Vol. 22, No. 2
INFECTION AND IMMUNITY, Nov. 1978, P. 480-485 0019-9567/78/0022-0480$02.00/0 Copyright i) 1978 American Society for Microbiology
Printed in U.S.A.
Dynamics of Escherichia coli Infection and Meningitis in Infant Rats ROBERT BORTOLUSSI,t PATRICIA FERRIERI,' * AND LEWIS W. WANNAMAKER1 2 Departments of Pediatrics' and Microbiology,2 University of Minnesota, Minneapolis, Minnesota 55455 Received for publication 31 August 1978
Escherichia coli strains isolated from newborn infants were injected intraperitoneally into infant rats. Strains possessing the Kl capsular polysaccharide antigen were significantly more virulent than strains lacking this antigen. When 5-day-old animals were injected with 1.2 x 101 colony forming units of a Kl E. coli strain (serotype 018ac:Kl:H7), about 80% had bacteria isolated from their blood. Forty-eight percent of bacteremic animals had positive cerebrospinal fluid cultures. The development of bacteremia with >10i colony-forming units per ml of blood correlated with positive cultures of cerebrospinal fluid. Some animals, studied with serial blood cultures, were able to clear bacteria spontaneously from their blood, whereas others succumbed to infection within 48 h of challenge. The susceptibility of infant rats to E. coli infection was age dependent and appeared related to the Kl antigen. Escherichia coli has been associated with a wide variety of human acute infections. Recent studies have shown that a high prevalence of E. coli strains recovered from newborn infants with meningitis possessed the Kl capsular polysaccharide antigen (5, 10, 12, 13). The Kl antigen of E. coli is a polysaccharide of sialic acid and is immunologically and biochemically identical or virtually identical to the capsular polysaccharide of group B meningococcus (6). These findings suggest that the Kl capsular antigen may be related to virulence of E. coli strains, particularly for the newborn infant. In attempting to compare virulence of Kl and non-Kl E. coli in an animal model, Wolberg and DeWitt in 1969 (15) and Robbins et al. in 1974 (10) found that Kl E. coli strains suspended in hog gastric mucin and injected intraperitoneally (i.p.) into adult mice were more virulent than non-Kl strains. Because the mechanism of action of hog gastric mucin is poorly understood, the relationship of E. coli with or without Kl capsular antigen to virulence is unclear. In 1977 Glode et al. described a model for E. coli meningitis using infant rats fed with Kl E. coli strains isolated from infants with meningitis (4). They were able to demonstrate that bacteremia and meningitis occurred more frequently in animals fed Kl than non-Kl E. coli strains. In this model intralitter transmission to water-fed controls was common, and the incidence of meningitis was relatively low. The time of onset of t Present address: I. W. Killam Children's Hospital, Halifax, Nova Scotia, Canada.
extraintestinal infection was not known, and studies of dynamics of bacterial spread were not possible. We have examined the effect on infant rats of i.p. injection of different strains of E. coli. Rats were chosen because of the extensive literature available on immune responsiveness in the newborn period and because they are well-adapted to handling at this time (1, 14). The use of adjuvants such as hog gastric mucin to enhance virulence was not found to be necessary. Introduction of bacteria into the peritoneum allowed us to assess the newborn rat's ability to resist Kl E. coli infection independent of intestinal colonization and to follow the dynamics of bacterial multiplication and spread. Intralitter transmission to controls was rare. Our results indicate that rats have an age-dependent susceptibility and that there is an association of virulence with Kl E. coli strains. MATERIALS AND METHODS Animals. Outbred, pathogen-free albino SpragueDawley pregnant rats were obtained from Bio-Lab Corp. (White Bear Lake, Minn.). Animals were housed under standardized conditions (250C; relative humidity, 40%) with a 7 a.m. to 7 p.m. light schedule. Purina rat chow and water were available ad libitum. Rat pups and their mothers were housed in solid polypropylene opaque cages with filter hoods made from spunbonded polyester (Lab Products Co., Garfield, N.J.) to prevent cross-infection by aerosolization. Care was taken to minimize trauma and separation of animals from their mothers. Rat pups from multiple litters were used in each phase of the study to minimize the effect of interlitter variability.
480
VOL. 22, 1978 Bacteria. Twenty-seven strains of E. coli isolated from cerebrospinal fluid (CSF), blood, urine, and stools were tested (Table 1). Strains were obtained from C. Krishnan (Department of Bacteriology, Hospital for Sick Children, Toronto, Canada), J. B. Robbins (Division of Bacterial Products, Bureau of Biologics, Food and Drug Administration, Bethesda, Md.), D. Blazevic (Diagnostic Bacteriology Department, University of Minnesota, Minneapolis, Minn.), and B. Bjorksten (University of Umea, Sweden). The opsonic requirements and sensitivity to bactericidal activity of serum of many of these strains have been described (2). One strain, serotype 018ac:Kl:H7 (isolated from the CSF of an infant), designated C5 in an earlier report (2), was studied in detail. All Kl E. coli strains were identified by an agarose halo technique (12). Serotyping was performed by F. 0rskov (Seruminstitut, Copenhagen, Denmark) and B. Bj6rksten. Of the 27 strains, 21 were serotyped: 018ac:Kl:H7 (3 strains), 07:K1:H? (3 strains), 016:K1:H6 (1 strain), SpAg:Kl:H7 (1 strain), SpAg:Kl:H34 (1 strain), 022:K1 (2 strains), 02:K1:H6 (1 strain), 07:H- (2 strains), 083:H- (1 strain), 015:K7:H- (1 strain), SpAg:H7 (1 strain), 09:K36A (1 strain), SpAg:H4 (2 strains), and 0137:H11 (1 strain). Two Kl and four non-Kl E. coli strains were of unknown 0 serotype. Strains were grown in brain heart infusion broth (Difco Laboratories, Detroit, Mich.) for 8 h, divided into small portions with glycerol (final concentration, 20%), rapidly frozen, and stored at -20°C. When needed, vials of frozen bacteria were thawed, inoculated into brain heart infusion broth, and grown at 35°C. After overnight incubation, the bacteria were suspended in brain heart infusion broth and diluted to an optical density of 0.06 at 460 nm by using a Coleman Junior II spectrophotometer (Coleman Instruments Division, Oak Brook, Ill.). A 102 dilution was prepared in fresh media from this. After incubation at 35°C for 2 h, the concentration of bacteria was 2.0 ± 0.5 x 107 colony-forming units (CFU) per ml (confirmed by pour-plate technique in triplicate). The broth suspension was diluted in 0.85% NaCl to the desired number of CFU for i.p. injection by using a final volume of 0.1 ml. Mortality after i.p. injection. No virulence-enhancing agents were employed in any of these studies. The virulence of 27 strains of E. coli for 5-day-old rats was assessed by studying mortality after i.p. injection of 5 x 104 bacteria. This inoculum was selected when initial studies showed that the median lethal dose (LD50) of most Kl E. coli strains was less than 103 organisms, while the LD50 of non-Kl E. coli strains was greater than 105. One inoculum was used to allow a rapid comparison of several E. coli strains. Animals from three litters were injected with each strain. Care was taken to compare Kl and non-Kl E. coli strains isolated from comparable sources in order to avoid selecting more virulent strains in one group. Animals were injected between 1800 and 1900 h, and observations were made every 12 h for the first 48 h and every 24 h thereafter. A total of 135 5-day-old rats were observed for 5 days after i.p. injection of 5 x 104 bacteria. Fourteen Kl and 13 non-Kl E. coli strains were tested (each strain was tested in five animals). Seventy-three control animals from these litters were
E. COLI MENINGITIS IN INFANT RATS
481
TABLE 1. Source of E. coli strainsa
a
Source
Kl
Blood and CSF Stool Urine
10 4 0
Non-Kl 7 4 2
Numbers indicate the number of strains isolated.
injected i.p. with saline or a sterile broth filtrate of the test strain. The relationship of animal age to susceptibility to Kl E. coli was examined by determining the LD5o of the Kl E. coli strain, C5, in 5-, 12-, and 19day-old and adult animals by the method of Reed and Muench 72 h after inoculation (9). Between 20 and 30 animals were used for each LD50 determination. Sampling of body fluids and organs. The Kl E. coli strain, C5, was used to follow the dynamics of infection. In this study infant rats were injected i.p with 0.8 x 10' to 1.6 x 10' bacteria. This low inoculum was chosen in order to follow infection after minimal bacterial invasion of the peritoneum. The animals were weighed daily, observed at frequent intervals, and sacrificed at a predetermined time by a lethal dose of ether. After exsanguination, CSF was obtained and the brain was removed, weighed, washed, and homogenized in cold saline. The spleen was handled in a similar manner. Pour plates were prepared from serial dilutions of homogenates of these organs and results expressed as CFU/g of tissue. Blood obtained by cardiac puncture and peritoneal cavity washings was diluted in heparinized cold saline and bacterial counts were determined. A wet count and differential of peritoneal washings were done with a Neubauer hemocytometer by using gentian violet as a nuclear stain. Because this method did not allow differentiation of cell granules, peritoneal cells were designated simply as multilobed or monolobed, based on microscopic appearance. CSF was obtained by cisternal puncture using a method similar to one previously described (8). CSF was drawn up by capillary action into a 10-1l micro-
pipette (Micropipet Disposable Pipettes, Clay Adams Division of Becton, Dickinson & CO., Parsippany, N.J.). Ten microliters of CSF and dilutions of it were inoculated onto brain heart infusion agar. Colonies were counted, and results were expressed as CFU per ml CSF. For serial quantitative blood cultures, the tail was cleansed with 70% acohol and amputated. Blood was collected in 5-p4 calibrated microcapillary pipettes and inoculated onto MacConkey agar (Difco Laboratories). Blood was obtained from 24 animals at 8, 24, 48, 72, and 96 h after i.p. injection or until the animal died. Bacteria recovered from animals were identified on the basis of colonial morphology, growth on MacConkey agar, and by demonstrating rapid agglutination of three to four colonies with anti-meningococcal group B horse antiserum diluted 1:5 (provided by J. B. Robbins). To establish that blood obtained from the tail vein accurately reflected the number of bacterial CFU in blood, samples were obtained simultaneously from the heart and tail in 25 other animals. Quantitative bacterial counts were within 1 log (or
482
BORTOLUSSI, FERRIERI, AND WANNAMAKER
were both negative) in the range of 102 to 105 bacteria per ml of blood. Statistical methods. The chi-square test was used to determine the significance of differences among groups over 20. When comparing groups of 20 or fewer animals, tables based on the Fisher exact test were used (11).
INFECT. IMMUN.
or control animals (12, 14, and 15%, respec-
tively).
The LD50 of strain C5 was determined on 5-, 12-, and 19-day-old and adult animals 72 h after i.p. injection of bacteria (Fig. 2). There was an increase in resistance to Kl E. coli infection beyond 5 days of age. Adult animals tolerated i.p. injection with 105 bacteria; however, when RESULTS injected with 106 or 107 bacteria, they appeared Mortality after i.p. injection: relation to overwhelmed with infection and died within 24 capsular type and age of animals. Survival h of challenge. In contrast, animals 5 to 19 days of Kl and non-Kl E. coli-injected animals and old challenged with bacteria at a concentration control animals is shown in Fig. 1. There was a near the LD50 for their age appeared well 12 to significant difference (P < 0.05) in survival of 24 h after i.p. injection and died only after 24 to Kl E. coli-injected animals compared to non-Ki 48 h. Weight ranged from 11 to 40 g for 5- to 19E. coli-injected or control animals 24 to 72 h day-old animals and from 300 to 400 g for adult after injection. Animals injected i.p. with non- animals. Bacterial multiplication and disseminaKl E. coli strains did not differ significantly in survival from controls (P > 0.05). Mortality tion of infection. To determine the progression rates in the two control groups (broth filtrate, of bacteremia in newborn animals, serial blood 12%; saline, 16%) were similar and have been cultures were taken from the tail vein of 24 5combined. Among the Kl strains, the two spon- day-old rats after i.p. injection of 1.2 x 101 CFU taneously agglutinating strains were associated of a Kl E. coli strain (C5). Actual CFU by pour with lower mortality than the 10 strains possess- plate varied ±0.4 x 10'. Overall, 19 of 24 animals ing a somatic antigen (20 versus 74%). However, (79%) developed bacteremia, and 17 animals the number of animals injected with the spon- (71%) died from 1 to 5 days after i.p. injection. taneously agglutinating Kl strains was small and The time of development and the severity of precludes a valid comparison. bacteremia were variable in these animals (Fig. Some animals dying later than 72 h after i.p. 3). The presence of bacteremia at 8 h was ordichallenge were found to have sterile postmortem narily predictive of rapid bacterial multiplication blood cultures, whereas all of the animals who to concentrations of >10' CFU/ml of blood and died within 72 h had positive cultures. After 72 was related significantly to the subsequent h, the number of new deaths was not different course. Of 11 animals who died within 48 h of among Kl and non-Kl E. coli-injected animals i.p. injection (group I), 10 were bacteremic at 8 h, while only 2 of 6 animals who died after 2 days (group II) and 1 of 7 animals who survived to 7 days (group III) were bacteremic at 8 h. This difference in incidence of bacteremia was significant (group I versus group II plus group 100
>
106
80
>
w~~~~
F
60
X
40
-
io4
z
I,,G _
w/ 3 /
a.
I 10
10
12
24
36 48 TIME (hr)
60
72
FIG. 1. Percentage of newborn rats surviving after i.p. injection of 5 x 104 CFU of Kl E. coli (0), non-Kl E. coli (A), and saline or sterile broth filtrate controls (0).
l 5 5
9'
/
12
19
ADULT
AGE (DAYS)
FIG. 2. Relationship of virulence of Kl E. coli to age of animal as demonstrated by change in LD50 determined 72 h after i.p. injection in 5-, 12-, or 19day-old or adult animals.
VOL. 22, 1978
E. COLI MENINGITIS IN INFANT RATS
0 0
o
-J
challenge. Of the 44 bacteremic animals, 21 (48%) had positive CSF cultures. The concentration of bacteria in CSF varied from 102 to 108 bacteria per ml of CSF. A relationship was found between the concentration of bacteria in blood and their isolation from CSF (Table 2). Animals with bacteremia of >104 CFU per ml were significantly more likely to have bacteria present
.p 103
-J
m L). U0
,
483
in their CSF than animals with
a
less intense
bacteremia (P < 0.05). It is possible that small quantities of blood may have contaminated CSF and accounted for bacterial isolation. Because of the small amount of CSF availabile, a cell count was not done. The
10
24
48
72
96
TIME AFTER IP INJECTION (hr) bacteremiP(geometric INJeomtriOmean of FIG. 3. Degree of bacteremia CFU/ml of blooai) after i.p. injection of a Kl E. coli
of
strain in newboyrn rats followed with serial blood cultures. Animal s who died within 48 h of challenge (Group I, 0) are contrasted with those dying after 48 h (Group II, l) oand those who survived (Group III A).
technique for collecting CSF has been shown by Moxon and
Ostrow to
provide
uncontaminated
CSF on most occasions (8). In our study, bacterial concentrations in CSF often
were
greater
104 CFU/ml of CSF, and *' CSF bacterial concentration was greater than that of blood. The isolation of bacteria from CSF than
two
occasions
was considered to indicate bacterial infection of this fluid. Peritoneal response to bacterial infec-
sacrifice, cold saline was III, P < 0.01). 1Preinfection weight of animals in .tion. these three grc rups did not differ sigoificantlyi injected i.p., and estimates of bacterial CFU and The clinical s igns of illness included inactivity, total and multilobed cells were made. Increased ulsmlerest m; {reeaimg, pantmg respiratnon, ana numbers of multilobed cells were seen in animals weight. loss. On occasion, animals had apparent myoclonic seizures. Not all animals with bacteriologically documented infection demonstrated these features, and some animals appeared clinically well at a time when they had >104 CFU bacteria per ml of blood. To further study the dynamics of infection in newbom animals, groups of 10 to 12 rats were m0L sacrificed 15 min, 3, 8, 24, or 48 h after i.p. 4) challenge. Counts of viable bacteria were made from samples of blood, CSF, and peritoneal cavity washings and also from saline suspensions of homogenized spleen and brain (Fig. 4). Fifteen minutes after i.p. injection, bacteria were iso0.25 3 8 24 48 lated from peritoneal washings and splenic hoTIME AFTER IP INJECTION (hr) mogenates. After 3 h, bacteria were isolated from number of CFU of bacteblood and from peritoneal fluid, but not from riaFIG. 4. Geometric mean per gram of spleen and brain or per ml of blood splenic or brain homogenates. At 8 h, 6 of 11 in newborn rats at various times after i.p. injection of animals had bacteria isolated from blood. Bac- 1.2 x 10' Kl E. coli. teria were also isolated from brain and splenic homogenates (three and seven animals, respecTABLE 2. Relationship of isolation of E. coli from tively). Twenty-four hours after i.p. injection, CSF to concentration of bacteria in blood after i.p. bacteria were found in high concentration in of a Kl E. coli strain into 5-day-old rats injection Bacand brain. spleen, peritoneal fluid, blood, Results of bacterial cultures from teria were recovered in highest titer in splenic CSF of rats (no. of animals) Bacterial concn at this sacrificed time. of animals homogenates (CFU/ml of blood) CSF infection. CSF was cultured from aniPositive Negative mals sacrificed at various times after i.p. injec20 2 104 quently from animals sacrificed 24 or 48 h after At
£-
the
time
of
484
BORTOLUSSI, FERRIERI, AND WANNAMAKER
sacrificed 3 h after i.p. challenge and later, whereas bacterial multiplication was not apparent until 8 h. Counts of monolobed cells were not different in infected or control animals. DISCUSSION In 1974 an association between E. coli strains with the Kl capsular polysaccharide antigen and neonatal meningitis was shown (10). Although 20 to 30% of infants are colonized with Kl E. coli strains, more than 75% of E. coli meningitis in newborns is caused by strains possessing the Kl antigen (12, 13). Both morbidity and mortality in newborn E. coli meningitis are significantly increased if the infecting strain possesses the Kl antigen (7). In addition, the quantity of Kl antigen in CSF measured by countercurrentimmunoelectrophoresis may be related to clinical outcome. Infants with the highest concentration of antigen are more likely to die or have major sequelae (7). Susceptibility to Kl E. coli strains appears to be greatest in the newborn because such strains are isolated less frequently among older age groups (12). When 5-day-old animals were injected i.p. with various E. coli strains, a marked difference was seen among strains. Mortality was greatest among animals challenged with Kl E. coli. Animals injected with non-Kl E. coli did not differ significantly in survival from control animals. Newborn rats appear to have an enhanced susceptibility to E. coli strains possessing the Kl antigen. The influence of somatic antigens of E. coli on virulence is not clear from our studies. The natural history of untreated Kl E. coli infection in humans is not known. In the animal model, the natural history of infection could be followed by serial tail vein bleeding and quantitative culture. Three groups of animal response to i.p. bacterial challenge were recognized. In the first group, bacteremia occurred within 8 h of i.p. challenge. Clinical illness, characterized by disinterest in feeding, lethargy, and poor weight gain, was apparent at 24 h, and death occurred before 48 h. A second group of animals developed bacteremia later and initially at a lower concentration than the first group. This group succumbed to infection 2 to 5 days after challenge. In the third group of animals, lowgrade bacteremia developed and was cleared with no clinical indication of infection. This variable response to identical i.p. bacterial challenge is of considerable interest. There was no significant difference in animal weight, litter size, or absolute number of bacteria injected in these groups. Animals who were able to prevent rapid dissemination of bacteria into the blood were likely to survive more than 2 days. In humans,
INFECT. IMMUN.
this delay might be crucial in allowing time for clinical recognition and therapeutic intervention. Moxon and Ostrow demonstrated that in newborn rats with Haemophilus influenzae infection, the development of intense bacteremia (>10i CFU/ml of blood) was correlated with the isolation of bacteria from CSF (8). Similarly, Glode et al. found that newborn rats who become infected after oral ingestion of Kl E. coli were more likely to have meningitis if bacteremia greater than 1025 CFU/ml of blood was present (4). In our study, animals with 104 or more CFU/ml of blood were significantly more likely to have bacteria isolated from CSF. Differences in the induction of infection or properties unique to the infecting bacteria may account for these differences. In naturally occurring E. coli infection in humans, it appears that infants with greater than 103 CFU/ml of blood are more likely to have meningitis (3). When not treated with appropriate antibiotics, E. coli meningitis is usually fatal in humans. In our study serial sampling of CSF was not possible. However, all animals with more than 104 CFU/ml of blood died, while none of the seven survivors fell into this category. Because animals with >104 CFU/ml of blood were likely to have meningitis, this would indicate that neonatal E. coli meningitis and intense bacteremia are ordinarily fatal in newborn rats. Animals sacrificed after i.p. injection of Kl E. coli showed widespread dissemination of bacteria by 8 h after injection. Isolation of bacteria in low numbers from brain homogenates does not confirm cerebral infection because small amounts of blood in the cerebral vascular system may have contributed to this finding. The bacterial counts in the spleen were higher than in the brain or any of the body fluids sampled. Because the concentration of bacteria in the spleen depends upon the rate of removal from the circulatory system and the rate of bacterial killing, further studies are required to understand the role of the spleen in host defense against Kl E. coli. The rapid multiplication and spread of bacteria in the animals indicate that the initial response to infection was inadequate. In spite of the appearance of increased numbers of multilobed cells in the peritoneal cavity by 3 h, bacteria were not cleared, suggesting that these phagocytic cells alone may not be able to effectively ingest and/or kill Kl E. coli. The infant rat model for E. coli infection has obvious similarities to human newborn E. coli infection. As in humans, the susceptibility of infant rats to E. coli infection is age dependent and appears related to the Kl capsular polysac-
E. COLI MENINGITIS IN INFANT RATS
VOL. 22, 1978
charide antigen. In rats, a small number of bacteria injected i.p. caused bacteremia within 24 h of injection. Bacteremia with >104 CFU/ml of blood was associated with the recovery of bacteria from the CSF. Although most animals eventually died from infection, some cleared bacteria from the blood and recovered spontaneously. It seems likely that future studies with the animal model will help to delineate the pathophysiology of naturally occurring infection. ACKNOWLEDGMENTS This work was supported by a grant-in-aid from the Graduate School, University of Minnesota. P. Ferrieri was supported in part by Public Health Service grants HL-06314-16 from the National Heart, Lung, and Blood Institute and AI13926-01 from the National Institute of Allergy and Infectious Diseases. R. Bortolussi is a recipient of the Hospital for Sick Children Foundation Fellowship, Toronto. L. W. Wannamaker is a Career Investigator of the American Heart Association.
LITERATURE CITED 1. Beer, A. E., and R. E. Billingham. 1976. The immunobiology of mammalan reproduction, p. 198-216. Prentice-Hall, Inc., Englewood Cliffs, N.J. 2. Bjorksten, B., R. Bortolussi, L Gothefors, and P. G. Quie. 1976. Interaction of E. coli strains with human serum: lack of a relationship to Kl antigen. J. Pediatr. 89:892-897. 3. Dietzman, D. E., G. W. Fischer, and F. D. Schoenknecht. 1974. Neonatal Escherichia coli septicemia-bacterial counts in blood. J. Pediatr. 86:128-130. 4. Glode, M. P., A. Sutton, E. R. Moxon, and J. B. Robbins. 1977. Pathogenesis of neonatal Escherichia coli meningitis: induction of bacteremia and meningitis in infant rats fed E. coli Kl. Infect. Immun. 16:75-80. 5. Glode, M. P., A. Sutton, J. B. Robbins, G. H. McCracken, E. C. Gotachlich, B. Kaijser, and IL A.
485
Hanson. 1977. Neonatal meningitis due to Escherichia coli Kl. J. Infect. Dis. 136 (Suppl.):S93-S97. 6. Kasper, D. L, J. L Winkelhake, W. D. Zollinger, B. L Brandt, and M. S. Artenstein. 1973. Immunochemical similarity between polysaccharide antigens of Escherichia coli 07:K1(L):NM and group B Neisseria meningitidis. J. Immunol. 110:262-268. 7. McCracken, G. H., Jr., L D. Sarff, M. P. Glode, S. G. Mize, M. S. Schiffer, J. B. Robbins, E. C. Gotschlich, L. Orskov, and F. Orskov. 1974. Relation between Escherichia coli Kl capsular polysaccharide antigen and clinical outcome in neonatal meningitis. Lancet ii:246-250. 8. Moxon, E. R., and P. T. Ostrow. 1977. Haemophilus influenzae meningitis in infant rats: role of bacteremia in pathogenesis of age-dependent inflammatory responses in cerebrospinal fluid. J. Infect. Dis. 135:303-307. 9. Reed, L. J., and H. Muench. 1938. A simple method of estimating fifty per cent endpoints. Am. J. Hyg. 27:493-497. 10. Robbins, J. B., G. H. McCracken, Jr., E. C. Gotachlich, F. 0rskov, L. 0rskov, and L. A. Hanson. 1974. Escherichia coli Ki capsular polysaccharide associated with neonatal meningitis. N. Engl. J. Med. 290:1216-1220. 11. St. Pierre, J., M. Cadieux, A. Guerault, and M. QuevMon. 1976. Statistical tables to detect significance between frequencies in two small samples with particular reference to biological assays. Rev. Can. Biol. 36:17-23. 12. Sarff, L D., G. H. McCracken, Jr., M. S. Schiffer, M. P. Glode, J. B. Robbins, L. 0rskov, and F. 0rskov. 1975. Epidemiology of Escherichia coli Kl in healthy and diseased newborns. Lancet i:1099-1104. 13. Schiffer, M. S., E. Oliveira, M. P. Glode, G. H. McCracken, Jr., L. M. Sarff, and J. B. Robbins. 1976. A review: relation between invasiveness and the Kl capsular polysaccharide of Escherichia coli. Pediatr. Res. 10:82-87. 14. Solomon, J. B. 1971. Foetal and neonatal immunology, p. 280-286; 349-361. North-Holland Publishing, Amsterdam. 15. Wolberg, G., and C. W. DeWitt. 1969. Mouse virulence of K(L) antigen-containing strains of Escherichia coli. J. Bacteriol. 100:730-737.
INFECTION AND IMMUNITY, Nov. 1978, P. 480-485 0019-9567/78/0022-0480$02.00/0 Copyright i) 1978 American Society for Microbiology
Printed in U.S.A.
Dynamics of Escherichia coli Infection and Meningitis in Infant Rats ROBERT BORTOLUSSI,t PATRICIA FERRIERI,' * AND LEWIS W. WANNAMAKER1 2 Departments of Pediatrics' and Microbiology,2 University of Minnesota, Minneapolis, Minnesota 55455 Received for publication 31 August 1978
Escherichia coli strains isolated from newborn infants were injected intraperitoneally into infant rats. Strains possessing the Kl capsular polysaccharide antigen were significantly more virulent than strains lacking this antigen. When 5-day-old animals were injected with 1.2 x 101 colony forming units of a Kl E. coli strain (serotype 018ac:Kl:H7), about 80% had bacteria isolated from their blood. Forty-eight percent of bacteremic animals had positive cerebrospinal fluid cultures. The development of bacteremia with >10i colony-forming units per ml of blood correlated with positive cultures of cerebrospinal fluid. Some animals, studied with serial blood cultures, were able to clear bacteria spontaneously from their blood, whereas others succumbed to infection within 48 h of challenge. The susceptibility of infant rats to E. coli infection was age dependent and appeared related to the Kl antigen. Escherichia coli has been associated with a wide variety of human acute infections. Recent studies have shown that a high prevalence of E. coli strains recovered from newborn infants with meningitis possessed the Kl capsular polysaccharide antigen (5, 10, 12, 13). The Kl antigen of E. coli is a polysaccharide of sialic acid and is immunologically and biochemically identical or virtually identical to the capsular polysaccharide of group B meningococcus (6). These findings suggest that the Kl capsular antigen may be related to virulence of E. coli strains, particularly for the newborn infant. In attempting to compare virulence of Kl and non-Kl E. coli in an animal model, Wolberg and DeWitt in 1969 (15) and Robbins et al. in 1974 (10) found that Kl E. coli strains suspended in hog gastric mucin and injected intraperitoneally (i.p.) into adult mice were more virulent than non-Kl strains. Because the mechanism of action of hog gastric mucin is poorly understood, the relationship of E. coli with or without Kl capsular antigen to virulence is unclear. In 1977 Glode et al. described a model for E. coli meningitis using infant rats fed with Kl E. coli strains isolated from infants with meningitis (4). They were able to demonstrate that bacteremia and meningitis occurred more frequently in animals fed Kl than non-Kl E. coli strains. In this model intralitter transmission to water-fed controls was common, and the incidence of meningitis was relatively low. The time of onset of t Present address: I. W. Killam Children's Hospital, Halifax, Nova Scotia, Canada.
extraintestinal infection was not known, and studies of dynamics of bacterial spread were not possible. We have examined the effect on infant rats of i.p. injection of different strains of E. coli. Rats were chosen because of the extensive literature available on immune responsiveness in the newborn period and because they are well-adapted to handling at this time (1, 14). The use of adjuvants such as hog gastric mucin to enhance virulence was not found to be necessary. Introduction of bacteria into the peritoneum allowed us to assess the newborn rat's ability to resist Kl E. coli infection independent of intestinal colonization and to follow the dynamics of bacterial multiplication and spread. Intralitter transmission to controls was rare. Our results indicate that rats have an age-dependent susceptibility and that there is an association of virulence with Kl E. coli strains. MATERIALS AND METHODS Animals. Outbred, pathogen-free albino SpragueDawley pregnant rats were obtained from Bio-Lab Corp. (White Bear Lake, Minn.). Animals were housed under standardized conditions (250C; relative humidity, 40%) with a 7 a.m. to 7 p.m. light schedule. Purina rat chow and water were available ad libitum. Rat pups and their mothers were housed in solid polypropylene opaque cages with filter hoods made from spunbonded polyester (Lab Products Co., Garfield, N.J.) to prevent cross-infection by aerosolization. Care was taken to minimize trauma and separation of animals from their mothers. Rat pups from multiple litters were used in each phase of the study to minimize the effect of interlitter variability.
480
VOL. 22, 1978 Bacteria. Twenty-seven strains of E. coli isolated from cerebrospinal fluid (CSF), blood, urine, and stools were tested (Table 1). Strains were obtained from C. Krishnan (Department of Bacteriology, Hospital for Sick Children, Toronto, Canada), J. B. Robbins (Division of Bacterial Products, Bureau of Biologics, Food and Drug Administration, Bethesda, Md.), D. Blazevic (Diagnostic Bacteriology Department, University of Minnesota, Minneapolis, Minn.), and B. Bjorksten (University of Umea, Sweden). The opsonic requirements and sensitivity to bactericidal activity of serum of many of these strains have been described (2). One strain, serotype 018ac:Kl:H7 (isolated from the CSF of an infant), designated C5 in an earlier report (2), was studied in detail. All Kl E. coli strains were identified by an agarose halo technique (12). Serotyping was performed by F. 0rskov (Seruminstitut, Copenhagen, Denmark) and B. Bj6rksten. Of the 27 strains, 21 were serotyped: 018ac:Kl:H7 (3 strains), 07:K1:H? (3 strains), 016:K1:H6 (1 strain), SpAg:Kl:H7 (1 strain), SpAg:Kl:H34 (1 strain), 022:K1 (2 strains), 02:K1:H6 (1 strain), 07:H- (2 strains), 083:H- (1 strain), 015:K7:H- (1 strain), SpAg:H7 (1 strain), 09:K36A (1 strain), SpAg:H4 (2 strains), and 0137:H11 (1 strain). Two Kl and four non-Kl E. coli strains were of unknown 0 serotype. Strains were grown in brain heart infusion broth (Difco Laboratories, Detroit, Mich.) for 8 h, divided into small portions with glycerol (final concentration, 20%), rapidly frozen, and stored at -20°C. When needed, vials of frozen bacteria were thawed, inoculated into brain heart infusion broth, and grown at 35°C. After overnight incubation, the bacteria were suspended in brain heart infusion broth and diluted to an optical density of 0.06 at 460 nm by using a Coleman Junior II spectrophotometer (Coleman Instruments Division, Oak Brook, Ill.). A 102 dilution was prepared in fresh media from this. After incubation at 35°C for 2 h, the concentration of bacteria was 2.0 ± 0.5 x 107 colony-forming units (CFU) per ml (confirmed by pour-plate technique in triplicate). The broth suspension was diluted in 0.85% NaCl to the desired number of CFU for i.p. injection by using a final volume of 0.1 ml. Mortality after i.p. injection. No virulence-enhancing agents were employed in any of these studies. The virulence of 27 strains of E. coli for 5-day-old rats was assessed by studying mortality after i.p. injection of 5 x 104 bacteria. This inoculum was selected when initial studies showed that the median lethal dose (LD50) of most Kl E. coli strains was less than 103 organisms, while the LD50 of non-Kl E. coli strains was greater than 105. One inoculum was used to allow a rapid comparison of several E. coli strains. Animals from three litters were injected with each strain. Care was taken to compare Kl and non-Kl E. coli strains isolated from comparable sources in order to avoid selecting more virulent strains in one group. Animals were injected between 1800 and 1900 h, and observations were made every 12 h for the first 48 h and every 24 h thereafter. A total of 135 5-day-old rats were observed for 5 days after i.p. injection of 5 x 104 bacteria. Fourteen Kl and 13 non-Kl E. coli strains were tested (each strain was tested in five animals). Seventy-three control animals from these litters were
E. COLI MENINGITIS IN INFANT RATS
481
TABLE 1. Source of E. coli strainsa
a
Source
Kl
Blood and CSF Stool Urine
10 4 0
Non-Kl 7 4 2
Numbers indicate the number of strains isolated.
injected i.p. with saline or a sterile broth filtrate of the test strain. The relationship of animal age to susceptibility to Kl E. coli was examined by determining the LD5o of the Kl E. coli strain, C5, in 5-, 12-, and 19day-old and adult animals by the method of Reed and Muench 72 h after inoculation (9). Between 20 and 30 animals were used for each LD50 determination. Sampling of body fluids and organs. The Kl E. coli strain, C5, was used to follow the dynamics of infection. In this study infant rats were injected i.p with 0.8 x 10' to 1.6 x 10' bacteria. This low inoculum was chosen in order to follow infection after minimal bacterial invasion of the peritoneum. The animals were weighed daily, observed at frequent intervals, and sacrificed at a predetermined time by a lethal dose of ether. After exsanguination, CSF was obtained and the brain was removed, weighed, washed, and homogenized in cold saline. The spleen was handled in a similar manner. Pour plates were prepared from serial dilutions of homogenates of these organs and results expressed as CFU/g of tissue. Blood obtained by cardiac puncture and peritoneal cavity washings was diluted in heparinized cold saline and bacterial counts were determined. A wet count and differential of peritoneal washings were done with a Neubauer hemocytometer by using gentian violet as a nuclear stain. Because this method did not allow differentiation of cell granules, peritoneal cells were designated simply as multilobed or monolobed, based on microscopic appearance. CSF was obtained by cisternal puncture using a method similar to one previously described (8). CSF was drawn up by capillary action into a 10-1l micro-
pipette (Micropipet Disposable Pipettes, Clay Adams Division of Becton, Dickinson & CO., Parsippany, N.J.). Ten microliters of CSF and dilutions of it were inoculated onto brain heart infusion agar. Colonies were counted, and results were expressed as CFU per ml CSF. For serial quantitative blood cultures, the tail was cleansed with 70% acohol and amputated. Blood was collected in 5-p4 calibrated microcapillary pipettes and inoculated onto MacConkey agar (Difco Laboratories). Blood was obtained from 24 animals at 8, 24, 48, 72, and 96 h after i.p. injection or until the animal died. Bacteria recovered from animals were identified on the basis of colonial morphology, growth on MacConkey agar, and by demonstrating rapid agglutination of three to four colonies with anti-meningococcal group B horse antiserum diluted 1:5 (provided by J. B. Robbins). To establish that blood obtained from the tail vein accurately reflected the number of bacterial CFU in blood, samples were obtained simultaneously from the heart and tail in 25 other animals. Quantitative bacterial counts were within 1 log (or
482
BORTOLUSSI, FERRIERI, AND WANNAMAKER
were both negative) in the range of 102 to 105 bacteria per ml of blood. Statistical methods. The chi-square test was used to determine the significance of differences among groups over 20. When comparing groups of 20 or fewer animals, tables based on the Fisher exact test were used (11).
INFECT. IMMUN.
or control animals (12, 14, and 15%, respec-
tively).
The LD50 of strain C5 was determined on 5-, 12-, and 19-day-old and adult animals 72 h after i.p. injection of bacteria (Fig. 2). There was an increase in resistance to Kl E. coli infection beyond 5 days of age. Adult animals tolerated i.p. injection with 105 bacteria; however, when RESULTS injected with 106 or 107 bacteria, they appeared Mortality after i.p. injection: relation to overwhelmed with infection and died within 24 capsular type and age of animals. Survival h of challenge. In contrast, animals 5 to 19 days of Kl and non-Kl E. coli-injected animals and old challenged with bacteria at a concentration control animals is shown in Fig. 1. There was a near the LD50 for their age appeared well 12 to significant difference (P < 0.05) in survival of 24 h after i.p. injection and died only after 24 to Kl E. coli-injected animals compared to non-Ki 48 h. Weight ranged from 11 to 40 g for 5- to 19E. coli-injected or control animals 24 to 72 h day-old animals and from 300 to 400 g for adult after injection. Animals injected i.p. with non- animals. Bacterial multiplication and disseminaKl E. coli strains did not differ significantly in survival from controls (P > 0.05). Mortality tion of infection. To determine the progression rates in the two control groups (broth filtrate, of bacteremia in newborn animals, serial blood 12%; saline, 16%) were similar and have been cultures were taken from the tail vein of 24 5combined. Among the Kl strains, the two spon- day-old rats after i.p. injection of 1.2 x 101 CFU taneously agglutinating strains were associated of a Kl E. coli strain (C5). Actual CFU by pour with lower mortality than the 10 strains possess- plate varied ±0.4 x 10'. Overall, 19 of 24 animals ing a somatic antigen (20 versus 74%). However, (79%) developed bacteremia, and 17 animals the number of animals injected with the spon- (71%) died from 1 to 5 days after i.p. injection. taneously agglutinating Kl strains was small and The time of development and the severity of precludes a valid comparison. bacteremia were variable in these animals (Fig. Some animals dying later than 72 h after i.p. 3). The presence of bacteremia at 8 h was ordichallenge were found to have sterile postmortem narily predictive of rapid bacterial multiplication blood cultures, whereas all of the animals who to concentrations of >10' CFU/ml of blood and died within 72 h had positive cultures. After 72 was related significantly to the subsequent h, the number of new deaths was not different course. Of 11 animals who died within 48 h of among Kl and non-Kl E. coli-injected animals i.p. injection (group I), 10 were bacteremic at 8 h, while only 2 of 6 animals who died after 2 days (group II) and 1 of 7 animals who survived to 7 days (group III) were bacteremic at 8 h. This difference in incidence of bacteremia was significant (group I versus group II plus group 100
>
106
80
>
w~~~~
F
60
X
40
-
io4
z
I,,G _
w/ 3 /
a.
I 10
10
12
24
36 48 TIME (hr)
60
72
FIG. 1. Percentage of newborn rats surviving after i.p. injection of 5 x 104 CFU of Kl E. coli (0), non-Kl E. coli (A), and saline or sterile broth filtrate controls (0).
l 5 5
9'
/
12
19
ADULT
AGE (DAYS)
FIG. 2. Relationship of virulence of Kl E. coli to age of animal as demonstrated by change in LD50 determined 72 h after i.p. injection in 5-, 12-, or 19day-old or adult animals.
VOL. 22, 1978
E. COLI MENINGITIS IN INFANT RATS
0 0
o
-J
challenge. Of the 44 bacteremic animals, 21 (48%) had positive CSF cultures. The concentration of bacteria in CSF varied from 102 to 108 bacteria per ml of CSF. A relationship was found between the concentration of bacteria in blood and their isolation from CSF (Table 2). Animals with bacteremia of >104 CFU per ml were significantly more likely to have bacteria present
.p 103
-J
m L). U0
,
483
in their CSF than animals with
a
less intense
bacteremia (P < 0.05). It is possible that small quantities of blood may have contaminated CSF and accounted for bacterial isolation. Because of the small amount of CSF availabile, a cell count was not done. The
10
24
48
72
96
TIME AFTER IP INJECTION (hr) bacteremiP(geometric INJeomtriOmean of FIG. 3. Degree of bacteremia CFU/ml of blooai) after i.p. injection of a Kl E. coli
of
strain in newboyrn rats followed with serial blood cultures. Animal s who died within 48 h of challenge (Group I, 0) are contrasted with those dying after 48 h (Group II, l) oand those who survived (Group III A).
technique for collecting CSF has been shown by Moxon and
Ostrow to
provide
uncontaminated
CSF on most occasions (8). In our study, bacterial concentrations in CSF often
were
greater
104 CFU/ml of CSF, and *' CSF bacterial concentration was greater than that of blood. The isolation of bacteria from CSF than
two
occasions
was considered to indicate bacterial infection of this fluid. Peritoneal response to bacterial infec-
sacrifice, cold saline was III, P < 0.01). 1Preinfection weight of animals in .tion. these three grc rups did not differ sigoificantlyi injected i.p., and estimates of bacterial CFU and The clinical s igns of illness included inactivity, total and multilobed cells were made. Increased ulsmlerest m; {reeaimg, pantmg respiratnon, ana numbers of multilobed cells were seen in animals weight. loss. On occasion, animals had apparent myoclonic seizures. Not all animals with bacteriologically documented infection demonstrated these features, and some animals appeared clinically well at a time when they had >104 CFU bacteria per ml of blood. To further study the dynamics of infection in newbom animals, groups of 10 to 12 rats were m0L sacrificed 15 min, 3, 8, 24, or 48 h after i.p. 4) challenge. Counts of viable bacteria were made from samples of blood, CSF, and peritoneal cavity washings and also from saline suspensions of homogenized spleen and brain (Fig. 4). Fifteen minutes after i.p. injection, bacteria were iso0.25 3 8 24 48 lated from peritoneal washings and splenic hoTIME AFTER IP INJECTION (hr) mogenates. After 3 h, bacteria were isolated from number of CFU of bacteblood and from peritoneal fluid, but not from riaFIG. 4. Geometric mean per gram of spleen and brain or per ml of blood splenic or brain homogenates. At 8 h, 6 of 11 in newborn rats at various times after i.p. injection of animals had bacteria isolated from blood. Bac- 1.2 x 10' Kl E. coli. teria were also isolated from brain and splenic homogenates (three and seven animals, respecTABLE 2. Relationship of isolation of E. coli from tively). Twenty-four hours after i.p. injection, CSF to concentration of bacteria in blood after i.p. bacteria were found in high concentration in of a Kl E. coli strain into 5-day-old rats injection Bacand brain. spleen, peritoneal fluid, blood, Results of bacterial cultures from teria were recovered in highest titer in splenic CSF of rats (no. of animals) Bacterial concn at this sacrificed time. of animals homogenates (CFU/ml of blood) CSF infection. CSF was cultured from aniPositive Negative mals sacrificed at various times after i.p. injec20 2 104 quently from animals sacrificed 24 or 48 h after At
£-
the
time
of
484
BORTOLUSSI, FERRIERI, AND WANNAMAKER
sacrificed 3 h after i.p. challenge and later, whereas bacterial multiplication was not apparent until 8 h. Counts of monolobed cells were not different in infected or control animals. DISCUSSION In 1974 an association between E. coli strains with the Kl capsular polysaccharide antigen and neonatal meningitis was shown (10). Although 20 to 30% of infants are colonized with Kl E. coli strains, more than 75% of E. coli meningitis in newborns is caused by strains possessing the Kl antigen (12, 13). Both morbidity and mortality in newborn E. coli meningitis are significantly increased if the infecting strain possesses the Kl antigen (7). In addition, the quantity of Kl antigen in CSF measured by countercurrentimmunoelectrophoresis may be related to clinical outcome. Infants with the highest concentration of antigen are more likely to die or have major sequelae (7). Susceptibility to Kl E. coli strains appears to be greatest in the newborn because such strains are isolated less frequently among older age groups (12). When 5-day-old animals were injected i.p. with various E. coli strains, a marked difference was seen among strains. Mortality was greatest among animals challenged with Kl E. coli. Animals injected with non-Kl E. coli did not differ significantly in survival from control animals. Newborn rats appear to have an enhanced susceptibility to E. coli strains possessing the Kl antigen. The influence of somatic antigens of E. coli on virulence is not clear from our studies. The natural history of untreated Kl E. coli infection in humans is not known. In the animal model, the natural history of infection could be followed by serial tail vein bleeding and quantitative culture. Three groups of animal response to i.p. bacterial challenge were recognized. In the first group, bacteremia occurred within 8 h of i.p. challenge. Clinical illness, characterized by disinterest in feeding, lethargy, and poor weight gain, was apparent at 24 h, and death occurred before 48 h. A second group of animals developed bacteremia later and initially at a lower concentration than the first group. This group succumbed to infection 2 to 5 days after challenge. In the third group of animals, lowgrade bacteremia developed and was cleared with no clinical indication of infection. This variable response to identical i.p. bacterial challenge is of considerable interest. There was no significant difference in animal weight, litter size, or absolute number of bacteria injected in these groups. Animals who were able to prevent rapid dissemination of bacteria into the blood were likely to survive more than 2 days. In humans,
INFECT. IMMUN.
this delay might be crucial in allowing time for clinical recognition and therapeutic intervention. Moxon and Ostrow demonstrated that in newborn rats with Haemophilus influenzae infection, the development of intense bacteremia (>10i CFU/ml of blood) was correlated with the isolation of bacteria from CSF (8). Similarly, Glode et al. found that newborn rats who become infected after oral ingestion of Kl E. coli were more likely to have meningitis if bacteremia greater than 1025 CFU/ml of blood was present (4). In our study, animals with 104 or more CFU/ml of blood were significantly more likely to have bacteria isolated from CSF. Differences in the induction of infection or properties unique to the infecting bacteria may account for these differences. In naturally occurring E. coli infection in humans, it appears that infants with greater than 103 CFU/ml of blood are more likely to have meningitis (3). When not treated with appropriate antibiotics, E. coli meningitis is usually fatal in humans. In our study serial sampling of CSF was not possible. However, all animals with more than 104 CFU/ml of blood died, while none of the seven survivors fell into this category. Because animals with >104 CFU/ml of blood were likely to have meningitis, this would indicate that neonatal E. coli meningitis and intense bacteremia are ordinarily fatal in newborn rats. Animals sacrificed after i.p. injection of Kl E. coli showed widespread dissemination of bacteria by 8 h after injection. Isolation of bacteria in low numbers from brain homogenates does not confirm cerebral infection because small amounts of blood in the cerebral vascular system may have contributed to this finding. The bacterial counts in the spleen were higher than in the brain or any of the body fluids sampled. Because the concentration of bacteria in the spleen depends upon the rate of removal from the circulatory system and the rate of bacterial killing, further studies are required to understand the role of the spleen in host defense against Kl E. coli. The rapid multiplication and spread of bacteria in the animals indicate that the initial response to infection was inadequate. In spite of the appearance of increased numbers of multilobed cells in the peritoneal cavity by 3 h, bacteria were not cleared, suggesting that these phagocytic cells alone may not be able to effectively ingest and/or kill Kl E. coli. The infant rat model for E. coli infection has obvious similarities to human newborn E. coli infection. As in humans, the susceptibility of infant rats to E. coli infection is age dependent and appears related to the Kl capsular polysac-
E. COLI MENINGITIS IN INFANT RATS
VOL. 22, 1978
charide antigen. In rats, a small number of bacteria injected i.p. caused bacteremia within 24 h of injection. Bacteremia with >104 CFU/ml of blood was associated with the recovery of bacteria from the CSF. Although most animals eventually died from infection, some cleared bacteria from the blood and recovered spontaneously. It seems likely that future studies with the animal model will help to delineate the pathophysiology of naturally occurring infection. ACKNOWLEDGMENTS This work was supported by a grant-in-aid from the Graduate School, University of Minnesota. P. Ferrieri was supported in part by Public Health Service grants HL-06314-16 from the National Heart, Lung, and Blood Institute and AI13926-01 from the National Institute of Allergy and Infectious Diseases. R. Bortolussi is a recipient of the Hospital for Sick Children Foundation Fellowship, Toronto. L. W. Wannamaker is a Career Investigator of the American Heart Association.
LITERATURE CITED 1. Beer, A. E., and R. E. Billingham. 1976. The immunobiology of mammalan reproduction, p. 198-216. Prentice-Hall, Inc., Englewood Cliffs, N.J. 2. Bjorksten, B., R. Bortolussi, L Gothefors, and P. G. Quie. 1976. Interaction of E. coli strains with human serum: lack of a relationship to Kl antigen. J. Pediatr. 89:892-897. 3. Dietzman, D. E., G. W. Fischer, and F. D. Schoenknecht. 1974. Neonatal Escherichia coli septicemia-bacterial counts in blood. J. Pediatr. 86:128-130. 4. Glode, M. P., A. Sutton, E. R. Moxon, and J. B. Robbins. 1977. Pathogenesis of neonatal Escherichia coli meningitis: induction of bacteremia and meningitis in infant rats fed E. coli Kl. Infect. Immun. 16:75-80. 5. Glode, M. P., A. Sutton, J. B. Robbins, G. H. McCracken, E. C. Gotachlich, B. Kaijser, and IL A.
485
Hanson. 1977. Neonatal meningitis due to Escherichia coli Kl. J. Infect. Dis. 136 (Suppl.):S93-S97. 6. Kasper, D. L, J. L Winkelhake, W. D. Zollinger, B. L Brandt, and M. S. Artenstein. 1973. Immunochemical similarity between polysaccharide antigens of Escherichia coli 07:K1(L):NM and group B Neisseria meningitidis. J. Immunol. 110:262-268. 7. McCracken, G. H., Jr., L D. Sarff, M. P. Glode, S. G. Mize, M. S. Schiffer, J. B. Robbins, E. C. Gotschlich, L. Orskov, and F. Orskov. 1974. Relation between Escherichia coli Kl capsular polysaccharide antigen and clinical outcome in neonatal meningitis. Lancet ii:246-250. 8. Moxon, E. R., and P. T. Ostrow. 1977. Haemophilus influenzae meningitis in infant rats: role of bacteremia in pathogenesis of age-dependent inflammatory responses in cerebrospinal fluid. J. Infect. Dis. 135:303-307. 9. Reed, L. J., and H. Muench. 1938. A simple method of estimating fifty per cent endpoints. Am. J. Hyg. 27:493-497. 10. Robbins, J. B., G. H. McCracken, Jr., E. C. Gotachlich, F. 0rskov, L. 0rskov, and L. A. Hanson. 1974. Escherichia coli Ki capsular polysaccharide associated with neonatal meningitis. N. Engl. J. Med. 290:1216-1220. 11. St. Pierre, J., M. Cadieux, A. Guerault, and M. QuevMon. 1976. Statistical tables to detect significance between frequencies in two small samples with particular reference to biological assays. Rev. Can. Biol. 36:17-23. 12. Sarff, L D., G. H. McCracken, Jr., M. S. Schiffer, M. P. Glode, J. B. Robbins, L. 0rskov, and F. 0rskov. 1975. Epidemiology of Escherichia coli Kl in healthy and diseased newborns. Lancet i:1099-1104. 13. Schiffer, M. S., E. Oliveira, M. P. Glode, G. H. McCracken, Jr., L. M. Sarff, and J. B. Robbins. 1976. A review: relation between invasiveness and the Kl capsular polysaccharide of Escherichia coli. Pediatr. Res. 10:82-87. 14. Solomon, J. B. 1971. Foetal and neonatal immunology, p. 280-286; 349-361. North-Holland Publishing, Amsterdam. 15. Wolberg, G., and C. W. DeWitt. 1969. Mouse virulence of K(L) antigen-containing strains of Escherichia coli. J. Bacteriol. 100:730-737.
