Analysis of the F Antigen-Specific papA Alleles of Extraintestinal Pathogenic Escherichia coli Using a Novel Multiplex PCR-Based Assay

Abstract

Polymorphisms in PapA, the major structural subunit and antigenic determinant of P fimbriae of extraintestinal pathogenic Escherichia coli, are of considerable epidemiological, phylogenetic, and immunotherapeutic importance. However, to date, no method other than DNA sequencing has been generally available for their detection. In the present study, we developed and rigorously validated a novel PCR-based assay for the 11 recognized variants of papA and then used the new assay to assess the prevalence, phylogenetic distribution, and bacteriological associations of the papA alleles among 75 E. coli isolates from patients with urosepsis. In comparison with conventional F serotyping, the assay was extremely sensitive and specific, evidence that papA sequences are highly conserved within each of the traditionally recognized F serotypes despite the diversity observed among F types. In certain strains, the assay detected serologically occult copies of papA, of which some were shown to represent false-negative serological results and others were shown to represent the presence of nonfunctional pap fragments. Among the urosepsis isolates, the assay revealed considerable segregation of papA alleles according to O:K:H serotype, consistent with vertical transmission within clones, but with exceptions which strongly suggested horizontal transfer of papA alleles between lineages. Sequencing of papA from two strains that were papA positive by probe and PCR but F negative in the new PCR assay led to the discovery of two novel papA variants, one of which was actually more prevalent among the urosepsis isolates than were several of the known papA alleles. These findings provide novel insights into the papA alleles of extraintestinal pathogenic E. coli and indicate that the F PCR assay represents a versatile new molecular tool for epidemiological and phylogenetic investigations which should make rapid, specific detection of papA alleles available to any laboratory with PCR capability.

P fimbriae, the principal mannose-resistant adherence organelles of extraintestinal pathogenic Escherichia coli, mediate Gal(α1-4)Gal-specific binding to glycolipid isoreceptors on host epithelial cells, thereby contributing to pathogenesis by promoting bacterial colonization of host tissues and stimulating an injurious host inflammatory response (11, 33, 34). P fimbriae are antigenically diverse, occurring in 11 known serological variants, which are termed F7-1, F7-2, and F8 to F16 according to the system of Ørskov et al. (18, 64, 68, 70). Whereas Gal(α1-4)Gal-specific binding is mediated by PapG, the fimbrial tip adhesin molecule (29, 56), the antigenic diversity of P fimbriae is attributable to peptide sequence variability within PapA, the major P fimbrial structural subunit (12, 13, 22, 55, 57, 96). PapA is present in hundreds to thousands of identical copies per fimbria and is encoded by the corresponding gene, papA (12, 57). Wild-type E. coli strains can contain up to three copies of the pap operon and, since each pap operon can have a different variant (allele) of papA, can express up to three different P fimbrial (F)-antigen types each (1, 18, 67, 68, 70, 75–77).

The evolution of antigenic diversity within PapA may have been driven by selective pressure from the host immune system (8, 12, 96). P fimbriae are expressed in vivo within the urinary tract during infection and appear to be physiologically relevant immunogens (2, 17, 49, 54, 73, 87). They also are effective antigenic targets for a protective host humoral immune response (48, 52, 71, 81). Thus, avoidance of the host immune attack would be predicted to favor a diverse PapA antigenic repertoire in the E. coli population as a whole and even within individual strains, which may be able to switch the expression of their various papA alleles on or off through phase variation (62, 68, 78, 94).

Knowledge of the F antigen status of wild-type E. coli strains can be valuable in several ways. First, independent of its pathogenetic significance, the F antigen can serve as a typing tool to help differentiate strains, much as do the O, K, and H antigens of conventional E. coli serotyping (18, 65, 70). Second, since F antigens are markers for the corresponding papA alleles, they can be used to trace the vertical or horizontal transmission of particular pap variants within the E. coli population, e.g., as defined by O:K:H serotypes, and in relation to other virulence factors (61, 70, 72, 74). Third, epidemiological analyses of F types among strains from defined clinical syndromes or host populations can reveal the clinical associations of particular F variants, which could help guide future efforts to develop and deploy an anti-PapA vaccine (12, 63, 64).

The current “gold standard” method for F determination involves rocket immunoelectrophoresis and crossed-line immunoelectrophoresis (67, 68), demanding techniques which require highly specialized equipment, skills, and F-specific antisera; hence, at present it is largely confined to a single laboratory (88). Consequently, few studies have been done regarding associations of F antigens with other bacterial properties, clonal markers, or clinical syndromes. F-specific monoclonal antibodies, which can be used in (more widely available) enzyme-linked immunosorbent assays (68), have been developed, but they have been developed for only a subset of the known F types (1, 15–18, 68) and have not come into general use.

Previous efforts to exploit DNA sequence polymorphisms for typing of papA variants by restriction endonuclease analysis (19) or with a battery of oligonucleotide probes (13) were handicapped by the unavailability of sequence data for the full range of papA alleles. Full-length sequencing of papA from wild-type strains, although highly informative and requiring no prior knowledge of existing papA sequences (8), is impractical for large-scale screening. The recently completed cloning and sequencing of papA variants corresponding to all 11 recognized F types of P fimbriae (7) suggested the possibility of exploiting amplification technology for papA allele determinations. In the present study, we sought to develop and validate a PCR-based assay for the 11 known F-type-specific papA alleles of E. coli and to use this assay to investigate the papA allele repertoire of a collection of well-characterized blood isolates from patients with urosepsis.

(This work was presented in part at the 98th General Meeting of the American Society for Microbiology, abstr. 12200.)

MATERIALS AND METHODS

Bacterial strains.

Control strains used in derivation and validation of the PCR assay included diverse wild-type strains plus, in many instances, their recombinant pap derivatives (e.g., Tables 1 and 2). The derivation set used in assay development consisted mainly of the source strains for the F type-specific papA sequences on which the PCR assay was based. The derivation set strains (and their respective F types and wild-type parent strains) were HB101/pPIL110-75 (F7-1, from AD110) (99), HB101/pPIL110-75 (F7-2, from AD110) (95), AM1727/pANN921 (F8, from 2980) (25), AM1727/pPIL288-10 (F9, from C1018) (14), MOSBLUE/pF10 (F10, from C1960-79) (this study), AM1727/pPIL291-15 (F11, from C1976) (15), 1442 (F12, wild type) (22), HB101/pRHU845 (F13, from J96) (30), MOSBLUE/pF14 (F14, from C1023-79) (this study), MOSBLUE/pF15 (F15, from C1805-79; present study), and MOSBLUE/pF16 (F16, from C83-83) (this study). The first validation set (Table 1) consisted primarily of strains from the International Escherichia and Klebsiella Center (IEKC), World Health Organization, Copenhagen, Denmark, and included, for most F types, that laboratory's type strain for the particular F type. These were supplemented with diverse other F-typed strains, as available. The second validation set (Table 2) consisted of nine strains, generously provided by Han de Ree, for which the O:K and F serotypes had been previously published (16–18, 68).

TABLE 1.

First validation set strains for F PCR assay

F serotype(s)		F PCR result	Strain name	O:K:H serotype (or source strain, if clone)	Source or reference
Putativea	Confirmatory
F7-1, F7-2	NDb	F7-1, F7-2	AD110	O6:K2:H1	18
F7-1d	ND	F7-1	MC400/pDAL201B	(AD110)	55
F7-2d	ND	F7-2	HB101/pDAL210B	(AD110)	55
F7-1, F7-2	ND	F7-1, F7-2	CFT073	O6:K2:H1	60
F7-1	ND	F7-1	DH5α/pKgg201	(CFT073)	60
F7-2	ND	F7-2	DH5α/pDIW101	(CFT073)	60
F7-2	ND	F7-2	C952-79	O6:K2:H1	IEKC
F7-2	ND	F7-2	C953-79	O6:K2:H1	IEKC
F7-2	ND	F7-2	C1898-79	O6:K2:H1	IEKC
F7-2	ND	F7-2	2H4	O6:K2:H1	41
F8	F8, F10	F8, F10	C1253-77	O18ac:K−:H−	IEKC
F8	F8	F8, F10	C1254-77	O75:K−:H5	IEKC
F8	F8, F10	F8, F10	C351-82	O18ac:K−:H−	IEKC
F8	F8, F10	F8, F10	C405-82	O18ac:K−:H5	IEKC
F8	F8, F10e	F8, F10	C805-83	O18ac:K−:H−	IEKC
F8	F−	F10	C659-81	O1:K51:H−	IEKC
F8	F8	F8, F10	C321-82	O18ac:K−:H−	IEKC
F8	F8	F8, F10	C825-83	O75:K5:H−	IEKC
F8	F8, F12	F8, F12	V30b	O16:K1:H−	41
F9	F9	F9, F10	3669	O2:K5:H4	12
F9d	ND	F9	HB101/pDAL200A	(3669)	12
F9	F9	F9, F10	C1018-79	O2:K5:H4	IEKC
F9	F9	F9, F10	C481-82	O1:K1:H−	IEKC
F9	F9, F10e	F8, F9	C483-82	O1?:K1:H−	IEKC
F9	F14, F16	F13, F14, F16	C878-83	O4:K12:H5	IEKC
F10	ND	F10	C1960-79	O7:K1:[H1]	IEKC
F10	ND	F10	C997-79	O7:K1:H−	IEKC
F10	ND	F10	C328-82	O7:K1:H−	IEKC
F10	ND	F10	C906-83	O7:K1:H−	IEKC
NAc	F−	F10	G1062a	O4:K7:HR	38
F11	F11, F16e	F11, F16	IA2	O4:K12:H−	10
F11	ND	F11	HB101/pDC1	(IA2)	16
F11	ND	F11	C1976-79	O1:K1:H7	IEKC
F11	ND	F11	C974-79	O1:K1:H7	IEKC
F11	F10e, F11	F10, F11	V31	O4:K12:H−	41
F11, F16	ND	F11, F16	AFR015	O4:K+:H−	45
F11, F16	F10e, F11, F16	F10, F11, F16	BOS040	O4:K+:H1	45
NA	F10, F11, F16	F10, F11, F16	JR1	O4:K7:H1	82
F12-1, F12-2	ND	F12	C1979-79	O16K1:H−	IEKC
F12-2	ND	F12	C493-82	O6:K13:H1	IEKC
F12-2	ND	F12	C438-82	O75:K5:H5	IEKC
F12-2	ND	F12	C469-82	O75:K95:H5	IEKC
F13	ND	F13	J96	O4:K−:H5	44
F13	ND	F13	DH5α/pJJ48	(J96: allele I)	47
F13	ND	F13	HB101/pJFK102	(J96: allele III)	51
F13	F13, F14	F13, F14	CP9	O4:K10,K54/96:H5	44
NA	ND	F13	NM554/pCP9I	(CP9)	This study
NA	ND	F14	NM554/pCP9III	(CP9)	This study
F13	F13, F14	F13, F14	BF1023	O4:K10,K54/96:H5	44
F13	F13, F14	F13, F14	BF1056	O4:K10,K54/96:H5	44
F13	F13, F14	F13, F14	BOS038	O4:K10,K54/96:H5	44
F13	F13, F14	F13, F14	BOS110	O4:K10,K54/96:H5	44
F13	F13, F14	F13, F14	518	O4:K10,K54/96:H5	45
F13	ND	F13, F14	18878	O4:K10,K54/96:H5	IEKC
F13	ND	F13	BF1040	O4:K3:H5	44
F13	ND	F13	BF9043	O4:K3:H5	44
F13	ND	F13	CA002	O4:K3:H5	45
F13	ND	F13	CA062	O4:K3:H5	45
F13	ND	F9, F13	CA022	O4:K3:H5	45
F13	F11,e F13	F11, F13	R28	O4:K3:H5	45
F13	F13, F16	F13, F16	3048	O4:K−:H−	31
F13, 16	ND	F13, 14, 16	C134-73	O4:K12:H5	IEKC
F13, F14, F16	ND	F13, F14, F16	20025	O4:K12:H−	IEKC
F14	ND	F14	C127-86f	(Unknown)	IEKC
F14	ND	F14	C 818-83	O25:H9	IEKC
F14	ND	F14	C1023-79	O83:K24:H31	IEKC
F15	ND	F10, F15	C1805-79	O75:K5:H−	IEKC
F15	F−	F10	C826-83	O75:K5:H−	IEKC
F15	F10e, F15	F10, F15	C372-82	O75:K5:H+	IEKC
F15	F10, F15	F10, F15	C394-82	O75:K5:H−	IEKC
F15	F10e, F15	F10, F15	C312-82	O75:K5:H−	IEKC
F16	ND	F16	C83-83	O157:K−:H45	IEKC
F16	F14, F16	F14, F16	PM8	O4:K12:H−	41
F16	F16	F7-1, F16	BOS021	O4:K+:H−	45
F16	ND	F16	BOS046	O4:K12:H−	45

^aAccording to records of the IEKC. 

^bND, not done. 

^cNA, not applicable. 

^dF type according to reference 12. 

^eAntigen detected only in second or third round of confirmatory serotyping. 

^fRecombinant. 

TABLE 2.

Second validation set of strains for F PCR assay

Straina	Published O:K serotype	Confirmatory O:K:H serotype	Published F type	Confirmatory F serotyping result			F PCR result
				Round 1	Round 2	Round 3
AD303	O25:K−	O4:K2:H1	F11, F16	F11	F10, F11	F10, F11, F16	F10, F11, F16
AD309	O77:K−	O77:K16:H4	F7-1, F11, F12	F11, F12	NDd	ND	F11b
AD314	O4:K12	O4:K12:H1	F11, F16	F11	F10?, F11	F11, F16	F11, F16
NS3	O14:K+	OR:K?:H6	F7-1, F13, F14	F7-2, F13	ND	ND	F7-2, F13
NS24	O1:K1	O1:K1:H7	F7-1, F11	F11	ND	ND	F11
NS25	O4:K12	O4:K12:H1	F11, F16	F11	F10, F11, F16?	F10, F11, F16	F10, F11, F16
NS26	O83:K24	O83:K−:H31	F7-2, F14	F14	ND	ND	F14
SP57	O157:K−	O157:K−:H45	F16 + ?	F−	F10?, F16	F16c	F16
SP88	O157:K−	O157:K−:H45	F16 + ?	F−	F10?, F16	F16	F16

^aSources of strains and serotypes: AD309, AD314, NS3, NS24, NS25, and NS26, reference 18; SP57 and SP88, reference 68. 

^bSubsequently found to be PCR positive for the new F48 papA allele (but still PCR negative for F12). 

^cA second colony type noted in the serotyping laboratory was serotyped as OR:K−:H45, F10, F16. (This variant was not retested with the F PCR assay.) 

^dND, not done. 

The application set consisted of 75 well-characterized blood culture isolates of E. coli collected from adults with urosepsis in Seattle, Wash., in the mid-1980s. The status of these strains with respect to multiple characteristics, including pap, diverse other virulence factors, O:K:H serotype, carboxylesterase B electrophoretic type, and host compromise status, has been reported previously (35, 39–42). Strains were considered to belong to a particular O:K:H serotype if they exhibited any two of the three corresponding antigens plus no other O, K, or H antigen (41).

Sequencing of papA alleles.

The nucleotide sequences of the F7-1, F7-2, F9, F11, F12, and F13 papA variants were as previously published (4, 22, 93, 97, 98) and as found in the GenBank database under accession no. X02921 (F7-1), M12861 (F7-2), M68059 (F9), L07420 (F11), X62157 (F12), and X61239 (F13). The nucleotide sequences of the F8, F10, F14, F15, and F16 papA variants were experimentally determined. papA was amplified from control strains AM1727/pANN921 (F8 clone, derivation set), C1960-79 (F10 wild type, Table 1), C1023-79 (F14 wild type, Table 1), C1805-79 (F15 wild type, Table 1), and C83-83 (F16 wild type, Table 1) using primers 5′-ctgagaattcaggttgaaattcgc-3′ (forward) and 5′-atgatgaattcggttattgccggtgcgg-3′ (reverse), which are modified from the corresponding pap sequences to provide EcoRI sites for cloning. The resulting papA amplicons were electrophoresed in agarose gel, and the DNA fragments were removed from the gel using the GeneClean Kit (Bio 101, Inc., La Jolla, Calif.). Purified fragments were ligated into the TA cloning vector (Invitrogen, Groningen, The Netherlands) to create plasmids pF8, pF10, pF14, pF15, and pF16, respectively, and transformed into the electrocompetent cells of the TA cloning kit to create the corresponding recombinant papA control strains. Sequencing of the inserts was done with the dideoxynucleotide chain termination method (88) using the Automated Laser Fluorescence sequencer (Pharmacia, Uppsala, Sweden). The M13 universal and reverse sequencing primers labeled with Cy5 were used in the AutoRead 1000 Sequencing Kit (Pharmacia, LKB Biotechnology). Three different clones of each fragment were sequenced.

Sequence alignments and dendrogram construction.

To illustrate the phylogenetic relationships between the PapA variants, an alignment of the predicted mature peptides was made using the CLUSTAL-X program (26). The alignment was then used to calculate a distance matrix from which an unrooted phylogenetic tree was inferred according to the neighbor-joining method (86) using the PHYLIP 3.5c package (21) (Fig. 1). (J. G. Kusters, Department of Medical Microbiology, Free University, Amsterdam, The Netherlands, generously constructed the tree.) For comparison with PapA, the two most closely related known pilin genes from other members of the family Enterobacteriaceae, i.e., MR/P of Proteus mirabilis (5, 6) and SMF of Serratia marcescens (58), were included in the tree (7) (Fig. 1).

FIG. 1.

Dendrogram of PapA variants. Included are the 11 recognized PapA variants (F7-1, F7-2, and F8 to F16), new PapA variants F40 and F48 from the present study, and SMF and MR/P pilins (from S. marcescens and P. mirabilis, respectively). The unrooted tree was based on (predicted) mature peptides and was constructed using the neighbor-joining method (86). The presumed evolutionary distance between any two members of the tree equals the sum of the lengths of the branches connecting them.

F PCR primers.

A universal forward F primer (Ff [5′-ggcagtggtgtcttttggtg-3′]) was selected from the consensus signal sequence region of papA, without regard for peptide structure. To guide selection of immunologically relevant and F-type-specific reverse primers, secondary-structure predictions were made for the 11 translated F-specific PapA peptides using the PlotStructure and PeptideStructure components of the Wisconsin Package (version 5.0; Genetics Computer Group, Madison, Wis.) (32). For each of the 11 PapA sequences, unique peptide regions were identified that were predicted to be hydrophilic (27, 28), surface exposed (20), and antigenic (32) and to contain beta (reverse) turns (9, 23), hence, to be putative F-specific epitopes (89). Candidate reverse primers for each F type were selected from the coding regions corresponding to these putative F-specific epitopes. Since all putative F-specific epitopes for the F12 peptide had corresponding DNA sequences which were suboptimal for use as primers, two compromise F12 reverse primers were chosen. One primer (F12/F15r) corresponded to a predicted antigenic epitope shared with F15, whereas the other (F12r) was unique to F12 at the nucleotide level but corresponded to a predicted nonantigenic peptide region shared by the F12 and F16 variants (not shown).

Reverse primers were sorted into three pools of four primers each while maintaining within each pool compatibility between primers and a distribution of product sizes that would allow ready resolution of products in agarose gels (Fig. 2 and 3). Reverse primers were tested in combination with the universal forward primer (Ff), first individually and then in multiples, against single and multiple positive and negative control DNAs. PCR conditions and primer sequences were adjusted as needed to achieve balanced and specific amplification of all F variants with the derivation set strains (Fig. 3).

FIG. 2.

Distribution of primers and amplicons for multiplex F PCR assay. The universal forward primer Ff (forward-pointing arrowhead above the 5′ end of papA), in combination with the reverse primers (backward-pointing arrowheads below papA), yields a PCR amplicon (wide bars) of specific length with template DNA representing each allele of papA. The allele-specific amplicons can be resolved by size in agarose gels when reverse primers are sorted into three pools, i.e., A (white), B (hatched), and C (black).

FIG. 3.

Gel electrophoresis of multiplex F PCR assay products. PCR was done using primer pools A, B, and C (plus universal forward primer Ff) with pooled template DNA from pap-positive control strains representing all 11 known F types (plus lanes) and pap-negative control DNA (minus lanes). Amplicons of the expected sizes appear in each of the positive control lanes. No PCR products appear in the negative control lanes. Lanes M, 100-bp ladder. The values on the left are molecular sizes in base pairs.

F PCR methods.

Amplification was done in a 25-μl reaction mixture containing template DNA (2 μl of boiled lysate [37]), 4 mM MgCl₂, the four deoxynucleoside triphosphates at 0.8 mM each, 0.6 μM concentration of each primer (except for those marked with an asterisk below, which were used at a concentration of 0.3 μM), and 2.5 U of AmpliTaq Gold in 1× PCR buffer (Perkin-Elmer, Branchburg, N.J.). The three primer pools, with the reverse primers listed in order of decreasing amplicon size within each pool, were as follows: pool A, *F15r (gctacattcttgccacttgc; 455 bp), F13r (gggtattagcatcaccttcggag; 400 bp), F14r (gcagcatatctttattgttccc; 320 bp), and *F11r (ggcccagtaaaagataattgaacc; 277 bp); pool B, *F9r (aaggccccgttgacgtttt; 416 bp), *F7-1r (tttcacccgttttccactcg; 375 bp), *F8 (gtaccacctacagcacttgg; 251 bp), and *F12/15r (aattcttgggcgttgaggatcca; 179 bp); pool C, *F12r (cccatcgacaagacttgaca; 393 bp), F10r (ctcctcattatgaccagaaaccct; 312 bp), F16r (gttcccgctttattaccagc; 239 bp), and F7-2r (tttgggttgactttccccatc; 183 bp). All three primer pools also contained the forward primer Ff (Fig. 2).

The reaction mixture was heated to 95°C in an automated thermal cycler (PTC-100-96; MJ Research, Watertown, Mass.) for 12 min to activate the AmpliTaq Gold. This initial denaturation was followed by 25 cycles of denaturation (94°C, 30 s), annealing (67°C, 1 min), and extension (72°C, 1.5 min) and a final extension step (72°C, 10 min). Samples were electrophoresed in 2% agarose gels, stained with ethidium bromide, destained with distilled water, and then photographed using a UV transilluminator and digital capture system (Gel Doc; Bio-Rad, Hercules, Calif.). The sizes of the amplicons were determined by comparing them to a 100-bp DNA ladder (Gibco-Bethesda Research Laboratories, Gaithersburg, Md.) which was run on the same 2% agarose gel (Fig. 2 and 3).

Confirmation of specificity of F PCR products.

To confirm that a PCR product of a size corresponding to one of the predicted F-specific amplicons actually represented that papA variant, the nucleotide sequence of at least one putative F-specific amplicon was determined for each F type as derived from a strain other than the source strain for the corresponding type-specific papA sequence (Table 3). Amplicons were purified using the QIAquick PCR purification kit (Qiagen Inc., Chatsworth, Calif.) or MicroSpin S-300 HR columns (Amersham Pharmacia Biotech, Piscataway, N.J.) and subsequently directly sequenced by the dideoxy-chain termination method (88). Experimentally determined sequences were compared with known papA sequences by using the BLAST algorithm (3) (http://www.ncbi.nlm.nih.gov/BLAST/) to find the closest match in the GenBank, EMBL, DDBJ, and PDB sequence data banks.

TABLE 3.

Confirmation of putative F-specific amplicon sequences

Source straina	Putative F type of amplicon	Unambiguous nucleotides in sequenced amplicon (no.)	BLAST search result
			Best match			2nd best match
			% Identity with best match	F type of best match	Accession no. of best match	% Identity with 2nd best match	F type of 2nd best match	Accession no. of 2nd best match
CFT073	F7-1	346	99.4	F7-1	X02921	39.6	F14	Y08928
CFT073	F7-2	158	96.2	F7-2	M12861	67.1	F16	Y08930
C351-82	F8	227	97.8	F8	Y08931	83.7	ECOR 48e	AF051811
C483-82	F9	339	92.6	F9	M68059	14.2	F14	Y08928
3669b	F10b	275	93.1	F10	Y08927	60.7	F13	X61239
IA2	F11	258	95.3	F11	L07420	77.1	ECOR 48e	AF051811
C1979-79	F12	359	98.3	F12	X62157	91.9	F15	Y08929
C1979-79	F12/15c	154	98.7	F12	X62157	98.1	F15	Y08929
CP9	F13	367	94.3	F13	X61239	45.0	F10	Y08927
CP9b	F14b	301	100	F14	Y08928	70.4	F7-1	X02921
C372-82	F15	427	99.7	F15	Y08929	92.7	ECOR 46e	AF051810
C372-82	F12/15d	156	97.4	F12	X62157	96.8	F15	Y08929
IA2b	F16b	220	97.3	F16	Y08930	89.1	F15	Y08929

^aStrains are as shown in Table 1. 

^bThe F10, F14, and F16 amplicons from strains 3669, CP9, and IA2, respectively, represented extra F types. Extra F10 amplicons from F8 and F15 control strains C1254-77 and C1805-79 also were confirmed as representing authentic F10 sequence (data not shown). 

^cFrom F12. 

^dFrom F15. 

^eF = antigen type not defined. 

Detection of papG alleles, other pap elements and virulence factor genes, hemolysin, and hemagglutination.

In selected strains, the three papG alleles were detected using an established allele-specific PCR assay (39). Other pap elements and non-pap virulence factors were detected by PCR and/or by dot blot hybridization (under high-stringency conditions) by using primers, probes, and experimental conditions previously described (43, 46). Hemolysin production was assessed by growth on 5% sheep blood agar. Mannose-resistant hemagglutination (MRHA) was assessed using human A₁P₁ and sheep erythrocytes, with pigeon egg white used as a specific inhibitor of P fimbriae to differentiate P pattern MRHA from non-P MRHA, as previously described (36).

Cosmid cloning of pap operons from strain CP9.

To isolate each of the two pap operons from control strain CP9 (Table 1) for separate analysis, transformants of pap-negative K-12 host strain NM554 containing a previously constructed cosmid library from strain CP9 (85) were incubated with human or sheep erythrocytes at 4°C in excess α-methyl-d-mannose to allow selective hemadsorption of bacterial clones expressing PapG allele I or III, respectively (38). Following hemadsorption, erythrocytes were extensively washed and then plated on selective agar. Colonies from overnight growth at 37°C which exhibited the appropriate MRHA phenotype were confirmed as containing the appropriate papG allele by PCR (37).

Sequencing of novel papA variants.

Certain strains were positive for papA by dot blot assay and by flanking-primer PCR but were negative with the F PCR assay. Consequently, papAH amplicons from these strains (generated using forward consensus primer Ff and reverse primer PapAr [5′-cgtcccaccatacgtgctcttc-3′] which is from the 5′ end of papH) were directly sequenced as described above for putative F-specific amplicons. Sequences were analyzed as described above for other papA sequences. A specific reverse primer (F48r [5′-gttcattggcttggattg-3′]) which was compatible with the pool A reverse primers was designed for one of these new papA variants (F48) using the approach described above for the other reverse primers and was used to screen selected strains in combination with the universal forward primer Ff.

RAPD fingerprinting.

Random amplified polymorphic DNA (RAPD) genomic fingerprints were generated for selected strains using boiled bacterial lysates as template DNA, Ready-To-Go PCR beads (Pharmacia), decamer primer 1290, and amplification conditions previously described (100). Fingerprints were compared side by side in ethidium bromide-stained 1% agarose gels.

Serological methods.

Historical O:K:H and F types were from the literature or from the records of the IEKC. Confirmatory serotyping was done for selected strains as part of the present study by the IEKC. Lipopolysaccharide (O), capsular (K), and flagellar (H) antigens were determined by using the established typing sera and methods specified by Ørskov and Ørskov (66). F determination was done by using rocket immunoelectrophoresis, followed by crossed-line immunoelectrophoresis, as previously described (68). For selected strains, 10× concentrated F10 antiserum was used to enhance detection of the F10 antigen.

Statistical methods.

Comparisons of proportions were tested using Fisher's exact test with a significance threshold of P < 0.05.

Nucleotide sequence accession numbers.

The sequences determined in this study have been deposited in GenBank under accession no. Y08931 (F8), Y08927 (F10), Y08923 (F14), Y08929 (F15), and Y08930 (F16).

RESULTS

Derivation set.

After optimization of assay conditions and primer sequences, the F PCR assay yielded the predicted F-specific amplicon(s) for each of the 11 control strains in the derivation set without cross-reactivity between F types. With pooled template DNAs, robust amplicons could be generated simultaneously in a single reaction mixture for all four of the F types represented within each primer pool and products could be readily resolved in agarose gels (Fig. 3).

First validation set.

Among the 73 strains in the first validation set, many PCR-serotype discrepancies were encountered (Table 1). Thirty-seven (51%) of the strains exhibited at least one such discrepancy. All 37 discrepant strains had an F type detected by PCR that was not expected based on known serological data (i.e., a putative false-positive PCR result); three also had a presumed serological F type that was not detected by PCR (i.e., a putative false-negative PCR result). The putative false-positive PCR results consisted mostly of F10, F14, or F16 PCR products appearing in strains which were not known to be serologically positive for these F types. Thus, although the PCR assay appeared to be highly sensitive for detection of serologically known F types, its specificity was less clear.

Three explanations were considered as possible explanations for the observed PCR-serology discrepancies, including (i) PCR contamination, (ii) nonspecific amplification due to primer cross-reactivity, and (iii) the true presence of (serologically occult) heterologous copies of papA. The first possibility, PCR contamination, was ruled out by a series of reproducibility experiments (data not shown). To differentiate between the second and third possibilities, DNA sequences were determined for representative (putative) F10-, F14-, and F16-specific amplicons from control strains 3669, CP9, and IA2, respectively, in which these products were serologically unexpected, and also for the expected F PCR amplicons from these strains (i.e., F9, F13, and F11, respectively). In each instance, the closest match in the sequence data banks for a strain's “extra” F amplicon corresponded to the strain's PCR-detected extra F type rather than with its known serological F type, whereas the closest match for the strain's serologically predicted F amplicon was the corresponding control papA sequence (Table 3). Similar direct sequence confirmation that the strain's unexpected putative F10 amplicon represented the authentic F10 papA sequence was obtained for F8 and F15 control strains C1254-77 and C1805-79 (data not shown). These results excluded primer cross-reactivity and, instead, strongly suggested the presence of true (serologically unrecognized) heterologous copies of papA in the extra-F strains.

As a second proof of the presence of serologically occult papA copies, the two pap operons from strain CP9 (which gave an unexpected F14 amplicon, as well as the expected F13 amplicon) (Table 1) were isolated as separate cosmid clones for independent analysis. Both cosmid clones expressed P pattern MRHA, evidence that the pap operons they contained were functional. One clone was nonhemolytic, contained papG allele III by PCR, and yielded the F13 product in the F PCR assay (Table 1). The other was strongly hemolytic, contained papG allele I by PCR, and yielded the F14 product in the F PCR assay (Table 1). Thus, the sum of the F PCR results for CP9's two pap cosmid clones equaled the F PCR result for the wild-type parent. These findings confirmed the presence in CP9 of two independent and fully functional copies of the pap operon, each with its own papA allele, as predicted by the F PCR assay.

Further evidence that detection by PCR of multiple F types correlates with the presence of multiple discrete copies of papA was provided by F PCR analysis of strain CFT073 and of cosmid clones representing its two pap operons, which were known to produce different-size pilins (60); hence, the two pap operons were hypothesized to contain distinct papA alleles. The wild-type parent was F7-1 plus F7-2 by PCR, and the two cosmid clones were F7-1 and F7-2, respectively. These results were subsequently confirmed by serology (Table 1). (Discovery of the same F profile in strain CFT073 as in F7-1, F7-2 control strain AD110 [Table 1] prompted RAPD fingerprinting of both of these O6:K2:H1 strains, which confirmed a close genomic relationship between them [data not shown].)

The above-described series of experiments suggested that most of the putative false-positive F PCR results actually represented the presence of true papA copies which either were not expressed or were expressed but not detected serologically. To differentiate between these two possibilities, selective confirmatory reserotyping was done using state-of-the-art methods (including, when indicated, 10× concentrated F10 antiserum) for 33 of the 37 strains from the first validation set for which discrepancies had been noted between the F PCR and historical F serotypes.

Repeat serological testing brought serology into agreement with PCR for 21 (64%) of the 33 previously discrepant strains, confirming that most of the initially observed PCR serology discrepancies represented serological false negatives. The 12 strains which, after confirmatory serotyping, had persisting discrepancies included 11 instances of a PCR-detected extra F type, mostly F10 in an F8-, F9-, or F15-positive control strain, and a single instance of a serological F type (a newly identified F10) not detected by PCR.

To clarify whether these residual PCR-serological discrepancies were due to nonexpression of papA versus expression of papA without serological detection, the three remaining extra-F strains (C659-81, G1062a, and C826-83) which were F PCR positive only for their extra F type (i.e., F10) and were serologically F negative, i.e., did not have a second serological or PCR-defined F type to confound the analysis (Table 1), were tested for multiple pap elements, for other adhesin genes, and for MRHA phenotype (Table 4). Two of the strains (C659-81 and C826-83) were found to have a fragmentary pap operon and not to express P-pattern MRHA and, hence, almost certainly did not express an F10 PapA pilin (Table 4). In contrast, strain G1062a had a complete copy of pap and expressed P pattern MRHA, consistent with P fimbrial expression. With F10 as its sole papA allele, this strain would be expected to express P fimbriae of the F10 type. Thus, this strain's F seronegativity seemed probably to represent a serological false negative.

TABLE 4.

Adhesin genotypes and phenotypes of three control strains positive for F10 by PCR but not by serology

Strainb	O:K:H serotype	F serotype		PCR	Adhesin genotypea							MRHA pattern
		Putativec	Confirmedd		papAH	papC	papEF	papG	papG allele	afa/draBC	sfaS
C659-81	O1:K51:H−	F8	F−	F10	+	+	+	−	−	+	−	Non-P
G1062a	O4:K7:Hr	NAe	F−	F10	+	+	+	+	III	−	−	P
C826-83	O75:K5:H−	F15	F−	F10	+	−	−	−	−	−	−	−f

^aAs determined by PCR, except for papAH in strain C826-83 (PCR negative, probe positive for papAH). +, wild type; −, mutant. 

^bStrains C659-81 (alias FS09) and C826-83 (alias FS30) were from the IEKC laboratory. Strain G1062a (alias C892-97) was from the J.R.J. laboratory (Table 1). 

^cBased on historical serotyping results for strains from the IEKC. 

^dIncluding with 10× concentrated F10 antiserum. 

^eNA, not applicable (strain G1062a was first serotyped during the present study). 

^f−, HA negative. 

Second validation set.

Among strains in the second validation set, although agreement was poor between F PCR and both the historical and the (initial) confirmatory F serotype results, correspondence of PCR to serology improved as serological F profiles were refined in successive rounds of confirmatory serotyping (Table 2). By the third round of serotyping, serology agreed with PCR with respect to the F profile of eight of the nine strains and for 15 (94%) of the 16 F types detected by one or both methods (Table 2). These findings confirmed that conventional F serotyping often substantially underestimates the potential F repertoire of wild-type P-fimbriated strains and that most PCR serotype discrepancies are due to false-negative serological results rather than to PCR false positivity.

Prevalence of papA variants among 75 urosepsis isolates.

With its sensitivity and specificity rigorously confirmed, the F PCR assay was next used to assess the papA allele status of 75 blood isolates from patients with urosepsis. Fifty-nine (79%) of the 75 strains were found by PCR to contain one or more of the 11 known papA alleles. Of the F PCR-positive strains, 44 had a single F type, 13 had two F types, and 2 had three F types, for a total of 76 detected F types (Fig. 4). Each of the 11 known F types occurred at least once in the population, in decreasing order of prevalence, as follows: F10, 27% (20 strains); F11, 21% (16 strains); F7-2, 17% (13 strains); F16, 11% (8 strains); F14, 8% (6 strains); F8 and F12, 4% (3 strains) each; F7-1, F9, and F13, 3% (2 strains) each; F15, 1% (1 strain). Thus, the three most prevalent F types (F10, F11, and F7-2) accounted for nearly two-thirds (64%) of the F types detected. F7-2, F11, and F16 tended to occur alone (33 [89%] of 37 occurrences), whereas F8, F9, F12, F13, F14, and F15 usually occurred in combination with another F type (15 [88%] of 17 occurrences). The remaining F types (F7-1 and F10) occurred both alone and in combination, each with approximately the same frequency (Fig. 4).

FIG. 4.

papA alleles among 75 E. coli isolates from patients with urosepsis. Strains are sorted according to O:K:H serotype within each carboxylesterase B type (CBT). Results for papAH are shown as positive if the blot and PCR assays were both positive, as negative if the blot and PCR assays were both negative, and as negative with an asterisk if the blot assay was positive but the PCR assay was negative. F PCR results are listed in the aggregate and also shown distributed according to individual F types. Serogroup O75 strains V21, V8, and PM3 (shown as papAH-negative) exhibited trace reactivity with the papAH probe in some blots. Carboxylesterase B data are from reference 39, O:K:H data are from reference 41, and papAH data are from reference 46. NA, not assayed.

Phylogenetic distribution of papA alleles among urosepsis isolates.

Several of the papA alleles exhibited a clear-cut phylogenetic distribution, as evidenced by their strong associations with particular O:K:H serotypes (which usually equate with genetic clonal groups) or carboxylesterase B types (Fig. 4; Table 5). It is noteworthy that although F10 was significantly associated with carboxylesterase B type B2 in general, it was most highly correlated specifically with serogroup O75 (Table 5). Similarly, although F16 was significantly associated with carboxylesterase B type B1 in general, this was due to its independent strong associations with three small subgroups among the B1 strains, i.e., O15:K52:H1, O25:K2:H2, and serogroup O157 (Table 5).

TABLE 5.

Association of F-specific papA alleles with serotype and carboxylesterase B type among 75 E. coli urosepsis isolates

papA allele (no. of isolates)	Group	No. with papA allele/total (%)		P value
		Group members	Other strains
F7-2 (13)	O6:K2:H1	10/11 (91)	3/64 (5)	<0.001
F10 (20)	Carboxylesterase B type B2	17/48 (35)	3/27 (11)	0.03
	Serogroup O75	5/5 (100)	15/70 (21)	<0.001
F11 (10)	O1:K1:H7, O2:K1:H7	10/10 (100)	56/65 (9)	<0.001
F12 (3)	O12:K1:H6	2/4 (50)	1/71 (1)	0.006
F13 (2)	O6:K5:H−	2/2 (100)	0/73 (0)	<0.001
F14 (6)	O2:K5:H1, O2:K7:H1	3/5 (60)	3/70 (4)	0.003
F16 (8)	O4:K12:H1	2/3 (67)	6/72 (8)	0.03
	Carboxylesterase B type B1	6/27 (22)	2/48 (4)	0.02
	O15:K52:H1	2/2 (100)	6/73 (8)	0.01
	O25:K2:H2	2/2 (100)	6/73 (8)	0.01
	O157	2/2 (100)	6/73 (8)	0.01
F48a	O12:K1:H6	2/4 (50)	2/71 (3)	0.01

^aThe F48 allele was discovered as part of present study. 

Associations of papA variants with other bacterial traits among urosepsis isolates.

Several of the papA variants exhibited statistically significant associations with specific bacterial traits, including other papA alleles (Table 6). F7-2 was highly correlated with probe positivity but PCR negativity for the group II capsule synthesis genes kpsMT, a genotype which largely equates with K2 capsule (46). F10 and F14 each exhibited multiple associations with other traits, including an association with one another (Table 6). The association of F11 with the K1 variant of kpsMT (Table 6) was consistent with its association with serotypes O1:K1:H7 and O2:K1:H7 (Table 5). In keeping with the association of F16 with serogroup O4 (Table 5), F16 was also associated with the O4 lipopolysaccharide synthesis gene rfc (Table 6).

TABLE 6.

Association of F-specific papA alleles with other bacterial traits among 75 E. coli urosepsis isolates

papA allele (no. of isolates)	Associated trait	No. with papA allele/total (%)		P value
		With trait	Without trait
F7-2 (13)	kpsMT II probe positive, PCR negative	12/13 (92)	1/62 (2)	<0.001
F10 (20)	cnf-1	8/12 (75)	12/63 (19)	0.001
	K5 kpsMT II	16/25 (64)	4/50 (18)	<0.001
	traT	9/50 (18)	11/25 (44)	0.03
F11 (16)	K1 kpsMT II	12/21 (57)	4/54 (7)	<0.001
F14 (6)	F10	4/20 (25)	2/55 (4)	0.04
	papG allele III	3/5 (60)	3/70 (4)	0.003
	cnf-1	4/12 (33)	2/63 (3)	0.005
	K5 kpsMT II	6/25 (24)	0/50 (0)	<0.001
F16 (8)	rfc	2/3 (67)	6/72 (8)	0.03
F48a (4)	F12	2/3 (67)	2/72 (3)	0.006

^aThe F48 allele was discovered as part of the present study. 

Discovery of novel papA variants.

Two of the urosepsis isolates, U7 and V29, were negative for the 11 known papA alleles according to the F PCR assay but still expressed P pattern MRHA and by both dot blot and PCR contained papAH, papC, papEF, and papG; hence, they were suspected of harboring novel variants of papA. Consequently, the papAH amplicons from these two strains were sequenced, translated into peptides, and compared with the 11 known PapA variants. Both new sequences appeared to represent novel papA alleles, since each was separated from its nearest neighbor in the PapA tree by a distance greater than that separating the two most closely related of the 11 known PapA variants (i.e., F12 and F15) (Fig. 1); one was actually as distant from its nearest neighbor as any of the known PapA variants was from its own nearest neighbor. The new PapA variant from strain V29 (which was termed F40 because of this strain's laboratory code number) was closely related to the F10 peptide (Fig. 1). PapA from strain U7 (which was termed F48) was an outlier member of the cluster that included the F40, F8, F13, F10, and F11 PapA variants (Fig. 1).

When combined with the universal forward primer (Ff), a reverse primer designed for the new F48 papA variant yielded the expected 176-bp amplicon from source strain U7 and did not react with control strains for 11 known F types (data not shown). Reamplification of all 75 urosepsis isolates with the new F48 primer detected a putative F48 papA allele in three additional strains, each of which was already known to contain at least one papA allele (Fig. 4). Direct sequencing of the putative F48 amplicons from these strains confirmed each as bona fide F48 (data not shown). The new F48 papA variant was found to be significantly associated with (but not confined to) serotype O12:K1:H6 (Table 5). It was statistically significantly associated with the F12 papA variant, from which it was quite distant at the nucleotide and peptide levels (Fig. 1 and 2), but not with other bacterial traits (Table 6). Because of this association of F48 with F12, the new F48 primer was used to test the F12 control strains. Although the F12 control strains from the first validation set were negative (data not shown), strain AD309 (second validation set), in which F12 had been detected serologically but not by PCR (which was one of only two instances, among all of the control strains, of a putative false-negative PCR result), was positive with the F48 primer (Table 2).

F10 amplicons from pap-negative O75 strains.

Among the urosepsis isolates, the F PCR assay showed all five representatives of serogroup O75 to contain the F10 papA allele (Fig. 4). This was surprising, since four of these strains had otherwise tested negative for all pap elements by both PCR and (except for one papAH probe-positive strain) dot blot assay. Direct sequencing of the putative F10 amplicons from these five strains confirmed that the amplicons represented the authentic F10 papA sequence (data not shown). Reexamination of the duplicate dot blots for the three putatively pap-negative O75 strains revealed variable faint reactivity with the papAH probe, consistent with the presence of a partial copy of papA, but no reactivity whatsoever with the probes for papC, papEF, and papG (data not shown). These findings, together with those for the serogroup O75 F15 control strain C826-83 (Table 4) and for several of the other serogroup O75 F8 and F15 control strains (first validation set [Table 1]), suggested that wild-type strains of serogroup O75 commonly contain isolated partial copies of the F10 papA allele, with or without a separate (complete) pap operon containing a different papA allele.

DISCUSSION

In the present study, we developed and rigorously validated a novel PCR-based assay for the 11 recognized variants of papA, the major pilin gene of E. coli P fimbriae, and then used the assay to assess the prevalence, phylogenetic distribution, and bacteriological associations of the papA alleles among 75 E. coli isolates from patients with urosepsis. The assay was extremely sensitive and specific, evidence that papA sequences are highly conserved within each of the traditionally recognized F serotypes despite the diversity observed among F types. The assay revealed considerable segregation of papA alleles according to O:K:H serotype, consistent with vertical transmission within clones, but with exceptions which strongly suggested horizontal transfer of papA alleles between lineages. Two novel papA variants were identified, one of which was actually more prevalent among the urosepsis isolates than were several of the known papA alleles.

The assay's sensitivity approached 100% for detection of serologically evident F types among the diverse control strains in the two validation sets, evidence that the sequences selected for use as F-specific primers are highly conserved within each serological F type. This suggests that the regions targeted by the primers either encode the actual epitopes responsible for F serospecificity (as intended by our primer design strategy) or, if they do not, are nonetheless tightly linked with them. The high degree of homology observed between entire F-specific amplicons and the corresponding F-specific papA source sequences (Table 3) suggests that type-specific homology is not limited to just the primer-binding regions but is broadly present throughout papA. Full-length sequence analysis of multiple representatives of each F type from phylogenetically distinct backgrounds (8) is needed to more thoroughly evaluate this hypothesis.

The observed patterns of discrepancy between PCR and serology with respect to detection of F types were highly informative. Historically determined serotypes, understandably, did not include certain F types which, at the time of the original serotyping, were uncharacterized or not recognized as corresponding to P fimbriae. This probably accounted for many of the initial discrepancies involving F10, F14, and F16 (F. Scheutz, unpublished data), which constituted the bulk of the extra F types detected by PCR but putatively not by serology (Table 1). Although confirmatory serotyping (which was done initially using standard techniques and reagents and then using intensified methods) did eventually confirm most of the PCR-detectable F types, a small subset remained serologically occult.

At least some of these serologically occult papA variants represented nonfunctional copies (or fragments) of papA, as documented for one of the F15 control strains (O75:K5:H5) (Table 4) and for four of the serogroup O75 urosepsis isolates (Fig. 4). What proportion of the serologically occult papA copies might still be functional under the appropriate conditions (with the possibility of expression, of course, limited to those strains which have an intact pap operon either in cis or in trans to the serologically inapparent papA copy) can only be speculated upon. However, the expression of P pattern MRHA by strain G1062a (Table 4) suggests that at least some of these occult papA copies are functional. In any event, the preponderance of evidence indicated that the residual PCR-positive, serotype-negative discrepancies did not represent false-positive PCR results, at least not in the conventional sense of nonspecific primer binding or contamination, but instead represented detection of true papA sequences.

Alternative explanations for serological nondetection of documented papA alleles, in addition to nonexpression, could include low-level expression (consistent with the demonstrated differences in expression level between different pap operons in a single strain) (55) or expression of an antigenically altered pilin. Either hypothesis would be consistent with the improved serological detection of F10 that was observed in the present study with the use of 10× concentrated antiserum.

The PCR assay provided abundant evidence among both the control strains and the urosepsis isolates of clonal segregation of papA alleles, as reflected by the distribution of certain papA alleles according to O:K:H serotype and carboxylesterase B type. This is consistent with previously reported serological findings (68, 70). Also noted were circumstances suggesting horizontal transfer of papA alleles. These included the appearance of diverse papA alleles in different members of a single O:K:H serotype, e.g., O2:K7:H−, O4:K12:H1, and O7:K1:H− (Fig. 4), and the local predominance of a particular papA allele in evolutionarily distant lineages, e.g., F16 in three non-B2 O:K:H serotypes (O15:K52:H1, O25:K2:H2, and serogroup O157), but also in B2 serotype O4:K12:H1 (Fig. 4; Table 5).

The new assay also facilitated the discovery of novel papA variants by revealing the absence of any of the 11 known papA alleles in two strains that expressed P pattern MRHA and were papAH positive by probe and PCR. Sequencing of the papAH amplicons from these two strains revealed a unique papA variant in each strain. The subsequent detection of one of the new papA variants (F48) in 3 of the 75 urosepsis isolates by using a primer based on the F48 papA sequence illustrates how the F PCR assay can be used to discover new papA alleles and then screen for them. Discovery of a statistically significant association of the new F48 papA variant with F12 prompted retesting of all of the F12 control strains with the new F48 primer, including the strain in which the PCR assay had putatively missed an F12 papA allele. Only the latter strain was F48 positive. We are currently investigating the phenotypic correlates of the F48 papA allele, including the relationship between F48 and F12 (J. R. Johnson, unpublished data).

Further evidence of the PCR assay's utility as a strain typing tool was provided by the close relationship the assay revealed between two archetypal uropathogenic strains, CFT073 and AD110. CFT073 has been used in studies of pathogenesis (59, 60) and is currently under intensive investigation with respect to the distribution and composition of its pathogenicity islands, including through complete sequencing of its genome by the Fred Blattner laboratory (24, 50; H. L. T. Mobley, personal communication). Venerable strain AD110 was the original source of the first two numbered P fimbrial F types (F7-1 and F7-2) (69) and of recombinant derivatives thereof (95, 97) and has been used in many subsequent studies regarding the regulation of the pap operon, P fimbrial structure, and immunological aspects of PapA (12, 13, 55, 79, 80, 101). The finding that CFT073 has the same distinctive F-1, F7-2 papA allele profile as does AD110 called attention to these two strains' shared O6:K2:H1 serotype, which suggested a clonal relationship, as subsequently confirmed by RAPD fingerprinting. Prior to the discovery of strain CFT073's papA allele profile, neither we nor H. L. T. Mobley (from whose laboratory CFT073 originated) were aware of the considerable similarities between these two archetypal uropathogenic strains (Mobley, personal communication). The discovery of these similarities is significant in that it reveals that there has been an (unrecognized) continuum of investigation over the past 2 decades in different laboratories regarding the pathogenetic mechanisms of what may be essentially the same model uropathogenic strain.

In view of the presence of both the F7-1 and F7-2 papA alleles in model pathogens CFT073 and AD110, and occasionally in other O6:K2:H1 strains (68, 70; Johnson, unpublished data), the absence of the F7-1 allele among the O6:K2:H1 urosepsis isolates in the present study is interesting. Almost all of the O6:K2:H1 urosepsis isolates, which constituted the population's single largest clonal group and quite probably equate with virulent clone 4 as described by Korhonen et al. and with clone V of Väisänen-Rhen et al. (53, 90, 92), were positive for F7-2 alone. Whether O6:K2:H1 strains with both F7-1 and F7-2 are simply uncommon enough that by chance alone none was represented in the present study population or whether there are selection factors that favor F7-2-only strains over F7-1-plus-F7-2 O6:K2:H1 strains as agents of urosepsis remains to be determined. It will be useful to survey O6:K2:H1 (and other) E. coli isolates from different extraintestinal syndromes for their papA allele content, in part to assess how generally representative are the strains that have emerged as model uropathogens (55, 60, 83, 84, 91) and in part to determine which F types would be most relevant to include in a syndrome-based PapA vaccine (12). The F PCR assay, which was simple, rapid, and easy to interpret, should greatly facilitate future studies of this sort.

This prediction is supported by the results of a “portability” assessment of the F PCR assay, in which 10 wild-type pap-positive strains of unknown F status plus strain J96 (an F13 control) were independently F typed in three different laboratories in the United States (J.R.J.), Canada (J. Fairbrother), and France (E. Oswald). The latter two laboratories had only just begun to use the assay when the testing was done. Consistent results were obtained for each strain in each laboratory (Johnson, unpublished data; E. Oswald and J. Fairbrother, personal communication, 1999).

In summary, the F-specific papA alleles of extraintestinal E. coli were found to be sufficiently conserved within each F type that highly sensitive and specific detection was possible by PCR using a small set of F-specific oligonucleotide primers. The multiplex PCR assay for the 11 known papA variants as developed and extensively validated in the present study arguably outperformed conventional F serotyping in accurately detecting papA variants. It detected serologically unrecognized papA copies, demonstrated both clonal segregation and horizontal transfer of papA alleles within the E. coli population and facilitated the discovery of two novel papA variants. The F PCR assay represents a versatile new molecular tool for epidemiological and phylogenetic investigation of the diverse papA alleles of extraintestinal pathogenic E. coli.