroxime, ceftazidime, aztreonam, imipenem, cotrimoxazole, chloramphenicol and gentamicin by an agar dilution method as previously described. 29 MIC values of <2 mg/l for ampicillin, < 8 mg/l for chloramphenicol and co-trimoxazole and < 4 mg/l for the other antibiotics were used to define susceptible strains.30 Statistical analyses x2 test, Fisher s exact test and the Mann-Whitney test were used when appropriate. Results Clinical characteristics The isolates of E. coli from blood and CSF of infants were associated with meningitis (21 %), urosepsis (29%), cryptogenic septicaemia (40%) or untreated bacteraemia (1 1%) (table I). Children with cryptogenic septicaemia had lower gestational age, birth weight and Apgar score and developed their infection earlier than the other groups of patients with invasive infection (table I). Infants with untreated bacteraemia tended to have similar characteristics whereas infants with meningitis and urosepsis were more mature, healthier at birth, and older when they developed the infection. A majority of infants with cryptogenic septicaemia (80%) or untreated bacteraemia (7 1 %) were classified as high-risk infants, compared to 46% and 33% of infants with meningitis and urosepsis, respectively. Of the patients in the cryptogenic and the untreated groups, 25% and 29%, respectively, had received antibiotics before the infection studied. Those who had received antibiotic treatment had a lower birth weight (average 948 g uersus 3099 g; Mann-Whitney test, p <0.001) and a higher proportion of risk factors (100% versus 52%; Mann-Whitney test, p<o.ol) than the other infants. Virulence-associated characteristics 0 serogroups and K antigens reported to be associated with virulence showed similar prevalences amongst isolates from clinical infection and faeces but were less prevalent in the untreated bacteraemia group (Fisher s exact test, p < 0-05 ; table 11). The K 1 antigen was identified in 69% of the meningitis isolates. P fimbriae were expressed more often by urosepsis isolates than by the other invasive or faecal isolates, but were found in none of the isolates from the untreated infants (table 11). Haemolysin and aerobactin production were least prevalent amongst isolates from the untreated group (table 11). Aerobactin production was associated with isolates from meningitis (table 11). P-fimbriate strains produced aerobactin more often than P-negative strains (78% versus 48%; x2 test, p < 0.05) whereas aerobactin production was evenly distributed amongst haemolysin-positive and -negative strains (58% and 58%, respectively).

3 ~ E. COLI BACTERAEMIA 205 Table I. Host factors in the different diagnostic groups Host factors Meningitis Numbers, median values (range) or percentages in each of the diagnostic groups Urosepsis Cryptogenic septicaemia Untreated bac teraemia Girls/ boys Age (days) at infection Gestational age (weeks) Birth weight (g) Apgar score at 5 min Percentage of highrisk infantst Percentage of patients receiving antibiotics before infection (1-240) 40 (27-42) 3040 ( ) 9 (5-10) 46% 0% 5/13 18 (5-436) 40 (37-42) 3480 ( ) 9 (8-10) 33% 0% 6/19 4/ 3 4* (0-134) 5 (1-40) 33* (25-41) 37 (25-42) 1900* 2720 (81 M 910) ( ) 7* (0-10) 8 (4-10) 80%* 71% 25%* 29% *Mann-Whitney test, pl0.01 versus the other isolates.?for definition of a high-risk infant see Materials and methods. Virulent clones Twenty-five isolates (40%) were classified as virulent strains. Uropathogenic strains were more common in urosepsis than in the other diagnostic groups and meningitis-associated strains were found in meningitis and a few cases of cryptogenic septicaemia (table 11). The serotypes of the uropathogenic strains found in this study were: 02:Kl:P+ :Hly-, 04:K3:P+ :Hly+, 06:K5:P+Hly+, 07:Kl:P+: Hly-, 016:Kl:P+ :Hly+ and 018:K5:P+Hly+. Both categories of virulent strains produced aerobactin more often than the other strains (85% and 84% versus 42%; x2 test, p < 0.001). Co-expression of virulence characteristics Isolates from urosepsis or cryptogenic septicaemia exhibited three or more virulence characteristics more often amongst those from normal-risk infants than amongst those from high-risk infants and faecal isolates (71% versus 39% and 43%, respectively; Fisher s exact test, p < 0.05 ; table 111). The isolates from the untreated infants were too few to split into two groups but they never exhibited more than two virulence characteristics (Fisher s exact test, p < 0.05 compared to faecal isolates). Normal-risk versus high-risk infants The occurrence of virulent 0 serogroups and K antigens and aerobactin production was similar amongst isolates from infections in infants with or without risk factors and those from the faecal flora (table IV). Expression of P fimbriae was twice as common amongst isolates from infants without appar- ent risk factors than amongst isolates from high-risk infants and faeces (table IV). Isolates from normalrisk infants with urosepsis, cryptogenic septicaemia or meningitis expressed P fimbriae in 83% 40% and 29% of instances, respectively, whereas isolates from high-risk infants in these diagnostic groups were P fimbriate in 17%, 30% and 33% of instances, respectively (Fisher s exact test, p<o.o5, p>o.o5 and p> 0.05, respectively). Haemolysin production was less common amongst isolates from normal-risk infants than amongst those from high-risk infants or faeces (table IV). Virulent strains caused 62% of the infections in normal-risk infants, 32% of the infections in high-risk infants (Fisher s exact test, p < 0-09, but represented only 9% of the faecal isolates (Fisher s exact test, p < ; table IV). Of the infecting strains, 24% were resistant to ampicillin, 14% to co-trimoxazole, 5% to cefuroxime, 5% to chloramphenicol and none to aztreonam, cefotaxime, cefpirome, ceftazidime, gentamicin or imipenem. All but one of the 10 isolates with antimicrobial multiresistance were collected at a referral centre. Antibiotic resistance correlated neither with virulence characteristics nor host factors. Discussion The present results indicated that bacterial factors played a significant causative role in neonatal meningitis and urosepsis, particularly in normal infants, whereas predisposing host factors contributed greatly to cryptogenic septicaemia and untreated bacteraemia.

4 ~~ 206 K. TULLUS ET AL. Table 11. Distribution of virulence characteristics among 63 isolates of E. coli from blood or CSF in relation to clinical diagnosis, and 35 faecal isolates Virulence characteristic Percentage of isolates from cryptogenic untreated meningitis urosepsis septicaemia bacteraemia (13) (18) (25) (7) faeces (35) 0 serogroups 1,2,4,6,7, 16, 18, 75 K antigens 1,2,3, 5, 12, 13 P fimbriae Haemolysin Aerobactin Virulent isolate$ uropathogenic meningitis O* * * * 60 77* * * 0 0 * Fisher s exact test, p < 0.05 compared to faecal isolates. t Fisher s exact test, p < 0401 compared to faecal isolates. $ For definition see Materiafs and methods. Table 111. Relationship between the number of bacterial virulence characteristics, diagnostic group and risk status of the infants Number of isolates from Number of virulence characteristics per isolate None One Two Three Four Five Median number of characteristics cryptogenic meningitis urosepsis septicaemia untreated bacteraemia faeces HRI NRI HRI NRI HRI NRI (7) (35) (6) (7) (6) (12) (20) (5) HRI, high-risk infant; NRI, normal-risk infant. However, the association between the classical virulence-associated 0 and K antigens and invasive infection was weak in the present study, partly due to an unusually high prevalence of these antigens amongst the faecal isolates. Such 0 serogroups occurred at prevalences of 3656% in the different diagnostic groups which is similar to published data from adults and children with septicaemia (36-83%) - and for urinary isolates from children with vesico-ureteric reflux cystitis and asymptomatic bacteriuria (8-59%), but lower than for isolates from pyelonephritis in children (58-80%).491 The virulence-associated K antigens have been found previously in 41-83% of blood is~lates~* ~ and 44% of faecal isolates.31 Although these studies did not determine the presence of the K5 antigen, the percentages found here were of the same order. The prevalence of the Kl antigen in meningitis isolates (69%) was somewhat lower than previously reported. 18~26-28 P fimbriation was more common amongst blood isolates from infants with urosepsis without risk factors (83%) than amongst the other blood or CSF isolates (049%) and the faecal isolates (29%). This is logical as P fimbriation is a virulence factor presumed to be of importance mainly in the urinary tract. Previous studies have shown P fimbriation (or mannose-resistant haemagglutination) in 62-74% of urosepsis isolates,**12 in 2-15% of isolates from cryptogenic or unspecified ba~teraemia,~.~~ and in 7% of neonatal meningitis isolates. Haemolysin production was surprisingly common amongst our faecal strains, 60% versus 5-1 0% in strains from other sources,1 1932i33 whereas our data for

5 ~~ E. COLI BACTERAEMIA 207 Table IV. Relationship between virulence characteristicsof the infecting strain and clinical characteristics of the infants with invasive infection (isolates from untreated bacteraemia are excluded) Percentage of isolates from Virulence characteristic blood or CSF high-risk normal-risk infants infants faeces (35) 0 serogroups 1,2,4,6,7,16,18,75 K antigens 1,2,3,5, 12, 13 P fimbriae Haemolysin Aerobactin Virulent isolates uropathogenic meningitis Non-virulent * *t t 29t $ *Fisher s exact test, p < 0-05 compared to isolates from normal-risk infants. TFisher s exact test, p < 0.05 compared to faecal isolates. $Fisher s exact test, p < compared to faecal isolates. invasive isolates were closer to those published previously; 36% versus 25% in neonatal septicaemia, and 50% in urosepsis versus 60-72% in pyelonephrito- Aerobactin genic urine isolates from children. 934 production occurred at frequencies similar to those previously reported for bacteraemic strains (56-77% versus %). * Combinations of factors were generally more predictive for defining virulent clones than the single factors, particularly in normal-risk infants. Thus, 62% of their isolates had characteristics typical of previously described uropathogenic or meningitis clones of E. coli compared to 32% of isolates from high-risk infants and only 9% of the faecal isolates. Previous studies have reported that uropathogenic clones comprise 43% of urine isolates from children with acute pyelonephritis 8 and 29% of isolates from neonatal septicaemia. To our surprise, as many as seven cases of bacteraemia had, by the clinician in charge of the child during hospitalisation, been regarded as not relevant for the infant s clinical condition and left untreated. Five of these seven cases occurred in highrisk infants. The isolates were unusual, also, in lacking nearly all of the virulence-associated factors and none belonged to a virulent clone. These isolates seem to represent a selection of low-virulent faecal strains that, even amongst high-risk infants, had been able to cause only transient bacteraemia. The invasive isolates rarely showed antibiotic multiresistance. The rate of ampicillin resistance (24%) was similar to that previously found in faecal E. coli from neonates in 22 Swedish neonatal wards (26%). The present study emphasises the growing importance of considering host factors when evaluating possible virulence factors in bacteria. This has not been done in previous studies of neonatal E. coli septicaemia. Conflicting data regarding the prevalence, and thus the significance, of virulence factors may be explained by differences between the patient populations studied. Our data also support the view that E. coli isolates causing bacteraemia in compromised patients are of lower virulence than other bacteraemic isolates,** similar to the strains which cause pyelonephritis in children with anatomical risk factors such as reflux. Recognising the relative importance of bacterial virulence and impaired host defence has significant implications for improved prevention of invasive gram-negative infections in the neonate. Thus, attempts to reduce colonisation with virulent E. coli strains in neonatal units would, during non-epidemic periods, potentially influence only half (in this study 45%) of the episodes of E. coli septicaemia. Current preventive measures directed at the individual infant at risk and future methods to enhance host defences will probably be of growing importance for the prevention of these neonatal diseases. This study was supported by grant B88-16X A from the Swedish Medical Research Council, and grants from the Swedish Society for Medical Sciences, the First May Flower Campaign for Children s Health, the Karolinska Institute, the General Maternity Hospital Foundation and the Samariten Foundation. We thank the staffs of the participating Departments of Pediatrics and Bacteriology for their kind interest and contribution to the study and Professor B. Kaijser for serotyping the bacterial isolates.

6 ~ 208 K. TULLUS ET AL. References 1. Bennet R, Eriksson M, Melen B, Zetterstrom R. Changes in the incidence and spectrum of neonatal septicaemia during a fifteen-year period. Acta Paediatr Scand 1985; 74: Vesikari T, Janas M, Gronroos P et af. Neonatal septicaemia. Arch Dis Child 1985; 60: Kallenius G, Miillby R, Svensson SB et a/. Occurrence of P- fimbriated Escherichia cofi in urinary tract infections. Lancet 1981 ; 2 : Lindberg U, Hanson LA, Jodal U et al. Asymptomatic bacteriuria in schoolgirls. 11. Differences in Escherichia cofi causing asymptomatic bacteriuria. Acta Paediatr Scand 1975; 64; Kaijser B, Hanson LA, Jodal U, Lidin-Janson G, Robbins JB. Frequency of E. cofi K antigens in urinary-tract infections in children. Lancet 1977; 1 : Hughes C, Hacker J, Roberts A, Goebel W. Hemolysin production as a virulence marker in symptomatic and asymptomatic urinary tract infections caused by Escherichia coli. Infect Immun 1983; 39: Jacobson SH, Tullus K, Wretlind B, Brauner A. Aerobactinmediated uptake of iron by strains of Escherichia cofi causing acute pyelonephritis and bacteremia. J Infect 1988 ; 16: Brauner A, Svenson SB, Wretlind B, Julander I, Kallenius G, Leissner M. Occurrence of P-fimbriated Escherichia cofi in patients with bacteremia. Eur J Clin Microbiof 1985; 4: McCabe WR, Kaijser B, Olling S, Uwaydah M, Hanson LA. Escherichiacofiin bacteremia: K and 0 antigens and serum sensitivity of strains from adults and neonates. J Infect Dis 1978; 138: Young LS, Stevens P, Kaijser B. Gram-negative pathogens in septicaemic infections. Scand J Infect Dis 1982; Suppl31: Korhonen TK, Valtonen MV, Parkkinen J et af. Serotypes, hemolysin production, and receptor recognition of Escherichia coli strains associated with neonatal sepsis and meningitis. Infect Immun 1985; 48: Johnson JR, Moseley SL, Roberts PL, Stamm WE. Aerobactin and other virulence factor genes among strains of Escherichia cofi causing urosepsis : association with patient characteristics. Infect Immun 1988; 56: rskov F, Orskov I. Summary of a workshop on the clone concept in the epidemiology, taxonomy, and evolution of the Enterobacteriaceae and other bacteria. J Infect Dis 1983; 148: Achtman M, Mercer A, Kusecek B et af. Six widespread bacterial clones among Escherichia coli K 1 isolates. Infect Immun 1983; 39: Ochman H, Selander RK. Evidence for a clonal population structure in Escherichia coli. Proc Natl Acad Sci USA 1984; 81: Eisenstein BI, Jones GW. The spectrum of infections and pathogenic mechanisms in Escherichia cofi. Adv Intern Med 1988 ; 33: Achtman M, Pluschke G. Clonal analysis of descent and virulence among selected Escherichia cofi. Annu Rev Microbiof 1986,; 40: Vaisanen-Rhen V, Elo J, Vaisanen E et af. P-fimbriated clones among uropathogenic Escherichia cofi strains. Infect Immun 1984;43: Lomberg H, Hellstrom M, Jodal U, Leffler H, Lincoln K, Svanborg Edh C. Virulence-associated traits in Eseherichia cofi causing first and recurrent episodes of urinary tract infection in children with or without vesicoureteral reflux. J Infect Dis 1984; 150: Lidin-Janson G, Falsen E, Jodal U, Kaijser B, Lincoln K. Characteristics of antibiotic-resistant Escherichia cofi in the rectum of healthy school-children. J Med Microbiol 1977; 10: Tullus K. Fecal colonization with P-fimbriated Escherichia coli in newborn children and relation to development of extraintestinal Escherichia cofi infections. Acta Paediatr Scand 1987; Suppl334: Kaijser B. A simple method for typing of acidic polysaccharide K antigens of Escherichia cofi. FEMS Microbiol Lett 1977 ; Kaijser B, Jodal U. Escherichia cofi K5 antigen in relation to various infections and in healthy individuals. J Clin Microbiol 1984; 19: Svenson SB, Kallenius G, Mollby R, Hultberg H, Winberg J. Rapid identification of P-fimbriated Escherichia coli by a receptor-specific particle agglutination test. Infection 1982 ; 10: Reissbrodt R, Rabsch W. Further differentiation of Enterobacteriaceae by means of siderophore-pattern analysis. Zentralbl Bakteriol Mikrobiof Hyg A 1988 ; 268 : 30& Robbins JB, McCracken GH, Gotschlich EC, 0rskov F, 0rskov I, Hanson LA. Escherichia coli K1 capsular polysaccharide associated with neonatal meningitis. N Engf J Med 1974; 290: Sarff LF, McCracken GH, Schiffer MS et al. Epidemiology of Escherichia coli K1 in healthy and diseased newborns. Lancet 1975; 1: Mulder CJJ, van Alphen L, Zanen HC. Neonatal meningitis caused by Escherichia cofi in the Netherlands. J Infect Dis 1984; 150: Tullus K, Olsson-Liljequist B, Lundstrom G, Burman LG. Antibiotic susceptibility of 629 bacterial blood and CSF isolates from Swedish infants and the therapeutic implications. Acta Paediatr Scand 1991 ; 80: The Swedish Reference Group for Antibiotics. A revised system for antibiotic sensitivity testing. Scand J Infect Dis 1981 ; 13: Kaijser B, Hanson LA, Jodal U, Lidin-Janson G, Robbins JB. Frequency of E. coli K antigens in urinary-tract infections in children. Lancet 1977; 1 : Opal SM, Cross A, Gemski P, Lyhte LW. Survey of purported virulence factors of Escherichia coli isolated from blood, urine and stool. Eur J Clin Microbiol Infect Dis 1988; 7: Cooke EM, Ewins SP. Properties of strains of Escherichia coli isolated from a variety of sources. J Med Microbiof 1975; 8: Brauner A, Katouli M, Tullus K, Jacobson SH. Production of cytotoxic necrotizing factor, verocytotoxin and hemolysin by pyelonephritogenic Escherichia coli. Eur J Cfin Microbiol Infect Dis 1990; 9: Tullus K, Berglund B, Fryklund B, Kuhn I, Burman LG. Influence of antibiotic therapy on faecal carriage of P- fimbriated Escherichia cofi and other gram-negative bacteria in neonates. J Antimicrob Chemother 1988; 22: