983
Discovery of Disseminated J96-like Strains of Uropathogenic Escherichia coli
O4:H5 Containing Genes for Both PapGJ96 (Class I) and PrsGJ96 (Class III)
Gal(a1-4)Gal–Binding Adhesins
James R. Johnson, Thomas A. Russo, Flemming Scheutz,
Jennifer J. Brown, Lixin Zhang, Karen Palin,
Christopher Rode, Craig Bloch, Carl F. Marrs,
and Betsy Foxman
Department of Medicine, University of Minnesota, and VA Medical
Center, Minneapolis; Department of Medicine, State University of New
York Clinical Center at Buffalo; International Escherichia and
Klebsiella Centre, World Health Organization, Copenhagen, Denmark;
Department of Epidemiology, University of Michigan School of Public
Health, and Department of Pediatrics and Communicable Diseases,
University of Michigan School of Medicine, Ann Arbor
The pyelonephritis-associated adhesin gene papG of Escherichia coli occurs in three variants.
Whereas the class II and class III variants are common among human urinary tract infection
isolates, the class I allele, despite being the first cloned, has previously been found only in source
strain J96. Five strains have been discovered from geographically diverse locales that, like J96,
contain both the class I and class III papG alleles. One strain caused bacteremia, whereas 4 caused
cystitis. Like J96, all 5 had group III capsule genes, expressed the H5 flagellar antigen and the F13
fimbrial antigen, and exhibited similar genomic patterns and virulence factor profiles. These findings
demonstrate that the class I papG allele is not unique to J96 but is present in a group of extraintestinal
isolates of E. coli O4:H5 that represent a disseminated virulent clonal group.
The PapG adhesin molecule of Escherichia coli P fimbriae,
which contributes to uropathogenesis by mediating bacterial
attachment to Gal(a1-4)Gal – containing receptors on host epithelial cells [1], occurs in three molecular variants, termed
classes I – III [2]. The corresponding alleles of papG (or prsG,
for the class III variant) differ sufficiently that they can be
resolved using allele-specific DNA probes [2] or polymerase
chain reaction (PCR) primers [3]. Because of their slightly
different receptor specificities, the three variants have been
predicted to have discrete host preferences and clinical associations. Epidemiologic evidence suggests that the class II allele
predominates in pyelonephritis and the class III allele in cystitis
[4, 5] or in canine urinary tract infections [6]. In contrast, the
class I allele’s clinical niche has remained mysterious because
until recently no example of this variant had been identified
outside of strain J96 [5], from which the class I papG variant
was cloned ú15 years ago.
Foxman et al. [7] recently reported 2 cystitis isolates with
pap restriction fragment polymorphism (RFLP) patterns similar
to strain J96. We undertook this study to determine whether
these and other selected strains might contain the class I papG
Received 16 July 1996; revised 21 October 1996.
Presented in part: American Society for Microbiology annual meeting, New
Orleans, 19 – 24 May 1996 (abstract B-251).
Grant support: NIH (DK-47504 to J.R.J. and J.J.B.; DK-35368 and DK47519 to C.F.M, L.Z., K.P., and B.F.); Research for Health in Erie County
(T.A.R.).
Reprints or correspondence: Dr. James R. Johnson, Infectious Diseases
(111F), VA Medical Center, 1 Veterans Dr., Minneapolis, MN 55417.
The Journal of Infectious Diseases 1997;175:983–8
q 1997 by The University of Chicago. All rights reserved.
0022–1899/97/7504–0041$01.00
allele and to explore possible similarities between these strains
and prototypic uropathogenic strain J96.
Methods
Strains. J96 is a pyelonephritis isolate from the late 1970s
from Seattle [3]. CP9 was isolated in the late 1980s from a patient
with bacteremia of unknown source at the National Institutes of
Health [8]. Isogenic transposon derivatives of CP9 deficient in
capsule (CP9.137), O polysaccharide (CP921), or both properties
(CP923), and resistant to kanamycin or chloramphenicol or both,
were constructed in one of our laboratories (T.A.R.) as previously
described [8]. Strains BF1023, BF1056, and BF1040 were isolated
in the early 1990s from the urine of otherwise healthy women
with first-episode acute cystitis (University of Texas at Austin
Student Health Service) [7]; BF9043 was isolated in the early
1990s from an otherwise healthy woman with three episodes of
urinary tract infection (UTI) during the previous 12 months (University of Michigan Student Health Service, Ann Arbor). Strains
V30b, PM8, and 2H4 are urosepsis isolates from Seattle [9]. Strain
IA2 contains the class II papG allele [2].
Serotyping. O:K:H serotyping and fimbrial (F) antigen determination were done by rocket immunoelectrophoresis followed by
crossed-line immunoelectrophoresis at the International Escherichia and Klebsiella Centre (WHO) in Copenhagen.
papG PCR assay. papG allele genotypes were determined using an allele-specific PCR assay [3]. All three primer pairs were
used simultaneously in single-tube 50-mL reactions as previously
described [3], except that each primer’s concentration was 0.45
mM and the thermocycler protocol was as follows: 957C for 7 min;
then 10 times at 947C for 1 min, 687C for 2 min, and 727C for 3
min; then 15 times at 947C for 1 min, 727C for 4 min; then 727C
for 10 min. Primers for the class I allele were j96-193f, 5*-TCGTGCTCAGGTCCGGAATTT-3*, and j96-653r, 5*-TGGCAT-
984
Concise Communications
JID 1997;175 (April)
Figure 1. Maps of 3 papG variants. Triangles indicate recognition sites for papGspecific primers. Solid bar below each map
indicates size and location of predicted
polymerase chain reaction product. Dashed
line between restriction sites indicates region used by others [2] as probe for particular papG variant (adapted from [3], with
permission).
CCCCCAACATTATCG-3* (461-bp product). For the class II allele, primers were ia2-383f, 5*-GGGATGAGCGGGCCTTTGAT3*, and ia2-572r, 5*-CGGGCCCCCAAGTAACTCG-3* (190-bp
product). For the class III allele, primers were prs-198f, 5*-GGCCTGCAATGGATTTACCTGG-3*, and prs-455r, 5*-CCACCAAATGACCATGCCAGAC-3* (258-bp product) (figure 1). (Primer
numbers identify the 5* base with respect to the 5* end of the corresponding papG allele.) Allele-specific products (figure 1) were resolved by agarose gel electrophoresis. In control experiments, strains
of known papG genotype yielded only the expected sized product(s),
which were confirmed as representing the corresponding papG region
by DNA sequencing and restriction analysis [3].
Repetitive element (rep)–PCR and pulsed-field gel electrophoresis (PFGE). Rep-PCR genomic patterns were generated from
whole cell boiled lysates using the BOX A1R primer or the
ERIC1R and ERIC2 primers, as previously described [10]. PFGE
of macrorestricted total DNA was done as previously described
using NotI [11] and XbaI [12].
Virulence factor genotype. Virulence-associated sequences
were detected using dot-blot hybridization with DNA probes specific for pap and prs (pap-2) operons (prf) [7], S fimbriae/F1-C
fimbriae (sfa), the aerobactin system (aer), group II capsule
(kpsMT), group III capsule (probe 2 [13]), hemolysin (hly), cytotoxic necrotizing factor 1 (cnf1), outer membrane protein T
(ompT), Dr family adhesins (drb), and type 1 fimbriae (fim). The
group III capsule probe was prepared from pRP7 [13]. Preparation
of other probes, and other blotting methods, were as previously
described [7].
Conventional and long-range Southern hybridization for prf and
hly. HindIII-digested total DNA was separated electrophoreti-
Figure 2. Typing methods using J96-like and other E. coli strains. A, Multiply primed papG polymerase chain reaction (PCR) using all 3
papG variant primer pairs combined. Sizes of PCR products: class I Å 461 bp; class II Å 190 bp; class III Å 258 bp. B, C, Repetitive element
(rep) – PCR. In B (BOX A1R primer), note bright band at Ç300 bp in all 6 J96-like strains that is absent from other strains. In C (ERIC
primers), note uniformity of J96-like strains, especially in lower (õ344 bp) region, compared with diverse patterns of other strains. D, E,
Pulsed-field gel electrophoresis (PFGE). In both D (XbaI digest) and E (NotI digest), strains CP9, BF1023, and BF1056 are indistinguishable
and very similar to J96. Strains BF1040 and BF9043 are distinct from one another and from J96. Strains V30b, PM8, and 2H4 differ even
more from one another and from J96-like strains (strain V30b did not digest with NotI, so was not included). F – H, Southern hybridization
(conventional: F, G; long-range: H) with prf and hly probes. HindIII-digested total DNA was separated by conventional electrophoresis and
hybridized with prf probe (F) or hly probe (G). For H, NotI-digested total DNA was separated by PFGE (figure 1E) and hybridized with prf
probe. Hybridization with hly probe gave results identical to H, except for absence of lower band in strain PM8 (not shown) (strain V30b did
not digest with NotI, so was not included). Lane designations: M, molecular size markers; a, J96; b, CP9; c, BF1023; d, BF1056; e, BF1040;
f, BF9043; g, V30b; h, PM8; i, 2H4; j, DNA-free blank; k, J96 / IA2 (combined).
JID 1997;175 (April)
Concise Communications
985
986
Concise Communications
JID 1997;175 (April)
Table 1. Characteristics of E. coli strains studied.
Genomic pattern‡
papG
allele(s)
O:K:H;F serotype†
Group,
strain
J96-like
J96
CP9
BF1023
BF1056
BF1040
BF9043
Other
V30b
PM8
2H4
Virulence factor signature
Syndrome*
Location
O
K†
H
F
PFGE
BOX
PCR
ERIC
PCR
I
II
III
prf
sfa
aer
kps
MT§
Group
III§
ompT
hly
cnf1
drb
fim
P
S
C
C
C
C
Seattle
NIH
Austin
Austin
Austin
Ann Arbor
O4
O4
O4
O4
O4
O4
K0
K10, K54/96†
K10, K54/96†
K10, K54/96†
K3
K3
H5
H5
H5
H5
H5
H5
F13
F13
F13
F13
F13
F13
A
B
B
B
C
D
A
A
A
A
A
A
A
B
B
B
C
D
/
/
/
/
/
/
0
0
0
0
0
0
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
0
0
0
0
/
0
0
0
0
0
0
0
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
/
0
0
0
0
0
0
/
/
/
/
/
/
U
U
U
Seattle
Seattle
Seattle
O16
O4
O6
K1
K12
K2
H0
H0
H1
F8
F16
F7
E
F
G
B
C
D
E
F
G
0
0
0
/
/
/
0
0
0
/
/
/
0
/
0
/
/
/
/
/
/
0
0
0
/
/
/
/
/
/
0
0
0
0
0
0
/
/
/
* Clinical syndrome: P Å pyelonephritis, S Å sepsis (unknown source), C Å cystitis, U Å urosepsis.
†
K10-positive strains also react with both K96 and K54 (which cross-react).
‡
Patterns with same letter are similar or indistinguishable if in same column, unrelated if in different columns. PFGE Å pulsed-field gel electrophoresis; BOX PCR and ERIC
PCR Å repetitive element – polymerase chain reaction fingerprinting using BOX A1R and ERIC primers, respectively.
§
kpsMT, group II capsule probe; group III, group III capsule probe (probe 2 in [13]).
cally and transferred to membranes, as was total DNA from NotI
PFGE [14]. Membranes were hybridized with the prf and hly
probes and results interpreted as previously described [7].
Results
Discovery of additional class I and class III papG – positive
strains. Strain CP9 was found serendipitously (J.R.J.) to yield
the class I and the class III papG PCR products, as does J96
(figure 2A) [2, 3]. To exclude contamination by (antibioticsusceptible) strain J96, three transposon derivatives of CP9
(CP9.137, CP921, and CP923) that are resistant to chloramphenicol or kanamycin or both were obtained from a different
laboratory (T.A.R.), grown with appropriate antibiotic selection, and subjected to papG PCR. All yielded the class I / III
papG pattern (not shown). This confirmed strain CP9 as the
first known example of a class I papG – positive strain other
than J96 [5].
Independently, Foxman et al. [7] noted that 2 urinary isolates
from part of their collection of first UTI isolates had prf RFLP
patterns and virulence factor profiles similar to those of strain
J96. For the present study, they searched the remainder of this
collection and their recurring UTI collection for strains with a
similar virulence factor profile. An additional 2 strains with
the identical prf RFLP pattern as J96 were identified. All 4 of
these J96-like strains (BF1023, BF1040, BF1056, and BF9043)
proved to have J96’s class I / III papG allele configuration
(figure 2A).
Serotyping. The similar reported O:K:H serotypes of J96
and CP9 (O4:K54:H5 and O4:K6:H5, respectively) suggested
that the class I / III papG configuration might characterize E.
coli O4:H5. As predicted, the J96-like cystitis strains (BF1023,
BF1056, BF1040, and BF9043) all exhibited serotype O4:H5
(table 1). K typing and F antigen determination showed that
each of the J96-like strains (except J96 itself, which was capsule-negative) expressed distinctive group III capsular antigens
(i.e., K3 or K10, K54/96) and the uncommon F13 fimbrial
antigen (table 1). In contrast, three arbitrarily selected urosepsis
isolates (of serogroups O4, O6, and O16) exhibited no commonality with the J96-like strains’ K, H, or F antigens (table
1) and contained only the class II papG allele (figure 2A).
Rep-PCR and PFGE. Rep-PCR and PFGE genomic patterns showed that the J96-like strains share a common evolutionary background. With the BOX A1R primer, the 6 J96-like
strains were essentially indistinguishable from one another,
whereas the 3 control strains were unique (figure 2B). With
the ERIC primers (figure 2C) and by PFGE (figure 2D, E), 3
of the J96-like strains (CP9, B1023, and BF1056) were indistinguishable from one another and were very similar to J96. The
other 2 J96-like strains (BF1040 and BF9043) were less similar,
and the 3 control strains differed markedly from one another
and the J96-like strains (figure 2C – E).
Virulence factors. The 6 J96-like strains exhibited identical
virulence signatures (table 1), except for strain BF1040, which
(unlike the others) was aer-positive. In contrast, the 3 control
strains (including the O4) had virulence signatures that differed
from those of the J96-like strains by several factors (table 1).
Capsule probe results, which cleanly differentiated the J96like strains from the controls, correlated closely with capsular
antigen typing in that all of the former (except capsule-negative
J96) but none of the controls expressed a group III capsule
[13] (table 1).
Southern hybridization for prf and hly. In conventional
HindIII Southern hybridization, the 6 J96-like strains exhibited
essentially identical prf and hly RFLP patterns (figure 2F, G),
whereas the control strains each had a unique prf RFLP pattern
(figure 2F) and shared a common hly RFLP pattern that ap-
JID 1997;175 (April)
Concise Communications
peared to be a subset of the J96-like strains’ hly RFLP pattern
(figure 2G). Long-range mapping (NotI digest) suggested the
presence of two genetically linked copies of prf and hly in the
J96-like strains, since both probes hybridized with the same
two large fragments in each strain (figure 2H). The otherwise
indistinguishable J96-like strains CP9, BF1023, and BF1056
had identical RFLP patterns for both prf and hly (figure 2H).
Discussion
We describe a group of clinically significant extraintestinal
E. coli isolates from diverse sites around the United States
that share many genetic and phenotypic characteristics with
prototypic uropathogenic strain J96, including the enigmatic
class I allele of papG. Among the several hundred strains previously studied with allele-specific papG probes, the class I
papG allele had been found only in source strain J96, evidence
that this gene might be unique to J96 [5]. That we encountered
5 strains other than J96 that contain the class I papG allele and
exhibit many other similarities to J96 permits several conclusions. First, the class I allele of papG may be more clinically
relevant than has been apparent previously, since it occurs
in strains that caused cystitis, pyelonephritis, and (probably)
urosepsis. Second, the class I papG allele may be restricted to
a specific subset of E. coli O4. Third, the absence of this papG
allele among the strain collections studied to date with allelespecific probes [5] requires explanation.
Our findings confirm that the 6 J96-like strains are closely
related and constitute a virulent clonal group [3] within E. coli
O4. They have similar (or indistinguishable) genomic fingerprints by multiple independent typing methods (figure 2B – E).
They all exhibit genetically related group III capsule determinants [13], which are expressed in all strains but J96 (table 1),
and all express the H5 flagellar and F13 fimbrial antigens (table
1). They all have sfa, ompT, and cnf1 sequences and (except
for BF1040) lack aer (table 1). They all contain the enigmatic
class I as well as the more common class III allele of papG
(figure 2A), and all contain two copies of pap, each of which
is genetically linked to an hly region, as has been described
previously for strain J96 (figure 2H). Their similar prf and hly
RFLP patterns indicate conservation of the entire prf and hly
operons and flanking sequences (figure 2F, G).
The absence of the class I papG allele from the strain collections examined to date with allele-specific DNA probes could
be due to technical differences between the assays used (which
is unlikely), to geographic segregation of the class I allele
(since the prior work involved mostly Scandinavian strains),
or to selection factors (since we specifically included strains
with known prf similarities to J96 [7]). Since Foxman et al.
[7] found just 2 J96-like strains among 216 cystitis isolates
screened, the J96-like group may constitute only a small fraction of uropathogenic E. coli, even in the United States. Examination of additional North American and European strains will
987
be needed to clarify this point and to determine whether the
class I allele of papG is confined to J96-like strains.
In every instance, we found the class I papG allele together
with the class III allele. Since the latter is often the sole papG
variant in pathogenic strains [5], the class III variant presumably is a sufficient P adhesin on its own. Studies in a mouse
UTI model that involved strains supplemented with cloned
fimbrial determinants [15] suggested that the class I adhesin
contributes to renal colonization only in the absence of the
class III adhesin. Thus, the role of the class I adhesin in uropathogenesis remains in question, notwithstanding the virulent
behavior of the J96-like strains studied here and the presence
of receptors for class I adhesin in the human kidney [16].
Immune avoidance by fimbrial phase variation has been proposed as a possible advantage for strains having both class I
and class III pap operons [15]. However, since PapA is the
major antigenic determinant [1], and since in the J96-like
strains both pap operons apparently express PapAJ96 (with its
F13 antigen) (table 1), this hypothesis is unsatisfying. It is
possible that because of their differing receptor specificities,
the class I and class III adhesins complement one another in
different stages of pathogenesis [15] or permit colonization of
a broader range of hosts and tissues [6]. Additional studies will
be needed to define the contribution to virulence of the class
I PapG allele and the clinical niche of this PapG variant and
of J96-like E. coli strains.
Acknowledgments
Parvia Ahmed helped in the laboratory of J.R.J.; Mary Hayes
helped prepare the manuscript; Rowan Pearce (University of
Leicester, UK) and Ian Roberts (University of Manchester, UK)
provided the plasmid from which the group III capsule probe was
prepared; Barbara Minshew (University of Washington), Steve
Clegg (University of Iowa), and Cindy Fennell, Cheryl Wobbe,
and Amy Denton (all from University of Washington) provided
strains; Joel N. Maslow (Boston University) provided helpful discussion and suggestions.
References
1. Johnson JR. Virulence factors in Escherichia coli urinary tract infection.
Clin Microbiol Rev 1991; 4:80 – 128.
2. Marklund BI, Tennent JM, Hamers GEA, et al. Horizontal gene transfer
of the Escherichia coli pap and prs pili operons as a mechanism for
the development of tissue-specific adhesive properties. Mol Microbiol
1992; 6:2225 – 42.
3. Johnson JR, Brown JJ. A novel multiply primed polymerase chain reaction
assay for identification of variant papG genes encoding the Gal(a14)Gal – binding papG adhesins of Escherichia coli. J Infect Dis 1996;
173:920 – 6.
4. Otto G, Sandberg T, Marklund BI, Ulleryd P, Svanborg Eden C. Virulence
factors and pap genotype in Escherichia coli isolates from women with
acute pyelonephritis, with or without bacteremia. Clin Infect Dis 1993;
17:448 – 56.
5. Johanson IM, Plos K, Marklund BI, Svanborg C. Pap, papG, and prsG
DNA sequences in Escherichia coli from the fecal flora and the urinary
tract. Microb Pathog 1993; 15:121 – 9.
988
Concise Communications
6. Strömberg M, Marklund BI, Lund B, et al. Host-specificity of uropathogenic Escherichia coli depends on differences in binding specificity to
Gala1-4Gal – containing isoreceptors. EMBO J 1990; 9:2001 – 10.
7. Foxman B, Zhang L, Palin K, Tallman P, Marrs CF. Bacterial virulence
characteristics of Escherichia coli isolates from first-time urinary tract
infection. J Infect Dis 1995; 171:1514 – 21.
8. Russo TA, Born JJ, Jodush ST, Johnson JR. The O4 specific antigen moiety
of lipopolysaccharide but not the K54 group 2 capsule is important
for urovirulence in an extraintestinal isolate of Escherichia coli. Infect
Immun 1996; 64:2343 – 8.
9. Johnson JR, Orskov I, Orskov F, et al. O, K, and H antigens predict
virulence factors, carboxylesterase B pattern, antimicrobial resistance,
and host compromise among Escherichia coli strains causing urosepsis.
J Infect Dis 1994; 169:119 – 26.
10. Versalovic J, Schneid M, de Bruijn FJ, Lupski JR. Genomic fingerprinting
of bacteria using repetitive sequence – based polymerase chain reaction.
Methods Mol Cell Biol 1994; 5:25 – 40.
11. Rode CK, Obreque VH, Bloch CA. New tools for integrated genetic and physical analyses of the Escherichia coli chromosome. Gene 1995;166:1–9.
JID 1997;175 (April)
12. Russo TA, Stapleton A, Wenderoth S, Hooton TM, Stamm WE. Chromosomal restriction fragment length polymorphism analysis of Escherichia
coli strains causing recurrent urinary tract infections in young women.
J Infect Dis 1995; 172:440 – 5.
13. Pearce R, Roberts IS. Cloning and analysis of gene clusters for production
of the Escherichia coli K10 and K54 antigens: identification of a new
group of serA-linked capsule gene clusters. J Bacteriol 1995; 177:3992 –
7.
14. Foxman B, Zhang L, Tallman P, et al. Virulence characteristics of Escherichia coli causing first urinary tract infection predict risk of second
infection. J Infect Dis 1995; 172:1536 – 41.
15. Hull RA, Nowicki B, Kaul A, Runyan R, Svanborg C, Hull SI. Effect of
pap copy number and receptor specificity on virulence of fimbriated
Escherichia coli in a murine urinary tract colonization model. Microb
Pathog 1994; 17:79 – 86.
16. Karr JF, Nowicki B, Truong LD, Hull RA, Hull SI. Purified P fimbriae
from two cloned gene clusters of a single pyelonephritogenic strain
adhere to unique structures in the human kidney. Infect Immun 1989;
57:3594 – 600.