Gestational pyelonephritis-associated Escherichia

represent a nonrandom, closely related population coli isolates

Audrey Hart, BA," Tuan Pham, BS, b Stella Nowicki, DDS," b Elbert B. Whorton, Jr., PhD, ~ Mark G. Martens, MD, a Garland D. Anderson, MD," and Bogdan J. Nowicld, MD, PhD " b

Galveston, Texas, and Minneapolis, Minnesota

OBdECTIVE: A select group of Escherichia coh strains known as uropathogenic cause pyelonephritis in nonpregnant individuals. We investigated whether Escherichia coli from gestational pyelonephritis represent a random population or possess common uropathogenic characteristics. STUDY DESlGN: Repetitive element sequence-based polymerase chNn reaction, plasmid profiles, hemolysin, and O serotypes were assayed from Escherichia cofi isolates of 57 pregnant patients with acute pyelonephritis at different gestational ages. RESULTS: The majodty of the first trimester isolates fell primarily into repetitive element sequence-based patterns 1 and 3 and 06, O15, and 075 serotyoes. Second-trimester isolates had muttiple pattems, with high-frequency repetitive element sequence-based polymerase chain reaction pattems 1 and 5 and an unknown (OX} serotype. Pattem 3, predominan+.ly 075 serotype, was found primarily among third-trimester isolates. CONCLUSlON: It is likely that Escherichia coli associated with acute pyelonephritis during different trimesters of pregnancy represents nonrandom closely related isolates, and some of these strains may be characteristic in pregnant patients only. (AM J O~sTE'r GYNECOL 1996;174:983-9.)

Key words: Pregnancy, pyelonephritis, deoxyribonucleic acid fingerprinting, Escherichia coli

Pregnant women seem to be more prone to urinary tract infections than nonpregnant women are. ~' 2 Entarg-

ing of the uterus results in anatomic obstruction of the urinary tract and is viewed as the major factor allowing infecdon to occur? Consequently, the Escherichia coli in-

fecting pregnant patients is likely to represent random, endogenous, nonpathogenic fecal strains. A random group of Escherichia coli fecal isolates would be character- ized by a lack of common features such as O serotypes, plasmid patterns, deoxyribonucleic acid (DNA) finger- prints, and virulence factors. However, if anatomic ob- struction is not the primary nr only cause of acute gesta- tional pyelonephritis, classic nephropathogenic E. coli

strains are likely to occur with pyelonephritis in pregnant individuals.

The primary cause of pyelonephritis in nonpregnant

From the Division of Gynecologic Infectious Disease. Departm«nt of Obstetrics and Gynecolog3, ~ the Department of Microbiology and Immu- nology, ~ and the Division of Biostatistics, Department of Preventive Medicine and Community Health/ University of Texas Medical Branch at Galveston, and the Department of Gyneeology and Obstetrics, Henne- pin County Medical Cent~ « Supported by National Institutes of Health grant No. NIH R01 DK42029 (B. f N). Received for puólication November 21, 1994; revised July 12, 1995; accepted August 11, 1995. Reprint requests: Bogdan Nowicki, MD, Ptu9, The Ur, iversity of Texas Medical Branch at Galveston, Department of Obstetrics and Gynecology, 301 University Blvd., Galveston, TX 77555-1062. Copyright © 1996 by Mosby-Year Book, Inc. 0002-9378/96 $5.00+ 0 6/1/68534

patients is the gram-negative bacilii E. coli, accounting for

about 70% to 90% o f in f ec t i onsS E. coli can be classified into more than 170 different groups on the basis of their O serotypes. Certain O serotypes, such as O1, 02, 04 , 06, 07, 08 , 09, O16, O18, 025, and 075, 4.7 are more orten

associated with pyelonephritis and are recognized as uro- pathogenic. E. coli produces a number of different viru- lence factors, including O andgens, hemolysins, and fim- briae, allowing infection to occur. 4' 5~ 8-~~ Limited studies characterizing O serotypes and virulence factors of E. coli

have been done on pregnant pafients with pyelonephri- tis." 14 It remains unclear whether pregnancy contributes to selective pressures allowing infection of the urinary

tract by random fecal isolates nr uropathogenic E. coli

strains carrying virulence factors.

Isolates causing nongestational pyelonephritis have been grouped by phenotypic properties into a limited number of clones. 7 0 n l y selected E. coli clones of com- mon ancestral origin seem to cause nongestational uri- nary tract infection. Others are presumably less virulent nr represent fecal commensal flora. DNA fingerprinting, including plasmid profiles and, more recently, repetitive element sequence-based polymerase chain reaction have been developed and are used to analyze clonal simi!ari- fies. '5 The repetitive element sequence-based polymerase chain reaction method uses primers composed of short, extragenic, repetitive sequences that generate a finger- print from genomic DNA, permitt ing identification of genetically related strains within a bacteriat species. ~517

983

984 Hart et al. March 1996 Am J Obstet Gynecol

lA

St. la lb 2 3a 3b 5 11

1B

St. 7 8 9 10

10D0 700

400

Fig. 1. Most ffequent repetitive element sequence-based polymerase chain reaction patterns among gestational E. coli isolates.

Table I. Occu r r ence of c o m m o n repetit ive

e l emen t sequence-based polymerase chain react ion

pat terns among gestational E. coli isolates

Pattern No. of strains %

1 15 26.3 2 8 14.0 3 11 19.3 5 5 8.8 Other t8 31.5

Repetit ive e l emen t sequence-based polymerase chain re-

action f ingerpr in t ing corre la ted weil with o the r typing

approaches in de t e rmin ing the clonal origin of isolates

and has been effectively used for molecu la r ep idemiology

of Enterobacteriaceae and o ther species.'»-17 These me thods

were used to characterize E. coli strains isolated f rom

p regnan t patients with pyelonephrit is .

Material and methods E. coli 0 serotyping. E. coli isolates f rom 57 p regnan t

patients with pyelonephri t is were sc reened against 173

classified (classification according to the World Heal th

Organiza t ion) and 15 nonclassified O serotype antibod-

ies. Isolates that did no t react with any of the antisera were

des ignated OX, indicat ing Unknown O serotype; the pro-

cedure used has been previously described. 1~

Repetitive element sequence-based polymerase chain reaction. E. coli strains were grown in 3 ml of Luria broth

overn igh t in a 37 ° C shaker and carbon dioxide incuba-

tor. Cultures were pel le ted in a microcent r i fuge at 15,000

r e v o l u t i o n s / m i n for 5 minutes and then resuspended in

200 lal of sterile watet. The suspensions were hea ted at

100 ° C for 10 minutes and then microcen t r i fuged for 5

minutes at 15,000 revo lu t ions /min . Supernatants con-

taining the bacterial DNA were collected and stored at

- 2 0 ° C until fur ther use.

Repetit ive sequence-based pr imers used were o rdered

f rom Bio-Synthesis (Lewisville, Tex.). The first primer,

Rep l , con ta ined the o l igonucleot ide sequence 5'-

IIIICGICGICATCIGGC-3, and the second primer, Rep2,

con ta ined the sequence 5 '-ICGICTTATCIGGCCTAC-3' ,

as previously desc r ibed) ~ Polymerase chain react ion am-

plification was p e r f o r m e d in an au tomated thermal cy-

cler (Perkin-Elmer-Cetus DNA thermal cycler, Branch-

burg, NJ . ) with an initial dena tura t ion at 94 ° C for 5

minutes, followed by 35 cycles of denatura t ion at 94 ° C

for 1 minute , anneal ing at 55 ° C for 2 rninutes, extension

at 72 ° C for 2 minutes, and a s tep-down hold ing cycle at

4 ° C. Twenty microli ters of polymerase chain react ion

sample was run on 2% agarose gel at 100 V for 2 hours.

Gels were stained with 0.5 tag/ml e th id ium bromide for

30 to 45 minutes. Gels were then pho tog raphed with a

15-second exposure to Polaroid type 57 film (Cambridge,

Mass.). Plasmid purifieaüon and profiles. Fifty-seven isolates

f rom patients with gestational pyelonephri t is were stored

at -70 ° C until fur ther characterization. E. coli strains

were grown in Luria bro th overnight in a 37 ° C shaker

and carbon dioxide incubator. Plasmids were harvested

according to pro tocol by use of Magic Minipreps DNA

purif icat ion system (Promega, Madison, Wis.). Ptasmids

were t un on 1% agarose gels at 100 V for 2 hours. Gels

were stained with 0.5 lag/ tal e th id ium bromide for 30 to

45 minutes. Gels were then pho tog raphed with a 15-

second exposure to Polaroid type 57 film.

Hemolytic activity. Hemolyt ic activity of the isolates was

tested by subcul tur ing the strains on trypticase soy agar

with 5% sheep b lood (TSA II, BDMS Microbiology Sys-

tems Products, Lenexa, Kan.) overnight in a 37 ° C carbon

Volume 174, Number 3 Ha~ et al, 985 Am J Obstet Gynecol

% Bacteriai Isolates 60 1st Trimester

40

30

2O

10

0 1 2 3 4 5 6 7 8 9 10 11

Rep-PCR Patterns

% Bacterial Isolates 60

5 0

40

3 0

20

10

0

2nd Trimester

1 2 3 4 5 6 7 8 ~ 10 11 Rep-PCR Pattems

% Bacterial

Isolates 60"~ 3 r d T r i m e s t e r

50

40

30

10

0 l 2 3 4 5 6 7 8 9 10 11

Rep-PCR Patterns

Fig. 2. Repetitive element sequence-based (Rep-PCR) patterns of E. coli isolates from patients with pyelonephritis during differ- ent trimesters of pregnancy.

d iox ide incubator . I f a zone of c lear ing a p p e a r e d a r o u n d

the colonies , the s t ra in was c o n s i d e r e d hemolyt ic .

S t a t i s t i c a l m e t h o d s . To d e t e r m i n e w h e t h e r Lhe distr ibu-

t ion of isolates with mos t f r e q u e n t O serotypes d i f fe red

statistically a m o n g the t h r e e t r imesters , a Z'-' test was per-

f o rmed . This test was also used to es t imate d i f fe rences in

the d i s t r ibu t ion of repet i t ive e l e m e n t s e q u e n c e - b a s e d

polymerase cha in r eac t ion pa t t e rn s across the t r imesters .

ResuRs

Fifty-seven p r e g n a n t pa t ien ts with acute pye lonephr i t i s

who h a d b e e n hospi ta l ized in the D e p a r t m e n t of Obste t -

rics a n d Gynecology at the Universi ty of Texas Medical

B r a n c h at Galves ton were invest igated. T h e s tud ied g r o u p

% B acterial Isolates 2 «

2

1

I

Ist Trimester

% Bacterial Isolates 30

20

10

5

0

02 04 06 O15 016 O21 025 075 OX O Serotype

2nd Trimester

02 04 OO O15 O16 O21 025 075 OX

O Serotype

% Bacterial 3rd Trimester Isolates 30 -~

25

20

15

I0

5

0 02 04 06 O15 O16 O21 025 075 OX

O Serotype

Fig. 3. Association of E. col~ 0 serotypes with first, second, and third trimesters of pregnancy.

r e p r e s e n t e d a b o u t 50% to 60% of p a d e n t s with gesta-

t ional pye lonephr i t i s hospi ta l ized in a p e r i o d of 9 m o n t h s

in 1993 to 1994. This se tec t ion was r a n d o m a n d based o n

the fact tha t app rox ima te ly 50% to 60% of u r i n e isotates

f rom pa t ien t s with s ignif icant bac te r iu r i a were sen t to us

by the clinical microb io logy laborator3« Protocol for the

inves t igat ion was app roved by the ins t i tu t ional review

b o a r d before the study. T h e ges ta t ionai ages of pa t i ems

i n c l u d e d 12 individuals in the first t r imester , 23 in the

second, a n d 22 in the third . E thn ic a n d ges ta t ional age

d i s t r ibu t ions were relatively even (one th i rd white, o n e

th i rd black, and one th i rd Hispanic) . Geograph ica l l> pa-

986 Hart et al. March 1996

Am J Obstet Gynecol

1 O0 o~ 6 075 o2s

' l i ~ ~ I l l 8O o 4

60 o

r4 40 02 om @ 20

0 I 1 I [ ~ 2 I L 3 / 4 5

Rep-PCR Pattern

Fig. 4. Association of predominant repetitive element se- quence-based polymerase chain reaction (rep-PCR) panerns with O serotypes.

tients represented a large area of South Texas, from the Louisiana border to Mexico. Pregnant patients with

pyelonephritis showed symptomatic bacteriuria of >105/ml microorganisms of urine and fever _>38.2 ° C, and

most individuals have costovertebral pain and dysuria.

Two patients were found to have positive blood cultures, one at the twenty-second week of gestation and one at the twenty-seventh week of gestation. Isolates from urine and

blood of both patients showed identical O serotypes and polymerase chain reaction patterns; therefore only urine isolates were included in the analysis. None of the pa- tients tested had renal diseases other than pyelonephritis. Two patients with recurrent pyelonephritis and diabetes were excluded from this study. Information on prior asymptomatic bacteriuria was not available.

To investigate our working hypothesis that gestational pyelonephritis develops with E. coli isolates of specific characteristics, we at tempted to characterize repetitive

e lement sequence-based polymerase chain reaction DNA

fingerprints of strains associated with acute pyelonephri- tis among 57 pregnant patients. Limited distinct repeti- tive e lement sequence-based polymerase chain reaction profiles (Fig. 1), 11 major patterns total, with minor varia- tions within some of the groups (designated as a, b, etc.),

were observed among the isolates. The predominant re- petitive e lement sequence-based polymerase chain reac- tion patterns 1 (15/57, 26.3%, 2 (8/57, 14%), 3 (11/57, 19.3%), and 5 (5/57, 8.8%) accounted for a total of

68.5% (39/57) of all isolates (Table I). Patterns 4 and 6, 7, 8, 9, and 10 contained three isolates or less each, for a total of 21% of strains. Examples of selected distinct re- petitive e lement sequence-based polymerase chain reac- tion patterns are shown in Fig. 1, B. Pattern 11 showed no distinguishable repetitive e lement sequence-based poly- merase chain reaction bands and characterized six iso- lates (11%) (Fig. 1, A). The repetitive e lement sequence- based polymerase chain reaetion data clearly indicated that a majority of gestational E. coli isolates were geneti- cally related, which was consistent with the hypothesis on

a clonal origin of the isolates. To further investigate association of E. coli repetitive

element sequence-based polymerase chain reaction types with particular gestational age, the occurrence of indi- vidual patterns was estimated in the first, second, and third trimesters. Repetitive element sequence-based poly-

merase chain reaction patterns appeared to correlate with gestational age (Fig. 2). In the first trimester the only patterns found were 1, 2, and 3. Several patterns were

associated with the second trimester, with patterns 1 and

ä predominating. The third-trimester isolates were orten found to belong to patterns 3, 4, 8, 9, and 11. The distri- bution of repetitive element sequence-based polymerase chain reaction types differed statistically among the three trimesters, with a pvalue <0.005 (p < 0.005). Pattern 1 was associated primarily with 54% of the first-trimester iso-

lates, 30% of the second, and only 9% of the third trimes- ter. Pattern 2 was found in all three trimesters. Pattern 3

was associated predominantly with the third trimester (30%) and the first trimester (27%). Only a single isolate from the second trimester (4%) belonged to pattern 3.

Patterns 4 and 5 occurred in the second and third trimes- ters but were absent in the first trimester. Three patterns, 6, 7, and 10, were found to be associated with the second trimester only. Patterns 8, 9, and 11 were found exclu- sively in the third trimester. The presence of characteris- tic patterns during individual trimesters further sup- ported our hypothesis that the nonrandom occurrence of genetically related isolates in pregnant individuals sug- gests a temporal pat tern in gestational pyelonephritis.

To analyze further E. coli clonal origin and temporal occurrence during pregnancy, we investigated whether

E. coli nephropathogenic serotypes were associated with pyelonephritis during pregnancy (Fig. 3). Classic neph- ropathogenic serotypes O1 (2%), 0 2 (14%), 0 4 (5%),

0 6 (9%), O16 (5%), 025 (4%), and 075 (17%) com- prised a large port ion (56%) of the detected serotypes. Almost half (44%) of the total isolates tested were of the nonuropathogenic serotypes restricted to a narrow group that included O15 (14%) and OX (16%) serotypes.

The predominant serotypes during the first trimester were 06 , O15, and 075 (27% each). The second trimes- ter was more diversified, with 14 (61%) isolates belonging to nonuropathogenic serotypes. The unknown, second- trimester serotype(s) designated as OX (30%) (non- typable by the E. coli Reference Center) and O15 (17%) were the most frequent serotypes. The third trimester isolates were predominantly serotypes 075 (26%) and 0 2 (24%), followed by O21 (17%), which is seldom asso- ciated with pyelonephritis. Serotypes 0 6 and O15 oc- curred most frequently in the first trimester and de- creased in frequency, whereas 0 2 increased in frequency with gestational age, especially in the third trimester. Other nonnephroPathogenic serotypes isolated con- sisted of single isolates O19/133, 065, and O81 from the second trimester and O 102 from the third trimester. Simi- lar to repetitive element sequence-based polymerase

Volume 174, Number 3 Ha r t e t a[. 9 8 7 A m J Obstet Gynecol

% Hemolytic

5A

1 2 3 a 5 6 7 8 9 !0 11 Rep PCR Pattern

% Hemolytic

5B lC~()

0 2 04 06 O15 O16 O21 025 075 OX O Serotypes

Fig. 5. Association of repetitive element sequence-based polymerase chain reaction (rep-PCR) parterres and O serotypes of gestational E. coli with hemolytic properties.

chain reaction, O serotyping indicated a close phenotypic relationship of gestational pyelonephritis isolates and suggested a temporal pat tern associated with gestational age. Distribution of the most frequent serotypes among trimesters was analyzed by a Z 2 test. Analysis per formed for the most frequent serotypes 06 , O75, and OX showed a nonrandom distribution of tested groups (p < 0.007) across trimesters.

To furtber investigate the interrelationship of E. coli

isolates, O serotypes of E. coli identified by serotype- specific antibody were characterized with respect to re-

petitive e lement sequence-based polymerase chain reac- tion patterns (Fig. 4). Serotypes O6 (80%) and O15 (63%) appeared to correlate predominantly with repeti- tive e lement sequence-based polymerase chain reacti0n pattern 1. Uropathogenic serotypes 0 4 (67%) and Ot6 (100%) were associated primarily with pattern 2. All the O75 (100%) isolates were found to belong to patterns 3a and 3b. Two patterns, 5 and 6, were associated predomi- n an tly witb the unknown OX serotype (s) (77 %). Serotype

0 2 was spread out among patterns 1, 2, 6, 8, and 11. Patterns 5, 7, 9, and 10 were found among nonuro- pathogenic serotypes only. Association of repetitive ele-

ment sequence-based polymerase chain reaction pat- terns with O serotypes clearly supports E. coli clonat nonrandom characteristics.

In addition, E. coli plasmid patterns were investigated. A single plasmid band of 23 kb was isolated from 7 of 12 (58%) of the first, 4 of 23 (18%) from the second, and 9 of 20 (45%) from the third trimester. The remainder of the isotates carried muhiple plasmids of various sizes, of which the second trimester predominated (19/23, 82%).

The majority of the uropathogenic serotypes carried the single 23 kb band (60%), and the nonuropathogenic isolates carried multiple plasmids (88%). Among the mul-

tiple plasmid isolates, E. coli OX showed related profiles (figure not shown).

Finally, hemolytic activity, which allows E. coli clinical isolates to damage a variety of human cell types, was examined. Strains in the first trimester exhibited this factor with a slightly higher freqnency, 46%, and de- creased with gestational age. Hemolytic activity was asso-

ciated with certain repetitive element seqnence-based polymerase chain reaction patterns and serotypes (Fig.

5, A and B). Patterns associated with hemolysis were 2 (88%), 4 (100%), 6 (100%), and 9 (100%). Isolates ot"O6,

988 Hart et N. March 1996 AmJ Obstet Gynecol

O16, and 025 serotypes were all hemolytic, whereas no

hemolytic activity was observed among the serotypes 021

and 075. Therefore toxic activity that was also associated

with both repetitive element sequence-based polymerase chain reacfion and O patterns further supported out

hypothesis that there is an existence of the same gesta-

tional nonrandom isolates in pregnant pyelonephritis

patients and the importance of E. coli is its role in the

pathogenesis of pyelonephritis in pregnant individuals.

Comment

In this article we attempted to cbaracterize various

E. coli isolates associated with pyelonephritis during preg-

nancy. The data obtained supported our hypothesis that

the development of gestational pyelonephritis is associ-

ated with specific E. coli strains, and among these strains

a novel group of gestational strains may occur.

During pregnancy both motber and developing fetus

are vulnerable to infectious complications such as pyelo-

nephritis that can resuh in low birth weight and preterm labor, infant mortality, 2' 3, i9, 20 and renal scarring. °' 21E. coli

clones bearing virulence factors have been shown to af-

fect the pathogenesis of nongestational urinary tract in-

fections.7, 22 Anatomic factors are likely to increase the

frequency of pregnancy-associated infection. However, it

is not clear whether gestation-associated obstruction of

the nrinary tract applies selective pressure on types o f

microorganisms, allowing either commensal or uro-

pathogenic strains to colonize in the pregnant female.

We found two major groups of isolates among gesta- tional pyelonephritis-associated E. coli. Orte group repre-

sented characterisücs of classic uropathogenic E. coli, pre- dominantly serotypes O1, 02, 04, 06, and O75," 2, 9, 10, 13

and the other group consisted of nonuropathogenic se-

rotypes, although they were mostly restricted to O15, O21, and OX. Occurrence of limited repetitive element

sequence-based polymerase chain reaction patterns,

O serotypes, and other associated factors among the ma- jority of E. coli gestational isolates was consistent with

hypothesis on the nonrandom origin of the isolates. The

limited number of patterns associated with serogroups

that predominated during given trimesters supports our

interpretation of the existence of gestational age-specific clones. O15, O21, and OX strains associated with the second trimester may represent high-risk gestational pyelonephritis strains; it is of interest that such strains

were highly virulent, causing patients infected with these

isolates to have a fever of >39.5 ° C (article in prepara- tion). Plasmid profiles indicated that the majority of sec- ond-trimester (83%) isolates carried the muhiple bands,

with OX isolates showing similarity among the multiple plasmid patterns. This further supported suggestions that specific OX strains could belong to the high-risk second- trimester gestational pyelonephritis clone(s). The above

associations indicated a nonrandom incidence of IL coli

strains. Therefore, in addition to anatomic factors, bacte-

rial clonal characteristics may be important in the patho- genesis of gestational pyelonephritis.

Genomic fingerprinting and O serotyping showed that

gestational infection of the genitourinary tract appears to be associated with E. coli of common genetic and pheno-

typic features, which is consistent with the clonal and the common ancestral origin. Clonality has been found

among isolates of nongestational, E. coli pyelonephritis in children. 7 It is therefore likely that a similar phenomenon

may account for the pathogenesis of renal infection in pregnant women.

Interestingly, serotypes predominating in about half the

isolates associated with gestational pyelonephritis were dif-

ferent than those observed in nongestational pyelonephri-

tis. Epidemiologic similarities were also found to be associ-

ated with gestational age. This may prove to be ofbenefit to the clinician during prenatal screening. E. coli isolated

from the urine of pregnant patients may be characterized by means of repetitive element sequence-based poly-

merase chain reaction, O serotype, and related expressed clonal factors. This information may be useful to identify

patients and strains that may represent high-risk factors

for the development of acute pyelonephritis and associ-

ated complications; polymerase chain reaction and sero- typing, however, would require experienced personnel to

perform them. Further studies are in progress to charac-

terize the origin and virulence ofgestational clones associ- ated with pyelonephritis and to explore potential clonal characteristics of bacterial isolates associated with gesta-

tional pyelonephritis complications, including preterm la-

bor and low birth weight.

REFERENCES

1. Stenqvist K, Sanberg T, Lidin-Janson G, Orskov F, Orskov I, Svanborg-Eden C. Virulence factors of Escherichia coli in uri- nary isolates from pregnant women. J Infect Dis 1987;156: 870-7.

2. Fihm SD. Urinary tract infecfions in primary care obstetrics and gynecology. Clin Obstet Gynecol 1988;31:1003-16.

3. Sweet RLS, Gibbs RS. Urinary tract infections. In: Brown CL, ed. Infectious diseases of the female genital tract. Baltimore, Maryland: Williams & Wilkins, 1985:293-313,

4. Johnson JR. Virulence factors in Escherichia coliurinary tract infection.J Clin Microbiol 1991;4:80-128.

5. Van der Auwera P, Denis R. Review of the immunological aspects of the pathogenesis, diagnosis and prognosis of uri- nary tract infections. Acta Clin Belg 1985;40:17903.

6. Sanberg T, Kaijser B, Lidin-Janson G, et al. Virulence of Escherichia coli in relation to host factors in women with symptomatic urinary tract infecfion. J Clin Microbiol 1988; 26:1471-6.

7. Vaisanen-Rhen V, EloJ, Vaisanen A, et al. P-fimbriated clones among uropathogenic Escherichia coli strains. Infect Immun 1984;43:149-55.

8. Schaeffer AZ, Amuedsen SK, Schmidt LN. Adherence of Escherichia coli to human urinary tract epithelial cells. Infect Immun 1979;24:753-9.

Volume 174, Number 3 [-[aß et af. 989 AmJ Obstet Gynecol

9. Leffer H, Svanborg-Edén C. Glycolipid receptors for uro- pathogenic Escherichia coli on human erythrocytes and uroepithe!ial cells. Infect Immun 1981;34:920-9.

10. O'Hanley P, Low D, Romero I, et al. Gal-gal binding and hemolysin phenotypes and genotypes associated with uro- pathogenic Escherichia coli. N Engl J Med 1985;313:414-20.

11. Hagberg L, Jodal U, Korhonen TK, Lidin-Janson G, Lind- berg U, Svanborg-Edén C. Adhesion, hemagglunnation and virulence of Escherichia coli causing urinary tract infections. Infect Immun 1981 ;31:564-70.

12. Ofek I, Mosek A, Sharon N. Mannose-speeific adherence of Esctwric,~ia coli freshly excreted in the urine of patients with urinary tract infections, and isolates subeulmred from the infected urine. Infect Immun 1981 ;34:708-11.

13. Nowicki B, Hart A, Coyne KE, Lublin DM, Nowicki S. Short consensus repeat-3 domain of recombinant decay-accelerat- ing factor is recognized by Escherichia coli recombinant Dr adhesin in a model of cell-ce!l interaction. J Exp Med 1993; 178:2115-21~

14. Nowicki B, Martens M, Hart A, Nowicki S. Gestational age dependent distribution of Escherichia coli fimbriae in preg- nant patients with pyelonephritis. In: Ades EW, Rest RF, Morse SA, eds. Microbial pathogenesis and immune re- sponse. New York: New York Academy of Science, 1994: 290-4.

15. Woods CRJr, Versalo~äc J, Koeuth T, LupskiJR. Analysis of relationships among isolates of Citrobacter diversus by using DNA fingerprints generated by repetitive sequenced-based

primers in the polymerase chain reaction. J Clin Microbiol 1992;30:2921-9.

16. Sharples GJ, Lloyd GL. A novel repeated DNA sequence located in the intergenic regions of bacterial chromosomes, Nucleic Acids Res 1990;18:6603-8.

17. Versalovic J, Koeuth T, Lupski JR. Distribution of repetitive DNA sequences in eubacteria and application to fingerprint- ing of bacterial genomes. Nucleic Adds Res 1991;19:6823- 31.

18. Glantz PJ. Serotypes of Escherichia coli associated with coli bacillosis in neonataI animals. Ann N YAcad Sci 1971;176: 67-79.

19. Schult.z R, Read AW, Straton JAY, Stanley FJ, Morich R Genito-urinary tract infections in pregnancy and low birth weight: case control smdy in Australian aboriginal women. BMJ 199I ;303:1369-73.

20. Romero R, Mazoc M, Oyarzum E, Sitori M, Hobbins JC, Bracken M, Metaanalysis of the relationship between asymp- tomatic bacteriuria and preterm delivery/low birth weight. Obstet Gynecol 1989;73:576-82.

21. Meyrier A, Condamin MC, Fernet M, et al. Frequency of development of early cortical scarring in acute primary pyelonephritis. Kidney Int 1989;35:696-703.

22. StameyAT, Kaufman ME Studies ofintroital colonization in women with recurrent urinary infections, II: a comparison of growth in normal vaginal fluid ofcommon versus uncom- mon serogroups of Escherichia coli. J Urol 1975; 1l 4:264-7.