INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

P fimbriae of extraintestinal pathogenic Escherichia coli (ExPEC) are heteropolymeric proteinaceous fibers that mediate adherence to Gal(1-4)Gal-containing isoreceptors on host cell surfaces (13). PapG, the P fimbrial tip adhesin molecule, is responsible for digalactoside-specific receptor recognition and binding (10, 13, 38). A thin, flexible fibrillum connects PapG to the main fimbrial shaft (13). The fimbrial shaft in turn is composed of hundreds to thousands of identical PapA structural subunits, linked end-to-end by "donor strand exchange" to form a rigid alpha-helix which is attached to an anchor protein in the outer membrane (13, 48). Adherence mediated by P fimbriae contributes to the pathogenesis of extraintestinal infections such as pyelonephritis (17, 46) and may promote intestinal colonization of the host (59, 60).

PapG occurs in three known molecular variants (I to III), which are encoded by the three alleles of the corresponding gene, papG (32, 40). The three PapG variants exhibit subtly different receptor binding preferences (15, 19, 32-34, 36, 37, 40, 50, 52-55). They also exhibit divergent associations with clinical syndromes, with specific antigenic variants of the major structural subunit PapA, and with phylogenetic groups. PapG variant III (PapG III) requires for binding terminal substitutions on the Gal(1-4)Gal consensus receptor (34, 36, 50, 53, 54). Thus, it mediates agglutination of sheep erythrocytes, which contain Forssman glycolipid [with its extended Gal(1-4)Gal-containing oligosaccharide chain] and human erythrocytes (which contain the extended digalactoside-containing glycolipid sialosyl-Gal-globoside), but not Gal(1-4)Gal-coated latex beads (P beads) or neuraminidase-treated human type O erythrocytes (25, 34, 36, 50, 53). Clinically, PapG III is associated with cystitis in humans and with genitourinary infections in dogs and cats (16, 25, 26). PapG III typically occurs in conjunction with the F12, F13, and F14 PapA variants (22, 31) and is concentrated within E. coli phylogenetic group B2 (22, 40).

PapG variant II (PapG II) binds well to both terminally substituted and nonsubstituted Gal(1-4)Gal-containing isoreceptors and hence mediates agglutination of sheep erythrocytes, human erythrocytes (irrespective of neuraminidase treatment), and P beads (25, 34, 36, 50, 53). Clinically, it is associated with pyelonephritis and bacteremia in humans (16, 18, 21, 43). It often occurs in conjunction with the F7-2, F10, and F11 PapA variants (22, 31). It is distributed across E. coli phylogenetic groups B2 and D and occurs sporadically in other phylogenetic groups (22, 40).

PapG variant I (PapG I), although able to bind both substituted and nonsubstituted Gal(1-4)Gal-containing isoreceptors, agglutinates sheep erythrocytes poorly but agglutinates human erythrocytes and P beads well (25, 36, 39, 40, 53, 54). Until recently, PapG I had been encountered only in pyelonephritis isolate J96 (O4:K:H5) (27), which has two pap operons, one with papG allele I and the other with papG allele III (33, 39); both operons contain the F13 papA allele (31, 39). Recently, however, a disseminated clonal group of "J96-like" strains of E. coli O4:H5 was discovered that includes archetypal extraintestinal pathogenic strains J96 and CP9 (27, 29). The members of this clonal group, like J96 and CP9, typically contain papG alleles I and III plus the F13 papA allele, with or without another papA allele (27, 29). These findings suggested that although papG allele I is not unique to strain J96, it nonetheless may be restricted to the J96-like clonal group and may occur only in conjunction with the F13 papA allele; hence, it may represent a comparatively recent evolutionary development within (or acquisition by) E. coli.

In this report we describe wild-type strains and variants of PapG I that contradict conventional assumptions regarding PapG I. The new findings include molecular variants of PapG I that differ both structurally and functionally from PapG I of strain J96, that occur either together with PapG II and PapG III or alone, that occur in the absence of the F13 papA allele, and that appear in lineages distant from the J96-like clonal group. Evidence also is presented that the PapG I line may actually represent the earliest evolutionary branch from a common PapG ancestor and that the J96-like clonal group may represent the original source of papG within E. coli.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Strains. Canine urinary tract infection (UTI) isolates 7U and L31 and canine rectal isolate 7R, which was from the same host as strain 7U (original designations 5557 [7U], 5808 [L31], and 5565 [7R]), were obtained from Gerald Ling. The collection from which 7U and 7R were derived comprised paired pap-positive urine and rectal isolates from eight dogs with UTI, representing serotypes O4, O6, O120, and O149 (25). Isolate L31 was from an unpublished collection of paired urine and rectal isolates from 60 dogs with UTI, representing serotypes O1, O2, O4, O6, O7, O8, and O25, plus 10 other non-UTI-associated O antigens. Strain C367-82, an O4 urine isolate from a human with UTI, was provided by Flemming Scheutz. C367-82 was derived from an unpublished collection comprising approximately 328 isolates of serogroup O4. Archetypal ExPEC strains J96 and CP9 (O4:H5), members of the J96-like clonal group, were used as controls for papG alleles I and III (25, 27). Strain IA2, a non-J96-like O4 isolate, was used as a control for papG allele II (5, 29). Other wild-type comparator strains included BOS080 and BOS060 (papG allele III) (21), CL021 (papG alleles II and III) (23), HU968-298 (sfa: S fimbriae) (25), COR149 (O4, papG allele II; provided by Jorge Blanco), 536 (O6, papG allele III) (9), BOS050 (O6, papG allele III) (21, 41), and R61 (O6, papG allele III). Recombinant comparator strains included P678-54/pRHU845 (papG allele I) (33), HB101/pDC1 (papG allele II) (5), and HB101/pJFK102 (papG allele III) (33).

Detection of pap elements and other putative E. coli virulence genes. The three papG alleles were detected both by PCR using allele-specific internal primers (Fig. 1) and by dot blot hybridization in which the allele-specific PCR products were digoxigenin labeled and used as probes, as previously described (20, 26). papG also was detected by using flanking primers, as previously described, including a consensus forward primer that detects all three papG alleles and two reverse primers, one specific for the region downstream from papG allele I in J96 and one specific for the consensus region downstream from papG alleles II and III in strains IA2 and J96, respectively (Fig. 1) (25, 30). Other pap regions (papAH, papC, and papEG) were detected by PCR, as previously described (Fig. 1) (30).

View larger version (26K): [in this window] [in a new window]   FIG. 1.   Map positions of pap primers and corresponding PCR products. Open boxes represent genes within the pap operon. Forward and reverse primers (right-pointing and left-pointing black triangles, respectively, above and below the pap operon) are used in combinations as shown to yield the indicated PCR products (thin rectangles below the pap operon). Heavily striped rectangles indicate papA and papG allele PCR products; solid black rectangles indicate pap gene PCR products; finely striped rectangles indicate long PCR operon fragments (as generated using either flanking or internal allele-specific papG reverse primers, as illustrated for allele I-I'). Adapted from Fig. 1 of reference 25.

Nucleotide sequencing and sequence analysis. After column purification, selected full-length papG amplicons underwent direct sequencing of both strands by using a dye terminator method and an automated DNA sequencer, as previously described (25, 31). Newly determined papG sequences were compared with all available papG sequences from the databanks, as previously described (25) and, to provide an outgroup, with the fimH sequence from strain J96 (GenBank accession no. AF089840) and the sfaS sequence from strain 536 (GenBank accession no. X16664). Sequences were edited and assembled, and alignments of predicted mature PapG, FimH, or SfaS peptides (and of the corresponding nucleotide sequences) were made using CLUSTAL-W (57). From these alignments, trees were inferred by the neighbor-joining (NJ) method (47) and the unweighted paired group method with averaging (UPGMA) (51) by using the application MEGA (S. Kumar, K. Tamura, and M. Nei, MEGA: Molecular Evolutionary Genetics Analysis, version 1.01, 1993). Hydrophilicity plots were generated by the method of Kyte and Doolittle (35) for a representative of each of the three allele-specific variants (classes) of PapG and for the newly identified PapG I' variant from strain 7U.

RAPD analysis. To assess the phylogenetic context within which the various papG variants occurred, selected wild-type strains were subjected to random amplified polymorphic DNA (RAPD) analysis with (separately) arbitrary decamer primers 1247, 1254, and 1281, as previously described (2, 58), and commercial PCR beads (24, 25). Single-primer RAPD fingerprints were digitally combined head-to-tail to create a "virtual" composite fingerprint for each strain by using a computerized image analysis system (Molecular Analyst; Bio-Rad, Hercules, Calif.). (24, 25). Composite fingerprints were then compared in a pairwise fashion by using Pearson's correlation coefficient to assess similarities between analog densitometric scans of each gel track. From the resulting similarity matrix, a dendrogram was inferred by using UPGMA (24, 25).

Hemagglutination phenotypes. Strains were tested for mannose-resistant hemagglutination (MRHA) of diverse erythrocyte types, with or without neuraminidase pretreatment of the erythrocytes, in the presence or absence of pigeon egg white (an inhibitor of P fimbriae) or fetuin (an inhibitor of S fimbriae), by using microscope slide assays as previously described (25, 45). MRHA inhibition and neuraminidase sensitivity were defined as a two-level decrease in the intensity of MRHA in the presence of inhibitor or following neuraminidase treatment of erythrocytes, respectively. P-pattern MRHA was defined as MRHA inhibition by pigeon egg white, and S-pattern MRHA was defined as inhibition by fetuin. P beads were from Chembiomed (Alberta, Canada; now defunct). Human A₁P₁ and OP₁ erythrocytes were from the investigators (J.R.J. and T.T.O., respectively). Human P^k₂ erythrocytes were provided by Jane Swanson.

Nucleotide sequences accession numbers. Newly determined papG allele I sequences were deposited in GenBank under accession numbers AF237471 (strain 7U), AF247505 (strain L31), AF237476 (strain C367-82), and AF237473 (CP9).

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Discovery of a molecular variant of papG allele I. Strains 7U and 7R, which are matching urine and rectal E. coli isolates from an individual canine host, were previously shown by allele-specific internal primer PCR to contain papG alleles II and III but not papG allele I (25). However, when amplified using a universal papG forward primer in combination with flanking reverse primers specific for papG alleles II and III versus papG allele I, respectively (Fig. 1), these strains yielded not only the expected ~1,100-bp PCR product with the papG allele II- and III-specific reverse primer but also (like strains J96 and CP9) a ~1,200-bp PCR product with the papG allele I-specific reverse primer. The possibility of a chimeric papG region consisting of papG allele II or III recombined with allele I-specific 3' flanking sequence, although consistent with the initial PCR results, was excluded by reassortment experiments in which each of three individual papG allele-specific internal forward primers was combined (separately) in PCR with the papG allele I-specific flanking reverse primer without producing an amplicon (data not shown).

View larger version (37K): [in this window] [in a new window]   FIG. 2.   Sequence comparisons of five papG allele I or I' variants and corresponding predicted PapG peptides, identified as to strain name and allele. Alignments were by CLUSTAL-W (57). Positions in the gene, as numbered from the putative start codon, or in predicted mature peptides, as numbered from the putative amino-terminal residue (31), are shown above polymorphisms. Only positions at which at least one variant differs from strain J96 are included. Dashes indicate identity to J96; asterisks indicate a gap. In the peptide alignment, symbols below the polymorphism data signify the folowing: colon, highly similar amino acids; period, somewhat similar amino acids; blank, different amino acids. The single-letter amino acid code is used.

Discovery of additional examples of papG allele I'. The modified allele-specific papG primer pool, now containing the new allele I'-specific primers, was used to screen a collection of approximately 60 canine UTI isolates of diverse serotypes and a collection of approximately 300 clinical isolates of serogroup O4 from animals and humans. Strains that were positive for an allele I-specific product were tested with the allele I and allele I' primers separately. Within each collection, a single strain was identified that yielded an appropriately sized PCR product (~470 bp) with the new allele I' primers but not the standard allele I primers (Table 1).

Analysis of predicted PapG peptides. The papG allele I sequence was determined for putative allele I'-containing strains L31 and C367-82 and for strain CP9, which was used as a second example (in addition to papG from J96) of the "orthodox" version of papG allele I (27). The four corresponding predicted PapG I peptides, and PapG I from strain 7U, were compared with all other available versions of PapG from the databases (25). In a similarity dendrogram that was rooted by using the type I fimbrial adhesin molecule FimH as an outgroup, the five predicted PapG I (I') peptides formed a cluster that was quite distant from the cluster containing representatives of PapG II and PapG III (Fig. 3). In both the NJ dendrogram (Fig. 3) and an UPGMA dendrogram (not shown), the PapG I (I') group was the first to branch off from an ancestral PapG trunk, suggesting that it represented the most basal or primitive version of PapG. Similar conformation trees were obtained when SfaS was used as the outgroup and when nucleotide data were used instead of peptide data (results not shown).

View larger version (19K): [in this window] [in a new window]   FIG. 3.   Dendrogram of predicted PapG and FimH peptides. Alignment of predicted mature peptides was done by CLUSTAL-W (57). Tree construction was done by the NJ method (47). PapG variants are identified as to class (brackets: I, I', II, or III) and individually according to strain name. FimH from strain J96 (shown) and SfaS (data not shown) were used to provide outgroups to root the tree. Similar tree conformations were obtained by UPGMA and the NJ method, whether nucleotide or peptide sequences were analyzed (data not shown). Numbers (in percentages) indicate reproducibility of nodes in bootstrap analysis.

Long-product PCR. We next sought to determine whether the newly identified papG allele I variants were integrated into intact pap operons, which should allow for expression, or instead occurred as isolated operon fragments, in which case expression would not be expected. Long-product PCR "operon walking" was done by using (flanking or internal) papG allele I- and I'-specific reverse primers, in combination (individually) with forward primers for papA, papC, and papE (Fig. 1). From each of the controls strains (J96 and CP9) and from strains 7U, C367-82, and L31, suboperonic amplicons representing papEFG, papCDJKEFG, and papAHCDJKEFG were obtained, in each instance only with the reverse primers corresponding to the variant of papG allele I known to be present in the particular strain (Fig. 4). This demonstrated that each of the papG allele I variants was contiguous with a seemingly intact pap operon. However, whereas strains C367-82 and L31 yielded papAHCDJKEFG amplicons of approximately the same size as that obtained from strain J96, strain 7U yielded an amplicon that was approximately 1.4 kb longer than those obtained from strains J96 and CP9 (Fig. 4). This suggested the possibility of an insertion or partial duplication within the allele I' pap operon of strains 7U that might account for the previously noted MRHA negativity of 7U (25). As predicted, long PCR products (i.e., papAHCDJKEFG and/or papCDJKEFG) also were obtained with the reverse papG allele II and III primers with strains 7U, 7R, and C367-82 but not with strain L31 (data not shown).

View larger version (63K): [in this window] [in a new window]   FIG. 4.   Long PCR products from strains containing papG allele I or I'. Amplification was done using the universal papA forward primer combined with the allele-specific papG reverse primer for either allele I (left, lanes 2 to 6) or allele I' (right, lanes 7 to 11). With the published papG allele I reverse primer, bands of the size expected for the papAHCDJKEFG amplicon from strain J96 (~7,540 bp [40]) were obtained with strains J96 (lane 2, ~7,430 bp) and CP9 (lane 4, ~7,540 bp) whereas no amplicon was obtained with strains 7U (lane 3), C367-82 (lane 5), or L31 (lane 6). In contrast, with the new papG allele I' reverse primer, comparable bands were obtained with strains 7U (lane 8, ~8,950 bp), C367-82 (lane 10, ~7,280 bp), and L31 (lane 11, ~6,920 bp), whereas no amplicon was obtained with strains J96 (lane 7) or CP9 (lane 10). Lanes 1 and 12, molecular weight marker (M, HindIII-restricted lambda DNA).

Agglutination phenotypes. Strain L31, which had PapG I' as its sole PapG variant and was MRHA positive, provided the only opportunity to examine the phenotype correlates of PapG I' in the absence of confounding from a coexpressed alternative PapG variant. The agglutination phenotype of L31 represented a unique combination of reaction patterns that included agglutination of P beads (characteristic of PapG I and II) and of sheep erythrocytes (characteristic of PapG II and III), but also NS-pattern MRHA of human OP₁ but not AP₁ erythrocytes (suggestive of PapG III only, according to previous data [25]; specific to PapG I' in the present study) (Table 2). Confounding by coexpression of S fimbriae, which was possible given the presence of sfa/focDE and sfaS in strain L31 (Table 1), was excluded by MRHA inhibition assays that showed P-pattern MRHA with L31 (i.e., inhibition of MRHA by pigeon egg white but not by fetuin) and S-pattern MRHA with the S fimbria control (i.e., inhibition of MRHA by fetuin but not by pigeon egg white) (Table 2).

Hydrophilicity analysis. To explore the possible structural basis for the phenotypic differences noted between PapG I' from 7U and conventional PapG I, II, and III (Table 2), hydrophilicity plots for these predicted PapG peptides were generated by the method of Kyte and Doolittle (35) and were aligned for comparative analysis (Fig. 5). The four PapG variants exhibited a consensus hydrophilicity profile characterized by six hydrophilic maxima (peaks 1 to 6) flanked by seven hydrophilic minima (valleys a to g), which included the amino- and carboxy-terminal regions of the peptides (Fig. 5). The only significant deviations from this pattern were exhibited by PapG III, which exhibited a secondary hydrophilicity minimum (designated a') and a secondary hydrophilicity peak (designated I') between minima a and b (Fig. 5). The hydrophilicity profiles of PapG I' and PapG I were similar, except that in PapG I' peaks 1 and 4 were higher and peak 5 was narrower (Fig. 5). Comparisons according to agglutination phenotypes revealed that the PapG I', II, and III peptides, which were associated with MRHA of sheep erythrocytes, exhibited higher maximal hydrophilicity values for peaks 1, 3, and 4 than did the PapG I peptide, which was not associated with this phenotype. Similarly, the PapG I', I, and II peptides, which conferred agglutination of human P^k₂ erythrocytes and P beads, exhibited higher hydrophilicity values for peak 2 than did the PapG III peptide, which was not associated with these phenotypes.

View larger version (58K): [in this window] [in a new window]   FIG. 5.   Hydrophilicity profiles of PapG peptides I', I, II, and III. Hydrophilicity was assessed by the method of Kyte and Doolittle (35), based on predicted mature PapG peptides from strains 7U (PapG I') (top), J96 (PapG I) (second from top), IA2 (PapG II) (second from bottom), and J96 (PapG III) (bottom). The horizontal axis shows residue number in the mature peptide; the vertical axis shows the hydrophilicity index. Shaded regions identify conserved hydrophilicity minima (a to g, present in all variants; a', unique to PapG III); intervening regions are hydrophilicity maxima (1 to 6, present in all variants; 1', unique to PapG III).

Clonal associations. The unusual occurrence in a serogroup O6 isolate (strain L31) of a version of papG allele I, which previously had been identified only within serogroup O4, suggested the possibility of horizontal transfer of O antigen determinants from an O6 donor into a papG allele I'-containing O4 recipient or, alternatively, transfer of papG allele I' from an O4 donor into an O6 recipient. To differentiate between these two hypotheses, strains 7U, C367-82, and L31 were compared with a panel of control O4 and O6 strains, which were selected on the basis of their status as archetypal representatives of the J96-like versus non-J96-like clonal groups within serogroup O4 (strains J96, CP9, and IA2, respectively) or their similarities to strain L31 with respect to O antigen, VF profiles, and/or papA allele content (Table 1). In a UPGMA-derived similarity dendrogram based on three-primer composite RAPD fingerprints, the papG allele I-containing O4 strains clustered closely within the larger O4 cluster (Fig. 6). Consistent with its O6 antigen status, strain L31 was placed among the O6 controls, well removed from the O4 cluster and the other papG allele I-containing strains, including the F15 papA allele-positive O4 strain COR149 (Fig. 6). This supported the hypothesis of horizontal transfer of papG allele I' from an O4 ancestor into a recipient within an O6 clonal group.

View larger version (38K): [in this window] [in a new window]   FIG. 6.   Phylogenetic analysis of PapG I'-containing strains and comparator strains. Pearson correlation coefficient analysis of composite RAPD fingerprints from primers 1281, 1247, and 1254 was used to establish pairwise similarity relationships between strains, which were then used construct a dendrogram by UPGMA (51). Strains that contain papG allele I or I' (asterisk) are shown in boldface. Strains of serogroups O4 (top) and O6 (bottom) are bracketed. Note that (papG allele I-containing) strain L31, which serologically is an O6 strain, is placed with the O6 comparator strains according to RAPD analysis. Strains J96, CP9, C367-82, and 7U constitute a "J96-like" subcluster within the larger O4 cluster. (Strains are shown in the same sequence in Table 1 as in this RAPD-based dendrogram.)

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

In the present study we used molecular and phenotypic methods to characterize novel variants of papG allele I and the corresponding wild-type E. coli strains. The new papG allele I variants exhibited unconventional agglutination phenotypes, papG and papA allele environments, and clonal associations. Comparative sequence analysis suggested that the papG allele I family may represent the earliest evolutionary branch from a common papG ancestor. In conjunction with previous findings, the data support the hypothesis that the J96-like clonal group of E. coli O4:H5 may represent the origin of papG within the species.

Our serendipitous discovery of the first known molecular variant of papG allele I led to the demonstration that genetic diversity is actually greater among representatives of papG allele I than among the known representatives of papG alleles II or III. We found that the various versions of papG allele I (or of the PapG I peptide) segregated into two clusters. One cluster included PapG I from strains J96, CP9, and C367-82, despite the reactivity of the latter strain with the alternate papG allele I' primers rather than with standard papG allele I primers. The other cluster included PapG I from canine isolates L31 and 7U. We termed this variant PapG I' and the corresponding gene papG allele I'. The novel agglutination phenotype of strain L31, which expressed papG allele I' in the absence of another papG allele, suggested that PapG I' confers a combination of the agglutination phenotypes associated with PapG I or II and those associated with PapG III. Comparison of the distinctive hydrophilicity profile of PapG I' with the hydrophilicity profiles of the three traditional PapG variants identified regions that could be responsible for the observed phenotypic similarities and differences between these four adhesin molecules. Ideally, the adherence phenotype conferred by papG allele I' should also be assessed using a recombinant strain to avoid artifacts from other adhesins and surface structures expressed by wild-type strains such as L31. However, recombinant strains may be associated with their own artifacts (11, 19).

In each strain analyzed, each copy of papG appeared to be part of an intact pap operon, as suggested by long-product PCR, which was done by using reverse primers specific for each allele of papG in combination with forward primers for upstream genes within the pap operon, as far back as papA. A possible explanation for the apparent nonexpression of papG allele I' by strains 7U and 7R was provided by the unusually long papAHCDJKEFG PCR products that were obtained from these strains with the papG allele I' reverse primer. These extra-long amplicons suggested the presence of insertions or partial duplications somewhere within the corresponding pap operons in these strains compared with strains J96 or CP9 (40).

The findings of the present study show that in contrast to what has been thought previously on the basis of available evidence, papG allele I (or I') can coexist with papG allele II; can occur alone, i.e., in the absence of another papG allele; and can appear in evolutionary lineages distant from the J96-like O4:H5 clonal group (21, 27, 29). Moreover, rather than being a recent evolutionary development within (or acquisition by) E. coli, the papG allele I family appears to represent the most primitive branch of the papG tree. These observations present a series of seeming paradoxes. If papG allele I (I') is ancestral within E. coli, is compatible with each of the other papG alleles, and is horizontally mobile between lineages, what can account for its comparative rarity (particularly papG allele I') within the E. coli population, its near-total confinement to the J96-like clonal group, and its consistent association with papG allele III, almost to the exclusion of papG allele II (18, 21, 26, 27, 29, 40)?

A possible scenario would be that the J96-like clonal group, which uniquely contains representatives of all three papG alleles (both collectively and within individual strains, as demonstrated here for the first time) (18), may be the ancestral home of papG within E. coli. papG allele I, putatively the most primitive form of papG, by itself may confer only a minor fitness advantage. However, it may synergize with (presumably later-evolving) papG allele III, but not with (presumably even later-evolving) papG allele II, to confer a greater fitness advantage than does papG allele III alone. The presumably minimal selective advantage conferred by papG allele I itself thus would limit the driving force for horizontal spread of this papG allele. In contrast, papG allele III, which presumably confers a greater fitness advantage than does papG allele I, has spread to (and has been maintained within) multiple lineages within phylogenetic group B2 (21, 30, 40). (The logistic difficulties of coordinate horizontal transfer of both papG alleles I and III, which occur on separate pathogenicity islands, may account for the limited diffusion of this papG allele combination, despite its presumably greater selective advantage than that of either papG allele I or III individually.) Finally, papG allele II, which hypothetically confers an even greater fitness advantage than does papG allele III, has been propelled by selective pressure horizontally from the J96-like clonal group not only throughout phylogenetic group B2 but also into phylogenetic group D and beyond (21, 30, 40). This scenario represents a substantial departure from the hypothesis that papG allele I is the most recently evolved (or acquired) papG variant (21, 27, 29).

The occurrence in strain L31 of papG allele I' by itself, and in a lineage other than the J96-like clonal group, suggests that this version of papG allele I may confer a greater selective advantage than does papG allele I from strain J96. If so, the even greater rarity of papG allele I' than of papG allele I per se presents another paradox to be explained.

In the present study, both examples of papG allele I' were from canine UTI isolates (strains 7U and L31). In contrast, a papG allele I variant that was initially detected only by the allele I' primers but that according to full-length DNA sequence actually represented "orthodox" papG allele I was from a human UTI isolate (strain C367-82). These data might be interpreted as suggesting that papG allele I' is specific to dogs. However, the limited sample size precludes firm conclusions. It must be remembered that for some time after its discovery, papG allele III was presumed to be specific to dogs (3, 6, 40, 53), whereas it is now known to be common also among human isolates from diverse clinical syndromes (1, 14, 16, 21, 26, 42). Additional investigation is needed to define the epidemiological associations and host range of papG allele I'.

The varied papA alleles encountered in the present study in conjunction with papG alleles I and I' provide additional evidence that horizontal transfer and recombination involving papA can proceed independently of papG. Thus, horizontal transfer of virulence genes can be envisioned as involving entire pathogenicity islands or large fragments thereof (i.e., of multiple virulence operons) (7, 8, 49, 56), single operons such as pap (4, 40, 44), or individual genes from within these operons, such as papA or papG (28, 31, 40). Possible driving forces for reassortment of papA and papG alleles through horizontal transfer may include the antigenic diversity provided by the various PapA structural subunit variants (31) and the diversity of receptor recognition capabilities provided by the various PapG adhesion variants (15, 19, 32, 36, 37, 53, 54).

Summary. In the present study we used molecular and phenotypic methods to characterize novel variants of papG allele I and the corresponding wild-type E. coli strains. The new papG allele I variants exhibited novel agglutination phenotypes, papG and papA allele environments, and clonal associations. Comparative sequence analysis suggested that the papG allele I family may represent the earliest evolutionary branch from a common papG ancestor. Together with previous findings, these data support the hypothesis that the J96-like clonal group of E. coli O4:H5 may represent the original source of papG within the species.